Peter Krempa [Thu, 31 Aug 2023 11:47:39 +0000 (13:47 +0200)]
qemuxml2argvtest: testAddCPUModels: Remove non-x86_64 fake cpu models
At this point only x86_64 is using fake machines, and for real machines
we don't populate the fake cpu models. Thus we can remove everything
non-x86_64.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Fri, 1 Sep 2023 13:41:10 +0000 (15:41 +0200)]
qemuxml2argvtest: Clean up host CPU if overriden
If a test uses ARG_CAPS_HOST_CPU_MODEL feature we override the global
host cpu model to the selected CPU but don't clear it afterwards. This
can trip up fake caps tests following a test which uses this feature.
This does not happen with real caps, because unless overriden, the host
cpu from capabilities is always populated as the host cpu.
Clear the CPU on cleanup.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Fri, 1 Sep 2023 19:43:22 +0000 (21:43 +0200)]
qemu: capabilities: Preserve location of the default machine type when stripping aliases for tests
Internally the preferred machine which is 'pc' for x86_64 must be kept
in the first place in the array of machines. This was not the case when
stripping the machine aliases for use in tests (so that test output
stays stable) where we've created a new entry for the alias. This means
that the original name (e.g. pc-i440fx-8.1) stayed in the first place.
To fix this we now swap the names around and create a new entry at the
end for the specific type. Additionally the default flag is not
propagated to the copy.
This is also visible now in the output of 'qemuxml2xmltest' as the test
cases which use the default machine type return to 'pc' instead of the
more specific name.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Peter Krempa [Fri, 1 Sep 2023 14:57:34 +0000 (16:57 +0200)]
qemu: Always assume support for QEMU_CAPS_PIIX_DISABLE_S* and QEMU_CAPS_ICH9_DISABLE_S*
The support for PIIX power management was added in qemu commit v1.0-3094-g459ae5ea5a and the suport for ICH9 power management was added
in qemu commit v2.2.0-542-g6ac0d8d44c and both can't be compiled out.
This means we can always assume support for these features. Remove the
validation and impossible tests. Move relevant bits from
'q35-pm-disable' to 'q35' test case.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
The sole purpose of getDeviceType() is to parse a file that
contains one integer (and a newline character). Well, we already
have a function for that: virFileReadValueInt(). Use the latter
and drop the former.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
ch: Use proper format directive for @i in virCHProcessSetupIOThreads()
The @i variable inside of virCHProcessSetupIOThreads() is a
typical loop counter - it's declared as size_t. But when passed
to VIR_DEBUG an invalid format directive is used (%ld). Fix that.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
ch: Use int for @niothreads in virCHMonitorGetIOThreads()
The @niothreads inside of virCHMonitorGetIOThreads() is declared
as of size_t type. This would work, except the variable is then
passed to VIR_DEBUG with incorrect format directive (%ld) and
returned. But the function returns an int not size_t. Fix the
variable declaration and format directive.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
ch: Use uint for @maxvcpus in virCHDomainRefreshThreadInfo()
The @maxvcpus variable inside of virCHDomainRefreshThreadInfo()
holds retval of virDomainDefGetVcpusMax() which returns an
unsigned int. Also, the variable is then passed to VIR_WARN()
with incorrect format directive (%ld). Switch variable to uint
and fix the format directive.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewing the sources, I noticed that, argumets order in
virCgroupGetMemoryStat() function call does not correspond
to the function declaration:
-instead of *activeAnon, &meminfo->inactive_anon is passed;
-instead of *inactiveAnon, &meminfo->active_anon is passed;
-instead of *activeFile, &meminfo->inactive_file is passed;
-instead of *inactiveFile, &meminfo->active_file is passed.
Fixes: e634c7cd0d ("lxc: Use virCgroupGetMemoryStat") Signed-off-by: Dmitry Frolov <frolov@swemel.ru> Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
conf: Generate MAC address instead of keeping all zeroes
When we parse <mac address="00:00:00:00:00:00"/> we keep that in memory
and pass it down to the hypervisor. However, that MAC address is not
strictly valid as it is not marked as locally administered (bit 0x02)
but it is not even globally unique. It is also used for loopback device
on Linux, for example. And QEMU sees such MAC address just as "not
specified" and generates a new one that libvirt does not even know
about. So to make the overall experience better we now generate it if
the supplied one is all clear.
Peter Krempa [Mon, 4 Sep 2023 12:30:48 +0000 (14:30 +0200)]
build: Fix logic bug determining whether running with optimization
The conversion from ternary to a 'if' clause was wrong and thus didn't
properly increase the stack size where needed but only where not
actually needed.
Fixes: b68faa99d9f16c2f504b23737040d25d072ee85d Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Peter Krempa [Mon, 4 Sep 2023 12:07:31 +0000 (14:07 +0200)]
build: Fix assignment into 'stack_frame_size' when sanitizer is enabled
Instead of an assignment into the 'stack_frame_size' variable when
sanitizers are enabled I've accidentally compared the value against the
requested size.
Fix the typo.
Fixes: b68faa99d9f16c2f504b23737040d25d072ee85d Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Peter Krempa [Mon, 4 Sep 2023 10:10:25 +0000 (12:10 +0200)]
build: Work around clang's stack size calculation without optimization
When building without optimization on clang, certain big functions trip
the stack size limit despite not actually reaching it. Relax the stack
limit size for clang without optimization.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Peter Krempa [Mon, 4 Sep 2023 10:09:11 +0000 (12:09 +0200)]
docs: compiling: Add a note about use of CFLAGS for optimization
Meson doesn't interpret what's set in CFLAGS, but rather simply appeds
it to the command line. Thus any logic which is based on the
optimization level will not work.
Note the caveat in the docs and instruct users to use
``--optimization=N`` instead.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Michal Privoznik [Mon, 28 Aug 2023 10:56:36 +0000 (12:56 +0200)]
syntax-check: Introduce a rule for one line error messages
Okay, this is a shortcut. Our coding style says that error
messages are exempt from '80 chars long lines' rule. But in the
very same paragraph it is said that all error messages need to be
marked for translation (as they might be presented to user).
Therefore, the syntax-check rule can check if _("...") is
formatted on one line. With exception of _("...\n" ...) (e.g.
various outputs from helper binaries like leaseshelper,
sshhelper, or daemons like lockd, logd). I believe nobody would
chose a substring that contains '\n' for git grep-ping the error
message.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Michal Privoznik [Mon, 28 Aug 2023 11:07:34 +0000 (13:07 +0200)]
tools: Reformat --help output of virsh and virt-admin
The --help output of virsh and virt-admin shows supported options
and commands and as such contains new lines. Both these strings
are marked for translation btw. But the way they are formatted
now ('\n' being at the start of new line instead at the end of
the previous) makes it hard to create a syntax-check rule for
'translation message on one line' (next commit).
Reformat both strings a bit (no user visible change though).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Michal Privoznik [Tue, 29 Aug 2023 14:25:38 +0000 (16:25 +0200)]
bhyve: Feed hook scripts with domain XML
Domain related hook scripts are all fed with domain XML on their
stdin, except for bhyve. Fix this.
Resolves: https://gitlab.com/libvirt/libvirt/-/issues/528 Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Michal Privoznik [Tue, 29 Aug 2023 14:13:26 +0000 (16:13 +0200)]
docs: Document that libxl hooks are also given full domain XML
Our hooks.rst document existence of libxl hook scripts, but
mentions only qemu and lxc as receivers of full domain XML. But
since their introduction in v2.2.0-rc1~201 they are also given
full domain XML. Fix our wording.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Change example logfile path and clarify how complicated all things passt
are. I chose not to create the non-existing directory because it could
open a whole new can of worms.
Also explain missing `dev` attribute of `<portForward/>`
Peter Krempa [Tue, 20 Sep 2022 11:16:58 +0000 (13:16 +0200)]
nss: aiforaf: Decrease stack size by scoping off large buffers
The 'buf', 'sa' and 'hints' stack allocated helper variables are never
used together. Decrease the stack memory usage by scoping them off into
do-while blocks.
In this instance we do not want to use dynamic allocation as this is the
NSS module.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Kristina Hanicova <khanicov@redhat.com>
Peter Krempa [Wed, 30 Aug 2023 10:19:23 +0000 (12:19 +0200)]
nss: aiforaf: Drop unused buffer 'port'
The 'port' buffer is passed to 'getnameinfo' which is supposed to fill
it but it's not actually later used. Drop the buffer as 'getnameinfo'
allows NULL arguments if they are not needed.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Kristina Hanicova <khanicov@redhat.com>
Peter Krempa [Mon, 28 Aug 2023 12:14:32 +0000 (14:14 +0200)]
qemuDomainSupportsVcpuHotplug: Base return value on virQEMUCapsGetMachineHotplugCpus
The QEMU_CAPS_QUERY_HOTPLUGGABLE_CPUS flag is always asserted as all
qemu versions support the command and selectively cleared when copying
the capabilities for VM use if given machine type does not support cpu
hotplug.
Rework this to directly probe the machine as we now populate the data
also when re-connecting to a qemu instance after daemon restart, so that
the capability can be removed.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Mon, 28 Aug 2023 09:35:07 +0000 (11:35 +0200)]
qemu: process: Probe machine type data on reconnect to qemu
When reconnecting we populate only the capability flags from the XML as
we need to know the exact flags that were present when starting the VM.
On the other hand the machine type data is not stored as it wasn't
really used after startup. While storing all of the data into the status
XML would be theoretically possible, with machine-type specific data it
makes no sense to do so, and thus the data can be re-probed from the
current instance.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Mon, 28 Aug 2023 10:56:40 +0000 (12:56 +0200)]
qemu: capabilities: Export functions necessary for probing machine types
Upcoming patch will re-probe machines from the current qemu instance to
populate the private copy of qemuCaps after reconnecting to a running
instance. This is needed to be able to access the machine type data,
while storing them in the status XML seems to be an overkill, for
information which can be easily reprobed.
Export 'virQEMUCapsInitQMPArch' needed to populate the 'arch' field and
'virQEMUCapsProbeQMPMachineTypes'.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Peter Krempa [Thu, 20 Jul 2023 15:55:33 +0000 (17:55 +0200)]
qemu: Rename qemuDomainSupportsNewVcpuHotplug to qemuDomainSupportsVcpuHotplug
Support for legacy cpu hotplug was removed a long time ago. At this
point this function only checks whether the current machine type
supports cpu hotplug.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
conf, schema: Switch iothread/poll values to unsignedLong
They represent nanoseconds, and we accept such values already. Not that
anyone would use such values in the wild, but even one person testing
QEMU could put in a bigger value and will be bothered with validation
errors after every `virsh edit`. Also add a test for it.
Erik Skultety [Mon, 28 Aug 2023 08:47:32 +0000 (10:47 +0200)]
ci: lcitool: Add libvirt-tck+runtime deps list
This change was supposed to be part of commit 120a674f , but was
proposed against the libvirt TCK project instead. Since we're running
the TCK test suite as part of this project, this is the right place for
the TCK runtime deps list config.
Signed-off-by: Erik Skultety <eskultet@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
K Shiva Kiran [Thu, 31 Aug 2023 18:16:57 +0000 (23:46 +0530)]
virsh: Fix net-desc --config output
Fixes the following bug:
Command: `net-desc --config [--title] my_network`
Expected Output: Title/Description of persistent config
Output: Title/Description of live config
This was caused due to the usage of a single `flags` variable in
`virshGetNetworkDescription()` which ended up in a wrong enum being
passed to `virNetworkGetMetadata()` (enum being that of LIVE instead of
CONFIG).
Although the domain object has the same code, this didn't cause a problem
there because the enum values of `VIR_DOMAIN_INACTIVE_XML` and
`VIR_DOMAIN_METADATA_CONFIG` turn out to be the same (1 << 1), whereas
they are not for network equivalent ones (1 << 0, 1 << 1).
Signed-off-by: K Shiva Kiran <shiva_kr@riseup.net> Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Andrea Bolognani [Wed, 30 Aug 2023 15:45:47 +0000 (17:45 +0200)]
rpm: Recommend libvirt-daemon for with_modular_daemons distros
A default deployment on modern distros uses modular daemons but
switching back to the monolithic daemon, while not recommended,
is still considered a perfectly valid option.
For a monolithic daemon deployment, the upgrade to libvirt 9.2.0
or newer works as expected; a subsequent call to dnf autoremove,
however, results in the libvirt-daemon package being removed and
the deployment no longer working.
In order to avoid that situation, mark the libvirt-daemon as
recommended.
This will unfortunately result in it being included in most
installations despite not being necessary, but considering that
the alternative is breaking existing setups on upgrade it feels
like a reasonable tradeoff.
Moreover, since the dependency on libvirt-daemon is just a weak
one, it's still possible for people looking to minimize the
footprint of their installation to manually remove the package
after installation, mitigating the drawbacks of this approach.
Signed-off-by: Andrea Bolognani <abologna@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Andrea Bolognani [Wed, 30 Aug 2023 15:41:14 +0000 (17:41 +0200)]
rpm: Fix typo in daemon name
The name of the virtsecretd daemon was misspelled, resulting
in multiple errors during installation:
Running scriptlet: libvirt-daemon-driver-secret-9.5.0-6.el9.x86_64
Failed to preset unit: Unit file virsecretd.socket does not exist.
Failed to preset unit: Unit file virsecretd-ro.socket does not exist.
Failed to preset unit: Unit file virsecretd-admin.socket does not exist.
Failed to preset unit: Unit file virsecretd.service does not exist.
Laura Hild [Tue, 15 Aug 2023 14:54:20 +0000 (10:54 -0400)]
Don't set cur=inf RLIM_NOFILE on macOS
virProcessActivateMaxFiles sets rlim_cur to rlim_max.
If rlim_max is RLIM_INFINITY,
2023-08-15 15:17:51.944+0000: 4456752640: debug :
virProcessActivateMaxFiles:1067 : Initial max files was 2560
2023-08-15 15:17:51.944+0000: 4456752640: debug :
virProcessActivateMaxFiles:1077 : Raised max files to 9223372036854775807
then when virCommandMassClose does `int openmax = sysconf(
_SC_OPEN_MAX)`, `openmax < 0` is true and virCommandMassClose
reports an error and bails. Setting rlim_cur instead to at most
OPEN_MAX, as macOS' documentation suggests, both avoids this problem
2023-08-18 16:01:44.366+0000: 4359562752: debug :
virProcessActivateMaxFiles:1072 : Initial max files was 256
2023-08-18 16:01:44.366+0000: 4359562752: debug :
virProcessActivateMaxFiles:1086 : Raised max files to 10240
and eliminates a case of what the documentation declares
to be invalid input to setrlimit anyway.
tools: fix VMSA construction with explicit CPU family/model/stepping
If the CPU family/model/stepping are provided on the command line, but
the firmware is being automatically extracted from the libvirt guest,
we try to build the VMSA too early. This leads to an exception trying
to parse the firmware that has not been loaded yet. We must delay
building the VMSA in that scenario.
Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
tools: fix handling of CPU family/model/stepping in SEV validation
The SEV-ES boot measurement includes the initial CPU register state
(VMSA) and one of the fields includes the CPU identification. When
building a VMSA blob we get the CPU family/model/stepping from the
host capabilities, however, the VMSA must reflect the guest CPU not
host CPU. Thus using host capabilities is only when whe the guest
has the 'host-passthrough' CPU mode active. With 'host-model' it is
cannot be assumed host and guest match, because QEMU may not (yet)
have a named CPU model for a given host CPU.
Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Erik Skultety <eskultet@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Andrea Bolognani [Thu, 24 Aug 2023 15:41:39 +0000 (17:41 +0200)]
ci: Fix quoting and option name
Multiple values passed to --meson-args need to be quoted so that
the shell will interpret them correctly. The option's name was
also reported incorrectly, so fix that as well.
Signed-off-by: Andrea Bolognani <abologna@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Andrea Bolognani [Thu, 24 Aug 2023 15:35:53 +0000 (17:35 +0200)]
ci: Fix precedence between arguments passed to meson
Commit 9c9848f955fd merged $MESON_OPTS into $MESON_ARGS, and
while doing so changed their behavior: while until then the
contents of $MESON_ARGS had precedence over those of $MESON_OPTS,
now the opposite is true. Restore the original behavior and
document it.
The argument for merging the two variables in the first place
was that having both present on the meson command line could be
confusing; however, that should no longer be the case now that
we have reasonably extensive comments explaining the role of
each of the variables and how they interact with each other, so
return the meson command line to its original form.
Signed-off-by: Andrea Bolognani <abologna@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Laine Stump [Fri, 25 Aug 2023 04:09:54 +0000 (00:09 -0400)]
docs: update description of virsh nodedev-detach --driver option
--driver can now be used to specify a specific driver to bind to the
device being detached from the host driver (e.g. vfio-pci-igbvf), not
just the *type* of driver (e.g. "vfio" or "xen", which are unnecessary
anyway, since they are implicit in which hypervisor driver is in use)
Signed-off-by: Laine Stump <laine@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This introduces the ability to set the discard granularity option
for a disk. It defines the smallest amount of data that can be
discarded in a single operation (useful for managing and
optimizing storage).
However, most hypervisors automatically set the proper discard
granularity and users usually do not need to change the default
setting.
Signed-off-by: Kristina Hanicova <khanicov@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Peter Krempa [Fri, 25 Aug 2023 12:16:12 +0000 (14:16 +0200)]
docs: Improve documentation of <disk type='dir'>
Note the implications and caveats of <disk type='dir'>.
Closes: https://gitlab.com/libvirt/libvirt/-/issues/519 Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
K Shiva Kiran [Wed, 16 Aug 2023 18:47:14 +0000 (00:17 +0530)]
Add Test driver and testcase for Network Metadata change APIs
This commit implements the newly defined Network Metadata Get and
Set APIs into the test driver.
It also adds a new testcase "networkmetadatatest" to test the APIs.
Signed-off-by: K Shiva Kiran <shiva_kr@riseup.net> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
projects / libvirt.git / log