Don't report OOM error on xmlCopyNode failure

Out of memory isn't the only reason the function can fail. Add a message
stating that copying of a XML node failed.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

util: xml: Add wrapper for 'xmlNewNode'

Add a wrapper that will handle the out of memory condition by abort()
and also prevents callers from having to typecast the argument.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

util: xml: Add virXMLBufferCreate wrapper

'xmlBufferCreate' returns NULL only on allocation failure. Add a wrapper
which will call 'abort()' in such case in a centralised spot. It doesn't
make much sense to continue execution from here.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

util: virnetlink: Add wrapper for 'nlmsg_alloc_simple'

The function is used in many places and fails only on allocation
failures. Since trying to recover from allocation failure of a small
buffer by reporting error doesn't make sense add a wrapper for
'nlmsg_alloc_simple' which will 'abort()' on failure and replace all
allocations of netlink message with the new helper.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virfirewall: Remove impossible OOM error reporting

There's nothing that would set the 'err' field of virFirewallPtr to
ENOMEM so we can remove the checks.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virfirewall: virFirewallAddRuleFullV: Remove OOM check from VIR_APPEND_ELEMENT

VIR_APPEND_ELEMENT_COPY will abort the program on OOM so there's no need
to check.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virfirewall: Remove OOM checks from virFirewallStartTransaction

Neither virFirewallGroupNew nor VIR_EXPAND_N can fail.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virfirewall: Don't check OOM in ADD_ARG macro

VIR_RESIZE_N can't fail nowadays, adjust the macro.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virCloseCallbacksGetForConn: Remove OOM handling

VIR_EXPAND_N will abort so we can simplify the hash iterator.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

util: vircommand: Remove OOM handling

The OOM error handling is dead code nowadays.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virDomainDefSetMetadata: Rework memory handling

Switch to use g_autoptr for 'doc' and 'new' local variables.
Additionally report proper error when 'xmlAddChild' fails because OOM is
not the only error it can report.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

lxc_process: Remove OOM handling from logging setup

'virLogGetFilters' doesn't return failure and 'virLogGetOutputs' reports
it's own errors.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virBuildPath: Remove return value

The function can't fail nowadays, remove the return value and adjust
callers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virhostcputest: linuxCPUStatsCompareFiles: Don't check return value of virBufferContentAndReset

The buffer won't encounter OOM condition nowadays

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virCPUx86DataParse: Don't check error from x86FeatureNames

x86FeatureNames uses virBuffer and thus can't fail nowadays.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virCommandAddArgBuffer: Simplify clearing of @buf

Get the buffer contents into a temporary variable with automatic
clearing so that the error branches don't have to reset the buffer.
Additionally handle the NULL string case before assignment.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virCommandAddEnv: Make stealing of argument more obvious

The function is supposed to always consume the passed environment
variable string. Use a temp variable with autofree and g_steal_pointer
to prevent having to free it manually.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virDomainDefSetMetadata: Avoid temporary variable for string copy

Since error checking was removed when switching to g_strdup, it doesn't
make much sense to have 'tmp' around.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

util: xml: Introduce autoptr cleanup support for 'xmlNode'

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

Remove useless comments for VIR_FROM_THIS definition

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

virtpm: Fix @path handling in virTPMEmulatorInit()

This function finds "swtmp", "swtpm_setup" and "swtpm_ioctl"
binaries in $PATH and stores resolved paths in global variables
so that they can be obtainer later. Anyway, the resolved path is
marked as g_autofree and to avoid its freeing later on in the
function the variable is set to NULL manually. Well, we have
g_steal_pointer() for that.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu_tpm: Generate log file path among with storage path

When starting a guest with TPM of type='emulator' an external
process is started with it (swtpm) to emulate TPM. This external
process is passed path to a log file via --logfile. The path to
the log file is generated in qemuTPMEmulatorPrepareHost() which
works, until the daemon is restarted. The problem is that the
path is not stored in private data or anywhere inside live XML
and thus later, when qemuExtTPMStop() is called (when shutting
off the guest) the stored logpath is NULL and thus its seclabel
is not cleaned up (see virSecuritySELinuxRestoreTPMLabels()).

Fortunately, qemuExtDevicesStop() (which calls qemuExtTPMStop()
eventually) does call qemuExtDevicesInitPaths() where the log
path can be generated again.

Basically, tpm->data.emulator.storagepath is generated in
qemuExtTPMInitPaths() and its seclabels are restored properly,
and this commit move logfile onto the same level.

This means, that the log path doesn't have to be generated in
qemuExtDevicesStart() because it was already done in
qemuExtDevicesPrepareHost().

This change also renders @vmname argument of
qemuTPMEmulatorPrepareHost() unused and thus is removed.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1769196
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu_tpm: Move logfile path generation into a separate function

Strictly not needed, but the rest of paths is generated in
separate functions. Helps with code readability.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

tools: Fix dry run of libvirt_recover_xattrs.sh

The libvirt_recover_xattrs.sh script can be used to remove stale
XATTRs that were left behind by secdrivers (which should happen
only if there's an imbalance between set and restore calls).
Anyway, the script has '-n' switch which is supposed to perform
just a dry run, i.e. just to report which files have XATTRs set
without any attempt to remove them.

But, when rewriting the script a few months ago a typo was
introduced which made the script report no files even if there
were files with XATTRs.

Fixes: 5377177f80da40ee7d47601400b50835f093715a
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

Remove redundant variables/labels

In files: src/lxc/lxc_native: in lxcAddNetworkRouteDefinition(),
src/conf/networkcommon_conf: in virNetDevIPRouteCreate() and
virNetDevIPRouteParseXML()

Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

Use g_autoptr instead of virNetDevIPRouteFree if possible

In files: src/conf/domain_conf: in virDomainNetIPInfoParseXML(),
src/lxc/lxc_native: in lxcAddNetworkRouteDefinition(),
src/vz/vz_sdk: in prlsdkGetRoutes(), src/conf/networkcommon_conf:
in virNetDevIPRouteCreate()

Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

networkcommon_conf: Use g_autofree where possible

Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>

cpumap: Add support for svme-addr-check CPU feature

Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

cpumap: Add support for ibrs CPU feature

Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

cpu_map/sync_qemu_i386.py: Add mapping for svme-addr-check

This was added to qemu in commit 5447089c2b3b084b51670af36fc86ee3979e04be.

Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

cpu_map/sync_qemu_i386.py: Add mapping for ibrs

This was added to qemu in commit 623972ceae091b31331ae4a1dc94fe5cbb891937.

Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

cpu_map/sync_qemu_i386.py: Add mapping for amd-ssbd

This was added to qemu in commit 623972ceae091b31331ae4a1dc94fe5cbb891937.

Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

hyperv: Fix 32bit compilation

Example:
../src/hyperv/hyperv_driver.c:3007:54: error: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 7 has type ‘size_t’ {aka ‘unsigned int’} [-Werror=format=]
 3007 |             virReportError(VIR_ERR_INTERNAL_ERROR, _("Could not attach serial port %lu"), i);

Signed-off-by: Cole Robinson <crobinso@redhat.com>

virhostdev.c: remove missing PCI devs from hostdev manager

virHostdevReAttachPCIDevices() is called when we want to re-attach
a list of hostdevs back to the host, either on the shutdown path or
via a 'virsh detach-device' call.  This function always count on the
existence of the device in the host to work, but this can lead to
problems. For example, a SR-IOV device can be removed via an admin
"echo 0 > /sys/bus/pci/devices/<addr>/sriov_numvfs", making the kernel
fire up and eventfd_signal() to the process, asking for the process to
release the device. The result might vary depending on the device driver
and OS/arch, but two possible outcomes are:

1) the hypervisor driver will detach the device from the VM, issuing a
delete event to Libvirt. This can be observed in QEMU;

2) the 'echo 0 > ...' will hang waiting for the device to be unplugged.
This means that the VM process failed/refused to release the hostdev back
to the host, and the hostdev will be detached during VM shutdown.

Today we don't behave well for both cases. We'll fail to remove the PCI device
reference from mgr->activePCIHostdevs and mgr->inactivePCIHostdevs because
we rely on the existence of the PCI device conf file in the sysfs. Attempting
to re-utilize the same device (assuming it is now present back in the host)
can result in an error like this:

$ ./run tools/virsh start vm1-sriov --console
error: Failed to start domain vm1-sriov
error: Requested operation is not valid: PCI device 0000:01:00.2 is in use by driver QEMU, domain vm1-sriov

For (1), a VM destroy/start cycle is needed to re-use the VF in the guest.
For (2), the effect is more nefarious, requiring a Libvirtd daemon restart
to use the VF again in any guest.

We can make it a bit better by checking, during virHostdevReAttachPCIDevices(),
if there is any missing PCI device that will be left behind in activePCIHostdevs
and inactivePCIHostdevs lists. Remove any missing device found from both lists,
unconditionally, matching the current state of the host. This change affects
the code path in (1) (processDeviceDeletedEvent into qemuDomainRemoveDevice, all
the way back to qemuHostdevReAttachPCIDevices) and also in (b) (qemuProcessStop
into qemuHostdevReAttachDomainDevices).

NB: Although this patch enables the possibility of 'outside Libvirt' SR-IOV
hotunplug of PCI devices, if the hypervisor and the PCI driver copes with it,
our goal is to mitigate what it is still considered a user oopsie. For all
supported purposes, the admin must remove the SR-IOV VFs from all running domains
before removing the VFs from the host.

Resolves:  https://gitlab.com/libvirt/libvirt/-/issues/72
Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virpci.c: use virPCIDeviceAddressPtr in virPCIDeviceListDel()

This change will allow us to remove PCI devices from a list
without the need of a PCI Device object, which will be need
in the next patch.

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virpci.c: use virPCIDeviceAddressPtr in virPCIDeviceListSteal()

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virpci.c: use virPCIDeviceAddressPtr in virPCIDeviceListFind()

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virpci.c: use virPCIDeviceAddressPtr in virPCIDeviceListFindIndex()

We're going to need a way to remove a PCI Device from a list without having
a valid virPCIDevicePtr, because the device is missing from the host. This
means that virPCIDevicesListDel() must operate with a PCI Device address
instead.

Turns out that virPCIDevicesListDel() and its related functions only use
the virPCIDeviceAddressPtr of the virPCIDevicePtr, so this change is
simple to do and will not cause hassle in all other callers. Let's
start adapting virPCIDeviceListFindIndex() and crawl our way up to
virPCIDevicesListDel().

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

qemu_cgroup.c: skip absent PCI devices in qemuTeardownHostdevCgroup()

There is no need to bother with cgroup tearing down for absent
PCI devices, given that their entries in the sysfs are already
gone.

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virhostdev.c: add virHostdevIsPCIDevice() helper

Add a helper to quickly determine if a hostdev is a PCI device,
instead of doing a tedious 'if' check with hostdev mode and
subsys type.

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virsh-domain.c: use g_auto* in cmdDetachDevice()

Use g_auto* pointers to avoid the need of a cleanup label. The
type of the pointer 'virDomainPtr dom' was changed to its alias
'virshDomainPtr' to allow the use of g_autoptr().

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

dac, selinux: skip setting/restoring label for absent PCI devices

If the underlying PCI device of a hostdev does not exist in the
host (e.g. a SR-IOV VF that was removed while the domain was
running), skip security label handling for it.

This will avoid errors that happens during qemuProcessStop() time,
where a VF that was being used by the domain is not present anymore.
The restore label functions of both DAC and SELinux drivers will
trigger errors in virPCIDeviceNew().

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virhostdev.c: virHostdevGetPCIHostDevice() now reports missing device

Gitlab issue #72 [1] reports that removing SR-IOVs VFs before
removing the devices from the running domains can have strange
consequences. QEMU might be able to hotunplug the device inside the
guest, but Libvirt will not be aware of that, and then the guest is
now inconsistent with the domain definition.

There's also the possibility of the VFs removal not succeeding
while the domain is running but then, as soon as the domain
is shutdown, all the VFs are removed. Libvirt can't handle
the removal of the PCI devices while trying to reattach the
hostdevs, and the Libvirt daemon can be left in an inconsistent
state (see [2]).

This patch starts to address the issue related in Gitlab #72, most
notably the issue described in [2]. When shutting down a domain
with SR-IOV hostdevs that got missing, virHostdevReAttachPCIDevices()
is failing the whole process and failing to reattach all the
PCI devices, including the ones that aren't related to the VFs that
went missing. Let's make it more resilient with host changes by
changing virHostdevGetPCIHostDevice() to return an exclusive error
code '-2' for this case. virHostdevGetPCIHostDeviceList() can then
tell when virHostdevGetPCIHostDevice() failed to find the PCI
device of a hostdev and continue to make the list of PCI devices.

virHostdevReAttachPCIDevices() will now be able to proceed reattaching
all other valid PCI devices, at least. The 'ghost hostdevs' will be
handled later on.

[1] https://gitlab.com/libvirt/libvirt/-/issues/72
[2] https://gitlab.com/libvirt/libvirt/-/issues/72#note_459032148

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

virpci: introduce virPCIDeviceExists()

We're going to add logic to handle the case where a previously
existing PCI device does not longer exist in the host.

The logic was copied from virPCIDeviceNew(), which verifies if a
PCI device exists in the host, returning NULL and throwing an
error if it doesn't. The NULL is used for other errors as well
(product/vendor id read errors, dev id overflow), meaning that we
can't re-use virPCIDeviceNew() for the purpose of detecting
if the device exists.

Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

Use g_steal_pointer where possible

Via coccinelle (not the handbag!)
spatches used:
@ rule1 @
identifier a, b;
symbol NULL;
@@

- b = a;
  ... when != a
- a = NULL;
+ b = g_steal_pointer(&a);

@@

- *b = a;
  ... when != a
- a = NULL;
+ *b = g_steal_pointer(&a);

Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>

qemuBackupJobTerminate: Don't calculate backup job stats if VM isn't active

If the VM isn't active calculating the job stats doesn't make sense.
Additionally this prevents a crash of libvirtd if qemu terminates while
libvirt wasn't running:

Thread 28 "init-backup-tes" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffb9310640 (LWP 3201116)]
qemuDomainJobInfoUpdateTime (jobInfo=0x0) at ../../../libvirt/src/qemu/qemu_domainjob.c:275
275     if (!jobInfo->started)
(gdb) bt
 #0  qemuDomainJobInfoUpdateTime (jobInfo=0x0) at ../../../libvirt/src/qemu/qemu_domainjob.c:275
 #1  0x00007fffcba1a12d in qemuBackupJobTerminate (vm=0x7fff9c1bc840, jobstatus=QEMU_DOMAIN_JOB_STATUS_CANCELED) at ../../../libvirt/src/qemu/qemu_backup.c:563
 #2  0x00007fffcbaefcae in qemuProcessStop
    (driver=0x7fff9c144ff0, vm=0x7fff9c1bc840, reason=VIR_DOMAIN_SHUTOFF_DAEMON, asyncJob=QEMU_ASYNC_JOB_NONE, flags=<optimized out>)
    at ../../../libvirt/src/qemu/qemu_process.c:7812
 #3  0x00007fffcbaf2a10 in qemuProcessReconnect (opaque=<optimized out>) at ../../../libvirt/src/qemu/qemu_process.c:8578
 #4  0x00007ffff7c46bb5 in virThreadHelper (data=<optimized out>) at ../../../libvirt/src/util/virthread.c:233
 #5  0x00007ffff6e453f9 in start_thread () at /lib64/libpthread.so.0
 #6  0x00007ffff766fb53 in clone () at /lib64/libc.so.6

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemuBackupJobTerminate: Move cleanup of temp files earlier

Upcoming patch will remove unnecessary actions if the VM crashed. The
cleanup needs to be performed always, thus needs to be moved earlier.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virStorageVolDefFormat: Don't format empty <source>

If there are no source extents the volume XML has an empty <source>
element. Remove it if there's nothing in it by using
virXMLFormatElement.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virStorageVolDefFormat: Extract formatting of source extents

Move the extent formatting code into
virStorageVolDefFormatSourceExtents.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

storageBackendProbeTarget: Don't fail if backing store can't be parsed

When the backing store of the image can't be parsed
virStorageSourceNewFromBacking returns -1. storageBackendProbeTarget
then also fails which makes the pool refresh fail or even the storage
pool becomes inactive after (re)start of libvirtd.

In situations when we can't access the backing store via network we
just report the backing store string, thus we can do the same thing for
unparsable backing store to prevent the pool from going offline.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

storageBackendProbeTarget: Check return value of virStorageSourceNewFromBacking

Commit bc3a78f61a2aaac3 errorneously removed the return value check from
virStorageSourceNewFromBacking. In cases when we e.g. can't parse the
backing store string this leads to a crash:

 #0  virStorageSourceGetActualType (def=0x0) at ../../../libvirt/src/conf/storage_source_conf.c:1014
 #1  0x00007ffff7cee4f9 in virStorageSourceIsLocalStorage (src=<optimized out>) at ../../../libvirt/src/conf/storage_source_conf.c:1026
 #2  0x00007ffff455c97c in storageBackendProbeTarget (encryption=0x7fff9c122ce8, target=0x7fff9c122c68) at ../../../libvirt/src/storage/storage_util.c:3443
 #3  virStorageBackendRefreshVolTargetUpdate (vol=0x7fff9c122c30) at ../../../libvirt/src/storage/storage_util.c:3519
 #4  0x00007ffff455cdc0 in virStorageBackendRefreshLocal (pool=0x7fff9c010ea0) at ../../../libvirt/src/storage/storage_util.c:3593
 #5  0x00007ffff454f0a1 in storagePoolRefreshImpl
    (backend=backend@entry=0x7ffff4711180 <virStorageBackendDirectory>, obj=obj@entry=0x7fff9c010ea0, stateFile=stateFile@entry=0x7fff9c111a90 "/var/run/libvirt/storage/tmp.xml") at ../../../libvirt/src/storage/storage_driver.c:103
 #6  0x00007ffff4550ea5 in storagePoolUpdateStateCallback (obj=0x7fff9c010ea0, opaque=<optimized out>) at ../../../libvirt/src/storage/storage_driver.c:165
 #7  0x00007ffff7cefef4 in virStoragePoolObjListForEachCb (payload=<optimized out>, name=<optimized out>, opaque=0x7fffc8a489c0)
    at ../../../libvirt/src/conf/virstorageobj.c:435
 #8  0x00007ffff7c03195 in virHashForEachSafe
    (table=<optimized out>, iter=iter@entry=0x7ffff7cefec0 <virStoragePoolObjListForEachCb>, opaque=opaque@entry=0x7fffc8a489c0)
    at ../../../libvirt/src/util/virhash.c:414
 #9  0x00007ffff7cf0520 in virStoragePoolObjListForEach
    (pools=<optimized out>, iter=iter@entry=0x7ffff4550e10 <storagePoolUpdateStateCallback>, opaque=opaque@entry=0x0)
    at ../../../libvirt/src/conf/virstorageobj.c:468
 #10 0x00007ffff454f43a in storagePoolUpdateAllState () at ../../../libvirt/src/storage/storage_driver.c:184
 #11 storageStateInitialize (privileged=<optimized out>, root=<optimized out>, callback=<optimized out>, opaque=<optimized out>)
    at ../../../libvirt/src/storage/storage_driver.c:315
 #12 0x00007ffff7e10c04 in virStateInitialize
    (opaque=0x555555621820, callback=0x55555557b1d0 <daemonInhibitCallback>, root=0x0, mandatory=<optimized out>, privileged=true)
    at ../../../libvirt/src/libvirt.c:656
 #13 virStateInitialize
    (privileged=<optimized out>, mandatory=mandatory@entry=false, root=root@entry=0x0, callback=callback@entry=0x55555557b1d0 <daemonInhibitCallback>, opaque=opaque@entry=0x555555621820) at ../../../libvirt/src/libvirt.c:638
 #14 0x000055555557b230 in daemonRunStateInit (opaque=0x555555621820) at ../../../libvirt/src/remote/remote_daemon.c:605
 #15 0x00007ffff7c46bb5 in virThreadHelper (data=<optimized out>) at ../../../libvirt/src/util/virthread.c:233
 #16 0x00007ffff6e453f9 in start_thread () at /lib64/libpthread.so.0
 #17 0x00007ffff766fb53 in clone () at /lib64/libc.so

An invalid image can be easily created by:

$ qemu-img create -f qcow2 -F qcow2 -b 'json:{' -u img.qcow2 10M

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

Post-release version bump to 7.2.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

Release of libvirt-7.1.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

Translated using Weblate (Finnish)

Currently translated at 15.5% (1638 of 10545 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/fi/

Co-authored-by: Ricky Tigg <ricky.tigg@gmail.com>
Signed-off-by: Ricky Tigg <ricky.tigg@gmail.com>

qemu_monitor: Document qemuMonitorUnregister()

The most important bit is that the caller is expected to pass
locked monitor.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>

qemu_domainjob: Make copy of owner API

Using the job owner API name directly works fine as long as it is a
static string or the owner's thread is still running. However, this is
not always the case. For example, when the owner API name is filled in a
job when we're reconnecting to existing domains after daemon restart,
the dynamically allocated owner name will disappear with the
reconnecting thread. Any follow up usage of the pointer will read random
memory.

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>

docs: fix bad cut/paste in <teaming> example

When the parser and docs were enhanced to support a <teaming> element
in a generic <hostdev>, the example XML for formatdomain.rst was
cut/pasted from the example for <interface type='hostdev'>. In my
haste I neglected to remove the <mac address='blah'/> element (which
is unused/ignored for generic <hostdev> and change the closing tag
from </interface> to </hostdev>

https://bugzilla.redhat.com/1927984

Fixes: db64acfbda59ad22b671580fda13968c60bb8c1a
Reported-by: Yalan Zhang <yalzhang@redhat.com>
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: allow migration of generic <hostdev> with <teaming>

Commit 010ed0856b and commit db64acfbda introduced the ability to use
the <teaming> element in a generic <hostdev> (previously it could only
be used with <interface type='hostdev'>). However, the patch omitted
one crucial detail - along with parsing the <teaming> element in
<hostdev>, and adding the necessary info to the qemu commandline, we
also need to modify qemuMigrationSrcIsAllowedHostdev() to allow
migration when the generic <hostdev> has a <teaming> element.

https://bugzilla.redhat.com/1927984

Fixes: 010ed0856bb06f439e6fdf44e4f529f53441c398
Reported-by: Yalan Zhang <yalzhang@redhat.com>
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

NEWS: Mention some bug fixes for the 7.1.0 dev cycle

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

qemu: Add missing lock in qemuProcessHandleMonitorEOF

qemuMonitorUnregister will be called in multiple threads (e.g. threads
in rpc worker pool and the vm event thread).  In some cases, it isn't
protected by the monitor lock, which may lead to call g_source_unref
more than one time and a use-after-free problem eventually.

Add the missing lock in qemuProcessHandleMonitorEOF (which is the only
position missing lock of monitor I found).

Suggested-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Peng Liang <liangpeng10@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

Translated using Weblate (Ukrainian)

Currently translated at 100.0% (10545 of 10545 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/uk/

Co-authored-by: Yuri Chornoivan <yurchor@ukr.net>
Signed-off-by: Yuri Chornoivan <yurchor@ukr.net>

Update translation files

Updated by "Update PO files to match POT (msgmerge)" hook in Weblate.

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/

Co-authored-by: Weblate <noreply@weblate.org>
Signed-off-by: Fedora Weblate Translation <i18n@lists.fedoraproject.org>

libxl: Add lock process indicator to libxlDomainObjPrivate object

The libvirt libxl driver has no access to FDs associated with VM disks.
The disks are opened by libxl.so and any related FDs are not exposed to
applications. The prevents using virtlockd's auto-release feature to
release locks when the FD is closed. Acquiring and releasing locks is
explicitly handled by the libxl driver.

The current logic is structured such that locks are acquired in
libxlDomainStart and released in libxlDomainCleanup. This works well
except for migration, where the locks must be released on the source
host before the domain can be started on the destination host, but the
domain cannot be cleaned up until the migration confirmation stage.
When libxlDomainCleanup if finally called in the confirm stage, locks
are again released resulting in confusing errors from virtlockd and
libvirtd

virtlockd[8095]: resource busy: Lockspace resource 'xxxxxx' is not locked
libvirtd[8050]: resource busy: Lockspace resource 'xxxxxx' is not locked
libvirtd[8050]: Unable to release lease on testvm

The error is also encountered in some error cases, e.g. when
libxlDomainStart fails before acquiring locks and libxlDomainCleanup
is still used for cleanup.

In lieu of a mechanism to check if a lock has been acquired, this patch
takes an easy approach to fixing the unnecessary lock releases by adding
an indicator to the libxlDomainPrivate object that can be set when the
lock is acquired and cleared when the lock is released. libxlDomainCleanup
can then skip releasing the lock in cases where it was previously released
or never acquired in the first place.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

NEWS: mention cgroups on hosts with systemd bug fix

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>

libxl: Fix domain shutdown

Commit fa30ee04a2 caused a regression in normal domain shutown.
Initiating a shutdown from within the domain or via 'virsh shutdown'
does cause the guest OS running in the domain to shutdown, but libvirt
never reaps the domain so it is always shown in a running state until
calling 'virsh destroy'.

The shutdown thread is also an internal user of the driver shutdown
machinery and eventually calls libxlDomainDestroyInternal where
the ignoreDeathEvent inhibitor is set, but running in a thread
introduces the possibility of racing with the death event from
libxl. This can be prevented by setting ignoreDeathEvent before
running the shutdown thread.

An additional improvement is to handle the destroy event synchronously
instead of spawning a thread. The time consuming aspects of destroying
a domain have been completed when the destroy event is delivered.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

NEWS: mention support for vhost-user-blk

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>

NEWS: Mention snapshot quiesce rollback bugfix

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

docs: use proper cpu quota value in our documentation

Commit <d505b8af58912ae1e1a211fabc9995b19bd40828> changed the cpu quota
value that reflects what kernel allows but did not update our
documentation.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

domain_validate: use defines for cpu period and quota limits

Commints <bc760f4d7c4f964fadcb2a73e126b0053e7a9b06> and
<98a09ca48ed4fc011abf2aa290e02ce1b8f1bb5f> fixed the code to use
defines instead of magic numbers but missed this place.

Following commit <ed1ba69f5a8132f8c1e73d2a1f142d70de0b564a> changed
the cpu quota limit to reflect what kernel actually allows so using
the defines fixes XML validations as well.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

po: Refresh potfile for v7.1.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

Translated using Weblate (Chinese (Simplified) (zh_CN))

Currently translated at 56.0% (5859 of 10451 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/zh_CN/

Co-authored-by: Pany <geekpany@gmail.com>
Signed-off-by: Pany <geekpany@gmail.com>

qemu*xml2*test: Cache capabilities between tests

Invoking the XML parser every time is quite expensive. Since we have a
deep copy function for 'virQEMUCapsPtr' object, we can cache the parsed
results lazily.

This brings significant speedup to qemuxml2argvtest:

real 0m2.234s
user 0m2.140s
sys 0m0.089s

vs.

real 0m1.161s
user 0m1.087s
sys 0m0.072s

qemuxml2xmltest benefits too:

real 0m0.879s
user 0m0.801s
sys 0m0.071s

vs.

real 0m0.466s
user 0m0.424s
sys 0m0.040s

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jonathon Jongsma <jjongsma@redhat.com>

testQemuInfoSetArgs: Use curly braces in else section

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jonathon Jongsma <jjongsma@redhat.com>

testCompareXMLToArgvValidateSchema: Improve and fix helper for testing everything

The schema validator has a comment which allows checking all xml2argv
input files for schema validity by forcing the latest schema onto files
which don't have any schema. Fix it so that it works properly with the
caching introduced in previous commit.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jonathon Jongsma <jjongsma@redhat.com>

qemuxml2argvtest: Cache QAPI schema between tests

It's quite wasteful to reparse the QAPI schema for each _CAPS_ test.

Add a simple cache filled lazily by encountered schemas.

The time saving on my box is quite significant:

real 0m3.318s
user 0m3.203s
sys 0m0.107s

vs

real 0m2.223s
user 0m2.134s
sys 0m0.084s

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jonathon Jongsma <jjongsma@redhat.com>

virJSONParserInsertValue: Take double pointer for @value

The function calls virJSONValueObjectAppend/virJSONValueArrayAppend, so
by taking a double pointer we can drop the pointer clearing from
callers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueNewNumber: Take ownership of passed string

Avoid pointless copies of temporary strings when constructing number
JSON objects.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONParserHandle*: Refactor memory cleanup and drop NULL checks

virJSONValueNew* won't return error nowadays so NULL checks are not
necessary. The memory can be cleared via g_autoptr.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemuAgentSetVCPUsCommand: Refactor cleanup

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemuMonitorJSONTransactionAdd: Refactor cleanup

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueArrayAppend: Clear pointer when taking ownership of passed value

The parent array takes ownership of the inserted value once all checks
pass. Don't make the callers second-guess when that happens and modify
the function to take a double pointer so that it can be cleared once the
ownership is taken.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

testQEMUSchemaValidateObjectMergeVariantMember: Fix theoretical leak

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virMACMapHashDumper: Refactor array addition

Use automatic memory freeing and don't check return value of
virJSONValueNewString as it can't fail.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemuAgentMakeStringsArray: Refactor cleanup

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueObjectAppend: Clear pointer when taking ownership of passed value

The parent object takes ownership of the inserted value once all checks
pass. Don't make the callers second-guess when that happens and modify
the function to take a double pointer so that it can be cleared once the
ownership is taken.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueObjectAddVArgs: Use autofree for the temporary bitmap

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueNewArrayFromBitmap: Refactor cleanup

Use g_autoptr for the JSON value objects and remove the cleanup label
and inline freeing of objects.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValue(Array|Object)Append*: Simplify handling of appended object

Use g_autofree for the pointer of the added object and remove the NULL
checks for values returned by virJSONValueNew* (except
virJSONValueNewNumberDouble) since they can't fail nowadays.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueCopy: Don't use virJSONValue(Object|Array)Append

We know the exact number of keys or array members for the copied objects
so we can pre-allocate the arrays rather than inserting into them in a
loop incurring realloc copy penalty.

Also virJSONValueCopy now can't fail since all of the functions
allocating the different cases use just g_new/g_strdup internally so we
can remove the NULL checks from the recursive calls.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virJSONValueObjectInsert: Clear @value on successful insertion

The function takes ownership of @value on success so the proper
semantics will be to clear out the @value pointer. Convert @value to a
double pointer to do this.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemuAgentMakeCommand: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virLockSpacePreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virNetServerPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virNetServerPreExecRestart: Drop error reporting from virJSONValueObjectAppend* calls

The functions report errors already and the error can nowadays only
happen on programmer errors (if the passed virJSONValue isn't an
object), which won't happen. Remove the reporting.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virNetServerClientPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virNetServerServicePreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virNetDaemonPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virLogHandlerPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virLogDaemonPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

virLockDaemonPreExecRestart: Refactor memory cleanup

Switch to using the 'g_auto*' helpers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>