The nwfilter conf update mutex previously serialized
updates to the internal data structures for firewall
rules, and updates to the firewall itself. The latter
was recently turned into a read/write lock, and filter
instantiation allowed to proceed in parallel. It was
believed that this was ok, since each filter is created
on a separate iptables/ebtables chain.
It turns out that there is a subtle lock ordering problem
on virNWFilterObjPtr instances. __virNWFilterInstantiateFilter
will hold a lock on the virNWFilterObjPtr it is instantiating.
This in turn invokes virNWFilterInstantiate which then invokes
virNWFilterDetermineMissingVarsRec which then invokes
virNWFilterObjFindByName. This iterates over every single
virNWFilterObjPtr in the list, locking them and checking their
name. So if 2 or more threads try to instantiate a filter in
parallel, they'll all hold 1 lock at the top level in the
__virNWFilterInstantiateFilter method which will cause the
other thread to deadlock in virNWFilterObjFindByName.
The fix is to add an exclusive mutex to serialize the
execution of __virNWFilterInstantiateFilter.
The NWFilter code has as a deadlock race condition between
the virNWFilter{Define,Undefine} APIs and starting of guest
VMs due to mis-matched lock ordering.
In the virNWFilter{Define,Undefine} codepaths the lock ordering
is
This has the effect of serializing VM startup once again, even if
no nwfilters are applied to the guest. There is also the possibility
of deadlock due to a call graph loop via virNWFilterInstantiate
and virNWFilterInstantiateFilterLate.
These two problems mean the lock must be turned into a read/write
lock instead of a plain mutex at the same time. The lock is used to
serialize changes to the "driver->nwfilters" hash, so the write lock
only needs to be held by the define/undefine methods. All other
methods can rely on a read lock which allows good concurrency.
Conflicts:
src/conf/nwfilter_conf.c
- virReportOOMError() in context of one hunk.
src/lxc/lxc_driver.c
- functions renamed, and lxc object locking changed, creating
a conflict in the context.
src/qemu/qemu_driver.c
- qemuDomainStartWithFlags (called qemuDomainCreateWithFlags
upstream) gets the domain object using
qemuDomObjFromDomain() upstream, but
virDomainObjListFindByUUID() in 1.0.4. This creates a
small conflict in context.
Remove use of virConnectPtr from all remaining nwfilter code
The virConnectPtr is passed around loads of nwfilter code in
order to provide it as a parameter to the callback registered
by the virt drivers. None of the virt drivers use this param
though, so it serves no purpose.
Avoiding the need to pass a virConnectPtr means that the
nwfilterStateReload method no longer needs to open a bogus
QEMU driver connection. This addresses a race condition that
can lead to a crash on startup.
The nwfilter driver starts before the QEMU driver and registers
some callbacks with DBus to detect firewalld reload. If the
firewalld reload happens while the QEMU driver is still starting
up though, the nwfilterStateReload method will open a connection
to the partially initialized QEMU driver and cause a crash.
Conflicts:
src/nwfilter/nwfilter_driver.c
- *EnsureACL*() was added after this branch was created, and
caused two small conflicts in the context around a hunk.
- nwfilterDriverReload() was renamed to nwfilterStateReload()
upstream.
Don't pass virConnectPtr in nwfilter 'struct domUpdateCBStruct'
The nwfilter driver only needs a reference to its private
state object, not a full virConnectPtr. Update the domUpdateCBStruct
struct to have a 'void *opaque' field instead of a virConnectPtr.
For inexplicable reasons, the nwfilter XML parser is intentionally
ignoring errors that arise during parsing. As well as meaning that
users don't get any feedback on their XML mistakes, this will lead
it to silently drop data in OOM conditions.
When writing commit 173c291, I missed the fact virNetServerClientClose
unlocks the client object before actually clearing client->sock and thus
it is possible to hit a window when client->keepalive is NULL while
client->sock is not NULL. I was thinking client->sock == NULL was a
better check for a closed connection but apparently we have to go with
client->keepalive == NULL to actually fix the crash.
When a client closes its connection to libvirtd early during
virConnectOpen, more specifically just after making
REMOTE_PROC_CONNECT_SUPPORTS_FEATURE call to check if
VIR_DRV_FEATURE_PROGRAM_KEEPALIVE is supported without even waiting for
the result, libvirtd may crash due to a race in keep-alive
initialization. Once receiving the REMOTE_PROC_CONNECT_SUPPORTS_FEATURE
call, the daemon's event loop delegates it to a worker thread. In case
the event loop detects EOF on the connection and calls
virNetServerClientClose before the worker thread starts to handle
REMOTE_PROC_CONNECT_SUPPORTS_FEATURE call, client->keepalive will be
disposed by the time virNetServerClientStartKeepAlive gets called from
remoteDispatchConnectSupportsFeature. Because the flow is common for
both authenticated and read-only connections, even unprivileged clients
may cause the daemon to crash.
To avoid the crash, virNetServerClientStartKeepAlive needs to check if
the connection is still open before starting keep-alive protocol.
Every libvirt release since 0.9.8 is affected by this bug.
Jiri Denemark [Fri, 20 Dec 2013 13:50:02 +0000 (14:50 +0100)]
qemu: Avoid using stale data in virDomainGetBlockInfo
CVE-2013-6458
Generally, every API that is going to begin a job should do that before
fetching data from vm->def. However, qemuDomainGetBlockInfo does not
know whether it will have to start a job or not before checking vm->def.
To avoid using disk alias that might have been freed while we were
waiting for a job, we use its copy. In case the disk was removed in the
meantime, we will fail with "cannot find statistics for device '...'"
error message.
When virDomainDetachDeviceFlags is called concurrently to
virDomainBlockStats: libvirtd may crash because qemuDomainBlockStats
finds a disk in vm->def before getting a job on a domain and uses the
disk pointer after getting the job. However, the domain in unlocked
while waiting on a job condition and thus data behind the disk pointer
may disappear. This happens when thread 1 runs
virDomainDetachDeviceFlags and enters monitor to actually remove the
disk. Then another thread starts running virDomainBlockStats, finds the
disk in vm->def, and while it's waiting on the job condition (owned by
the first thread), the first thread finishes the disk removal. When the
second thread gets the job, the memory pointed to be the disk pointer is
already gone.
That said, every API that is going to begin a job should do that before
fetching data from vm->def.
Eric Blake [Tue, 17 Dec 2013 23:28:20 +0000 (16:28 -0700)]
tests: be more explicit on qcow2 versions in virstoragetest
While working on v1.0.5-maint (the branch in use on Fedora 19)
with the host at Fedora 20, I got a failure in virstoragetest.
I traced it to the fact that we were using qemu-img to create a
qcow2 file, but qemu-img changed from creating v2 files by
default in F19 to creating v3 files in F20. Rather than leaving
it up to qemu-img, it is better to write the test to force
testing of BOTH file formats (better code coverage and all).
This patch alone does not fix all the failures in v1.0.5-maint;
for that, we must decide to either teach the older branch to
understand v3 files, or to reject them outright as unsupported.
But for upstream, making the test less dependent on changing
qemu-img defaults is always a good thing.
* tests/virstoragetest.c (testPrepImages): Simplify creation of
raw file; check if qemu supports compat and if so use it.
Jim Fehlig [Thu, 16 May 2013 14:42:21 +0000 (08:42 -0600)]
libxl: fix build with Xen4.3
Xen 4.3 fixes a mistake in the libxl event handler signature where the
event owned by the application was defined as const. Detect this and
define the libvirt libxl event handler signature appropriately.
(cherry picked from commit 43b0ff5b1eb7fbcdc348b2a53088a7db939d5e48)
Zhou Yimin [Thu, 17 Oct 2013 07:59:21 +0000 (15:59 +0800)]
remote: fix regression in event deregistration
Introduced by 7b87a3
When I quit the process which only register VIR_DOMAIN_EVENT_ID_REBOOT,
I got error like:
"libvirt: XML-RPC error : internal error: domain event 0 not registered".
Then I add the following code, it fixed.
Signed-off-by: Zhou Yimin <zhouyimin@huawei.com> Signed-off-by: Eric Blake <eblake@redhat.com>
(cherry picked from commit 9712c2510ec87a87578576a407768380e250a6a4)
Osier Yang [Fri, 24 May 2013 03:59:14 +0000 (11:59 +0800)]
virsh: Fix regression of vol-resize
Introduced by commit 1daa4ba33acf. vshCommandOptStringReq returns
0 on *success* or the option is not required && not present, both
are right result. Error out when returning 0 is not correct.
the caller, it doesn't have to check wether it
(cherry picked from commit 2a3a725c33aba2046443d33eb473eb54517f61c8)
Fix crash in remoteDispatchDomainMemoryStats (CVE-2013-4296)
The 'stats' variable was not initialized to NULL, so if some
early validation of the RPC call fails, it is possible to jump
to the 'cleanup' label and VIR_FREE an uninitialized pointer.
This is a security flaw, since the API can be called from a
readonly connection which can trigger the validation checks.
Add support for using 3-arg pkcheck syntax for process (CVE-2013-4311)
With the existing pkcheck (pid, start time) tuple for identifying
the process, there is a race condition, where a process can make
a libvirt RPC call and in another thread exec a setuid application,
causing it to change to effective UID 0. This in turn causes polkit
to do its permission check based on the wrong UID.
To address this, libvirt must get the UID the caller had at time
of connect() (from SO_PEERCRED) and pass a (pid, start time, uid)
triple to the pkcheck program.
Signed-off-by: Colin Walters <walters@redhat.com> Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
(cherry picked from commit 922b7fda77b094dbf022d625238262ea05335666)
Conflicts:
src/access/viraccessdriverpolkit.c
Resolution:
Dropped file that does not exist in this branch.
Include process start time when doing polkit checks
Since PIDs can be reused, polkit prefers to be given
a (PID,start time) pair. If given a PID on its own,
it will attempt to lookup the start time in /proc/pid/stat,
though this is subject to races.
It is safer if the client app resolves the PID start
time itself, because as long as the app has the client
socket open, the client PID won't be reused.
Peter Krempa [Fri, 16 Aug 2013 10:22:32 +0000 (12:22 +0200)]
virbitmap: Refactor virBitmapParse to avoid access beyond bounds of array
The virBitmapParse function was calling virBitmapIsSet() function that
requires the caller to check the bounds of the bitmap without checking
them. This resulted into crashes when parsing a bitmap string that was
exceeding the bounds used as argument.
This patch refactors the function to use virBitmapSetBit without
checking if the bit is set (this function does the checks internally)
and then counts the bits in the bitmap afterwards (instead of keeping
track while parsing the string).
This patch also changes the "parse_error" label to a more common
"error".
The refactor should also get rid of the need to call sa_assert on the
returned variable as the callpath should allow coverity to infer the
possible return values.
Dennis Chen [Fri, 28 Jun 2013 09:59:51 +0000 (11:59 +0200)]
Fix vPort management: FC vHBA creation
When creating a virtual FC HBA with virsh/libvirt API, an error message
will be returned: "error: Node device not found",
also the 'nodedev-dumpxml' shows wrong information of wwpn & wwnn
for the new created device.
Signed-off-by: xschen@tnsoft.com.cn
This reverts f90af69 which switched wwpn & wwwn in the wrong place.
Ján Tomko [Wed, 26 Jun 2013 11:07:24 +0000 (13:07 +0200)]
Fix invalid read in virCgroupGetValueStr
Don't check for '\n' at the end of file if zero bytes were read.
Found by valgrind:
==404== Invalid read of size 1
==404== at 0x529B09F: virCgroupGetValueStr (vircgroup.c:540)
==404== by 0x529AF64: virCgroupMoveTask (vircgroup.c:1079)
==404== by 0x1EB475: qemuSetupCgroupForEmulator (qemu_cgroup.c:1061)
==404== by 0x1D9489: qemuProcessStart (qemu_process.c:3801)
==404== by 0x18557E: qemuDomainObjStart (qemu_driver.c:5787)
==404== by 0x190FA4: qemuDomainCreateWithFlags (qemu_driver.c:5839)
Ján Tomko [Tue, 25 Jun 2013 13:10:27 +0000 (15:10 +0200)]
virsh: edit: don't leak XML string on reedit or redefine
Free the old XML strings before overwriting them if the user
has chosen to reedit the file or force the redefinition.
Found by Alex Jia trying to reproduce another bug:
https://bugzilla.redhat.com/show_bug.cgi?id=977430#c3
(cherry picked from commit 1e3a252974c8e5c650f1d84dc2b167f0ae8cee3c)
As a consequence of the cgroup layout changes from commit 'cfed9ad4', the
lxcDomainGetSchedulerParameters[Flags]()' and lxcGetSchedulerType() APIs
failed to return data for a non running domain. This can be seen through
a 'virsh schedinfo <domain>' command which returns:
Scheduler : Unknown
error: Requested operation is not valid: cgroup CPU controller is not mounted
Prior to that change a non running domain would return:
This patch will restore the capability to return configuration only data
for a non running domain regardless of whether cgroups are available.
Conflicts:
src/lxc/lxc_driver.c
* Resolved conflict by using former lxcCgroupHasController() rather than
virCgroupHasController()
* Needed to add the code to fetch the 'vm'
vm = virDomainObjListFindByUUID(driver->domains, domain->uuid);
if (vm == NULL) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("No such domain %s"), domain->uuid);
goto cleanup;
}
* Used 'ret = strdup("posix");' rather than VIR_STRDUP(ret, "posix");
and added the virReportOOMError(); on failure.
As a consequence of the cgroup layout changes from commit '632f78ca', the
qemuDomainGetSchedulerParameters[Flags]()' and qemuGetSchedulerType() APIs
failed to return data for a non running domain. This can be seen through
a 'virsh schedinfo <domain>' command which returns:
Scheduler : Unknown
error: Requested operation is not valid: cgroup CPU controller is not mounted
Prior to that change a non running domain would return:
This patch will restore the capability to return configuration only data
for a non running domain regardless of whether cgroups are available.
Conflicts:
src/qemu/qemu_driver.c
* Resolved conflict by using former qemuCgroupHasController() rather than
virCgroupHasController()
* Needed to add the code to fetch the 'vm'
vm = virDomainObjListFindByUUID(driver->domains, dom->uuid);
if (vm == NULL) {
virReportError(VIR_ERR_INTERNAL_ERROR,
_("No such domain %s"), dom->uuid);
goto cleanup;
}
* Used 'ret = strdup("posix");' rather than VIR_STRDUP(ret, "posix");
and added the virReportOOMError(); on failure.
The problem was that qemuUpdateActivePciHostdevs was returning 0
(success) when no hostdevs were present, but would otherwise return -1
(failure) even when it completed successfully. It is only called from
qemuProcessReconnect(), and when qemuProcessReconnect got back an
error, it would not only stop reconnecting, but would terminate the
guest qemu process "to remove danger of it ending up running twice if
user tries to start it again later".
(This bug was introduced in commit 011cf7ad, which was pushed between
v1.0.2 and v1.0.3, so all maintenance branches from v1.0.3 up to 1.0.5
will need this one line patch applied.)
(cherry picked from commit 2ea45647bcde23cff5da48f725561ff5ba3fba39)
Ján Tomko [Fri, 12 Apr 2013 15:30:56 +0000 (17:30 +0200)]
daemon: fix leak after listing all volumes
CVE-2013-1962
remoteDispatchStoragePoolListAllVolumes wasn't freeing the pool.
The pool also held a reference to the connection, preventing it from
getting freed and closing the netcf interface driver, which held two
sockets open.
(cherry picked from commit ca697e90d5bd6a6dfb94bfb6d4438bdf9a44b739)
Eric Blake [Mon, 15 Apr 2013 14:54:53 +0000 (08:54 -0600)]
maint: update to latest gnulib
Upstream gnulib determined that we were needlessly compiling in
gnulib's regex instead of glibc's when targetting new-enough glibc,
because the m4 test was being too strict in requiring a particular
answer to undefined behavior.
https://lists.gnu.org/archive/html/bug-gnulib/2013-04/msg00032.html
Eric Blake [Mon, 1 Apr 2013 18:49:05 +0000 (12:49 -0600)]
maint: update to latest gnulib
While this update doesn't address any reported problems in libvirt,
doing a post-release update to latest gnulib makes it easier to
stay in sync with best upstream practices.
Commit d04916fa introduced a regression in audit quality - even
though the code was computing the proper escaped name for a
path, it wasn't feeding that escaped name on to the audit message.
As a result, /var/log/audit/audit.log would mention a pair of
fields class=path path=/dev/hpet instead of the intended
class=path path="/dev/hpet", which in turn caused ausearch to
format the audit log with path=(null).
* src/conf/domain_audit.c (virDomainAuditCgroupPath): Use
constructed encoding.
Since commit b8a32e0e94d75702714167539310f0df4d268e0f, all man pages
depend on configure.ac so that they are properly regenerated whenever
libvirt version changes. Thus libvirt.spec needs to have a build
dependency on pod2man when %{enable_autotools} is set.
(cherry picked from commit 6f1b9c8d2ab3f28d1b94f6aca0e2695632fa7019)
Eric Blake [Fri, 5 Apr 2013 17:09:32 +0000 (11:09 -0600)]
build: check correct protocol.o file
By default, libtool builds two .o files for every .lo rule:
src/foo.o - static builds
src/.libs/foo.o - shared library builds
But since commit ad42b34b disabled static builds, src/foo.o is
no longer built by default. On a fresh checkout, this means our
protocol check rules using pdwtags were testing a missing file,
and thanks to a lousy behavior of pdwtags happily giving no output
and 0 exit status (http://bugzilla.redhat.com/949034), we were
merely claiming that "dwarves is too old" and skipping the test.
However, if you swap between branches and do incremental builds,
such as building v0.10.2-maint and then switching back to master,
you end up with src/foo.o being leftover from its 0.10.2 state,
and then 'make check' fails because the .o file does not match
the protocol-structs file due to API additions in the meantime.
A simpler fix would be to always look in .libs for the .o to
be parsed; but since it is possible to pass ./configure options
to tell libtool to do a static-only build with no shared .o,
I went with the approach of finding the newest of the two files,
whenever both exist.
The linker will ignore LD_PRELOAD libraries which do not
exist, just printing a warning message. This is not helpful
for the test suite which will be utterly fubar without the
preload library present. Add an explicit test for existence
of the library to protect against this
Peter Krempa [Fri, 29 Mar 2013 17:21:19 +0000 (18:21 +0100)]
rpc: Fix connection close callback race condition and memory corruption/crash
The last Viktor's effort to fix the race and memory corruption unfortunately
wasn't complete in the case the close callback was not registered in an
connection. At that time, the trail of event's that I'll describe later could
still happen and corrupt the memory or cause a crash of the client (including
the daemon in case of a p2p migration).
Consider the following prerequisities and trail of events:
Let's have a remote connection to a hypervisor that doesn't have a close
callback registered and the client is using the event loop. The crash happens in
cooperation of 2 threads. Thread E is the event loop and thread W is the worker
that does some stuff. R denotes the remote client.
1.) W - The client finishes everything and sheds the last reference on the client
2.) W - The virObject stuff invokes virConnectDispose that invokes doRemoteClose
3.) W - the remote close method invokes the REMOTE_PROC_CLOSE RPC method.
4.) W - The thread is preempted at this point.
5.) R - The remote side receives the close and closes the socket.
6.) E - poll() wakes up due to the closed socket and invokes the close callback
7.) E - The event loop is preempted right before remoteClientCloseFunc is called
8.) W - The worker now finishes, and frees the conn object.
9.) E - The remoteClientCloseFunc accesses the now-freed conn object in the
attempt to retrieve pointer for the real close callback.
10.) Kaboom, corrupted memory/segfault.
This patch tries to fix this by introducing a new object that survives the
freeing of the connection object. We can't increase the reference count on the
connection object itself or the connection would never be closed, as the
connection is closed only when the reference count reaches zero.
The new object - virConnectCloseCallbackData - is a lockable object that keeps
the pointers to the real user registered callback and ensures that the
connection callback is either not called if the connection was already freed or
that the connection isn't freed while this is being called.
(cherry picked from commit 8ad126e695e5cef5da9d62ccfde7338317041e84)
Peter Krempa [Wed, 27 Mar 2013 13:37:01 +0000 (14:37 +0100)]
virsh: Register and unregister the close callback also in cmdConnect
This patch improves the error message after disconnecting from the
hypervisor and adds the close callback operations required not to leak
the callback reference.
(cherry picked from commit 69ab07560a134e82e36b6391be3c806d3dbdb16c)
Peter Krempa [Wed, 27 Mar 2013 13:22:47 +0000 (14:22 +0100)]
virsh: Move cmdConnect from virsh-host.c to virsh.c
The function is used to establish connection so it should be in the main
virsh file. This movement also enables further improvements done in next
patches.
Note that the "connect" command has moved from the host section of virsh to the
main section. It is now listed by 'virsh help virsh' instead of 'virsh help
host'.
(cherry picked from commit ca9e73ebb60e2efb1ea835e9a394a8b64ecb97c1)
When creating a logical volume with virStorageVolCreateXMLFrom,
"qemu-img convert" is called internally if clonevol is a file volume.
Then, vol->target.format is used as output_fmt parameter but the
target.format of logical volumes is always 0 because logical volumes
haven't the volume format type element.
Fortunately, 0 was treated as RAW file format before commit f772b3d9,
so there was no problem. But now, 0 is treated as the type of none,
qemu-img fails with "Unknown file format 'none'".
This patch fixes this issue by treating output block devices as RAW
file format like for input block devices.
By passing the flags -z relro -z now to the linker, we can force
it to resolve all library symbols at startup, instead of on-demand.
This allows it to then make the global offset table (GOT) read-only,
which makes some security attacks harder.
PIE (position independent executable) adds security to executables
by composing them entirely of position-independent code (PIC. The
.so libraries already build with -fPIC. This adds -fPIE which is
the equivalent to -fPIC, but for executables. This for allows Exec
Shield to use address space layout randomization to prevent attackers
from knowing where existing executable code is during a security
attack using exploits that rely on knowing the offset of the
executable code in the binary, such as return-to-libc attacks.
The virsh domfstrim command was not freeing allocated domain,
leaving leaked references behind.
(cherry picked from commit deb86ee9123ef47dce80dd77a9bc583f2b0214db)
TEST: qemuxml2argvtest
........................................ 40
........................................ 80
........................................ 120
........................................ 160
........................................ 200
........................................ 240
................................. 273 OK
==30993== 39 bytes in 1 blocks are definitely lost in loss record 33 of 87
==30993== at 0x4A0887C: malloc (vg_replace_malloc.c:270)
==30993== by 0x41E501: fakeSecretGetValue (qemuxml2argvtest.c:33)
==30993== by 0x427591: qemuBuildDriveURIString (qemu_command.c:2571)
==30993== by 0x42C502: qemuBuildDriveStr (qemu_command.c:2627)
==30993== by 0x4335FC: qemuBuildCommandLine (qemu_command.c:6443)
==30993== by 0x41E8A0: testCompareXMLToArgvHelper (qemuxml2argvtest.c:154
==30993== by 0x41FE8F: virtTestRun (testutils.c:157)
==30993== by 0x418BE3: mymain (qemuxml2argvtest.c:506)
==30993== by 0x4204CA: virtTestMain (testutils.c:719)
==30993== by 0x38D6821A04: (below main) (in /usr/lib64/libc-2.16.so)
==30993==
==30993== 46 bytes in 1 blocks are definitely lost in loss record 64 of 87
==30993== at 0x4A0887C: malloc (vg_replace_malloc.c:270)
==30993== by 0x38D690A167: __vasprintf_chk (in /usr/lib64/libc-2.16.so)
==30993== by 0x4CB28E7: virVasprintf (stdio2.h:210)
==30993== by 0x4CB29A3: virAsprintf (virutil.c:2017)
==30993== by 0x4275B4: qemuBuildDriveURIString (qemu_command.c:2580)
==30993== by 0x42C502: qemuBuildDriveStr (qemu_command.c:2627)
==30993== by 0x4335FC: qemuBuildCommandLine (qemu_command.c:6443)
==30993== by 0x41E8A0: testCompareXMLToArgvHelper (qemuxml2argvtest.c:154
==30993== by 0x41FE8F: virtTestRun (testutils.c:157)
==30993== by 0x418BE3: mymain (qemuxml2argvtest.c:506)
==30993== by 0x4204CA: virtTestMain (testutils.c:719)
==30993== by 0x38D6821A04: (below main) (in /usr/lib64/libc-2.16.so)
==30993==
==30993== 385 (56 direct, 329 indirect) bytes in 1 blocks are definitely los
==30993== at 0x4A06B6F: calloc (vg_replace_malloc.c:593)
==30993== by 0x4C6B2CF: virAllocN (viralloc.c:152)
==30993== by 0x4C9C7EB: virObjectNew (virobject.c:191)
==30993== by 0x4D21810: virGetSecret (datatypes.c:642)
==30993== by 0x41E5D5: fakeSecretLookupByUsage (qemuxml2argvtest.c:51)
==30993== by 0x4D4BEC5: virSecretLookupByUsage (libvirt.c:15295)
==30993== by 0x4276A9: qemuBuildDriveURIString (qemu_command.c:2565)
==30993== by 0x42C502: qemuBuildDriveStr (qemu_command.c:2627)
==30993== by 0x4335FC: qemuBuildCommandLine (qemu_command.c:6443)
==30993== by 0x41E8A0: testCompareXMLToArgvHelper (qemuxml2argvtest.c:154
==30993== by 0x41FE8F: virtTestRun (testutils.c:157)
==30993== by 0x418BE3: mymain (qemuxml2argvtest.c:506)
==30993==
PASS: qemuxml2argvtest
Interesting side note is that running the test singularly via 'make -C tests
check TESTS=qemuxml2argvtest' didn't trip the valgrind error; however,
running during 'make -C tests valgrind' did cause the error to be seen.
(cherry picked from commit 9a80050e523da636a11d32d507ede11af764b9e0)
Laine Stump [Tue, 9 Apr 2013 18:06:51 +0000 (14:06 -0400)]
Fix crash in virNetDevGetVirtualFunctions
Commit 9a3ff01d7f16cc280ce3176620c0714f55511a65 (which was ACKed at
the end of January, but for some reason didn't get pushed until during
the 1.0.4 freeze) fixed the logic in virPCIGetVirtualFunctions().
Unfortunately, a typo in the fix (replacing VIR_REALLOC_N with
VIR_ALLOC_N during code movement) caused not only a memory leak, but
also resulted in most of the elements of the result array being
replaced with NULL. virNetDevGetVirtualFunctions() assumed (and I think
rightly so) that virPCIGetVirtualFunctions() wouldn't return any NULL
elements in the array, so it ended up segfaulting.
This was found when attempting to use a virtual network with an
auto-created pool of SRIOV VFs, e.g.:
Ján Tomko [Fri, 29 Mar 2013 11:55:38 +0000 (12:55 +0100)]
virsh: don't call virSecretFree on NULL
Since the refactoring in fbe2d49 we call virSecretFree even if
virSecretDefineXML fails, which leads to overwriting the error
message with:
error: Invalid secret: virSecretFree
storage: Avoid double virCommandFree in virStorageBackendLogicalDeletePool
When logical pool has no PVs associated with itself (user-created),
virCommandFree(cmd) is called twice with the same pointer and that
causes a segfault in daemon.
Michal Privoznik [Thu, 28 Mar 2013 15:13:01 +0000 (16:13 +0100)]
security_manager.c: Append seclabel iff generated
With my previous patches, we unconditionally appended a seclabel,
even if it wasn't generated but found in array of defined seclabels.
This resulted in double free later when doing virDomainDefFree
and iterating over the array of defined seclabels.
Moreover, there was another possibility of double free, if the
seclabel was generated in the last iteration of the process of
walking trough security managers array.
One of my previous patches manipulated virSecurityLabel* APIs,
some were added to header files, and some were renamed. However,
these changes were not reflected in libvirt_private.syms.
The <seclabel type='none'/> should be added iff there is no other
seclabel defined within a domain. This bug can be easily reproduced:
1) configure selinux seclabel for a domain
2) disable system's selinux and restart libvirtd
3) observe <seclabel type='none'/> being appended to a domain on its
startup
Michal Privoznik [Thu, 21 Mar 2013 15:12:55 +0000 (16:12 +0100)]
security_manager: Don't manipulate domain XML in virDomainDefGetSecurityLabelDef
The virDomainDefGetSecurityLabelDef was modifying the domain XML.
It tried to find a seclabel corresponding to given sec driver. If the
label wasn't found, the function created one which is wrong. In fact
it's security manager which should modify this part of domain XML.
Guannan Ren [Thu, 28 Mar 2013 04:10:05 +0000 (12:10 +0800)]
conf: fix memory leak of class_id bitmap
When libvirtd loads active network configs from network state directory,
it should release the class_id memory block which was allocated
at the time of loading xml from network config directory.
virBitmapParse will create a new memory block of bitmap class_id which
causes a memory leak.
This happens when at least one virtual network is active before.
==12234== 8,216 (24 direct, 8,192 indirect) bytes in 1 blocks are definitely \
lost in loss record 702 of 709
==12234== at 0x4A06B2F: calloc (vg_replace_malloc.c:593)
==12234== by 0x37AB04D77D: virAlloc (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x37AB04EF89: virBitmapNew (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x37AB0BFB37: virNetworkAssignDef (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x37AB0BFD31: ??? (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x37AB0BFE92: virNetworkLoadAllConfigs (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x10650E5A: ??? (in /usr/lib64/libvirt/connection-driver/libvirt_driver_network.so)
==12234== by 0x37AB0EB72F: virStateInitialize (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x40DE04: ??? (in /usr/sbin/libvirtd)
==12234== by 0x37AB0832E8: ??? (in /usr/lib64/libvirt.so.0.1000.3)
==12234== by 0x3796807D14: start_thread (in /usr/lib64/libpthread-2.16.so)
==12234== by 0x37960F246C: clone (in /usr/lib64/libc-2.16.so)
Stefan Seyfried [Mon, 25 Mar 2013 19:39:40 +0000 (20:39 +0100)]
net: use newer iptables syntax
iptables-1.4.18 removed the long deprecated "state" match.
Use "conntrack" instead in forwarding rules.
Fixes openSUSE bug https://bugzilla.novell.com/811251 #811251.
Jiri Denemark [Tue, 26 Mar 2013 14:50:38 +0000 (15:50 +0100)]
rpc: Fix client crash when server drops connection
Despite the comment stating virNetClientIncomingEvent handler should
never be called with either client->haveTheBuck or client->wantClose
set, there is a sequence of events that may lead to both booleans being
true when virNetClientIncomingEvent is called. However, when that
happens, we must not immediately close the socket as there are other
threads waiting for the buck and they would cause SIGSEGV once they are
woken up after the socket was closed. Another thing is we should clear
all remaining calls in the queue after closing the socket.
The situation that can lead to the crash involves three threads, one of
them running event loop and the other two calling libvirt APIs. The
event loop thread detects an event on client->sock and calls
virNetClientIncomingEvent handler. But before the handler gets a chance
to lock client, the other two threads (T1 and T2) start calling some
APIs. T1 gets the buck and detects EOF on client->sock while processing
its RPC call. Since T2 is waiting for its own call, T1 passes the buck
on to it and unlocks client. But before T2 gets the signal, the event
loop thread wakes up, does its job and closes client->sock. The crash
happens when T2 actually wakes up and tries to do its job using a closed
client->sock.
Jiri Denemark [Mon, 25 Mar 2013 10:35:27 +0000 (11:35 +0100)]
log: Separate thread ID from timestemp in ring buffer
When we write a log message into a log, we separate thread ID from
timestamp using ": ". However, when storing the message into the ring
buffer, we omitted the separator, e.g.:
Guannan Ren [Tue, 26 Mar 2013 14:17:43 +0000 (22:17 +0800)]
conf: fix a failure when detaching a usb device
#virsh detach-device $guest usb.xml
error: Failed to detach device from usb2.xml
error: operation failed: host usb device vendor=0x0951 \
product=0x1625 not found
This regresstion is due to a typo in matching function. The first
argument is always the usb device that we are checking for. If the
usb xml file provided by user contains bus and device info, we try
to search it by them, otherwise, we use vendor and product info.
The bug occurred only when detaching a usb device with no bus and
device info provided in the usb xml file.
Guido Günther [Fri, 22 Mar 2013 09:25:42 +0000 (10:25 +0100)]
qemu: Don't set address type too early during virtio disk hotplug
f946462e14ac036357b7c11ce5c23f94a3ee4e49 changed behavior by settings
VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI upfront. If we do so before invoking
qemuDomainPCIAddressEnsureAddr we merely try to set the PCI slot via
qemuDomainPCIAddressReserveSlot instead reserving a new address via
qemuDomainPCIAddressSetNextAddr which fails with
$ ~/run-tck-test domain/200-disk-hotplug.t
./scripts/domain/200-disk-hotplug.t .. # Creating a new transient domain
./scripts/domain/200-disk-hotplug.t .. 1/5 # Attaching the new disk /var/lib/jenkins/jobs/libvirt-tck-build/workspace/scratchdir/200-disk-hotplug/extra.img
# Failed test 'disk has been attached'
# at ./scripts/domain/200-disk-hotplug.t line 67.
# died: Sys::Virt::Error (libvirt error code: 1, message: internal error unable to reserve PCI address 0:0:0.0
# )
Michal Privoznik [Tue, 26 Mar 2013 14:45:16 +0000 (15:45 +0100)]
qemu: Set migration FD blocking
Since we switched from direct host migration scheme to the one,
where we connect to the destination and then just pass a FD to a
qemu, we have uncovered a qemu bug. Qemu expects migration FD to
block. However, we are passing a nonblocking one which results in
cryptic error messages like:
qemu: warning: error while loading state section id 2
load of migration failed
The bug is already known to Qemu folks, but we should workaround
already released Qemus. Patch has been originally proposed by Stefan
Hajnoczi <stefanha@gmail.com>
virConnectOpenAuth didn't require 'name' to be specified (VIR_DEBUG
used NULLSTR() for the output) and by default, if name == NULL, the
default connection uri is used. This was not indicated in the
documentation and wasn't checked for in other API's VIR_DEBUG outputs.
Guannan Ren [Tue, 26 Mar 2013 04:34:49 +0000 (12:34 +0800)]
python: set default value to optional arguments
When prefixing with string (optional) or optional in the description
of arguments to libvirt C APIs, in python, these arguments will be
set as optional arugments, for example:
* virDomainSaveFlags:
* @domain: a domain object
* @to: path for the output file
* @dxml: (optional) XML config for adjusting guest xml used on restore
* @flags: bitwise-OR of virDomainSaveRestoreFlags
the corresponding python APIs is
restoreFlags(self, frm, dxml=None, flags=0)
After further discussions with Alon Levy, I learned the following:
The use of '-vga qxl' vs. '-device qxl-vga' is completely orthogonal
to whether ram_size can be exposed. Downstream distros are interested
in backporting support for multi-head qxl, but this can be done in
one of two ways:
1. Support one head per PCI device. If you do this, then it makes
sense to have full control over the PCI address of each device. For
full control, you need '-device qxl-vga' instead of '-vga qxl'.
2. Support multiple heads through a single PCI device. If you do
this, then you need to allocate more RAM to that PCI device (enough
ram to cover the multiple screens). Here, the device is hard-coded
to 0:0:2.0, both in qemu and libvirt code.
Apparently, backporting ram_size changes to allow multiple heads in
a single device is much easier than backporting multiple device
support. Furthermore, the presence or absence of qxl-vga.surfaces
is no different than the presence or absence of qxl-vga.ram_size;
both properties can be applied regardless of whether you have one
PCI device (-vga qxl) or multiple (-device qxl-vga), so this property
is NOT a good witness of whether '-device qxl-vga' support has been
backported.
Downstream RHEL will NOT be using this patch; and worse, leaving this
patch in risks doing the wrong thing if compiling upstream libvirt
on RHEL, so the best course of action is to revert it. That means
that libvirt will go back to only using '-device qxl-vga' for qemu
>= 1.2, but this is just fine because we know of no distros that plan
on backporting multiple PCI address support to any older version of
qemu. Meanwhile, downstream can still use ram_size to pack multiple
heads through a single PCI device.
$ service libvirt-guests restart
Running guests on default URI: a, b, d, c
Shutting down guests on default URI...
Starting shutdown on guest: a
Shutdown of guest a failed to complete in time.Starting shutdown on guest: b
Shutdown of guest b failed to complete in time.Starting shutdown on guest: d
Shutdown of guest d failed to complete in time.Starting shutdown on guest: c
Shutdown of guest c failed to complete in time.libvirt-guests is configured not to start any guests on boot
* tools/libvirt-guests.sh.in (shutdown_guest): Add missing newline.
Reported by Xuesong Zhang.
to tag the PCI device with "virtual_function" cap.
However, this results in a VF will aslo get "virtual_function" cap.
This patch fixes it by:
* Ignoring the VF which has failure of getting PCI config space
(given that the successfully detected VFs are kept , it makes
sense to not give up on the failure of one VF too) with a warning,
so virPCIGetVirtualFunctions will not return -1 except out of memory.
* Free the allocated *virtual_functions when out of memory
And thus the logic can be changed to:
/* Out of memory */
int ret = virPCIGetVirtualFunctions(syspath,
&data->pci_dev.virtual_functions,
&data->pci_dev.num_virtual_functions);
if (ret < 0 )
goto out;
if (data->pci_dev.num_virtual_functions > 0)
data->pci_dev.flags |= VIR_NODE_DEV_CAP_FLAG_PCI_VIRTUAL_FUNCTION;
Osier Yang [Mon, 7 Jan 2013 17:05:34 +0000 (01:05 +0800)]
nodedev: Abstract nodeDeviceVportCreateDelete as util function
This abstracts nodeDeviceVportCreateDelete as an util function
virManageVport, which can be further used by later storage patches
(to support persistent vHBA, I don't want to create the vHBA
using the public API, which is not good).
* docs/formatnode.html.in: (Document the new XML)
* docs/schemas/nodedev.rng: (Add the schema)
* src/conf/node_device_conf.h: (New member for data.scsi_host)
* src/node_device/node_device_linux_sysfs.c: (Collect the value of
max_vports and vports)
Osier Yang [Mon, 7 Jan 2013 17:05:31 +0000 (01:05 +0800)]
nodedev: Refactor the helpers
This adds two util functions (virIsCapableFCHost and virIsCapableVport),
and rename helper check_fc_host_linux as detect_scsi_host_caps,
check_capable_vport_linux is removed, as it's abstracted to the util
function virIsCapableVport. detect_scsi_host_caps nows detect both
the fc_host and vport_ops capabilities. "stat(2)" is replaced with
"access(2)" for saving.
* src/util/virutil.h:
- Declare virIsCapableFCHost and virIsCapableVport
* src/util/virutil.c:
- Implement virIsCapableFCHost and virIsCapableVport
* src/node_device/node_device_linux_sysfs.c:
- Remove check_capable_vport_linux
- Rename check_fc_host_linux as detect_scsi_host_caps, and refactor
it a bit to detect both fc_host and vport_os capabilities
* src/node_device/node_device_driver.h:
- Change/remove the related declarations
* src/node_device/node_device_udev.c: (Use detect_scsi_host_caps)
* src/node_device/node_device_hal.c: (Likewise)
* src/node_device/node_device_driver.c (Likewise)
Osier Yang [Mon, 7 Jan 2013 17:05:30 +0000 (01:05 +0800)]
nodedev: Use access instead of stat
The use of 'stat' in nodeDeviceVportCreateDelete is only to check
if the file exists or not, it's a bit overkill, and safe to replace
with the wrapper of access(2) (virFileExists).
Osier Yang [Mon, 7 Jan 2013 17:05:29 +0000 (01:05 +0800)]
util: Add one helper virReadFCHost to read the value of fc_host entry
"open_wwn_file" in node_device_linux_sysfs.c is redundant, on one
hand it duplicates work of virFileReadAll, on the other hand, it's
waste to use a function for it, as there is no other users of it.
So I don't see why the file opening work cannot be done in
"read_wwn_linux".
"read_wwn_linux" can be abstracted as an util function. As what all
it does is to read the sysfs entry.
So this patch removes "open_wwn_file", and abstract "read_wwn_linux"
as an util function "virReadFCHost" (a more general name, because
after changes, it can read each of the fc_host entry now).
* src/util/virutil.h: (Declare virReadFCHost)
* src/util/virutil.c: (Implement virReadFCHost)
* src/node_device/node_device_linux_sysfs.c: (Remove open_wwn_file,
and read_wwn_linux)
src/node_device/node_device_driver.h: (Remove the declaration of
read_wwn_linux, and the related macros)
src/libvirt_private.syms: (Export virReadFCHost)
Osier Yang [Mon, 7 Jan 2013 17:05:28 +0000 (01:05 +0800)]
nodedev: Introduce two new flags for listAll API
VIR_CONNECT_LIST_NODE_DEVICES_CAP_FC_HOST to filter the FC HBA,
and VIR_CONNECT_LIST_NODE_DEVICES_CAP_VPORTS to filter the FC HBA
which supports vport.
Osier Yang [Mon, 7 Jan 2013 17:05:27 +0000 (01:05 +0800)]
nodedev: Remove the unused enum
Guess it was created for the fc_host and vports_ops capabilities
purpose, but there is enum virNodeDevScsiHostCapFlags for them,
and enum virNodeDevHBACapType is unused, and actually both
VIR_ENUM_DECL and VIR_ENUM_IMPL use the wrong enum name
"virNodeDevHBACap".
Peter Krempa [Fri, 22 Mar 2013 10:05:36 +0000 (11:05 +0100)]
virsh: Fix docs for "virsh setmaxmem"
The docs assumed the command works always for QEMU and other
hypervisors. As this is done using the balloon mechainism live increase
of the maximum memory limit isn't supported. Fix the docs to mention
this limitation.
projects / thirdparty / libvirt.git / log
