This makes it possible to enable stable NIC device names in most modern
Linux distros.
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This property is exposed by QEMU on any PCI device, but we have to pick
some specific device(s) to probe it against. We expect that at least one
of the virtio devices will be present, so probe against them.
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The compiler can more easily optimize a switch, and more importantly can
also warn when new address types are added which are not handled.
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Enable '-compat' if requested in qemu.conf and supported by qemu to
instruct qemu to crash when a deprecated command is used and stop
returning deprecated fields.
This setting is meant for libvirt developers and such.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Similar to the qemu.conf knob 'deprecation_behavior' add a per-VM knob
in the QEMU namespace:
<qemu:deprecation behavior='...'/>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
New QEMU supports a harsh, but hard to ignore way to notify that the
QMP user used a deprecated command. This is useful e.g. for developers
to see that something needs to be fixed.
This patch introduces a qemu.conf option to enable the setting in cases
when qemu supports it so that developers and continiuous integration
efforts are notified about use of deprecated fields before it's too
late.
The option is deliberately stored as string and not validated to prevent
failures when downgrading qemu or libvirt versions. While we don't
support this, the knob isn't meant for public consumption anyways.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
The capability is asserted if qemu supports the -compat
deprecated-input= and deprecated-output= settings to control what should
happen if deprecated fields are used in QMP.
This will be used for a developer/tester-oriented setting which will
aid us in catching use of deprecated settings sooner.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
This is available in QEMU with "ide-hd" and "scsi-hd" device
types. It was originally mistakenly added to the "scsi-block"
device type too, but later removed. This doesn't affect libvirt
since we restrict usage to device=disk.
When this property is not set then QEMU's default behaviour
is to not report any rotation rate information, which
causes most guest OS to assume rotational storage.
https://bugzilla.redhat.com/show_bug.cgi?id=1498955
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This API talks to QEMU and changes its internal state. Therefore,
it should acquire QEMU_JOB_MODIFY instead of QEMU_JOB_QUERY.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Recently, a few commits back I've switched bunch of code to
g_steal_pointer() using coccinelle. Problem was that the semantic
patch used was slightly off:
@@
expression a, b;
@@
+ b = g_steal_pointer(&a);
- b = a;
... when != a
- a = NULL;
Problem is that, "... when != a" is supposed to jump over those
lines, which don't contain expression a. My idea was to replace
the following pattern too:
ptrX = ptrY;
if (something(ptrZ) < 0) goto error;
ptrY = NULL;
But what I missed is that the following pattern is also matched
and replaced:
ptrX = ptrY;
if (something(ptrX) < 0) goto error;
ptrY = NULL;
This is not necessarily correct - as demonstrated by our hotplug
code. The real problem is ambiguous memory ownership transfer
(functions which add device to domain def take ownership only on
success), but to not tackle the real issue let's revert those
parts.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
The order in which virNetworkUpdate() accepts @section and
@command arguments is not the same as in which it passes them
onto networkUpdate() callback. Until recently, it did not really
matter, because calling the API on client side meant arguments
were encoded in reversed order (compared to the public API), but
then on the server it was fixed again - because the server
decoded RPC (still swapped), called public API (still swapped)
and in turn called the network driver callback (with reversing
the order - so magically fixing the order).
Long story short, if the public API is called even number of
times those swaps cancel each other out. The problem is when the
API is called an odd numbed of times - which happens with split
daemons and the right URI. There's one call in the client (e.g.
virsh net-update), the other in a hypervisor daemon (say
virtqemud) which ends up calling the API in the virnetworkd.
The fix is obvious - fix the order in which arguments are passed
to the callback.
But, to maintain compatibility with older, yet unfixed, daemons
new connection feature is introduced. The feature is detected
just before calling the callback and allows client to pass
arguments in correct order (talking to fixed daemon) or in
reversed order (talking to older daemon).
Unfortunately, older client talking to newer daemon can't be
fixed. Let's hope that it's less frequent scenario.
Fixes: 574b9bc66b
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1870552
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Just like VFIO devices, vDPA devices may need to have all guest memory
pages locked/pinned in order to operate properly. In the case of VFIO
devices (including mdev and NVME, which also use VFIO) libvirt
automatically increases the locked memory limit when one of those
devices is present. This patch modifies that code to also increase the
limit if there are any vDPA devices present.
Resolves: https://bugzilla.redhat.com/1939776
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
This function is a specialized version of
qemuDomainGetMemLockLimitBytes() for PPC64. Simplifying it in the same
manner as the previous patch has the nice side effect of accounting
for the possibility of an mdev device
(I don't know if mdev devices are supported on PPC, but even if not
then a) the additional check for mdev devices gained by using
qemuDomainNeedsVFIO() in place of open coding will be an effective
NOP, and b) if mdev devices are supported on PPC64 in the future, this
function will be prepared for it).
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
This function goes through a loop checking if each hostdev is a VFIO
or mdev device, and then later it calls virDomainDefHasNVMEDisk(). The
function qemuDomainNeedsVFIO() does exactly the same thing, so let's
just call that instead.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Generated by the following spatch:
@@
expression a, b;
@@
+ b = g_steal_pointer(&a);
- b = a;
... when != a
- a = NULL;
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
When a checkpoint is redefined without providing the domain XML, we
might end up with a definition where the per-disk bitmap name is not
set. Trying to delete such checkpoint would lead to a crash.
Refuse such deletion.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1941600
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Base the detection on the presence of the 'secret' qom-type entry, which
isn't conditionally compiled in qemu.
All caps-based test now switch to using JSON for -object.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Skip the lossy conversion to legacy commandline arguments by using the
JSON props directly when -object is QAPIfied. This avoids issues with
conversion of bitmaps and also allows validation of the generated JSON
against the QMP schema in the tests.
Since the new approach is triggered by a qemu capability the code
from 'virQEMUBuildObjectCommandlineFromJSON' in util/virqemu.c was moved
to 'qemuBuildObjectCommandlineFromJSON' in qemu/qemu_command.c which has
the virQEMUCaps type.
Some functions needed to be modified to propagate qemuCaps.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Set 'objectAddNoWrap' when the capability is present.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
There's just one caller left. Since qemuBuildMemoryBackendProps is too
complex to be modified for now, just move the adding of 'id' and 'qom'
type directly into the function.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Construct the JSON object which is used for object-add without the
'props' wrapper and add the wrapper only in the monitor code.
This simplifies the JSON->commandline generator in the first place and
also prepares for upcoming qemu where 'props' will be removed.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Starting from qemu-6.0 the parameters of -object/object-add are formally
described by the QAPI schema. Additionally this changes the nesting of
the properties as the 'props' nested object will be flattened to the
parent.
We'll need to detect whether qemu switched to this new approach to
generate the objects with proper nesting and also allow testing.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
virQEMUCapsGet checks for qemuCaps itself, no need to do it explicitly.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
In case an async job spans multiple APIs (e.g., incoming migration) the
API that started the job is recorded as the asyncOwnerAPI even though it
is no longer running and the owner thread is updated properly to the one
currently handling the job. Let's also update asyncOwnerAPI to make it
more obvious which is the current (or the most recent) API involved in
the job.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Attempting to set the memlock limit might fail if we're running
in a containerized environment where CAP_SYS_RESOURCE is not
available, and if the limit is already high enough there's no
point in trying to raise it anyway.
https://bugzilla.redhat.com/show_bug.cgi?id=1916346
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Store the current memory locking limit and the desired one
separately, which will help with later changes.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Now that we've implemented a fallback for the function that
obtains the information from /proc, there is no reason we would
get a failure unless there's something seriously wrong with the
environment we're running in, in which case we're better off
reporting the issue to the user rather than pretending
everything is fine.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
When the backup job is terminated normally the security label is
restored by the blockjob finishing handler.
If the VM dies or is destroyed that wouldn't happen as the blockjob
handler wouldn't be called.
Restore the security label on disk store where we remember that the job
was running at the point when 'qemuBackupJobTerminate' was called.
Not resetting the security label means that we also leak the xattr
attributes remembering the label which prevents any further use of the
file, which is a problem for block devices.
This also requires that the call to 'qemuBackupJobTerminate' from
'qemuProcessStop' happens only after 'vm->pid' was reset as otherwise
the security subdrivers attempt to enter the process namespace which
fails if the process isn't running any more.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1939082
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
qemuBackupBegin can take a full backup of the disks (excluding any
operations with bitmaps) without the need to wait for the
blockdev-reopen support in qemu.
Add a check that no checkpoint creation is required and the disk backup
mode isn't VIR_DOMAIN_BACKUP_DISK_BACKUP_MODE_INCREMENTAL.
Call to virDomainBackupAlignDisks is moved earlier as it initializes the
disk backup mode if not present in user config.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Upcoming commit will enable full backup support (incremental part
requires blockdev-reopen, which won't happen in qemu for at least
another release).
Add a capability that the 'blockdev-backup' job is supported by qemu
capped, but limited to when qemu supports QEMU_CAPS_BLOCKDEV.
We can also use it in the expression to enable
QEMU_CAPS_INCREMENTAL_BACKUP since it's a pre-requisite too.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Extend dirtyrate statistics for domGetStats to display the information
of a domain's memory dirty rate produced by domainStartDirtyRateCalc.
Signed-off-by: Hao Wang <wanghao232@huawei.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The g_idle_add function adds a callback to the primary GMainContext.
To workaround the GSource unref bugs, we need to add our callbacks
to the GMainContext that is associated with the GSource being
unref'd. Thus code using the per-VM virEventThread must use its
private GMainContext.
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
When connecting to the monitor, a timeout is calculated that is
bigger the more memory guest has (because QEMU has to allocate
and possibly zero out the memory and what not, empirically
deducted). However, when computing the timeout the @total_memory
mmember is accessed directly even though
virDomainDefGetMemoryTotal() should have been used.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
When switching to g_autoptr this was incorrectly changed from
'continue;' into 'return -1;' resulting into an error when user tries
to set vcpu_quota of running VM:
error: An error occurred, but the cause is unknown
Fixes: e4a8bbfaf2
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
In short, virXXXPtr type is going away. With big bang. And to
help us rewrite the code with a sed script, it's better if each
variable is declared on its own line.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The qemu shim spawns a separate thread in which the event loop is
ran. The virEventRunDefaultImpl() call is wrapped in a while()
loop, just like it should. There are few lines of code around
which try to ensure that domain is destroyed (when quitting) and
that the last round of event loop is ran after the
virDomainDestroy() call. Only after that the loop is quit from
and the thread quits.
However, if domain creation fails, there is no @dom to call
destroy over, the @quit flag is never set and while() never
exits. Set the flag regardless of @dom pointer.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1920337
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
The commandline generator for 'iothread' objects has a private
implementation of the properties. Convert it to JSON so that it can be
later validated.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
While the 'sev0' sev-guest object will never be hotplugged, but we want
to generate it through JSON so that we'll be able to validate all
parameters of '-object' against the QAPI schema once 'object-add' is
qapified in qemu.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
While the 'masterKey0' secret object will never be hotplugged we want to
generate it through JSON so that we'll be able to validate all
parameters of '-object' against the QAPI schema once 'object-add' is
qapified in qemu.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This problem is reproducible only with secret driver. When
starting a domain via virt-qemu-run and both secret and
(nonexistent) root directory specified this is what happens:
1) virt-qemu-run opens "secret:///embed?root=$rootdir"
connection, which results in the secret driver initialization
(done in secretStateInitialize()). During this process, the
driver creates its own configDir (derived from $rootdir)
including those parents which don't exists yet. This is all
done with the mode S_IRWXU and thus results in the $rootdir
being created with very restrictive mode (specifically, +x is
missing for group and others).
2) now, virt-qemu-run opens "qemu:///embed?root=$rootdir" and
calls virDomainCreateXML(). This results in the master-key.aes
being written somewhere under the $rootdir and telling qemu
where to find it.
But because the secret driver created $rootdir with too
restrictive mode, qemu can't access the file (even though it
knows the full path) and fails to start.
It looks like the best solution is to pre-create the root
directory before opening any connection (letting any driver
initialize itself) and set its mode to something less
restrictive.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1859873
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Commit cb29e4e801 didn't take into account that the VM can be inactive
when it's destroyed. This means that the job would remain active also
when the VM became inactive.
To fix this properly:
1) Remove the bogus VM liveness check and early return
(reverts the aforementioned commit)
2) Conditionalize the stats assignment only when the stats object is
present
(properly fix the crash when VM dies when reconnecting)
3) end the asyncjob only when it was already set
(prevent corruption of priv->jobs_queued)
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1937598
Fixes: cb29e4e801
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
'qemuBackupJobTerminate' needs the API flags to see whether
VIR_DOMAIN_BACKUP_BEGIN_REUSE_EXTERNAL. Unfortunately when called via
qemuProcessReconnect()->qemuProcessStop() early (e.g. if the qemu
process died while we were reconnecting) the job is cleared temporarily
so that other APIs can be called. This would mean that we couldn't clean
up the files in some cases.
Save the 'apiFlags' inside the backup object and set it from the
'qemuDomainJobObj' 'apiFlags' member when reconnecting to a VM.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Switch to using the 'g_auto*' helpers.
Signed-off-by: Yi Li <yili@winhong.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
The -audiodev argument is replacing the QEMU_AUDIO_DRV env variable (and
its relations).
Sadly we still have to use the SDL_AUDIODRIVER env variable because that
wasn't mapped into QAPI schema.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The -audiodev arg is a new way to configure audio devices in QEMU to
replace the QEMU_AUDIO_DRV env variable. This arg is not visible in
the "query-command-line-options" output since it is entirely QAPI
driven, not QemuOpts. It also isn't in "query-qmp-schema" though
since there's no QMP command that uses the Audiodev type yet.
So probe for the existance of this feature by looking for the
-vnc "audiodev" property. This won't let us determine which
precise audio backends QEMU has been built with, but for now
that's no worse than with env variables today.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Currently the QEMU driver secretly sets the QEMU_AUDIO_DRV env variable
- VNC - set to "none", unless passthrough of host env variable is set
- SPICE - always set to "spice"
- SDL - always passthrough host env
- No graphics - set to "none", unless passthrough of host env variable is set
The setting of the QEMU_AUDIO_DRV env variable is done in the code which
configures graphics.
If no <audio> element is present, we now auto-populate <audio> elements
to reflect this historical default config. This avoids need to set audio
env when processing graphics.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Currently the QEMU driver secretly sets the QEMU_AUDIO_DRV env variable
depending on how <graphics> are configured.
This introduces support for configuring audio backends from the <audio>
elements in the XML config.
The existing default behaviour is now only used if no <audio> element is
present.
All except the 'jack' audio driver are supported via QEMU's old env
variable config.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The check for ICH6 || ICH9 is repeated in many places in the code. The
new virDomainSoundModelSupportsCodecs() method provides a helper to
standardize this check.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This patch adds delay time (steal time inside guest) to libvirt
domain per-vcpu stats. Delay time is an important performance metric.
It is a consequence of the overloaded CPU. Knowledge of the delay
time of a virtual machine helps to understand if it is affected and
estimate the impact.
As a result, it is possible to react exactly when needed and
rebalance the load between hosts. This is used by cloud providers
to provide quality of service, especially when the CPU is
oversubscribed.
It's more convenient to work with this metric in a context of a
libvirt domain. Any monitoring software may use this information.
Signed-off-by: Aleksei Zakharov <zaharov@selectel.ru>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Failure of 'qemuMigrationSetDBusVMState' would jump to 'exit_monitor'
but the function isn't called inside of the monitor context.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
That's more consistent with our usual naming convention.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The current code is written under the assumption that, for all
limits except the core size, asking for the limit to be set to
zero is a no-op, and so the operation is performed
unconditionally.
While this is the behavior we want for the QEMU driver, the
virCommand and virProcess facilities are generic, and should not
implement this kind of policy: asking for a limit to be set to
zero should result in that limit being set to zero every single
time.
Add some checks in the QEMU driver, effectively moving the
policy where it belongs.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
qemuProcessLaunch() is the correct place to set process limits,
and in fact is where we were dealing with almost all of them,
but the memory locking limit was handled in
qemuBuildCommandLine() instead for some reason.
The code is rewritten so that the desired limit is calculated
and applied in separated steps, which will help with further
changes, but this doesn't alter the behavior.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Doing this now will make the next changes nicer.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This allows the VNC client user to perform a shutdown, reboot and reset
of the VM from the host side.
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The function is used to automatically feed a buffer into a pipe which
can be used by the command to read contents of the buffer.
Rather than passing in a pipe, let's create the pipe inside
virCommandSetSendBuffer and directly associate the reader end with the
command. This way the ownership of both ends of the pipe will end up
with the virCommand right away reducing the need of cleanup in callers.
The returned value then can be used just to format the appropriate
arguments without worrying about cleanup or failure.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The function is constructing an error message from a prefix and the
contents of the qemu log file. Marking just two string modifiers as
translatable is pointless and will certainly confuse translators.
Remove the marking and add a comment which bypasses the
sc_libvirt_unmarked_diagnostics check.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Now that error message formatting doesn't use fixed size buffers we can
drop the math for calculating the maximum chunk of log to report in the
error message and use a round number. This also makes it obvious that
the chosen number is arbitrary.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Similarly to the crash workaround:
commit 0db4743645
Author: Daniel P. Berrangé <berrange@redhat.com>
Date: Tue Jul 28 16:52:47 2020 +0100
util: avoid crash due to race in glib event loop code
we need to do this in the other event loop as crash in that one was also
reported:
https://bugzilla.redhat.com/show_bug.cgi?id=1931331
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Generated using the following spatch:
@@
expression path;
@@
- virFileMakePath(path)
+ g_mkdir_with_parents(path, 0777)
However, 14 occurrences were not replaced, e.g. in
virHostdevManagerNew(). I don't really understand why.
Fixed by hand afterwards.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
These functions are identical. Made using this spatch:
@@
expression path, mode;
@@
- virFileMakePathWithMode(path, mode)
+ g_mkdir_with_parents(path, mode)
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Now that we enforce the cpu.shares range kernel will no longer silently
change the value that libvirt configures so there is no need to read
the value back to get the actual configuration.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
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>
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>
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>
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>
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>
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>
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>
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>
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: 010ed0856b
Reported-by: Yalan Zhang <yalzhang@redhat.com>
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
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>
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>
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>
For incremental backup we need QEMU_CAPS_BLOCKDEV,
QEMU_CAPS_BLOCKDEV_REOPEN, QEMU_CAPS_MIGRATION_PARAM_BLOCK_BITMAP_MAPPING.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Preserve block dirty bitmaps after migration with
QEMU_MONITOR_MIGRATE_NON_SHARED_(DISK|INC).
This patch implements functions which offer the bitmaps to the
destination, check for eligibility on destination and then configure
source for the migration.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
In case when the block migration job required temporary bitmaps for
merging the appropriate checkpoints we need to clean them up when
cancelling the job. On success we don't need to do that though as the
bitmaps are just temporary thus are not written to disk.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Add status XML infrastructure for storing a list of block dirty bitmaps
which are temporarily used when migrating a VM with
VIR_MIGRATE_NON_SHARED_DISK for cleanup after a libvirtd restart during
migration.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
'qemuMigrationCookieBlockDirtyBitmapsMatchDisks' maps the bitmaps from
the migration cookie to actual disk objects definition pointers.
'qemuMigrationCookieBlockDirtyBitmapsToParams' converts the bitmap
definitions from the migration cookie into parameters for the
'block-bitmap-mapping' migration parameter.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
In cases where we are copying the storage we need to ensure that also
bitmaps are copied properly. This patch adds migration cookie XML
infrastructure which will allow the migration sides reach consensus on
which bitmaps to migrate.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Add the migration capability flag and the propagation of the
corresponding mapping configuration. The mapping will be produced from
the bitmaps on disk depending on both sides of the migration and the
necessity to perform merges.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
There's no need in the cleanup steps to invoke a transaction to delete a
single bitmap.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
The non-transaction wrapper is useful for code paths which want to
delete individual bitmaps or for cleanup after a failed job where we
want to attempt to delete every bitmap individually to prevent a failure
from cleaning up the rest.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Use the new format when pre-creating the image for the user. Users
wishing to use the legacy format can always provide their own images or
use shared storage.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Such images don't support stuff like dirty bitmaps. Note that the
synthetic test for detecting bitmaps is used as an example to prevent
adding additional test cases.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
The capability represents qemu's ability to setup mappings for migrating
block dirty bitmaps and is based on presence of the 'transform' property
of the 'block-bitmap-mapping' property of 'migrate-set-parameters' QMP
command.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This patch partially reverts commit 5cde9dee where the qemuExtDevicesStop()
was moved to a location before the QEMU process is stopped. It may be
alright to tear down some devices before QEMU is stopped, but it doesn't work
for the external TPM (swtpm) which assumes that QEMU sends it a signal to stop
it before libvirt may try to clean it up. So this patch moves the
virFileDeleteTree() calls after the call to qemuExtDevicesStop() so that the
pid file of virtiofsd is not deleted before that call.
Afftected libvirt versions are 6.10 and 7.0.
Fixes: 5cde9dee8c
Cc: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Signed-off-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
A VM defined similar to:
...
<features><kvm><hint-dedicated state='on'/></kvm></features>
<cpu mode="host-model"/>
...
is currently invalid, as hint-dedicated is only allowed if cpu mode
is host-passthrough or maximum. This restriction is unnecessary, see
https://bugzilla.redhat.com/show_bug.cgi?id=1857671
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Stay true to the name of the function and clear the pointer
after freeing it.
This also silences a bogus Coverity report about a double
free in qemuMonitorGetCPUInfo where qemuMonitorCPUInfoClear
is called right after allocating a new qemuMonitorCPUInfo
to fill out the non-zero defaults.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
The comment and the caller assume virQEMUSaveDataNew only steals
domXML on success, but it is copied even on failure.
Also remove the misleading g_steal_pointer call on a local variable.
Reported by coverity.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Commit 76f4788932 made qemuNodeDeviceDetachFlags() unusable due to an
'if then else if' chain that will always results in a 'return -1',
regardless of 'driverName' input.
Found by Coverity.
Fixes: 76f4788932
Reported-by: John Ferlan <jferlan@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Just return when alias is null and Remove the 'ret' variable.
Signed-off-by: Yi Li <yili@winhong.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Ján Tomko <jtomko@redhat.com>
libxlNodeDeviceDetachFlags() and qemuNodeDeviceDetachFlags() are mostly
equal, aside from how the virHostdevmanager pointer is retrieved and
the PCI stub driver used.
Now that the PCI stub driver verification is done early in both functions,
we can use the virDomainDriverNodeDeviceDetachFlags() helper to reduce
code duplication between them. 'driverName' is checked inside the helper
to set the appropriate stub driver.
The helper is named with the 'Flags' suffix, even when the helper itself
isn't receiving the flags from the callers, to be compliant with the
ACL function virNodeDeviceDetachFlagsEnsureACL() that is being called
inside it and was called from the original functions. Renaming the helper
would implicate in renaming REMOTE_PROC_NODE_DEVICE_DETACH_FLAGS, and all the
related structs inside remote_protocol.x, to be compliant with the ACL
rules.
This is not being checked at this moment, but we'll fix check-aclrules.py to
verify all the helpers that calls ACL functions in domain_driver.c shortly.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The validation of 'driverName' does not depend on any other state and can be
done right on the start of the function. We can fail earlier while avoiding
a cleanup jump.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
libxlNodeDeviceReAttach() and qemuNodeDeviceReAttach() are mostly equal,
differing only how the virHostdevManager pointer is retrieved.
Put the common code into virDomainDriverNodeDeviceReAttach() to reduce
code duplication.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
libxlNodeDeviceReset() and qemuNodeDeviceReset() are mostly equal,
differing only how the virHostdevManager pointer is retrieved.
Put the common code into virDomainDriverNodeDeviceReset() to reduce
code duplication.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The preferred syntax for boolean options is to set the value "on" or
"off". QEMU 7.1.0 will deprecate the short format we currently use.
The long format has been supported with -vnc since the change to use
QemuOpts in 2.2.0, so we check based on the new capability flag.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This was introduced in QEMU 2.2.0, and is visible by -vnc appearing in
the "query-command-line-options" data.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
When virDomainGetFSInfo() is called over a QEMU/KVM domain it
results into calling of 'guest-get-fsinfo' guest agent command to
which it replies with info on guest (mounted) filesystems. When
filling return structure we also try to do basic lookup and
translate guest agent provided disk address into disk target (as
seen in domain XML). This can of course fail - guest can have
variety of disks not recorded in domain XML (iSCSI, scsi_debug,
NFS to name a few). If that's the case, a debug message is logged
and no disk target is added into the return structure.
However, due to the way our code is written the caller is led to
believe that the target was added into the structure. This may
lead to a situation where the array of disk targets (strings)
contains NULL. But our RPC structure says the array contains only
non-NULL strings. This results in somewhat 'cryptic' (at least to
users) error message:
error: Unable to get filesystem information
error: Unable to encode message payload
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1919783
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
After previous commit, the freeing of @info_ret inside of
virDomainFSInfoFormat() looks like this:
for () {
if (info_ret)
virDomainFSInfoFree(info_ret[i]);
}
It is needless to compare @info_ret against NULL in each
iteration. We can switch the order and do the comparison first
followed by the loop.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
When qemuDomainGetFSInfo() is called it calls
qemuDomainGetFSInfoAgent() which executes 'guest-get-fsinfo'
guest agent command, parses returned JSON and returns an array of
qemuAgentFSInfo structures (well, pointers to those structs).
Then it grabs a domain job and tries to do some matching of guest
returned info against domain definition. This matching is done in
virDomainFSInfoFormat() which also frees the array of
qemuAgentFSInfo structures allocated earlier.
But this is not just. If acquiring the domain job fails (or
domain activeness check executed right after that fails) then
virDomainFSInfoFormat() is not called, leaking the array of
structs.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
As the very first thing, this function checks whether the number
of items inside @agentinfo array is not negative. This is
redundant as the only caller - qemuDomainGetFSInfo() already
checked for that and would not even call this function if that
was the case.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The preferred syntax for boolean options is to set the value "on" or
"off". QEMU 7.1.0 will deprecate the short format we currently use.
The long format has been supported with -spice since at least 1.5.3,
so we don't need to check for it.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The preferred syntax for boolean options is to set the value "on" or
"off". QEMU 7.1.0 will deprecate the short format we currently use.
The long format has been supported with -chardev since at least 1.5.3,
so we don't need to check for it.
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The -2 value is misleading because if 'qemuAgentFSFreeze' fails it
doesn't necessarily mean that the command was sent to the agent.
Since callers don't care about the -2 value specifically, remove it.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
If we didn't freeze any filesystems we should not even attempt thawing
them. Additionally 'guest-fsfreeze-freeze' fails if the filesystems are
already frozen, where thawing them may break users data integrity if
they used VIR_DOMAIN_SNAPSHOT_CREATE_QUIESCE accidentally after an
explicit virDomainFSFreeze and the next snapshot without that flag would
be taken with already thawed filesystems.
This effectively reverts 7c736bab06 .
Libvirt nowadays checks whether the guest agent is connected and pings
it before issuing an command so it's very unlikely that we'd end up in a
situation where qemuSnapshotCreateActiveExternal froze filesystems and
didn't thaw them.
Additionally we now discourage the use of
VIR_DOMAIN_SNAPSHOT_CREATE_QUIESCE since users have better control if
they freeze the FS themselves.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Use g_autofree for 'dom_xml' to free it on some of the (unlikely) code
paths jumping to cleanup prior to the deallocation which is done right
after it's not needed any more since it's a big string.
Noticed when running under valgrind:
==2204780== 8,192 bytes in 1 blocks are definitely lost in loss record 2,539 of 2,551
==2204780== at 0x483BCE8: realloc (vg_replace_malloc.c:834)
==2204780== by 0x4D890DF: g_realloc (in /usr/lib64/libglib-2.0.so.0.6600.4)
==2204780== by 0x4DA3AF0: g_string_append_vprintf (in /usr/lib64/libglib-2.0.so.0.6600.4)
==2204780== by 0x4917293: virBufferAsprintf (virbuffer.c:307)
==2204780== by 0x49B0B75: virDomainChrDefFormat (domain_conf.c:26109)
==2204780== by 0x49E25EF: virDomainDefFormatInternalSetRootName (domain_conf.c:28956)
==2204780== by 0x15F81D24: qemuDomainDefFormatBufInternal (qemu_domain.c:6204)
==2204780== by 0x15F8270D: qemuDomainDefFormatXMLInternal (qemu_domain.c:6229)
==2204780== by 0x15F8270D: qemuDomainDefFormatLive (qemu_domain.c:6279)
==2204780== by 0x15FD8100: qemuMigrationSrcBeginPhase (qemu_migration.c:2395)
==2204780== by 0x15FE0F0D: qemuMigrationSrcPerformPeer2Peer3 (qemu_migration.c:4640)
==2204780== by 0x15FE0F0D: qemuMigrationSrcPerformPeer2Peer (qemu_migration.c:5093)
==2204780== by 0x15FE0F0D: qemuMigrationSrcPerformJob (qemu_migration.c:5168)
==2204780== by 0x15FE280E: qemuMigrationSrcPerform (qemu_migration.c:5372)
==2204780== by 0x15F9BA3D: qemuDomainMigratePerform3Params (qemu_driver.c:11841)
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
In one of my previous commits I've made an attempt to restore the
noqueue qdisc on a TAP corresponding to domain's <interface/> if
QoS is cleared out. The commit consisted of two almost identical
hunks. In both the pointer is dereferenced. But in one of them,
the pointer to new bandwidth can't be NULL while in the other it
can leading to a crash.
Fixes: d53b092353
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1919619
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The root directory can be provided by user (or a temporary one is
generated) and is always formatted into connection URI for both
secret driver and QEMU driver, like this:
qemu:///embed?root=$root
But if it so happens that there is an URI unfriendly character in
root directory or path to it (say a space) then invalid URI is
formatted which results in unexpected results. We can trust
g_dir_make_tmp() to generate valid URI but we can't trust user.
Escape user provided root directory. Always.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1920400
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
These messages are only valid while the domain is running.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
To make it easier to split out the parsing/formatting of the <teaming>
element into separate functions (so we can more easily add the
<teaming> element to <hostdev>, change its virDomainNetDef so that it
points to a virDomainNetTeamingInfo rather than containing one.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
TPM devices with model='tpm-tis' are only valid with x86 and aarch64
virt machines. Add a check to qemuValidateDomainDeviceDefTPM() to
ensure VIR_DOMAIN_TPM_MODEL_TIS is only used with these architectures.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Starting a VM with swtpm device fails with qemu-system-aarch64.
E.g. with TPM device config
<tpm model='tpm-tis'>
<backend type='emulator' version='2.0'/>
</tpm>
QEMU reports the following error
error: internal error: process exited while connecting to monitor:
2021-02-07T05:15:35.378927Z qemu-system-aarch64: -device
tpm-tis,tpmdev=tpm-tpm0,id=tpm0: 'tpm-tis' is not a valid device model name
Indeed the TPM device name is 'tpm-tis-device' [1][2] for aarch64,
versus the shorter 'tpm-tis' for x86. The devices are the same from
a functional POV, i.e. they both emulate a TPM device conforming to
the TIS specification. Account for the unfortunate name difference
when building the TPM device option in qemuBuildTPMDevStr(). Also
include a test case for 'tpm-tis-device'.
[1] https://qemu.readthedocs.io/en/latest/specs/tpm.html
[2] c294ac327c
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Our implementation was inspired by glib anyways. The difference is only
the order of arguments.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Our implementation was heavily inspired by the glib version so it's a
drop-in replacement.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The glib implementation doesn't tolerate NULL but in most cases we check
before anyways. The rest of the callers adds a NULL check.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Some callers don't need to know the actual lenght of the list but only
care whether the required element is present or the list is non-empty.
Don't calculate the list length in those cases.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The glib variant doesn't accept NULL list, but there's just one caller
where it wasn't checked explicitly, thus there's no need for our own
wrapper.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Use automatic memory freeing and remove the 'cleanup' label. Also make
it a bit more obvious that nothing happens if the 'old' list wasn't
present.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The conversion removes the use of virStringListAdd/virStringListRemove
which try to add dynamic properties to a string list which is really
inefficient.
Storing the dbus VMState ids in a GSList is pretty straightforward and
the slightly increased complexity of the code will be paid back by
removing the string list helpers later.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
'virHashGetItems' already returns the number of entries which will be
considered for addition to the list so we can allocate it to the upper
bound upfront rather than growing it in a loop. This avoids the
quadratic complexity of 'virStringListAdd'.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
'virStringListAdd' calculates the string list length on every invocation
so constructing a string list using it results in O(n^2) complexity.
Use a GSList which has cheap insertion and iteration and doesn't need
failure handling.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Some code paths return -1 directly while others jump to 'cleanup' which
cleans the list of mounts. Since qemuDomainGetPreservedMounts now
returns a NULL-terminated list, convert devMountsPath to g_auto(GStrv)
and remove the cleanup altoghether.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
'i' is used in both outer and inner loop. Since 'devMountsPath' is now a
NULL-terminated list, we can use a GStrv to iterate it;
Additionally rewrite the conditional of adding to the 'unlinkPaths'
array so that it's more clear what's happening.
Fixes: 5c86fbb72d
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Refactor the handling of internals so that NULL-terminated lists are
always returned.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
In commit 88957116c9 I've adapted
libvirt to QEMU's deprecation of -mem-path and -mem-prealloc and
switched to memory-backend-* even for system memory. My claim was
that that's what QEMU does under the hood anyway. And indeed it
was: see QEMU commit 900c0ba373aada4c13d47d95330aa72ec4067ba5 and
look at function create_default_memdev().
However, then commit d96c4d5f193e0e45beec80a6277728b32875bddb was
merged into QEMU. While it was fixing a bug, it also changed the
create_default_memdev() function in which it started turning off
use of canonical path (by setting
"x-use-canonical-path-for-ramblock-id" attribute to false). This
wasn't documented until QEMU commit
8db0b20415c129cf5e577a593a4a0372d90b7cc9. The path affects
migration - the same path has to be used on the source and on the
destination. Therefore, if there is old guest started with '-m X'
it has "pc.ram" block which doesn't use canonical path and thus
when migrating to newer QEMU which uses memory-backend-* we have
to turn off the canonical path explicitly. Otherwise,
"/objects/pc.ram" path would be expected by QEMU which doesn't
match the source.
Ideally, we would need to set it only for some machine types
(4.0 and older) because newer machine types already do what we
are doing. However, we treat machine types as opaque strings and
therefore we don't want to parse nor inspect their versions. But
then again, newer machine types already do what we are doing in
this commit, so when old machine types are deprecated and removed
we can remove our hack and forget it ever happened.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1912201
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
This capability tracks whether memory-backend-file has
"x-use-canonical-path-for-ramblock-id" attribute. Introduced into
QEMU by commit fa0cb34d2210cc749b9a70db99bb41c56ad20831. As of
QEMU commit 8db0b20415c129cf5e577a593a4a0372d90b7cc9 the property
is considered stable by qemu despite the 'x-' prefix to preserve
compatibility with released qemu versions.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
The 'conflict' key in a virt_daemon_unit dictionary is not used when
generating systemd service and socket files. The comment associated
with the key claims the default is 'true', and a few build files
needlessly set it to 'true' when defining their virt_daemon_unit.
Remove the 'conflict' key and its use in the affect build files.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The "max" model can be treated the same way as "host" model in general.
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This is a special CPU model similar to "-cpu host", so won't use our
normal CPU model detection logic.
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
For hardware virtualization this is functionally identical to the
existing host-passthrough mode so the same caveats apply.
For emulated guest this exposes the maximum featureset supported by
the emulator. Note that despite being emulated this is not guaranteed
to be migration safe, especially if different emulator software versions
are used on each host.
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
A <machine/> element can have "deprecated" attribute that
corresponds to 'deprecated' member of _virQEMUCapsMachineType
struct. But the member is of boolean type. Therefore, the string
returned by virXMLPropString() must be freed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
If parsing "maxCpus" attribute of <machine/> element fails an
error is printed but the corresponding string is not freed. While
it is very unlikely to happen (parsed XML is not user provided
and we are the ones generating it), it is possible. Instead of
freeing the variable in the error path explicitly, let's declare
it as g_autofree. And while I'm at it, let's bring it into the
loop where it's used.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
A few commits back I've introduced new 'virtio-pmem' <memory/>
device. Since it's virtio it goes onto PCI bus. Therefore, on
hotplug new PCI address is generated (or provided one is
reserved). However, if hotplug fails (for whatever reason) the
address needs to be released. This is different to 'dimm' type of
address because for that type we don't keep a map of used slots
rather generate one on each address assign request. The map is
then thrown away. But for PCI addresses we keep internal state
and thus has to keep it updated. Therefore, this new
qemuDomainReleaseMemoryDeviceSlot() function is NOP for those
models which use 'dimm' address type ('dimm' and 'nvdimm').
While I'm at it, let's release the address in case of hot unplug.
Not that is supported (any such attempt fails with the following
error:
"virtio based memory devices cannot be unplugged"
But if QEMU ever implements hot unplug then we don't have to
remember to fix our code.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Get rid of the 'need_release' variable. The code can be rewritten
so that it is not needed.
Signed-off-by: Yi Li <yili@winhong.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
These functions are all only called as a part of qemuFirmwareFree(),
which frees the qemuFirmware object before return, so we can be sure
none of the pointers is referenced after freeing (and thus there is no
need to clear any of them).
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
These functions all cooperate to free memory pointed to by a single
object that contains (doesn't *point to*, but actually contains)
several sub-objects. They were written to send copies of these
sub-objects to subordinate functions, rather than just sending
pointers to the sub-objects.
Let's change these functions to just send pointers to the objects
they're cleaning out rather than all the wasteful and pointless
copying.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Several functions had the names virFirmware[something]Free(), but they
aren't taking a pointer to some object and freeing it. Instead, they
are making a copy of the content of an entire object, then Freeing the
objects pointed to by that content.
As a first step in a too-complicated cleanup just to eliminate a few
occurrences of VIR_FREE(), this patch renames those functions to more
accurately reflect what they do - they Free the *Content* of their
arguments.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Commit <318d807a0bd3372b634d1952b559c5c627ccfa5b> added a fix to skip
most of the block stat code to not log error message for missing storage
sources but forgot to increase the recordnr counter.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The correct backend type is 'vc', same as in qemuBuildChrChardevStr()
where we generate qemu command line.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Implements QEMU support for vhost-user-blk together with live
hotplug/unplug.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Make the function reusable by other vhost-user based devices.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
QEMU has the ability to mark machine types as deprecated. This should be
exposed to management applications in the capabilities.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
QEMU has the ability to mark CPUs as deprecated. This should be exposed
to management applications in the domain capabilities.
This attribute is only set when the model is actually deprecated.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
In this instance attempting to be correct is really pointless since the
secret is formatted into another string which is not erased securely and
then put on the commandline.
Keep the secure handling for correctness.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Phase out use of VIR_DISPOSE_N from the qemu driver. Use memset in the
appropriate cases.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
When virRandomBytes fails we don't get any random bytes and even if we
did they don't have to be treated as secret as they weren't used in any
way.
Add a temporary variable with automatic freeing for the secret buffer
and assign it only on success.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Pass the parameter clock rt to qemu to ensure that the
virtual machine is not synchronized with the host time
Signed-off-by: gongwei <gongwei@smartx.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This was found by clang-tidy's "readability-misleading-indentation"
check.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Back in commit 2c71d3826, which appeared in libvirt-1.2.3 in April
2014, the location used to store saved MAC addresses and vlan tags of
SRIOV VFs was changed from /var/run/libvirt/qemu to
/var/run/libvirt/hostdevmgr. For backward compatibility the code was
made to continue looking in the old location for the files when it
didn't find them in the new location.
It's now been 6 years, and even if there was somebody still running
libvirt-1.2.3 on their system, that system would now be out of support
for libvirt, so there would be no way for them to upgrade to a new
libvirt that no longer looks in "oldStateDir" for the files. So
let's no longer look in "oldStateDir" for the files!
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The current virPCIDeviceNew() signature, receiving 4 uints in sequence
(domain, bus, slot, function), is not neat.
We already have a way to represent a PCI address in virPCIDeviceAddress
that is used in the code. Aside from the test files, most of
virPCIDeviceNew() callers have access to a virPCIDeviceAddress reference,
but then we need to retrieve the 4 required uints (addr.domain, addr.bus,
addr.slot, addr.function) to satisfy virPCIDeviceNew(). The result is
that we have extra verbosity/boilerplate to retrieve an information that
is already available in virPCIDeviceAddress.
A better way is presented by virNVMEDeviceNew(), where the caller just
supplies a virPCIDeviceAddress pointer and the function handles the
details internally.
This patch changes virPCIDeviceNew() to receive a virPCIDeviceAddress
pointer instead of 4 uints.
Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Instead of receiving 4 uints in order and write domain/bus/slot/function,
receive a virPCIDeviceAddressPtr instead and write into it.
This change will allow us to simplify the API for virPCIDeviceNew()
in the next patch.
Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
libxlNodeDeviceGetPCIInfo() and qemuNodeDeviceGetPCIInfo() are equal.
Let's move the logic to a new virDomainDriverNodeDeviceGetPCIInfo()
info to be used by libxl_driver.c and qemu_driver.c.
Reviewed-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Wire up the QEMU command line for this option.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Tested-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Add virtio related options iommu, ats and packed as driver element attributes
to vsock devices. Ex:
<vsock model='virtio'>
<cid auto='no' address='3'/>
<driver iommu='on'/>
</vsock>
Signed-off-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The qemuDomainObjFromDomain() API must be paired with
the virDomainObjEndAPI API. The qemuDomainAuthorizedSSHKeysGet
method simply did 'return -1' leaking a reference and lock
in two paths.
The qemuDomainAuthorizedSSHKeysSet method marked the object
as an autoptr while also have some code paths that will call
virDomainObjEndAPI. As a result the object will be released
but not unlocked in error paths.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
If QEMU driver fails to initialize for whatever reason (it can be
as trivial as a typo on qemu.conf), the control jumps to error
label in qemuStateInitialize() where qemuStateCleanup() is called
which frees the driver. But the daemon then asks drivers to
prepare for shutdown, which in case of QEMU driver is implemented
in qemuStateShutdownPrepare(). In here, the driver is
dereferenced but since it was freed earlier, the pointer is NULL
which leads to instant crash.
Solution is simple - just check if qemu_driver is not NULL. But
doing so only in qemuStateShutdownPrepare() would push the
problem down to virStateShutdownWait(), well
qemuStateShutdownWait(). Therefore, duplicate the trick there
too.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1895359#c14
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
All callers of this function called virStorageFileParseChainIndex
before. Internalize the logic of that function to prevent multiple calls
and passing around unnecessary temporary variables.
This is achieved by calling virStorageFileParseBackingStoreStr and using
it to fill the values internally.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Rename the function to virStorageSourceFetchRelativeBackingPath and
return relative paths only. The function is only used to restore the
relative relationship between images so there's no need for it to be
universal.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
dtrace invokes the C compiler, so when cross-building we need
to make sure that $CC is set in the environment and that it
points to the cross-compiler rather than the native one.
Until https://github.com/mesonbuild/meson/issues/266
is addressed, the workaround is to call dtrace via env(1).
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=980334
Signed-off-by: Helmut Grohne <helmut@subdivi.de>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The code handles XML bits and internal definition and should be
in conf directory.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Same as virStorageFileBackend, it doesn't belong into util directory.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Up until now we had a runtime code and XML related code in the same
source file inside util directory.
This patch takes the runtime part and extracts it into the new
storage_file directory.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
This code is not directly relevant to virStorageSource so move it to
separate file.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
