docs: Render API and XML sections one per line in HTML

Finding the correct link to a XML description or API reference section
in a big blob of links concatenated in a paragraph is unpleasand and
especially for 'capabilities' and 'domain capabilities' following each
other.

Turn the API and XML reference sections into a list in RST and add CSS
to fromat it a bit more compact.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

docs: Re-order XML schema after API documentation on 'docs' page

Swap the order of links to XML schema docs and to the other language
docs. The XML schema is usually accessed more.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

NEWS: Document ARM CPU map bug fix

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

cpu_map: Install Ampere-1 ARM CPU models

Fixes: 282f38ad3d3960b24aa97d75cae95c8c3d08a218
Fixes: https://gitlab.com/libvirt/libvirt/-/issues/762
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

tests: Add a basic CPU test for aarch64

The purpose of this test is to enforce loading and parsing of ARM CPU
map so that possible issues are found earlier.

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

meson: Drop workaround for -Wframe-larger-than and clang

After previous cleanups, all functions have their stack smaller
than 2048 bytes and thus the workaround is no longer needed.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

vbox: Move parts of vboxSnapshotRedefine() into a separate function

There's too much happening inside of vboxSnapshotRedefine(). Not
only it makes the function hard to read, but it also increases
stack size of the function. Move one part into a separate
function: vboxSnapshotCreateFakeDiffStorage()

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

vbox: Move parts of vboxSnapshotRedefine() into a separate function

There's too much happening inside of vboxSnapshotRedefine(). Not
only it makes the function hard to read, but it also increases
stack size of the function. Move one part into a separate
function: vboxSnapshotAddRWDisks()

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

vbox: Move parts of vboxSnapshotRedefine() into a separate function

There's too much happening inside of vboxSnapshotRedefine(). Not
only it makes the function hard to read, but it also increases
stack size of the function. Move one part into a separate
function: vboxSnapshotAddDisksToMediaRegistry()

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

vbox: Move parts of vboxSnapshotRedefine() into a separate function

There's too much happening inside of vboxSnapshotRedefine(). Not
only it makes the function hard to read, but it also increases
stack size of the function. Move one part into a separate
function: vboxSnapshotReplaceRWDisks()

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

doRemoteOpen: Drop needless typecast of @transport

The @transport variable is already pass into the function with
proper type. There's no need to typecast it to its very same type
inside the function.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

doRemoteOpen: Move RPC talk to a separate function

When opening a connection, the client does some RPC talk
(most notably REMOTE_PROC_CONNECT_OPEN, and in some cases
REMOTE_PROC_CONNECT_GET_URI even).

Now, calling RPC means that local variables must be created.
Having them in doRemoteOpen() increases its stack size which goes
against our effort in bringing the size down (see one of previous
commits).

Move that part of the code into a separate function.

This brings the stack size of doRemoteOpen() even further: from
1320 bytes to 1272.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

remote_driver: Move URI arg extraction into a separate function

There's a problem with glib: what we might consider functions are
in fact macros and to make things worse - they do declare local
variables. For instance here's the declaration of
g_clear_pointer() macro:

#define g_clear_pointer(pp, destroy)                     \
  G_STMT_START                                           \
  {                                                      \
    G_STATIC_ASSERT (sizeof *(pp) == sizeof (gpointer)); \
    glib_typeof ((pp)) _pp = (pp);                       \
    glib_typeof (*(pp)) _ptr = *_pp;                     \
    *_pp = NULL;                                         \
    if (_ptr)                                            \
      (destroy) (_ptr);                                  \
  }                                                      \
  G_STMT_END                                             \

Now, as of v6.2.0-rc1~267 our VIR_FREE() macro is in fact a
redeclaration of g_clear_pointer(). Thus, calling VIR_FREE()
increases stack size!

Ideally, this wouldn't be a problem, because those variables
(_pp, _ptr) live in their own block. And clever compiler can just
reuse space created for one block.

But then there's clang where we are hitting this exact problem in
functions like doRemoteOpen() where either g_clear_pointer() is
called directly, or there are macros like EXTRACT_URI_ARG_STR()
which hide the call away.

That's why despite our previous efforts decreasing stack size we
still needed v9.8.0-rc1~208.

Well, moving URI argument extraction (those calls to
EXTRACT_URI_ARG_* macros) into a separate function helps us
decrease stack size from 2296 bytes to 1320.

Even after this there are still more possibilities for
improvements, but those will be addressed in future commits.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Roman Bogorodskiy <bogorodskiy@gmail.com>

ch: Fix printf format strings wrt size_t argument

In a few places, when a size_t typed argument is passed to a
printf-like function the corresponding specifier is %ld instead
of %zu. Fix those places.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

ch: Drop pid from monitor

The domain object already has a member that allows storing
hypervisor's PID (vm->pid). There's no need to duplicate it in
_virCHMonitor struct. Switch CH code to use the former.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

ch: Use CH_DOMAIN_PRIVATE() more

There are two instances where vm->privateData is typecasted only
so that it can be dereferenced further. Well, that's exactly what
CH_DOMAIN_PRIVATE() macro is for. Use that instead.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

Add NEWS item for <hyperv/> flags with Xen

Signed-off-by: Will <tcosprojects@gmail.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

libxl_conf: Implement hyperv domain flags for Xen

Adds support for configuring <hyperv/> flags for domains
running under Xen.

The following flags, making use of QEMU's existing flags, are now
configurable for Xen: vapic, synic, stimer, frequencies, tlbflush and
ipi.

Tests have been added validating translation to libxl's viridian flags

Updated docs section on <hyperv/> flags to note support and to specify
which flags work with Xen.

Signed-off-by: Will <tcosprojects@gmail.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

tools: virsh: metadata: do not report error on missing metadata

Similarly to `desc` and `net-desc`, return an empty string if
there is no metadata to be returned.

https://issues.redhat.com/browse/RHEL-27172

Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

Translated using Weblate (Ukrainian)

Currently translated at 100.0% (10917 of 10917 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/uk/

Signed-off-by: Yuri Chornoivan <yurchor@ukr.net>

docs: add table showing guest IP/DNS/gateway settings when using SLIRP

When using the default SLIRP backend for <interface type='user'>, the
<ip address='blah' prefix='blur'/> setting doesn't behave as might be
expected (i.e. it doesn't set the guest interface IP/prefix to exactly
the provided values). This *should* have created questions when users
originally encountered it, but instead it has become more apparent as
people are contemplating switching from using the SLIRP backend to
using passt instead (with passt, the <ip> settings do behave "as
expected").

In order to make this difference in behavior less mysterious, Yalan
Zhang kindly took the time to test and document the effect of various
representative <ip> settings on guest interface config when SLIRP is
used (see https://issues.redhat.com/browse/RHEL-46601); this patch
adds that same table to libvirt's documentation.

Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

ch: virCHProcessEvent() vm shutdown event handler fix

When the domain shutdown was executed from virsh, only the VM
process (a child of the CH monitor) was terminated. Since we assume
only one VM per monitor, the monitor process should also be
terminated.

Modified the VM shutdown event handler to match the VMM shutdown
behavior, ensuring the VM monitor stops along with the VM. Also
updated the virCHEventStopProcess job type, as it only destroys the
domain rather than modifying anything.

Signed-off-by: Kirill Shchetiniuk <kshcheti@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

ch: virCHProcessEvent() update domain info after reboot

When the domain was rebooted, some of its properties were changed but
not updated in the transient domain definition. This led to the
inability to connect to the serial console as its path had changed
during the reboot but was not updated in the domain definition.

Added VIR_CH_EVENT_VM_REBOOTED event handling to update the
information in transient domain definition after domain's reboot is
completed to maintain it in consistent state.

Signed-off-by: Kirill Shchetiniuk <kshcheti@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

ch: virCHMonitorNew() run new CH monitor daemonized

When the new CH monitor was started, it ran as a non-daemonized
process and was a child of the CH driver process. This led to a
situation where if the CH driver died, the monitor process were
killed too, terminating the running VM under the monitor. This
led to termination of all VM started under the libvirt.

Make new monitor running daemonized to avoid VMs shutdown when
driver dies. Also added a pidfile its preparetion to be able
to aquire daemon's PID.

Signed-off-by: Kirill Shchetiniuk <kshcheti@redhat.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

qemu: capabilities: Retire QEMU_CAPS_SNAPSHOT_INTERNAL_QMP

The code now always assumes support for the QMP internal snapshot
commands so the capability is no longer needed.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: monitor: Remove HMP monitor code

Now that we've replaced the final two HMP commands used by libvirt we
can fully drop the 'text' monitor support.

The only thing we keep is the HMP passtrhough with
'virsh qemu-monitor-command'.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: Always revert internal snapshots via QMP rather than '-loadvm'

As all supported qemu versions now support the QMP internal snapshot
commands (QEMU_CAPS_SNAPSHOT_INTERNAL_QMP is always present) we can
remove the code for loading snapshots during startup via '-loadvm'.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: snapshot: Always assume support for QEMU_CAPS_SNAPSHOT_INTERNAL_QMP

The 'snapshot-save' QMP command was introduced in 'qemu-6.0' and libvirt
now requires at least 'qemu-6.2'. Thus we can assume that the QMP
command can be used always.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemuSnapshotCreateActiveInternal: Fix error logic

The 'ret' variable is set to 0 before a call which can theoretically
fail. Not in practice really as the failure scenarion includes only
object initialization.

Since the code already has another variable for checking monitor returns
use that one properly so that the code makes sense.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: domain: Don't check return value of 'virCPUDefCopy'

As 'virCPUDefCopy' can't fail any more (without aborting) remove the
last two return value checks.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: monitor: Improve field annotations in QEMU_CHECK_MONITOR

The macro checking monitor object state also logs information such as
the monitor object pointer and the number of the monitor FD.

Name the field 'monfd' instead of 'fd' as it's confusing when debugging
FD pasing via monitor.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

tests: domainbackupxml2xml: Add test case for pull-mode backup with NBD transport='fd'

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

backup: Add support for passing server socket file descriptor to backup NBD server

In deployments where libvirt is containerized together with the VM it
may be hard for the management application to access listening sockets
inside the container from the outside.

This patch implements "transport='fd'" for the NBD server definition for
backups which allows to use the existing "virDomainFDAssociate()" to
pass FD to a pre-opened server socket to qemu instead of trying to
create it by qemu.

Add schema, enable the parser, add formatter and implement the actual
passing for the qemu backup code.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Spellchecked-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: monitor: Support FD passing of sockets to 'qemuMonitorJSONNBDServerStart'

Upcoming patches will extend the FD passing infrastructure to the backup
job so that users can pass an opened socket instead of qemu opening it
themself to bypass difficulities caused by containerizing libvirt.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

virStorageNetHostDefClear: Move into virStorageNetHostDefFree

There are no other callers.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

conf: Introduce VIR_STORAGE_NET_HOST_TRANS_FD

Prepare the parser code and anything using 'virStorageNetHostTransport'
to support passing a FD instead of opening the connection by qemu
itself.

For now this just prepares the parser and data structures, but the code
is dormant.

Only code paths which will actually support FD passing will then enable
it in the future.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Spellchecked-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

virDomainStorageNetworkParseHost: Refactor cleanup

Use a 'switch' statement instead of a bunch of if statements to do
validation and selection what to parse.

Remove the pre-clearing of the struct as we always allocate cleared
memory for it and we can reorder assignments to avoid the need for
cleanup.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

virDomainStorageNetworkParseHost: Remove unpopulated 'transport' variable

Since the refactor to use proper enum type for the network transport the
'transport' variable is no longer filled. Remove it and fix the error
message which references it without using NULLSTR.

Fixes: 452695926dc
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

virDomainDiskDefValidateSourceChainOne: Fix validation of 'data-file' nesting

As the 'dataStore' is internally represented as a virStorageSource
object it has provisions for nesting which is not supported.

When I've reviewed and modified the commit adding data file parsing
support I've added code that was supposed to reject any 'backingStore'
and 'dataStore' structures nested in a source of a 'dataStore'.

Unfortunately the check was broken as one of the terms checked the
presence of parent's 'backingStore' instead of the nesting.

Fix it and add tests.

Fixes: b3171cf8da3
Resolves: https://issues.redhat.com/browse/RHEL-85320
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

esxConnectListAllDomains: Don't propagate failure to lookup a single domain

In esxConnectListAllDomains if the lookup of the VM name and UUID fails
for a single VM (possible e.g. with broken storage) the whole API would
return failure even when there are working VMs.

Rework the lookup so that if a subset fails we ignore the failure on
those. We report an error only if lookup of all of the objects failed.
Failure is reported from the last one.

Resolves: https://issues.redhat.com/browse/RHEL-80606
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

news: document dropping apparmor < 3.0.0

Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

meson: drop remaining checks for apparmor version

Now that we mandate version 3, any remaining conditional checks
in meson/source code can be removed.

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

Revert "apparmor: Allow version-specific bits in profiles"

This reverts commit 19eb8abc9a4d15190852d644b773a2348f11c9da.

There is no longer any need to dynamically generate version specific
rules. This revert can be reverted, if the need ever arises again
in the future.

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

Revert "apparmor: Allow version-specific bits in abstractions too"

This reverts commit 63a312fa2d3be0e34a8989deddd39792fc9badf6.

There is no longer any need to dynamically generate version specific
rules. This revert can be reverted, if the need ever arises again
in the future.

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

apparmor: assume at least apparmor >= 3

By assuming version 3, we can drop all the conditional version
substitutions from the profiles.

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

meson: mandate apparmor >= 3.0.0

We can now assume at least version three:

 * Debian 12: 3.0.8
 * openSUSE Leap 15.5: 3.0.4
 * openSUSE Leap 15.6: 3.1.7
 * Ubuntu 22.04: 3.0.4
 * Ubuntu 24.04: 4.0.0

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

Post-release version bump to 11.3.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

Release of libvirt-11.2.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

Translated using Weblate (Korean)

Currently translated at 98.6% (10775 of 10917 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/ko/

Signed-off-by: 김인수 <simmon@nplob.com>

Translated using Weblate (English (United Kingdom))

Currently translated at 48.2% (5270 of 10917 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/en_GB/

Signed-off-by: Andi Chandler <andi@gowling.com>

NEWS: qemu: support block disk along with throttle filters

Signed-off-by: Harikumar Rajkumar <harirajkumar230@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

NEWS: Document features/improvements/bug fixes I've participated in

There are some features/improvements/bug fixes I've either
contributed or reviewed/merged. Document them for upcoming
release.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

NEWS: Mention bump to qemu-6.2, vhost-user shmem check and stats params discoverabilty

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

doc: remove wrong comment

This comment is wrong as later qemuMigrationSrcRun() is called which
checks if TLS should be used and activated. QEMU has built-in support
for TLS, which this refers to.

The comment originates from a time when tunneled support was the only
way to get encryption.

Signed-off-by: Philipp Schuster <philipp.schuster@cyberus-technology.de>

lib: virDomain{Save,Restore}Params: Ensure absolute path

When invoking virDomainSaveParams with a relative path, the image is
saved to the daemon's CWD. Similarly, when providing virDomainRestoreParams
with a relative path, it attempts to restore from the daemon's CWD. In most
configurations, the daemon's CWD is set to '/'. Ensure a relative path is
converted to absolute before invoking the driver domain{Save,Restore}Params
functions.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

Update translation files

Updated by "Update PO files to match POT (msgmerge)" hook in Weblate.

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/

Signed-off-by: Fedora Weblate Translation <i18n@lists.fedoraproject.org>

qemu_driver: Fix virDomainSaveImageDefineXML

Commit 28a06215280b99708ed8dc2d183f62ba7b34ccf8 added support to restore
sparse images but changed the boolean that controls if we open the file
as read-only or read-write. Editing XML in the save image resulted in
following error message:

    failed to write header to domain save file '/data/images/fedora40.save': Bad file descriptor

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

po: Refresh potfile for v11.2.0

Signed-off-by: Jiri Denemark <jdenemar@redhat.com>

qemu: Emit NIC_MAC_CHANGE event

So far, we only process NIC_RX_FILTER_CHANGED event when the
corresponding device has 'trustGuestRxFilters' enabled. And the
event is emitted only for virtio model. IOW, this is fairly
limited situation and other scenarios don't emit any event (e.g.
change of MAC address on a PCI passthrough device).

Resolves: https://issues.redhat.com/browse/RHEL-7035
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

Introduce NIC_MAC_CHANGE event

The aim off this event is to notify management application that
guest changed MAC address on one of its vNICs so the app can
update its internal records, e.g. for finding match between
guest/host view of vNICs.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

qemu: Reflect MAC address change in live domain XML

If a guest changes MAC address on its vNIC, then QEMU emits
NIC_RX_FILTER_CHANGED event (the event is emitted in other cases
too, but that's not important right now). Now, domain XML allows
users to chose whether to trust these events or not:

  <interface trustGuestRxFilters='yes|no'/>

For the 'no' case no action is performed and the event is
ignored. But for the 'yes' case, some host side features of
corresponding vNIC (well tap/macvtap device) are tweaked to
reflect changed MAC address. But what is missing is reflecting
this new MAC address in domain XML.

Basically, what happens is: the host sees traffic with new MAC
address, all tools inside the guest see the new MAC address
(including 'virsh domifaddr --source agent') which makes it
harder to match device in the guest with the one in the domain
XML.

Therefore, report this new MAC address as another attribute of
the <mac/> element:

  <mac address="52:54:00:a4:6f:91" currentAddress="00:11:22:33:44:55"/>

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

NEWS: Mention virsh hypervisor-cpu-models command

Signed-off-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>

Translated using Weblate (Spanish)

Currently translated at 52.6% (5708 of 10838 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/es/

Signed-off-by: Nicolás Gal <nialegal@yandex.com>

NEWS: Mention guest agent error code improvements

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemuAgentCheckError: Rewort error if neither return nor error is found

Disambiguate the case from other types of error.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemuAgentCheckError: Use 'VIR_ERR_AGENT_COMMAND_FAILED'

In the two cases when we know that the command returned failure switch
to the new error code so that management applications can
programatically detect failure of the guest agent command.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

lib: error: Introduce 'VIR_ERR_AGENT_COMMAND_FAILED'

Add a special error code for when the guest agent returned a failure
message.

Allow management applications to deterministically detect failure of the
guest agent command.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

docs: Point to VIR_ERR_AGENT_COMMAND_TIMEOUT when setting timeout

In addition to the error constant appearing add docs hinting that this
new error code can be produced on timeouts.

The most relevant place is to do it when setting the timeout.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemuAgentCommandFull: Use VIR_ERR_AGENT_COMMAND_TIMEOUT when agent disappears

When the agent disappears after geting a proper command we ought to
report the same error code as if we timed out as it's uncertain whether
the guest agent did anything.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: agent: Differentiate timeouts when syncing from command timeout

As the guest agent code uses timeouts it is possible that we stop
waiting before the guest agent replies. If this happens while syncing
everything is okay because we didn't send any state-changing command.

In case when the timeout happens after a real command was transmitted
it's unknown if the guest-agent processed it or not.

Use the new special error code VIR_ERR_AGENT_COMMAND_TIMEOUT for cases
when we sent non-sync commands, so that the management applications or
users have possibility to react to this situation.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

lib: error: Introduce 'VIR_ERR_AGENT_COMMAND_TIMEOUT'

Introduce a new special error code for guest agent commands.

The error code will be specifically reported only when an actual command
(not a sync) was issued to the guest agent and the timeout time was
reached.

This will allow users and management applications to differentiate
between the cases when the sync timed out and thus there's no risk in
the agent actually having executed the command and when the actual
command was sent.

Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

NEWS: fix one more typo

s/ an / a /

Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reported-by: Jiri Denemark <jdenemar@redhat.com>

NEWS: hotplug virtio-mem device

Signed-off-by: Boris Fiuczynski <fiuczy@linux.ibm.com>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>

Translated using Weblate (Chinese (Traditional) (zh_TW))

Currently translated at 2.4% (264 of 10838 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/zh_TW/

Signed-off-by: hsu zangmen <chzang55@gmail.com>

Translated using Weblate (English (United Kingdom))

Currently translated at 48.5% (5262 of 10838 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/en_GB/

Signed-off-by: Andi Chandler <andi@gowling.com>

Translated using Weblate (Russian)

Currently translated at 83.8% (9086 of 10838 strings)

Translation: libvirt/libvirt
Translate-URL: https://translate.fedoraproject.org/projects/libvirt/libvirt/ru/

Signed-off-by: "Sergey A." <Ser82-png@yandex.ru>

NEWS: Mention support for mapped-ram

Add a 'New features' entry for mapped-ram itself, and another
for the parallel save/restore feature built on top.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

qemu: Fix build on 32-bit platforms

Commits c2518f7bc7 and 28a0621528 introduced build failures on 32-bit
platforms by using incorrect format specifiers with g_strdup_printf.

In one case, an 'unsigned long' format specifier is used with a
'long long int' variable. Fix by changing the format specifier to
'uintmax_t', and casting the variable likewise.

In a second case, an 'unsigned long' format specifier is used with a
'size_t' variable, which is 'unsigned int' on 32-bit systems. Fix by
changing the format specifier to use the 'z' modifier.

Fixes: c2518f7bc7dd4f8ab8655a12ec3a000e1eb5b232
Fixes: 28a06215280b99708ed8dc2d183f62ba7b34ccf8
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

virsh: add --image-format option to the save command

Option --parallel-channels would require changing configuration file to
be used so introduce this option as well to make it convenient for
users.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

tools: use virDomainSaveParams only when necessary

We should use the newest API only when user sets parallel-channels.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

tools: use virDomainRestoreParams only when necessary

We should use the newest API only when user sets parallel-channels.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

qemu: remove VIR_DOMAIN_SAVE_PARALLEL flag

There is no need to use extra flag in addition to the new
"parallel.channels" param.

Using the flag without param would result in using uninitialized
variable. Fixing it would result in error that parallel channels cannot
be less then 1 or setting 1 as default.

Using the param without the flag is ignored.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

tools: remote --parallel from virsh save command

There is no need to have --parallel and --parallel-channels especially
when --parallel on its own is the same as not used at all. In both cases
libvirt will default to single channel.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

tools: remove --parallel from virsh restore command

There is no need to have --parallel and --parallel-channels especially
when --parallel on its own is the same as not used at all. In both cases
libvirt will default to single channel.

Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Jim Fehlig <jfehlig@suse.com>

NEWS: fix even more typos

The 'editting one' was:
Reported-by: Martin Kletzander <mkletzan@redhat.com>
I found the rest myself.

Signed-off-by: Ján Tomko <jtomko@redhat.com>

qemu/dbus: Allow connections from root to the dbus-daemon

In commit dbfb96d18c04 libvirt started connecting to the daemon to set
RDP credentials, but our configuration file did not allow connections
from the root user, so the connection failed and the VM failed to start.

In order to avoid such issue allow root to connect if the daemon is
running privileged.

Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

Fix typo in NEWS files

Reported-by: Laine Stump <laine@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

Add NEWS item for auto-shutdown feature

Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

tools: add parallel parameter to virsh restore command

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

tools: add parallel parameter to virsh save command

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Add support for parallel save and restore

Add support for parallel save and restore by mapping libvirt's
"parallel-channels" parameter to QEMU's "multifd-channels"
migration parameter.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

include: Define constants for parallel save/restore

Add a new VIR_DOMAIN_SAVE_PARALLEL flag to the save and restore APIs,
which can be used to specify the use of multiple, parallel channels
for saving and restoring a domain. The number of parallel channels
can be set using the VIR_DOMAIN_SAVE_PARAM_PARALLEL_CHANNELS
typed parameter.

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Support O_DIRECT with mapped-ram on restore

When using the mapped-ram migration capability, direct IO is
enabled by setting the "direct-io" migration parameter to
"true" and passing QEMU an additional fd with O_DIRECT set.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Support O_DIRECT with mapped-ram on save

When using the mapped-ram migration capability, direct IO is
enabled by setting the "direct-io" migration parameter to
"true" and passing QEMU an additional fd with O_DIRECT set.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Add support for mapped-ram on restore

Add support for the mapped-ram migration capability on restore.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Apply migration parameters in qemuMigrationDstRun

Similar to qemuMigrationSrcRun, apply migration parameters in
qemuMigrationDstRun. This allows callers to create customized
migration parameters, but delegates their application to the
function performing the migration.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Move creation of qemuProcessIncomingDef struct

qemuProcessStartWithMemoryState() is the only caller of qemuProcessStart()
that uses the qemuProcessIncomingDef struct. Move creation of the struct
to qemuProcessStartWithMemoryState().

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Add support for mapped-ram on save

Introduce support for QEMU's new mapped-ram stream format [1].
mapped-ram can be enabled by setting the 'save_image_format'
setting in qemu.conf to 'sparse'.

To use mapped-ram with QEMU:
- The 'mapped-ram' migration capability must be set to true
- The 'multifd' migration capability must be set to true and
  the 'multifd-channels' migration parameter must set to 1
- QEMU must be provided an fdset containing the migration fd
- The 'migrate' qmp command is invoked with a URI referencing the
  fdset and an offset where to start reading or writing the data
  stream, e.g.

  {"execute":"migrate",
   "arguments":{"detach":true,"resume":false,
                "uri":"file:/dev/fdset/0,offset=0x11921"}}

The mapped-ram stream, in conjunction with direct IO and multifd
support provided by subsequent patches, can significantly improve
the time required to save VM memory state. The following tables
compare mapped-ram with the existing, sequential save stream. In
all cases, the save and restore operations are to/from a block
device comprised of two NVMe disks in RAID0 configuration with
xfs (~8600MiB/s). The values in the 'save time' and 'restore time'
columns were scraped from the 'real' time reported by time(1). The
'Size' and 'Blocks' columns were provided by the corresponding
outputs of stat(1).

VM: 32G RAM, 1 vcpu, idle (shortly after boot)

                       | save    | restore |
       | time    | time    | Size         | Blocks
-----------------------+---------+---------+--------------+--------
legacy                 | 6.193s  | 4.399s  | 985744812    | 1925288
-----------------------+---------+---------+--------------+--------
mapped-ram             | 5.109s  | 1.176s  | 34368554354  | 1774472
-----------------------+---------+---------+--------------+--------
legacy + direct IO     | 5.725s  | 4.512s  | 985765251    | 1925328
-----------------------+---------+---------+--------------+--------
mapped-ram + direct IO | 4.627s  | 1.490s  | 34368554354  | 1774304
-----------------------+---------+---------+--------------+--------
mapped-ram + direct IO |         |         |              |
 + multifd-channels=8  | 4.421s  | 0.845s  | 34368554318  | 1774312
-------------------------------------------------------------------

VM: 32G RAM, 30G dirty, 1 vcpu in tight loop dirtying memory

                       | save    | restore |
       | time    | time    | Size         | Blocks
-----------------------+---------+---------+--------------+---------
legacy                 | 25.800s | 14.332s | 33154309983  | 64754512
-----------------------+---------+---------+--------------+---------
mapped-ram             | 18.742s | 15.027s | 34368559228  | 64617160
-----------------------+---------+---------+--------------+---------
legacy + direct IO     | 13.115s | 18.050s | 33154310496  | 64754520
-----------------------+---------+---------+--------------+---------
mapped-ram + direct IO | 13.623s | 15.959s | 34368557392  | 64662040
-----------------------+-------- +---------+--------------+---------
mapped-ram + direct IO |         |         |              |
 + multifd-channels=8  | 6.994s  | 6.470s  | 34368554980  | 64665776
--------------------------------------------------------------------

As can be seen from the tables, one caveat of mapped-ram is the logical
file size of a saved image is basically equivalent to the VM memory size.
Note however that mapped-ram typically uses fewer blocks on disk, hence
the name 'sparse' for 'save_image_format'.

Also note the mapped-ram stream is incompatible with the existing stream
format, hence mapped-ram cannot be used to restore an image saved with
the existing format and vice versa.

[1] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/migration/mapped-ram.rst?ref_type=heads

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: Add helper function for creating save image fd

Move the code in qemuSaveImageCreate that opens, labels, and wraps the
save image fd to a helper function, providing more flexibility for
upcoming mapped-ram support.

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

qemu_saveimage: add "sparse" to supported save image formats

Extend the list of formats to include "sparse", which uses QEMU's mapped-ram
stream format [1] to write guest memory blocks at fixed offsets in the save
image file.

[1] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/migration/mapped-ram.rst?ref_type=heads

Signed-off-by: Jim Fehlig <jfehlig@suse.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>