We only set up host for VIR_DOMAIN_TPM_TYPE_EMULATOR and thus
similarly, we should do cleanup for the same type. This also
fixes a crasher, in which qemuTPMEmulatorCleanupHost() accesses
tpm->data.emulator.storagepath which is NULL for
VIR_DOMAIN_TPM_TYPE_EXTERNAL.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2168762
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
I initially had the passt process being started in an identical
fashion to the slirp-helper - libvirt was daemonizing the new process
and recording its pid in a pidfile. The problem with this is that,
since it is daemonized immediately, any startup error in passt happens
after the daemonization, and thus isn't seen by libvirt - libvirt
believes that the process has started successfully and continues on
its merry way. The result was that sometimes a guest would be started,
but there would be no passt process for qemu to use for network
traffic.
Instead, we should be starting passt in the same manner we start
dnsmasq - we just exec it as normal (along with a request that passt
create the pidfile, which is just another option on the passt
commandline) and wait for the child process to exit; passt then has a
chance to parse its commandline and complete all the setup prior to
daemonizing itself; if it encounters an error and exits with a non-0
code, libvirt will see the code and know about the failure. We can
then grab the output from stderr, log that so the "user" has some idea
of what went wrong, and then fail the guest startup.
Signed-off-by: Laine Stump <laine@redhat.com>
Commit 5ef2582646eb98 added emitting of even when refreshign disk state,
where it wanted to avoid sending the event if disk state didn't change.
This was achieved by using 'continue' in the loop filling the
information. Unfortunately this skips extraction of whether the device
has a tray which is propagated into internal structures, which in turn
broke cdrom media change as the code thought there's no tray for the
device.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2166411
Fixes: 5ef2582646eb98af208ce37355f82bdef39931fa
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Kristina Hanicova <khanicov@redhat.com>
The virURIFormat() function either returns a string, or aborts
(on OOM). There's no way this function can return NULL (as of
v7.2.0-rc1~277). Therefore, it doesn't make sense to check its
retval against NULL.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
In one of recent commits (v9.0.0-rc1~106) I've made our QEMU
namespace code umount the original /dev. One of the reasons was
enhanced security, because previously we just mounted a tmpfs
over the original /dev. Thus a malicious QEMU could just
umount("/dev") and it would get to the original /dev with all
nodes.
Now, on some systems this introduced a regression:
failed to umount devfs on /dev: Device or resource busy
But how this could be? We've moved all file systems mounted under
/dev to a temporary location. Or have we? As it turns out, not
quite. If there are two file systems mounted on the same target,
e.g. like this:
mount -t tmpfs tmpfs /dev/shm/ && mount -t tmpfs tmpfs /dev/shm/
then only the top most (i.e. the last one) is moved. See
qemuDomainUnshareNamespace() for more info.
Now, we could enhance our code to deal with these "doubled" mount
points. Or, since it is the top most file system that is
accessible anyways (and this one is preserved), we can
umount("/dev") in a recursive fashion.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2167302
Fixes: 379c0ce4bfed8733dfbde557c359eecc5474ce38
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>
When going through debug log of a domain startup process, one can
meet the following line:
debug : qemuProcessLaunch:7668 : Building mount namespace
But this is in fact wrong. Firstly, domain namespaces are just
enabled in domain's privateData. Secondly, the debug message says
nothing about actual state of namespace - whether it was enabled
or not.
Therefore, move the debug printing into
qemuProcessEnableDomainNamespaces() and tweak it so that the
actual value is reflected.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>
Similar to other error paths in qemuDomainUnshareNamespace(), jump to
the cleanup label on umount error instead of directly returning -1.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
When starting a guest, helper processes are started first. But
they need a bit of special handling. Just consider a regular cold
boot and an incoming migration. For instance, in case of swtpm
with its state on a shared volume, we want to set label on the
state for the cold boot case, but don't want to touch the label
in case of incoming migration (because the source very
specifically did not restore it either).
Until now, these two cases were differentiated by testing
@incoming against NULL. And while that makes sense for other
aspects of domain startup, for external devices we need a bit
more, because a restore from a save file is also 'incoming
migration'.
Now, there is a difference between regular migration and restore
from a save file. In the former case we do not want to set
seclabels in the save state. BUT, in the latter case we do need
to set them, because the code that saves the machine restored
seclabels.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2161557
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
When stopping swtpm we can restore the label either on just the
swtpm's domain specific logfile (/var/log/swtpm/libvirt/qemu/...),
or on the logfile and the state too (/var/lib/libvirt/swtpm/...).
The deciding factor is whether the guest is stopped because of
outgoing migration OR the state is on a shared filesystem.
But this is not correct condition, because for instance saving the
guest into a file (virsh save) is also an outgoing migration.
Alternatively, when the swtpm state is stored on a shared
filesystem, but the guest is destroyed (virsh destroy), i.e.
stopped because of different reason than migration, we want to
restore the seclabels.
The correct condition is: skip restoring the state on outgoing
migration AND shared filesystem.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2161557
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
When cleaning up host in qemuProcessStop(), our external helper
processes (e.g. swtpm) want to know whether the domain is being
migrated out or not (so that they restore seclabels on a device
state that's on a shared storage).
This fact is reflected in the @outgoingMigration variable which
is set to true if asyncJob is anything but
VIR_ASYNC_JOB_MIGRATION_IN. Well, we have a specific job for
outgoing migration (VIR_ASYNC_JOB_MIGRATION_OUT) and thus we
should check for that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Automatically free 'cpuinfo' and remove the cleanup label and ret
variable.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Replace virJSONValueObjectGet + virJSONValueIsArray by the single API
which returns only an array.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Replace instances of virJSONValueObjectGet + virJSONValueIsArray by
virJSONValueObjectGetArray.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Simplify construction of a single provider by using
virJSONValueObjectAdd and restructuring the code.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
If a domain is configured to create a macvtap/macvlan but the
target link already exists, startup fails (as expected) with:
error: error creating macvtap interface test@eth0 (52:54:00:d9:0b:db): File exists
Okay, we could make that error message better, but that's not the
point. Since this error originated while generating cmd line
(the caller is qemuProcessStart(), transitively), the cleanup
after failed start is performed (qemuProcessStop()). Here,
virNetDevMacVLanDeleteWithVPortProfile() is called which removes
the macvtap interface we did not create (as it made us fail in
the first place).
Therefore, we need to track which macvtap/macvlan interface was
created successfully and remove only those.
You'll notice that only qemuProcessStop() has the new check. For
the (failed) hotplug case (qemuDomainAttachNetDevice()) this
function is already in place (the @iface_connected variable), or
not needed (qemuDomainRemoveNetDevice() - we're removing an
interface that was already attached to QEMU).
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2166235
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Every single caller of the
virNetDevMacVLanDeleteWithVPortProfile() function is calling it
wrapped inside of ignore_value() macro. This is because the
function is annotated as G_GNUC_WARN_UNUSED_RESULT. This makes no
sense. Drop the annotation and the macro envelope.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The hotplug code paths need to be able to pass the FDs to the monitor to
ensure that hotplug works.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
To ensure that we can hot-unplug the disk including the associated fdset
we need to store the fdset ID in the status XML.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Rollback of FD sets passed to qemu is also needed after possible restart
of libvirtd when we need to serialize the data into status XML. For this
purpose we need to access the fdset ID once it was passed to qemu and
potentially re-create a 'qemuFDPass' struct in passed state.
Introduce 'qemuFDPassNewPassed' and 'qemuFDPassIsPassed'.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Until now the code didn't expect that we'd want to rollback/detach a FD
passed on the commandline, but whith disk backend FD passing this can
happen.
Properly mark the 'qemuFDPass' object as passed to qemu even when it was
done on the commandline.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Copy the pointer to qemuFDPass into struct qemuBlockStorageSourceAttachData
so that it can be used from qemuBuildBlockStorageSourceAttachDataCommandline
rather than looping again in qemuBuildDiskSourceCommandLineFDs.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Be consistent with other children buffer variable naming scheme.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
The iTCO watchdog is part of the q35 machine type since its inception,
we just did not add it implicitly.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2137346
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
In order for the iTCO watchdog to be operational we must disable the
noreboot pin strap in qemu. This is the default starting from 8.0
machine types, but desirable for older ones as well. And we can safely
do that since that is not guest-visible.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This is already possible with qemu, and actually already happening with
q35 machines and a specified watchdog since q35 already includes a
watchdog we do not include in the XML. In order to express such
posibility multiple watchdogs need to be supported.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The g_hash_table_unref() function does not accept NULL. Passing
NULL results in a glib warning being triggered. Check whether the
hash table is not NULL and unref it only then.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Support virtio-crypto device, also support cryptodev types:
- builtin
- lkcf
Finally, we can launch a VM(QEMU) with one or more crypto devices by
libvirt.
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Changes in this commit:
- docs: formatdomaincaps.rst
- conf: crypto related domain caps
- qemu: crypto related
- tests: crypto related test
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Introduce crypto device like:
<crypto model='virtio' type='qemu'>
<backend model='builtin' queues='1'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x0a' function='0x0'/>
</crypto>
<crypto model='virtio' type='qemu'>
<backend model='lkcf'/>
<address type='pci' domain='0x0000' bus='0x00' slot='0x0b' function='0x0'/>
</crypto>
Currently, crypto model supports virtio only, type supports qemu only
(vhost-user in the plan). For the qemu type, backend supports modle
builtin/lkcf, and the queues is optional.
Changes in this commit:
- docs: formatdomain.rst
- schemas: domaincommon.rng
- conf: crypto related domain conf
- qemu: crypto related
- tests: crypto related test
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The field is no longer used so we can remove it and the code filling it.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
All callers pass 'false' so we no longer need it.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Commit b7798a07f93 (in fall of 2016) changed the way we generate aliases
for 'dimm' memory devices as the alias itself is part of the migration
stream section naming and thus must be treated as ABI.
The code added compatibility layer for VMs with memory hotplug started
with the old scheme to prevent from generating wrong aliases. The
compatibility layer broke though later when 'nvdimm' and 'pmem' devices
were introduced as it wrongly detected them as old configuration.
Now rather than attempting to fix the legacy compat layer to treat other
devices properly we'll be better off simply removing it as it's
extremely unlikely that somebody has a VM started in 2016 running with
today's libvirt and attempts to hotplug more memory.
This fixes a corner case when a user hot-adds a 'dimm' into a VM with a
'dimm' and a 'nvdimm' after restart of libvirtd and then attempts to
migrate the VM.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2158701
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
In error case, unref event->vm instead of vm. This makes it
easier for the reader to understand as it is the event struct
that's holding the reference.
Signed-off-by: Shaleen Bathla <shaleen.bathla@oracle.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
We have virDomainGetCPUStats() API which offers querying
statistics on host CPU usage by given guest. And it works in two
modes: getting overall stats (@start_cpu == -1, @ncpus == 1) or
getting per host CPU usage.
For the QEMU driver it is implemented by looking into values
stored in corresponding cpuacct CGroup controller. Well, this
works for system instances, where libvirt has permissions to
create CGroups and place QEMU process into them. But it does not
fly for session connection, where no CGroups are set up.
Fortunately, we can do something similar to v8.8.0-rc1~95 and use
virProcessGetStatInfo() to fill the overall stats. Unfortunately,
I haven't found any source of per host CPU usage, so we just
continue throwing an error in that case.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Firstly, the virProcessGetStatInfo() does not fail really. But
even if it did, it sets correct errno only sometimes (and even
that is done in a helper it's calling - virProcessGetStat() and
even there it's the case only in very few error paths).
Therefore, using virReportSystemError() to report errors is very
misleading. Use plain virReportError() instead. Luckily, there
are only two places where the former was used:
chDomainHelperGetVcpus() and qemuDomainHelperGetVcpus() (not a
big surprise since CH driver is heavily inspired by QEMU driver).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
In a recent commit of v9.0.0-rc1~192 I've tried to forbid case
where a TAP device already exists, but at the same time it's
managed by Libvirt (<interface type='ethernet'> <target
dev='tap0' managed='yes'/> </interface>). NB, if @managed
attribute is missing then it's assumed to be managed by Libvirt.
Anyway, I've mistakenly put setting of
VIR_NETDEV_TAP_CREATE_ALLOW_EXISTING flag into managed='yes'
branch instead of managed='no' branch in
qemuInterfaceEthernetConnect().
Move the setting of the flag into the correct branch.
Fixes: a2ae3d299cf9c5ada8aa42ec4271748eb479dc27
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
The systemd service files of the qemu and libxl driver currently have a
'Requires' dependency on virtlockd, which is too strong since virtlockd
is not enabled by default in either driver. Change the dependency to a
'Wants' to avoid a package dependency between the driver subpackages and
the new libvirt-daemon-lock subpackage.
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
This commented-out option was pointed out by jtomko during review, but
I missed taking it out when addressing his comments.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This attribute was added to support setting the --interface option for
passt, but in a post-push/pre-9.0-release review, danpb pointed out
that it would be better to use the existing <source dev='xxx'/>
attribute to set --interface rather than creating a new attribute (in
the wrong place). So we remove backend/upstream, and change the passt
commandline creation to grab the name for --interface from source/dev.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Currently, the ThreadContext object is generated whenever we see
.host-nodes attribute for a memory-backend-* object. The idea was
that when the backend is pinned to a specific set of host NUMA
nodes, then the allocation could be happening on CPUs from those
nodes too. But this may not be always possible.
Users might configure their guests in such way that vCPUs and
corresponding guest NUMA nodes are on different host NUMA nodes
than emulator thread. In this case, ThreadContext won't work,
because ThreadContext objects live in context of the emulator
thread (vCPU threads are moved around by us later, when emulator
thread finished its setup and spawned vCPU threads - see
qemuProcessSetupVcpus()). Therefore, memory allocation is done by
emulator thread which is pinned to a subset of host NUMA nodes,
but tries to create a ThreadContext object with a disjoint subset
of host NUMA nodes, which fails.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2154750
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Jumping to the error label and reading the pidfile does not make sense
until we reached qemuSecurityCommandRun which creates the pidfile.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
The pidfile is guaranteed to be non-NULL (thanks to glib allocation
functions) and it's dereferenced two lines above anyway.
Reported by coverity:
/src/qemu/qemu_passt.c: 278 in qemuPasstStart()
272 return 0;
273
274 error:
275 ignore_value(virPidFileReadPathIfLocked(pidfile, &pid));
276 if (pid != -1)
277 virProcessKillPainfully(pid, true);
>>> CID 404360: Null pointer dereferences (REVERSE_INULL)
>>> Null-checking "pidfile" suggests that it may be null, but it
>>> has already been dereferenced on all paths leading to the check.
278 if (pidfile)
279 unlink(pidfile);
280
281 return -1;
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
To ensure same behaviour when remote driver is or is not used we must
not steal the FDs and array holding them passed to qemuDomainFDAssociate
but rather duplicate them. At the same time the remote driver must close
and free them to prevent leak.
Pointed out by Coverity as FD leak on error path:
*** CID 404348: Resource leaks (RESOURCE_LEAK)
/src/remote/remote_daemon_dispatch.c: 7484 in remoteDispatchDomainFdAssociate()
7478 rv = 0;
7479
7480 cleanup:
7481 if (rv < 0)
7482 virNetMessageSaveError(rerr);
7483 virObjectUnref(dom);
>>> CID 404348: Resource leaks (RESOURCE_LEAK)
>>> Variable "fds" going out of scope leaks the storage it points to.
7484 return rv;
Fixes: abd9025c2fd
Fixes: f762f87534e
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This consists of (1) adding the necessary args to the qemu commandline
netdev option, and (2) starting a passt process prior to starting
qemu, and making sure that it is terminated when it's no longer
needed. Under normal circumstances, passt will terminate itself as
soon as qemu closes its socket, but in case of some error where qemu
is never started, or fails to startup completely, we need to terminate
passt manually.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
