This patch adds the implementation of the ccf-assist pSeries
feature, based on the QEMU_CAPS_MACHINE_PSERIES_CAP_CCF_ASSIST
capability that was added in the previous patch.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Linux kernel 5.1 added a new PPC KVM capability named
KVM_PPC_CPU_CHAR_BCCTR_FLUSH_ASSIST, which is exposed to the QEMU guest
since QEMU commit 8ff43ee404d under a new sPAPR capability called
SPAPR_CAP_CCF_ASSIST. This cap indicates whether the processor supports
hardware acceleration for the count cache flush workaround, which
is a software workaround that flushes the count cache on context
switch. If the processor has this hardware acceleration, the software
flush can be shortened, resulting in performance gain.
This hardware acceleration is defaulted to 'off' in QEMU. The reason
is that earlier versions of the Power 9 processor didn't support
it (it is available on Power 9 DD2.3 and newer), and defaulting this
option to 'on' would break migration compatibility between the Power 9
processor class.
However, the user running a P9 DD2.3+ hypervisor might want to create
guests with ccf-assist=on, accepting the downside of only being able
to migrate them only between other P9 DD2.3+ hosts running upstream
kernel 5.1+, to get a performance boost.
This patch adds this new capability to Libvirt, with the name of
QEMU_CAPS_MACHINE_PSERIES_CAP_CCF_ASSIST.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
This device is a very simple framebuffer device supported by qemu that
is mostly intended to use as a boot framebuffer in conjunction with a
vgpu. However, there is also a standalone ramfb device that can be used
as a primary display device and is useful for e.g. aarch64 guests where
different memory mappings between the host and guest can prevent use of
other devices with framebuffers such as virtio-vga.
https://bugzilla.redhat.com/show_bug.cgi?id=1679680 describes the
issues in more detail.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Add a qemu capbility to see if the standalone ramfb device is available.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
When the bochs display type was added, the capability was never checked.
Add that check in the same place as the other video device capability
checks.
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
This reverts commit a5a777a8ba.
After previous commit the domain won't disappear while connecting
to monitor. There's no need to ref monitor config then.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
When connecting to qemu's monitor the @vm object is unlocked.
This is justified - connecting may take a long time and we don't
want to wait with the domain object locked. However, just before
the domain object is locked again, the monitor's FD is registered
in the event loop. Therefore, there is a small window where the
event loop has a chance to call a handler for an event that
occurred on the monitor FD but vm is not initalized properly just
yet (i.e. priv->mon is not set). For instance, if there's an
incoming migration, qemu creates its socket but then fails to
initialize (for various reasons, I'm reproducing this by using
hugepages but leaving the HP pool empty) then the following may
happen:
1) qemuConnectMonitor() unlocks @vm
2) qemuMonitorOpen() connects to the monitor socket and by
calling qemuMonitorOpenInternal() which subsequently calls
qemuMonitorRegister() the event handler is installed
3) qemu fails to initialize and exit()-s, which closes the
monitor
4) The even loop sees EOF on the monitor and the control gets to
qemuProcessEventHandler() which locks @vm and calls
processMonitorEOFEvent() which then calls
qemuMonitorLastError(priv->mon). But priv->mon is not set just
yet.
5) qemuMonitorLastError() dereferences NULL pointer
The solution is to unlock the domain object for a shorter time
and most importantly, register event handler with domain object
locked so that any possible event processing is done only after
@vm's private data was properly initialized.
This issue is also mentioned in v4.2.0-99-ga5a777a8ba.
Since we are unlocking @vm and locking it back, another thread
might have destroyed the domain meanwhile. Therefore we have to
check if domain is still active, and we have to do it at the
same place where domain lock is acquired back, i.e. in
qemuMonitorOpen(). This creates a small problem for our test
suite which calls qemuMonitorOpen() directly and passes @vm which
has no definition. This makes virDomainObjIsActive() call crash.
Fortunately, allocating empty domain definition is sufficient.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
QEMU 2.11 for ppc64 changed all CPU model names to lower case. Since
libvirt can't change the model names for compatibility reasons, we need
to translate the matching lower case models to the names known by
libvirt.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1755303
With the recent work in daemon split and socket activation
daemons can come and go. They can and will be started many times
during a session which results in objects being autostarted
multiple times. This is not optimal. Use
virDriverShouldAutostart() to determine if autostart should be
done or not.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The comment says that the function kills domains and networks.
This is obviously not the case.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
086c19d69 added bochs-display capability but didn't fill in the info for
domain capabilities.
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This command is hooked into the virsh hypervisor-cpu-compare command.
As such, the CPU model XML provided to the command will be compared
to the hypervisor CPU contained in the QEMU capabilities file for the
appropriate QEMU binary (for s390x, this CPU definition can be observed
via virsh domcapabilities).
QMP will report that the XML CPU is either identical to, a subset of,
or incompatible with the hypervisor CPU. s390 can also report that
the XML CPU is a "superset" of the hypervisor CPU. This response is
presented as incompatible, as this CPU model would not be able to run
on the hypervisor.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Message-Id: <1568924706-2311-15-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This capability enables comparison of CPU models via QMP.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Message-Id: <1568924706-2311-13-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Interfaces with QEMU to compare CPU models. The command takes two CPU
models, A and B, that are given a model name and an optional list of
CPU features. Through the query-cpu-model-comparison command issued
via QMP, a result is produced that contains the comparison evaluation
string (identical, superset, subset, incompatible).
The list of properties (aka CPU features) that is returned from the QMP
response is ignored.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Message-Id: <1568924706-2311-12-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Perform a full CPU model expansion on the result of the baselined
model name when the features flag is present.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-11-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This command is hooked into the virsh hypervisor-cpu-baseline command.
The CPU models provided in the XML sent to the command will be baselined
via the query-cpu-model-baseline QMP command. The resulting CPU model
will be reported.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Message-Id: <1568924706-2311-10-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This capability enables baselining of CPU models via QMP.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Message-Id: <1568924706-2311-9-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Interfaces with QEMU to baseline CPU models. The command takes two
CPU models, A and B, that are given a model name and an optional list
of CPU features. Through the query-cpu-model-baseline command issued
via QMP, a result is produced that contains a new baselined CPU model
that is guaranteed to run on both A and B.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielh413@gmail.com>
Message-Id: <1568924706-2311-8-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Modify the error messages in qemuMonitorJSONParseCPUModelData to print
the command name provided to the function.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-7-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Some older s390 CPU models (e.g. z900) will not report props as a
response from query-cpu-model-expansion. As such, we should make the
props field optional when parsing the return data from the QMP response.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-6-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
query-cpu-model-baseline/comparison will accept a list of features
as part of the command. Since CPUs may be defined with CPU feature
policies, let's parse it to the appropriate boolean that the QMP
command expects.
A feature that is set to required, force, or if it is a hypervisor
CPU feature (-1), then set the property value to true. Otherwise
(optional, disabled) set the value to false.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-5-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
When expanding a CPU model via query-cpu-model-expansion, any features
that were a part of the original model are discarded. For exmaple,
when expanding modelA with features f1, f2, a full expansion may reveal
feature f3, but the expanded model will not include f1 or f2.
Let's pass a virCPUDefPtr to the expansion function in preparation for
taking features into consideration.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-4-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
With refactoring most of the expansion function, let's take care of
some additional cleanups.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Message-Id: <1568924706-2311-3-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Refactor some code in qemuMonitorJSONGetCPUModelExpansion to be later
used for the comparison and baseline functions.
Signed-off-by: Collin Walling <walling@linux.ibm.com>
Reviewed-by: Bjoern Walk <bwalk@linux.ibm.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <1568924706-2311-2-git-send-email-walling@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Parseability of disk name is now checked in qemuDomainDeviceDefValidateDisk().
Signed-off-by: Pavel Mores <pmores@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The way in which the qemu driver generates aliases for disks involves
ignoring the partition number part of a target dev name. This means that
all partitions of a block device and the device itself all end up with the
same alias. If multiple such disks are specified in XML, the resulting
name clash makes qemu invocation fail.
Since attaching partitions to qemu VMs doesn't seem to make much sense
anyway, disallow partitions in target specifications altogether.
https://bugzilla.redhat.com/show_bug.cgi?id=1346265
Signed-off-by: Pavel Mores <pmores@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
In the domain capabilities XML there are FW image paths printed.
There are two sources for the image paths (in order of
preference):
1) firmware descriptor files - as returned by
qemuFirmwareGetSupported()
2) a compile time list of FW:NRAM pairs which can be overridden
in qemu.conf
If either of those contains a duplicate FW image path (which is
a valid use case) it is printed twice in the capabilities XML.
While it's technically not a bug, it doesn't look good.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Kashyap Chamarthy <kchamart@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Similarly to the snapshot code there's no reason to modify current
checkpoint until we are done creating the new one.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Since commit f105627992 we store whether a snapshot is current globally
rather than locally in the snapshot object.
This means that we don't have to unset the current snapshot prior to
taking/reverting the snapshot and we can do it only when everything is
done successfully.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Ensure that the FD we're passing to QEMU is actually open, so we get a
sane error message upfront instead of telling QEMU to use a closed FD.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The video private data was not initializing the vhostuser FD
causing us to attempt to close FD 0 many times over.
Fixes
commit ca60ecfa8c
Author: Marc-André Lureau <marcandre.lureau@redhat.com>
Date: Mon Sep 23 14:44:36 2019 +0400
qemu: add qemuDomainVideoPrivate
Since the test suite does not invoke qemuExtDevicesStart(), no
vhost_user_fd will be present when generating test XML. To deal
with this we can must a fake FD number. While the current XML
is using FD == 0, we pick a very interesting number that's unlikely
to be a real FD, so that we're more likely to see any mistakes
closing the invalid FD.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Use the new generator residing in the monitor code rather than directly
using qemuMonitorJSONTransactionAdd.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Unify with other code that generates parameters for the 'transaction'
command.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Rather than generating the transaction contents in random places add a
unified set of APIs to generate the contents for a 'transaction' for the
dirty bitmap APIs.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The QEMU_CAPS_INCREMENTAL_BACKUP will be enabled once all bits of the
incremental backup feature work as expected which means also properly
interacting with blockjobs and snapshots.
Thus we can allow blockjobs and snapshots if QEMU_CAPS_INCREMENTAL_BACKUP
is present even when checkpoints exist.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Rather than having to fix 5 places once we support the combination, add
a function called by all the blockjob/snapshot APIs.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Checkpoints by themselves are not very useful for anything else than
testing the few bitmap interactions that are currently implemented.
It's very unlikely that anybody used this feature and thus we can
disable it until we have a more complete implementation ready.
Additionally the code for deleting checkpoints has many broken failure
scenarios which should be fixed first. This will require support of
deleting a bitmap in a qemu 'transaction' which was not released yet.
Curious users obviously can use the qemu namespace in the XML to enable
this for experiments:
<domain type='kvm' xmlns:qemu='http://libvirt.org/schemas/domain/qemu/1.0'>
...
<qemu:capabilities>
<qemu:add capability='incremental-backup'/>
</qemu:capabilities>
</domain>
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Add a new all-covering capability which will be used to interlock
incremental backup support until all bits are ready.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Add a 'cleanup' label and use jumps as we do in other places.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Once somebody is motivated enough to add the support for the quiesce
flag or offline checkpoint deletion they are welcome to do so but we
don't need to have a reminder.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
There's nothing that uses it directly now. Also not allowing direct use
will promote our layering.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Finish the refactor by moving and renaming functions from qemu_domain.c
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Move all extensive functions to a new file so that we don't just pile
everything in the common files. This obviously isn't possible with
straight code movement as we still need stubs in qemu_driver.c
Additionally some functions e.g. for looking up a checkpoint by name
were so short that moving the impl didn't make sense.
Note that in the move the new file also doesn't use
virQEMUMomentReparent but rather an stripped down copy. As I plan to
split out snapshot code into a separate file the unification doesn't
make sense any more.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The interlocking with snapshots is executed prior to the ACL check so if
a VM has snapshots invoking the checkpoint API may leak it's existance.
Introduced with the qemuDomainCheckpointCreateXML API implementation in
commit 5f4e079650.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
The code that gets the job to refresh disk sizes was not merged yet so
remove this artifact.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
'vm' is passed in which contains the definition which contains the UUID
so we don't need another parameter for this.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
'vm' is passed in which contains the definition which contains the UUID
so we don't need another parameter for this.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
