And use the newly added caps->host.netprefix (if it exists) for
interface names that match the autogenerated target names.
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
On the host when we start a container, it will be
placed in a cgroup path of
/machine.slice/machine-lxc\x2ddemo.scope
under /sys/fs/cgroup/*
Inside the containers' namespace we need to setup
/sys/fs/cgroup mounts, and currently will bind
mount /machine.slice/machine-lxc\x2ddemo.scope on
the host to appear as / in the container.
While this may sound nice, it confuses applications
dealing with cgroups, because /proc/$PID/cgroup
now does not match the directory in /sys/fs/cgroup
This particularly causes problems for systems and
will make it create repeated path components in
the cgroup for apps run in the container eg
/machine.slice/machine-lxc\x2ddemo.scope/machine.slice/machine-lxc\x2ddemo.scope/user.slice/user-0.slice/session-61.scope
This also causes any systemd service that uses
sd-notify to fail to start, because when systemd
receives the notification it won't be able to
identify the corresponding unit it came from.
In particular this break rabbitmq-server startup
Future kernels will provide proper cgroup namespacing
which will handle this problem, but until that time
we should not try to play games with hiding parent
cgroups.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
'free' on fedora23 wants to use the Slab field for calculated used
memory. The equation is:
used = MemTotal - MemFree - (Cached + Slab) - Buffers
We already set Cached and Buffers to 0, do the same for Slab and its
related values
https://bugzilla.redhat.com/show_bug.cgi?id=1300781
'free' on Fedora 23 will use MemAvailable to calculate its 'available'
field, but we are passing through the host's value. Set it to match
MemFree, which is what 'free' will do for older linux that don't have
MemAvailable
https://bugzilla.redhat.com/show_bug.cgi?id=1300781
We virtualize bits of /proc/meminfo by replacing host values with
values specific to the container.
However for calculating the final size of the returned data, we are
using the size of the original file and not the altered copy, which
could give garbelled output.
Some of the protocol files already include handing of the missing int
types such as xdr_uint64_t, some don't. To fix it everywhere, move out
of the appropriate defines to the utils/virxdrdefs.h file and include
it where needed.
Signed-off-by: Roman Bogorodskiy <bogorodskiy@gmail.com>
Commit id '71ce4759' altered the cgroup processing with respect to the
call to virCgroupAddTask being moved out from lower layers into the calling
layers especially for qemu processing of emulator and vcpu threads. The
movement affected lxc insomuch as it is possible for a code path to
return a NULL cgroup *and* a 0 return status via virCgroupNewPartition
failure when virCgroupNewIgnoreError succeeded when virCgroupNewMachineManual
returns. Coverity pointed out that would cause virCgroupAddTask to core.
This patch will check for a NULL cgroup as well as the negative return
and just return the NULL cgroup to the caller (as it would have previously)
virCgroupNewMachine used to add the pidleader to the newly created
machine cgroup. Do not do this implicit anymore.
Signed-off-by: Henning Schild <henning.schild@siemens.com>
For the multiqueue on macvtaps we are going to need to open
the device multiple times. Currently, this is not supported.
Rework the function, so that upper layers can be reworked too.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So yet again one of integer arguments that we use as a boolean.
Since the argument count of the function is unbearably long
enough, lets turn those booleans into flags.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This change ensures to call driver specific post-parse code to modify
domain definition after parsing hypervisor config the same way we do
after parsing XML.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Instead of creating symlinks, bind mount the devices to
/dev/pts/XY.
Using bind mounts it is no longer needed to add pts devices
to files like /etc/securetty.
Signed-off-by: Richard Weinberger <richard@nod.at>
Userspace does not expect that the initial console
is a controlling TTY. systemd can deal with that, others not.
On sysv init distros getty will fail to spawn a controlling on
/dev/console or /dev/tty1. Which will cause to whole container
to reboot upon ctrl-c.
This patch changes the behavior of libvirt to match the kernel
behavior where the initial TTY is also not controlling.
The only user visible change should be that a container with
bash as PID 1 would complain. But this matches exactly the kernel
be behavior with init=/bin/bash.
To get a controlling TTY for bash just run "setsid /bin/bash".
Signed-off-by: Richard Weinberger <richard@nod.at>
We have macros for both positive and negative string matching.
Therefore there is no need to use !STREQ or !STRNEQ. At the same
time as we are dropping this, new syntax-check rule is
introduced to make sure we won't introduce it again.
Signed-off-by: Ishmanpreet Kaur Khera <khera.ishman@gmail.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So imagine you want to crate new security manager:
if (!(mgr = virSecurityManagerNew("selinux", "QEMU", false, true, false, true)));
Hard to parse, right? What about this:
if (!(mgr = virSecurityManagerNew("selinux", "QEMU",
VIR_SECURITY_MANAGER_DEFAULT_CONFINED |
VIR_SECURITY_MANAGER_PRIVILEGED)));
Now that's better! This is what the commit does.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=871452
So, you want to create a domain from XML. The domain already
exists in libvirt's database of domains. It's okay, because name
and UUID matches. However, on domain startup, internal
representation of the domain is overwritten with your XML even
though we claim that the XML you've provided is a transient one.
The bug is to be found across nearly all the drivers.
Le sigh.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=871452
Okay, so we allow users to 'virsh create' an already existing
domain, providing completely different XML than the one stored in
Libvirt. Well, as long as name and UUID matches. However, in some
drivers the code that handles errors unconditionally removes the
domain that failed to start even though the domain might have
been persistent. Fortunately, the domain is removed just from the
internal list of domains and the config file is kept around.
Steps to reproduce:
1) virsh dumpxml $dom > /tmp/dom.xml
2) change XML so that it is still parse-able but won't boot, e.g.
change guest agent path to /foo/bar
3) virsh create /tmp/dom.xml
4) virsh dumpxml $dom
5) Observe "No such domain" error
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
We may want to do some decisions in drivers based on fact if we
are running as privileged user or not. Propagate this info there.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Commit id '692e9fac7' used virProcessSetNamespaces instead of inlining
the similar functionality; however, Coverity notes that the function
prototype expects a size_t value and not an enum and complains. So,
just typecast the enum as a size_t to avoid the noise.
Adds a new interface type using UDP sockets, this seems only applicable
to QEMU but have edited tree-wide to support the new interface type.
The interface type required the addition of a "localaddr" (local
address), this then maps into the following xml and qemu call.
<interface type='udp'>
<mac address='52:54:00:5c:67:56'/>
<source address='127.0.0.1' port='11112'>
<local address='127.0.0.1' port='22222'/>
</source>
<model type='virtio'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x07' function='0x0'/>
</interface>
QEMU call:
-net socket,udp=127.0.0.1:11112,localaddr=127.0.0.1:22222
Notice the xml "local" entry becomes the "localaddr" for the qemu call.
reference:
http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg00629.html
Signed-off-by: Jonathan Toppins <jtoppins@cumulusnetworks.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
This reverts commit 1ce7c1d20c,
which introduced a significant semantic change to the
virDomainGetInfo() API. Additionally, the change was only
made to 2 of the 15 virt drivers.
Conflicts:
src/qemu/qemu_driver.c
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Now that virProcessSetNamespaces() does accept FD list in the
correct format, we can simply turn lxcAttachNS into calling
virProcessSetNamespaces().
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
So far, if libvirt_lxc binary (usually to be found under
/usr/libexec/) is run with --help, due to a missing line
and our usual functions pattern, an 'uknown' error is returned.
Yeah, the help is printed out, but we should not claim error.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Older versions of glibc don't provide the setns() syscall
function wrapper, so we must define it ourselves to prevent
build failure on old distros.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
This patch adds feature for lxc containers to inherit namespaces.
This is very similar to what lxc-tools or docker provides. Look
for "man lxc-start" and you will find that you can pass command
args as [ --share-[net|ipc|uts] name|pid ]. Or check out docker
networking option in which you can give --net=container:NAME_or_ID
as an option for sharing +namespace.
>From this patch you can add extra libvirt option to share
namespace in following way.
<lxc:namespace>
<lxc:sharenet type='netns' value='red'/>
<lxc:shareipc type='pid' value='12345'/>
<lxc:shareuts type='name' value='container1'/>
</lxc:namespace>
The netns option is specific to sharenet. It can be used to
inherit from existing network namespace.
Co-authored: Daniel P. Berrange <berrange@redhat.com>
The problem here is that there are some values that kernel accepts, but
does not set them, for example 18446744073709551615 which acts the same
way as zero. Let's do the same thing we do with other tuning options
and re-read them right after they are set in order to keep our internal
structures up-to-date.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Since its introduction in 2011 (particularly in commit f4324e3292),
the option doesn't work. It just effectively disables all incoming
connections. That's because the client private data that contain the
'keepalive_supported' boolean, are initialized to zeroes so the bool is
false and the only other place where the bool is used is when checking
whether the client supports keepalive. Thus, according to the server,
no client supports keepalive.
Removing this instead of fixing it is better because a) apparently
nobody ever tried it since 2011 (4 years without one month) and b) we
cannot know whether the client supports keepalive until we get a ping or
pong keepalive packet. And that won't happen until after we dispatched
the ConnectOpen call.
Another two reasons would be c) the keepalive_required was tracked on
the server level, but keepalive_supported was in private data of the
client as well as the check that was made in the remote layer, thus
making all other instances of virNetServer miss this feature unless they
all implemented it for themselves and d) we can always add it back in
case there is a request and a use-case for it.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Rather than provide a somewhat generic error message when the API
returns false, allow the caller to supply a "report = true" option
in order to cause virReportError's to describe which of the 3 paths
that can cause failure.
Some callers don't care about what caused the failure, they just want
to have a true/false - for those, calling with report = false should
be sufficient.
virDomainDeleteConfig is meant to delete the persistent config and thus
it resets vm->autostart. Copy parts of qemuProcessRemoveDomainStatus to
a new helper to avoid using the incorrect function.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1230071
The scope name, even according to our docs is
"machine-$DRIVER\x2d$VMNAME.scope" virSystemdMakeScopeName would use the
resource partition name instead of "machine-" if it was specified thus
creating invalid scope paths.
This makes libvirt drop cgroups for a VM that uses custom resource
partition upon reconnecting since the detected scope name would not
match the expected name generated by virSystemdMakeScopeName.
The error is exposed by the following log entry:
debug : virCgroupValidateMachineGroup:302 : Name 'machine-qemu\x2dtestvm.scope' for controller 'cpu' does not match 'testvm', 'testvm.libvirt-qemu' or 'machine-test-qemu\x2dtestvm.scope'
for a "/machine/test" resource and "testvm" vm.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1238570
Add the sysfs_prefix argument to the call to allow for setting the
path for tests to something other than SYSFS_CPU_PATH which is a
derivative of SYSFS_SYSTEM_PATH
Use cpupath for nodeCapsInitNUMAFake and remove SYSFS_CPU_PATH
Explicit 'enum' keyword does not work with portablexdr-rpcgeb, causing its
parser to fail. Fix method is borrowed from virnetprotocol.x
Signed-off-by: Pavel Fedin <p.fedin@samsung.com>
So, recently I was testing the LXC driver. You know, startup some
domains. But to my surprise, I was not able to start a single one:
virsh # start --console test
error: Reconnected to the hypervisor
error: Failed to start domain test
error: internal error: guest failed to start: unexpected exit status 125
So I've start digging. It turns out, that in virExec(), when I printed
out the @cmd, I got strange values: *(cmd->outfdptr) was certainly not
valid FD number: it has random value of several millions. This
obviously made prepareStdFd(childout, STDOUT_FILENO) fail (line 611).
But outfdptr is set in virCommandSetOutputFD(). The only place within
LXC driver where the function is called is in
virLXCProcessBuildControllerCmd(). If you take a closer look at the
function it looks like this:
static virCommandPtr
virLXCProcessBuildControllerCmd(virLXCDriverPtr driver,
..
int logfd,
const char *pidfile)
{
...
virCommandSetOutputFD(cmd, &logfd);
virCommandSetErrorFD(cmd, &logfd);
...
}
Yes, you guessed it. @logfd is passed into the function by value.
However, in the function we try to get its address (an address of a
local variable) which is no longer valid once function is finished and
stack is cleaned. Therefore when cmd->outfdptr is evaluated at any
point after this function, we may get a random number, depending on
what's currently on the stack. Of course, this may work sometimes too
- it depends on the compiler how it arranges the code, when the stack
is wiped out.
In order to fix this, lets pass a pointer to @logfd instead of
figuring out (wrong) its value in a function.
The bug was introduced in e1de5521.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Although defined the same way, fortunately there hadn't been any deviation.
Ensure any assignments to onCrash use VIR_DOMAIN_LIFECYCLE_CRASH_* defs and
not VIR_DOMAIN_LIFECYCLE_* defs
Functions like virDomainOpenConsole() and virDomainOpenChannel() accept
NULL as a dev_name parameter. Try using alias for the error message if
dev_name is not specified.
Before:
error: internal error: character device <null> is not using a PTY
After:
error: internal error: character device serial0 is not using a PTY
Signed-off-by: Luyao Huang <lhuang@redhat.com>
So that they can format private data (e.g., disk private data) stored
elsewhere in the domain object.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Future kernels will mandate the use of nosuid+nodev+noexec
flags when mounting the /proc/sys filesystem. Unconditionally
add them now since they don't harm things regardless and could
mitigate future security attacks.
Add the qemu-nbd tasks to the container cgroup to make sure those will
be killed when the container is stopped. In order to reliably get the
qemu-nbd tasks PIDs, we use /sys/devices/virtual/block/<DEV>/pid as
qemu-nbd is daemonizing itself.
Most virDomainDiskIndexByName callers do not care about the index; what
they really want is a disk def pointer.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
So far, we are not reporting if numatune was even defined. The
value of zero is blindly returned (which maps onto
VIR_DOMAIN_NUMATUNE_MEM_STRICT). Unfortunately, we are making
decisions based on this value. Instead, we should not only return
the correct value, but report to the caller if the value is valid
at all.
For better viewing of this patch use '-w'.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Upping an interface for no reason and not configuring it is a cardinal sin.
With the default addrgenmode if eui64 it sticks a link-local address to the
interface. That is not good, as NetworkManager would see an address configured,
assume the interface is already configured and won't touch it iself and the
interface might stay unconfigured until the end of the days.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1124721
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Every domain that grabs a domain object to work over should
reference it to make sure it won't disappear meanwhile.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This needs to specified in way too many places for a simple validation
check. The ostype/arch/virttype validation checks later in
DomainDefParseXML should catch most of the cases that this was covering.
# virsh -c lxc:/// start helloworld
error: Failed to start domain helloworld
error: internal error: guest failed to start: Unknown
failure in libvirt_lxc startup
Return success when there are no cpuset.mems to be set,
instead of failing without setting an error.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
# virsh -c lxc:/// start helloworld
error: Failed to start domain helloworld
error: internal error: guest failed to start: Invalid value '1-3'
for 'cpuset.mems': Invalid argument
Free the cpu mask to avoid reusing it as a mem mask
in virCgroupSetCpusetMems
if virDomainNumatuneMaybeFormatNodeset does not set a mask.
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
/var/run may reside on a tmpfs and we fail to create the PID file if
/var/run/lxc does not exist.
Since commit 0a8addc1, the lxc driver's state directory isn't
automatically created before starting a domain. Now, the lxc driver
makes sure the state directory exists when it initializes.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Like we are doing in qemu driver (ea576ee543), lets call
virNumaSetupMemoryPolicy() only if really needed. Problem is, if
we numa_set_membind() child, there's no way to change it from the
daemon afterwards. So any later attempts to change the pinning
will fail. But in very weird way - CGroups will be set, but due
to membind child will not allocate memory from any other node.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
This patch adds code that parses and formats configuration for memory
devices.
A simple configuration would be:
<memory model='dimm'>
<target>
<size unit='KiB'>524287</size>
<node>0</node>
</target>
</memory>
A complete configuration of a memory device:
<memory model='dimm'>
<source>
<pagesize unit='KiB'>4096</pagesize>
<nodemask>1-3</nodemask>
</source>
<target>
<size unit='KiB'>524287</size>
<node>1</node>
</target>
</memory>
This patch preemptively forbids use of the <memory> device in individual
drivers so the users are warned right away that the device is not
supported.
Add a XML element that will allow to specify maximum supportable memory
and the count of memory slots to use with memory hotplug.
To avoid possible confusion and misuse of the new element this patch
also explicitly forbids the use of the maxMemory setting in individual
drivers's post parse callbacks. This limitation will be lifted when the
support is implemented.
virnetdevopenvswitch.h declares a few functions that can be called to
add ports to and remove them from OVS bridges, and retrieve the
migration data for a port. It does not contain any data definitions
that are used by domain_conf.h. But for some reason, domain_conf.h
virnetdevopenvswitch.h should be directly #including it. This adds a
few lines to the project, but saves all the files that don't need it
from the extra computing, and makes the dependencies more clear cut.
As there are two possible approaches to define a domain's memory size -
one used with legacy, non-NUMA VMs configured in the <memory> element
and per-node based approach on NUMA machines - the user needs to make
sure that both are specified correctly in the NUMA case.
To avoid this burden on the user I'd like to replace the NUMA case with
automatic totaling of the memory size. To achieve this I need to replace
direct access to the virDomainMemtune's 'max_balloon' field with
two separate getters depending on the desired size.
The two sizes are needed as:
1) Startup memory size doesn't include memory modules in some
hypervisors.
2) After startup these count as the usable memory size.
Note that the comments for the functions are future aware and document
state that will be present after a few later patches.
There was a mess in the way how we store unlimited value for memory
limits and how we handled values provided by user. Internally there
were two possible ways how to store unlimited value: as 0 value or as
VIR_DOMAIN_MEMORY_PARAM_UNLIMITED. Because we chose to store memory
limits as unsigned long long, we cannot use -1 to represent unlimited.
It's much easier for us to say that everything greater than
VIR_DOMAIN_MEMORY_PARAM_UNLIMITED means unlimited and leave 0 as valid
value despite that it makes no sense to set limit to 0.
Remove unnecessary function virCompareLimitUlong. The update of test
is to prevent the 0 to be miss-used as unlimited in future.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1146539
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
libvirt was unconditionally calling virNetDevBandwidthClear() for
every interface (and network bridge) of a type that supported
bandwidth, whether it actually had anything set or not. This doesn't
hurt anything (unless ifname == NULL!), but is wasteful.
This patch makes sure that all calls to virNetDevBandwidthClear() are
qualified by checking that the interface really had some bandwidth
setup done, and checks for a null ifname inside
virNetDevBandwidthClear(), silently returning success if it is null
(as well as removing the ATTRIBUTE_NONNULL from that function's
prototype, since we can't guarantee that it is never null,
e.g. sometimes a type='ethernet' interface has no ifname as it is
provided on the fly by qemu).
As virDomainNumatuneSet now doesn't allocate the virDomainNuma object
any longer it's not necessary to pass the pointer to a pointer to store
the object as it will not change any longer.
While touching the parameter definitions I've also changed the name of
the parameter to "numa".
Not all files we want to find using virFileFindResource{,Full} are
generated when libvirt is built, some of them (such as RNG schemas) are
distributed with sources. The current API was not able to find source
files if libvirt was built in VPATH.
Both RNG schemas and cpu_map.xml are distributed in source tarball.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Jumping to the cleanup label prior to starting the container failed to
properly clean everything up that is handled by the virLXCProcessCleanup
which is called if virLXCProcessStop is called on failure after the
container properly starts. Most importantly is prior to this patch none
of the stop/release hooks, host device reattachment, and network cleanup
(that is reverse of virLXCProcessSetupInterfaces).
Signed-off-by: Luyao Huang <lhuang@redhat.com>
Modify the VIR_DEBUG message in virLXCProcessCleanup to make it clearer
about the path. Also add some more VIR_DEBUG messages in virLXCProcessStart
in order to help debug error flow.
https://bugzilla.redhat.com/show_bug.cgi?id=1176503
Move the two console checks - one for zero nconsoles present and the
other for an invalid console type to earlier in the processing rather than
getting after performing some setup that has to be undone for what amounts
to an invalid configuration.
This resolves the above bug since it's not not possible to have changed
the security labels when we cause the configuration check failure.
We do have a check for valid per-domain security model, however we still
do permit an invalid security model for a domain's device (those which
are specified with <source> element).
This patch introduces a new function virSecurityManagerCheckAllLabel
which compares user specified security model against currently
registered security drivers. That being said, it also permits 'none'
being specified as a device security model.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1165485
Signed-off-by: Ján Tomko <jtomko@redhat.com>
commit a58e1cb4 didn't fix the bug if the security_default_confined is
not set to 1. We now clean up even if there is no seclabel defined or
the default one.
Return 0 instead of ERR_NO_SUPPORT in each driver
where we don't support managed save or -1 if
the domain does not exist.
This avoids spamming daemon logs when 'virsh dominfo' is run.
https://bugzilla.redhat.com/show_bug.cgi?id=1095637
Record the index of each host-side veth device created and report
them to systemd, so they show up in machinectl status for the VM.
lxc-shell(95449419f969d649d9962566ec42af7d)
Since: Fri 2015-01-16 16:53:37 GMT; 3s ago
Leader: 28085 (sh)
Service: libvirt-lxc; class container
Iface: vnet0
Address: fe80::216:3eff:fe00:c317%124
OS: Fedora 21 (Twenty One)
Unit: machine-lxc\x2dshell.scope
└─28085 /bin/sh
Add more logging to the lxc controller and container files to
facilitate debugging startup problems. Also make it clear when
the container is going to close stdout and thus no longer do
any logging.
Don't create the cgroups ahead of launching the container since
there is no need for the limits to apply during initial bootstrap.
Create the cgroup after the container PID is known and tell
systemd the initpid is the leader, instead of the controller
pid.
Currently when launching the LXC controller we first write out
the plain, inactive XML configuration, then launch the controller,
then replace the file with the live status XML configuration.
By good fortune this hasn't caused any problems other than some
misleading error messages during failure scenarios.
This simplifies the code so it only writes out the XML once and
always writes the live status XML. To do this we need to handshake
with the child process, to make execution pause just before exec()
so we can write the XML status with the child PID present.
Currently the lxc controller process itself is responsible for
daemonizing itself into the background and writing out its pid
file. The lxc driver would fork the controller and then attempt
to connect to the lxc monitor. This connection would only
succeed after the controller has backgrounded itself, setup
cgroups and written its pid file, so startup was race free.
The problem is that we need to delay create of the cgroups to
much later, such that we can tell systemd the container init
pid when we create the cgroups. If we delay cgroup creation
though the current synchronization won't work.
A second problem is that the controller needs the XML config
of the guest. Currently we write out the plain virDomainDefPtr
XML before starting the controller, and then later replace it
with the full virDomainObjPtr status XML. This is kind of gross
and also means that the controller doesn't get a record of the
live XML config right away. This means it doesn't have a record
of the veth device names either and so can't give that info
to systemd when creating the cgroups.
To address this we change the startup sequencing. The goal
is that we want to get the PID as soon as possible, before
the LXC controller even starts. So we stop letting the LXC
controller daemonize itself, and instead use virCommand's
built-in capabilities. This daemonizes and writes the PID
before LXC controller is exec'd. So the driver can read
the PID as soon as virCommandRun returns. It is no longer
safe to connect to the monitor or detect the cgroups though.
Fortunately the LXC controller already has a second point
of synchronization. Immediately before its event loop
starts running, it performs a handshake with the driver.
So we move the opening of the monitor connection and cgroup
detection after this synchronization point.
Build the pidfile string once when starting a guest and then
use the same string thereafter. This will benefit following
patches which need the pidfile string in more situations.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
For stateless, client side drivers, it is never correct to
probe for secondary drivers. It is only ever appropriate to
use the secondary driver that is associated with the
hypervisor in question. As a result the ESX & HyperV drivers
have both been forced to do hacks where they register no-op
drivers for the ones they don't implement.
For stateful, server side drivers, we always just want to
use the same built-in shared driver. The exception is
virtualbox which is really a stateless driver and so wants
to use its own server side secondary drivers. To deal with
this virtualbox has to be built as 3 separate loadable
modules to allow registration to work in the right order.
This can all be simplified by introducing a new struct
recording the precise set of secondary drivers each
hypervisor driver wants
struct _virConnectDriver {
virHypervisorDriverPtr hypervisorDriver;
virInterfaceDriverPtr interfaceDriver;
virNetworkDriverPtr networkDriver;
virNodeDeviceDriverPtr nodeDeviceDriver;
virNWFilterDriverPtr nwfilterDriver;
virSecretDriverPtr secretDriver;
virStorageDriverPtr storageDriver;
};
Instead of registering the hypervisor driver, we now
just register a virConnectDriver instead. This allows
us to remove all probing of secondary drivers. Once we
have chosen the primary driver, we immediately know the
correct secondary drivers to use.
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
The virDomainDefineXMLFlags and virDomainCreateXML APIs both
gain new flags allowing them to be told to validate XML.
This updates all the drivers to turn on validation in the
XML parser when the flags are set
systemd-machined introduced a new method CreateMachineWithNetwork
that obsoletes CreateMachine. It expects to be given a list of
VETH/TAP device indexes for the host side device(s) associated
with a container/machine.
This falls back to the old CreateMachine method when the new
one is not supported.
