There are two types of vhostuser ports:
dpdkvhostuser - OVS creates the socket and QEMU connects to it
dpdkvhostuserclient - QEMU creates the socket and OVS connects to it
But of course ovs-vsctl syntax for fetching ifname is different.
So far, we've implemented the former. The lack of implementation
for the latter means that we are not detecting the interface name
and thus not reporting it in domain XML, or failing to get
interface statistics.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1767013
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
This function always returns zero, so it might as well be void.
Signed-off-by: Matt Coleman <matt@datto.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Currently all errors from qemuInterfacePrepareSlirp() are completely
ignored by the callers. The intention is that missing qemu-slirp binary
should cause the caller to fallback to the built-in slirp impl.
Many of the possible errors though should indeed be considered fatal.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
By using the new qemu monitor functions to handle passing and removing
file descriptors, we can support hotplug of vdpa devices.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Enable <interface type='vdpa'> for qemu domains. This provides basic
support and does not support hotplug or migration.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
This patch adds new schema and adds support for parsing and formatting
domain configurations that include vdpa devices.
vDPA network devices allow high-performance networking in a virtual
machine by providing a wire-speed data path. These devices require a
vendor-specific host driver but the data path follows the virtio
specification.
When a device on the host is bound to an appropriate vendor-specific
driver, it will create a chardev on the host at e.g. /dev/vhost-vdpa-0.
That chardev path can then be used to define a new interface with
type='vdpa'.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
SCSI hostdev setup requires querying the host os for the actual path of
the configured hostdev. This was historically done in the command line
formatter. Our new approach is to split out this part into
'qemuProcessPrepareHost' which is designed to be skipped in tests.
Refactor the hostdev code to use this new semantics, and add appropriate
handlers filling in the data for tests and the qemuConnectDomainXMLToNative
users.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
These variables seem to be left over from a previous refactoring and
they don't add anything to the code.
Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Laine Stump <laine@redhat.com>
By default, pfifo_fast queueing discipline (qdisc) is set on
newly created interfaces (including TAPs). This qdisc has three
queues and packets that want to be sent through given NIC are
placed into one of the queues based on TOS field. Queues are then
emptied based on their priority allowing interactive sessions
stay interactive whilst something else is downloading a large
file.
Obviously, this means that kernel has to be involved and some
locking has to happen (when placing packets into queues). If
virtualization is taken into account then the above algorithm
happens twice - once in the guest and the second time in the
host.
This is arguably not optimal as it burns host CPU cycles
needlessly. Guest already made it choice and sent packets in the
order it wants.
To resolve this, Linux kernel offers 'noqueue' qdisc which can be
applied on virtual interfaces and in fact for 'lo' it is by
default:
lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue
Set it for other TAP devices we create for domains too. With this
change I was able to squeeze 1Mbps more from a macvtap attached
to a guest and to my 1Gbps LAN (as measured by iperf3).
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1329644
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
I left in a 'return' or 'goto cleanup' in a few places
where I did the conversion manually.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reported-by: John Ferlan <jferlan@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
After recent refactors the function can be refactored to remove the
'cleanup' label by using autoptr for the 'map' variable.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Remove the overlay if the disk was <transient/>. Note that even if we'd
forbid unplug of such a disk through the API, the disk can still be
ejected from the guest.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Tested-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Tested-by: Ján Tomko <jtomko@redhat.com>
For now we disable disk hotplug of transient disk as it requires
creating an overlay prior to adding the frontend.
Signed-off-by: Masayoshi Mizuma <m.mizuma@jp.fujitsu.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Tested-by: Ján Tomko <jtomko@redhat.com>
In [1], changes were made to remove the existing auto-alignment
for pSeries NVDIMM devices. That design promotes strange situations
where the NVDIMM size reported in the domain XML is different
from what QEMU is actually using. We removed the auto-alignment
and relied on standard size validation.
However, this goes against Libvirt design philosophy of not
tampering with existing guest behavior, as pointed out by Daniel
in [2]. Since we can't know for sure whether there are guests that
are relying on the auto-alignment feature to work, the changes
made in [1] are a direct violation of this rule.
This patch reverts [1] entirely, re-enabling auto-alignment for
pSeries NVDIMM as it was before. Changes will be made to ease
the limitations of this design without hurting existing
guests.
This reverts the following commits:
- commit 2d93cbdea9
Revert "formatdomain.html.in: mention pSeries NVDIMM 'align down' mechanic"
- commit 0ee56369c8
qemu_domain.c: change qemuDomainMemoryDeviceAlignSize() return type
- commit 07de813924
qemu_domain.c: do not auto-align ppc64 NVDIMMs
- commit 0ccceaa57c
qemu_validate.c: add pSeries NVDIMM size alignment validation
- commit 4fa2202d88
qemu_domain.c: make qemuDomainGetMemorySizeAlignment() public
[1] https://www.redhat.com/archives/libvir-list/2020-July/msg02010.html
[2] https://www.redhat.com/archives/libvir-list/2020-September/msg00572.html
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Historically we've prepared secrets for all objects in one place. This
doesn't make much sense and it's semantically more appealing to prepare
everything for a single device type in one place.
Move the setup of the (iSCSI|SCSI) hostdev secrets into a new function
which will be used to setup other things as well in the future.
This is a similar approach we do for disks.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Introduce a new device element "<audio>" which allows
to map guest sound device specified using the "<sound>"
element to specific audio backend.
Example:
<sound model='ich7'>
<audio id='1'/>
</sound>
<audio id='1' type='oss'>
<input dev='/dev/dsp0'/>
<output dev='/dev/dsp0'/>
</audio>
This block maps to OSS audio backend on the host using
/dev/dsp0 device for both input (recording)
and output (playback).
OSS is the only backend supported so far.
Signed-off-by: Roman Bogorodskiy <bogorodskiy@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
After the recent changes, this function is now always returning
zero. Turn it to 'void' to relieve callers from checking it.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The qemu_domain.c file is big as is and we should split it into
separate semantic blocks. Start with code that handles domain
namespaces.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Similarly to previous commits, modify the hostdev detach code to use
blockdev infrastructure to detach (i)SCSI hostdevs.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Similarly to command line creation, use the blockdev helpers when
hotplugging an (i)SCSI hostdev.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Don't (re)generate the backend alias (alias of the -drive backend for
now) internally but rather pass it in. Later on it will be replaced by
the nodename when blockdev is used depending on the capabilities.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
We now store the alias of the secrets in the status XML so there's no
need to generate it again.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
During disk hot plugging, qemuDomainAttachDeviceLive() adds the new
disk to the device list of the VM object. However, hot plugging
cdroms and floppies only updates the src variable of the original
disk device, so the newly generated disk object needs to be freed.
Signed-off-by: Jin Yan <jinyan12@huawei.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Let us fix the issues with zPCI address validation and auto-generation
on s390.
Currently, there are two issues with handling the ZPCI address
extension. Firstly, when the uid is to be auto-generated with a
specified fid, .i.e.:
...
<address type='pci'>
<zpci fid='0x0000001f'/>
</address>
...
we expect uid='0x0001' (or the next available uid for the domain).
However, we get a parsing error:
$ virsh define zpci.xml
error: XML error: Invalid PCI address uid='0x0000', must be > 0x0000
and <= 0xffff
Secondly, when the uid is specified explicitly with the invalid
numerical value '0x0000', we actually expect the parsing error above.
However, the domain is being defined and the uid value is silently
changed to a valid value.
The first issue is a bug and the second one is undesired behaviour, and
both issues are related to how we (in-band) signal invalid values for
uid and fid. So let's fix the XML parsing to do validation based on what
is actually specified in the XML.
The first issue is also related to the current code behaviour, which
is, if either uid or fid is specified by the user, it is incorrectly
assumed that both uid and fid are specified. This bug is fixed by
identifying when the user specified ZPCI address is incomplete and
auto-generating the missing ZPCI address.
Signed-off-by: Bjoern Walk <bwalk@linux.ibm.com>
Signed-off-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Signed-off-by: Shalini Chellathurai Saroja <shalini@linux.ibm.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
If virDomainUpdateDeviceFlags() was used to update an <interface>, and
the interface type changed from type='network' where the network was
an unmanaged bridge (so actualType == bridge) to type='bridge'
(i.e. actualType *also* == bridge), the update would fail due to the
perceived change in type.
In practice it is okay to switch between any interface types that end
up using a tap device, since libvirt just needs to attach the device
to a new bridge. But in this case we were erroneously rejecting it due
to a conditional that was too restrictive. This is what the code was doing:
if (old->type != new->type)
[allow update]
else
if ((oldActual == bridge and newActual == network)
|| (oldActual == network and newActual == bridge)) {
[allow update]
else
[error]
In the case described above though, old->type and new->type don't match,
but oldActual and newActual are both 'bridge', so we get an error.
This patch changes the inner conditional so that any combination of
'network' and 'bridge' for oldActual and newActual, since they both
use a tap device connected to a bridge.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
According to the context, here we are checking net->downscript's validity,
Signed-off-by: Liao Pingfang <liao.pingfang@zte.com.cn>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Support downscript for booting vm,
and hotunplug interface device.
Signed-off-by: Chen Hanxiao <chen_han_xiao@126.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Now that all code paths generate JSON props we can remove the conversion
to command line arguments and back in the monitor code.
Note that the test which is removed in this commit will be replaced by a
stronger testsuite later.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Syntax of guestfwd channel also needs to be modified to conform to the
QAPI schema.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The 'netdev_add' command was recently formally described in qemu via the
QMP schema. This means that it also requires the arguments to be
properly formatted. Our current approach is to generate the command line
and then use qemuMonitorJSONKeywordStringToJSON to get the JSON
properties for the monitor. This will not work if we need to pass some
fields as numbers or booleans.
In this step we re-do internals of qemuBuildHostNetStr to format a JSON
object which is converted back via virQEMUBuildNetdevCommandlineFromJSON
to the equivalent command line. This will later allow fixing of the
monitor code to use the JSON object directly rather than rely on the
conversion.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
The output of the function is fed as argument to '-device' command line
argument or 'device_add' monitor command except for 'guestfwd' channels
where it needs to be fed to -netdev/netdev_add. This is confusing and
error prone. Split it up since the caller needs to know which
command/option to use anyways, so the caller can call the appropriate
function without any magic.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Although the original patches to support controllers with
hotplug='off' were checking during hotplug/attach requests that the
device was being plugged into a PCI controller that didn't have
hotplug disabled, but I forgot to do the same for device detach (the
main impetus for adding the feature was to prevent unplugs originating
from within the guest, so it slipped my mind). So although the guest
OS was ultimately unable to honor the unplug request, libvirt could
still be used to make such a request, and since device attach/detach
are asynchronous operations, the caller to libvirt would receive a
success status back (the device would stubbornly/correctly remain in
the domain status XML however)
This patch remedies that, by looking at the controller for the device
in the detach request, and immediately failing the operation if that
controller has hotplug=off.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>
SD cards need to be instantiated via -drive if=sd. This means that all
cases where we use the blockdev path need to be special-cased for SD
cards.
Note that at this point QEMU_CAPS_BLOCKDEV is still cleared if the VM
config has a SD card.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The function effectively boils down to whether the disk is 'SD'. Since
we'll need to make more decisions based on the fact whether the disk is
on the SD bus, rename the function.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Remove the function and passing of 'def' through the callers.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The function is mocked in qemuhotplugmock.so. Recent clang versions
decided to inline it so the mock stopped working resulting in
qemuhotplugtest wasting 15 seconds waiting for timeouts.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Helper processes may have their state migrated with QEMU data stream
thanks to the QEMU "dbus-vmstate".
libvirt maintains the list of helpers to be migrated. The
"dbus-vmstate" is added when required, and given the list of helper
Ids that must be migrated, on save & load sides.
See also:
https://git.qemu.org/?p=qemu.git;a=blob;f=docs/interop/dbus-vmstate.rst
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This code was based on a per-helper instance and peer-to-peer
connections. The code that landed in qemu master for v5.0 is relying
on a single instance and DBus bus.
Instead of trying to adapt the existing dbus-vmstate code, let's
remove it and resubmit. That should make reviewing easier.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Using the 'uuid' element for ppc64 NVDIMM memory added in the
previous patch, use it in qemuBuildMemoryDeviceStr() to pass
it over to QEMU.
Another ppc64 restriction is the necessity of a mem->labelsize,
given than ppc64 only support label-area backed NVDIMMs.
Finally, we don't want ppc64 NVDIMMs to align up due to the
high risk of going beyond the end of file with a 256MiB
increment that the user didn't predict. Align it down
instead. If target size is less than the minimum of
256MiB + labelsize, error out since QEMU will error out
if we attempt to round it up to the minimum.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Many calls of qemuMonitorDelObject don't actually check the return value
or report the error from the object deletion itself since they are on
cleanup paths. In some cases this can lead to reporting of spurious
errors e.g. when qemuMonitorDelObject is used to clean up a possibly
pre-existing objects.
Add a new argument for qemuMonitorDelObject which controls whether
the internals report errors from qemu and fix all callers accordingly.
Note that some of the cases on device unplug which check the error code
don't in fact propagate the error to the user, but in this case it is
important to add the log entry anyways for tracing that the device
deletion failed.
https://bugzilla.redhat.com/show_bug.cgi?id=1784040
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
When changing media we'd attempt to remove the managed pr daemon even if
neither of the images involved in the media change used it. This caused
libvirtd to log a spurious error:
2020-03-18 01:41:19.832+0000: 643207: error : qemuMonitorJSONCheckError:412 : internal error: unable to execute QEMU command 'object-del': object 'pr-helper0' not found
With this patch we completely avoid calling the deletion code.
https://bugzilla.redhat.com/show_bug.cgi?id=1814486
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The loop which checks whether the vcpus are in proper configuration for
the requested hot(un)plug skips the first modified vcpu. This means
that 'firstvcpu' which is used to print the error message in case the
configuration is not suitable would never point to the first modified
vcpu.
In cases such as:
<vcpu placement='auto' current='5'>8</vcpu>
<vcpus>
<vcpu id='0' enabled='yes' hotpluggable='no'/>
<vcpu id='1' enabled='yes' hotpluggable='no'/>
<vcpu id='2' enabled='yes' hotpluggable='no'/>
<vcpu id='3' enabled='yes' hotpluggable='no'/>
<vcpu id='4' enabled='yes' hotpluggable='no'/>
<vcpu id='5' enabled='no' hotpluggable='yes'/>
<vcpu id='6' enabled='no' hotpluggable='yes'/>
<vcpu id='7' enabled='no' hotpluggable='yes'/>
</vcpus>
# virsh setvcpu --config --disable upstream 1
error: invalid argument: vcpu '-1' can't be modified as it is followed by non-hotpluggable online vcpus
After this fix the proper vcpu is reported in the error message:
# virsh setvcpu --config --disable upstream 1
error: invalid argument: vcpu '1' can't be modified as it is followed by non-hotpluggable online vcpu
https://bugzilla.redhat.com/show_bug.cgi?id=1611061
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
There are two last callers of this function. Replace them by
qemuAliasForSecret and delete qemuDomainGetSecretAESAlias.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
