libvirt/src/network/bridge_driver.c

5651 lines
184 KiB
C
Raw Normal View History

/*
2010-06-19 18:08:25 +00:00
* bridge_driver.c: core driver methods for managing network
*
* Copyright (C) 2006-2016 Red Hat, Inc.
* Copyright (C) 2006 Daniel P. Berrange
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library. If not, see
* <http://www.gnu.org/licenses/>.
*
* Author: Daniel P. Berrange <berrange@redhat.com>
*/
#include <config.h>
#include <sys/types.h>
#include <sys/poll.h>
#include <limits.h>
#include <string.h>
#include <stdio.h>
#include <stdarg.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <sys/utsname.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <signal.h>
#include <pwd.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
#include <net/if.h>
#include <dirent.h>
#if HAVE_SYS_SYSCTL_H
# include <sys/sysctl.h>
#endif
#include "virerror.h"
#include "datatypes.h"
#include "bridge_driver.h"
#include "bridge_driver_platform.h"
#include "device_conf.h"
#include "driver.h"
#include "virbuffer.h"
#include "virpidfile.h"
#include "vircommand.h"
2012-12-12 18:06:53 +00:00
#include "viralloc.h"
2012-12-13 18:01:25 +00:00
#include "viruuid.h"
#include "viriptables.h"
2012-12-12 17:59:27 +00:00
#include "virlog.h"
#include "virdnsmasq.h"
maint: use gnulib configmake rather than open-coding things * bootstrap.conf (gnulib_modules): Add configmake. * daemon/Makefile.am (libvirtd_CFLAGS): Drop defines provided by gnulib. * src/Makefile.am (INCLUDES): Likewise. * tests/Makefile.am (INCLUDES): Likewise. * tools/Makefile.am (virsh_CFLAGS): Likewise. * daemon/libvirtd.c (qemudInitPaths, usage, main): Update clients. * src/cpu/cpu_map.c (CPUMAPFILE): Likewise. * src/driver.c (DEFAULT_DRIVER_DIR): Likewise. * src/internal.h (_): Likewise. * src/libvirt.c (virInitialize): Likewise. * src/lxc/lxc_conf.h (LXC_CONFIG_DIR, LXC_STATE_DIR, LXC_LOG_DIR): Likewise. * src/lxc/lxc_conf.c (lxcCapsInit, lxcLoadDriverConfig): Likewise. * src/network/bridge_driver.c (NETWORK_PID_DIR) (NETWORK_STATE_DIR, DNSMASQ_STATE_DIR, networkStartup): Likewise. * src/nwfilter/nwfilter_driver.c (nwfilterDriverStartup): Likewise. * src/qemu/qemu_conf.c (qemudLoadDriverConfig): Likewise. * src/qemu/qemu_driver.c (qemudStartup): Likewise. * src/remote/remote_driver.h (LIBVIRTD_PRIV_UNIX_SOCKET) (LIBVIRTD_PRIV_UNIX_SOCKET_RO, LIBVIRTD_CONFIGURATION_FILE) (LIBVIRT_PKI_DIR): Likewise. * src/secret/secret_driver.c (secretDriverStartup): Likewise. * src/security/security_apparmor.c (VIRT_AA_HELPER): Likewise. * src/security/virt-aa-helper.c (main): Likewise. * src/storage/storage_backend_disk.c (PARTHELPER): Likewise. * src/storage/storage_driver.c (storageDriverStartup): Likewise. * src/uml/uml_driver.c (TEMPDIR, umlStartup): Likewise. * src/util/hooks.c (LIBVIRT_HOOK_DIR): Likewise. * tools/virsh.c (main): Likewise. * docs/hooks.html.in: Likewise.
2010-11-16 14:54:17 +00:00
#include "configmake.h"
#include "virnetlink.h"
#include "virnetdev.h"
#include "virnetdevip.h"
#include "virnetdevbridge.h"
#include "virnetdevtap.h"
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
#include "virnetdevvportprofile.h"
#include "virpci.h"
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
#include "virdbus.h"
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
#include "virfile.h"
#include "virstring.h"
#include "viraccessapicheck.h"
#include "network_event.h"
#include "virhook.h"
#include "virjson.h"
#define VIR_FROM_THIS VIR_FROM_NETWORK
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
#define MAX_BRIDGE_ID 256
/**
* VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX:
*
* Macro providing the upper limit on the size of leases file
*/
#define VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX (32 * 1024 * 1024)
#define SYSCTL_PATH "/proc/sys"
VIR_LOG_INIT("network.bridge_driver");
static virNetworkDriverStatePtr network_driver;
static virNetworkDriverStatePtr
networkGetDriver(void)
{
/* Maybe one day we can store @network_driver in the
* connection object, but until then, it's just a global
* variable which is returned. */
return network_driver;
}
static void
networkDriverLock(virNetworkDriverStatePtr driver)
{
2009-01-15 19:56:05 +00:00
virMutexLock(&driver->lock);
}
static void
networkDriverUnlock(virNetworkDriverStatePtr driver)
{
2009-01-15 19:56:05 +00:00
virMutexUnlock(&driver->lock);
}
static dnsmasqCapsPtr
networkGetDnsmasqCaps(virNetworkDriverStatePtr driver)
{
dnsmasqCapsPtr ret;
networkDriverLock(driver);
ret = virObjectRef(driver->dnsmasqCaps);
networkDriverUnlock(driver);
return ret;
}
static int
networkDnsmasqCapsRefresh(virNetworkDriverStatePtr driver)
{
dnsmasqCapsPtr caps;
if (!(caps = dnsmasqCapsNewFromBinary(DNSMASQ)))
return -1;
networkDriverLock(driver);
virObjectUnref(driver->dnsmasqCaps);
driver->dnsmasqCaps = caps;
networkDriverUnlock(driver);
return 0;
}
static int
networkStateCleanup(void);
static int
networkStartNetwork(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj);
static int
networkShutdownNetwork(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj);
static int
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj);
static int
networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static int
networkStartNetworkExternal(virNetworkObjPtr obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static int
networkShutdownNetworkExternal(virNetworkObjPtr obj);
static void
networkReloadFirewallRules(virNetworkDriverStatePtr driver);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static void
networkRefreshDaemons(virNetworkDriverStatePtr driver);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static int
networkPlugBandwidth(virNetworkObjPtr obj,
virDomainNetDefPtr iface);
static int
networkUnplugBandwidth(virNetworkObjPtr obj,
virDomainNetDefPtr iface);
static void
networkNetworkObjTaint(virNetworkObjPtr obj,
virNetworkTaintFlags taint);
static virNetworkObjPtr
networkObjFromNetwork(virNetworkPtr net)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
char uuidstr[VIR_UUID_STRING_BUFLEN];
obj = virNetworkObjFindByUUID(driver->networks, net->uuid);
if (!obj) {
virUUIDFormat(net->uuid, uuidstr);
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching uuid '%s' (%s)"),
uuidstr, net->name);
}
return obj;
}
static int
networkRunHook(virNetworkObjPtr obj,
virDomainDefPtr dom,
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
virDomainNetDefPtr iface,
int op,
int sub_op)
{
virNetworkDefPtr def;
virBuffer buf = VIR_BUFFER_INITIALIZER;
char *xml = NULL, *net_xml = NULL, *dom_xml = NULL;
int hookret;
int ret = -1;
if (virHookPresent(VIR_HOOK_DRIVER_NETWORK)) {
if (!obj) {
VIR_DEBUG("Not running hook as @obj is NULL");
ret = 0;
goto cleanup;
}
def = virNetworkObjGetDef(obj);
virBufferAddLit(&buf, "<hookData>\n");
virBufferAdjustIndent(&buf, 2);
if (iface && virDomainNetDefFormat(&buf, iface, NULL, 0) < 0)
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
goto cleanup;
if (virNetworkDefFormatBuf(&buf, def, 0) < 0)
goto cleanup;
if (dom && virDomainDefFormatInternal(dom, NULL, 0, &buf, NULL) < 0)
goto cleanup;
virBufferAdjustIndent(&buf, -2);
virBufferAddLit(&buf, "</hookData>");
if (virBufferCheckError(&buf) < 0)
goto cleanup;
xml = virBufferContentAndReset(&buf);
hookret = virHookCall(VIR_HOOK_DRIVER_NETWORK, def->name,
op, sub_op, NULL, xml, NULL);
/*
* If the script raised an error, pass it to the callee.
*/
if (hookret < 0)
goto cleanup;
networkNetworkObjTaint(obj, VIR_NETWORK_TAINT_HOOK);
}
ret = 0;
cleanup:
virBufferFreeAndReset(&buf);
VIR_FREE(xml);
VIR_FREE(net_xml);
VIR_FREE(dom_xml);
return ret;
}
static char *
networkDnsmasqLeaseFileNameDefault(virNetworkDriverStatePtr driver,
const char *netname)
{
char *leasefile;
ignore_value(virAsprintf(&leasefile, "%s/%s.leases",
driver->dnsmasqStateDir, netname));
return leasefile;
}
static char *
networkDnsmasqLeaseFileNameCustom(virNetworkDriverStatePtr driver,
const char *bridge)
{
char *leasefile;
ignore_value(virAsprintf(&leasefile, "%s/%s.status",
driver->dnsmasqStateDir, bridge));
return leasefile;
}
static char *
networkDnsmasqConfigFileName(virNetworkDriverStatePtr driver,
const char *netname)
{
char *conffile;
ignore_value(virAsprintf(&conffile, "%s/%s.conf",
driver->dnsmasqStateDir, netname));
return conffile;
}
static char *
networkRadvdPidfileBasename(const char *netname)
{
/* this is simple but we want to be sure it's consistently done */
char *pidfilebase;
ignore_value(virAsprintf(&pidfilebase, "%s-radvd", netname));
return pidfilebase;
}
static char *
networkRadvdConfigFileName(virNetworkDriverStatePtr driver,
const char *netname)
{
char *configfile;
ignore_value(virAsprintf(&configfile, "%s/%s-radvd.conf",
driver->radvdStateDir, netname));
return configfile;
}
/* do needed cleanup steps and remove the network from the list */
static int
networkRemoveInactive(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
char *leasefile = NULL;
char *customleasefile = NULL;
char *radvdconfigfile = NULL;
char *configfile = NULL;
char *radvdpidbase = NULL;
char *statusfile = NULL;
char *macMapFile = NULL;
dnsmasqContext *dctx = NULL;
virNetworkDefPtr def = virNetworkObjGetPersistentDef(obj);
int ret = -1;
/* remove the (possibly) existing dnsmasq and radvd files */
if (!(dctx = dnsmasqContextNew(def->name,
driver->dnsmasqStateDir))) {
goto cleanup;
}
if (!(leasefile = networkDnsmasqLeaseFileNameDefault(driver, def->name)))
goto cleanup;
if (!(customleasefile = networkDnsmasqLeaseFileNameCustom(driver, def->bridge)))
goto cleanup;
if (!(radvdconfigfile = networkRadvdConfigFileName(driver, def->name)))
goto cleanup;
if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
goto cleanup;
if (!(configfile = networkDnsmasqConfigFileName(driver, def->name)))
goto cleanup;
if (!(statusfile = virNetworkConfigFile(driver->stateDir, def->name)))
goto cleanup;
if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir, def->bridge)))
goto cleanup;
/* dnsmasq */
dnsmasqDelete(dctx);
unlink(leasefile);
unlink(customleasefile);
unlink(configfile);
/* MAC map manager */
unlink(macMapFile);
/* radvd */
unlink(radvdconfigfile);
virPidFileDelete(driver->pidDir, radvdpidbase);
/* remove status file */
unlink(statusfile);
/* remove the network definition */
virNetworkObjRemoveInactive(driver->networks, obj);
ret = 0;
cleanup:
VIR_FREE(leasefile);
VIR_FREE(configfile);
VIR_FREE(customleasefile);
VIR_FREE(radvdconfigfile);
VIR_FREE(radvdpidbase);
VIR_FREE(statusfile);
VIR_FREE(macMapFile);
dnsmasqContextFree(dctx);
return ret;
}
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
static char *
networkBridgeDummyNicName(const char *brname)
{
static const char dummyNicSuffix[] = "-nic";
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
char *nicname;
if (strlen(brname) + sizeof(dummyNicSuffix) > IFNAMSIZ) {
/* because the length of an ifname is limited to IFNAMSIZ-1
* (usually 15), and we're adding 4 more characters, we must
* truncate the original name to 11 to fit. In order to catch
* a possible numeric ending (eg virbr0, virbr1, etc), we grab
* the first 8 and last 3 characters of the string.
*/
ignore_value(virAsprintf(&nicname, "%.*s%s%s",
/* space for last 3 chars + "-nic" + NULL */
(int)(IFNAMSIZ - (3 + sizeof(dummyNicSuffix))),
brname, brname + strlen(brname) - 3,
dummyNicSuffix));
} else {
ignore_value(virAsprintf(&nicname, "%s%s", brname, dummyNicSuffix));
}
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
return nicname;
}
static int
networkUpdateState(virNetworkObjPtr obj,
void *opaque)
{
virNetworkDefPtr def;
virNetworkDriverStatePtr driver = opaque;
dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
virMacMapPtr macmap;
char *macMapFile = NULL;
int ret = -1;
virObjectLock(obj);
if (!virNetworkObjIsActive(obj)) {
ret = 0;
goto cleanup;
}
def = virNetworkObjGetDef(obj);
switch (def->forward.type) {
case VIR_NETWORK_FORWARD_NONE:
case VIR_NETWORK_FORWARD_NAT:
case VIR_NETWORK_FORWARD_ROUTE:
case VIR_NETWORK_FORWARD_OPEN:
/* If bridge doesn't exist, then mark it inactive */
if (!(def->bridge && virNetDevExists(def->bridge) == 1))
virNetworkObjSetActive(obj, false);
if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir,
def->bridge)))
goto cleanup;
if (!(macmap = virMacMapNew(macMapFile)))
goto cleanup;
virNetworkObjSetMacMap(obj, macmap);
break;
case VIR_NETWORK_FORWARD_BRIDGE:
if (def->bridge) {
if (virNetDevExists(def->bridge) != 1)
virNetworkObjSetActive(obj, false);
network: set macvtap/hostdev networks active if their state file exists libvirt attempts to determine at startup time which networks are already active, and set their active flags. Previously it has done this by assuming that all networks are inactive, then setting the active flag if the network has a bridge device associated with it and that bridge device exists. This is not useful for macvtap and hostdev based networks, since they do not use a bridge device. Of course the reason that such a check had to be done was that the presence of a status file in the network "stateDir" couldn't be trusted as an indicator of whether or not a network was active. This was due to the network driver mistakenly using /var/lib/libvirt/network to store the status files, rather than /var/run/libvirt/network (similar to what is done by every other libvirt driver that stores status xml for its objects). The difference is that /var/run is cleared out when the host reboots, so you can be assured that the state file you are seeing isn't just left over from a previous boot of the host. Now that the network driver has been switched to using /var/run/libvirt/network for status, we can also modify it to assume that any network with an existing status file is by definition active - we do this when reading the status file. To fine tune the results, networkFindActiveConfigs() is changed to networkUpdateAllState(), and only sets active = 0 if the conditions for particular network types are *not* met. The result is that during the first run of libvirtd after the host boots, there are no status files, so no networks are active. Any time libvirtd is restarted, any network with a status file will be marked as active (unless the network uses a bridge device and that device for some reason doesn't exist).
2014-04-09 14:16:45 +00:00
break;
}
/* intentionally drop through to common case for all
* macvtap networks (forward='bridge' with no bridge
* device defined is macvtap using its 'bridge' mode)
*/
case VIR_NETWORK_FORWARD_PRIVATE:
case VIR_NETWORK_FORWARD_VEPA:
case VIR_NETWORK_FORWARD_PASSTHROUGH:
/* so far no extra checks */
break;
case VIR_NETWORK_FORWARD_HOSTDEV:
/* so far no extra checks */
break;
}
/* Try and read dnsmasq/radvd pids of active networks */
if (virNetworkObjIsActive(obj) && def->ips && (def->nips > 0)) {
pid_t radvdPid;
pid_t dnsmasqPid;
char *radvdpidbase;
ignore_value(virPidFileReadIfAlive(driver->pidDir,
def->name,
&dnsmasqPid,
dnsmasqCapsGetBinaryPath(dnsmasq_caps)));
virNetworkObjSetDnsmasqPid(obj, dnsmasqPid);
radvdpidbase = networkRadvdPidfileBasename(def->name);
if (!radvdpidbase)
goto cleanup;
ignore_value(virPidFileReadIfAlive(driver->pidDir,
radvdpidbase,
&radvdPid, RADVD));
virNetworkObjSetRadvdPid(obj, radvdPid);
VIR_FREE(radvdpidbase);
}
ret = 0;
cleanup:
virObjectUnlock(obj);
virObjectUnref(dnsmasq_caps);
VIR_FREE(macMapFile);
return ret;
}
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
static int
networkAutostartConfig(virNetworkObjPtr obj,
void *opaque)
{
virNetworkDriverStatePtr driver = opaque;
int ret = -1;
virObjectLock(obj);
if (virNetworkObjIsAutostart(obj) &&
!virNetworkObjIsActive(obj) &&
networkStartNetwork(driver, obj) < 0)
goto cleanup;
ret = 0;
cleanup:
virObjectUnlock(obj);
return ret;
}
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
#if HAVE_FIREWALLD
static DBusHandlerResult
firewalld_dbus_filter_bridge(DBusConnection *connection ATTRIBUTE_UNUSED,
DBusMessage *message,
void *user_data)
{
virNetworkDriverStatePtr driver = user_data;
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
if (dbus_message_is_signal(message, DBUS_INTERFACE_DBUS,
"NameOwnerChanged") ||
dbus_message_is_signal(message, "org.fedoraproject.FirewallD1",
"Reloaded"))
{
VIR_DEBUG("Reload in bridge_driver because of firewalld.");
networkReloadFirewallRules(driver);
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
}
return DBUS_HANDLER_RESULT_NOT_YET_HANDLED;
}
#endif
static int
networkMigrateStateFiles(virNetworkDriverStatePtr driver)
{
/* Due to a change in location of network state xml beginning in
* libvirt 1.2.4 (from /var/lib/libvirt/network to
* /var/run/libvirt/network), we must check for state files in two
* locations. Anything found in the old location must be written
* to the new location, then erased from the old location. (Note
* that we read/write the file rather than calling rename()
* because the old and new state directories are likely in
* different filesystems).
*/
int ret = -1;
const char *oldStateDir = LOCALSTATEDIR "/lib/libvirt/network";
DIR *dir;
int direrr;
struct dirent *entry;
char *oldPath = NULL, *newPath = NULL;
char *contents = NULL;
int rc;
if ((rc = virDirOpenIfExists(&dir, oldStateDir)) <= 0)
return rc;
if (virFileMakePath(driver->stateDir) < 0) {
virReportSystemError(errno, _("cannot create directory %s"),
driver->stateDir);
goto cleanup;
}
while ((direrr = virDirRead(dir, &entry, oldStateDir)) > 0) {
if (entry->d_type != DT_UNKNOWN &&
entry->d_type != DT_REG)
continue;
if (virAsprintf(&oldPath, "%s/%s",
oldStateDir, entry->d_name) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
if (entry->d_type == DT_UNKNOWN) {
struct stat st;
if (lstat(oldPath, &st) < 0) {
virReportSystemError(errno,
_("failed to stat network status file '%s'"),
oldPath);
goto cleanup;
}
if (!S_ISREG(st.st_mode)) {
VIR_FREE(oldPath);
continue;
}
}
if (virFileReadAll(oldPath, 1024*1024, &contents) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
if (virAsprintf(&newPath, "%s/%s",
driver->stateDir, entry->d_name) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
if (virFileWriteStr(newPath, contents, S_IRUSR | S_IWUSR) < 0) {
virReportSystemError(errno,
_("failed to write network status file '%s'"),
newPath);
goto cleanup;
}
unlink(oldPath);
VIR_FREE(oldPath);
VIR_FREE(newPath);
VIR_FREE(contents);
}
if (direrr < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
ret = 0;
cleanup:
VIR_DIR_CLOSE(dir);
VIR_FREE(oldPath);
VIR_FREE(newPath);
VIR_FREE(contents);
return ret;
}
/**
* networkStateInitialize:
*
* Initialization function for the QEMU daemon
*/
static int
networkStateInitialize(bool privileged,
virStateInhibitCallback callback ATTRIBUTE_UNUSED,
void *opaque ATTRIBUTE_UNUSED)
{
int ret = -1;
char *configdir = NULL;
char *rundir = NULL;
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
#ifdef HAVE_FIREWALLD
DBusConnection *sysbus = NULL;
#endif
if (VIR_ALLOC(network_driver) < 0)
goto error;
if (virMutexInit(&network_driver->lock) < 0) {
VIR_FREE(network_driver);
2009-01-15 19:56:05 +00:00
goto error;
}
network_driver->privileged = privileged;
/* configuration/state paths are one of
* ~/.config/libvirt/... (session/unprivileged)
* /etc/libvirt/... && /var/(run|lib)/libvirt/... (system/privileged).
*/
if (privileged) {
if (VIR_STRDUP(network_driver->networkConfigDir,
SYSCONFDIR "/libvirt/qemu/networks") < 0 ||
VIR_STRDUP(network_driver->networkAutostartDir,
SYSCONFDIR "/libvirt/qemu/networks/autostart") < 0 ||
VIR_STRDUP(network_driver->stateDir,
LOCALSTATEDIR "/run/libvirt/network") < 0 ||
VIR_STRDUP(network_driver->pidDir,
LOCALSTATEDIR "/run/libvirt/network") < 0 ||
VIR_STRDUP(network_driver->dnsmasqStateDir,
LOCALSTATEDIR "/lib/libvirt/dnsmasq") < 0 ||
VIR_STRDUP(network_driver->radvdStateDir,
LOCALSTATEDIR "/lib/libvirt/radvd") < 0)
goto error;
/* migration from old to new location is only applicable for
* privileged mode - unprivileged mode directories haven't
* changed location.
*/
if (networkMigrateStateFiles(network_driver) < 0)
goto error;
} else {
configdir = virGetUserConfigDirectory();
rundir = virGetUserRuntimeDirectory();
if (!(configdir && rundir))
2009-01-22 19:41:48 +00:00
goto error;
if ((virAsprintf(&network_driver->networkConfigDir,
"%s/qemu/networks", configdir) < 0) ||
(virAsprintf(&network_driver->networkAutostartDir,
"%s/qemu/networks/autostart", configdir) < 0) ||
(virAsprintf(&network_driver->stateDir,
"%s/network/lib", rundir) < 0) ||
(virAsprintf(&network_driver->pidDir,
"%s/network/run", rundir) < 0) ||
(virAsprintf(&network_driver->dnsmasqStateDir,
"%s/dnsmasq/lib", rundir) < 0) ||
(virAsprintf(&network_driver->radvdStateDir,
"%s/radvd/lib", rundir) < 0)) {
goto error;
2009-01-22 19:41:48 +00:00
}
}
if (virFileMakePath(network_driver->stateDir) < 0) {
virReportSystemError(errno,
_("cannot create directory %s"),
network_driver->stateDir);
goto error;
}
/* if this fails now, it will be retried later with dnsmasqCapsRefresh() */
network_driver->dnsmasqCaps = dnsmasqCapsNewFromBinary(DNSMASQ);
if (!(network_driver->networks = virNetworkObjListNew()))
goto error;
if (virNetworkObjLoadAllState(network_driver->networks,
network_driver->stateDir) < 0)
goto error;
if (virNetworkObjLoadAllConfigs(network_driver->networks,
network_driver->networkConfigDir,
network_driver->networkAutostartDir) < 0)
goto error;
/* Update the internal status of all allegedly active
* networks according to external conditions on the host
* (i.e. anything that isn't stored directly in each
* network's state file). */
virNetworkObjListForEach(network_driver->networks,
networkUpdateState,
network_driver);
virNetworkObjListPrune(network_driver->networks,
VIR_CONNECT_LIST_NETWORKS_INACTIVE |
VIR_CONNECT_LIST_NETWORKS_TRANSIENT);
networkReloadFirewallRules(network_driver);
networkRefreshDaemons(network_driver);
network_driver->networkEventState = virObjectEventStateNew();
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
#ifdef HAVE_FIREWALLD
if (!(sysbus = virDBusGetSystemBus())) {
VIR_WARN("DBus not available, disabling firewalld support "
"in bridge_network_driver: %s", virGetLastErrorMessage());
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
} else {
/* add matches for
* NameOwnerChanged on org.freedesktop.DBus for firewalld start/stop
* Reloaded on org.fedoraproject.FirewallD1 for firewalld reload
*/
dbus_bus_add_match(sysbus,
"type='signal'"
",interface='"DBUS_INTERFACE_DBUS"'"
",member='NameOwnerChanged'"
",arg0='org.fedoraproject.FirewallD1'",
NULL);
dbus_bus_add_match(sysbus,
"type='signal'"
",interface='org.fedoraproject.FirewallD1'"
",member='Reloaded'",
NULL);
dbus_connection_add_filter(sysbus, firewalld_dbus_filter_bridge,
network_driver, NULL);
network: use firewalld instead of iptables, when available * configure.ac, spec file: firewalld defaults to enabled if dbus is available, otherwise is disabled. If --with_firewalld is explicitly requested and dbus is not available, configure will fail. * bridge_driver: add dbus filters to get the FirewallD1.Reloaded signal and DBus.NameOwnerChanged on org.fedoraproject.FirewallD1. When these are encountered, reload all the iptables reuls of all libvirt's virtual networks (similar to what happens when libvirtd is restarted). * iptables, ebtables: use firewall-cmd's direct passthrough interface when available, otherwise use iptables and ebtables commands. This decision is made once the first time libvirt calls iptables/ebtables, and that decision is maintained for the life of libvirtd. * Note that the nwfilter part of this patch was separated out into another patch by Stefan in V2, so that needs to be revised and re-reviewed as well. ================ All the configure.ac and specfile changes are unchanged from Thomas' V3. V3 re-ran "firewall-cmd --state" every time a new rule was added, which was extremely inefficient. V4 uses VIR_ONCE_GLOBAL_INIT to set up a one-time initialization function. The VIR_ONCE_GLOBAL_INIT(x) macro references a static function called vir(Ip|Eb)OnceInit(), which will then be called the first time that the static function vir(Ip|Eb)TablesInitialize() is called (that function is defined for you by the macro). This is thread-safe, so there is no chance of any race. IMPORTANT NOTE: I've left the VIR_DEBUG messages in these two init functions (one for iptables, on for ebtables) as VIR_WARN so that I don't have to turn on all the other debug message just to see these. Even if this patch doesn't need any other modification, those messages need to be changed to VIR_DEBUG before pushing. This one-time initialization works well. However, I've encountered problems with testing: 1) Whenever I have enabled the firewalld service, *all* attempts to call firewall-cmd from within libvirtd end with firewall-cmd hanging internally somewhere. This is *not* the case if firewall-cmd returns non-0 in response to "firewall-cmd --state" (i.e. *that* command runs and returns to libvirt successfully.) 2) If I start libvirtd while firewalld is stopped, then start firewalld later, this triggers libvirtd to reload its iptables rules, however it also spits out a *ton* of complaints about deletion failing (I suppose because firewalld has nuked all of libvirt's rules). I guess we need to suppress those messages (which is a more annoying problem to fix than you might think, but that's another story). 3) I noticed a few times during this long line of errors that firewalld made a complaint about "Resource Temporarily unavailable. Having libvirtd access iptables commands directly at the same time as firewalld is doing so is apparently problematic. 4) In general, I'm concerned about the "set it once and never change it" method - if firewalld is disabled at libvirtd startup, causing libvirtd to always use iptables/ebtables directly, this won't cause *terrible* problems, but if libvirtd decides to use firewall-cmd and firewalld is later disabled, libvirtd will not be able to recover.
2012-08-14 18:59:52 +00:00
}
#endif
ret = 0;
cleanup:
VIR_FREE(configdir);
VIR_FREE(rundir);
return ret;
error:
networkStateCleanup();
goto cleanup;
}
/**
* networkStateAutoStart:
*
* Function to AutoStart the bridge configs
*/
static void
networkStateAutoStart(void)
{
if (!network_driver)
return;
virNetworkObjListForEach(network_driver->networks,
networkAutostartConfig,
network_driver);
}
/**
* networkStateReload:
*
* Function to restart the QEMU daemon, it will recheck the configuration
* files and update its state and the networking
*/
static int
networkStateReload(void)
{
if (!network_driver)
return 0;
virNetworkObjLoadAllState(network_driver->networks,
network_driver->stateDir);
virNetworkObjLoadAllConfigs(network_driver->networks,
network_driver->networkConfigDir,
network_driver->networkAutostartDir);
networkReloadFirewallRules(network_driver);
networkRefreshDaemons(network_driver);
virNetworkObjListForEach(network_driver->networks,
networkAutostartConfig,
network_driver);
return 0;
}
/**
* networkStateCleanup:
*
* Shutdown the QEMU daemon, it will stop all active domains and networks
*/
static int
networkStateCleanup(void)
{
if (!network_driver)
return -1;
virObjectUnref(network_driver->networkEventState);
/* free inactive networks */
virObjectUnref(network_driver->networks);
VIR_FREE(network_driver->networkConfigDir);
VIR_FREE(network_driver->networkAutostartDir);
VIR_FREE(network_driver->stateDir);
VIR_FREE(network_driver->pidDir);
VIR_FREE(network_driver->dnsmasqStateDir);
VIR_FREE(network_driver->radvdStateDir);
virObjectUnref(network_driver->dnsmasqCaps);
virMutexDestroy(&network_driver->lock);
VIR_FREE(network_driver);
return 0;
}
static virDrvOpenStatus
networkConnectOpen(virConnectPtr conn,
virConnectAuthPtr auth ATTRIBUTE_UNUSED,
virConfPtr conf ATTRIBUTE_UNUSED,
unsigned int flags)
{
virCheckFlags(VIR_CONNECT_RO, VIR_DRV_OPEN_ERROR);
/* Verify uri was specified */
if (conn->uri == NULL) {
/* Only hypervisor drivers are permitted to auto-open on NULL uri */
return VIR_DRV_OPEN_DECLINED;
} else {
if (STRNEQ_NULLABLE(conn->uri->scheme, "network"))
return VIR_DRV_OPEN_DECLINED;
/* Leave for remote driver */
if (conn->uri->server != NULL)
return VIR_DRV_OPEN_DECLINED;
if (network_driver == NULL) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("network state driver is not active"));
return VIR_DRV_OPEN_ERROR;
}
if (network_driver->privileged) {
if (STRNEQ(conn->uri->path, "/system")) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("unexpected network URI path '%s', try network:///system"),
conn->uri->path);
return VIR_DRV_OPEN_ERROR;
}
} else {
if (STRNEQ(conn->uri->path, "/session")) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("unexpected network URI path '%s', try network:///session"),
conn->uri->path);
return VIR_DRV_OPEN_ERROR;
}
}
}
if (virConnectOpenEnsureACL(conn) < 0)
return VIR_DRV_OPEN_ERROR;
return VIR_DRV_OPEN_SUCCESS;
}
static int networkConnectClose(virConnectPtr conn ATTRIBUTE_UNUSED)
{
return 0;
}
static int networkConnectIsSecure(virConnectPtr conn ATTRIBUTE_UNUSED)
{
/* Trivially secure, since always inside the daemon */
return 1;
}
static int networkConnectIsEncrypted(virConnectPtr conn ATTRIBUTE_UNUSED)
{
/* Not encrypted, but remote driver takes care of that */
return 0;
}
static int networkConnectIsAlive(virConnectPtr conn ATTRIBUTE_UNUSED)
{
return 1;
}
/* networkKillDaemon:
*
* kill the specified pid/name, and wait a bit to make sure it's dead.
*/
static int
networkKillDaemon(pid_t pid,
const char *daemonName,
const char *networkName)
{
size_t i;
int ret = -1;
const char *signame = "TERM";
/* send SIGTERM, then wait up to 3 seconds for the process to
* disappear, send SIGKILL, then wait for up to another 2
* seconds. If that fails, log a warning and continue, hoping
* for the best.
*/
for (i = 0; i < 25; i++) {
int signum = 0;
if (i == 0) {
signum = SIGTERM;
} else if (i == 15) {
signum = SIGKILL;
signame = "KILL";
}
if (kill(pid, signum) < 0) {
if (errno == ESRCH) {
ret = 0;
} else {
char ebuf[1024];
VIR_WARN("Failed to terminate %s process %d "
"for network '%s' with SIG%s: %s",
daemonName, pid, networkName, signame,
virStrerror(errno, ebuf, sizeof(ebuf)));
}
goto cleanup;
}
/* NB: since networks have no reference count like
* domains, there is no safe way to unlock the network
* object temporarily, and so we can't follow the
* procedure used by the qemu driver of 1) unlock driver
* 2) sleep, 3) add ref to object 4) unlock object, 5)
* re-lock driver, 6) re-lock object. We may need to add
* that functionality eventually, but for now this
* function is rarely used and, at worst, leaving the
* network driver locked during this loop of sleeps will
* have the effect of holding up any other thread trying
* to make modifications to a network for up to 5 seconds;
* since modifications to networks are much less common
* than modifications to domains, this seems a reasonable
* tradeoff in exchange for less code disruption.
*/
usleep(20 * 1000);
}
VIR_WARN("Timed out waiting after SIG%s to %s process %d "
"(network '%s')",
signame, daemonName, pid, networkName);
cleanup:
return ret;
}
2014-06-23 09:51:38 +00:00
/* the following does not build a file, it builds a list
* which is later saved into a file
*/
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
static int
networkBuildDnsmasqDhcpHostsList(dnsmasqContext *dctx,
virNetworkIPDefPtr ipdef)
{
size_t i;
bool ipv6 = false;
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
ipv6 = true;
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
for (i = 0; i < ipdef->nhosts; i++) {
virNetworkDHCPHostDefPtr host = &(ipdef->hosts[i]);
if (VIR_SOCKET_ADDR_VALID(&host->ip))
if (dnsmasqAddDhcpHost(dctx, host->mac, &host->ip,
host->name, host->id, ipv6) < 0)
return -1;
}
return 0;
}
static int
networkBuildDnsmasqHostsList(dnsmasqContext *dctx,
virNetworkDNSDefPtr dnsdef)
{
size_t i, j;
if (dnsdef) {
for (i = 0; i < dnsdef->nhosts; i++) {
virNetworkDNSHostDefPtr host = &(dnsdef->hosts[i]);
if (VIR_SOCKET_ADDR_VALID(&host->ip)) {
for (j = 0; j < host->nnames; j++)
if (dnsmasqAddHost(dctx, &host->ip, host->names[j]) < 0)
return -1;
}
}
}
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
return 0;
}
static int
networkDnsmasqConfLocalPTRs(virBufferPtr buf,
virNetworkDefPtr def)
{
virNetworkIPDefPtr ip;
size_t i;
char *ptr = NULL;
int rc;
for (i = 0; i < def->nips; i++) {
ip = def->ips + i;
if (ip->localPTR != VIR_TRISTATE_BOOL_YES)
continue;
if ((rc = virSocketAddrPTRDomain(&ip->address,
virNetworkIPDefPrefix(ip),
&ptr)) < 0) {
if (rc == -2) {
int family = VIR_SOCKET_ADDR_FAMILY(&ip->address);
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("PTR domain for %s network with prefix %u "
"cannot be automatically created"),
(family == AF_INET) ? "IPv4" : "IPv6",
virNetworkIPDefPrefix(ip));
}
return -1;
}
virBufferAsprintf(buf, "local=/%s/\n", ptr);
VIR_FREE(ptr);
}
return 0;
}
int
networkDnsmasqConfContents(virNetworkObjPtr obj,
const char *pidfile,
char **configstr,
dnsmasqContext *dctx,
dnsmasqCapsPtr caps ATTRIBUTE_UNUSED)
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virBuffer configbuf = VIR_BUFFER_INITIALIZER;
int r, ret = -1;
int nbleases = 0;
size_t i;
virNetworkDNSDefPtr dns = &def->dns;
bool wantDNS = dns->enable != VIR_TRISTATE_BOOL_NO;
virNetworkIPDefPtr tmpipdef, ipdef, ipv4def, ipv6def;
bool ipv6SLAAC;
char *saddr = NULL, *eaddr = NULL;
*configstr = NULL;
/*
* All dnsmasq parameters are put into a configuration file, except the
* command line --conf-file=parameter which specifies the location of
* configuration file.
*
* All dnsmasq conf-file parameters must be specified as "foo=bar"
* as oppose to "--foo bar" which was acceptable on the command line.
*/
/*
* Needed to ensure dnsmasq uses same algorithm for processing
* multiple namedriver entries in /etc/resolv.conf as GLibC.
*/
/* create dnsmasq config file appropriate for this network */
/* Don't forget to update networkxml2conftest :-) */
virBufferAsprintf(&configbuf,
"##WARNING: THIS IS AN AUTO-GENERATED FILE. "
"CHANGES TO IT ARE LIKELY TO BE\n"
"##OVERWRITTEN AND LOST. Changes to this "
"configuration should be made using:\n"
"## virsh net-edit %s\n"
"## or other application using the libvirt API.\n"
"##\n## dnsmasq conf file created by libvirt\n"
"strict-order\n",
def->name);
/* if dns is disabled, set its listening port to 0, which
* tells dnsmasq to not listen
*/
if (!wantDNS)
virBufferAddLit(&configbuf, "port=0\n");
if (wantDNS && def->dns.forwarders) {
/* addNoResolv should be set to true if there are any entries
* that specify an IP address for requests, but no domain
* qualifier (implying that all requests otherwise "unclaimed"
* should be sent to that address). if it is still false when
* we've looked at all entries, it means we still need the
* host's resolv.conf for some cases.
*/
bool addNoResolv = false;
for (i = 0; i < def->dns.nfwds; i++) {
virNetworkDNSForwarderPtr fwd = &def->dns.forwarders[i];
virBufferAddLit(&configbuf, "server=");
if (fwd->domain)
virBufferAsprintf(&configbuf, "/%s/", fwd->domain);
if (VIR_SOCKET_ADDR_VALID(&fwd->addr)) {
char *addr = virSocketAddrFormat(&fwd->addr);
if (!addr)
goto cleanup;
virBufferAsprintf(&configbuf, "%s\n", addr);
VIR_FREE(addr);
if (!fwd->domain)
addNoResolv = true;
} else {
/* "don't forward requests for this domain" */
virBufferAddLit(&configbuf, "#\n");
}
}
if (addNoResolv)
virBufferAddLit(&configbuf, "no-resolv\n");
}
if (def->domain) {
if (def->domainLocalOnly == VIR_TRISTATE_BOOL_YES) {
virBufferAsprintf(&configbuf,
"local=/%s/\n",
def->domain);
}
virBufferAsprintf(&configbuf,
"domain=%s\n"
"expand-hosts\n",
def->domain);
}
if (wantDNS &&
networkDnsmasqConfLocalPTRs(&configbuf, def) < 0)
goto cleanup;
if (wantDNS && def->dns.forwardPlainNames == VIR_TRISTATE_BOOL_NO) {
network: only prevent forwarding of DNS requests for unqualified names In commit f386825 we began adding the options --domain-needed --local=/$mydomain/ to all dnsmasq commandlines with the stated reason of preventing forwarding of DNS queries for names that weren't fully qualified domain names ("FQDN", i.e. a name that included some "."s and a domain name). This was later changed to domain-needed local=/$mydomain/ when we moved the options from the dnsmasq commandline to a conf file. The original patch on the list, and discussion about it, is here: https://www.redhat.com/archives/libvir-list/2012-August/msg01594.html When a domain name isn't specified (mydomain == ""), the addition of "domain-needed local=//" will prevent forwarding of domain-less requests to the virtualization host's DNS resolver, but if a domain *is* specified, the addition of "local=/domain/" will prevent forwarding of any requests for *qualified* names within that domain that aren't resolvable by libvirt's dnsmasq itself. An example of the problems this causes - let's say a network is defined with: <domain name='example.com'/> <dhcp> .. <host mac='52:54:00:11:22:33' ip='1.2.3.4' name='myguest'/> </dhcp> This results in "local=/example.com/" being added to the dnsmasq options. If a guest requests "myguest" or "myguest.example.com", that will be resolved by dnsmasq. If the guest asks for "www.example.com", dnsmasq will not know the answer, but instead of forwarding it to the host, it will return NOT FOUND to the guest. In most cases that isn't the behavior an admin is looking for. A later patch (commit 4f595ba) attempted to remedy this by adding a "forwardPlainNames" attribute to the <dns> element. The idea was that if forwardPlainNames='yes' (default is 'no'), we would allow unresolved names to be forwarded. However, that patch was botched, in that it only removed the "domain-needed" option when forwardPlainNames='yes', and left the "local=/mydomain/". Really we should have been just including the option "--domain-needed --local=//" (note the lack of domain name) regardless of the configured domain of the network, so that requests for names without a domain would be treated as "local to dnsmasq" and not forwarded, but all others (including those in the network's configured domain) would be forwarded. We also shouldn't include *either* of those options if forwardPlainNames='yes'. This patch makes those corrections. This patch doesn't remedy the fact that default behavior was changed by the addition of this feature. That will be handled in a subsequent patch.
2013-12-06 10:55:37 +00:00
virBufferAddLit(&configbuf, "domain-needed\n");
/* need to specify local=// whether or not a domain is
* specified, unless the config says we should forward "plain"
* names (i.e. not fully qualified, no '.' characters)
*/
network: only prevent forwarding of DNS requests for unqualified names In commit f386825 we began adding the options --domain-needed --local=/$mydomain/ to all dnsmasq commandlines with the stated reason of preventing forwarding of DNS queries for names that weren't fully qualified domain names ("FQDN", i.e. a name that included some "."s and a domain name). This was later changed to domain-needed local=/$mydomain/ when we moved the options from the dnsmasq commandline to a conf file. The original patch on the list, and discussion about it, is here: https://www.redhat.com/archives/libvir-list/2012-August/msg01594.html When a domain name isn't specified (mydomain == ""), the addition of "domain-needed local=//" will prevent forwarding of domain-less requests to the virtualization host's DNS resolver, but if a domain *is* specified, the addition of "local=/domain/" will prevent forwarding of any requests for *qualified* names within that domain that aren't resolvable by libvirt's dnsmasq itself. An example of the problems this causes - let's say a network is defined with: <domain name='example.com'/> <dhcp> .. <host mac='52:54:00:11:22:33' ip='1.2.3.4' name='myguest'/> </dhcp> This results in "local=/example.com/" being added to the dnsmasq options. If a guest requests "myguest" or "myguest.example.com", that will be resolved by dnsmasq. If the guest asks for "www.example.com", dnsmasq will not know the answer, but instead of forwarding it to the host, it will return NOT FOUND to the guest. In most cases that isn't the behavior an admin is looking for. A later patch (commit 4f595ba) attempted to remedy this by adding a "forwardPlainNames" attribute to the <dns> element. The idea was that if forwardPlainNames='yes' (default is 'no'), we would allow unresolved names to be forwarded. However, that patch was botched, in that it only removed the "domain-needed" option when forwardPlainNames='yes', and left the "local=/mydomain/". Really we should have been just including the option "--domain-needed --local=//" (note the lack of domain name) regardless of the configured domain of the network, so that requests for names without a domain would be treated as "local to dnsmasq" and not forwarded, but all others (including those in the network's configured domain) would be forwarded. We also shouldn't include *either* of those options if forwardPlainNames='yes'. This patch makes those corrections. This patch doesn't remedy the fact that default behavior was changed by the addition of this feature. That will be handled in a subsequent patch.
2013-12-06 10:55:37 +00:00
virBufferAddLit(&configbuf, "local=//\n");
}
if (pidfile)
virBufferAsprintf(&configbuf, "pid-file=%s\n", pidfile);
network: prevent dnsmasq from listening on localhost This patch resolves the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=886663 The source of the problem was the fix for CVE 2011-3411: https://bugzilla.redhat.com/show_bug.cgi?id=833033 which was originally committed upstream in commit 753ff83a50263d6975f88d6605d4b5ddfcc97560. That commit improperly removed the "--except-interface lo" from dnsmasq commandlines when --bind-dynamic was used (based on comments in the latter bug). It turns out that the problem reported in the CVE could be eliminated without removing "--except-interface lo", and removing it actually caused each instance of dnsmasq to listen on localhost on port 53, which created a new problem: If another instance of dnsmasq using "bind-interfaces" (instead of "bind-dynamic") had already been started (or if another instance started later used "bind-dynamic"), this wouldn't have any immediately visible ill effects, but if you tried to start another dnsmasq instance using "bind-interfaces" *after* starting any libvirt networks, the new dnsmasq would fail to start, because there was already another process listening on port 53. (Subsequent to the CVE fix, another patch changed the network driver to put dnsmasq options in a conf file rather than directly on the dnsmasq commandline, but preserved the same options.) This patch changes the network driver to *always* add "except-interface=lo" to dnsmasq conf files, regardless of whether we use bind-dynamic or bind-interfaces. This way no libvirt dnsmasq instances are listening on localhost (and the CVE is still fixed). The actual code change is miniscule, but must be propogated through all of the test files as well.
2012-12-13 06:46:40 +00:00
/* dnsmasq will *always* listen on localhost unless told otherwise */
#ifdef __linux__
network: prevent dnsmasq from listening on localhost This patch resolves the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=886663 The source of the problem was the fix for CVE 2011-3411: https://bugzilla.redhat.com/show_bug.cgi?id=833033 which was originally committed upstream in commit 753ff83a50263d6975f88d6605d4b5ddfcc97560. That commit improperly removed the "--except-interface lo" from dnsmasq commandlines when --bind-dynamic was used (based on comments in the latter bug). It turns out that the problem reported in the CVE could be eliminated without removing "--except-interface lo", and removing it actually caused each instance of dnsmasq to listen on localhost on port 53, which created a new problem: If another instance of dnsmasq using "bind-interfaces" (instead of "bind-dynamic") had already been started (or if another instance started later used "bind-dynamic"), this wouldn't have any immediately visible ill effects, but if you tried to start another dnsmasq instance using "bind-interfaces" *after* starting any libvirt networks, the new dnsmasq would fail to start, because there was already another process listening on port 53. (Subsequent to the CVE fix, another patch changed the network driver to put dnsmasq options in a conf file rather than directly on the dnsmasq commandline, but preserved the same options.) This patch changes the network driver to *always* add "except-interface=lo" to dnsmasq conf files, regardless of whether we use bind-dynamic or bind-interfaces. This way no libvirt dnsmasq instances are listening on localhost (and the CVE is still fixed). The actual code change is miniscule, but must be propogated through all of the test files as well.
2012-12-13 06:46:40 +00:00
virBufferAddLit(&configbuf, "except-interface=lo\n");
#else
/* BSD family OSes and Solaris call loopback interface as lo0 */
virBufferAddLit(&configbuf, "except-interface=lo0\n");
#endif
network: prevent dnsmasq from listening on localhost This patch resolves the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=886663 The source of the problem was the fix for CVE 2011-3411: https://bugzilla.redhat.com/show_bug.cgi?id=833033 which was originally committed upstream in commit 753ff83a50263d6975f88d6605d4b5ddfcc97560. That commit improperly removed the "--except-interface lo" from dnsmasq commandlines when --bind-dynamic was used (based on comments in the latter bug). It turns out that the problem reported in the CVE could be eliminated without removing "--except-interface lo", and removing it actually caused each instance of dnsmasq to listen on localhost on port 53, which created a new problem: If another instance of dnsmasq using "bind-interfaces" (instead of "bind-dynamic") had already been started (or if another instance started later used "bind-dynamic"), this wouldn't have any immediately visible ill effects, but if you tried to start another dnsmasq instance using "bind-interfaces" *after* starting any libvirt networks, the new dnsmasq would fail to start, because there was already another process listening on port 53. (Subsequent to the CVE fix, another patch changed the network driver to put dnsmasq options in a conf file rather than directly on the dnsmasq commandline, but preserved the same options.) This patch changes the network driver to *always* add "except-interface=lo" to dnsmasq conf files, regardless of whether we use bind-dynamic or bind-interfaces. This way no libvirt dnsmasq instances are listening on localhost (and the CVE is still fixed). The actual code change is miniscule, but must be propogated through all of the test files as well.
2012-12-13 06:46:40 +00:00
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
if (dnsmasqCapsGet(caps, DNSMASQ_CAPS_BIND_DYNAMIC)) {
/* using --bind-dynamic with only --interface (no
* --listen-address) prevents dnsmasq from responding to dns
* queries that arrive on some interface other than our bridge
* interface (in other words, requests originating somewhere
* other than one of the virtual guests connected directly to
* this network). This was added in response to CVE 2012-3411.
*/
virBufferAsprintf(&configbuf,
"bind-dynamic\n"
"interface=%s\n",
def->bridge);
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
} else {
network: prevent dnsmasq from listening on localhost This patch resolves the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=886663 The source of the problem was the fix for CVE 2011-3411: https://bugzilla.redhat.com/show_bug.cgi?id=833033 which was originally committed upstream in commit 753ff83a50263d6975f88d6605d4b5ddfcc97560. That commit improperly removed the "--except-interface lo" from dnsmasq commandlines when --bind-dynamic was used (based on comments in the latter bug). It turns out that the problem reported in the CVE could be eliminated without removing "--except-interface lo", and removing it actually caused each instance of dnsmasq to listen on localhost on port 53, which created a new problem: If another instance of dnsmasq using "bind-interfaces" (instead of "bind-dynamic") had already been started (or if another instance started later used "bind-dynamic"), this wouldn't have any immediately visible ill effects, but if you tried to start another dnsmasq instance using "bind-interfaces" *after* starting any libvirt networks, the new dnsmasq would fail to start, because there was already another process listening on port 53. (Subsequent to the CVE fix, another patch changed the network driver to put dnsmasq options in a conf file rather than directly on the dnsmasq commandline, but preserved the same options.) This patch changes the network driver to *always* add "except-interface=lo" to dnsmasq conf files, regardless of whether we use bind-dynamic or bind-interfaces. This way no libvirt dnsmasq instances are listening on localhost (and the CVE is still fixed). The actual code change is miniscule, but must be propogated through all of the test files as well.
2012-12-13 06:46:40 +00:00
virBufferAddLit(&configbuf, "bind-interfaces\n");
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
/*
* --interface does not actually work with dnsmasq < 2.47,
* due to DAD for ipv6 addresses on the interface.
*
* virCommandAddArgList(cmd, "--interface", def->bridge, NULL);
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
*
* So listen on all defined IPv[46] addresses
*/
for (i = 0;
(tmpipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i));
i++) {
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
char *ipaddr = virSocketAddrFormat(&tmpipdef->address);
if (!ipaddr)
goto cleanup;
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
/* also part of CVE 2012-3411 - if the host's version of
* dnsmasq doesn't have bind-dynamic, only allow listening on
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
* private/local IP addresses (see RFC1918/RFC3484/RFC4193)
*/
2012-12-17 17:49:18 +00:00
if (!dnsmasqCapsGet(caps, DNSMASQ_CAPS_BINDTODEVICE) &&
!virSocketAddrIsPrivate(&tmpipdef->address)) {
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
unsigned long version = dnsmasqCapsGetVersion(caps);
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Publicly routable address %s is prohibited. "
"The version of dnsmasq on this host (%d.%d) "
2012-12-17 17:49:18 +00:00
"doesn't support the bind-dynamic option or "
"use SO_BINDTODEVICE on listening sockets, "
"one of which is required for safe operation "
"on a publicly routable subnet "
"(see CVE-2012-3411). You must either "
"upgrade dnsmasq, or use a private/local "
"subnet range for this network "
"(as described in RFC1918/RFC3484/RFC4193)."),
ipaddr, (int)version / 1000000,
(int)(version % 1000000) / 1000);
VIR_FREE(ipaddr);
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
goto cleanup;
}
virBufferAsprintf(&configbuf, "listen-address=%s\n", ipaddr);
network: use dnsmasq --bind-dynamic when available This bug resolves CVE-2012-3411, which is described in the following bugzilla report: https://bugzilla.redhat.com/show_bug.cgi?id=833033 The following report is specifically for libvirt on Fedora: https://bugzilla.redhat.com/show_bug.cgi?id=874702 In short, a dnsmasq instance run with the intention of listening for DHCP/DNS requests only on a libvirt virtual network (which is constructed using a Linux host bridge) would also answer queries sent from outside the virtualization host. This patch takes advantage of a new dnsmasq option "--bind-dynamic", which will cause the listening socket to be setup such that it will only receive those requests that actually come in via the bridge interface. In order for this behavior to actually occur, not only must "--bind-interfaces" be replaced with "--bind-dynamic", but also all "--listen-address" options must be replaced with a single "--interface" option. Fully: --bind-interfaces --except-interface lo --listen-address x.x.x.x ... (with --listen-address possibly repeated) is replaced with: --bind-dynamic --interface virbrX Of course libvirt can't use this new option if the host's dnsmasq doesn't have it, but we still want libvirt to function (because the great majority of libvirt installations, which only have mode='nat' networks using RFC1918 private address ranges (e.g. 192.168.122.0/24), are immune to this vulnerability from anywhere beyond the local subnet of the host), so we use the new dnsmasqCaps API to check if dnsmasq supports the new option and, if not, we use the "old" option style instead. In order to assure that this permissiveness doesn't lead to a vulnerable system, we do check for non-private addresses in this case, and refuse to start the network if both a) we are using the old-style options, and b) the network has a publicly routable IP address. Hopefully this will provide the proper balance of not being disruptive to those not practically affected, and making sure that those who *are* affected get their dnsmasq upgraded. (--bind-dynamic was added to dnsmasq in upstream commit 54dd393f3938fc0c19088fbd319b95e37d81a2b0, which was included in dnsmasq-2.63)
2012-11-22 02:21:02 +00:00
VIR_FREE(ipaddr);
}
}
network driver: don't send default route to clients on isolated networks Normally dnsmasq will send a default route (the address of the host in the network definition) to any client requesting an address via DHCP. On an isolated network this makes no sense, as we have iptables to prevent any traffic going out via that interface, so anything sent that way would be dropped anyway. This extra/unusable default route becomes problematic if you have setup a guest with multiple network interfaces, with one connected to an isolated network and another that provides connectivity to the outside (example - one interface directly connecting to a physical interface via macvtap, with a second connected to an isolated network so that the host and guest can communicate (macvtap doesn't support guest<->host communication without an external switch that supports vepa, or reflecting all traffic back)). In this case, if the guest chooses the default route of the isolated network, the guest will not be able to get network traffic beyond the host. To prevent dnsmasq from sending a default route, you can tell it to send 0 bytes of data for the default route option (option number 3) with --dhcp-option=3 (normally the data to send for the option would follow the option number; no extra data means "don't send this option"). I have checked on RHEL5 (a good representative of the oldest supported libvirt platforms) and its version of dnsmasq (2.45) does support --dhcp-option, so this shouldn't create any compatibility problems.
2011-03-13 08:42:58 +00:00
/* If this is an isolated network, set the default route option
* (3) to be empty to avoid setting a default route that's
* guaranteed to not work, and set no-resolv so that no dns
* requests are forwarded on to the dns server listed in the
* host's /etc/resolv.conf (since this could be used as a channel
* to build a connection to the outside).
* IPv6 RA always contains an implicit default route
* via the sender's link-local address. The only thing we can do
* is set the lifetime of this route to 0, i.e. disable it.
network driver: don't send default route to clients on isolated networks Normally dnsmasq will send a default route (the address of the host in the network definition) to any client requesting an address via DHCP. On an isolated network this makes no sense, as we have iptables to prevent any traffic going out via that interface, so anything sent that way would be dropped anyway. This extra/unusable default route becomes problematic if you have setup a guest with multiple network interfaces, with one connected to an isolated network and another that provides connectivity to the outside (example - one interface directly connecting to a physical interface via macvtap, with a second connected to an isolated network so that the host and guest can communicate (macvtap doesn't support guest<->host communication without an external switch that supports vepa, or reflecting all traffic back)). In this case, if the guest chooses the default route of the isolated network, the guest will not be able to get network traffic beyond the host. To prevent dnsmasq from sending a default route, you can tell it to send 0 bytes of data for the default route option (option number 3) with --dhcp-option=3 (normally the data to send for the option would follow the option number; no extra data means "don't send this option"). I have checked on RHEL5 (a good representative of the oldest supported libvirt platforms) and its version of dnsmasq (2.45) does support --dhcp-option, so this shouldn't create any compatibility problems.
2011-03-13 08:42:58 +00:00
*/
if (def->forward.type == VIR_NETWORK_FORWARD_NONE) {
virBufferAddLit(&configbuf, "dhcp-option=3\n"
"no-resolv\n");
if (dnsmasqCapsGet(caps, DNSMASQ_CAPS_RA_PARAM)) {
/* interface=* (any), interval=0 (default), lifetime=0 (seconds) */
virBufferAddLit(&configbuf, "ra-param=*,0,0\n");
}
}
network driver: don't send default route to clients on isolated networks Normally dnsmasq will send a default route (the address of the host in the network definition) to any client requesting an address via DHCP. On an isolated network this makes no sense, as we have iptables to prevent any traffic going out via that interface, so anything sent that way would be dropped anyway. This extra/unusable default route becomes problematic if you have setup a guest with multiple network interfaces, with one connected to an isolated network and another that provides connectivity to the outside (example - one interface directly connecting to a physical interface via macvtap, with a second connected to an isolated network so that the host and guest can communicate (macvtap doesn't support guest<->host communication without an external switch that supports vepa, or reflecting all traffic back)). In this case, if the guest chooses the default route of the isolated network, the guest will not be able to get network traffic beyond the host. To prevent dnsmasq from sending a default route, you can tell it to send 0 bytes of data for the default route option (option number 3) with --dhcp-option=3 (normally the data to send for the option would follow the option number; no extra data means "don't send this option"). I have checked on RHEL5 (a good representative of the oldest supported libvirt platforms) and its version of dnsmasq (2.45) does support --dhcp-option, so this shouldn't create any compatibility problems.
2011-03-13 08:42:58 +00:00
if (wantDNS) {
for (i = 0; i < dns->ntxts; i++) {
virBufferAsprintf(&configbuf, "txt-record=%s,%s\n",
dns->txts[i].name,
dns->txts[i].value);
network: fix problems with SRV records A patch submitted by Steven Malin last week pointed out a problem with libvirt's DNS SRV record configuration: https://www.redhat.com/archives/libvir-list/2014-March/msg00536.html When searching for that message later, I found another series that had been posted by Guannan Ren back in 2012 that somehow slipped between the cracks: https://www.redhat.com/archives/libvir-list/2012-July/msg00236.html That patch was very much out of date, but also pointed out some real problems. This patch fixes all the noted problems by refactoring virNetworkDNSSrvDefParseXML() and networkDnsmasqConfContents(), then verifies those fixes by added several new records to the test case. Problems fixed: * both service and protocol now have an underscore ("_") prepended on the commandline, as required by RFC2782. <srv service='sip' protocol='udp' domain='example.com' target='tests.example.com' port='5060' priority='10' weight='150'/> before: srv-host=sip.udp.example.com,tests.example.com,5060,10,150 after: srv-host=_sip._udp.example.com,tests.example.com,5060,10,150 * if "domain" wasn't specified in the <srv> element, the extra trailing "." will no longer be added to the dnsmasq commandline. <srv service='sip' protocol='udp' target='tests.example.com' port='5060' priority='10' weight='150'/> before: srv-host=sip.udp.,tests.example.com,5060,10,150 after: srv-host=_sip._udp,tests.example.com,5060,10,150 * when optional attributes aren't specified, the separating comma is also now not placed on the dnsmasq commandline. If optional attributes in the middle of the line are not specified, they are replaced with a default value in the commandline (1 for port, 0 for priority and weight). <srv service='sip' protocol='udp' target='tests.example.com' port='5060'/> before: srv-host=sip.udp.,tests.example.com,5060,, after: srv-host=_sip._udp,tests.example.com,5060 (actually the would have generated an error, because "optional" attributes weren't really optional.) * The allowed characters for both service and protocol are now limited to alphanumerics, plus a few special characters that are found in existing names in /etc/services and /etc/protocols. (One exception is that both of these files contain names with an embedded ".", but "." can't be used in these fields of an SRV record because it is used as a field separator and there is no method to escape a "." into a field.) (Previously only the strings "tcp" and "udp" were allowed for protocol, but this restriction has been removed, since RFC2782 specifically says that it isn't limited to those, and that anyway it is case insensitive.) * the "domain" attribute is no longer required in order to recognize the port, priority, and weight attributes during parsing. Only "target" is required for this. * if "target" isn't specified, port, priority, and weight are not allowed (since they are meaningless - an empty target means "this service is *not available* for this domain"). * port, priority, and weight are now truly optional, as the comments originally suggested, but which was not actually true.
2014-03-18 01:16:38 +00:00
}
for (i = 0; i < dns->nsrvs; i++) {
/* service/protocol are required, and should have been validated
* by the parser.
*/
if (!dns->srvs[i].service) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Missing required 'service' "
"attribute in SRV record of network '%s'"),
def->name);
goto cleanup;
}
if (!dns->srvs[i].protocol) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Missing required 'service' "
"attribute in SRV record of network '%s'"),
def->name);
goto cleanup;
}
/* RFC2782 requires that service and protocol be preceded by
* an underscore.
*/
virBufferAsprintf(&configbuf, "srv-host=_%s._%s",
dns->srvs[i].service, dns->srvs[i].protocol);
/* domain is optional - it defaults to the domain of this network */
if (dns->srvs[i].domain)
virBufferAsprintf(&configbuf, ".%s", dns->srvs[i].domain);
/* If target is empty or ".", that means "the service is
* decidedly not available at this domain" (RFC2782). In that
* case, any port, priority, or weight is irrelevant.
network: fix problems with SRV records A patch submitted by Steven Malin last week pointed out a problem with libvirt's DNS SRV record configuration: https://www.redhat.com/archives/libvir-list/2014-March/msg00536.html When searching for that message later, I found another series that had been posted by Guannan Ren back in 2012 that somehow slipped between the cracks: https://www.redhat.com/archives/libvir-list/2012-July/msg00236.html That patch was very much out of date, but also pointed out some real problems. This patch fixes all the noted problems by refactoring virNetworkDNSSrvDefParseXML() and networkDnsmasqConfContents(), then verifies those fixes by added several new records to the test case. Problems fixed: * both service and protocol now have an underscore ("_") prepended on the commandline, as required by RFC2782. <srv service='sip' protocol='udp' domain='example.com' target='tests.example.com' port='5060' priority='10' weight='150'/> before: srv-host=sip.udp.example.com,tests.example.com,5060,10,150 after: srv-host=_sip._udp.example.com,tests.example.com,5060,10,150 * if "domain" wasn't specified in the <srv> element, the extra trailing "." will no longer be added to the dnsmasq commandline. <srv service='sip' protocol='udp' target='tests.example.com' port='5060' priority='10' weight='150'/> before: srv-host=sip.udp.,tests.example.com,5060,10,150 after: srv-host=_sip._udp,tests.example.com,5060,10,150 * when optional attributes aren't specified, the separating comma is also now not placed on the dnsmasq commandline. If optional attributes in the middle of the line are not specified, they are replaced with a default value in the commandline (1 for port, 0 for priority and weight). <srv service='sip' protocol='udp' target='tests.example.com' port='5060'/> before: srv-host=sip.udp.,tests.example.com,5060,, after: srv-host=_sip._udp,tests.example.com,5060 (actually the would have generated an error, because "optional" attributes weren't really optional.) * The allowed characters for both service and protocol are now limited to alphanumerics, plus a few special characters that are found in existing names in /etc/services and /etc/protocols. (One exception is that both of these files contain names with an embedded ".", but "." can't be used in these fields of an SRV record because it is used as a field separator and there is no method to escape a "." into a field.) (Previously only the strings "tcp" and "udp" were allowed for protocol, but this restriction has been removed, since RFC2782 specifically says that it isn't limited to those, and that anyway it is case insensitive.) * the "domain" attribute is no longer required in order to recognize the port, priority, and weight attributes during parsing. Only "target" is required for this. * if "target" isn't specified, port, priority, and weight are not allowed (since they are meaningless - an empty target means "this service is *not available* for this domain"). * port, priority, and weight are now truly optional, as the comments originally suggested, but which was not actually true.
2014-03-18 01:16:38 +00:00
*/
if (dns->srvs[i].target && STRNEQ(dns->srvs[i].target, ".")) {
virBufferAsprintf(&configbuf, ",%s", dns->srvs[i].target);
/* port, priority, and weight are optional, but are
* identified by their position in the line. If an item is
* unspecified, but something later in the line *is*
* specified, we need to give the default value for the
* unspecified item. (According to the dnsmasq manpage,
* the default for port is 1).
*/
if (dns->srvs[i].port ||
dns->srvs[i].priority || dns->srvs[i].weight)
virBufferAsprintf(&configbuf, ",%d",
dns->srvs[i].port ? dns->srvs[i].port : 1);
if (dns->srvs[i].priority || dns->srvs[i].weight)
virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].priority);
if (dns->srvs[i].weight)
virBufferAsprintf(&configbuf, ",%d", dns->srvs[i].weight);
}
virBufferAddLit(&configbuf, "\n");
}
}
/* Find the first dhcp for both IPv4 and IPv6 */
for (i = 0, ipv4def = NULL, ipv6def = NULL, ipv6SLAAC = false;
(ipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i));
i++) {
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
if (ipdef->nranges || ipdef->nhosts) {
if (ipv4def) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("For IPv4, multiple DHCP definitions "
"cannot be specified."));
goto cleanup;
} else {
ipv4def = ipdef;
}
}
}
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6)) {
if (ipdef->nranges || ipdef->nhosts) {
if (!DNSMASQ_DHCPv6_SUPPORT(caps)) {
unsigned long version = dnsmasqCapsGetVersion(caps);
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("The version of dnsmasq on this host "
"(%d.%d) doesn't adequately support "
"IPv6 dhcp range or dhcp host "
"specification. Version %d.%d or later "
"is required."),
(int)version / 1000000,
(int)(version % 1000000) / 1000,
DNSMASQ_DHCPv6_MAJOR_REQD,
DNSMASQ_DHCPv6_MINOR_REQD);
goto cleanup;
}
if (ipv6def) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("For IPv6, multiple DHCP definitions "
"cannot be specified."));
goto cleanup;
} else {
ipv6def = ipdef;
}
} else {
ipv6SLAAC = true;
}
}
}
if (ipv6def && ipv6SLAAC) {
VIR_WARN("For IPv6, when DHCP is specified for one address, then "
"state-full Router Advertising will occur. The additional "
"IPv6 addresses specified require manually configured guest "
"network to work properly since both state-full (DHCP) "
"and state-less (SLAAC) addressing are not supported "
"on the same network interface.");
}
ipdef = ipv4def ? ipv4def : ipv6def;
while (ipdef) {
int prefix;
prefix = virNetworkIPDefPrefix(ipdef);
if (prefix < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("bridge '%s' has an invalid prefix"),
def->bridge);
goto cleanup;
}
for (r = 0; r < ipdef->nranges; r++) {
int thisRange;
if (!(saddr = virSocketAddrFormat(&ipdef->ranges[r].start)) ||
!(eaddr = virSocketAddrFormat(&ipdef->ranges[r].end)))
goto cleanup;
virBufferAsprintf(&configbuf, "dhcp-range=%s,%s",
saddr, eaddr);
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
virBufferAsprintf(&configbuf, ",%d", prefix);
virBufferAddLit(&configbuf, "\n");
VIR_FREE(saddr);
VIR_FREE(eaddr);
thisRange = virSocketAddrGetRange(&ipdef->ranges[r].start,
&ipdef->ranges[r].end,
&ipdef->address,
virNetworkIPDefPrefix(ipdef));
if (thisRange < 0)
goto cleanup;
nbleases += thisRange;
}
/*
* For static-only DHCP, i.e. with no range but at least one
* host element, we have to add a special --dhcp-range option
* to enable the service in dnsmasq. (this is for dhcp-hosts=
* support)
*/
if (!ipdef->nranges && ipdef->nhosts) {
char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
if (!bridgeaddr)
goto cleanup;
virBufferAsprintf(&configbuf, "dhcp-range=%s,static",
bridgeaddr);
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
virBufferAsprintf(&configbuf, ",%d", prefix);
virBufferAddLit(&configbuf, "\n");
VIR_FREE(bridgeaddr);
}
if (networkBuildDnsmasqDhcpHostsList(dctx, ipdef) < 0)
goto cleanup;
/* Note: the following is IPv4 only */
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
if (ipdef->nranges || ipdef->nhosts) {
virBufferAddLit(&configbuf, "dhcp-no-override\n");
virBufferAddLit(&configbuf, "dhcp-authoritative\n");
}
if (ipdef->tftproot) {
virBufferAddLit(&configbuf, "enable-tftp\n");
virBufferAsprintf(&configbuf, "tftp-root=%s\n", ipdef->tftproot);
}
if (ipdef->bootfile) {
if (VIR_SOCKET_ADDR_VALID(&ipdef->bootserver)) {
char *bootserver = virSocketAddrFormat(&ipdef->bootserver);
if (!bootserver)
goto cleanup;
virBufferAsprintf(&configbuf, "dhcp-boot=%s%s%s\n",
ipdef->bootfile, ",,", bootserver);
VIR_FREE(bootserver);
} else {
virBufferAsprintf(&configbuf, "dhcp-boot=%s\n", ipdef->bootfile);
}
}
}
ipdef = (ipdef == ipv6def) ? NULL : ipv6def;
}
if (nbleases > 0)
virBufferAsprintf(&configbuf, "dhcp-lease-max=%d\n", nbleases);
/* this is done once per interface */
if (networkBuildDnsmasqHostsList(dctx, dns) < 0)
goto cleanup;
/* Even if there are currently no static hosts, if we're
* listening for DHCP, we should write a 0-length hosts
* file to allow for runtime additions.
*/
if (ipv4def || ipv6def)
virBufferAsprintf(&configbuf, "dhcp-hostsfile=%s\n",
dctx->hostsfile->path);
/* Likewise, always create this file and put it on the
* commandline, to allow for runtime additions.
*/
if (wantDNS) {
virBufferAsprintf(&configbuf, "addn-hosts=%s\n",
dctx->addnhostsfile->path);
}
/* Are we doing RA instead of radvd? */
if (DNSMASQ_RA_SUPPORT(caps)) {
if (ipv6def) {
virBufferAddLit(&configbuf, "enable-ra\n");
} else {
for (i = 0;
(ipdef = virNetworkDefGetIPByIndex(def, AF_INET6, i));
i++) {
if (!(ipdef->nranges || ipdef->nhosts)) {
char *bridgeaddr = virSocketAddrFormat(&ipdef->address);
if (!bridgeaddr)
goto cleanup;
virBufferAsprintf(&configbuf,
"dhcp-range=%s,ra-only\n", bridgeaddr);
VIR_FREE(bridgeaddr);
}
}
}
network: add 'bootp' and 'tftp' config Currently, libvirtd will start a dnsmasq process for the virtual network, but (aside from killing the dnsmasq process and replacing it), there's no way to define tftp boot options. This change introduces the appropriate tags to the dhcp configuration: <network> <name>default</name> <bridge name="virbr%d" /> <forward/> <ip address="192.168.122.1" netmask="255.255.255.0"> <tftp root="/var/lib/tftproot" /> <dhcp> <range start="192.168.122.2" end="192.168.122.254" /> <bootp file="pxeboot.img"/> </dhcp> </ip> </network> When the attributes are present, these are passed to the arguments to dnsmasq: dnsmasq [...] --enable-tftp --tftp-root /srv/tftp --dhcp-boot pxeboot.img ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^ from <tftp /> from <bootp /> At present, only local tftp servers are supported (ie, dnsmasq runs as the tftp server), but we could improve this in future by adding a server= attribute. Signed-off-by: Jeremy Kerr <jk@ozlabs.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> 2009-09-21 Paolo Bonzini <pbonzini@redhat.com> Jeremy Kerr <jk@ozlabs.org> * docs/formatnetwork.html.in: Document new tags. * docs/formatnetwork.html: Regenerate. * docs/schemas/network.rng: Update. * src/network_conf.c (virNetworkDefFree): Free new fields. (virNetworkDHCPRangeDefParseXML): Parse <bootp>. (virNetworkIPParseXML): New, parsing <dhcp> and <tftp>. (virNetworkDefParseXML): Use virNetworkIPParseXML instead of virNetworkDHCPRangeDefParseXML. (virNetworkDefFormat): Pretty print new fields. * src/network_conf.h (struct _virNetworkDef): Add netboot fields. * src/network_driver.c (networkBuildDnsmasqArgv): Add TFTP and BOOTP arguments. * tests/Makefile.am (EXTRA_DIST): Add networkschemadata. * tests/networkschematest: Look in networkschemadata. * tests/networkschemadata/netboot-network.xml: New.
2009-09-21 20:50:25 +00:00
}
if (!(*configstr = virBufferContentAndReset(&configbuf)))
goto cleanup;
ret = 0;
cleanup:
VIR_FREE(saddr);
VIR_FREE(eaddr);
virBufferFreeAndReset(&configbuf);
return ret;
}
/* build the dnsmasq command line */
static int ATTRIBUTE_NONNULL(3)
networkBuildDhcpDaemonCommandLine(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj,
virCommandPtr *cmdout,
char *pidfile,
dnsmasqContext *dctx)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
virCommandPtr cmd = NULL;
int ret = -1;
char *configfile = NULL;
char *configstr = NULL;
char *leaseshelper_path = NULL;
virNetworkObjSetDnsmasqPid(obj, -1);
if (networkDnsmasqConfContents(obj, pidfile, &configstr,
dctx, dnsmasq_caps) < 0)
goto cleanup;
if (!configstr)
goto cleanup;
/* construct the filename */
if (!(configfile = networkDnsmasqConfigFileName(driver, def->name)))
goto cleanup;
/* Write the file */
if (virFileWriteStr(configfile, configstr, 0600) < 0) {
virReportSystemError(errno,
2014-06-23 09:51:38 +00:00
_("couldn't write dnsmasq config file '%s'"),
configfile);
goto cleanup;
}
/* This helper is used to create custom leases file for libvirt */
if (!(leaseshelper_path = virFileFindResource("libvirt_leaseshelper",
abs_topbuilddir "/src",
LIBEXECDIR)))
goto cleanup;
cmd = virCommandNew(dnsmasqCapsGetBinaryPath(dnsmasq_caps));
virCommandAddArgFormat(cmd, "--conf-file=%s", configfile);
/* Libvirt gains full control of leases database */
virCommandAddArgFormat(cmd, "--leasefile-ro");
virCommandAddArgFormat(cmd, "--dhcp-script=%s", leaseshelper_path);
virCommandAddEnvPair(cmd, "VIR_BRIDGE_NAME", def->bridge);
*cmdout = cmd;
ret = 0;
cleanup:
virObjectUnref(dnsmasq_caps);
VIR_FREE(configfile);
VIR_FREE(configstr);
VIR_FREE(leaseshelper_path);
return ret;
}
static int
networkStartDhcpDaemon(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virNetworkIPDefPtr ipdef;
size_t i;
bool needDnsmasq = false;
virCommandPtr cmd = NULL;
char *pidfile = NULL;
pid_t dnsmasqPid;
int ret = -1;
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
dnsmasqContext *dctx = NULL;
/* see if there are any IP addresses that need a dhcp server */
i = 0;
while ((ipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i))) {
i++;
if (ipdef->nranges || ipdef->nhosts)
needDnsmasq = true;
}
if (i == 0) {
/* no IP addresses at all, so we don't need to run */
ret = 0;
goto cleanup;
}
if (!needDnsmasq && def->dns.enable == VIR_TRISTATE_BOOL_NO) {
/* no DHCP services needed, and user disabled DNS service */
ret = 0;
goto cleanup;
}
if (virFileMakePath(driver->pidDir) < 0) {
virReportSystemError(errno,
_("cannot create directory %s"),
driver->pidDir);
goto cleanup;
}
if (!(pidfile = virPidFileBuildPath(driver->pidDir, def->name)))
goto cleanup;
if (virFileMakePath(driver->dnsmasqStateDir) < 0) {
virReportSystemError(errno,
_("cannot create directory %s"),
driver->dnsmasqStateDir);
goto cleanup;
}
dctx = dnsmasqContextNew(def->name, driver->dnsmasqStateDir);
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
if (dctx == NULL)
goto cleanup;
if (networkDnsmasqCapsRefresh(driver) < 0)
goto cleanup;
ret = networkBuildDhcpDaemonCommandLine(driver, obj, &cmd, pidfile, dctx);
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
if (ret < 0)
goto cleanup;
ret = dnsmasqSave(dctx);
if (ret < 0)
goto cleanup;
ret = virCommandRun(cmd, NULL);
if (ret < 0)
goto cleanup;
/*
* There really is no race here - when dnsmasq daemonizes, its
* leader process stays around until its child has actually
* written its pidfile. So by time virCommandRun exits it has
* waitpid'd and guaranteed the proess has started and written a
* pid
*/
ret = virPidFileRead(driver->pidDir, def->name, &dnsmasqPid);
if (ret < 0)
goto cleanup;
virNetworkObjSetDnsmasqPid(obj, dnsmasqPid);
ret = 0;
cleanup:
VIR_FREE(pidfile);
virCommandFree(cmd);
network: Fix dnsmasq hostsfile creation logic and related tests networkSaveDnsmasqHostsfile was added in 8fa9c2214247 (Apr 2010). It has a force flag. If the dnsmasq hostsfile already exists force needs to be true to overwrite it. networkBuildDnsmasqArgv sets force to false, networkDefine sets it to true. This results in the hostsfile being written only in networkDefine in the common case. If no error occurred networkSaveDnsmasqHostsfile returns true and networkBuildDnsmasqArgv adds the --dhcp-hostsfile to the dnsmasq command line. networkSaveDnsmasqHostsfile was changed in 89ae9849f744 (24 Jun 2011) to return a new dnsmasqContext instead of reusing one. This change broke the logic of the force flag as now networkSaveDnsmasqHostsfile returns NULL on error, but the early return -- if force was not set and the hostsfile exists -- returns 0. This turned the early return in an error case and networkBuildDnsmasqArgv didn't add the --dhcp-hostsfile option anymore if the hostsfile already exists. It did because networkDefine created the hostsfile already. Then 9d4e2845d498 fixed the return 0 case in networkSaveDnsmasqHostsfile but didn't apply the force option correctly to the new addnhosts file. Now force doesn't control an early return anymore, but influences the handling of the hostsfile context creation and dnsmasqSave is always called now. This commit also added test cases that reveal several problems. First, the tests now calls functions that try to write the dnsmasq config files to disk. If someone runs this tests as root this might overwrite actively used dnsmasq config files, this is a no-go. Also the tests depend on configure --localstatedir, this needs to be fixed as well, because it makes the tests fail when localstatedir is different from /var. This patch does several things to fix this: 1) Move dnsmasqContext creation and saving out of networkBuildDnsmasqArgv to the caller to separate the command line generation from the config file writing. This makes the command line generation testable without the risk of interfering with system files, because the tests just don't call dnsmasqSave. 2) This refactoring of networkSaveDnsmasqHostsfile makes the force flag useless as the saving happens somewhere else now. This fixes the wrong usage of the force flag in combination with then newly added addnhosts file by removing the force flag. 3) Adapt the wrong test cases to the correct behavior, by adding the missing --dhcp-hostsfile option. Both affected tests contain DHCP host elements but missed the necessary --dhcp-hostsfile option. 4) Rename networkSaveDnsmasqHostsfile to networkBuildDnsmasqHostsfile, because it doesn't save the dnsmasqContext anymore. 5) Move all directory creations in dnsmasq context handling code from the *New functions to dnsmasqSave to avoid directory creations in system paths in the test cases. 6) Now that networkBuildDnsmasqArgv doesn't create the dnsmasqContext anymore the test case can create one with the localstatedir that is expected by the tests instead of the configure --localstatedir given one.
2011-06-28 11:07:59 +00:00
dnsmasqContextFree(dctx);
return ret;
}
/* networkRefreshDhcpDaemon:
* Update dnsmasq config files, then send a SIGHUP so that it rereads
* them. This only works for the dhcp-hostsfile and the
* addn-hosts file.
*
* Returns 0 on success, -1 on failure.
*/
static int
networkRefreshDhcpDaemon(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
int ret = -1;
size_t i;
pid_t dnsmasqPid;
virNetworkIPDefPtr ipdef, ipv4def, ipv6def;
dnsmasqContext *dctx = NULL;
/* if no IP addresses specified, nothing to do */
if (!virNetworkDefGetIPByIndex(def, AF_UNSPEC, 0))
return 0;
/* if there's no running dnsmasq, just start it */
dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
if (dnsmasqPid <= 0 || (kill(dnsmasqPid, 0) < 0))
return networkStartDhcpDaemon(driver, obj);
VIR_INFO("Refreshing dnsmasq for network %s", def->bridge);
if (!(dctx = dnsmasqContextNew(def->name,
driver->dnsmasqStateDir))) {
goto cleanup;
}
/* Look for first IPv4 address that has dhcp defined.
* We only support dhcp-host config on one IPv4 subnetwork
* and on one IPv6 subnetwork.
*/
ipv4def = NULL;
for (i = 0;
(ipdef = virNetworkDefGetIPByIndex(def, AF_INET, i));
i++) {
if (!ipv4def && (ipdef->nranges || ipdef->nhosts))
ipv4def = ipdef;
}
ipv6def = NULL;
for (i = 0;
(ipdef = virNetworkDefGetIPByIndex(def, AF_INET6, i));
i++) {
if (!ipv6def && (ipdef->nranges || ipdef->nhosts))
ipv6def = ipdef;
}
if (ipv4def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv4def) < 0))
2014-06-23 09:51:38 +00:00
goto cleanup;
if (ipv6def && (networkBuildDnsmasqDhcpHostsList(dctx, ipv6def) < 0))
2014-06-23 09:51:38 +00:00
goto cleanup;
if (networkBuildDnsmasqHostsList(dctx, &def->dns) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
if ((ret = dnsmasqSave(dctx)) < 0)
goto cleanup;
dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
ret = kill(dnsmasqPid, SIGHUP);
cleanup:
dnsmasqContextFree(dctx);
return ret;
}
/* networkRestartDhcpDaemon:
*
* kill and restart dnsmasq, in order to update any config that is on
* the dnsmasq commandline (and any placed in separate config files).
*
* Returns 0 on success, -1 on failure.
*/
static int
networkRestartDhcpDaemon(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
pid_t dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
/* if there is a running dnsmasq, kill it */
if (dnsmasqPid > 0) {
networkKillDaemon(dnsmasqPid, "dnsmasq", def->name);
virNetworkObjSetDnsmasqPid(obj, -1);
}
/* now start dnsmasq if it should be started */
return networkStartDhcpDaemon(driver, obj);
}
static char radvd1[] = " AdvOtherConfigFlag off;\n\n";
static char radvd2[] = " AdvAutonomous off;\n";
static char radvd3[] = " AdvOnLink on;\n"
" AdvAutonomous on;\n"
" AdvRouterAddr off;\n";
static int
networkRadvdConfContents(virNetworkObjPtr obj,
char **configstr)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virBuffer configbuf = VIR_BUFFER_INITIALIZER;
int ret = -1;
size_t i;
virNetworkIPDefPtr ipdef;
bool v6present = false, dhcp6 = false;
*configstr = NULL;
/* Check if DHCPv6 is needed */
for (i = 0; (ipdef = virNetworkDefGetIPByIndex(def, AF_INET6, i)); i++) {
v6present = true;
if (ipdef->nranges || ipdef->nhosts) {
dhcp6 = true;
break;
}
}
/* If there are no IPv6 addresses, then we are done */
if (!v6present) {
ret = 0;
goto cleanup;
}
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
/* create radvd config file appropriate for this network;
* IgnoreIfMissing allows radvd to start even when the bridge is down
*/
virBufferAsprintf(&configbuf, "interface %s\n"
"{\n"
" AdvSendAdvert on;\n"
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
" IgnoreIfMissing on;\n"
" AdvManagedFlag %s;\n"
"%s",
def->bridge,
dhcp6 ? "on" : "off",
dhcp6 ? "\n" : radvd1);
/* add a section for each IPv6 address in the config */
for (i = 0; (ipdef = virNetworkDefGetIPByIndex(def, AF_INET6, i)); i++) {
int prefix;
char *netaddr;
prefix = virNetworkIPDefPrefix(ipdef);
if (prefix < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("bridge '%s' has an invalid prefix"),
def->bridge);
goto cleanup;
}
if (!(netaddr = virSocketAddrFormat(&ipdef->address)))
goto cleanup;
virBufferAsprintf(&configbuf,
" prefix %s/%d\n"
" {\n%s };\n",
netaddr, prefix,
dhcp6 ? radvd2 : radvd3);
VIR_FREE(netaddr);
}
virBufferAddLit(&configbuf, "};\n");
if (virBufferCheckError(&configbuf) < 0)
goto cleanup;
*configstr = virBufferContentAndReset(&configbuf);
ret = 0;
cleanup:
virBufferFreeAndReset(&configbuf);
return ret;
}
/* write file and return its name (which must be freed by caller) */
static int
networkRadvdConfWrite(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj,
char **configFile)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
int ret = -1;
char *configStr = NULL;
char *myConfigFile = NULL;
if (!configFile)
configFile = &myConfigFile;
*configFile = NULL;
if (networkRadvdConfContents(obj, &configStr) < 0)
goto cleanup;
if (!configStr) {
ret = 0;
goto cleanup;
}
/* construct the filename */
if (!(*configFile = networkRadvdConfigFileName(driver, def->name)))
goto cleanup;
/* write the file */
if (virFileWriteStr(*configFile, configStr, 0600) < 0) {
virReportSystemError(errno,
_("couldn't write radvd config file '%s'"),
*configFile);
goto cleanup;
}
ret = 0;
cleanup:
VIR_FREE(configStr);
VIR_FREE(myConfigFile);
return ret;
}
static int
networkStartRadvd(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
pid_t radvdPid;
char *pidfile = NULL;
char *radvdpidbase = NULL;
char *configfile = NULL;
virCommandPtr cmd = NULL;
int ret = -1;
virNetworkObjSetRadvdPid(obj, -1);
/* Is dnsmasq handling RA? */
if (DNSMASQ_RA_SUPPORT(dnsmasq_caps)) {
ret = 0;
goto cleanup;
}
if (!virNetworkDefGetIPByIndex(def, AF_INET6, 0)) {
/* no IPv6 addresses, so we don't need to run radvd */
ret = 0;
goto cleanup;
}
if (!virFileIsExecutable(RADVD)) {
virReportSystemError(errno,
_("Cannot find %s - "
"Possibly the package isn't installed"),
RADVD);
goto cleanup;
}
if (virFileMakePath(driver->pidDir) < 0) {
virReportSystemError(errno,
_("cannot create directory %s"),
driver->pidDir);
goto cleanup;
}
if (virFileMakePath(driver->radvdStateDir) < 0) {
virReportSystemError(errno,
_("cannot create directory %s"),
driver->radvdStateDir);
goto cleanup;
}
/* construct pidfile name */
if (!(radvdpidbase = networkRadvdPidfileBasename(def->name)))
goto cleanup;
if (!(pidfile = virPidFileBuildPath(driver->pidDir, radvdpidbase)))
goto cleanup;
if (networkRadvdConfWrite(driver, obj, &configfile) < 0)
goto cleanup;
/* prevent radvd from daemonizing itself with "--debug 1", and use
* a dummy pidfile name - virCommand will create the pidfile we
* want to use (this is necessary because radvd's internal
* daemonization and pidfile creation causes a race, and the
* virPidFileRead() below will fail if we use them).
* Unfortunately, it isn't possible to tell radvd to not create
* its own pidfile, so we just let it do so, with a slightly
* different name. Unused, but harmless.
*/
cmd = virCommandNewArgList(RADVD, "--debug", "1",
"--config", configfile,
"--pidfile", NULL);
virCommandAddArgFormat(cmd, "%s-bin", pidfile);
virCommandSetPidFile(cmd, pidfile);
virCommandDaemonize(cmd);
if (virCommandRun(cmd, NULL) < 0)
goto cleanup;
if (virPidFileRead(driver->pidDir, radvdpidbase, &radvdPid) < 0)
goto cleanup;
virNetworkObjSetRadvdPid(obj, radvdPid);
ret = 0;
cleanup:
virObjectUnref(dnsmasq_caps);
virCommandFree(cmd);
VIR_FREE(configfile);
VIR_FREE(radvdpidbase);
VIR_FREE(pidfile);
return ret;
}
static int
networkRefreshRadvd(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
dnsmasqCapsPtr dnsmasq_caps = networkGetDnsmasqCaps(driver);
char *radvdpidbase;
pid_t radvdPid;
/* Is dnsmasq handling RA? */
if (DNSMASQ_RA_SUPPORT(dnsmasq_caps)) {
virObjectUnref(dnsmasq_caps);
radvdPid = virNetworkObjGetRadvdPid(obj);
if (radvdPid <= 0)
return 0;
/* radvd should not be running but in case it is */
if ((networkKillDaemon(radvdPid, "radvd", def->name) >= 0) &&
((radvdpidbase = networkRadvdPidfileBasename(def->name))
!= NULL)) {
virPidFileDelete(driver->pidDir, radvdpidbase);
VIR_FREE(radvdpidbase);
}
virNetworkObjSetRadvdPid(obj, -1);
return 0;
}
virObjectUnref(dnsmasq_caps);
/* if there's no running radvd, just start it */
radvdPid = virNetworkObjGetRadvdPid(obj);
if (radvdPid <= 0 || (kill(radvdPid, 0) < 0))
return networkStartRadvd(driver, obj);
if (!virNetworkDefGetIPByIndex(def, AF_INET6, 0)) {
/* no IPv6 addresses, so we don't need to run radvd */
return 0;
}
if (networkRadvdConfWrite(driver, obj, NULL) < 0)
return -1;
return kill(radvdPid, SIGHUP);
}
#if 0
/* currently unused, so it causes a build error unless we #if it out */
static int
networkRestartRadvd(virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
char *radvdpidbase;
pid_t radvdPid = virNeworkObjGetRadvdPid(obj);
/* if there is a running radvd, kill it */
if (radvdPid > 0) {
/* essentially ignore errors from the following two functions,
* since there's really no better recovery to be done than to
* just push ahead (and that may be exactly what's needed).
*/
if ((networkKillDaemon(radvdPid, "radvd", def->name) >= 0) &&
((radvdpidbase = networkRadvdPidfileBasename(def->name))
!= NULL)) {
virPidFileDelete(driver->pidDir, radvdpidbase);
VIR_FREE(radvdpidbase);
}
virNetworkObjSetRadvdPid(obj, -1);
}
/* now start radvd if it should be started */
return networkStartRadvd(obj);
}
#endif /* #if 0 */
static int
networkRefreshDaemonsHelper(virNetworkObjPtr obj,
void *opaque)
{
virNetworkDefPtr def;
virNetworkDriverStatePtr driver = opaque;
virObjectLock(obj);
def = virNetworkObjGetDef(obj);
if (virNetworkObjIsActive(obj) &&
((def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
(def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
(def->forward.type == VIR_NETWORK_FORWARD_ROUTE) ||
(def->forward.type == VIR_NETWORK_FORWARD_OPEN))) {
/* Only the three L3 network types that are configured by
* libvirt will have a dnsmasq or radvd daemon associated
* with them. Here we send a SIGHUP to an existing
* dnsmasq and/or radvd, or restart them if they've
* disappeared.
*/
networkRefreshDhcpDaemon(driver, obj);
networkRefreshRadvd(driver, obj);
}
virObjectUnlock(obj);
return 0;
}
/* SIGHUP/restart any dnsmasq or radvd daemons.
* This should be called when libvirtd is restarted.
*/
static void
networkRefreshDaemons(virNetworkDriverStatePtr driver)
{
VIR_INFO("Refreshing network daemons");
virNetworkObjListForEach(driver->networks,
networkRefreshDaemonsHelper,
driver);
}
static int
networkReloadFirewallRulesHelper(virNetworkObjPtr obj,
void *opaque ATTRIBUTE_UNUSED)
{
virNetworkDefPtr def;
virObjectLock(obj);
def = virNetworkObjGetDef(obj);
if (virNetworkObjIsActive(obj) &&
((def->forward.type == VIR_NETWORK_FORWARD_NONE) ||
(def->forward.type == VIR_NETWORK_FORWARD_NAT) ||
(def->forward.type == VIR_NETWORK_FORWARD_ROUTE))) {
/* Only three of the L3 network types that are configured by
* libvirt need to have iptables rules reloaded. The 4th L3
* network type, forward='open', doesn't need this because it
* has no iptables rules.
*/
networkRemoveFirewallRules(def);
if (networkAddFirewallRules(def) < 0) {
/* failed to add but already logged */
}
}
virObjectUnlock(obj);
return 0;
}
static void
networkReloadFirewallRules(virNetworkDriverStatePtr driver)
{
VIR_INFO("Reloading iptables rules");
virNetworkObjListForEach(driver->networks,
networkReloadFirewallRulesHelper,
NULL);
}
/* Enable IP Forwarding. Return 0 for success, -1 for failure. */
static int
networkEnableIPForwarding(bool enableIPv4,
bool enableIPv6)
{
int ret = 0;
#ifdef HAVE_SYSCTLBYNAME
int enabled = 1;
if (enableIPv4)
ret = sysctlbyname("net.inet.ip.forwarding", NULL, 0,
2014-06-23 09:51:38 +00:00
&enabled, sizeof(enabled));
if (enableIPv6 && ret == 0)
ret = sysctlbyname("net.inet6.ip6.forwarding", NULL, 0,
2014-06-23 09:51:38 +00:00
&enabled, sizeof(enabled));
#else
if (enableIPv4)
ret = virFileWriteStr(SYSCTL_PATH "/net/ipv4/ip_forward", "1\n", 0);
if (enableIPv6 && ret == 0)
ret = virFileWriteStr(SYSCTL_PATH "/net/ipv6/conf/all/forwarding", "1\n", 0);
#endif
return ret;
}
static int
networkSetIPv6Sysctls(virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
char *field = NULL;
int ret = -1;
bool enableIPv6 = !!virNetworkDefGetIPByIndex(def, AF_INET6, 0);
/* set disable_ipv6 if there are no ipv6 addresses defined for the
* network. But also unset it if there *are* ipv6 addresses, as we
* can't be sure of its default value.
*/
if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/disable_ipv6",
def->bridge) < 0)
goto cleanup;
if (access(field, W_OK) < 0 && errno == ENOENT) {
if (!enableIPv6)
VIR_DEBUG("ipv6 appears to already be disabled on %s",
def->bridge);
ret = 0;
goto cleanup;
}
if (virFileWriteStr(field, enableIPv6 ? "0" : "1", 0) < 0) {
virReportSystemError(errno,
_("cannot write to %s to enable/disable IPv6 "
"on bridge %s"), field, def->bridge);
goto cleanup;
}
VIR_FREE(field);
/* The rest of the ipv6 sysctl tunables should always be set the
* same, whether or not we're using ipv6 on this bridge.
*/
/* Prevent guests from hijacking the host network by sending out
* their own router advertisements.
*/
if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/accept_ra",
def->bridge) < 0)
goto cleanup;
if (virFileWriteStr(field, "0", 0) < 0) {
virReportSystemError(errno,
_("cannot disable %s"), field);
goto cleanup;
}
VIR_FREE(field);
/* All interfaces used as a gateway (which is what this is, by
* definition), must always have autoconf=0.
*/
if (virAsprintf(&field, SYSCTL_PATH "/net/ipv6/conf/%s/autoconf",
def->bridge) < 0)
goto cleanup;
if (virFileWriteStr(field, "0", 0) < 0) {
virReportSystemError(errno,
_("cannot disable %s"), field);
goto cleanup;
}
ret = 0;
cleanup:
VIR_FREE(field);
return ret;
}
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
/* add an IP address to a bridge */
static int
networkAddAddrToBridge(virNetworkObjPtr obj,
virNetworkIPDefPtr ipdef)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
int prefix = virNetworkIPDefPrefix(ipdef);
if (prefix < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("bridge '%s' has an invalid netmask or IP address"),
def->bridge);
return -1;
}
if (virNetDevIPAddrAdd(def->bridge, &ipdef->address, NULL, prefix) < 0)
return -1;
return 0;
}
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
static int
networkStartHandleMACTableManagerMode(virNetworkObjPtr obj,
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
const char *macTapIfName)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
const char *brname = def->bridge;
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
if (brname &&
def->macTableManager == VIR_NETWORK_BRIDGE_MAC_TABLE_MANAGER_LIBVIRT) {
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
if (virNetDevBridgeSetVlanFiltering(brname, true) < 0)
return -1;
if (macTapIfName) {
if (virNetDevBridgePortSetLearning(brname, macTapIfName, false) < 0)
return -1;
if (virNetDevBridgePortSetUnicastFlood(brname, macTapIfName, false) < 0)
return -1;
}
}
return 0;
}
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
/* add an IP (static) route to a bridge */
static int
networkAddRouteToBridge(virNetworkObjPtr obj,
virNetDevIPRoutePtr routedef)
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
int prefix = virNetDevIPRouteGetPrefix(routedef);
unsigned int metric = virNetDevIPRouteGetMetric(routedef);
virSocketAddrPtr addr = virNetDevIPRouteGetAddress(routedef);
virSocketAddrPtr gateway = virNetDevIPRouteGetGateway(routedef);
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
if (prefix < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' has an invalid netmask "
"or IP address in route definition"),
def->name);
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
return -1;
}
if (virNetDevIPRouteAdd(def->bridge, addr, prefix, gateway, metric) < 0)
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
return -1;
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
return 0;
}
static int
networkWaitDadFinish(virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virNetworkIPDefPtr ipdef;
virSocketAddrPtr *addrs = NULL, addr = NULL;
size_t naddrs = 0;
int ret = -1;
VIR_DEBUG("Begin waiting for IPv6 DAD on network %s", def->name);
while ((ipdef = virNetworkDefGetIPByIndex(def, AF_INET6, naddrs))) {
addr = &ipdef->address;
if (VIR_APPEND_ELEMENT_COPY(addrs, naddrs, addr) < 0)
goto cleanup;
}
ret = (naddrs == 0) ? 0 : virNetDevIPWaitDadFinish(addrs, naddrs);
cleanup:
VIR_FREE(addrs);
VIR_DEBUG("Finished waiting for IPv6 DAD on network %s with status %d",
def->name, ret);
return ret;
}
static int
networkStartNetworkVirtual(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
size_t i;
bool v4present = false, v6present = false;
virErrorPtr save_err = NULL;
virNetworkIPDefPtr ipdef;
virNetDevIPRoutePtr routedef;
char *macTapIfName = NULL;
virMacMapPtr macmap;
char *macMapFile = NULL;
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
int tapfd = -1;
pid_t dnsmasqPid;
/* Check to see if any network IP collides with an existing route */
if (networkCheckRouteCollision(def) < 0)
return -1;
/* Create and configure the bridge device */
if (!def->bridge) {
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
/* bridge name can only be empty if the config files were
* edited directly. Otherwise networkValidate() (called after
* parsing the XML from networkCreateXML() and
* networkDefine()) guarantees we will have a valid bridge
* name before this point. Since hand editing of the config
* files is explicitly prohibited we can, with clear
* conscience, log an error and fail at this point.
*/
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' has no bridge name defined"),
def->name);
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
return -1;
}
if (virNetDevBridgeCreate(def->bridge) < 0)
return -1;
if (def->mac_specified) {
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
/* To set a mac for the bridge, we need to define a dummy tap
* device, set its mac, then attach it to the bridge. As long
* as its mac address is lower than any other interface that
* gets attached, the bridge will always maintain this mac
* address.
*/
macTapIfName = networkBridgeDummyNicName(def->bridge);
if (!macTapIfName)
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
goto err0;
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
/* Keep tun fd open and interface up to allow for IPv6 DAD to happen */
if (virNetDevTapCreateInBridgePort(def->bridge,
&macTapIfName, &def->mac,
NULL, NULL, &tapfd, 1, NULL, NULL,
NULL, def->mtu, NULL,
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
VIR_NETDEV_TAP_CREATE_USE_MAC_FOR_BRIDGE |
VIR_NETDEV_TAP_CREATE_IFUP |
VIR_NETDEV_TAP_CREATE_PERSIST) < 0) {
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
VIR_FREE(macTapIfName);
goto err0;
}
}
if (!(macMapFile = virMacMapFileName(driver->dnsmasqStateDir,
def->bridge)) ||
!(macmap = virMacMapNew(macMapFile)))
goto err1;
virNetworkObjSetMacMap(obj, macmap);
/* Set bridge options */
/* delay is configured in seconds, but virNetDevBridgeSetSTPDelay
* expects milliseconds
*/
if (virNetDevBridgeSetSTPDelay(def->bridge, def->delay * 1000) < 0)
goto err1;
if (virNetDevBridgeSetSTP(def->bridge, def->stp ? true : false) < 0)
goto err1;
/* Disable IPv6 on the bridge if there are no IPv6 addresses
* defined, and set other IPv6 sysctl tunables appropriately.
*/
if (networkSetIPv6Sysctls(obj) < 0)
goto err1;
/* Add "once per network" rules */
if (def->forward.type != VIR_NETWORK_FORWARD_OPEN &&
networkAddFirewallRules(def) < 0)
goto err1;
for (i = 0; (ipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i)); i++) {
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET))
v4present = true;
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6))
v6present = true;
/* Add the IP address/netmask to the bridge */
if (networkAddAddrToBridge(obj, ipdef) < 0)
goto err2;
}
if (networkStartHandleMACTableManagerMode(obj, macTapIfName) < 0)
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
goto err2;
/* Bring up the bridge interface */
if (virNetDevSetOnline(def->bridge, 1) < 0)
goto err2;
for (i = 0; i < def->nroutes; i++) {
virSocketAddrPtr gateway = NULL;
routedef = def->routes[i];
gateway = virNetDevIPRouteGetGateway(routedef);
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
/* Add the IP route to the bridge */
/* ignore errors, error msg will be generated */
/* but libvirt will not know and net-destroy will work. */
if (VIR_SOCKET_ADDR_VALID(gateway)) {
if (networkAddRouteToBridge(obj, routedef) < 0) {
Support for static routes on a virtual bridge network: static route support for <network> This patch adds the <route> subelement of <network> to define a static route. the address and prefix (or netmask) attribute identify the destination network, and the gateway attribute specifies the next hop address (which must be directly reachable from the containing <network>) which is to receive the packets destined for "address/(prefix|netmask)". These attributes are translated into an "ip route add" command that is executed when the network is started. The command used is of the following form: ip route add <address>/<prefix> via <gateway> \ dev <virbr-bridge> proto static metric <metric> Tests are done to validate that the input data are correct. For example, for a static route ip definition, the address must be a network address and not a host address. Additional checks are added to ensure that the specified gateway is directly reachable via this network (i.e. that the gateway IP address is in the same subnet as one of the IP's defined for the network). prefix='0' is supported for both family='ipv4' address='0.0.0.0' netmask='0.0.0.0' or prefix='0', and for family='ipv6' address='::', prefix=0', although care should be taken to not override a desired system default route. Anytime an attempt is made to define a static route which *exactly* duplicates an existing static route (for example, address=::, prefix=0, metric=1), the following error message will be sent to syslog: RTNETLINK answers: File exists This can be overridden by decreasing the metric value for the route that should be preferred, or increasing the metric for the route that shouldn't be preferred (and is thus in place only in anticipation that the preferred route may be removed in the future). Caution should be used when manipulating route metrics, especially for a default route. Note: The use of the command-line interface should be replaced by direct use of libnl so that error conditions can be handled better. But, that is being left as an exercise for another day. Signed-off-by: Gene Czarcinski <gene@czarc.net> Signed-off-by: Laine Stump <laine@laine.org>
2013-05-07 17:42:55 +00:00
/* an error occurred adding the static route */
continue; /* for now, do nothing */
}
}
}
/* If forward.type != NONE, turn on global IP forwarding */
if (def->forward.type != VIR_NETWORK_FORWARD_NONE) {
if (v6present && !virNetDevIPCheckIPv6Forwarding())
goto err3; /* Precise error message already provided */
if (networkEnableIPForwarding(v4present, v6present) < 0) {
virReportSystemError(errno, "%s",
_("failed to enable IP forwarding"));
goto err3;
}
}
network driver: Start dnsmasq even if no dhcp ranges/hosts are specified. This fixes a regression introduced in commit ad48df, and reported on the libvirt-users list: https://www.redhat.com/archives/libvirt-users/2011-March/msg00018.html The problem in that commit was that we began searching a list of ip address definitions (rather than just having one) to look for a dhcp range or static host; when we didn't find any, our pointer (ipdef) was left at NULL, and when ipdef was NULL, we returned without starting up dnsmasq. Previously dnsmasq was started even without any dhcp ranges or static entries, because it's still useful for DNS services. Another problem I noticed while investigating was that, if there are IPv6 addresses, but no IPv4 addresses of any kind, we would jump out at an ever higher level in the call chain. This patch does the following: 1) networkBuildDnsmasqArgv() = all uses of ipdef are protected from NULL dereference. (this patch doesn't change indentation, to make review easier. The next patch will change just the indentation). ipdef is intended to point to the first IPv4 address with DHCP info (or the first IPv4 address if none of them have any dhcp info). 2) networkStartDhcpDaemon() = if the loop looking for an ipdef with DHCP info comes up empty, we then grab the first IPv4 def from the list. Also, instead of returning if there are no IPv4 defs, we just return if there are no IP defs at all (either v4 or v6). This way a network that is IPv6-only will still get dnsmasq listening for DNS queries. 3) in networkStartNetworkDaemon() - we will startup dhcp not just if there are any IPv4 addresses, but also if there are any IPv6 addresses.
2011-03-11 16:47:58 +00:00
/* start dnsmasq if there are any IP addresses (v4 or v6) */
if ((v4present || v6present) &&
networkStartDhcpDaemon(driver, obj) < 0)
goto err3;
/* start radvd if there are any ipv6 addresses */
if (v6present && networkStartRadvd(driver, obj) < 0)
goto err4;
/* dnsmasq does not wait for DAD to complete before daemonizing,
* so we need to wait for it ourselves.
*/
if (v6present && networkWaitDadFinish(obj) < 0)
goto err4;
/* DAD has finished, dnsmasq is now bound to the
* bridge's IPv6 address, so we can set the dummy tun down.
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
*/
if (tapfd >= 0) {
if (virNetDevSetOnline(macTapIfName, false) < 0)
goto err4;
VIR_FORCE_CLOSE(tapfd);
}
if (virNetDevBandwidthSet(def->bridge, def->bandwidth, true, true) < 0)
goto err5;
VIR_FREE(macTapIfName);
VIR_FREE(macMapFile);
return 0;
err5:
if (def->bandwidth)
virNetDevBandwidthClear(def->bridge);
err4:
if (!save_err)
save_err = virSaveLastError();
dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
if (dnsmasqPid > 0) {
kill(dnsmasqPid, SIGTERM);
virNetworkObjSetDnsmasqPid(obj, -1);
}
err3:
if (!save_err)
save_err = virSaveLastError();
ignore_value(virNetDevSetOnline(def->bridge, 0));
err2:
if (!save_err)
save_err = virSaveLastError();
if (def->forward.type != VIR_NETWORK_FORWARD_OPEN)
networkRemoveFirewallRules(def);
err1:
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
if (!save_err)
save_err = virSaveLastError();
if (macTapIfName) {
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
VIR_FORCE_CLOSE(tapfd);
ignore_value(virNetDevTapDelete(macTapIfName, NULL));
VIR_FREE(macTapIfName);
}
VIR_FREE(macMapFile);
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
err0:
if (!save_err)
save_err = virSaveLastError();
ignore_value(virNetDevBridgeDelete(def->bridge));
if (save_err) {
virSetError(save_err);
virFreeError(save_err);
}
/* coverity[leaked_handle] - 'tapfd' is not leaked */
return -1;
}
static int
networkShutdownNetworkVirtual(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
pid_t radvdPid;
pid_t dnsmasqPid;
if (def->bandwidth)
virNetDevBandwidthClear(def->bridge);
virNetworkObjUnrefMacMap(obj);
radvdPid = virNetworkObjGetRadvdPid(obj);
if (radvdPid > 0) {
char *radvdpidbase;
kill(radvdPid, SIGTERM);
/* attempt to delete the pidfile we created */
if ((radvdpidbase = networkRadvdPidfileBasename(def->name))) {
virPidFileDelete(driver->pidDir, radvdpidbase);
VIR_FREE(radvdpidbase);
}
}
dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
if (dnsmasqPid > 0)
kill(dnsmasqPid, SIGTERM);
if (def->mac_specified) {
char *macTapIfName = networkBridgeDummyNicName(def->bridge);
if (macTapIfName) {
ignore_value(virNetDevTapDelete(macTapIfName, NULL));
Give each virtual network bridge its own fixed MAC address This fixes https://bugzilla.redhat.com/show_bug.cgi?id=609463 The problem was that, since a bridge always acquires the MAC address of the connected interface with the numerically lowest MAC, as guests are started and stopped, it was possible for the MAC address to change over time, and this change in the network was being detected by Windows 7 (it sees the MAC of the default route change), so on each reboot it would bring up a dialog box asking about this "new network". The solution is to create a dummy tap interface with a MAC guaranteed to be lower than any guest interface's MAC, and attach that tap to the bridge as soon as it's created. Since all guest MAC addresses start with 0xFE, we can just generate a MAC with the standard "0x52, 0x54, 0" prefix, and it's guaranteed to always win (physical interfaces are never connected to these bridges, so we don't need to worry about competing numerically with them). Note that the dummy tap is never set to IFF_UP state - that's not necessary in order for the bridge to take its MAC, and not setting it to UP eliminates the clutter of having an (eg) "virbr0-nic" displayed in the output of the ifconfig command. I chose to not auto-generate the MAC address in the network XML parser, as there are likely to be consumers of that API that don't need or want to have a MAC address associated with the bridge. Instead, in bridge_driver.c when the network is being defined, if there is no MAC, one is generated. To account for virtual network configs that already exist when upgrading from an older version of libvirt, I've added a %post script to the specfile that searches for all network definitions in both the config directory (/etc/libvirt/qemu/networks) and the state directory (/var/lib/libvirt/network) that are missing a mac address, generates a random address, and adds it to the config (and a matching address to the state file, if there is one). docs/formatnetwork.html.in: document <mac address.../> docs/schemas/network.rng: add nac address to schema libvirt.spec.in: %post script to update existing networks src/conf/network_conf.[ch]: parse and format <mac address.../> src/libvirt_private.syms: export a couple private symbols we need src/network/bridge_driver.c: auto-generate mac address when needed, create dummy interface if mac address is present. tests/networkxml2xmlin/isolated-network.xml tests/networkxml2xmlin/routed-network.xml tests/networkxml2xmlout/isolated-network.xml tests/networkxml2xmlout/routed-network.xml: add mac address to some tests
2011-02-09 08:28:12 +00:00
VIR_FREE(macTapIfName);
}
}
ignore_value(virNetDevSetOnline(def->bridge, 0));
if (def->forward.type != VIR_NETWORK_FORWARD_OPEN)
networkRemoveFirewallRules(def);
ignore_value(virNetDevBridgeDelete(def->bridge));
/* See if its still alive and really really kill it */
dnsmasqPid = virNetworkObjGetDnsmasqPid(obj);
if (dnsmasqPid > 0 &&
(kill(dnsmasqPid, 0) == 0))
kill(dnsmasqPid, SIGKILL);
virNetworkObjSetDnsmasqPid(obj, -1);
radvdPid = virNetworkObjGetRadvdPid(obj);
if (radvdPid > 0 &&
(kill(radvdPid, 0) == 0))
kill(radvdPid, SIGKILL);
virNetworkObjSetRadvdPid(obj, -1);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
return 0;
}
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
static int
networkStartNetworkBridge(virNetworkObjPtr obj)
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
{
/* put anything here that needs to be done each time a network of
* type BRIDGE, is started. On failure, undo anything you've done,
* and return -1. On success return 0.
*/
return networkStartHandleMACTableManagerMode(obj, NULL);
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
}
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
static int
networkShutdownNetworkBridge(virNetworkObjPtr obj ATTRIBUTE_UNUSED)
network: setup bridge devices for macTableManager='libvirt' When the bridge device for a network has macTableManager='libvirt' the intent is that all kernel management of the bridge's MAC table (Forwarding Database, or fdb, in the case of a Linux Host Bridge) be disabled, with libvirt handling updates to the table instead. The setup required for the bridge itself is: 1) set the "vlan_filtering" property of the bridge device to 1. 2) If the bridge has a "Dummy" tap device used to set a fixed MAC address on the bridge (which is always the case for a bridge created by libvirt, and never the case for a bridge created by the host system network config), turn off learning and unicast_flood on this tap (this is needed even though this tap is never IFF_UP, because the kernel ignores the IFF_UP flag of devices when using their settings to automatically decide whether or not to turn off promiscuous mode for any attached device). (1) is done both for libvirt-created/managed bridges, and for bridges that are created by the host system config, while (2) is done only for bridges created by libvirt (i.e. for forward modes of nat, routed, and isolated bridges) There is no attempt to turn vlan_filtering off when destroying the network because in the case of a libvirt-created bridge, the bridge is about to be destroyed anyway, and in the case of a system bridge, if the other devices attached to the bridge could operate properly before destroying libvirt's network object, they will continue to operate properly (this is similar to the way that libvirt will enable ip_forwarding whenever a routed/natted network is started, but will never attempt to disable it if they are stopped).
2014-11-20 20:44:19 +00:00
{
/* put anything here that needs to be done each time a network of
* type BRIDGE is shutdown. On failure, undo anything you've done,
* and return -1. On success return 0.
*/
return 0;
}
/* networkCreateInterfacePool:
* @netdef: the original NetDef from the network
*
* Creates an implicit interface pool of VF's when a PF dev is given
*/
static int
networkCreateInterfacePool(virNetworkDefPtr netdef)
{
size_t numVirtFns = 0;
unsigned int maxVirtFns = 0;
char **vfNames = NULL;
virPCIDeviceAddressPtr *virtFns;
int ret = -1;
size_t i;
if (netdef->forward.npfs == 0 || netdef->forward.nifs > 0)
return 0;
if ((virNetDevGetVirtualFunctions(netdef->forward.pfs->dev, &vfNames,
&virtFns, &numVirtFns, &maxVirtFns)) < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Could not get Virtual functions on %s"),
netdef->forward.pfs->dev);
goto cleanup;
}
if (VIR_ALLOC_N(netdef->forward.ifs, numVirtFns) < 0)
goto cleanup;
for (i = 0; i < numVirtFns; i++) {
virPCIDeviceAddressPtr thisVirtFn = virtFns[i];
const char *thisName = vfNames[i];
virNetworkForwardIfDefPtr thisIf
= &netdef->forward.ifs[netdef->forward.nifs];
switch (netdef->forward.type) {
case VIR_NETWORK_FORWARD_BRIDGE:
case VIR_NETWORK_FORWARD_PRIVATE:
case VIR_NETWORK_FORWARD_VEPA:
case VIR_NETWORK_FORWARD_PASSTHROUGH:
if (thisName) {
if (VIR_STRDUP(thisIf->device.dev, thisName) < 0)
goto cleanup;
thisIf->type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV;
netdef->forward.nifs++;
} else {
VIR_WARN("VF %zu of SRIOV PF %s couldn't be added to the "
"interface pool because it isn't bound "
"to a network driver - possibly in use elsewhere",
i, netdef->forward.pfs->dev);
}
break;
case VIR_NETWORK_FORWARD_HOSTDEV:
/* VF's are always PCI devices */
thisIf->type = VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI;
thisIf->device.pci.domain = thisVirtFn->domain;
thisIf->device.pci.bus = thisVirtFn->bus;
thisIf->device.pci.slot = thisVirtFn->slot;
thisIf->device.pci.function = thisVirtFn->function;
netdef->forward.nifs++;
break;
case VIR_NETWORK_FORWARD_NONE:
case VIR_NETWORK_FORWARD_NAT:
case VIR_NETWORK_FORWARD_ROUTE:
case VIR_NETWORK_FORWARD_OPEN:
case VIR_NETWORK_FORWARD_LAST:
/* by definition these will never be encountered here */
break;
}
}
if (netdef->forward.nifs == 0) {
/* If we don't get at least one interface in the pool, declare
* failure
*/
virReportError(VIR_ERR_INTERNAL_ERROR,
_("No usable Vf's present on SRIOV PF %s"),
netdef->forward.pfs->dev);
goto cleanup;
}
ret = 0;
cleanup:
if (ret < 0) {
/* free all the entries made before error */
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].type
== VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV)
VIR_FREE(netdef->forward.ifs[i].device.dev);
}
netdef->forward.nifs = 0;
}
if (netdef->forward.nifs == 0)
VIR_FREE(netdef->forward.ifs);
for (i = 0; i < numVirtFns; i++) {
VIR_FREE(vfNames[i]);
VIR_FREE(virtFns[i]);
}
VIR_FREE(vfNames);
VIR_FREE(virtFns);
return ret;
}
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static int
networkStartNetworkExternal(virNetworkObjPtr obj)
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
{
/* put anything here that needs to be done each time a network of
* type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is started. On
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
* failure, undo anything you've done, and return -1. On success
* return 0.
*/
return networkCreateInterfacePool(virNetworkObjGetDef(obj));
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
}
static int
networkShutdownNetworkExternal(virNetworkObjPtr obj ATTRIBUTE_UNUSED)
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
{
/* put anything here that needs to be done each time a network of
* type BRIDGE, PRIVATE, VEPA, HOSTDEV or PASSTHROUGH is shutdown. On
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
* failure, undo anything you've done, and return -1. On success
* return 0.
*/
return 0;
}
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
static int
networkStartNetwork(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
int ret = -1;
VIR_DEBUG("driver=%p, network=%p", driver, obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
if (virNetworkObjIsActive(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID,
"%s", _("network is already active"));
return ret;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
}
VIR_DEBUG("Beginning network startup process");
VIR_DEBUG("Setting current network def as transient");
if (virNetworkObjSetDefTransient(obj, true) < 0)
goto cleanup;
/* Run an early hook to set-up missing devices.
* If the script raised an error abort the launch. */
if (networkRunHook(obj, NULL, NULL,
VIR_HOOK_NETWORK_OP_START,
VIR_HOOK_SUBOP_BEGIN) < 0)
goto cleanup;
switch (def->forward.type) {
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
case VIR_NETWORK_FORWARD_NONE:
case VIR_NETWORK_FORWARD_NAT:
case VIR_NETWORK_FORWARD_ROUTE:
case VIR_NETWORK_FORWARD_OPEN:
if (networkStartNetworkVirtual(driver, obj) < 0)
goto cleanup;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
break;
case VIR_NETWORK_FORWARD_BRIDGE:
if (def->bridge) {
if (networkStartNetworkBridge(obj) < 0)
goto cleanup;
break;
}
/* intentionally fall through to the macvtap/direct case for
* VIR_NETWORK_FORWARD_BRIDGE with no bridge device defined
* (since that is macvtap bridge mode).
*/
ATTRIBUTE_FALLTHROUGH;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
case VIR_NETWORK_FORWARD_PRIVATE:
case VIR_NETWORK_FORWARD_VEPA:
case VIR_NETWORK_FORWARD_PASSTHROUGH:
case VIR_NETWORK_FORWARD_HOSTDEV:
if (networkStartNetworkExternal(obj) < 0)
goto cleanup;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
break;
}
/* finally we can call the 'started' hook script if any */
if (networkRunHook(obj, NULL, NULL,
VIR_HOOK_NETWORK_OP_STARTED,
VIR_HOOK_SUBOP_BEGIN) < 0)
goto cleanup;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
/* Persist the live configuration now that anything autogenerated
* is setup.
*/
VIR_DEBUG("Writing network status to disk");
if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0)
goto cleanup;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
virNetworkObjSetActive(obj, true);
VIR_INFO("Network '%s' started up", def->name);
ret = 0;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
cleanup:
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
if (ret < 0) {
virNetworkObjUnsetDefTransient(obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
virErrorPtr save_err = virSaveLastError();
int save_errno = errno;
networkShutdownNetwork(driver, obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
virSetError(save_err);
virFreeError(save_err);
errno = save_errno;
}
return ret;
}
static int
networkShutdownNetwork(virNetworkDriverStatePtr driver,
virNetworkObjPtr obj)
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
int ret = 0;
char *stateFile;
VIR_INFO("Shutting down network '%s'", def->name);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
if (!virNetworkObjIsActive(obj))
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
return 0;
stateFile = virNetworkConfigFile(driver->stateDir, def->name);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
if (!stateFile)
return -1;
unlink(stateFile);
VIR_FREE(stateFile);
switch (def->forward.type) {
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
case VIR_NETWORK_FORWARD_NONE:
case VIR_NETWORK_FORWARD_NAT:
case VIR_NETWORK_FORWARD_ROUTE:
case VIR_NETWORK_FORWARD_OPEN:
ret = networkShutdownNetworkVirtual(driver, obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
break;
case VIR_NETWORK_FORWARD_BRIDGE:
if (def->bridge) {
ret = networkShutdownNetworkBridge(obj);
break;
}
/* intentionally fall through to the macvtap/direct case for
* VIR_NETWORK_FORWARD_BRIDGE with no bridge device defined
* (since that is macvtap bridge mode).
*/
ATTRIBUTE_FALLTHROUGH;
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
case VIR_NETWORK_FORWARD_PRIVATE:
case VIR_NETWORK_FORWARD_VEPA:
case VIR_NETWORK_FORWARD_PASSTHROUGH:
case VIR_NETWORK_FORWARD_HOSTDEV:
ret = networkShutdownNetworkExternal(obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
break;
}
/* now that we know it's stopped call the hook if present */
networkRunHook(obj, NULL, NULL, VIR_HOOK_NETWORK_OP_STOPPED,
VIR_HOOK_SUBOP_END);
virNetworkObjSetActive(obj, false);
virNetworkObjUnsetDefTransient(obj);
network: separate Start/Shutdown functions for new network types Previously all networks were composed of bridge devices created and managed by libvirt, and the same operations needed to be done for all of them when they were started and stopped (create and start the bridge device, configure its MAC address and IP address, add iptables rules). The new network types are (for now at least) managed outside of libvirt, and the network object is used only to contain information about the network, which is then used as each individual guest connects itself. This means that when starting/stopping one of these new networks, we really want to do nothing, aside from marking the network as active/inactive. This has been setup as toplevel Start/Shutdown functions that do the small bit of common stuff, then have a switch statement to execute network type-specific start/shutdown code, then do a bit more common code. The type-specific functions called for the new host bridge and macvtap based types are currently empty. In the future these functions may actually do something, and we will surely add more functions that are similarly patterned. Once everything has settled, we can make a table of "sub-driver" function pointers for each network type, and store a pointer to that table in the network object, then we can replace the switch statements with calls to functions in the table. The final step in this will be to add a new table (and corresponding new functions) for new network types as they are added.
2011-06-30 21:05:07 +00:00
return ret;
}
static virNetworkPtr
networkLookupByUUID(virConnectPtr conn,
const unsigned char *uuid)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
virNetworkPtr net = NULL;
obj = virNetworkObjFindByUUID(driver->networks, uuid);
if (!obj) {
char uuidstr[VIR_UUID_STRING_BUFLEN];
virUUIDFormat(uuid, uuidstr);
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching uuid '%s'"),
uuidstr);
goto cleanup;
}
def = virNetworkObjGetDef(obj);
if (virNetworkLookupByUUIDEnsureACL(conn, def) < 0)
goto cleanup;
net = virGetNetwork(conn, def->name, def->uuid);
cleanup:
virNetworkObjEndAPI(&obj);
return net;
}
static virNetworkPtr
networkLookupByName(virConnectPtr conn,
const char *name)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
virNetworkPtr net = NULL;
obj = virNetworkObjFindByName(driver->networks, name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"), name);
goto cleanup;
}
def = virNetworkObjGetDef(obj);
if (virNetworkLookupByNameEnsureACL(conn, def) < 0)
goto cleanup;
net = virGetNetwork(conn, def->name, def->uuid);
cleanup:
virNetworkObjEndAPI(&obj);
return net;
}
static int
networkConnectNumOfNetworks(virConnectPtr conn)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int nactive;
if (virConnectNumOfNetworksEnsureACL(conn) < 0)
return -1;
nactive = virNetworkObjListNumOfNetworks(driver->networks,
true,
virConnectNumOfNetworksCheckACL,
conn);
return nactive;
}
static int
networkConnectListNetworks(virConnectPtr conn,
char **const names,
int maxnames)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int got = 0;
if (virConnectListNetworksEnsureACL(conn) < 0)
return -1;
got = virNetworkObjListGetNames(driver->networks,
true, names, maxnames,
virConnectListNetworksCheckACL,
conn);
return got;
}
static int
networkConnectNumOfDefinedNetworks(virConnectPtr conn)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int ninactive = 0;
if (virConnectNumOfDefinedNetworksEnsureACL(conn) < 0)
return -1;
ninactive = virNetworkObjListNumOfNetworks(driver->networks,
false,
virConnectNumOfDefinedNetworksCheckACL,
conn);
return ninactive;
}
static int
networkConnectListDefinedNetworks(virConnectPtr conn,
char **const names,
int maxnames)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int got = 0;
if (virConnectListDefinedNetworksEnsureACL(conn) < 0)
return -1;
got = virNetworkObjListGetNames(driver->networks,
false, names, maxnames,
virConnectListDefinedNetworksCheckACL,
conn);
return got;
}
static int
networkConnectListAllNetworks(virConnectPtr conn,
virNetworkPtr **nets,
unsigned int flags)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int ret = -1;
virCheckFlags(VIR_CONNECT_LIST_NETWORKS_FILTERS_ALL, -1);
if (virConnectListAllNetworksEnsureACL(conn) < 0)
goto cleanup;
ret = virNetworkObjListExport(conn, driver->networks, nets,
virConnectListAllNetworksCheckACL,
flags);
cleanup:
return ret;
}
static int
networkConnectNetworkEventRegisterAny(virConnectPtr conn,
virNetworkPtr net,
int eventID,
virConnectNetworkEventGenericCallback callback,
void *opaque,
virFreeCallback freecb)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int ret = -1;
if (virConnectNetworkEventRegisterAnyEnsureACL(conn) < 0)
goto cleanup;
if (virNetworkEventStateRegisterID(conn, driver->networkEventState,
net, eventID, callback,
opaque, freecb, &ret) < 0)
ret = -1;
cleanup:
return ret;
}
static int
networkConnectNetworkEventDeregisterAny(virConnectPtr conn,
int callbackID)
{
virNetworkDriverStatePtr driver = networkGetDriver();
int ret = -1;
if (virConnectNetworkEventDeregisterAnyEnsureACL(conn) < 0)
goto cleanup;
event: make deregister return value match docs Ever since their introduction (commit 1509b80 in v0.5.0 for virConnectDomainEventRegister, commit 4445723 in v0.8.0 for virConnectDomainEventDeregisterAny), the event deregistration functions have been documented as returning 0 on success; likewise for older registration (only the newer RegisterAny must return a non-zero callbackID). And now that we are adding virConnectNetworkEventDeregisterAny for v1.2.1, it should have the same semantics. Fortunately, all of the stateful drivers have been obeying the docs and returning 0, thanks to the way the remote_driver tracks things (in fact, the RPC wire protocol is unable to send a return value for DomainEventRegisterAny, at least not without adding a new RPC number). Well, except for vbox, which was always failing deregistration, due to failure to set the return value to anything besides its initial -1. But for local drivers, such as test:///default, we've been returning non-zero numbers; worse, the non-zero numbers have differed over time. For example, in Fedora 12 (libvirt 0.8.2), calling Register twice would return 0 and 1 [the callbackID generated under the hood]; while in Fedora 20 (libvirt 1.1.3), it returns 1 and 2 [the number of callbacks registered for that event type]. Since we have changed the behavior over time, and since it differs by local vs. remote, we can safely argue that no one could have been reasonably relying on any particular behavior, so we might as well obey the docs, as well as prepare callers that might deal with older clients to not be surprised if the docs are not strictly followed. For consistency, this patch fixes the code for all drivers, even though it only makes an impact for vbox and for local drivers. By fixing all drivers, future copy and paste from a remote driver to a local driver is less likely to reintroduce the bug. Finally, update the testsuite to gain some coverage of the issue for local drivers, including the first test of old-style domain event registration via function pointer instead of event id. * src/libvirt.c (virConnectDomainEventRegister) (virConnectDomainEventDeregister) (virConnectDomainEventDeregisterAny): Clarify docs. * src/libxl/libxl_driver.c (libxlConnectDomainEventRegister) (libxlConnectDomainEventDeregister) (libxlConnectDomainEventDeregisterAny): Match documentation. * src/lxc/lxc_driver.c (lxcConnectDomainEventRegister) (lxcConnectDomainEventDeregister) (lxcConnectDomainEventDeregisterAny): Likewise. * src/test/test_driver.c (testConnectDomainEventRegister) (testConnectDomainEventDeregister) (testConnectDomainEventDeregisterAny) (testConnectNetworkEventDeregisterAny): Likewise. * src/uml/uml_driver.c (umlConnectDomainEventRegister) (umlConnectDomainEventDeregister) (umlConnectDomainEventDeregisterAny): Likewise. * src/vbox/vbox_tmpl.c (vboxConnectDomainEventRegister) (vboxConnectDomainEventDeregister) (vboxConnectDomainEventDeregisterAny): Likewise. * src/xen/xen_driver.c (xenUnifiedConnectDomainEventRegister) (xenUnifiedConnectDomainEventDeregister) (xenUnifiedConnectDomainEventDeregisterAny): Likewise. * src/network/bridge_driver.c (networkConnectNetworkEventDeregisterAny): Likewise. * tests/objecteventtest.c (testDomainCreateXMLOld): New test. (mymain): Run it. (testDomainCreateXML): Check return values. Signed-off-by: Eric Blake <eblake@redhat.com>
2014-01-03 21:21:17 +00:00
if (virObjectEventStateDeregisterID(conn,
driver->networkEventState,
events: Avoid double free possibility on remote call failure If a remote call fails during event registration (more than likely from a network failure or remote libvirtd restart timed just right), then when calling the virObjectEventStateDeregisterID we don't want to call the registered @freecb function because that breaks our contract that we would only call it after succesfully returning. If the @freecb routine were called, it could result in a double free from properly coded applications that free their opaque data on failure to register, as seen in the following details: Program terminated with signal 6, Aborted. #0 0x00007fc45cba15d7 in raise #1 0x00007fc45cba2cc8 in abort #2 0x00007fc45cbe12f7 in __libc_message #3 0x00007fc45cbe86d3 in _int_free #4 0x00007fc45d8d292c in PyDict_Fini #5 0x00007fc45d94f46a in Py_Finalize #6 0x00007fc45d960735 in Py_Main #7 0x00007fc45cb8daf5 in __libc_start_main #8 0x0000000000400721 in _start The double dereference of 'pyobj_cbData' is triggered in the following way: (1) libvirt_virConnectDomainEventRegisterAny is invoked. (2) the event is successfully added to the event callback list (virDomainEventStateRegisterClient in remoteConnectDomainEventRegisterAny returns 1 which means ok). (3) when function remoteConnectDomainEventRegisterAny is hit, network connection disconnected coincidently (or libvirtd is restarted) in the context of function 'call' then the connection is lost and the function 'call' failed, the branch virObjectEventStateDeregisterID is therefore taken. (4) 'pyobj_conn' is dereferenced the 1st time in libvirt_virConnectDomainEventFreeFunc. (5) 'pyobj_cbData' (refered to pyobj_conn) is dereferenced the 2nd time in libvirt_virConnectDomainEventRegisterAny. (6) the double free error is triggered. Resolve this by adding a @doFreeCb boolean in order to avoid calling the freeCb in virObjectEventStateDeregisterID for any remote call failure in a remoteConnect*EventRegister* API. For remoteConnect*EventDeregister* calls, the passed value would be true indicating they should run the freecb if it exists; whereas, it's false for the remote call failure path. Patch based on the investigation and initial patch posted by fangying <fangying1@huawei.com>.
2017-06-14 11:32:15 +00:00
callbackID, true) < 0)
event: make deregister return value match docs Ever since their introduction (commit 1509b80 in v0.5.0 for virConnectDomainEventRegister, commit 4445723 in v0.8.0 for virConnectDomainEventDeregisterAny), the event deregistration functions have been documented as returning 0 on success; likewise for older registration (only the newer RegisterAny must return a non-zero callbackID). And now that we are adding virConnectNetworkEventDeregisterAny for v1.2.1, it should have the same semantics. Fortunately, all of the stateful drivers have been obeying the docs and returning 0, thanks to the way the remote_driver tracks things (in fact, the RPC wire protocol is unable to send a return value for DomainEventRegisterAny, at least not without adding a new RPC number). Well, except for vbox, which was always failing deregistration, due to failure to set the return value to anything besides its initial -1. But for local drivers, such as test:///default, we've been returning non-zero numbers; worse, the non-zero numbers have differed over time. For example, in Fedora 12 (libvirt 0.8.2), calling Register twice would return 0 and 1 [the callbackID generated under the hood]; while in Fedora 20 (libvirt 1.1.3), it returns 1 and 2 [the number of callbacks registered for that event type]. Since we have changed the behavior over time, and since it differs by local vs. remote, we can safely argue that no one could have been reasonably relying on any particular behavior, so we might as well obey the docs, as well as prepare callers that might deal with older clients to not be surprised if the docs are not strictly followed. For consistency, this patch fixes the code for all drivers, even though it only makes an impact for vbox and for local drivers. By fixing all drivers, future copy and paste from a remote driver to a local driver is less likely to reintroduce the bug. Finally, update the testsuite to gain some coverage of the issue for local drivers, including the first test of old-style domain event registration via function pointer instead of event id. * src/libvirt.c (virConnectDomainEventRegister) (virConnectDomainEventDeregister) (virConnectDomainEventDeregisterAny): Clarify docs. * src/libxl/libxl_driver.c (libxlConnectDomainEventRegister) (libxlConnectDomainEventDeregister) (libxlConnectDomainEventDeregisterAny): Match documentation. * src/lxc/lxc_driver.c (lxcConnectDomainEventRegister) (lxcConnectDomainEventDeregister) (lxcConnectDomainEventDeregisterAny): Likewise. * src/test/test_driver.c (testConnectDomainEventRegister) (testConnectDomainEventDeregister) (testConnectDomainEventDeregisterAny) (testConnectNetworkEventDeregisterAny): Likewise. * src/uml/uml_driver.c (umlConnectDomainEventRegister) (umlConnectDomainEventDeregister) (umlConnectDomainEventDeregisterAny): Likewise. * src/vbox/vbox_tmpl.c (vboxConnectDomainEventRegister) (vboxConnectDomainEventDeregister) (vboxConnectDomainEventDeregisterAny): Likewise. * src/xen/xen_driver.c (xenUnifiedConnectDomainEventRegister) (xenUnifiedConnectDomainEventDeregister) (xenUnifiedConnectDomainEventDeregisterAny): Likewise. * src/network/bridge_driver.c (networkConnectNetworkEventDeregisterAny): Likewise. * tests/objecteventtest.c (testDomainCreateXMLOld): New test. (mymain): Run it. (testDomainCreateXML): Check return values. Signed-off-by: Eric Blake <eblake@redhat.com>
2014-01-03 21:21:17 +00:00
goto cleanup;
ret = 0;
cleanup:
return ret;
}
static int
networkIsActive(virNetworkPtr net)
{
virNetworkObjPtr obj;
int ret = -1;
if (!(obj = networkObjFromNetwork(net)))
return ret;
if (virNetworkIsActiveEnsureACL(net->conn, virNetworkObjGetDef(obj)) < 0)
goto cleanup;
ret = virNetworkObjIsActive(obj);
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkIsPersistent(virNetworkPtr net)
{
virNetworkObjPtr obj;
int ret = -1;
if (!(obj = networkObjFromNetwork(net)))
return ret;
if (virNetworkIsPersistentEnsureACL(net->conn, virNetworkObjGetDef(obj)) < 0)
goto cleanup;
ret = virNetworkObjIsPersistent(obj);
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
/*
* networkFindUnusedBridgeName() - try to find a bridge name that is
* unused by the currently configured libvirt networks, as well as by
* the host system itself (possibly created by someone/something other
* than libvirt). Set this network's name to that new name.
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
*/
static int
networkFindUnusedBridgeName(virNetworkObjListPtr nets,
virNetworkDefPtr def)
{
int ret = -1, id = 0;
char *newname = NULL;
const char *templ = "virbr%d";
const char *p;
if (def->bridge &&
(p = strchr(def->bridge, '%')) == strrchr(def->bridge, '%') &&
p && p[1] == 'd')
templ = def->bridge;
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
do {
if (virAsprintf(&newname, templ, id) < 0)
goto cleanup;
/* check if this name is used in another libvirt network or
* there is an existing device with that name. ignore errors
* from virNetDevExists(), just in case it isn't implemented
* on this platform (probably impossible).
*/
if (!(virNetworkObjBridgeInUse(nets, newname, def->name) ||
virNetDevExists(newname) == 1)) {
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
VIR_FREE(def->bridge); /*could contain template */
def->bridge = newname;
ret = 0;
goto cleanup;
}
VIR_FREE(newname);
} while (++id <= MAX_BRIDGE_ID);
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Bridge generation exceeded max id %d"),
MAX_BRIDGE_ID);
ret = 0;
cleanup:
if (ret < 0)
VIR_FREE(newname);
return ret;
}
/*
* networkValidateBridgeName() - if no bridge name is set, or if the
* bridge name contains a %d (indicating that this is a template for
* the actual name) try to set an appropriate bridge name. If a
* bridge name *is* set, make sure it doesn't conflict with any other
* network's bridge name.
*/
static int
networkBridgeNameValidate(virNetworkObjListPtr nets,
virNetworkDefPtr def)
{
int ret = -1;
if (def->bridge && !strstr(def->bridge, "%d")) {
if (virNetworkObjBridgeInUse(nets, def->bridge, def->name)) {
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
virReportError(VIR_ERR_INTERNAL_ERROR,
_("bridge name '%s' already in use."),
def->bridge);
goto cleanup;
}
} else {
/* Allocate a bridge name */
if (networkFindUnusedBridgeName(nets, def) < 0)
goto cleanup;
}
ret = 0;
cleanup:
return ret;
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
static int
networkValidate(virNetworkDriverStatePtr driver,
virNetworkDefPtr def)
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
{
size_t i, j;
bool vlanUsed, vlanAllowed, badVlanUse = false;
virPortGroupDefPtr defaultPortGroup = NULL;
virNetworkIPDefPtr ipdef;
bool ipv4def = false, ipv6def = false;
bool bandwidthAllowed = true;
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
bool usesInterface = false, usesAddress = false;
if (virXMLCheckIllegalChars("name", def->name, "\n") < 0)
return -1;
/* Only the three L3 network types that are configured by libvirt
* need to have a bridge device name / mac address provided
*/
if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
def->forward.type == VIR_NETWORK_FORWARD_NAT ||
def->forward.type == VIR_NETWORK_FORWARD_ROUTE ||
def->forward.type == VIR_NETWORK_FORWARD_OPEN) {
network: move auto-assign of bridge name from XML parser to net driver We already check that any auto-assigned bridge device name for a virtual network (e.g. "virbr1") doesn't conflict with the bridge name for any existing libvirt network (via virNetworkSetBridgeName() in conf/network_conf.c). We also want to check that the name doesn't conflict with any bridge device created on the host system outside the control of libvirt (history: possibly due to the ploriferation of references to libvirt's bridge devices in HOWTO documents all around the web, it is not uncommon for an admin to manually create a bridge in their host's system network config and name it "virbrX"). To add such a check to virNetworkBridgeInUse() (which is called by virNetworkSetBridgeName()) we would have to call virNetDevExists() (from util/virnetdev.c); this function calls ioctl(SIOCGIFFLAGS), which everyone on the mailing list agreed should not be done from an XML parsing function in the conf directory. To remedy that problem, this patch removes virNetworkSetBridgeName() from conf/network_conf.c and puts an identically functioning networkBridgeNameValidate() in network/bridge_driver.c (because it's reasonable for the bridge driver to call virNetDevExists(), although we don't do that yet because I wanted this patch to have as close to 0 effect on function as possible). There are a couple of inevitable changes though: 1) We no longer check the bridge name during virNetworkLoadConfig(). Close examination of the code shows that this wasn't necessary anyway - the only *correct* way to get XML into the config files is via networkDefine(), and networkDefine() will always call networkValidate(), which previously called virNetworkSetBridgeName() (and now calls networkBridgeNameValidate()). This means that the only way the bridge name can be unset during virNetworkLoadConfig() is if someone edited the config file on disk by hand (which we explicitly prohibit). 2) Just on the off chance that somebody *has* edited the file by hand, rather than crashing when they try to start their malformed network, a check for non-NULL bridge name has been added to networkStartNetworkVirtual(). (For those wondering why I don't instead call networkValidateBridgeName() there to set a bridge name if one wasn't present - the problem is that during networkStartNetworkVirtual(), the lock for the network being started has already been acquired, but the lock for the network list itself *has not* (because we aren't adding/removing a network). But virNetworkBridgeInuse() iterates through *all* networks (including this one) and locks each network as it is checked for a duplicate entry; it is necessary to lock each network even before checking if it is the designated "skip" network because otherwise some other thread might acquire the list lock and delete the very entry we're examining. In the end, permitting a setting of the bridge name during network start would require that we lock the entire network list during any networkStartNetwork(), which eliminates a *lot* of parallelism that we've worked so hard to achieve (it can make a huge difference during libvirtd startup). So rather than try to adjust for someone playing against the rules, I choose to instead give them the error they deserve.) 3) virNetworkAllocateBridge() (now removed) would leak any "template" string set as the bridge name. Its replacement networkFindUnusedBridgeName() doesn't leak the template string - it is properly freed.
2015-04-23 16:49:59 +00:00
/* if no bridge name was given in the config, find a name
* unused by any other libvirt networks and assign it.
*/
if (networkBridgeNameValidate(driver->networks, def) < 0)
return -1;
virNetworkSetBridgeMacAddr(def);
} else {
/* They are also the only types that currently support setting
network: disallow <bandwidth>/<mac> for bridged/macvtap/hostdev networks https://bugzilla.redhat.com/show_bug.cgi?id=1057321 pointed out that we weren't honoring the <bandwidth> element in libvirt networks using <forward mode='bridge'/>. In fact, these networks are just a method of giving a libvirt network name to an existing Linux host bridge on the system, and libvirt doesn't have enough information to know where to set such limits. We are working on a method of supporting network bandwidths for some specific cases of <forward mode='bridge'/>, but currently libvirt doesn't support it. So the proper thing to do now is just log an error when someone tries to put a <bandwidth> element in that type of network. (It's unclear if we will be able to do proper bandwidth limiting for macvtap networks, and most definitely we will not be able to support it for hostdev networks). While looking through the network XML documentation and comparing it to the networkValidate function, I noticed that we also ignore the presence of a mac address in the config in the same cases, rather than failing so that the user will understand that their desired action has not been taken. This patch updates networkValidate() (which is called any time a persistent network is defined, or a transient network created) to log an error and fail if it finds either a <bandwidth> or <mac> element and the network forward mode is anything except 'route'. 'nat', or nothing. (Yes, neither of those elements is acceptable for any macvtap mode, nor for a hostdev network). NB: This does *not* cause failure to start any existing network that contains one of those elements, so someone might have erroneously defined such a network in the past, and that network will continue to function unmodified. I considered it too disruptive to suddenly break working configs on the next reboot after a libvirt upgrade.
2014-01-24 11:58:05 +00:00
* a MAC or IP address for the host-side device (bridge), DNS
* configuration, or network-wide bandwidth limits.
*/
network: disallow <bandwidth>/<mac> for bridged/macvtap/hostdev networks https://bugzilla.redhat.com/show_bug.cgi?id=1057321 pointed out that we weren't honoring the <bandwidth> element in libvirt networks using <forward mode='bridge'/>. In fact, these networks are just a method of giving a libvirt network name to an existing Linux host bridge on the system, and libvirt doesn't have enough information to know where to set such limits. We are working on a method of supporting network bandwidths for some specific cases of <forward mode='bridge'/>, but currently libvirt doesn't support it. So the proper thing to do now is just log an error when someone tries to put a <bandwidth> element in that type of network. (It's unclear if we will be able to do proper bandwidth limiting for macvtap networks, and most definitely we will not be able to support it for hostdev networks). While looking through the network XML documentation and comparing it to the networkValidate function, I noticed that we also ignore the presence of a mac address in the config in the same cases, rather than failing so that the user will understand that their desired action has not been taken. This patch updates networkValidate() (which is called any time a persistent network is defined, or a transient network created) to log an error and fail if it finds either a <bandwidth> or <mac> element and the network forward mode is anything except 'route'. 'nat', or nothing. (Yes, neither of those elements is acceptable for any macvtap mode, nor for a hostdev network). NB: This does *not* cause failure to start any existing network that contains one of those elements, so someone might have erroneously defined such a network in the past, and that network will continue to function unmodified. I considered it too disruptive to suddenly break working configs on the next reboot after a libvirt upgrade.
2014-01-24 11:58:05 +00:00
if (def->mac_specified) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported <mac> element in network %s "
"with forward mode='%s'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
if (virNetworkDefGetIPByIndex(def, AF_UNSPEC, 0)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported <ip> element in network %s "
"with forward mode='%s'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
if (def->dns.ntxts || def->dns.nhosts || def->dns.nsrvs) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported <dns> element in network %s "
"with forward mode='%s'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
if (def->domain) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported <domain> element in network %s "
"with forward mode='%s'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
network: disallow <bandwidth>/<mac> for bridged/macvtap/hostdev networks https://bugzilla.redhat.com/show_bug.cgi?id=1057321 pointed out that we weren't honoring the <bandwidth> element in libvirt networks using <forward mode='bridge'/>. In fact, these networks are just a method of giving a libvirt network name to an existing Linux host bridge on the system, and libvirt doesn't have enough information to know where to set such limits. We are working on a method of supporting network bandwidths for some specific cases of <forward mode='bridge'/>, but currently libvirt doesn't support it. So the proper thing to do now is just log an error when someone tries to put a <bandwidth> element in that type of network. (It's unclear if we will be able to do proper bandwidth limiting for macvtap networks, and most definitely we will not be able to support it for hostdev networks). While looking through the network XML documentation and comparing it to the networkValidate function, I noticed that we also ignore the presence of a mac address in the config in the same cases, rather than failing so that the user will understand that their desired action has not been taken. This patch updates networkValidate() (which is called any time a persistent network is defined, or a transient network created) to log an error and fail if it finds either a <bandwidth> or <mac> element and the network forward mode is anything except 'route'. 'nat', or nothing. (Yes, neither of those elements is acceptable for any macvtap mode, nor for a hostdev network). NB: This does *not* cause failure to start any existing network that contains one of those elements, so someone might have erroneously defined such a network in the past, and that network will continue to function unmodified. I considered it too disruptive to suddenly break working configs on the next reboot after a libvirt upgrade.
2014-01-24 11:58:05 +00:00
if (def->bandwidth) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported network-wide <bandwidth> element "
"in network %s with forward mode='%s'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
bandwidthAllowed = false;
}
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
/* we support configs with a single PF defined:
* <pf dev='eth0'/>
* or with a list of netdev names:
* <interface dev='eth9'/>
* OR a list of PCI addresses
* <address type='pci' domain='0' bus='4' slot='0' function='1'/>
* but not any combination of those.
*
* Since <interface> and <address> are for some strange reason
* stored in the same array, we need to cycle through it and check
* the type of each.
*/
for (i = 0; i < def->forward.nifs; i++) {
virNetworkForwardIfDefPtr iface = &def->forward.ifs[i];
char *sysfs_path = NULL;
switch ((virNetworkForwardHostdevDeviceType) iface->type) {
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV:
usesInterface = true;
if (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("hostdev network '%s' lists '%s' "
"in the device pool, but hostdev "
"networks require all devices to "
"be listed by PCI address, not "
"network device name"),
def->name, iface->device.dev);
return -1;
}
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
break;
case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI: {
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
usesAddress = true;
if (def->forward.type != VIR_NETWORK_FORWARD_HOSTDEV) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("network '%s' has forward mode '%s' "
" but lists a device by PCI address "
"in the device pool. This is only "
"supported for networks with forward "
"mode 'hostdev'"),
def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
if (virPCIDeviceAddressGetSysfsFile(&iface->device.pci, &sysfs_path) < 0)
return -1;
if (!virPCIIsVirtualFunction(sysfs_path)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("device '%s' in network '%s' is not "
"an SR-IOV Virtual Function"),
sysfs_path, def->name);
VIR_FREE(sysfs_path);
return -1;
}
VIR_FREE(sysfs_path);
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
break;
}
network: allow <pf> together with <interface>/<address> in network status The function that parses the <forward> subelement of a network used to fail/log an error if the network definition contained both a <pf> element as well as at least one <interface> or <address> element. That check was present because the configuration of a network should have either one <pf>, one or more <interface>, or one or more <address>, but never combinations of multiple kinds. This caused a problem when libvirtd was restarted with a network already active - when a network with a <pf> element is started, the referenced PF (Physical Function of an SRIOV-capable network card) is checked for VFs (Virtual Functions), and the <forward> is filled in with a list of all VFs for that PF either in the form of their PCI addresses (a list of <address>) or their netdev names (a list of <interface>); the <pf> element is not removed though. When libvirtd is restarted, it parses the network status and finds both the original <pf> from the config, as well as the list of either <address> or <interface>, fails the parse, and the network is not added to the active list. This failure is often obscured because the network is marked as autostart so libvirt immediately restarts it. It seems odd to me that <interface> and <address> are stored in the same array rather than keeping two separate arrays, and having separate arrays would have made the check much simpler. However, changing to use two separate arrays would have required changes in more places, potentially creating more conflicts and (more importantly) more possible regressions in the event of a backport, so I chose to keep the existing data structure in order to localize the change. It appears that this problem has been in the code ever since support for <pf> was added (0.9.10), but until commit 34cc3b2f106e296df5e64309620c79d16fd76c85 (first in libvirt 1.2.4) networks with interface pools were not properly marked as active on restart anyway, so there is no point in backporting this patch any further than that.
2015-02-15 03:43:16 +00:00
case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NONE:
case VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_LAST:
break;
}
}
if ((def->forward.npfs > 0) + usesInterface + usesAddress > 1) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("<address>, <interface>, and <pf> elements of "
"<forward> in network %s are mutually exclusive"),
def->name);
return -1;
}
/* We only support dhcp on one IPv4 address and
* on one IPv6 address per defined network
*/
for (i = 0;
(ipdef = virNetworkDefGetIPByIndex(def, AF_UNSPEC, i));
i++) {
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET)) {
if (ipdef->nranges || ipdef->nhosts) {
if (ipv4def) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("Multiple IPv4 dhcp sections found -- "
"dhcp is supported only for a "
"single IPv4 address on each network"));
return -1;
} else {
ipv4def = true;
}
}
}
if (VIR_SOCKET_ADDR_IS_FAMILY(&ipdef->address, AF_INET6)) {
if (ipdef->nranges || ipdef->nhosts) {
if (ipv6def) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("Multiple IPv6 dhcp sections found -- "
"dhcp is supported only for a "
"single IPv6 address on each network"));
return -1;
} else {
ipv6def = true;
}
}
}
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
/* The only type of networks that currently support transparent
* vlan configuration are those using hostdev sr-iov devices from
* a pool, and those using an Open vSwitch bridge.
*/
vlanAllowed = (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV ||
def->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH ||
(def->forward.type == VIR_NETWORK_FORWARD_BRIDGE &&
2014-06-23 09:51:38 +00:00
def->virtPortProfile &&
def->virtPortProfile->virtPortType
== VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH));
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
vlanUsed = def->vlan.nTags > 0;
for (i = 0; i < def->nPortGroups; i++) {
if (vlanUsed || def->portGroups[i].vlan.nTags > 0) {
/* anyone using this portgroup will get a vlan tag. Verify
* that they will also be using an openvswitch connection,
* as that is the only type of network that currently
* supports a vlan tag.
*/
if (def->portGroups[i].virtPortProfile) {
if (def->forward.type != VIR_NETWORK_FORWARD_BRIDGE ||
def->portGroups[i].virtPortProfile->virtPortType
!= VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
badVlanUse = true;
}
} else if (!vlanAllowed) {
/* virtualport taken from base network definition */
badVlanUse = true;
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
}
if (def->portGroups[i].isDefault) {
if (defaultPortGroup) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("network '%s' has multiple default "
"<portgroup> elements (%s and %s), "
"but only one default is allowed"),
def->name, defaultPortGroup->name,
def->portGroups[i].name);
return -1;
}
defaultPortGroup = &def->portGroups[i];
}
for (j = i + 1; j < def->nPortGroups; j++) {
if (STREQ(def->portGroups[i].name, def->portGroups[j].name)) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("multiple <portgroup> elements with the "
"same name (%s) in network '%s'"),
def->portGroups[i].name, def->name);
return -1;
}
}
if (def->portGroups[i].bandwidth && !bandwidthAllowed) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("Unsupported <bandwidth> element in network '%s' "
"in portgroup '%s' with forward mode='%s'"),
def->name, def->portGroups[i].name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
}
if (badVlanUse ||
(vlanUsed && !vlanAllowed && !defaultPortGroup)) {
/* NB: if defaultPortGroup is set, we don't directly look at
* vlanUsed && !vlanAllowed, because the network will never be
* used without having a portgroup added in, so all necessary
* checks were done in the loop above.
*/
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("<vlan> element specified for network %s, "
"whose type doesn't support vlan configuration"),
def->name);
return -1;
}
if (def->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
for (i = 0; i < def->nPortGroups; i++) {
if (def->portGroups[i].bandwidth) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("unsupported <bandwidth> element "
"in <portgroup name='%s'> of "
"network '%s' with forward mode='%s'"),
def->portGroups[i].name, def->name,
virNetworkForwardTypeToString(def->forward.type));
return -1;
}
}
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
return 0;
}
static virNetworkPtr
networkCreateXML(virConnectPtr conn,
const char *xml)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkDefPtr newDef;
virNetworkObjPtr obj = NULL;
virNetworkDefPtr def;
virNetworkPtr net = NULL;
virObjectEventPtr event = NULL;
if (!(newDef = virNetworkDefParseString(xml)))
goto cleanup;
if (virNetworkCreateXMLEnsureACL(conn, newDef) < 0)
goto cleanup;
if (networkValidate(driver, newDef) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
network: fix virNetworkObjAssignDef and persistence Experimentation showed that if virNetworkCreateXML() was called for a network that was already defined, and then the network was subsequently shutdown, the network would continue to be persistent after the shutdown (expected/desired), but the original config would be lost in favor of the transient config sent in with virNetworkCreateXML() (which would then be the new persistent config) (obviously unexpected/not desired). To fix this, virNetworkObjAssignDef() has been changed to 1) properly save/free network->def and network->newDef for all the various combinations of live/active/persistent, including some combinations that were previously considered to be an error but didn't need to be (e.g. setting a "live" config for a network that isn't yet active but soon will be - that was previously considered an error, even though in practice it can be very useful). 2) automatically set the persistent flag whenever a new non-live config is assigned to the network (and clear it when the non-live config is set to NULL). the libvirt network driver no longer directly manipulates network->persistent, but instead relies entirely on virNetworkObjAssignDef() to do the right thing automatically. After this patch, the following sequence will behave as expected: virNetworkDefineXML(X) virNetworkCreateXML(X') (same name but some config different) virNetworkDestroy(X) At the end of these calls, the network config will remain as it was after the initial virNetworkDefine(), whereas previously it would take on the changes given during virNetworkCreateXML(). Another effect of this tighter coupling between a) setting a !live def and b) setting/clearing the "persistent" flag, is that future patches which change the details of network lifecycle management (e.g. upcoming patches to fix detection of "active" networks when libvirtd is restarted) will find it much more difficult to break persistence functionality.
2014-04-22 13:48:54 +00:00
/* NB: even though this transient network hasn't yet been started,
* we assign the def with live = true in anticipation that it will
* be started momentarily.
*/
if (!(obj = virNetworkObjAssignDef(driver->networks, newDef,
VIR_NETWORK_OBJ_LIST_ADD_LIVE |
VIR_NETWORK_OBJ_LIST_ADD_CHECK_LIVE)))
goto cleanup;
newDef = NULL;
def = virNetworkObjGetDef(obj);
if (networkStartNetwork(driver, obj) < 0) {
virNetworkObjRemoveInactive(driver->networks, obj);
goto cleanup;
}
event = virNetworkEventLifecycleNew(def->name,
def->uuid,
VIR_NETWORK_EVENT_STARTED,
0);
VIR_INFO("Creating network '%s'", def->name);
net = virGetNetwork(conn, def->name, def->uuid);
cleanup:
virNetworkDefFree(newDef);
if (event)
virObjectEventStateQueue(driver->networkEventState, event);
virNetworkObjEndAPI(&obj);
return net;
}
static virNetworkPtr
networkDefineXML(virConnectPtr conn,
const char *xml)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkDefPtr def = NULL;
bool freeDef = true;
virNetworkObjPtr obj = NULL;
virNetworkPtr net = NULL;
virObjectEventPtr event = NULL;
if (!(def = virNetworkDefParseString(xml)))
goto cleanup;
if (virNetworkDefineXMLEnsureACL(conn, def) < 0)
goto cleanup;
if (networkValidate(driver, def) < 0)
2014-06-23 09:51:38 +00:00
goto cleanup;
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
if (!(obj = virNetworkObjAssignDef(driver->networks, def, 0)))
2014-06-23 09:51:38 +00:00
goto cleanup;
network: fix virNetworkObjAssignDef and persistence Experimentation showed that if virNetworkCreateXML() was called for a network that was already defined, and then the network was subsequently shutdown, the network would continue to be persistent after the shutdown (expected/desired), but the original config would be lost in favor of the transient config sent in with virNetworkCreateXML() (which would then be the new persistent config) (obviously unexpected/not desired). To fix this, virNetworkObjAssignDef() has been changed to 1) properly save/free network->def and network->newDef for all the various combinations of live/active/persistent, including some combinations that were previously considered to be an error but didn't need to be (e.g. setting a "live" config for a network that isn't yet active but soon will be - that was previously considered an error, even though in practice it can be very useful). 2) automatically set the persistent flag whenever a new non-live config is assigned to the network (and clear it when the non-live config is set to NULL). the libvirt network driver no longer directly manipulates network->persistent, but instead relies entirely on virNetworkObjAssignDef() to do the right thing automatically. After this patch, the following sequence will behave as expected: virNetworkDefineXML(X) virNetworkCreateXML(X') (same name but some config different) virNetworkDestroy(X) At the end of these calls, the network config will remain as it was after the initial virNetworkDefine(), whereas previously it would take on the changes given during virNetworkCreateXML(). Another effect of this tighter coupling between a) setting a !live def and b) setting/clearing the "persistent" flag, is that future patches which change the details of network lifecycle management (e.g. upcoming patches to fix detection of "active" networks when libvirtd is restarted) will find it much more difficult to break persistence functionality.
2014-04-22 13:48:54 +00:00
/* def was assigned to network object */
freeDef = false;
if (virNetworkSaveConfig(driver->networkConfigDir, def) < 0) {
if (!virNetworkObjIsActive(obj)) {
virNetworkObjRemoveInactive(driver->networks, obj);
goto cleanup;
}
network: fix virNetworkObjAssignDef and persistence Experimentation showed that if virNetworkCreateXML() was called for a network that was already defined, and then the network was subsequently shutdown, the network would continue to be persistent after the shutdown (expected/desired), but the original config would be lost in favor of the transient config sent in with virNetworkCreateXML() (which would then be the new persistent config) (obviously unexpected/not desired). To fix this, virNetworkObjAssignDef() has been changed to 1) properly save/free network->def and network->newDef for all the various combinations of live/active/persistent, including some combinations that were previously considered to be an error but didn't need to be (e.g. setting a "live" config for a network that isn't yet active but soon will be - that was previously considered an error, even though in practice it can be very useful). 2) automatically set the persistent flag whenever a new non-live config is assigned to the network (and clear it when the non-live config is set to NULL). the libvirt network driver no longer directly manipulates network->persistent, but instead relies entirely on virNetworkObjAssignDef() to do the right thing automatically. After this patch, the following sequence will behave as expected: virNetworkDefineXML(X) virNetworkCreateXML(X') (same name but some config different) virNetworkDestroy(X) At the end of these calls, the network config will remain as it was after the initial virNetworkDefine(), whereas previously it would take on the changes given during virNetworkCreateXML(). Another effect of this tighter coupling between a) setting a !live def and b) setting/clearing the "persistent" flag, is that future patches which change the details of network lifecycle management (e.g. upcoming patches to fix detection of "active" networks when libvirtd is restarted) will find it much more difficult to break persistence functionality.
2014-04-22 13:48:54 +00:00
/* if network was active already, just undo new persistent
* definition by making it transient.
* XXX - this isn't necessarily the correct thing to do.
*/
virNetworkObjUpdateAssignDef(obj, NULL, false);
goto cleanup;
}
event = virNetworkEventLifecycleNew(def->name, def->uuid,
VIR_NETWORK_EVENT_DEFINED,
0);
VIR_INFO("Defining network '%s'", def->name);
net = virGetNetwork(conn, def->name, def->uuid);
cleanup:
if (event)
virObjectEventStateQueue(driver->networkEventState, event);
if (freeDef)
2014-06-23 09:51:38 +00:00
virNetworkDefFree(def);
virNetworkObjEndAPI(&obj);
return net;
}
static int
networkUndefine(virNetworkPtr net)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
int ret = -1;
bool active = false;
virObjectEventPtr event = NULL;
if (!(obj = networkObjFromNetwork(net)))
goto cleanup;
def = virNetworkObjGetDef(obj);
if (virNetworkUndefineEnsureACL(net->conn, def) < 0)
goto cleanup;
if (virNetworkObjIsActive(obj))
active = true;
if (!virNetworkObjIsPersistent(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID, "%s",
_("can't undefine transient network"));
goto cleanup;
}
network: fix virNetworkObjAssignDef and persistence Experimentation showed that if virNetworkCreateXML() was called for a network that was already defined, and then the network was subsequently shutdown, the network would continue to be persistent after the shutdown (expected/desired), but the original config would be lost in favor of the transient config sent in with virNetworkCreateXML() (which would then be the new persistent config) (obviously unexpected/not desired). To fix this, virNetworkObjAssignDef() has been changed to 1) properly save/free network->def and network->newDef for all the various combinations of live/active/persistent, including some combinations that were previously considered to be an error but didn't need to be (e.g. setting a "live" config for a network that isn't yet active but soon will be - that was previously considered an error, even though in practice it can be very useful). 2) automatically set the persistent flag whenever a new non-live config is assigned to the network (and clear it when the non-live config is set to NULL). the libvirt network driver no longer directly manipulates network->persistent, but instead relies entirely on virNetworkObjAssignDef() to do the right thing automatically. After this patch, the following sequence will behave as expected: virNetworkDefineXML(X) virNetworkCreateXML(X') (same name but some config different) virNetworkDestroy(X) At the end of these calls, the network config will remain as it was after the initial virNetworkDefine(), whereas previously it would take on the changes given during virNetworkCreateXML(). Another effect of this tighter coupling between a) setting a !live def and b) setting/clearing the "persistent" flag, is that future patches which change the details of network lifecycle management (e.g. upcoming patches to fix detection of "active" networks when libvirtd is restarted) will find it much more difficult to break persistence functionality.
2014-04-22 13:48:54 +00:00
/* remove autostart link */
if (virNetworkObjDeleteConfig(driver->networkConfigDir,
driver->networkAutostartDir,
obj) < 0)
goto cleanup;
event = virNetworkEventLifecycleNew(def->name,
def->uuid,
VIR_NETWORK_EVENT_UNDEFINED,
0);
VIR_INFO("Undefining network '%s'", def->name);
if (!active) {
if (networkRemoveInactive(driver, obj) < 0)
goto cleanup;
network: fix virNetworkObjAssignDef and persistence Experimentation showed that if virNetworkCreateXML() was called for a network that was already defined, and then the network was subsequently shutdown, the network would continue to be persistent after the shutdown (expected/desired), but the original config would be lost in favor of the transient config sent in with virNetworkCreateXML() (which would then be the new persistent config) (obviously unexpected/not desired). To fix this, virNetworkObjAssignDef() has been changed to 1) properly save/free network->def and network->newDef for all the various combinations of live/active/persistent, including some combinations that were previously considered to be an error but didn't need to be (e.g. setting a "live" config for a network that isn't yet active but soon will be - that was previously considered an error, even though in practice it can be very useful). 2) automatically set the persistent flag whenever a new non-live config is assigned to the network (and clear it when the non-live config is set to NULL). the libvirt network driver no longer directly manipulates network->persistent, but instead relies entirely on virNetworkObjAssignDef() to do the right thing automatically. After this patch, the following sequence will behave as expected: virNetworkDefineXML(X) virNetworkCreateXML(X') (same name but some config different) virNetworkDestroy(X) At the end of these calls, the network config will remain as it was after the initial virNetworkDefine(), whereas previously it would take on the changes given during virNetworkCreateXML(). Another effect of this tighter coupling between a) setting a !live def and b) setting/clearing the "persistent" flag, is that future patches which change the details of network lifecycle management (e.g. upcoming patches to fix detection of "active" networks when libvirtd is restarted) will find it much more difficult to break persistence functionality.
2014-04-22 13:48:54 +00:00
} else {
/* if the network still exists, it was active, and we need to make
* it transient (by deleting the persistent def)
*/
virNetworkObjUpdateAssignDef(obj, NULL, false);
}
ret = 0;
cleanup:
if (event)
virObjectEventStateQueue(driver->networkEventState, event);
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkUpdate(virNetworkPtr net,
unsigned int command,
unsigned int section,
int parentIndex,
const char *xml,
unsigned int flags)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj = NULL;
virNetworkDefPtr def;
int isActive, ret = -1;
size_t i;
virNetworkIPDefPtr ipdef;
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
bool oldDhcpActive = false;
bool needFirewallRefresh = false;
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
virCheckFlags(VIR_NETWORK_UPDATE_AFFECT_LIVE |
VIR_NETWORK_UPDATE_AFFECT_CONFIG,
-1);
if (!(obj = networkObjFromNetwork(net)))
goto cleanup;
def = virNetworkObjGetDef(obj);
if (virNetworkUpdateEnsureACL(net->conn, def, flags) < 0)
goto cleanup;
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
/* see if we are listening for dhcp pre-modification */
for (i = 0;
(ipdef = virNetworkDefGetIPByIndex(def, AF_INET, i));
i++) {
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
if (ipdef->nranges || ipdef->nhosts) {
oldDhcpActive = true;
break;
}
}
/* VIR_NETWORK_UPDATE_AFFECT_CURRENT means "change LIVE if network
* is active, else change CONFIG
2014-06-23 09:51:38 +00:00
*/
isActive = virNetworkObjIsActive(obj);
network: fix dnsmasq/radvd binding to IPv6 on recent kernels I hit this problem recently when trying to create a bridge with an IPv6 address on a 3.2 kernel: dnsmasq (and, further, radvd) would not bind to the given address, waiting 20s and then giving up with -EADDRNOTAVAIL (resp. exiting immediately with "error parsing or activating the config file", without libvirt noticing it, BTW). This can be reproduced with (I think) any kernel >= 2.6.39 and the following XML (to be used with "virsh net-create"): <network> <name>test-bridge</name> <bridge name='testbr0' /> <ip family='ipv6' address='fd00::1' prefix='64'> </ip> </network> (it happens even when you have an IPv4, too) The problem is that since commit [1] (which, ironically, was made to “help IPv6 autoconfiguration”) the linux bridge code makes bridges behave like “real” devices regarding carrier detection. This makes the bridges created by libvirt, which are started without any up devices, stay with the NO-CARRIER flag set, and thus prevents DAD (Duplicate address detection) from happening, thus letting the IPv6 address flagged as “tentative”. Such addresses cannot be bound to (see RFC 2462), so dnsmasq fails binding to it (for radvd, it detects that "interface XXX is not RUNNING", thus that "interface XXX does not exist, ignoring the interface" (sic)). It seems that this behavior was enhanced somehow with commit [2] by avoiding setting NO-CARRIER on empty bridges, but I couldn't reproduce this behavior on my kernel. Anyway, with the “dummy tap to set MAC address” trick, this wouldn't work. To fix this, the idea is to get the bridge's attached device to be up so that DAD can happen (deactivating DAD altogether is not a good idea, I think). Currently, libvirt creates a dummy TAP device to set the MAC address of the bridge, keeping it down. But even if we set this device up, it is not RUNNING as soon as the tap file descriptor attached to it is closed, thus still preventing DAD. So, we must modify the API a bit, so that we can get the fd, keep the tap device persistent, run the daemons, and close it after DAD has taken place. After that, the bridge will be flagged NO-CARRIER again, but the daemons will be running, even if not happy about the device's state (but we don't really care about the bridge's daemons doing anything when no up interface is connected to it). Other solutions that I envisioned were: * Keeping the *-nic interface up: this would waste an fd for each bridge during all its life. May be acceptable, I don't really know. * Stop using the dummy tap trick, and set the MAC address directly on the bridge: it is possible since quite some time it seems, even if then there is the problem of the bridge not being RUNNING when empty, contrary to what [2] says, so this will need fixing (and this fix only happened in 3.1, so it wouldn't work for 2.6.39) * Using the --interface option of dnsmasq, but I saw somewhere that it's not used by libvirt for backward compatibility. I am not sure this would solve this problem, though, as I don't know how dnsmasq binds itself to it with this option. This is why this patch does what's described earlier. This patch also makes radvd start even if the interface is “missing” (i.e. it is not RUNNING), as it daemonizes before binding to it, and thus sometimes does it after the interface has been brought down by us (by closing the tap fd), and then originally stops. This also makes it stop yelling about it in the logs when the interface is down at a later time. [1] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=1faa4356a3bd89ea11fb92752d897cff3a20ec0e [2] http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=commit;h=b64b73d7d0c480f75684519c6134e79d50c1b341
2012-09-26 19:02:20 +00:00
if ((flags & (VIR_NETWORK_UPDATE_AFFECT_LIVE |
VIR_NETWORK_UPDATE_AFFECT_CONFIG)) ==
VIR_NETWORK_UPDATE_AFFECT_CURRENT) {
if (isActive)
flags |= VIR_NETWORK_UPDATE_AFFECT_LIVE;
else
flags |= VIR_NETWORK_UPDATE_AFFECT_CONFIG;
}
if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
/* Take care of anything that must be done before updating the
* live NetworkDef.
*/
if (def->forward.type == VIR_NETWORK_FORWARD_NONE ||
def->forward.type == VIR_NETWORK_FORWARD_NAT ||
def->forward.type == VIR_NETWORK_FORWARD_ROUTE) {
switch (section) {
case VIR_NETWORK_SECTION_FORWARD:
case VIR_NETWORK_SECTION_FORWARD_INTERFACE:
case VIR_NETWORK_SECTION_IP:
case VIR_NETWORK_SECTION_IP_DHCP_RANGE:
case VIR_NETWORK_SECTION_IP_DHCP_HOST:
/* these could affect the firewall rules, so remove the
* old rules (and remember to load new ones after the
* update).
*/
if (def->forward.type != VIR_NETWORK_FORWARD_OPEN) {
networkRemoveFirewallRules(def);
needFirewallRefresh = true;
}
break;
default:
break;
}
}
}
/* update the network config in memory/on disk */
if (virNetworkObjUpdate(obj, command, section, parentIndex, xml, flags) < 0) {
if (needFirewallRefresh)
ignore_value(networkAddFirewallRules(def));
goto cleanup;
}
/* @def is replaced */
def = virNetworkObjGetDef(obj);
if (needFirewallRefresh && networkAddFirewallRules(def) < 0)
goto cleanup;
if (flags & VIR_NETWORK_UPDATE_AFFECT_CONFIG) {
/* save updated persistent config to disk */
if (virNetworkSaveConfig(driver->networkConfigDir,
virNetworkObjGetPersistentDef(obj)) < 0) {
goto cleanup;
}
}
if (isActive && (flags & VIR_NETWORK_UPDATE_AFFECT_LIVE)) {
/* rewrite dnsmasq host files, restart dnsmasq, update iptables
* rules, etc, according to which section was modified. Note that
* some sections require multiple actions, so a single switch
* statement is inadequate.
*/
if (section == VIR_NETWORK_SECTION_BRIDGE ||
section == VIR_NETWORK_SECTION_DOMAIN ||
section == VIR_NETWORK_SECTION_IP ||
section == VIR_NETWORK_SECTION_IP_DHCP_RANGE ||
section == VIR_NETWORK_SECTION_DNS_TXT ||
section == VIR_NETWORK_SECTION_DNS_SRV) {
/* these sections all change things on the dnsmasq
* commandline (i.e. in the .conf file), so we need to
* kill and restart dnsmasq, because dnsmasq sets its uid
* to "nobody" after it starts, and is unable to re-read
* the conf file (owned by root, mode 600)
*/
if (networkRestartDhcpDaemon(driver, obj) < 0)
goto cleanup;
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
} else if (section == VIR_NETWORK_SECTION_IP_DHCP_HOST) {
/* if we previously weren't listening for dhcp and now we
* are (or vice-versa) then we need to do a restart,
* otherwise we just need to do a refresh (redo the config
* files and send SIGHUP)
*/
bool newDhcpActive = false;
for (i = 0; (ipdef = virNetworkDefGetIPByIndex(def, AF_INET, i));
i++) {
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
if (ipdef->nranges || ipdef->nhosts) {
newDhcpActive = true;
break;
}
}
if ((newDhcpActive != oldDhcpActive &&
networkRestartDhcpDaemon(driver, obj) < 0) ||
networkRefreshDhcpDaemon(driver, obj) < 0) {
network: always create dnsmasq hosts and addnhosts files, even if empty This fixes the problem reported in: https://bugzilla.redhat.com/show_bug.cgi?id=868389 Previously, the dnsmasq hosts file (used for static dhcp entries, and addnhosts file (used for additional dns host entries) were only created/referenced on the dnsmasq commandline if there was something to put in them at the time the network was started. Once we can update a network definition while it's active (which is now possible with virNetworkUpdate), this is no longer a valid strategy - if there were 0 dhcp static hosts (resulting in no reference to the hosts file on the commandline), then one was later added, the commandline wouldn't have linked dnsmasq up to the file, so even though we create it, dnsmasq doesn't pay any attention. The solution is to just always create these files and reference them on the dnsmasq commandline (almost always, anyway). That way dnsmasq can notice when a new entry is added at runtime (a SIGHUP is sent to dnsmasq by virNetworkUdpate whenever a host entry is added or removed) The exception to this is that the dhcp static hosts file isn't created if there are no lease ranges *and* no static hosts. This is because in this case dnsmasq won't be setup to listen for dhcp requests anyway - in that case, if the count of dhcp hosts goes from 0 to 1, dnsmasq will need to be restarted anyway (to get it listening on the dhcp port). Likewise, if the dhcp hosts count goes from 1 to 0 (and there are no dhcp ranges) we need to restart dnsmasq so that it will stop listening on port 67. These special situations are handled in the bridge driver's networkUpdate() by checking for ((bool) nranges||nhosts) both before and after the update, and triggering a dnsmasq restart if the before and after don't match.
2012-10-19 20:15:44 +00:00
goto cleanup;
}
} else if (section == VIR_NETWORK_SECTION_DNS_HOST) {
/* this section only changes data in an external file
* (not the .conf file) so we can just update the config
* files and send SIGHUP to dnsmasq.
*/
if (networkRefreshDhcpDaemon(driver, obj) < 0)
goto cleanup;
}
if (section == VIR_NETWORK_SECTION_IP) {
/* only a change in IP addresses will affect radvd, and all of radvd's
* config is stored in the conf file which will be re-read with a SIGHUP.
*/
if (networkRefreshRadvd(driver, obj) < 0)
goto cleanup;
}
/* save current network state to disk */
if ((ret = virNetworkObjSaveStatus(driver->stateDir, obj)) < 0)
goto cleanup;
}
/* call the 'updated' network hook script */
if (networkRunHook(obj, NULL, NULL, VIR_HOOK_NETWORK_OP_UPDATED,
VIR_HOOK_SUBOP_BEGIN) < 0)
goto cleanup;
ret = 0;
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkCreate(virNetworkPtr net)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
int ret = -1;
virObjectEventPtr event = NULL;
if (!(obj = networkObjFromNetwork(net)))
goto cleanup;
def = virNetworkObjGetDef(obj);
if (virNetworkCreateEnsureACL(net->conn, def) < 0)
goto cleanup;
if ((ret = networkStartNetwork(driver, obj)) < 0)
goto cleanup;
event = virNetworkEventLifecycleNew(def->name,
def->uuid,
VIR_NETWORK_EVENT_STARTED,
0);
cleanup:
if (event)
virObjectEventStateQueue(driver->networkEventState, event);
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkDestroy(virNetworkPtr net)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
int ret = -1;
virObjectEventPtr event = NULL;
if (!(obj = networkObjFromNetwork(net)))
goto cleanup;
def = virNetworkObjGetDef(obj);
if (virNetworkDestroyEnsureACL(net->conn, def) < 0)
goto cleanup;
if (!virNetworkObjIsActive(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID,
_("network '%s' is not active"),
def->name);
goto cleanup;
}
if ((ret = networkShutdownNetwork(driver, obj)) < 0)
goto cleanup;
/* @def replaced in virNetworkObjUnsetDefTransient*/
def = virNetworkObjGetDef(obj);
event = virNetworkEventLifecycleNew(def->name,
def->uuid,
VIR_NETWORK_EVENT_STOPPED,
0);
if (!virNetworkObjIsPersistent(obj) &&
networkRemoveInactive(driver, obj) < 0) {
ret = -1;
goto cleanup;
}
cleanup:
if (event)
virObjectEventStateQueue(driver->networkEventState, event);
virNetworkObjEndAPI(&obj);
return ret;
}
static char *
networkGetXMLDesc(virNetworkPtr net,
unsigned int flags)
drivers: prefer unsigned int for flags Now that the public APIs always use unsigned flags, the internal driver callbacks might as well do likewise. * src/driver.h (vrDrvOpen, virDrvDomainCoreDump) (virDrvDomainGetXMLDesc, virDrvNetworkGetXMLDesc) (virDrvNWFilterGetXMLDesc): Update type. * src/remote/remote_protocol.x (remote_open_args) (remote_domain_core_dump_args, remote_domain_get_xml_desc_args) (remote_network_get_xml_desc_args) (remote_nwfilter_get_xml_desc_args): Likewise. * src/test/test_driver.c: Update clients. * src/remote/remote_driver.c: Likewise. * src/xen/xen_hypervisor.c: Likewise. * src/xen/xen_hypervisor.h: Likewise. * src/xen/xen_driver.c: Likewise. * src/xen/xend_internal.c: Likewise. * src/xen/xend_internal.h: Likewise. * src/xen/xm_internal.c: Likewise. * src/xen/xm_internal.h: Likewise. * src/xen/xs_internal.c: Likewise. * src/xen/xs_internal.h: Likewise. * src/xen/xen_inotify.c: Likewise. * src/xen/xen_inotify.h: Likewise. * src/phyp/phyp_driver.c: Likewise. * src/openvz/openvz_driver.c: Likewise. * src/vmware/vmware_driver.c: Likewise. * src/vbox/vbox_driver.c: Likewise. * src/vbox/vbox_tmpl.c: Likewise. * src/xenapi/xenapi_driver.c: Likewise. * src/esx/esx_driver.c: Likewise. * src/esx/esx_interface_driver.c: Likewise. * src/esx/esx_network_driver.c: Likewise. * src/esx/esx_storage_driver.c: Likewise. * src/esx/esx_device_monitor.c: Likewise. * src/esx/esx_secret_driver.c: Likewise. * src/esx/esx_nwfilter_driver.c: Likewise. * src/interface/netcf_driver.c: Likewise. * src/nwfilter/nwfilter_driver.c: Likewise. * src/libxl/libxl_driver.c: Likewise. * src/qemu/qemu_driver.c: Likewise. * src/lxc/lxc_driver.c: Likewise. * src/uml/uml_driver.c: Likewise. * src/network/bridge_driver.c: Likewise. * src/secret/secret_driver.c: Likewise. * src/storage/storage_driver.c: Likewise. * src/node_device/node_device_hal.c: Likewise. * src/node_device/node_device_udev.c: Likewise. * src/remote_protocol-structs: Likewise.
2011-07-06 20:40:19 +00:00
{
virNetworkObjPtr obj;
virNetworkDefPtr curDef;
virNetworkDefPtr def;
virNetworkDefPtr newDef;
char *ret = NULL;
virCheckFlags(VIR_NETWORK_XML_INACTIVE, NULL);
if (!(obj = networkObjFromNetwork(net)))
return ret;
def = virNetworkObjGetDef(obj);
newDef = virNetworkObjGetNewDef(obj);
if (virNetworkGetXMLDescEnsureACL(net->conn, def) < 0)
goto cleanup;
if ((flags & VIR_NETWORK_XML_INACTIVE) && newDef)
curDef = newDef;
else
curDef = def;
ret = virNetworkDefFormat(curDef, flags);
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static char *
networkGetBridgeName(virNetworkPtr net)
{
virNetworkObjPtr obj;
virNetworkDefPtr def;
char *bridge = NULL;
if (!(obj = networkObjFromNetwork(net)))
return bridge;
def = virNetworkObjGetDef(obj);
if (virNetworkGetBridgeNameEnsureACL(net->conn, def) < 0)
goto cleanup;
if (!(def->bridge)) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' does not have a bridge name."),
def->name);
goto cleanup;
}
ignore_value(VIR_STRDUP(bridge, def->bridge));
cleanup:
virNetworkObjEndAPI(&obj);
return bridge;
}
static int
networkGetAutostart(virNetworkPtr net,
int *autostart)
{
virNetworkObjPtr obj;
int ret = -1;
if (!(obj = networkObjFromNetwork(net)))
return ret;
if (virNetworkGetAutostartEnsureACL(net->conn, virNetworkObjGetDef(obj)) < 0)
goto cleanup;
*autostart = virNetworkObjIsAutostart(obj) ? 1 : 0;
ret = 0;
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkSetAutostart(virNetworkPtr net,
int autostart)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj;
virNetworkDefPtr def;
char *configFile = NULL, *autostartLink = NULL;
bool new_autostart;
bool cur_autostart;
int ret = -1;
if (!(obj = networkObjFromNetwork(net)))
goto cleanup;
def = virNetworkObjGetDef(obj);
if (virNetworkSetAutostartEnsureACL(net->conn, def) < 0)
goto cleanup;
if (!virNetworkObjIsPersistent(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID,
"%s", _("cannot set autostart for transient network"));
goto cleanup;
}
new_autostart = (autostart != 0);
cur_autostart = virNetworkObjIsAutostart(obj);
if (cur_autostart != new_autostart) {
if ((configFile = virNetworkConfigFile(driver->networkConfigDir,
def->name)) == NULL)
goto cleanup;
if ((autostartLink = virNetworkConfigFile(driver->networkAutostartDir,
def->name)) == NULL)
goto cleanup;
if (new_autostart) {
if (virFileMakePath(driver->networkAutostartDir) < 0) {
virReportSystemError(errno,
_("cannot create autostart directory '%s'"),
driver->networkAutostartDir);
goto cleanup;
}
if (symlink(configFile, autostartLink) < 0) {
virReportSystemError(errno,
_("Failed to create symlink '%s' to '%s'"),
autostartLink, configFile);
goto cleanup;
}
} else {
if (unlink(autostartLink) < 0 && errno != ENOENT && errno != ENOTDIR) {
virReportSystemError(errno,
_("Failed to delete symlink '%s'"),
autostartLink);
goto cleanup;
}
}
virNetworkObjSetAutostart(obj, new_autostart);
}
ret = 0;
cleanup:
VIR_FREE(configFile);
VIR_FREE(autostartLink);
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkGetDHCPLeases(virNetworkPtr net,
const char *mac,
virNetworkDHCPLeasePtr **leases,
unsigned int flags)
{
virNetworkDriverStatePtr driver = networkGetDriver();
size_t i, j;
size_t nleases = 0;
int rv = -1;
ssize_t size = 0;
int custom_lease_file_len = 0;
bool need_results = !!leases;
long long currtime = 0;
long long expirytime_tmp = -1;
bool ipv6 = false;
char *lease_entries = NULL;
char *custom_lease_file = NULL;
const char *ip_tmp = NULL;
const char *mac_tmp = NULL;
virJSONValuePtr lease_tmp = NULL;
virJSONValuePtr leases_array = NULL;
virNetworkIPDefPtr ipdef_tmp = NULL;
virNetworkDHCPLeasePtr lease = NULL;
virNetworkDHCPLeasePtr *leases_ret = NULL;
virNetworkObjPtr obj;
virNetworkDefPtr def;
virMacAddr mac_addr;
virCheckFlags(0, -1);
/* only to check if the MAC is valid */
if (mac && virMacAddrParse(mac, &mac_addr) < 0) {
virReportError(VIR_ERR_INVALID_MAC, "%s", mac);
return -1;
}
if (!(obj = networkObjFromNetwork(net)))
return -1;
def = virNetworkObjGetDef(obj);
if (virNetworkGetDHCPLeasesEnsureACL(net->conn, def) < 0)
goto cleanup;
/* Retrieve custom leases file location */
custom_lease_file = networkDnsmasqLeaseFileNameCustom(driver, def->bridge);
/* Read entire contents */
if ((custom_lease_file_len = virFileReadAll(custom_lease_file,
VIR_NETWORK_DHCP_LEASE_FILE_SIZE_MAX,
&lease_entries)) < 0) {
/* Even though src/network/leaseshelper.c guarantees the existence of
* leases file (even if no leases are present), and the control reaches
* here, instead of reporting error, return 0 leases */
rv = 0;
goto error;
}
if (custom_lease_file_len) {
if (!(leases_array = virJSONValueFromString(lease_entries))) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("invalid json in file: %s"), custom_lease_file);
goto error;
}
if ((size = virJSONValueArraySize(leases_array)) < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("couldn't fetch array of leases"));
goto error;
}
}
currtime = (long long) time(NULL);
for (i = 0; i < size; i++) {
if (!(lease_tmp = virJSONValueArrayGet(leases_array, i))) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("failed to parse json"));
goto error;
}
if (!(mac_tmp = virJSONValueObjectGetString(lease_tmp, "mac-address"))) {
/* leaseshelper program guarantees that lease will be stored only if
* mac-address is known otherwise not */
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("found lease without mac-address"));
goto error;
}
if (mac && virMacAddrCompare(mac, mac_tmp))
continue;
if (virJSONValueObjectGetNumberLong(lease_tmp, "expiry-time", &expirytime_tmp) < 0) {
/* A lease cannot be present without expiry-time */
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("found lease without expiry-time"));
goto error;
}
/* Do not report expired lease */
if (expirytime_tmp < currtime)
continue;
if (need_results) {
if (VIR_ALLOC(lease) < 0)
goto error;
lease->expirytime = expirytime_tmp;
if (!(ip_tmp = virJSONValueObjectGetString(lease_tmp, "ip-address"))) {
/* A lease without ip-address makes no sense */
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("found lease without ip-address"));
goto error;
}
/* Unlike IPv4, IPv6 uses ':' instead of '.' as separator */
ipv6 = strchr(ip_tmp, ':') ? true : false;
lease->type = ipv6 ? VIR_IP_ADDR_TYPE_IPV6 : VIR_IP_ADDR_TYPE_IPV4;
/* Obtain prefix */
for (j = 0; j < def->nips; j++) {
ipdef_tmp = &def->ips[j];
if (ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
AF_INET6)) {
lease->prefix = ipdef_tmp->prefix;
break;
}
if (!ipv6 && VIR_SOCKET_ADDR_IS_FAMILY(&ipdef_tmp->address,
AF_INET)) {
lease->prefix = virSocketAddrGetIPPrefix(&ipdef_tmp->address,
&ipdef_tmp->netmask,
ipdef_tmp->prefix);
break;
}
}
if ((VIR_STRDUP(lease->mac, mac_tmp) < 0) ||
(VIR_STRDUP(lease->ipaddr, ip_tmp) < 0) ||
(VIR_STRDUP(lease->iface, def->bridge) < 0))
goto error;
/* Fields that can be NULL */
if ((VIR_STRDUP(lease->iaid,
virJSONValueObjectGetString(lease_tmp, "iaid")) < 0) ||
(VIR_STRDUP(lease->clientid,
virJSONValueObjectGetString(lease_tmp, "client-id")) < 0) ||
(VIR_STRDUP(lease->hostname,
virJSONValueObjectGetString(lease_tmp, "hostname")) < 0))
goto error;
if (VIR_INSERT_ELEMENT(leases_ret, nleases, nleases, lease) < 0)
goto error;
} else {
nleases++;
}
VIR_FREE(lease);
}
if (leases_ret) {
/* NULL terminated array */
ignore_value(VIR_REALLOC_N(leases_ret, nleases + 1));
*leases = leases_ret;
leases_ret = NULL;
}
rv = nleases;
cleanup:
VIR_FREE(lease);
VIR_FREE(lease_entries);
VIR_FREE(custom_lease_file);
virJSONValueFree(leases_array);
virNetworkObjEndAPI(&obj);
return rv;
error:
if (leases_ret) {
for (i = 0; i < nleases; i++)
virNetworkDHCPLeaseFree(leases_ret[i]);
VIR_FREE(leases_ret);
}
goto cleanup;
}
/* A unified function to log network connections and disconnections */
static void
networkLogAllocation(virNetworkDefPtr netdef,
virDomainNetType actualType,
virNetworkForwardIfDefPtr dev,
virDomainNetDefPtr iface,
bool inUse)
{
char macStr[VIR_MAC_STRING_BUFLEN];
const char *verb = inUse ? "using" : "releasing";
if (!dev) {
VIR_INFO("MAC %s %s network %s (%d connections)",
virMacAddrFormat(&iface->mac, macStr), verb,
netdef->name, netdef->connections);
} else {
if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) {
VIR_INFO("MAC %s %s network %s (%d connections) "
"physical device %04x:%02x:%02x.%x (%d connections)",
virMacAddrFormat(&iface->mac, macStr), verb,
netdef->name, netdef->connections,
dev->device.pci.domain, dev->device.pci.bus,
dev->device.pci.slot, dev->device.pci.function,
dev->connections);
} else {
VIR_INFO("MAC %s %s network %s (%d connections) "
"physical device %s (%d connections)",
virMacAddrFormat(&iface->mac, macStr), verb,
netdef->name, netdef->connections,
dev->device.dev, dev->connections);
}
}
}
/* Private API to deal with logical switch capabilities.
* These functions are exported so that other parts of libvirt can
* call them, but are not part of the public API and not in the
* driver's function table. If we ever have more than one network
* driver, we will need to present these functions via a second
* "backend" function table.
*/
/* networkAllocateActualDevice:
* @dom: domain definition that @iface belongs to
* @iface: the original NetDef from the domain
*
* Looks up the network reference by iface, allocates a physical
* device from that network (if appropriate), and returns with the
* virDomainActualNetDef filled in accordingly. If there are no
* changes to be made in the netdef, then just leave the actualdef
* empty.
*
* Returns 0 on success, -1 on failure.
*/
static int
networkAllocateActualDevice(virDomainDefPtr dom,
virDomainNetDefPtr iface)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virDomainNetType actualType = iface->type;
virNetworkObjPtr obj = NULL;
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
virNetworkDefPtr netdef = NULL;
virNetDevBandwidthPtr bandwidth = NULL;
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
virPortGroupDefPtr portgroup = NULL;
virNetDevVPortProfilePtr virtport = iface->virtPortProfile;
virNetDevVlanPtr vlan = NULL;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
virNetworkForwardIfDefPtr dev = NULL;
size_t i;
int ret = -1;
if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
goto validate;
virDomainActualNetDefFree(iface->data.network.actual);
iface->data.network.actual = NULL;
obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"),
iface->data.network.name);
goto error;
}
netdef = virNetworkObjGetDef(obj);
if (!virNetworkObjIsActive(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID,
_("network '%s' is not active"),
netdef->name);
goto error;
}
if (VIR_ALLOC(iface->data.network.actual) < 0)
goto error;
/* portgroup can be present for any type of network, in particular
* for bandwidth information, so we need to check for that and
* fill it in appropriately for all forward types.
2014-06-23 09:51:38 +00:00
*/
portgroup = virPortGroupFindByName(netdef, iface->data.network.portgroup);
/* If there is already interface-specific bandwidth, just use that
* (already in NetDef). Otherwise, if there is bandwidth info in
* the portgroup, fill that into the ActualDef.
*/
if (iface->bandwidth)
bandwidth = iface->bandwidth;
else if (portgroup && portgroup->bandwidth)
bandwidth = portgroup->bandwidth;
if (bandwidth && virNetDevBandwidthCopy(&iface->data.network.actual->bandwidth,
bandwidth) < 0)
goto error;
/* copy appropriate vlan info to actualNet */
if (iface->vlan.nTags > 0)
vlan = &iface->vlan;
else if (portgroup && portgroup->vlan.nTags > 0)
vlan = &portgroup->vlan;
else if (netdef->vlan.nTags > 0)
vlan = &netdef->vlan;
if (vlan && virNetDevVlanCopy(&iface->data.network.actual->vlan, vlan) < 0)
goto error;
if (iface->trustGuestRxFilters)
iface->data.network.actual->trustGuestRxFilters
= iface->trustGuestRxFilters;
else if (portgroup && portgroup->trustGuestRxFilters)
iface->data.network.actual->trustGuestRxFilters
= portgroup->trustGuestRxFilters;
else if (netdef->trustGuestRxFilters)
iface->data.network.actual->trustGuestRxFilters
= netdef->trustGuestRxFilters;
if ((netdef->forward.type == VIR_NETWORK_FORWARD_NONE) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_NAT) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_OPEN)) {
/* for these forward types, the actual net type really *is*
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
* NETWORK; we just keep the info from the portgroup in
* iface->data.network.actual
2014-06-23 09:51:38 +00:00
*/
iface->data.network.actual->type = VIR_DOMAIN_NET_TYPE_NETWORK;
/* we also store the bridge device and macTableManager settings
network: save bridge name in ActualNetDef when actualType==network too When the actualType of a virDomainNetDef is "network", it means that we are connecting to a libvirt-managed network (routed, natted, or isolated) which does use a bridge device (created by libvirt). In the past we have required drivers such as qemu to call the public API to retrieve the bridge name in this case (even though it is available in the NetDef's ActualNetDef if the actualType is "bridge" (i.e., an externally-created bridge that isn't managed by libvirt). There is no real reason for this difference, and as a matter of fact it complicates things for qemu. Also, there is another bridge-related attribute (macTableManager) that will need to be available in both cases, so this makes things consistent. In order to avoid problems when restarting libvirtd after an update from an older version that *doesn't* store the network's bridgename in the ActualNetDef, we also need to put it in place during networkNotifyActualDevice() (this function is run for each interface of each domain whenever libvirtd is restarted). Along with making the bridge name available in the internal object, it is also now reported in the <source> element of the <interface> state XML (or the <actual> subelement in the internally-stored format). The one oddity about this change is that usually there is a separate union for every different "type" in a higher level object (e.g. in the case of a virDomainNetDef there are separate "network" and "bridge" members of the union that pivots on the type), but in this case network and bridge types both have exactly the same attributes, so the "bridge" member is used for both type==network and type==bridge.
2014-11-21 17:20:37 +00:00
* in iface->data.network.actual->data.bridge for later use
* after the domain's tap device is created (to attach to the
* bridge and set flood/learning mode on the tap device)
*/
if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
netdef->bridge) < 0)
goto error;
iface->data.network.actual->data.bridge.macTableManager
= netdef->macTableManager;
network: save bridge name in ActualNetDef when actualType==network too When the actualType of a virDomainNetDef is "network", it means that we are connecting to a libvirt-managed network (routed, natted, or isolated) which does use a bridge device (created by libvirt). In the past we have required drivers such as qemu to call the public API to retrieve the bridge name in this case (even though it is available in the NetDef's ActualNetDef if the actualType is "bridge" (i.e., an externally-created bridge that isn't managed by libvirt). There is no real reason for this difference, and as a matter of fact it complicates things for qemu. Also, there is another bridge-related attribute (macTableManager) that will need to be available in both cases, so this makes things consistent. In order to avoid problems when restarting libvirtd after an update from an older version that *doesn't* store the network's bridgename in the ActualNetDef, we also need to put it in place during networkNotifyActualDevice() (this function is run for each interface of each domain whenever libvirtd is restarted). Along with making the bridge name available in the internal object, it is also now reported in the <source> element of the <interface> state XML (or the <actual> subelement in the internally-stored format). The one oddity about this change is that usually there is a separate union for every different "type" in a higher level object (e.g. in the case of a virDomainNetDef there are separate "network" and "bridge" members of the union that pivots on the type), but in this case network and bridge types both have exactly the same attributes, so the "bridge" member is used for both type==network and type==bridge.
2014-11-21 17:20:37 +00:00
if (networkPlugBandwidth(obj, iface) < 0)
goto error;
} else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) &&
netdef->bridge) {
/* <forward type='bridge'/> <bridge name='xxx'/>
* is VIR_DOMAIN_NET_TYPE_BRIDGE
*/
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_BRIDGE;
if (VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
netdef->bridge) < 0)
goto error;
iface->data.network.actual->data.bridge.macTableManager
= netdef->macTableManager;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
/* merge virtualports from interface, network, and portgroup to
* arrive at actual virtualport to use
*/
if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
iface->virtPortProfile,
netdef->virtPortProfile,
portgroup
? portgroup->virtPortProfile : NULL) < 0) {
goto error;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
}
virtport = iface->data.network.actual->virtPortProfile;
if (virtport) {
/* only type='openvswitch' is allowed for bridges */
if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("<virtualport type='%s'> not supported for network "
"'%s' which uses a bridge device"),
virNetDevVPortTypeToString(virtport->virtPortType),
netdef->name);
goto error;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
}
}
} else if (netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
virDomainHostdevSubsysPCIBackendType backend;
iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_HOSTDEV;
if (networkCreateInterfacePool(netdef) < 0)
goto error;
/* pick first dev with 0 connections */
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].connections == 0) {
dev = &netdef->forward.ifs[i];
break;
}
}
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' requires exclusive access "
"to interfaces, but none are available"),
netdef->name);
goto error;
}
iface->data.network.actual->data.hostdev.def.parent.type = VIR_DOMAIN_DEVICE_NET;
iface->data.network.actual->data.hostdev.def.parent.data.net = iface;
iface->data.network.actual->data.hostdev.def.info = &iface->info;
iface->data.network.actual->data.hostdev.def.mode = VIR_DOMAIN_HOSTDEV_MODE_SUBSYS;
iface->data.network.actual->data.hostdev.def.managed = netdef->forward.managed ? 1 : 0;
iface->data.network.actual->data.hostdev.def.source.subsys.type = dev->type;
iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.addr = dev->device.pci;
switch (netdef->forward.driverName) {
case VIR_NETWORK_FORWARD_DRIVER_NAME_DEFAULT:
backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_DEFAULT;
break;
case VIR_NETWORK_FORWARD_DRIVER_NAME_KVM:
backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_KVM;
break;
case VIR_NETWORK_FORWARD_DRIVER_NAME_VFIO:
backend = VIR_DOMAIN_HOSTDEV_PCI_BACKEND_VFIO;
break;
default:
virReportError(VIR_ERR_INTERNAL_ERROR,
_("unrecognized driver name value %d "
" in network '%s'"),
netdef->forward.driverName, netdef->name);
goto error;
}
iface->data.network.actual->data.hostdev.def.source.subsys.u.pci.backend
= backend;
/* merge virtualports from interface, network, and portgroup to
* arrive at actual virtualport to use
*/
if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
iface->virtPortProfile,
netdef->virtPortProfile,
portgroup
? portgroup->virtPortProfile : NULL) < 0) {
goto error;
}
virtport = iface->data.network.actual->virtPortProfile;
if (virtport) {
/* make sure type is supported for hostdev connections */
if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBG &&
virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBH) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("<virtualport type='%s'> not supported for network "
"'%s' which uses an SR-IOV Virtual Function "
"via PCI passthrough"),
virNetDevVPortTypeToString(virtport->virtPortType),
netdef->name);
goto error;
}
}
} else if ((netdef->forward.type == VIR_NETWORK_FORWARD_BRIDGE) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_VEPA) ||
(netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH)) {
/* <forward type='bridge|private|vepa|passthrough'> are all
* VIR_DOMAIN_NET_TYPE_DIRECT.
*/
/* Set type=direct and appropriate <source mode='xxx'/> */
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
iface->data.network.actual->type = actualType = VIR_DOMAIN_NET_TYPE_DIRECT;
switch (netdef->forward.type) {
case VIR_NETWORK_FORWARD_BRIDGE:
iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_BRIDGE;
break;
case VIR_NETWORK_FORWARD_PRIVATE:
iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PRIVATE;
break;
case VIR_NETWORK_FORWARD_VEPA:
iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_VEPA;
break;
case VIR_NETWORK_FORWARD_PASSTHROUGH:
iface->data.network.actual->data.direct.mode = VIR_NETDEV_MACVLAN_MODE_PASSTHRU;
break;
}
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
/* merge virtualports from interface, network, and portgroup to
* arrive at actual virtualport to use
*/
if (virNetDevVPortProfileMerge3(&iface->data.network.actual->virtPortProfile,
iface->virtPortProfile,
netdef->virtPortProfile,
portgroup
? portgroup->virtPortProfile : NULL) < 0) {
goto error;
}
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
virtport = iface->data.network.actual->virtPortProfile;
if (virtport) {
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
/* make sure type is supported for macvtap connections */
if (virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBG &&
virtport->virtPortType != VIR_NETDEV_VPORT_PROFILE_8021QBH) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("<virtualport type='%s'> not supported for network "
"'%s' which uses a macvtap device"),
virNetDevVPortTypeToString(virtport->virtPortType),
netdef->name);
goto error;
}
}
/* If there is only a single device, just return it (caller will detect
* any error if exclusive use is required but could not be acquired).
*/
if ((netdef->forward.nifs <= 0) && (netdef->forward.npfs <= 0)) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' uses a direct mode, but "
"has no forward dev and no interface pool"),
netdef->name);
goto error;
} else {
/* pick an interface from the pool */
if (networkCreateInterfacePool(netdef) < 0)
goto error;
/* PASSTHROUGH mode, and PRIVATE Mode + 802.1Qbh both
* require exclusive access to a device, so current
* connections count must be 0. Other modes can share, so
* just search for the one with the lowest number of
* connections.
*/
if ((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
iface->data.network.actual->virtPortProfile &&
(iface->data.network.actual->virtPortProfile->virtPortType
== VIR_NETDEV_VPORT_PROFILE_8021QBH))) {
/* pick first dev with 0 connections */
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].connections == 0) {
dev = &netdef->forward.ifs[i];
break;
}
}
} else {
/* pick least used dev */
dev = &netdef->forward.ifs[0];
for (i = 1; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].connections < dev->connections)
dev = &netdef->forward.ifs[i];
}
}
/* dev points at the physical device we want to use */
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' requires exclusive access "
"to interfaces, but none are available"),
netdef->name);
goto error;
}
if (VIR_STRDUP(iface->data.network.actual->data.direct.linkdev,
dev->device.dev) < 0)
goto error;
}
}
if (virNetworkObjMacMgrAdd(obj, driver->dnsmasqStateDir,
dom->name, &iface->mac) < 0)
goto error;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
if (virNetDevVPortProfileCheckComplete(virtport, true) < 0)
goto error;
network: merge relevant virtualports rather than choosing one One of the original ideas behind allowing a <virtualport> in an interface definition as well as in the <network> definition *and*one or more <portgroup>s within the network, was that guest-specific parameteres (like instanceid and interfaceid) could be given in the interface's virtualport, and more general things (portid, managerid, etc) could be given in the network and/or portgroup, with all the bits brought together at guest startup time and combined into a single virtualport to be used by the guest. This was somehow overlooked in the implementation, though - it simply picks the "most specific" virtualport, and uses the entire thing, with no attempt to merge in details from the others. This patch uses virNetDevVPortProfileMerge3() to combine the three possible virtualports into one, then uses virNetDevVPortProfileCheck*() to verify that the resulting virtualport type is appropriate for the type of network, and that all the required attributes for that type are present. An example of usage is this: assuming a <network> definitions on host ABC of: <network> <name>testA</name> ... <virtualport type='openvswitch'/> ... <portgroup name='engineering'> <virtualport> <parameters profileid='eng'/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters profileid='sales'/> </virtualport> </portgroup> </network> and the same <network> on host DEF of: <network> <name>testA</name> ... <virtualport type='802.1Qbg'> <parameters typeid="1193047" typeidversion="2"/> </virtualport> ... <portgroup name='engineering'> <virtualport> <parameters managerid="11"/> </virtualport> </portgroup> <portgroup name='sales'> <virtualport> <parameters managerid="55"/> </virtualport> </portgroup> </network> and a guest <interface> definition of: <interface type='network'> <source network='testA' portgroup='sales'/> <virtualport> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" interfaceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f"\> </virtualport> ... </interface> If the guest was started on host ABC, the <virtualport> used would be: <virtualport type='openvswitch'> <parameters interfaceid='09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f' profileid='sales'/> </virtualport> but if that guest was started on host DEF, the <virtualport> would be: <virtualport type='802.1Qbg'> <parameters instanceid="09b11c53-8b5c-4eeb-8f00-d84eaa0aaa4f" typeid="1193047" typeidversion="2" managerid="55"/> </virtualport> Additionally, if none of the involved <virtualport>s had a specified type (this includes cases where no virtualport is given at all),
2012-08-02 18:10:00 +00:00
validate:
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
/* make sure that everything now specified for the device is
* actually supported on this type of network. NB: network,
* netdev, and iface->data.network.actual may all be NULL.
*/
if (virDomainNetGetActualVlan(iface)) {
/* vlan configuration via libvirt is only supported for PCI
* Passthrough SR-IOV devices (hostdev or macvtap passthru
* mode) and openvswitch bridges. Otherwise log an error and
* fail
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
*/
if (!(actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV ||
(actualType == VIR_DOMAIN_NET_TYPE_DIRECT &&
virDomainNetGetActualDirectMode(iface)
== VIR_NETDEV_MACVLAN_MODE_PASSTHRU) ||
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
(actualType == VIR_DOMAIN_NET_TYPE_BRIDGE &&
virtport && virtport->virtPortType
== VIR_NETDEV_VPORT_PROFILE_OPENVSWITCH))) {
if (netdef) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("an interface connecting to network '%s' "
"is requesting a vlan tag, but that is not "
"supported for this type of network"),
netdef->name);
} else {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED,
_("an interface of type '%s' "
"is requesting a vlan tag, but that is not "
"supported for this type of connection"),
virDomainNetTypeToString(iface->type));
}
goto error;
}
}
if (virDomainNetGetActualBandwidth(iface)) {
/* bandwidth configuration via libvirt is not supported for
* hostdev network devices
*/
if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) {
virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
_("bandwidth settings are not supported "
"for hostdev interfaces"));
goto error;
}
}
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
if (netdef) {
netdef->connections++;
if (dev)
dev->connections++;
/* finally we can call the 'plugged' hook script if any */
if (networkRunHook(obj, dom, iface,
VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
VIR_HOOK_SUBOP_BEGIN) < 0) {
/* adjust for failure */
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
netdef->connections--;
if (dev)
dev->connections--;
goto error;
}
networkLogAllocation(netdef, actualType, dev, iface, true);
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
}
ret = 0;
network: make network driver vlan-aware The network driver now looks for the vlan element in network and portgroup objects, and logs an error at network define time if a vlan is requested for a network type that doesn't support it. (Currently vlan configuration is only supported for openvswitch networks, and networks used to do hostdev assignment of SR-IOV VFs.) At runtime, the three potential sources of vlan information are examined in this order: interface, chosen portgroup, network, and the first that is non-empty is used. Another check for valid network type is made at this time, since the interface may have requested a vlan (a legal thing to have in the interface config, since it's not known until runtime if the chosen network will actually support it). Since we must also check for domains requesting vlans for unsupported connection types even if they are type='network', and since networkAllocateActualDevice() is being called in exactly the correct places, and has all of the necessary information to check, I slightly modified the logic of that function so that interfaces that aren't type='network' don't just return immediately. Instead, they also perform all the same validation for supported features. Because of this, it's not necessary to make this identical check in the other three places that would normally require it: 1) qemu domain startup, 2) qemu device hotplug, 3) lxc domain startup. This can be seen as a first step in consolidating network-related functionality into the network driver, rather than having copies of the same code spread around in multiple places; this will make it easier to split the network parts off into a separate daemon, as we've discussed recently.
2012-08-13 02:46:27 +00:00
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
error:
if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
virDomainActualNetDefFree(iface->data.network.actual);
iface->data.network.actual = NULL;
}
goto cleanup;
}
/* networkNotifyActualDevice:
* @dom: domain definition that @iface belongs to
* @iface: the domain's NetDef with an "actual" device already filled in.
*
* Called to notify the network driver when libvirtd is restarted and
* finds an already running domain. If appropriate it will force an
* allocation of the actual->direct.linkdev to get everything back in
* order, or re-attach the interface's tap device to the network's
* bridge.
*
* No return value (but does log any failures)
*/
static void
networkNotifyActualDevice(virDomainDefPtr dom,
virDomainNetDefPtr iface)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virDomainNetType actualType = virDomainNetGetActualType(iface);
virNetworkObjPtr obj;
virNetworkDefPtr netdef;
virNetworkForwardIfDefPtr dev = NULL;
size_t i;
char *master = NULL;
if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
return;
obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"),
iface->data.network.name);
goto error;
}
netdef = virNetworkObjGetDef(obj);
if (!virNetworkObjIsActive(obj)) {
virReportError(VIR_ERR_OPERATION_INVALID,
_("network '%s' is not active"),
netdef->name);
goto error;
}
network: save bridge name in ActualNetDef when actualType==network too When the actualType of a virDomainNetDef is "network", it means that we are connecting to a libvirt-managed network (routed, natted, or isolated) which does use a bridge device (created by libvirt). In the past we have required drivers such as qemu to call the public API to retrieve the bridge name in this case (even though it is available in the NetDef's ActualNetDef if the actualType is "bridge" (i.e., an externally-created bridge that isn't managed by libvirt). There is no real reason for this difference, and as a matter of fact it complicates things for qemu. Also, there is another bridge-related attribute (macTableManager) that will need to be available in both cases, so this makes things consistent. In order to avoid problems when restarting libvirtd after an update from an older version that *doesn't* store the network's bridgename in the ActualNetDef, we also need to put it in place during networkNotifyActualDevice() (this function is run for each interface of each domain whenever libvirtd is restarted). Along with making the bridge name available in the internal object, it is also now reported in the <source> element of the <interface> state XML (or the <actual> subelement in the internally-stored format). The one oddity about this change is that usually there is a separate union for every different "type" in a higher level object (e.g. in the case of a virDomainNetDef there are separate "network" and "bridge" members of the union that pivots on the type), but in this case network and bridge types both have exactly the same attributes, so the "bridge" member is used for both type==network and type==bridge.
2014-11-21 17:20:37 +00:00
/* if we're restarting libvirtd after an upgrade from a version
* that didn't save bridge name in actualNetDef for
* actualType==network, we need to copy it in so that it will be
* available in all cases
*/
if (actualType == VIR_DOMAIN_NET_TYPE_NETWORK &&
!iface->data.network.actual->data.bridge.brname &&
(VIR_STRDUP(iface->data.network.actual->data.bridge.brname,
netdef->bridge) < 0))
goto error;
/* see if we're connected to the correct bridge */
if (netdef->bridge) {
if (virNetDevGetMaster(iface->ifname, &master) < 0)
goto error;
if (STRNEQ_NULLABLE(netdef->bridge, master)) {
/* disconnect from current (incorrect) bridge */
if (master)
ignore_value(virNetDevBridgeRemovePort(master, iface->ifname));
/* attach/reattach to correct bridge.
* NB: we can't notify the guest of any MTU change anyway,
* so there is no point in trying to learn the actualMTU
* (final arg to virNetDevTapAttachBridge())
*/
if (virNetDevTapAttachBridge(iface->ifname, netdef->bridge,
&iface->mac, dom->uuid,
virDomainNetGetActualVirtPortProfile(iface),
virDomainNetGetActualVlan(iface),
iface->mtu, NULL) < 0) {
goto error;
}
}
}
if (!iface->data.network.actual ||
(actualType != VIR_DOMAIN_NET_TYPE_DIRECT &&
actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV)) {
VIR_DEBUG("Nothing to claim from network %s", iface->data.network.name);
goto success;
}
if (networkCreateInterfacePool(netdef) < 0)
goto error;
if (netdef->forward.nifs == 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' uses a direct or hostdev mode, "
"but has no forward dev and no interface pool"),
netdef->name);
goto error;
}
if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
const char *actualDev;
actualDev = virDomainNetGetActualDirectDev(iface);
if (!actualDev) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("the interface uses a direct mode, "
"but has no source dev"));
goto error;
}
/* find the matching interface and increment its connections */
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].type
== VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
dev = &netdef->forward.ifs[i];
break;
}
}
/* dev points at the physical device we want to use */
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' doesn't have dev='%s' "
"in use by domain"),
netdef->name, actualDev);
goto error;
}
/* PASSTHROUGH mode and PRIVATE Mode + 802.1Qbh both require
* exclusive access to a device, so current connections count
* must be 0 in those cases.
*/
if ((dev->connections > 0) &&
((netdef->forward.type == VIR_NETWORK_FORWARD_PASSTHROUGH) ||
((netdef->forward.type == VIR_NETWORK_FORWARD_PRIVATE) &&
iface->data.network.actual->virtPortProfile &&
(iface->data.network.actual->virtPortProfile->virtPortType
== VIR_NETDEV_VPORT_PROFILE_8021QBH)))) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' claims dev='%s' is already in "
"use by a different domain"),
netdef->name, actualDev);
goto error;
}
} else /* if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) */ {
virDomainHostdevDefPtr hostdev;
hostdev = virDomainNetGetActualHostdev(iface);
if (!hostdev) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("the interface uses a hostdev mode, "
"but has no hostdev"));
goto error;
}
/* find the matching interface and increment its connections */
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].type
== VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
virPCIDeviceAddressEqual(&hostdev->source.subsys.u.pci.addr,
&netdef->forward.ifs[i].device.pci)) {
dev = &netdef->forward.ifs[i];
break;
}
}
/* dev points at the physical device we want to use */
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' doesn't have "
"PCI device %04x:%02x:%02x.%x in use by domain"),
netdef->name,
hostdev->source.subsys.u.pci.addr.domain,
hostdev->source.subsys.u.pci.addr.bus,
hostdev->source.subsys.u.pci.addr.slot,
hostdev->source.subsys.u.pci.addr.function);
2014-06-23 09:51:38 +00:00
goto error;
}
/* PASSTHROUGH mode, PRIVATE Mode + 802.1Qbh, and hostdev (PCI
* passthrough) all require exclusive access to a device, so
* current connections count must be 0 in those cases.
*/
if ((dev->connections > 0) &&
netdef->forward.type == VIR_NETWORK_FORWARD_HOSTDEV) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' claims the PCI device at "
"domain=%d bus=%d slot=%d function=%d "
"is already in use by a different domain"),
netdef->name,
dev->device.pci.domain, dev->device.pci.bus,
dev->device.pci.slot, dev->device.pci.function);
goto error;
}
}
success:
netdef->connections++;
if (dev)
dev->connections++;
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
/* finally we can call the 'plugged' hook script if any */
if (networkRunHook(obj, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_PLUGGED,
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
VIR_HOOK_SUBOP_BEGIN) < 0) {
/* adjust for failure */
if (dev)
dev->connections--;
netdef->connections--;
goto error;
}
networkLogAllocation(netdef, actualType, dev, iface, true);
network: include plugged interface XML in "plugged" network hook The network hook script gets called whenever an interface is plugged into or unplugged from a network, but even though the full XML of both the network and the domain is included, there is no reasonable way to determine what exact resources the plugged interface is using: 1) Prior to a recent patch which modified the status XML of interfaces to include the information about actual hardware resources used, it would be possible to scan through the domain XML output sent to the hook, and from there find the correct interface, but that interface definition would not include any runtime info (e.g. bandwidth or vlan taken from a portgroup, or which physdev was used in case of a macvtap network). 2) After the patch modifying the status XML of interfaces, the network name would no longer be included in the domain XML, so it would be completely impossible to determine which interface was the one being plugged. To solve that problem, this patch includes a single <interface> element at the beginning of the XML sent to the network hook for "plugged" and "unplugged" (just inside <hookData>) that is the status XML of the interface being plugged. This XML will include all info gathered from the chosen network and portgroup. NB: due to hardcoded spaces in all of the device *Format() functions, the <interface> element inside the <hookData> will be indented by 6 spaces rather than 2. I had intended to fix this, but it turns out that to make virDomainNetDefFormat() indentation relative, I would have to do the same to virDomainDeviceInfoFormat(), and that function is called from 19 places - making that a prerequisite of this patch would cause too many merge difficulties if we needed to backport network hooks, so I chose to ignore the problem here and fix the problem for *all* devices in a followup later.
2014-02-21 12:12:00 +00:00
cleanup:
virNetworkObjEndAPI(&obj);
VIR_FREE(master);
return;
error:
goto cleanup;
}
/* networkReleaseActualDevice:
* @dom: domain definition that @iface belongs to
* @iface: a domain's NetDef (interface definition)
*
* Given a domain <interface> element that previously had its <actual>
* element filled in (and possibly a physical device allocated to it),
* free up the physical device for use by someone else, and free the
* virDomainActualNetDef.
*
* Returns 0 on success, -1 on failure.
*/
static int
networkReleaseActualDevice(virDomainDefPtr dom,
virDomainNetDefPtr iface)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virDomainNetType actualType = virDomainNetGetActualType(iface);
virNetworkObjPtr obj;
virNetworkDefPtr netdef;
virNetworkForwardIfDefPtr dev = NULL;
size_t i;
int ret = -1;
if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK)
2014-06-23 09:51:38 +00:00
return 0;
obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"),
iface->data.network.name);
goto error;
}
netdef = virNetworkObjGetDef(obj);
if (iface->data.network.actual &&
(netdef->forward.type == VIR_NETWORK_FORWARD_NONE ||
netdef->forward.type == VIR_NETWORK_FORWARD_NAT ||
netdef->forward.type == VIR_NETWORK_FORWARD_ROUTE ||
netdef->forward.type == VIR_NETWORK_FORWARD_OPEN) &&
networkUnplugBandwidth(obj, iface) < 0)
goto error;
if ((!iface->data.network.actual) ||
((actualType != VIR_DOMAIN_NET_TYPE_DIRECT) &&
(actualType != VIR_DOMAIN_NET_TYPE_HOSTDEV))) {
VIR_DEBUG("Nothing to release to network %s", iface->data.network.name);
goto success;
}
if (netdef->forward.nifs == 0) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' uses a direct/hostdev mode, but "
"has no forward dev and no interface pool"),
netdef->name);
goto error;
}
if (actualType == VIR_DOMAIN_NET_TYPE_DIRECT) {
const char *actualDev;
actualDev = virDomainNetGetActualDirectDev(iface);
if (!actualDev) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("the interface uses a direct mode, "
"but has no source dev"));
goto error;
}
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].type
== VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_NETDEV &&
STREQ(actualDev, netdef->forward.ifs[i].device.dev)) {
dev = &netdef->forward.ifs[i];
break;
}
}
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' doesn't have dev='%s' "
"in use by domain"),
netdef->name, actualDev);
goto error;
}
} else /* if (actualType == VIR_DOMAIN_NET_TYPE_HOSTDEV) */ {
virDomainHostdevDefPtr hostdev;
hostdev = virDomainNetGetActualHostdev(iface);
if (!hostdev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
"%s", _("the interface uses a hostdev mode, but has no hostdev"));
goto error;
}
for (i = 0; i < netdef->forward.nifs; i++) {
if (netdef->forward.ifs[i].type
== VIR_NETWORK_FORWARD_HOSTDEV_DEVICE_PCI &&
virPCIDeviceAddressEqual(&hostdev->source.subsys.u.pci.addr,
&netdef->forward.ifs[i].device.pci)) {
dev = &netdef->forward.ifs[i];
break;
}
}
if (!dev) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("network '%s' doesn't have "
"PCI device %04x:%02x:%02x.%x in use by domain"),
netdef->name,
hostdev->source.subsys.u.pci.addr.domain,
hostdev->source.subsys.u.pci.addr.bus,
hostdev->source.subsys.u.pci.addr.slot,
hostdev->source.subsys.u.pci.addr.function);
2014-06-23 09:51:38 +00:00
goto error;
}
2014-06-23 09:51:38 +00:00
}
success:
virNetworkObjMacMgrDel(obj, driver->dnsmasqStateDir, dom->name, &iface->mac);
if (iface->data.network.actual) {
netdef->connections--;
if (dev)
dev->connections--;
/* finally we can call the 'unplugged' hook script if any */
networkRunHook(obj, dom, iface, VIR_HOOK_NETWORK_OP_IFACE_UNPLUGGED,
VIR_HOOK_SUBOP_BEGIN);
networkLogAllocation(netdef, actualType, dev, iface, false);
}
ret = 0;
cleanup:
virNetworkObjEndAPI(&obj);
if (iface->type == VIR_DOMAIN_NET_TYPE_NETWORK) {
virDomainActualNetDefFree(iface->data.network.actual);
iface->data.network.actual = NULL;
}
return ret;
error:
goto cleanup;
}
/**
* networkCheckBandwidth:
* @net: network QoS
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
* @ifaceBand: interface QoS (may be NULL if no QoS)
* @oldBandwidth: new interface QoS (may be NULL if no QoS)
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
* @ifaceMac: interface MAC (used in error messages for identification)
* @new_rate: new rate for non guaranteed class
*
* Function checks if @ifaceBand can be satisfied on @net. However, sometimes it
* may happen that the interface that @ifaceBand corresponds to is already
* plugged into the @net and the bandwidth is to be updated. In that case we
* need to check if new bandwidth can be satisfied. If that's the case
* @ifaceBand should point to new bandwidth settings and @oldBandwidth to
* current ones. If you want to suppress this functionality just pass
* @oldBandwidth == NULL.
*
* Returns: -1 if plugging would overcommit network QoS
* 0 if plugging is safe (@new_rate updated)
* 1 if no QoS is set (@new_rate untouched)
*/
static int
networkCheckBandwidth(virNetworkObjPtr obj,
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
virNetDevBandwidthPtr ifaceBand,
virNetDevBandwidthPtr oldBandwidth,
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
virMacAddr ifaceMac,
unsigned long long *new_rate)
{
int ret = -1;
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virNetDevBandwidthPtr netBand = def->bandwidth;
unsigned long long tmp_floor_sum = virNetworkObjGetFloorSum(obj);
unsigned long long tmp_new_rate = 0;
char ifmac[VIR_MAC_STRING_BUFLEN];
network: Resolve Coverity FORWARD_NULL The following is a long winded way to say this patch is avoiding a false positive. Coverity complains that calling networkPlugBandwidth() could eventually end up with a NULL dereference on iface->bandwidth because in the networkAllocateActualDevice there's a check of 'iface->bandwidth' before deciding to try to use the 'portgroup' if it exists or to not perferm the virNetDevBandwidthCopy if 'bandwidth' is not NULL. Later in networkPlugBandwidth the 'iface->bandwidth' is sourced from virDomainNetGetActualBandwidth - which would be either iface->bandwidth or (preferably) iface->data.network.actual->bandwidth which would have been filled in from either 'iface->bandwidth' or 'portgroup->bandwidth' back in networkAllocateActualDevice There *is* a check in networkCheckBandwidth for the result of the virDomainNetGetActualBandwidth being NULL and a return 1 based on that which would cause networkPlugBandwidth to exit properly and thus never hit the condition that Coverity complains about. However, since Coverity checks all paths - it somehow believes that a return of 0 by networkCheckBandwidth in this condition would end up causing the possible NULL dereference. The "fix" to silence Coverity is to not have networkCheckBandwidth also call virDomainNetGetActualBandwidth in order to get the ifaceBand, but rather have it accept it as an argument which causes Coverity to "see" that it's the exit condition of 1 that won't have the possible NULL dereference. Since we're passing that, I added the passing of iface->mac rather than passing iface as well. This just hopefully makes sure someone doesn't undo this in the future...
2015-03-16 12:50:11 +00:00
virMacAddrFormat(&ifaceMac, ifmac);
if (ifaceBand && ifaceBand->in && ifaceBand->in->floor &&
!(netBand && netBand->in)) {
virReportError(VIR_ERR_OPERATION_UNSUPPORTED,
_("Invalid use of 'floor' on interface with MAC "
"address %s - network '%s' has no inbound QoS set"),
ifmac, def->name);
return -1;
}
if (((!ifaceBand || !ifaceBand->in || !ifaceBand->in->floor) &&
(!oldBandwidth || !oldBandwidth->in || !oldBandwidth->in->floor)) ||
!netBand || !netBand->in) {
/* no QoS required, claim success */
return 1;
}
tmp_new_rate = netBand->in->average;
if (oldBandwidth && oldBandwidth->in)
tmp_floor_sum -= oldBandwidth->in->floor;
if (ifaceBand && ifaceBand->in)
tmp_floor_sum += ifaceBand->in->floor;
/* check against peak */
if (netBand->in->peak) {
tmp_new_rate = netBand->in->peak;
if (tmp_floor_sum > netBand->in->peak) {
virReportError(VIR_ERR_OPERATION_INVALID,
_("Cannot plug '%s' interface into '%s' because it "
"would overcommit 'peak' on network '%s'"),
ifmac,
def->bridge,
def->name);
goto cleanup;
}
} else if (tmp_floor_sum > netBand->in->average) {
/* tmp_floor_sum can be between 'average' and 'peak' iff 'peak' is set.
* Otherwise, tmp_floor_sum must be below 'average'. */
virReportError(VIR_ERR_OPERATION_INVALID,
_("Cannot plug '%s' interface into '%s' because it "
"would overcommit 'average' on network '%s'"),
ifmac,
def->bridge,
def->name);
goto cleanup;
}
if (new_rate)
*new_rate = tmp_new_rate;
ret = 0;
cleanup:
return ret;
}
/**
* networkNextClassID:
* @net: network object
*
* Find next free class ID. @net is supposed
* to be locked already. If there is a free ID,
* it is marked as used and returned.
*
* Returns next free class ID or -1 if none is available.
*/
static ssize_t
networkNextClassID(virNetworkObjPtr obj)
{
ssize_t ret = 0;
virBitmapPtr classIdMap = virNetworkObjGetClassIdMap(obj);
if ((ret = virBitmapNextClearBit(classIdMap, -1)) < 0)
ret = virBitmapSize(classIdMap);
if (virBitmapSetBitExpand(classIdMap, ret) < 0)
return -1;
return ret;
}
static int
networkPlugBandwidthImpl(virNetworkObjPtr obj,
virDomainNetDefPtr iface,
virNetDevBandwidthPtr ifaceBand,
unsigned long long new_rate)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virBitmapPtr classIdMap = virNetworkObjGetClassIdMap(obj);
unsigned long long tmp_floor_sum = virNetworkObjGetFloorSum(obj);
ssize_t class_id = 0;
int plug_ret;
int ret = -1;
/* generate new class_id */
if ((class_id = networkNextClassID(obj)) < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("Could not generate next class ID"));
goto cleanup;
}
plug_ret = virNetDevBandwidthPlug(def->bridge, def->bandwidth,
&iface->mac, ifaceBand, class_id);
if (plug_ret < 0) {
ignore_value(virNetDevBandwidthUnplug(def->bridge, class_id));
goto cleanup;
}
/* QoS was set, generate new class ID */
iface->data.network.actual->class_id = class_id;
/* update sum of 'floor'-s of attached NICs */
tmp_floor_sum += ifaceBand->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
/* update status file */
if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
ignore_value(virBitmapClearBit(classIdMap, class_id));
tmp_floor_sum -= ifaceBand->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
iface->data.network.actual->class_id = 0;
ignore_value(virNetDevBandwidthUnplug(def->bridge, class_id));
goto cleanup;
}
/* update rate for non guaranteed NICs */
new_rate -= tmp_floor_sum;
if (virNetDevBandwidthUpdateRate(def->bridge, 2,
def->bandwidth, new_rate) < 0)
VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
def->bridge);
ret = 0;
cleanup:
return ret;
}
static int
networkPlugBandwidth(virNetworkObjPtr obj,
virDomainNetDefPtr iface)
{
int ret = -1;
int plug_ret;
unsigned long long new_rate = 0;
char ifmac[VIR_MAC_STRING_BUFLEN];
virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
if ((plug_ret = networkCheckBandwidth(obj, ifaceBand, NULL,
iface->mac, &new_rate)) < 0) {
/* helper reported error */
goto cleanup;
}
if (plug_ret > 0) {
/* no QoS needs to be set; claim success */
ret = 0;
goto cleanup;
}
virMacAddrFormat(&iface->mac, ifmac);
if (iface->type != VIR_DOMAIN_NET_TYPE_NETWORK ||
!iface->data.network.actual) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("Cannot set bandwidth on interface '%s' of type %d"),
ifmac, iface->type);
goto cleanup;
}
if (networkPlugBandwidthImpl(obj, iface, ifaceBand, new_rate) < 0)
goto cleanup;
ret = 0;
cleanup:
return ret;
}
static int
networkUnplugBandwidth(virNetworkObjPtr obj,
virDomainNetDefPtr iface)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
virBitmapPtr classIdMap = virNetworkObjGetClassIdMap(obj);
unsigned long long tmp_floor_sum = virNetworkObjGetFloorSum(obj);
virNetworkDriverStatePtr driver = networkGetDriver();
int ret = 0;
unsigned long long new_rate;
virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
if (iface->data.network.actual &&
iface->data.network.actual->class_id) {
if (!def->bandwidth || !def->bandwidth->in) {
VIR_WARN("Network %s has no bandwidth but unplug requested",
def->name);
goto cleanup;
}
/* we must remove class from bridge */
new_rate = def->bandwidth->in->average;
if (def->bandwidth->in->peak > 0)
new_rate = def->bandwidth->in->peak;
ret = virNetDevBandwidthUnplug(def->bridge,
iface->data.network.actual->class_id);
if (ret < 0)
goto cleanup;
/* update sum of 'floor'-s of attached NICs */
tmp_floor_sum -= ifaceBand->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
/* return class ID */
ignore_value(virBitmapClearBit(classIdMap,
iface->data.network.actual->class_id));
/* update status file */
if (virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
tmp_floor_sum += ifaceBand->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
ignore_value(virBitmapSetBit(classIdMap,
iface->data.network.actual->class_id));
goto cleanup;
}
/* update rate for non guaranteed NICs */
new_rate -= tmp_floor_sum;
if (virNetDevBandwidthUpdateRate(def->bridge, 2,
def->bandwidth, new_rate) < 0)
VIR_WARN("Unable to update rate for 1:2 class on %s bridge",
def->bridge);
/* no class is associated any longer */
iface->data.network.actual->class_id = 0;
}
cleanup:
return ret;
}
static void
networkNetworkObjTaint(virNetworkObjPtr obj,
virNetworkTaintFlags taint)
{
virNetworkDefPtr def = virNetworkObjGetDef(obj);
if (virNetworkObjTaint(obj, taint)) {
char uuidstr[VIR_UUID_STRING_BUFLEN];
virUUIDFormat(def->uuid, uuidstr);
VIR_WARN("Network name='%s' uuid=%s is tainted: %s",
def->name, uuidstr, virNetworkTaintTypeToString(taint));
}
}
static bool
networkBandwidthGenericChecks(virDomainNetDefPtr iface,
virNetDevBandwidthPtr newBandwidth)
{
virNetDevBandwidthPtr ifaceBand;
unsigned long long old_floor, new_floor;
if (virDomainNetGetActualType(iface) != VIR_DOMAIN_NET_TYPE_NETWORK) {
/* This is not an interface that's plugged into a network.
* We don't care. Thus from our POV bandwidth change is allowed. */
return false;
}
ifaceBand = virDomainNetGetActualBandwidth(iface);
old_floor = new_floor = 0;
if (ifaceBand && ifaceBand->in)
old_floor = ifaceBand->in->floor;
if (newBandwidth && newBandwidth->in)
new_floor = newBandwidth->in->floor;
return new_floor != old_floor;
}
static bool
networkBandwidthChangeAllowed(virDomainNetDefPtr iface,
virNetDevBandwidthPtr newBandwidth)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj = NULL;
virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
bool ret = false;
if (!networkBandwidthGenericChecks(iface, newBandwidth))
return true;
obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"),
iface->data.network.name);
return false;
}
if (networkCheckBandwidth(obj, newBandwidth, ifaceBand, iface->mac, NULL) < 0)
goto cleanup;
ret = true;
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static int
networkBandwidthUpdate(virDomainNetDefPtr iface,
virNetDevBandwidthPtr newBandwidth)
{
virNetworkDriverStatePtr driver = networkGetDriver();
virNetworkObjPtr obj = NULL;
virNetworkDefPtr def;
unsigned long long tmp_floor_sum;
virNetDevBandwidthPtr ifaceBand = virDomainNetGetActualBandwidth(iface);
unsigned long long new_rate = 0;
int plug_ret;
int ret = -1;
if (!networkBandwidthGenericChecks(iface, newBandwidth))
return 0;
obj = virNetworkObjFindByName(driver->networks, iface->data.network.name);
if (!obj) {
virReportError(VIR_ERR_NO_NETWORK,
_("no network with matching name '%s'"),
iface->data.network.name);
return ret;
}
def = virNetworkObjGetDef(obj);
if ((plug_ret = networkCheckBandwidth(obj, newBandwidth, ifaceBand,
iface->mac, &new_rate)) < 0) {
/* helper reported error */
goto cleanup;
}
if (plug_ret > 0) {
/* no QoS needs to be set; claim success */
ret = 0;
goto cleanup;
}
/* Okay, there are three possible scenarios: */
if (ifaceBand && ifaceBand->in && ifaceBand->in->floor &&
newBandwidth->in && newBandwidth->in->floor) {
/* Either we just need to update @floor .. */
if (virNetDevBandwidthUpdateRate(def->bridge,
iface->data.network.actual->class_id,
def->bandwidth,
newBandwidth->in->floor) < 0)
goto cleanup;
tmp_floor_sum = virNetworkObjGetFloorSum(obj);
tmp_floor_sum -= ifaceBand->in->floor;
tmp_floor_sum += newBandwidth->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
new_rate -= tmp_floor_sum;
if (virNetDevBandwidthUpdateRate(def->bridge, 2,
def->bandwidth, new_rate) < 0 ||
virNetworkObjSaveStatus(driver->stateDir, obj) < 0) {
/* Ouch, rollback */
tmp_floor_sum -= newBandwidth->in->floor;
tmp_floor_sum += ifaceBand->in->floor;
virNetworkObjSetFloorSum(obj, tmp_floor_sum);
ignore_value(virNetDevBandwidthUpdateRate(def->bridge,
iface->data.network.actual->class_id,
def->bandwidth,
ifaceBand->in->floor));
goto cleanup;
}
} else if (newBandwidth->in && newBandwidth->in->floor) {
/* .. or we need to plug in new .. */
if (networkPlugBandwidthImpl(obj, iface, newBandwidth, new_rate) < 0)
goto cleanup;
} else {
/* .. or unplug old. */
if (networkUnplugBandwidth(obj, iface) < 0)
goto cleanup;
}
ret = 0;
cleanup:
virNetworkObjEndAPI(&obj);
return ret;
}
static virNetworkDriver networkDriver = {
.name = "bridge",
.connectNumOfNetworks = networkConnectNumOfNetworks, /* 0.2.0 */
.connectListNetworks = networkConnectListNetworks, /* 0.2.0 */
.connectNumOfDefinedNetworks = networkConnectNumOfDefinedNetworks, /* 0.2.0 */
.connectListDefinedNetworks = networkConnectListDefinedNetworks, /* 0.2.0 */
.connectListAllNetworks = networkConnectListAllNetworks, /* 0.10.2 */
.connectNetworkEventRegisterAny = networkConnectNetworkEventRegisterAny, /* 1.2.1 */
.connectNetworkEventDeregisterAny = networkConnectNetworkEventDeregisterAny, /* 1.2.1 */
.networkLookupByUUID = networkLookupByUUID, /* 0.2.0 */
.networkLookupByName = networkLookupByName, /* 0.2.0 */
.networkCreateXML = networkCreateXML, /* 0.2.0 */
.networkDefineXML = networkDefineXML, /* 0.2.0 */
.networkUndefine = networkUndefine, /* 0.2.0 */
.networkUpdate = networkUpdate, /* 0.10.2 */
.networkCreate = networkCreate, /* 0.2.0 */
.networkDestroy = networkDestroy, /* 0.2.0 */
.networkGetXMLDesc = networkGetXMLDesc, /* 0.2.0 */
.networkGetBridgeName = networkGetBridgeName, /* 0.2.0 */
.networkGetAutostart = networkGetAutostart, /* 0.2.1 */
.networkSetAutostart = networkSetAutostart, /* 0.2.1 */
.networkIsActive = networkIsActive, /* 0.7.3 */
.networkIsPersistent = networkIsPersistent, /* 0.7.3 */
.networkGetDHCPLeases = networkGetDHCPLeases, /* 1.2.6 */
};
static virHypervisorDriver networkHypervisorDriver = {
.name = "network",
.connectOpen = networkConnectOpen, /* 4.1.0 */
.connectClose = networkConnectClose, /* 4.1.0 */
.connectIsEncrypted = networkConnectIsEncrypted, /* 4.1.0 */
.connectIsSecure = networkConnectIsSecure, /* 4.1.0 */
.connectIsAlive = networkConnectIsAlive, /* 4.1.0 */
};
static virConnectDriver networkConnectDriver = {
.hypervisorDriver = &networkHypervisorDriver,
.networkDriver = &networkDriver,
};
static virStateDriver networkStateDriver = {
.name = "bridge",
.stateInitialize = networkStateInitialize,
.stateAutoStart = networkStateAutoStart,
.stateCleanup = networkStateCleanup,
.stateReload = networkStateReload,
};
int
networkRegister(void)
{
if (virRegisterConnectDriver(&networkConnectDriver, false) < 0)
return -1;
if (virSetSharedNetworkDriver(&networkDriver) < 0)
return -1;
if (virRegisterStateDriver(&networkStateDriver) < 0)
return -1;
virDomainNetSetDeviceImpl(
networkAllocateActualDevice,
networkNotifyActualDevice,
networkReleaseActualDevice,
networkBandwidthChangeAllowed,
networkBandwidthUpdate);
return 0;
}