Commit Graph

16398 Commits

Author SHA1 Message Date
Ján Tomko
a65871aaae Fix vlan ID detection in udev interface driver
Instead of guessing it from the interface name, look into
/proc/net/vlan/<interface>.

This works for devices not named <real_device>.<vlan ID>,
avoiding an error flood when virt-manager keeps asking about
them every second:

https://bugzilla.redhat.com/show_bug.cgi?id=966329
2014-05-04 15:08:14 +02:00
John Ferlan
5c2f782f76 Bump version to 1.2.5 for new dev cycle 2014-05-04 08:57:00 -04:00
Daniel Veillard
791fb3f668 Release of libvirt-1.2.4
* docs/news.html.in libvirt.spec.in: updates for release
* po/*.po*: fetched new localization and regenerated
2014-05-04 14:41:29 +08:00
Guido Günther
4bdb50376f Explicitly link virfirewalltest and virsystemdtest against dbus
This fixes link failures like:

  CCLD     virfirewalltest
  /usr/bin/ld: virfirewalltest-virfirewalltest.o: undefined reference to
  symbol 'dbus_message_iter_init_append'
2014-05-03 16:24:16 +02:00
Guido Günther
3cee4c05b7 qemuxml2argvtest: Don't use privileged mode upfront
When building packages in a clean chroot the QEMU_USER and QEMU_GROUP
don't exist making VirQemuDriverConfigNew fail with privileged=true.

Avoid that by not requiring privileged mode upfront but setting it later
so we skip the user/group existence check.

This solution was suggested by Daniel P. Berrange and tested by Martin
Kletzander.
2014-05-03 16:23:59 +02:00
Roman Bogorodskiy
064f49a050 tests: skip virfirewalltest on non-Linux systems
Currently firewalling is supported on Linux only, so skip the
virfirewalltest on other platforms.
2014-05-02 16:19:34 +04:00
John Ferlan
f1856eb622 Restore skipping of setting capacity
Commit id 'ac9a0963' refactored out the 'withCapacity' for the
virStorageBackendUpdateVolInfo() API.  See:

http://www.redhat.com/archives/libvir-list/2014-April/msg00043.html

This resulted in a difference in how 'virsh vol-info --pool <poolName>
<volume>' or 'virsh vol-list vol-list --pool <poolName> --details' outputs
the capacity information for a directory pool with a qcow2 sparse file.

For example, using the following XML

mkdir /home/TestPool
cat testpool.xml
<pool type='dir'>
  <name>TestPool</name>
  <uuid>6bf80895-10b6-75a6-6059-89fdea2aefb7</uuid>
  <source>
  </source>
  <target>
    <path>/home/TestPool</path>
    <permissions>
      <mode>0755</mode>
      <owner>0</owner>
      <group>0</group>
    </permissions>
  </target>
</pool>

virsh pool-create testpool.xml
virsh vol-create-as --pool TestPool temp_vol_1 \
      --capacity 1048576 --allocation 1048576 --format qcow2
virsh vol-info --pool TestPool temp_vol_1

Results in listing a Capacity value.  Prior to the commit, the value would
be '1.0 MiB' (1048576 bytes). However, after the commit the output would be
(for example) '192.50 KiB', which for my system was the size of the volume
in my file system (eg 'ls -l TestPool/temp_vol_1' results in '197120' bytes
or 192.50 KiB). While perhaps technically correct, it's not necessarily
what the user expected (certainly virt-test didn't expect it).

This patch restores the code to not update the target capacity for this path
2014-05-02 07:11:05 -04:00
Martin Kletzander
4cbc15d037 tests: don't fail with newer gnutls
gnutls-3.3.0 and newer leaves 2 FDs open in order to be backwards
compatible when it comes to chrooted binaries [1].  Linking
commandhelper with gnutls then leaves these two FDs open and
commandtest fails thanks to that.  This patch does not link
commandhelper with libvirt.la, but rather only the utilities making
the test pass.

Based on suggestion from Daniel [2].

[1] http://lists.gnutls.org/pipermail/gnutls-help/2014-April/003429.html
[2] https://www.redhat.com/archives/libvir-list/2014-April/msg01119.html

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
2014-05-02 09:56:33 +02:00
Ján Tomko
1055852aef fix build with older gcc
Older gcc (4.1.2-55.el5, 4.2.1 on FreeBSD) reports bogus warnings:
../../src/conf/nwfilter_conf.c:2111: warning: 'protocol' may be used
uninitialized in this function
../../src/conf/nwfilter_conf.c:2110: warning: 'dataProtocolID' may be
used uninitialized in this function

Initialize them to NULL to make the compiler happy.
2014-05-02 09:37:34 +02:00
Eric Blake
5c05e2b126 storage: reject negative indices
Commit f22b7899 stumbled across a difference between 32-bit and
64-bit platforms when parsing "-1" as an int.  Now that we've
fixed that difference, it's time to fix the testsuite.

* src/util/virstoragefile.c (virStorageFileParseChainIndex):
Require a positive index.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-05-01 15:11:02 -06:00
Eric Blake
7b045c8ce9 util: new stricter unsigned int parsing
strtoul() is required to parse negative numbers as their
twos-complement positive counterpart.  But sometimes we want
to reject negative numbers.  Add new functions to do this.
The 'p' suffix is a mnemonic for 'positive' (technically it
also parses 0, but 'non-negative' doesn't lend itself to a
nice one-letter suffix).

* src/util/virstring.h (virStrToLong_uip, virStrToLong_ulp)
(virStrToLong_ullp): New prototypes.
* src/util/virstring.c (virStrToLong_uip, virStrToLong_ulp)
(virStrToLong_ullp): New functions.
* src/libvirt_private.syms (virstring.h): Export them.
* tests/virstringtest.c (testStringToLong): Test them.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-05-01 15:11:02 -06:00
Eric Blake
f18c02ec22 util: fix uint parsing on 64-bit platforms
Commit f22b7899 called to light a long-standing latent bug: the
behavior of virStrToLong_ui was different on 32-bit platforms
than on 64-bit platforms.  Curse you, C type promotion and
narrowing rules, and strtoul specification.  POSIX says that for
a 32-bit long, strtol handles only 2^32 values [LONG_MIN to
LONG_MAX] while strtoul handles 2^33 - 1 values [-ULONG_MAX to
ULONG_MAX] with twos-complement wraparound for negatives.  Thus,
parsing -1 as unsigned long produces ULONG_MAX, rather than a
range error.  We WANT[1] this same shortcut for turning -1 into
UINT_MAX when parsing to int; and get it for free with 32-bit
long.  But with 64-bit long, ULONG_MAX is outside the range
of int and we were rejecting it as invalid; meanwhile, we were
silently treating -18446744073709551615 as 1 even though it
textually exceeds INT_MIN.  Too bad there's not a strtoui() in
libc that does guaranteed parsing to int, regardless of the size
of long.

The bug has been latent since 2007, introduced by Jim Meyering
in commit 5d25419 in the attempt to eradicate unsafe use of
strto[u]l when parsing ints and longs.  How embarrassing that we
are only discovering it now - so I'm adding a testsuite to ensure
that it covers all the corner cases we care about.

[1] Ideally, we really want the caller to be able to choose whether
to allow negative numbers to wrap around to their 2s-complement
counterpart, as in strtoul, or to force a stricter input range
of [0 to UINT_MAX] by rejecting negative signs; this will be added
in a later patch for all three int types.

This patch is tested on both 32- and 64-bit; the enhanced
virstringtest passes on both platforms, while virstoragetest now
reliably fails on both platforms instead of just 32-bit platforms.
That test will be fixed later.

* src/util/virstring.c (virStrToLong_ui): Ensure same behavior
regardless of platform long size.
* tests/virstringtest.c (testStringToLong): New function.
(mymain): Comprehensively test string to long parsing.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-05-01 15:11:02 -06:00
Daniel P. Berrange
dca027a9b7 Misc error reporting bugs in QEMU cli builder
A couple of places in the QEMU XML -> ARGV conversion code
raised an error but then forgot to return an error status
due to missing gotos. While fixing this also tweak style
of a couple of other error reports

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-05-01 17:24:45 +01:00
Laine Stump
0eac9d1e90 qemu: fix crash when removing <filterref> from interface with update-device
If a domain network interface that contains a <filterref> is modified
"live" using "virsh update-device --live", libvirtd would crash. This
was because the code supporting live update of an interface's
filterref was assuming that a filterref might be added or modified,
but didn't account for removing the filterref, resulting in a null
dereference of the filter name.

Introduced with commit 258fb278, which was first in libvirt v1.0.1.

This addresses https://bugzilla.redhat.com/show_bug.cgi?id=1093301
2014-05-01 15:46:10 +03:00
Peter Krempa
b47668233e storage: Clear all data allocated about backing store before reparsing
To avoid memory leak of the "backingStoreRaw" field when reparsing
backing chains a new function is being introduced by this patch that
shall be used to clear backing store information.

The memory leak was introduced in commit 8823272d41.
2014-04-30 19:59:20 +02:00
Stefan Berger
cdebc2f74c nwfilter: Tear down temp. filters when tearing all filters
Refactor the ebiptablesTearNewRules function so that the teardown of temporary
filters can also be called by the ebiptablesAllTeardown function.

This fixes a problem that leaves temporary filters behind when a VM shuts down
while its filters are modified.

Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>

v1->v2:
  - test cases adjusted to expect more commands
2014-04-30 12:51:38 -04:00
Michal Privoznik
7a6ae35dbe Device{Attach,Detach}: Document S4 limitations
https://bugzilla.redhat.com/show_bug.cgi?id=808463

Well, libvirt doesn't distinguish between domain poweroff and
hibernation (S4). It's hard to differentiate these two on a real
machine anyway. As a result, any device that is hot(un-)plugged is
lost (appears again) when domain is started again as from our POV
it is a fresh cold boot. Instead of doing anything wise here, we
should just document this as known limitation.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
2014-04-30 18:25:11 +02:00
Daniel P. Berrange
916b147cc8 Set mknod permission in device ACL for LXC USB devices
The LXC controller itself needs to mknod the USB device
node in /dev/bus/usb, so we can't block mknod permission
from the cgroup.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-30 16:54:13 +01:00
Stefan Berger
a457fc6550 nwfilter: Validate rule after parsing
An IP or IPv6 rule with port specification but without protocol
specification cannot be instantiated by ebtables. The documentation
points to 'protocol' being required but implementation does not
enforce it to be given.

Implement a rule validation function that checks whether the rule is
valid when it is defined. This for example prevents the definition
of rules like:

<ip dstportstart='53'>

where a protocol attribute would be required for it to be valid and for
ebtables to be able to instantiate it. A valid rule then is:

<ip protocol='udp' dstportstart='53'>

Signed-off-by: Stefan Berger <stefanb@linux.vnet.ibm.com>
2014-04-30 11:41:18 -04:00
Sahid Orentino Ferdjaoui
b21795bb5b Add a new example to illustrate domain migration
This commit adds a new example to illustrate peer to
peer domain migration with virDomainMigrateToURI.

Signed-off-by: Sahid Orentino Ferdjaoui <sahid.ferdjaoui@cloudwatt.com>
2014-04-30 14:29:07 +01:00
Sahid Orentino Ferdjaoui
f393c4603a Add support for QEMU migration to use SASL authentication
This commit provides the ability to virDomainMigrateToURI to
check for SASL credentials when attempts to migrate a domain
with the driver QEMU.

Signed-off-by: Sahid Orentino Ferdjaoui <sahid.ferdjaoui@cloudwatt.com>
2014-04-30 14:07:53 +01:00
Pavel Hrdina
e8eb5146de networkxml2firewalltest: fix build failure on freebsd
We need to include the testutils.h also for freebsd.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
2014-04-30 10:16:09 +02:00
Pavel Hrdina
620e16760e virfirewall: fix build on freebsd
Freebsd doesn't know ENODATA so we have to use different EINVAL
error code.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
2014-04-30 08:51:29 +02:00
Eric Blake
c9679edca9 conf: avoid null deref during storage probe
Commit 5c43e2e introduced a NULL deref if there is a failure
in virStorageFileGetMetadataInternal.

* src/util/virstoragefile.c (virStorageFileGetMetadataFromBuf):
Fix error handling.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-29 08:13:29 -06:00
Daniel P. Berrange
0e285a588a Disable libvirtd by default when building on Win32
We don't support building libvirtd on Win32 since we lack the
fork/exec feature needed for the stateful drivers. Disable this
by default, so users can just do 'mingw32-configure' with no
special args required.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-29 11:30:32 +01:00
Daniel P. Berrange
37697d828b Don't use SO_REUSEADDR on Win32 platforms
SO_REUSEADDR on Windows is actually akin to SO_REUSEPORT
on Linux/BSD. ie it allows 2 apps to listen to the same
port at once. Thus we must not set it on Win32 platforms

See http://msdn.microsoft.com/en-us/library/windows/desktop/ms740621.aspx

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-29 11:30:32 +01:00
Martin Kletzander
837154a151 qemu: properly quit migration with abort_on_error
When EIO comes to qemu while it's replying to
qemuMigrationUpdateJobStatus(), qemu blocks, the migration of RAM can
complete in the meantime, and when qemu unblocks, it sends us
BLOCK_IO_ERROR plus migrations "status": "complete".  Even though we
act upon the BLOCK_IO_ERROR by setting the proper state of the domain,
the call still waits for the proper reply on monitor for query_migrate
and after it gets it, it checks that migration is completed and the
migration is finished.  This is what abort_on_error flag was meant for
(we can migrate with these errors, but this flag must inhibit such
behaviour).  Changing the order of the steps guarantees the flag works
properly.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1045833

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
2014-04-29 10:40:44 +02:00
Martin Kletzander
f6ec6af8a6 qemu: don't call virFileExists() for network type disks
Based on suggestion from Eric [1], because it might not get cleaned up
before the release, so to avoid potential errors.

[1] https://www.redhat.com/archives/libvir-list/2014-April/msg00929.html

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
2014-04-29 10:32:27 +02:00
Chen Hanxiao
6c19f0de49 docs: update docs for setting the QEMU BIOS path
commit ddf2dfa1f7
provided a way to determine which bios files to use.
But we need to update related docs.

disccussed at:
http://www.redhat.com/archives/libvir-list/2012-June/msg01286.html

Signed-off-by: Chen Hanxiao <chenhanxiao@cn.fujitsu.com>
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
2014-04-29 10:18:12 +02:00
Steven McDonald
4cd508ba4f storage_backend_rbd: Correct argument order to rbd_create3
The stripe_unit and stripe_count arguments are passed to rbd_create3 in
the wrong order, resulting in a stripe size of 1 byte with 4194304
stripes on newly created RBD volumes.

https://bugzilla.redhat.com/show_bug.cgi?id=1092208
Signed-off-by: Steven McDonald <steven.mcdonald@anchor.net.au>
2014-04-28 22:11:09 -06:00
Eric Blake
eae00fb59a enforce sane readdir usage
Now that all clients have been adjusted, ensure that no future
misuse of readdir is introduced into the code base.

* cfg.mk (sc_prohibit_readdir): New rule.
* src/util/virfile.c (virDirRead): Exempt the wrapper.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:46 -06:00
Laine Stump
a431523444 network: use virDirRead in networkMigrateStateFiles
This attempts to follow the same variable name and usage patterns as
the other conversions to virDirRead().
2014-04-28 17:52:46 -06:00
Eric Blake
56a03233ab storage: use virDirRead API
More instances of failure to report (unlikely) readdir errors.
In one case, I chose to ignore them, given that a readdir error
would be no different than timing out on the loop, where the
fallback path behaves correctly either way.

* src/storage/storage_backend.c (virStorageBackendStablePath):
Ignore readdir errors.
* src/storage/storage_backend_fs.c
(virStorageBackendFileSystemRefresh): Report readdir errors.
* src/storage/storage_backend_iscsi.c
(virStorageBackendISCSIGetHostNumber): Likewise.
* src/storage/storage_backend_scsi.c (getNewStyleBlockDevice)
(getBlockDevice, virStorageBackendSCSIFindLUs): Likewise.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:46 -06:00
Eric Blake
ddcf4730ce drivers: use virDirRead API
Convert all remaining clients of readdir to use the new
interface, so that we can ensure (unlikely) errors while
reading a directory are reported.

* src/openvz/openvz_conf.c (openvzAssignUUIDs): Use new
interface.
* src/parallels/parallels_storage.c (parallelsFindVolumes)
(parallelsFindVmVolumes): Report readdir failures.
* src/qemu/qemu_driver.c (qemuDomainSnapshotLoad): Ignore readdir
failures.
* src/secret/secret_driver.c (loadSecrets): Likewise.
* src/qemu/qemu_hostdev.c
(qemuHostdevHostSupportsPassthroughVFIO): Report readdir failures.
* src/xen/xen_inotify.c (xenInotifyOpen): Likewise.
* src/xen/xm_internal.c (xenXMConfigCacheRefresh): Likewise.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:45 -06:00
Eric Blake
ac1d42ac72 util: use virDirRead API
In making the conversion to the new API, I fixed a couple bugs:
virSCSIDeviceGetSgName would leak memory if a directory
unexpectedly contained multiple entries;
virNetDevTapGetRealDeviceName could report a spurious error
from a stale errno inherited before starting the readdir search.

The decision on whether to store the result of virDirRead into
a variable is based on whether the end of the loop falls through
to cleanup code automatically.  In some cases, we have loops that
are documented to return NULL on failure, and which raise an
error on most failure paths but not in the case where the directory
was unexpectedly empty; it may be worth a followup patch to
explicitly report an error if readdir was successful but the
directory was empty, so that a NULL return always has an error set.

* src/util/vircgroup.c (virCgroupRemoveRecursively): Use new
interface.
(virCgroupKillRecursiveInternal, virCgroupSetOwner): Report
readdir failures.
* src/util/virfile.c (virFileLoopDeviceOpenSearch)
(virFileNBDDeviceFindUnused, virFileDeleteTree): Use new
interface.
* src/util/virnetdevtap.c (virNetDevTapGetRealDeviceName):
Properly check readdir errors.
* src/util/virpci.c (virPCIDeviceIterDevices)
(virPCIDeviceFileIterate, virPCIGetNetName): Report readdir
failures.
(virPCIDeviceAddressIOMMUGroupIterate): Use new interface.
* src/util/virscsi.c (virSCSIDeviceGetSgName): Report readdir
failures, and avoid memory leak.
(virSCSIDeviceGetDevName): Report readdir failures.
* src/util/virusb.c (virUSBDeviceSearch): Report readdir
failures.
* src/util/virutil.c (virGetFCHostNameByWWN)
(virFindFCHostCapableVport): Report readdir failures.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:45 -06:00
Eric Blake
6b9f40e856 conf: use virDirRead API
When reading configuration files, we were silently ignoring
directory read failures.  While unlikely, we might as well
report them.

* src/conf/domain_conf.c (virDomainObjListLoadAllConfigs): Report
readdir errors.
* src/conf/network_conf.c (virNetworkLoadAllState)
(virNetworkLoadAllConfigs): Likewise.
* src/conf/nwfilter_conf.c (virNWFilterLoadAllConfigs): Likewise.
* src/conf/storage_conf.c (virStoragePoolLoadAllConfigs):
Likewise.

Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:45 -06:00
Natanael Copa
92cf75df8e nodeinfo: use virDirRead API
This makes sure that errno is reset before readdir is called, even if
the loop does a 'continue'.

This fixes issue with musl libc which sets errno on sscanf. The
following 'continue' makes the errno be set before calling readdir.

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:45 -06:00
Natanael Copa
1ce2f1a434 util: introduce virDirRead wrapper for readdir
Introduce a wrapper for readdir. This helps us make sure that we always
set errno before calling readdir and it will make sure errors are
properly logged.

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Eric Blake <eblake@redhat.com>
2014-04-28 17:52:45 -06:00
Martin Kletzander
fbcd9d3252 tests: remove hostdevmgr directory on cleanup
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
2014-04-28 22:58:10 +02:00
Jim Fehlig
4f70ae2409 xen: ensure /usr/sbin/xend exists before checking status
With xend on the way out, installations may not even have
/usr/sbin/xend, which results in the following error when the
drivers are probed

2014-04-28 18:21:19.271+0000: 22129: error : virCommandWait:2426 :
internal error: Child process (/usr/sbin/xend status) unexpected exit
status 127: libvirt:  error : cannot execute binary /usr/sbin/xend:
No such file or directory

Check for existence of /usr/sbin/xend before trying to run it with
the 'status' option.
2014-04-28 13:58:33 -06:00
Daniel P. Berrange
04515a3438 Remove bogus ATTRIBUTE_NONNULL from virFirewallAddRuleFull
The virFirewallAddRuleFull method originally had a single
compulsory virFirewallQueryCallback parameter. During dev
work though the ignoreErrors parameter was added and the
callback parameter made optional. The ATTRIBUTE_NONNULL
annotation was never removed though.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-28 14:27:40 +01:00
Daniel P. Berrange
727efb9696 Make autostart of virtlockd actually work
The virnetsocket.c API is hardcoded to pass --timeout=30 to
any daemon it auto-starts. For inexplicable reasons the virtlockd
daemon did not implement the --timeout option, so it would
immediately exit on autostart with an error.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-28 13:30:41 +01:00
Daniel P. Berrange
e18a80a3a0 Use virFileFindResource to locate virtlockd daemon
Make the lock plugin use virFileFindResource to find the
virtlockd daemon path, so that it executes the in-builddir
daemon if run from source tree.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-28 13:30:41 +01:00
Daniel P. Berrange
57e5c3c2ae Use virFileFindResource to locate libvirtd daemon
Make the remote driver use virFileFindResource to find the
libvirt daemon path, so that it executes the in-builddir
daemon if run from source tree.

Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
2014-04-28 13:30:41 +01:00
Jiri Denemark
eec2f65946 Recheck disk backing chains after snapshot
When a snapshot operation finishes we have to recheck the backing chain
of all disks involved in the snapshot. And we need to do that even if
the operation failed because some of the disks might have changed if
QEMU did not support transactions.
2014-04-27 21:19:12 +02:00
Laine Stump
34cc3b2f10 network: centralize check for active network during interface attach
The check for a network being active during interface attach was being
done individually in several places (by both the lxc driver and the
qemu driver), but those places were too specific, leading to it *not*
being checked when allocating a connection/device from a macvtap or
hostdev network.

This patch puts a single check in networkAllocateActualDevice(), which
is always called before the any network interface is attached to any
type of domain. It also removes all the other now-redundant checks
from the lxc and qemu drivers.

NB: the following patches are prerequisites for this patch, in the
case that it is backported to any branch:

  440beeb network: fix virNetworkObjAssignDef and persistence
  8aaa5b6 network: create statedir during driver initialization
  b9e9549 network: change location of network state xml files
  411c548 network: set macvtap/hostdev networks active if their state
          file exists

This fixes:

  https://bugzilla.redhat.com/show_bug.cgi?id=880483
2014-04-27 12:22:36 +03:00
Laine Stump
411c548668 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).
2014-04-27 12:20:39 +03:00
Laine Stump
b9e95491d1 network: change location of network state xml files
For some reason these have been stored in /var/lib, although other
drivers (e.g. qemu and lxc) store their state files in /var/run.

It's much nicer to store state files in /var/run because it is
automatically cleared out when the system reboots. We can then use
existence of the state file as a convenient indicator of whether or
not a particular network is active.

Since changing the location of the state files by itself will cause
problems in the case of a *live* upgrade from an older libvirt that
uses /var/lib (because current status of active networks will be
lost), the network driver initialization has been modified to migrate
any network state files from /var/lib to /var/run.

This will not help those trying to *downgrade*, but in practice this
will only be problematic in two cases

1) If there are networks with network-wide bandwidth limits configured
   *and in use* by a guest during a downgrade to "old" libvirt. In this
   case, the class ID's used for that network's tc rules, as well as
   the currently in-use bandwidth "floor" will be forgotten.

2) If someone does this: 1) upgrade libvirt, 2) downgrade libvirt, 3)
   modify running state of network (e.g. add a static dhcp host, etc),
   4) upgrade. In this case, the modifications to the running network
   will be lost (but not any persistent changes to the network's
   config).
2014-04-27 12:19:57 +03:00
Laine Stump
8aaa5b68ef network: create statedir during driver initialization
This directory should be created when the network driver is first
started up, not just when a dhcp daemon is run. This hasn't posed a
problem in the past, because the directory has always been
pre-existing.
2014-04-27 12:19:13 +03:00
Laine Stump
440beeb7ac 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.
2014-04-27 11:02:05 +03:00