Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

This improves commit 706b5b62 in a way that we check qemu capabilities
instead of what architecture we are running on to detect whether we can
use *virtio-vga* model or not.  This is not a case only for arm/aarch64.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Commit 21373feb added support for primary virtio-vga device but it was
checking for virtio-gpu.  Let's check for existence of virtio-vga if we
want to use it.

Virtio video device is currently represented by three different models
*virtio-gpu-device*, *virtio-gpu-pci* and *virtio-vga*.  The first two
models are tied together and if virtio video devices is compiled in they
both exist.  However, the *virtio-vga* model doesn't have to exist on
some architectures even if the first two models exist.  So we cannot
group all three together.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Before this patch we've checked qemu capabilities for video devices
only while constructing qemu command line using "-device" option.

Since we support qemu only if "-device" option is present we can use
the same capabilities to check also video devices while using "-vga"
option to construct qemu command line.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Runtime validation that depend on qemu capabilities should be moved
into qemuProcessStartValidateXML.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

All definition validation that doesn't depend on qemu capabilities
and was allowed previously as valid definition should be placed into
qemuDomainDefValidate.

The check whether video type is supported or not was based on an enum
that translates type into model.  Use switch to ensure that if new
video type is added, it will be properly handled.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

We generally uses QEMU_CAPS_DEVICE_$NAME to probe for existence of some
device and QEMU_CAPS_$NAME_$PROP to probe for existence of some property
of that device.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

If QEMU in question supports QMP, this capability is set if
QEMU_CAPS_DEVICE_QXL was set based on existence of "-device qxl". If
libvirt needs to parse *help*, because there is no QMP support, it
checks for existence of "-vga qxl", but it also parses output of
"-device ?" and sets QEMU_CAPS_DEVICE_QXL too.

Now that libvirt supports only QEMU that has "-device" implemented it's
safe to drop this capability and stop using it.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

This patch simplifies QEMU capabilities for QXL video device.  QEMU
exposes this device as *qxl-vga* and *qxl* and they are both the same
device with the same set of parameters, the only difference is that
*qxl-vga* includes VGA compatibility.

Based on QEMU code they are tied together so it's safe to check only for
presence of only one of them.

This patch also removes an invalid test case "video-qxl-sec-nodevice"
where there is only *qxl-vga* device and *qxl* device is not present.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

If one of QEMU_CAPS_DEVICE_QXL_VGA or QEMU_CAPS_DEVICE_QXL is set the
other one will always be set as well because both devices are tied
together in QEMU.

The change of args files is caused by the presence of capability
QEMU_CAPS_DEVICE_VIDEO_PRIMARY which means it's safe to use
"-device qxl-vga" instead of "-vga qxl", see commit (e3f2686b) and
by the fact that if QEMU_CAPS_VGA_QXL is set QEMU_CAPS_DEVICE_QXL_VGA
and QEMU_CAPS_DEVICE_QXL would be set too (since we support only qemu
with "-device" option).
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Qemu supports *xen* video device only with XEN and this code was part
of xenner code.  We dropped support for xenner in commit de9be0ab.

Before this patch if you used 'xen' video type you ended up with
domain without any video device at all.  Now we don't allow to start
such domain.
Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

Signed-off-by: NPavel Hrdina <phrdina@redhat.com>

If the VM is offline virsh attempted to at least report the pinning
information for the VM. This would not work properly now that the vcpus
can be sparse. Fix it by getting the vcpu states from the XML.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1375920

Put it into a separate function so that more fallback handling can be
added without making a mess.

cmdVcpuinfo will be split in upcomming patches thus extract the common
code that formats pinning cpumaps for the vcpus.

The fallback code used if virDomainGetVcpusFlags is not supported used
wrong XPath queries and basically did not work at all. Fix them to point
to the <domain> <vcpu> element instead of <vcpus> which was not present
until lately.

I managed to space most of the code by 5 spaces instead of 4 when
orignally implementing this function.

The functions just read the passed pointer so it can be marked as const.

The Requires relationship is very strong, in that it prevents
a unit from running unless all the units it Requires are
running as well.

This turns out to be a problem because we want to be able to
restart libvirtd at any time without having libvirt-guests
suspend or shutdown running guests.

Turn the Requires relationship into a Wants relationship:
this way starting libvirt-guests will cause systemd to (attempt
to) start libvirtd as well, but stopping or restarting libvirtd
will not alter libvirt-guests' running state.

It was never safe anyway and as such shouldn't have been enabled in the
first place.  Future patches will allow hot-(un)pluging of some ivshmem
devices as a workaround.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

Due to the switch of parameters in a call to virDomainShmemDefEquals()
no device was found when looking for device with all the information
except address.  Also fix the indentation.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

If the last event callback is unregistered while the event loop is
dispatching, it is only marked as deleted, but not removed.  The number
of callbacks is more than zero in that case, so the timer is not
removed.  Because it can be removed in this function now (but also
accessed afterwards so that we set 'isDispatching = false' and have it
locked), we need to temporarily increase the reference counter of the
state for the duration of this function.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

There is a repeating pattern of code that removes the timer if it's not
needed.  So let's move it to a new function.  We'll also use it later.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

There should be one more reference because it is being kept in the list
of callbacks as an opaque.  We also unref it properly using
virObjectFreeCallback.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

This way we get reference counting and we can get rid of locking
function.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>

Currently Libvirt allows attempts to migrate read only disks. Qemu
cannot handle this as read only disks cannot be written to on the
destination system. The end result is a cryptic error message and a
failed migration.

This patch causes migration to fail earlier and provides a meaningful
error message stating that migrating read only disks is not supported.
Signed-off-by: NCorey S. McQuay <csmcquay@linux.vnet.ibm.com>
Reviewed-by: NJason J. Herne <jjherne@linux.vnet.ibm.com>
Reviewed-by: NBoris Fiuczynski <fiuczy@linux.vnet.ibm.com>

Rather than multiplying sockets, cores, and threads use the new helper
for getting the vcpu count resulting from the topology.

Make sure that the topology results into a sane number of cpus (up to
UINT_MAX) so that it can be sanely compared to the vcpu count of the VM.

Additionally the helper added in this patch allows to fetch the total
number the topology results to so that it does not have to be
reimplemented later.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1378290

Since it's only used in loop - just go direct.

If not initialized and the virAsprintf to jsonreply or fulllablel fails,
then the call to qemuMonitorTestFree will take stack data.

1. Move the declaration of const vshCmdDef *help - it should be at the
   top of the "if" rather than in the middle.

2. Change a comparison from && to || - without doing so we could crash
   on commands like 'virsh list' which would allow completion of some
   non -- option based on whatever was found in the current working
   directory and then as soon as that was completed, the next <tab>
   would crash since "opt" would be returned as NULL, but the check
   was dereferencing "&& opt->type"

3. Before dereferencing opt->completer, be sure opt isn't NULL.

In both virLogParseOutput and virLogParseFilter, rather than returning
NULL, goto cleanup since it's possible that for each the first condition
passes, but the || condition doesn't and thus we leak memory.

Yet another case of not needing virGetLastError processing

Remove the !size check since size is initialized to NULL and thus
causing the condition to always be true

The dnsmasq man page recommends that dhcp-authoritative "should be
set when dnsmasq is definitely the only DHCP server on a network".
This is the case for libvirt-managed virtual networks.

The effect of this is that VMs that fail to renew their DHCP lease
in time (e.g. if the VM or host is suspended) will be able to
re-acquire the lease even if it's expired, unless the IP address has
been taken by some other host. This avoids various annoyances caused
by changing VM IP addresses.

Sometimes virObjectEventStateFlush can be called without timer (if the
last event was unregistered right when the timer fired).  There is a
check for timer == -1, but that triggers warning and other log messages,
which is unnecessary.
Signed-off-by: NMartin Kletzander <mkletzan@redhat.com>