Currently, after we removed the qemu driver lock, it may happen
that two or more threads will start up a machine with macvlan and
race over virNetDevMacVLanCreateWithVPortProfile(). However,
there's a racy section in which we are generating a sequence of
possible device names and detecting if they exits. If we found
one which doesn't we try to create a device with that name.
However, the other thread is doing just the same. Assume it will
succeed and we must therefore fail. If this happens more than 5
times (which in massive parallel startup surely will) we return
-1 without any error reported. This patch is a simple hack to
both of these problems. It introduces a mutex, so only one thread
will enter the section, and if it runs out of possibilities,
error is reported. Moreover, the number of retries is raised to 20.

This reverts the hack done in

commit 568a6cda
Author: Jiri Denemark <jdenemar@redhat.com>
Date:   Fri Feb 15 15:11:47 2013 +0100

    qemu: Avoid deadlock in autodestroy

since we now have a fix which avoids the deadlock scenario
entirely

There is a lock ordering problem in the QEMU close callback
APIs.

When starting a guest we have a lock on the VM. We then
set a autodestroy callback, which acquires a lock on the
close callbacks.

When running auto-destroy, we obtain a lock on the close
callbacks, then run each callbacks - which obtains a lock
on the VM.

This causes deadlock if anyone tries to start a VM, while
autodestroy is taking place.

The fix is to do autodestroy in 2 phases. First obtain
all the callbacks and remove them from the list under
the close callback lock. Then invoke each callback
from outside the close callback lock.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

When the auto-destroy callback runs it is supposed to return
NULL if the virDomainObjPtr is no longer valid. It was not
doing this for transient guests, so we tried to virObjectUnlock
a mutex which had been freed. This often led to a crash.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

qemuProcessStart expects to be run with a job already set and every
caller except for qemuMigrationPrepareAny use it correctly. This bug can
be observed in libvirtd logs during incoming migration as

    warning : qemuDomainObjEnterMonitorInternal:979 : This thread seems
    to be the async job owner; entering monitor without asking for a
    nested job is dangerous

With the apparmor security driver enabled, qemu instances fail
to start

# grep ^security_driver /etc/libvirt/qemu.conf
security_driver = "apparmor"
# virsh start test-kvm
error: Failed to start domain test-kvm
error: internal error security label already defined for VM

The model field of virSecurityLabelDef object is always populated
by virDomainDefGetSecurityLabelDef(), so remove the check for a
NULL model when verifying if a label is already defined for the
instance.

Checking for a NULL model and populating it later in
AppArmorGenSecurityLabel() has been left in the code to be
consistent with virSecuritySELinuxGenSecurityLabel().

As pointed out in
https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1034661
The sentence

"The function of PCI device addresses must less than 8"

does not quite make sense.  Update that to read

"The function of PCI device addresses must be less than 8"
Signed-off-by: NSerge Hallyn <serge.hallyn@ubuntu.com>

Currently, qemuDomainShutdownFlags() chooses the agent method of
shutdown whenever the agent is configured. However, this
assumption is not enough as the guest agent may be unresponsive
at the moment. So unless guest agent method has been explicitly
requested, we should fall back to the ACPI method.

The unitialized local variable qemuVersion can cause an random value
to be returned for the hypervisor version, observable with virsh version.
Introduced by commit b46f7f4aSigned-off-by: NViktor Mihajlovski <mihajlov@linux.vnet.ibm.com>

disk->src is still used for disks->hosts->name, do not free it.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NEric Blake <eblake@redhat.com>

* docs/formatnetwork.html.in: Spell variation correctly.

The QEMU driver has a list of devices nodes that are whitelisted
for all guests. The kernel has recently started returning an
error if you try to whitelist a device which does not exist.
This causes a warning in libvirt logs and an audit error for
any missing devices. eg

2013-02-27 16:08:26.515+0000: 29625: warning : virDomainAuditCgroup:451 : success=no virt=kvm resrc=cgroup reason=allow vm="vm031714" uuid=9d8f1de0-44f4-a0b1-7d50-e41ee6cd897b cgroup="/sys/fs/cgroup/devices/libvirt/qemu/vm031714/" class=path path=/dev/kqemu rdev=? acl=rw
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

s/VIR_QEMU_PROCESS_START_AUTODESROY/VIR_QEMU_PROCESS_START_AUTODESTROY/
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

The code for putting the emulator threads in a separate cgroup
would spam the logs with warnings

2013-02-27 16:08:26.731+0000: 29624: warning : virCgroupMoveTask:887 : no vm cgroup in controller 3
2013-02-27 16:08:26.731+0000: 29624: warning : virCgroupMoveTask:887 : no vm cgroup in controller 4
2013-02-27 16:08:26.732+0000: 29624: warning : virCgroupMoveTask:887 : no vm cgroup in controller 6

This is because it has only created child cgroups for 3 of the
controllers, but was trying to move the processes from all the
controllers. The fix is to only try to move threads in the
controllers we actually created. Also remove the warning and
make it return a hard error to avoid such lazy callers in the
future.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

The virQEMUCloseCallbacksRunOne method was passing a uuid string
to virDomainObjListFindByUUID, when it actually expected to get
a raw uuid buffer. This was not caught by the compiler because
the method was using a 'void *uuid' instead of first casting
it to the expected type.

This regression was accidentally caused by refactoring in

  commit 568a6cda
  Author: Jiri Denemark <jdenemar@redhat.com>
  Date:   Fri Feb 15 15:11:47 2013 +0100

    qemu: Avoid deadlock in autodestroy
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

https://bugzilla.redhat.com/show_bug.cgi?id=896092 mentions that
qemu 1.4 and earlier only accept a simple start-stop range for
the cpu=... argument of -numa.  Libvirt would attempt to use
-numa cpu=1,3 for a disjoint range, which did not work as intended.

Upstream qemu will be adding a new syntax for disjoint cpu ranges
in 1.5; but the design for that syntax is still under discussion
at the time of this patch.  So for libvirt 1.0.3, it is safest to
just reject attempts to build an invalid qemu command line; in the
future, we can add a capability bit and translate to the final
accepted design for selecting a disjoint cpu range in numa.

* src/qemu/qemu_command.c (qemuBuildNumaArgStr): Reject disjoint
ranges.

This reverts commit 383ebc46.

We decided the xml for this feature needed more thought to make sure
we are doing it the best way, in particular wrt option values that
have multiple items.

To avoid confusion about usage of this function explicitly document that
this function returns copy of the attribute string.

Users may want to specify XML entities in paths to devices. Ensure they
are parsed and used properly.

Reported by John Ferlan.

These two flags in fact are mutually exclusive. Requesting them both
doesn't make any sense regardless of hypervisor driver. Hence, we have
to make it within libvirt.c file instead of fixing it in each driver.

Add qed for dirfs pool.
Add ocfs2 for disk pool.
Add lvm2 for disk and logical pool.
Add cifs and glusterfs for netfs pool.

Note: POOL_DISK_LVM2 can not be created by "parted mklabel", but is only
returned from auto-detection on disk pools.
Signed-off-by: NPhilipp Hahn <hahn@univention.de>

Eugene Marcotte reported that if gcrypt-devel (a prereq of
gnutls-devel) is not present, then compilation fails due to
an unconditional use of <gcrypt.h>.

* src/libvirt.c (includes): Properly guard use of gcrypt.h.

This reverts commit 0bbbd42c.

The design for this feature is not complete, and may change the
name of the 'schid' attribute.  Revert requested by Viktor Mihajlovski.

This reverts commit 24aa7f8d.

The implementation to match the documentation is not complete yet,
and the final design might change the name of the 'schid' attribute.

virStrToLong(..., 8, ...) already requires the mode to be octal.
Change the relax-ng schema to check for octal as well.
Signed-off-by: NPhilipp Hahn <hahn@univention.de>

virstoragetest was failing on RHEL 5, but with no good error message:

TEST: virstoragetest
                                        0   FAIL

It turns out that qemu-img was so old, that it lacked support for
-o backing_file.  It didn't help that the test was also using
qemu-img from PATH, even after first probing for kvm-img.

* tests/virstoragetest.c (testPrepImages): Consistently use
discovered binary.  Skip instead of fail if qemu-img fails during
setup.

This fixes a potential NULL deref identified by John Ferlan
<jferlan@redhat.com> if scandir() didn't return an expected value.

The parallels storage driver declared some loop variables
inside the for(;;). This is not allowed by libvirt coding
standards
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

This change tried to fix a crash with changing CDROM media but
failed to actually do so

  commit d0172d2b
  Author: Osier Yang <jyang@redhat.com>
  Date:   Tue Feb 19 20:27:45 2013 +0800

    qemu: Remove the shared disk entry if the operation is ejecting or updating

It was still accessing disk->src, when the entire 'disk' object
has been free'd already. Even if it weren't free'd, accessing
the 'src' value of virDomainDiskDef is not allowed without
first validating disk->type is file or block. Just remove the
broken code entirely.
Signed-off-by: NDaniel P. Berrange <berrange@redhat.com>

On RHEL 5, 'make check' included failures such as:

TEST: virstoragetest
unable to create directory /virstoragedata/sub
unable to return to correct directory, refusing to clean up /virstoragedata

It turns out that with automake 1.9.x, $(abs_builddir) is not
automatically provided.  We have previously worked around this
by using `pwd` before, but because we did not do it everywhere,
we had a number of broken tests.

This patch brings RHEL 5 from 8 failed tests down to 5 (the
remaining failures may be due to bugs in the older libxml2 and
RNG schema validation available in RHEL 5, so I'm not sure if
they can be fixed in libvirt, but I'm still investigating).

* tests/Makefile.am (AM_CFLAGS): Reliably set abs_builddir.
(*_la_CFLAGS): Factor out common settings; delete when nothing
remains to be added.

On RHEL 5, I noticed this test failure message:

TEST: qemumonitorjsontest
libvirt not compiled with yajl, skippingSKIP: qemumonitorjsontest

* tests/virstoragetest.c (testPrepImages): Use simpler fputs.
* tests/qemumonitorjsontest.c (mymain): Ensure trailing newline.

VIR_ERR_NO_CONNECT already contains "no connection driver available".

This patch changes:
no connection driver available for No connection for URI hello
to:
no connection driver available for hello

Bug: https://bugzilla.redhat.com/show_bug.cgi?id=851413

Currently we call virSetDeviceUnprivSGIO with val == 0 if a block device
has an sgio attribute.  But for sgio='filtered', we know that a
kernel with no unpriv_sgio support will always behave as the user
wanted.  In this case, there is no need to call the function and
report a (bogus) error.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

testutils.c likes to print summaries after a test completes,
including if it failed.  But if the test outright exit()s,
this summary is skipped.  Enforce that we return instead of exit.

* cfg.mk (sc_prohibit_exit_in_tests): New syntax check.
* tests/commandhelper.c (main): Fix offenders.
* tests/qemumonitorjsontest.c (mymain): Likewise.
* tests/seclabeltest.c (main): Likewise.
* tests/securityselinuxlabeltest.c (mymain): Likewise.
* tests/securityselinuxtest.c (mymain): Likewise.
* tests/testutils.h (VIRT_TEST_MAIN_PRELOAD): Likewise.
* tests/testutils.c (virtTestMain): Likewise.
(virtTestCaptureProgramOutput): Use symbolic name.

Now that the segfault is solved, we can skip instead of fail
the test when yajl is not present.

* tests/qemumonitorjsontest.c (mymain): Skip if no yajl.

On a machine without yajl headers, I was seeing random segfaults
from qemumonitorjsontest (about 90% of the runs on my particular
machine).  The segfault was inside virClassIsDerivedFrom, which
points to a case of a race leading to unreferencing a stale
pointer to an object that had already been freed.  I also noticed
that if I got the segfault, I was seeing messages such as:

2013-02-22 16:12:37.504+0000: 19833: error : virNetSocketWriteWire:1361 : Cannot write data: Bad file descriptor

which is also evidence of deferencing a stale pointer.  I traced it
to a race where qemuMonitorTestIO could execute late, after the
main thread had already called qemuMonitorTestFree and called
virNetSocketClose(test->client) but not clearing it out to NULL.
Sure enough, after test->client has been closed, fd is -1, which
causes an attempt to write to the socket to fail, which in turn
triggers the error code of qemuMonitorTestIO that tries to re-close
test->client.

* tests/qemumonitortestutils.c (qemuMonitorTestIO): Don't attempt
to free client again if test already quit.