Move the majority of the packet-libvirt.h content into
packet-libvirt.c and expose only register functions which are the
only ones that are not static.
The rationale behind is that packet-libvirt.h will be included
from packet.c and therefore the header file needs to be as clean
as possible.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Cole Robinson [Wed, 6 Feb 2019 23:26:05 +0000 (18:26 -0500)]
docs: storage: owner/group default to libvirtd UID/GID
Commit fafcc818f changed the docs to say that when creating a
pool directory or file volume with no owner/group specified, they
will be inherited from the parent directory. This isn't correct
now and doesn't seem to have ever been correct
In reality default owner/group is whatever UID/GID libvirtd is
running as
rpc: client: stream: fix multi thread abort/finish
If 2 threads call abort for example then one of them
will hang because client will send 2 abort messages and
server will reply only on first of them, the second will be
ignored. And on server reply client changes the state only
one of abort message to complete, the second will hang forever.
There are other similar issues.
We should complete all messages waiting reply if we got
error or expected abort/finish reply from server. Also if one
thread send finish and another abort one of them will win
the race and server will either abort or finish stream. If
stream is aborted then thread requested finishing should report
error. In order to archive this let's keep stream closing reason
in @closed field. If we receive VIR_NET_OK message for stream
then stream is finished if oldest (closest to queue end) message
in stream queue is finish message and stream is aborted if oldest
message is abort message. Otherwise it is protocol error.
By the way we need to fix case of receiving VIR_NET_CONTINUE
message. Now we take oldest message in queue and check if
this is dummy message. If one thread first sends abort and
second thread then receives data then oldest message is abort
message and second thread won't be notified when data arrives.
Let's find oldest dummy message instead.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
rpc: client stream: dispose private data on stream dispose
If we call virStreamFinish and virStreamAbort from 2 distinct
threads for example we can have access to freed memory.
Because when virStreamFinish finishes for example virStreamAbort
yet to be finished and it access virNetClientStreamPtr object
in stream->privateData.
Also it does not make sense to clear @driver field. After
stream is finished/aborted it is better to have appropriate
error message instead of "unsupported error".
This commit reverts [1] or virNetClientStreamPtr and
virStreamPtr will never be unrefed due to cyclic dependency.
Before this patch we don't have leaks because all execution
paths we call virStreamFinish or virStreamAbort.
Checking virNetClientStreamRaiseError without client lock
is racy which is fixed in [1] for example. Thus let's remove such checks
when we are sending message to server. And in other cases
(like virNetClientStreamRecvHole for example) let's move the check
into client stream code.
virNetClientStreamRecvPacket already have stream lock so we could
introduce another error checking function like virNetClientStreamRaiseErrorLocked
but as error is set when both client and stream lock are hold we
can remove locking from virNetClientStreamRaiseError because all
callers hold either client or stream lock.
Also let's split virNetClientStreamRaiseErrorLocked into checking
state function and checking message send status function. They are
same yet.
[1] 1b6a29c21: rpc: fix race on stream abort/finish and server side abort
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Stream server error is not propagated if thread does not have the buck.
In case we have the buck we are ok due to the code added in [1].
Let's check for stream error on all paths. Now we don't need
to raise error in virNetClientCallDispatchStream.
Old code reported error only if the first message in wait
queue awaits reply. It is odd as depends on wait queue
situation. For example if we have only TX
message in queue and in one iteration loop both send the
message and receive error then thread sending TX message did
not receive the error. Next if we have RX message (first)
and TX message (second) in queue and in one iteration
loop both send the TX message and receive error then
thread sending TX message received error. In short
it was inconsistent. Let's report error whenever
we received it and for every type of message as it makes
sense to report errors as early as possible.
[1] 16c6e2b41: Fix propagation of RPC errors from streams
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
In next patches we'll add stream state checks to this
function that applicable to all call paths. This is handy
place because we hold client lock here.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
rpc: fix race on stream abort/finish and server side abort
Stream abort/finish can hang because we can receive abort message
from server and yet sent abort/finish message to server. The latter
will not be answered ever because after server sends abort message
it forgets the stream and messages for unknown stream are simply ignored.
We check for stream error at the very beginning of remoteStreamFinish/remoteStreamAbort
but stream error can be set after the check in another thread operating
on stream. Let's check for stream error under client lock similar
to what's done in [1].
These functions do mostly the same things, and it would be
preferrable if they did them in mostly the same ways. This
also fixes a few violations to our code style guidelines.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Andrea Bolognani [Wed, 28 Nov 2018 16:45:37 +0000 (17:45 +0100)]
qemu: Move qemuDomainSupportsPCI() to qemu_domain
The function operates on a virDomainDef and is not tied to
device address assignment in any way, so it makes more sense
for it to live along with qemuDomainIs*() and the like.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Andrea Bolognani [Wed, 28 Nov 2018 15:56:35 +0000 (16:56 +0100)]
qemu: Add arch checks to qemuDomainMachine*()
There is very little overlap in the machine types available
on different architectures, so broadly speaking checking the
machine type is usually enough; regardless, it's better to
check the architecture as well.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Andrea Bolognani [Wed, 28 Nov 2018 15:15:31 +0000 (16:15 +0100)]
qemu: Move functions around
Make sure related functions, eg. all qemuDomainIs*(), are
close together instead of being sprinkled throughout both
the header and implementation file, and also that all
qemuDomainMachine*() functions are declared first since
we're going to make a bunch of them static later on.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Andrea Bolognani [Wed, 28 Nov 2018 14:04:51 +0000 (15:04 +0100)]
qemu: Use more specific prefixes
While the chances of the current checks resulting in false
positives are basically zero, it's still nicer to check for
the full prefix instead of the prefix's prefix.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Erik Skultety [Fri, 8 Feb 2019 10:42:14 +0000 (11:42 +0100)]
docs: Update the AMD SEV's spec URL
Luckily, the new URL still points to the same location, the only change
is in the document name where an escaped space (%20) was replaced by an
underscore.
Signed-off-by: Erik Skultety <eskultet@redhat.com> Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Peter Krempa [Thu, 7 Feb 2019 11:50:27 +0000 (12:50 +0100)]
qemu: caps: Probe events from 'query-qmp-schema' rather than 'query-events'
QEMU plans to deprecate 'query-events' as it's non-extensible. Events
are also described by 'query-qmp-schema' so we can use that one instead.
This patch adds detection of events to
virQEMUCapsProbeQMPSchemaCapabilities using the same structure declaring
them for the old approach (virQEMUCapsEvents). This is possible as the
name is the same in the QMP schema and our detector supports that
trivially.
For any complex queries virQEMUCapsQMPSchemaQueries can be used in the
future.
For now we still call 'query-events' and discard the result so that it's
obvious that the tests pass. This will be cleaned up later.
Peter Krempa [Mon, 28 Jan 2019 08:30:41 +0000 (09:30 +0100)]
qemu: caps: Introduce capability for 'device_id' property of 'scsi-disk'
The property allows to control the guest-visible content of the vendor
specific designator of the 'Device Identification' page of a SCSI
device's VPD (vital product data).
QEMU was leaking the id string of -drive as the value if the 'serial' of
the disk was not specified. Switching to -blockdev would impose an ABI
change.
Peter Krempa [Mon, 28 Jan 2019 16:18:56 +0000 (17:18 +0100)]
qemu: command: Drop formatting of 'media=cdrom' from -drive
For SCSI, IDE, and AHCI cdroms the appropriate device types which select
the correct media are used. In qemu there's one other code path that
looks at -drive media=cdrom in the XEN pv code. Thankfully we don't
support it with qemu (see qemuBuildDiskDeviceStr). All other devices
ignore it as the comment states, thus we can drop that code.
The test fallout is expectedly only in the test added for uncommon cdrom
types.
Peter Krempa [Mon, 28 Jan 2019 16:18:55 +0000 (17:18 +0100)]
qemu: Forbid cdroms on virtio bus
Attempting to create an empty virtio-blk drive results into:
-device virtio-blk-pci,scsi=off,bus=pci.0,addr=0xc,drive=drive-virtio-disk1,id=virtio-disk1: Device needs media, but drive is empty
Attempting to eject media from virtio-blk based drive results into:
error: internal error: unable to execute QEMU command 'eject': Device 'drive-virtio-disk0' is not removable
Forbid configurations where users would attempt to use cdroms in virtio
bus.
Fix few wrong examples which are not really relevant to the tested code.
Peter Krempa [Mon, 28 Jan 2019 12:17:10 +0000 (13:17 +0100)]
tests: qemu: Remove 'disk-scsi-virtio-scsi' test
Now that we have a specific test for testing the 'virtio-scsi'
controller and other tests which test a combination of scsi and non-scsi
devices this test no longer makes sense.
Peter Krempa [Mon, 28 Jan 2019 13:52:22 +0000 (14:52 +0100)]
qemu: capabilities: Probe caps for 'ide-hd' instead of 'ide-drive'
Since commit a4cda054e7 we are using 'ide-hd' and 'ide-cd' instead of
'ide-drive'. We also should probe capabilities for 'ide-hd' instead of
'ide-drive'. It is safe to do as 'ide-drive' is the common denominator
of both 'ide-hd' and 'ide-cd' so all the properties were common.
For now the test data are modified by just changing the appropriate type
when probing for caps.
Peter Krempa [Mon, 28 Jan 2019 09:03:25 +0000 (10:03 +0100)]
qemu: capabilities: Probe caps for 'scsi-hd' instead of 'scsi-disk'
Since commit 02e8d0cfdf8 we are using 'scsi-hd' and 'scsi-cd' instead of
'scsi-disk'. We also should probe capabilities for 'scsi-hd' instead of
'scsi-disk'. It is safe to do as 'scsi-disk' is the common denominator
of both 'scsi-hd' and 'scsi-cd' so all the properties were common.
For now the test data are modified by just changing the appropriate type
when probing for caps.
Jiri Denemark [Mon, 4 Feb 2019 16:08:36 +0000 (17:08 +0100)]
Public API for post-copy migration bandwidth
This patch adds a new VIR_MIGRATE_PARAM_BANDWIDTH_POSTCOPY typed
parameter for virDomainMigrate3 and virDomainMigrateToURI3 for setting
maximum post-copy migration bandwidth.
In case the initial VIR_MIGRATE_PARAM_BANDWIDTH_POSTCOPY value turns out
to be suboptimal a new VIR_DOMAIN_MIGRATE_MAX_SPEED_POSTCOPY flag for
virDomainMigrateSetMaxSpeed and virDomainMigrateGetMaxSpeed may be used
to set/get the maximum post-copy migration bandwidth while migration is
already running.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Jiri Denemark [Mon, 4 Feb 2019 16:11:42 +0000 (17:11 +0100)]
qemu: Make migration params usable outside migration
So far migration parameters were changed only at the beginning of
migration mostly via an automatic translation from flags and typed
parameters. We need to export a few more functions to support APIs which
may set migration parameters while migration is already running.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Jiri Denemark [Thu, 31 Jan 2019 09:02:17 +0000 (10:02 +0100)]
qemu: Add optional unit to qemuMigrationParamsTPMapItem
Some migration parameters supported by libvirt may use units that differ
from the units used by QEMU for the corresponding parameters. For
example, libvirt defines migration bandwidth in MiB/s while QEMU expects
B/s. Let's add a unit field to qemuMigrationParamsTPMapItem for
automatic conversion when translating between libvirt's migration typed
parameters and QEMU's migration paramteres.
This patch is a preparation for future parameters as the existing
VIR_MIGRATE_PARAM_BANDWIDTH parameter is set using "migrate_set_speed"
QMP command rather than "migrate-set-parameters" for backward
compatibility.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Thu, 24 Jan 2019 16:50:59 +0000 (17:50 +0100)]
qemu: Use data in qemuBlockJobDataPtr instead of re-generating job name
qemuDomainBlockPivot and qemuDomainBlockJobAbort need the job name for
cancelling or pivoting but were generating it locally instead of
accessing the existing copy in the job data structure.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Thu, 24 Jan 2019 16:24:56 +0000 (17:24 +0100)]
qemu: Move shareable disk check for block copy
The writing to an image actually starts when the copy job is initiated,
so checking this at the time of the pivot operation is too late.
Move the check to qemuDomainBlockCopyCommon. Note that modern qemu would
have prevented two writers with qcow2 so the slim possibility of a job
started with libvirtd without this patch missing the check is not really
worth worrying about.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Thu, 24 Jan 2019 16:13:58 +0000 (17:13 +0100)]
qemu: Always save status XML in qemuDomainBlockJobAbort
For copy and active commit jobs we record the state of the mirror so
that we can recover. The status XML was not saved in case of
qemuDomainBlockPivot due to an oversight.
Save the XML always when invoking qemuDomainBlockJobAbort even if
the job is not currently tracking any state. This will change later and
also this is not a particularly hot code path.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Michal Privoznik [Fri, 25 Jan 2019 11:42:54 +0000 (12:42 +0100)]
lxc: Don't reboot host on virDomainReboot
If the container is really a simple one (init is just bash and
the whole root is passed through) then virDomainReboot and
virDomainShutdown will talk to the actual init within the host.
Therefore, 'virsh shutdown $dom' will result in shutting down the
host. True, at that point the container is shut down too but
looks a bit harsh to me.
The solution is to check if the init inside the container is or
is not the same as the init running on the host.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com>
Michal Privoznik [Fri, 25 Jan 2019 11:37:53 +0000 (12:37 +0100)]
virinitctl: Expose fifo paths and allow caller to chose one
So far the virInitctlSetRunLevel() is fully automatic. It finds
the correct fifo to use to talk to the init and it will set the
desired runlevel. Well, callers (so far there is just one) will
need to inspect the fifo a bit just before the runlevel is set.
Therefore, expose the internal list of fifos and also allow
caller to explicitly use one.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com>
Michal Privoznik [Thu, 24 Jan 2019 16:20:58 +0000 (17:20 +0100)]
vircgroup: Try harder to kill cgroup
Prior to rewrite of cgroup code we only had one backend to try.
After the rewrite the virCgroupBackendGetAll() returns both
backends (for v1 and v2). However, not both have to really be
present on the system which results in killRecursive callback
failing which in turn might mean we won't try the other backend.
At the same time, this function reports no error as it should.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com>
Jie Wang [Thu, 31 Jan 2019 12:50:31 +0000 (20:50 +0800)]
conf: Remove iothreads restriction in virDomainDefCheckABIStabilityFlags
The number of iothreads is not part of the vm state sent during
migration, nor exposed to the guest ABI, so this restriction is
a mistake in libvirt. Let's remove that bit of code.
Reviewed-by: Cole Robinson <crobinso@redhat.com> Signed-off-by: Jie Wang <wangjie88@huawei.com>
dosc: schema: fix usb source address device attribute format
Device attribute does not have dotted "portAddr" format. Instead it
has single number format described but "usbAddr" which corresponds
to device parsing code in virDomainHostdevSubsysUSBDefParseXML.
Looks like [1] mistakenly changed device format for hostdev devices.
And [2] copy-n-paste this for hostdev network interfaces.
[1] 31710a53 Modify USB port to be defined as a port path
[2] 3b1c191f conf: parse/format type='hostdev' network interfaces
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Reviewed-by: John Ferlan <jferlan@redhat.com>
Turns out different versions of QEMU on the same architecture
produce the same output, so we can have a single output file
per architecture instead of duplicating the same data over and
over again.
Spotted-by: Ján Tomko <jtomko@redhat.com> Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Marc Hartmayer [Mon, 4 Feb 2019 12:36:24 +0000 (13:36 +0100)]
qemu: Refresh state before starting the VCPUs
For normal starts (no incoming migration) the refresh of the QEMU
state must be done before the VCPUs getting started since otherwise
there might be a race condition between a possible shutdown of the
guest OS and the QEMU monitor queries.
This fixes "qemu: migration: Refresh device information after
transferring state" (93db7eea1b864).
Signed-off-by: Marc Hartmayer <mhartmay@linux.ibm.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Michal Privoznik [Thu, 24 Jan 2019 08:58:38 +0000 (09:58 +0100)]
qemu: Assume migration with a network disk migration is safe
If a domain has a disk that is type='network' we require specific
cache mode to allow migration with it (either 'directsync' or
'none'). This doesn't make much sense since network disks are
supposed to be safe to migrate by default.
At the same time, we should be checking for the actual source
type, not apparent type set in the domain XML.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com>
Peter Krempa [Thu, 4 Oct 2018 12:43:46 +0000 (14:43 +0200)]
qemu: domain: Assume 'raw' default storage format also for network storage
Post parse callback adds the 'raw' type only for local files. Remote
files can also have backing store (even local) so we should do this also
for network backed storage.
Note that virStorageFileGetMetadata always considers files with no type
as raw so we will not accidentally traverse the backing chain and allow
unexpected files being labelled with svirt labels.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Fri, 1 Feb 2019 16:54:46 +0000 (17:54 +0100)]
qemu: command: Don't skip 'readonly' and throttling info for empty drive
In commit f80eae8c2ae I was too agresive in removing properties of
-drive for empty drives. It turns out that qemu actually persists the
state of 'readonly' and the throttling information even for the empty
drive.
Removing 'readonly' thus made qemu open any subsequent images added via
the 'change' command as RW which was forbidden by selinux thanks to the
restrictive sVirt label for readonly media.
Fix this by formating the property again and bump the tests and leave a
note detailing why the rest of the properties needs to be skipped.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Cole Robinson [Sun, 20 Jan 2019 17:23:29 +0000 (12:23 -0500)]
Require a semicolon for VIR_ONCE_GLOBAL_INIT calls
Missing semicolon at the end of macros can confuse some analyzers
(like cppcheck <filename>). VIR_ONCE_GLOBAL_INIT is almost
exclusively called without an ending semicolon, but let's
standardize on using one like the other macros.
Add a dummy struct definition at the end of the macro, so
the compiler will require callers to add a semicolon.
Reviewed-by: John Ferlan <jferlan@redhat.com> Signed-off-by: Cole Robinson <crobinso@redhat.com>
Cole Robinson [Sun, 20 Jan 2019 16:32:42 +0000 (11:32 -0500)]
Require a semicolon for VIR_LOG_INIT calls
Missing semicolon at the end of macros can confuse some analyzers
(like cppcheck <filename>), and we have a mix of semicolon and
non-semicolon usage through the code. Let's standardize on using
a semicolon for VIR_LOG_INIT calls.
Drop the semicolon from the final statement of the macro, so
the compiler will require callers to add a semicolon.
Reviewed-by: John Ferlan <jferlan@redhat.com> Signed-off-by: Cole Robinson <crobinso@redhat.com>
projects / libvirt.git / log