Since the functions provided by libpciaccess are not thread-safe,
when the udev-event and nodedev-init threads of libvirt call the
pci_get_strings function provided by libpaciaccess at the same
time the following can happen:
nodedev-init thread:
nodeStateInitializeEnumerate ->
udevEnumerateDevices->
udevProcessDeviceListEntry ->
udevAddOneDevice ->
udevGetDeviceDetails->
udevProcessPCI ->
udevTranslatePCIIds ->
pci_get_strings -> (libpciaccess)
find_device_name ->
populate_vendor ->
d = realloc( vend->devices, (vend->num_devices + 1), * sizeof( struct pci_device_leaf ) );
vend->num_devices++;
udev-event thread:
udevEventHandleThread ->
udevHandleOneDevice ->
udevAddOneDevice->
udevGetDeviceDetails->
udevProcessPCI ->
udevTranslatePCIIds ->
pci_get_strings -> (libpciaccess)
find_device_name ->
populate_vendor ->
d = realloc( vend->devices, (vend->num_devices + 1), * sizeof( struct pci_device_leaf ) );
vend->num_devices++;
Signed-off-by: WangJian <wangjian161@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@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>
Turns out, the way that glib implements g_steal_pointer() is not
compatible with function callbacks. And that's what my recent
patch did in virNetSocketEventFree(). Revert that part.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Notable changes:
* HAL is no longer installed on FreeBSD;
* the native version of libwsman is no longer installed in
containers intended for cross-compilation;
* Meson 0.55 rather than 0.54 is requested when installing
it from PyPI;
* GNU sed and GNU grep are installed explicitly everywhere.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
In one of my recent patches I've introduced new connection
feature VIR_DRV_FEATURE_NETWORK_UPDATE_HAS_CORRECT_ORDER.
However, I forgot to add corresponding case into a switch in
vzConnectSupportsFeature().
Signed-off-by: Michal Privoznik <mprivozn@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>
So far, it was not needed, but shortly a client will want to know
whether virNetworkUpdate() API is fixed or not. See next commits
for more info.
Side note, this driver's implementation is called only when using
sub-driver's connection, i.e. "network:///system". For any other
URI the corresponding hypervisor's driver callback is called.
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>
This function returns true if the domain has any interfaces that are
type='vdpa'.
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>
The ESX implementation of virConnectListAllDomains() follows
pretty much implementations in other drivers: it has local array
of virDomainPtr-s which (if requested by caller) is filled by
actual domains or not (if the caller is interested only in the
count of domains).
Anyway, in case of the former, the passed @domains argument is
set to the local array, which is then set to NULL to prevent it
from freeing under cleanup label. Pretty standard pattern.
Except, the local array is set to NULL always. Even if the local
array is not stolen. Fortunately, this doesn't lead to a memory
leak, because if caller is not interested in the array, none is
allocated. But it doesn't set good example and also breaks my
spatch rules.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Some SRIOV PFs don't have a netdev associated with them (the spec
apparently doesn't require it). In most cases when libvirt is dealing
with an SRIOV VF, that VF must have a PF, and the PF *must* have an
associated netdev (the only way to set the MAC address of a VF is by
sending a netlink message to the netdev of that VF's PF). But there
are times when we don't need for the PF to have a netdev; in
particular, when we're just getting the Switchdev Features for a VF,
we don't need the PF netdev - the netdev of the VF (apparently) works
just as well.
Commit 6452e2f5 (libvirt 5.1.0) *kind of* made libvirt work around PFs
with no netdevs in this case - if virNetDevGetPhysicalFunction
returned an error when setting up to retrieve Switchdev feature info,
it would ignore the error, and then check if the PF netdev name was
NULL and, if so it would reset the error object and continue on rather
than returning early with a failure. The problem is that by the time
this special handling occured, the error message about missing netdev
had already been logged, which was harmless to proper operation, but
confused the user.
Fortunately there are only 2 users of virNetDevGetPhysicalFunction, so
it is easy to redefine it's API to state that a missing netdev name is
*not* an error - in that case it will still return success, but the
caller must be prepared for the PF netdev name to be NULL. After
making this change, we can modify the two callers to behave properly
with the new semantics (for one of the callers it *is* still an error,
so the error message is moved there, but for the other it is okay to
continue), and our spurious error messages are a thing of the past.
Resolves: https://bugzilla.redhat.com/1924616
Fixes: 6452e2f5e1
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
These were the result of the conversion to RST by commit
97f21a82b2.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
The page isn't linked from anywhere and the contents is dated.
Images related to the page are also dropped.
Signed-off-by: Erik Skultety <eskultet@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
With this, incomplete XML without <frames/> for <rx/> in coalesce
won't raise error as before. It will leave the coalesce parameter
empty, thanks to passing it as a parameter and return an integer
to indicate error state - previously it returned pointer (or NULL
for both error and incomplete XML).
I also added a test case to test this functionality in the
qemuxml2xmltest.
The code went through some refactoring:
* change of a condition
* addition of a parameter
* change of order, that allowed removal of VIR_FREE
* removal of redundant labels and variables
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1535930
Signed-off-by: Kristina Hanicova <khanicov@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Commit ac87d3520a consolidated common cgroup code between the QEMU and
lxc drivers in domain_cgroup.c. In this process, in
virDomainCgroupSetupDomainBlkioParameters(), a call to
virCgroupGetBlkioWeight() went missing.
The result is that 'virsh blkiotune' is setting the blkio.weight for the
guest in the host cgroup, but not on the domain XML, because
virCgroupGetBlkioWeight() is also used to write the blkio.weight value
in the domain object.
Fix it by adding the virCgroupGetBlkioWeight() call in the
virDomainCgroupSetupDomainBlkioParameters() helper.
Fixes: ac87d3520a
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1941407
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Previously we'd assign the default checkpoint bitmap names in
virDomainCheckpointAlignDisks. In cases when the checkpoint is redefined
without a domain XML virDomainCheckpointAlignDisks is not called.
Add an explicit call to virDomainCheckpointDefAssignBitmapNames to
restore functionality.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@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>
Add a selection of tests making exapmple use of -object prior to change
to the JSON format for -object.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The test has interesting config of the memory backend object. Preserve
the 5.2 output too since it's prior to JSONification.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
qemu qapified object-add, which means that it's introspectable via
query-qmp-schema. Update the qemu-6.0 capabilities to commit
v5.2.0-3205-g92566947b3
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Similarly to the validation for blockdev-add and netdev_add, use the
qemuxml2argv test repository to drive validation of props for
object-add.
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>
The file is unused since commit e34097750a split
the test file for VXHS and NBD protocols.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The firmwareFeatures member of virDomainOSDef struct is allocated
in virDomainDefParseBootFirmwareOptions() but never freed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
The virDomainDefFree() function frees individual members of
virDomainDef struct. The function is already long enough, move
code that handles def->os member into a separate function.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
FreeBSD 12 was released in December 2018, so according to our
platform support policy we can now drop support for the previous
major release. It would be going EOL in September anyway.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Our reallocation APIs already abort on OOM and thus can only return 0.
There's no need to force callers to check the result.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Erik Skultety <eskultet@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>
