Peter Krempa [Tue, 10 Aug 2021 14:28:15 +0000 (16:28 +0200)]
qemu: capabilities: Add alternative detection of QEMU_CAPS_NUMA
'set-numa-node' is the command which can set the equivalent parameters
to '-numa' in preconfig mode, so we can use it as witness to see that
-numa is supported.
To ensure that the old detection method is removed once we'll be bumping
qemu support add a comment with the appropriate version check.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
They are no longer used as we now assume that all tuning caps are
present and in case some will be removed we'll need to use different
probing methods.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Tue, 10 Aug 2021 12:56:31 +0000 (14:56 +0200)]
qemu: Always assume that qemu supports drive throttling
All currently supported qemu versions support all throttling
capabilities. It is unlikely that any of the fields will be removed in
the future and if it will we will need to do specific probing which is
possible via the 'throttle' object which is the replacement for the
legacy way to configure throttling.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Mon, 26 Jul 2021 12:03:07 +0000 (14:03 +0200)]
qemu: capabilities: Probe 'discard' and 'detect-zeroes' from 'blockdev-add'
Probing QEMU_CAPS_DRIVE_DISCARD and QEMU_CAPS_DRIVE_DETECT_ZEROES can be
replaced by looking into the QMP schema rather than looking at -drive
which isn't in use any more.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Mon, 26 Jul 2021 12:03:07 +0000 (14:03 +0200)]
qemu: capabilities: Probe 'reconnect', 'logfile' and 'append' from 'chardev-add'
QEMU_CAPS_CHARDEV_RECONNECT, QEMU_CAPS_CHARDEV_LOGFILE and
QEMU_CAPS_CHARDEV_FILE_APPEND can be probed from the appropriate fields
in 'chardev-add' probed via the QMP schema instead of the command line
parameters.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Tue, 27 Jul 2021 11:14:51 +0000 (13:14 +0200)]
scripts: group-qemu-caps: Add capability name as comment to capability string
Add a cross reference of the enum value name with the string
representation. This allows a quick cross-reference of the values
without having to open the header and implementation files separately.
To achieve this the checker code at first obtains a list of the
flags and cross-references them when checking the grouping in
syntax-check, thus we are guaranteed to stay in sync.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Fri, 6 Aug 2021 07:43:47 +0000 (09:43 +0200)]
virQEMUCapsCommandLine: Group by commandline argument name alphabetically
Further commits will be refactoring and minimizing capabilities being
parsed from 'query-command-line-options'. Group the struct driving the
detection by argument name so it's easier to spot options belonging
together.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Pavel Hrdina [Mon, 2 Aug 2021 18:26:57 +0000 (20:26 +0200)]
qemu: implement support for Fibre Channel VMID
Based on kernel commit messages the interface is
/sys/class/fc/fc_udev_device/appid_store
where we need to write the following string "$INODE:$APPID".
$INODE is the VM root cgroup inode in hexadecimal and $APPID is user
provided string that will be attached to each FC frame for the VM
within the cgroup identified by inode and has limit 128 bytes.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Pavel Hrdina [Thu, 5 Aug 2021 13:27:17 +0000 (15:27 +0200)]
conf: refactor virDomainResourceDefParse
There is no need to error out for empty <partition></partition> element
as we can just simply ignore it. This allows to simplify the function
and prepare it for new sub-elements of <resource>.
It makes the <partition> element optional so we need to reflect the
change in schema as well.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Michal Privoznik [Thu, 29 Jul 2021 14:09:58 +0000 (16:09 +0200)]
vmx: Support super wide SCSI bus
Since its 6.7 release, vSphere allows up to 64 units on a SCSI
bus [1]. The release version translates to virtualHW_version 13
and thus if we are dealing with sufficiently new version we can
enable the feature.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1738392 Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Libvirt assumes that a SCSI bus can fit up to 8 devices
(including controller itself), except for so called wide bus
which can accommodate up to 16 devices (again, including
controller). This plays important role when computing 'drive'
address in virDomainDiskDefAssignAddress(). So far, the only
driver that enables wide SCSI bus is VMX. But with newer
releases, ESX is capable of "super wide" bus (64 devices).
We can blindly bump the limit in our code because then we would
compute address that's invalid for older ESX versions that we
still want to support.
Unfortunately, I haven't found a better place where to store this
than virDomainDef.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Michal Privoznik [Thu, 29 Jul 2021 13:39:08 +0000 (15:39 +0200)]
vmx: Rework disk def allocation
The way we parse VMX configuration is rather unfortunate,
especially when it comes to disks. We allocate an array that can
handle all possible disks but leave the array counter (ndisks) at
zero and increase it only after successful parsing. But, we never
size the array down to release unneeded chunks of memory.
We can do better: we can use VIR_APPEND_ELEMENT() to allocate
array as needed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Michal Privoznik [Fri, 30 Jul 2021 09:06:04 +0000 (11:06 +0200)]
vmx: Fill virtualHW.version to ESX version mapping
At the beginning of vmx.c we have a comment that maps
virtualHW.version field onto ESX version. However, it wasn't
updated in a while. Fill it in using the following kbase article:
If the QEMU driver is configured to use the old "file" stdio
handler (meaning virtlogd is out of the picture) and a chardev
has a log file configured we rely on QEMU being able to create
the file itself. This may not be always possible (e.g. if the
logfile is set to a directory that QEMU process can't reach).
In such case we should create the file and just pass its FD to
QEMU.
We could do that unconditionally and just either pass FD from
virtlogd or the one we opened, because we bumped QEMU version
and are now requiring new enough QEMU. However, I'm keeping the
old style where logfile is appended on the cmd line for the tests
sake.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1989457 Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com> Tested-by: Ján Tomko <jtomko@redhat.com>
The function doesn't write to domain definition really so make
@def argument as const. This allows us to call it from functions
where the domain definition is already const.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com> Tested-by: Ján Tomko <jtomko@redhat.com>
util: add xml validation against schema in virXMLParseHelper()
We need this in order to validate XML against schema at one
place, rather than have the same code for validation in different
functions.
I will add '--validate' option to more virsh commands soon and
this makes it easier as virXMLParse() is called in every one I
plan to change.
Signed-off-by: Kristina Hanicova <khanicov@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com> Signed-off-by: Ján Tomko <jtomko@redhat.com>
Ján Tomko [Wed, 11 Aug 2021 14:04:45 +0000 (16:04 +0200)]
virsh: refactor cmdDomblkinfo
Use automatic memory cleanup to get rid of the cleanup section,
and of the memory leak that happens inside the loop, because
cap, alloc and phy are only freed once per function.
Signed-off-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
qemu: save status xml after generating taint message
We didn't always save status xml after generating new taint message
which resulted in it being deleted in case of a libvirtd restart.
Some taint messages were preserved thanks to saving status xml
separately at the end of the calling functions. With this, every taint
message is saved, regardless of the calling functions.
Turn 'mounts' into a proper GStrv after sorting so that automatic
cleanup can be used and shuffle around the cleanup steps so that jumps
can be avoided in favor of direct return of error code.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Thu, 5 Aug 2021 15:00:45 +0000 (17:00 +0200)]
qemuNamespacePrepareOneItem: Restructure code to avoid temporary variables
The value of 'next' is copied into 'item.file' so we can move the update
to the 'next' pointer earlier and move the VIR_APPEND_ELEMENT call to
where we figure out that we need to append the value.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Mon, 14 Jun 2021 12:56:15 +0000 (14:56 +0200)]
remote: dispatch: Don't use virStringListFreeCount for NULL terminated lists
Both virDomainAuthorizedSSHKeysGet and virDomainGetMessages return a
NULL-terminated string-list, so we can use g_auto(GStrv) to clear the
used memory on failures.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Wed, 4 Aug 2021 15:37:44 +0000 (17:37 +0200)]
virPCIGetVirtualFunctions: Fetch also network interface name if needed
'virNetDevGetVirtualFunctions' calls 'virPCIGetVirtualFunctions' and
then re-iterates the returned list to fetch the interface names for the
returned virtual functions.
If we move the fetching of the interface name into
virPCIGetVirtualFunctions we can simplify the code and remove a bunch of
impossible error states.
To accomplish this the function is renamed to
'virPCIGetVirtualFunctionsFull' while keeping a wrapper with original
name and if the physical port ID is passed the interface name is fetched
too without the need to re-convert the address into a sysfs link.
For now 'virNetDevGetVirtualFunctions' still converts the returned data
into two lists.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Tue, 3 Aug 2021 14:45:36 +0000 (16:45 +0200)]
virPCIGetVirtualFunctions: Simplify cleanup of returned data
Introduce a struct for holding the list of VFs returned by
virPCIGetVirtualFunctions so that we can employ automatic memory
clearing and also allow querying more information at once.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
projects / thirdparty / libvirt.git / log
