qemuBlockJobProcessEventCompletedPull: Avoid dangling pointer after blockpull

When doing a full block pull job (base == NULL) and the config XML
contains a compatible disk, the completer function would leave a
dangling pointer in 'cfgdisk->src->backingStore' as cfgdisk->src would
be set to the value of 'cfgbase' which was always set to
'cfgdisk->src->backingStore'.

This is wrong though since for the live definition XML we set the
respective counterpart to 'job->data.pull.base' which is NULL in the
above scenario.

This leads to a invalid pointer read when saving the config XML and may
end up in a crash.

Resolve it by setting 'cfgbase' only when 'job->data.pull.base' is
non-NULL.

Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1946918
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

nodedev: Only set up mdevctl monitors if mdevctl.d exist

During its initialization, the nodedev driver tries to set up
monitors for /etc/mdevctl.d directory, so that it can register
mdevs as they come and go. However, if the file doesn't exist
there is nothing to monitor and therefore we can exit early. In
fact, we have to otherwise monitorFileRecursively() fails and
whole driver initialization fails with it.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

nodedev: Separate mdevctl monitor setup into a function

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

nodedev: Don't join not spawned threads

During the nodedev driver initialization two threads are created:
one for listening on udev events (like device plug/unplug) and
the other for enumerating devices (so that the main thread doing
the driver init is not blocked). If something goes wrong at any
point then nodeStateCleanup() is called which joins those two
threads (possibly) created before. But it tries to join them even
they weren't created which is undefined behaviour (and it just so
happens that it crashes on my system).

If those two virThread variables are turned into pointers then we
can use comparison against NULL to detect whether threads were
created.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

nodedev: Lock @priv sooner

The nodedev driver private data object @priv is created by
calling udevEventDataNew(). After that, driver->privateData
pointer is set to the freshly allocated object and only a few
lines after all of this the object is locked. Technically it is
safe because there should not be any other thread at this point,
but defensive style of programming says it's better if the object
is locked before driver's privateData is set.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

nodedev: Unlock @priv if initialization of mdevctlMonitors fails

If initialization of priv->mdevctlMonitors fails, then the
control jumps over to cleanup label where nodeStateCleanup() is
called which tries to lock @priv. But since @priv was already
locked before taking the jump a deadlock occurs. The solution is
to jump onto @unlock label, just like the code around is doing.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Pavel Hrdina <phrdina@redhat.com>

bhyve: Fix declaration of 'params' in 'bhyveParsePCIFbuf'

In commit ad80bba90a3 I mistakenly didn't delete '**' from the
variable declaration when converting it to 'GStrv' and deleted the
'separator' variable since it was declared on the same line as a
different variable.

Fixes: ad80bba90a3
Signed-off-by: Peter Krempa <pkrempa@redhat.com>

ci: README: Mention necessary step for cirrus to pick up the github project

Unless you create such an commit, cirrus-ci.com will not pick up the
github project and cirrus-run will fail.

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

tests: qemucapabilitiesdata: Fix wrong command identifier in caps_4.0.0.riscv64.replies

Commit cbd4ab4cc69 added a hunk with a broken reply ID.

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

tests: qemucapabilitiesdata: Fix formatting of manually added hunk

Commit 66c5674e797 added a query for the device properties of 'usb-host'
but the command header isn't formated the same way as if it were
autogenerated. Reformat all the files.

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

qemuxml2xmltest: Always include basic set of capabilities

Use virQEMUCapsInitQMPBasicArch to add the basic set of capabilities
which all qemu versions will get.

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

qemuxml2argvtest: Parse 'arch' from XML early

If we want to provide correct (fake) caps already for the XML parser we
need to be able to parse the arch early so that we can properly
initialize the caps cache prior to calling the XML parser.

This patch adds code which parses the arch and updates the caps cache
prior to the parse step.

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

qemuxml2argvtest: Rewrite parsing of XMLs to provide earlier parsing

In upcoming patches we'll need to parse a certain bit of XML before
calling the full XML parser. This effectively open-codes what
virDomainDefParseFile to reach virDomainDefParseNode.

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

virQEMUCapsInitQMPBasicArch: Use switch for arch-based decisions

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

qemuxml2(argv|xml)test: Don't exit early when testQemuInfoSetArgs fails

Set the return code to failure and continue testing.

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

qemuxml2xmltest: testInfoSetPaths: Remove return value

The function doesn't fail. Remove the return value and checks.

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

docs/platforms: Drop separate link to qemu-kvm on repology.org

The qemu-kvm page redirects to qemu on repology.

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

Remove test 'args' file rewrapping infrastructure

All tests which use files with 'ldargs' and 'args' suffix as output now
use the internal and better line splitting.

Remove the test-wrap-argv.py script, the syntax check which used it and
the helper rewrapping the output when regenerating test output.

For any further use, we require code to use virCommand anyways and thus
it has internal wrapping now.

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

nodedevmdevctltest: Use internal linewrapping and variable command path stripping

Remove `nodedevCompareToFile` which was stripping the path to mdevctl
since it's no longer needed if we use the new features of
virCommandSetDryRun.

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

bhyvexml2argvtest: Use internal wrapping of command line arguments

virCommandToString has the possibility to return an already wrapped
string with better format than what we get from the test wrapper script.

The main advantage is that arguments for an option are always on the
same line which makes it more easy to see what changed in a diff and
prevents re-wrapping of the line if a wrapping point moves over the
threshold.

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

nwfilterxml2firewalldata: Use internal wrapping to wrap output files

As with previous commits use virCommandSetDryRun to invoke
virCommandToString so that it returns pre-wrapped string.

Since virCommand is better aware of where the arguments terminate we can
see an improvement where comments are no longer line-wrapped.

The changes to the 'commonRules' strings were done with the following
regex:

 s/ -/ \\\\\\n-/

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

networkxml2firewalltest: Use internal wrapping of command line arguments

virCommandSetDryRun allows to invoke virCommandToString so that the
command string is already wrapped.

We now also need to load the base arguments file without unwrapping the
arguments.

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

storagevolxml2argvtest: Use internal wrapping of command line arguments

virCommandToString has the possibility to return an already wrapped
string with better format than what we get from the test wrapper script.

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

storagepoolxml2argvtest: Use internal wrapping of command line arguments

virCommandToString has the possibility to return an already wrapped
string with better format than what we get from the test wrapper script.

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

storage: Format mount options before positional arguments

Move calls to virStorageBackendFileSystemMountAddOptions earlier so that
the options are formatted before the positional arguments.

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

qemuxml2argvtest: Use internal wrapping of command line arguments

virCommandToString has the possibility to return an already wrapped
string with better format than what we get from the test wrapper script.

The main advantage is that arguments for an option are always on the
same line which makes it more easy to see what changed in a diff and
prevents re-wrapping of the line if a wrapping point moves over the
threshold.

Additionally the used output is the same we have in the VM log file when
a VM is starting.

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

syntax-check: Remove 90 char limit for 'argv' files

Splitting lines with arguments causes in many cases a rewrap if the
arguments are modified making it harder to see what actually changed.

In upcoming patches some rewrapping of 'args' files will be removed so
remove this check first.

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

virCommandToStringFull: Improve linebreaking behaviour

Put multiple values for an option if followed by another option as used
in certain iptables arguments.

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

virTestLoadFileJSON: Don't try to unwrap JSON documents

Use virFileReadAll to load the file instead of virTestLoadFile which
tries to unwrap the file.

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

virTestCompareToFile: Add possibility to skip unwrapping of input file

In some cases we might want to compare already wrapped data against a
wrapped file. Introduce virTestCompareToFileFull with a 'unwrap' boolean
which will control the unwrapping.

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

testutils: Remove unused virTestClearCommandPath

virCommand(ToString) now provides the functionality internally so we
don't have to keep the string-munging function around.

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

nwfilterebiptablestest: Strip path from test output via virCommandSetDryRun

Enable the internal path clearing instead of using
virTestClearCommandPath.

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

nwfilterxml2firewalltest: Strip path from test output via virCommandSetDryRun

Enable the internal path clearing instead of using
virTestClearCommandPath.

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

networkxml2firewalltest: Strip path from test output via virCommandSetDryRun

Enable the internal path clearing instead of using
virTestClearCommandPath.

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

virCommandSetDryRun: Add flags to linebreak and strip prefix from the command buffer

virCommandToStringFull used internally when virCommandSetDryRun is
requested allows to strip command path and wrap lines nicely. Expose
these via virCommandSetDryRun so that tests can use those features
instead of local hacks.

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

virCommandSetDryRun: Rework resetting of the dry run data

While virCommandSetDryRun is used in tests only, there were some cases
when error paths would not call the function with NULL arguments to
reset the dry run infrastructure.

Introduce virCommandDryRunToken type which must be allocated via
virCommandDryRunTokenNew and passed to virCommandSetDryRun.

This way we can use automatic variable cleaning to trigger the cleanup
of virCommandSetDryRun parameters and also the use of the token variable
ensures that all callers of virCommandSetDryRun clean up after
themselves and also that the token isn't left unused in the code.

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

tests: storagepoolxml2argvtest: Use integrated command path stripping

Replace virTestClearCommandPath by virCommandToStringFull which allows
to strip the command prefix internally.

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

virCommandToString: Allow stripping command path

In tests we don't want to use the full path to commands as it's
unpleasant to keep that working on all systems.

Add an integrated way to strip the prefix which will be used to replace
virTestClearCommandPath() as a more systemic solution.

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

qemuxml2xmloutdata: Remove unused symlinks to test XML files

The symlinks are not used by the test.

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

qemuxml2xmloutdata: Remove unused output XML files

The files are no longer referenced by existing test cases. Remove them.

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

qemuxml2argvdata: Remove unused 'xml' files

The files are no longer referenced by either qemuxml2argvtest or
qemuxml2xmltest. Remove them.

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

qemuxml2argvdata: Remove unused 'args' files

The files were added in error (audio-*) for test cases which produce an
error, left over after converting to DO_TEST_CAPS_LATEST
(disk-detect-zeroes), or left over after splitting test cases
(disk-network-tlsx509).

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

tests: string: Remove pointless test for virStringListFreeCount

It's way more useful to run valgrind against the rest of the code than
this test to see whether virStringListFreeCount works. Remove the test.

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

util: virstring: Remove the virStringSplitCount wrapper funcion

Callers which need the count of elements now count it in place.

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

virVMXParseConfig: Replace virStringSplitCount by g_strsplit

Remove the last usage of virStringSplitCount

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

virSystemdActivationInitFromNames: Replace virStringSplit by g_strsplit

While the code invokes the string list length calculation twice, it
happens only on error path, which by itself should never happen.

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

openvzParseBarrierLimit: Rework string handling

Use g_strsplit instead of virStringSplitCount and automatically free the
temporary string list.

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

xenParseXLVnuma: Replace virStringSplitCount by g_strsplit

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

xenParsePCI: Replace virStringSplitCount by g_strsplit

Count the number of elements in place just for the check.

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

util: virresctrl: Use g_strsplit instead of virStringSplitCount

In 3 of 4 instances the code didn't even need the count of the elements.

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

util: virresctrl: Remove empty 'cleanup' sections

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

util: virresctrl: Use automatic memory freeing

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

virResctrlAllocGetUnused: Use g_autoptr for variables of virResctrlAlloc type

Refactor the handling of variables so that the cleanup section can be
sanitized.

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

virResctrlAllocNewFromInfo: Use g_autoptr for 'ret'

Remove 'cleanup' and 'error' labels by switching 'ret' to automatic
pointer and stealing it in the return statement.

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

virResctrlAllocNewFromInfo: Restrict variable scope and use automatic freeing

Move variables into the loop which uses them and use automatic freeing
for temporarily allocated variables.

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

virResctrlGetCacheInfo: Restrict variable scope and use automatic freeing

Move variables into the loop which uses them and use automatic freeing
for temporarily allocated variables.

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

storage: zfs: Use g_strsplit instead of virStringSplitCount

Both instances just check the length once. Replicate that faithfully.

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

virStorageBackendZFSRefreshPool: Reduce scope of 'tokens'

Declare it in the loop that actually uses it.

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

storage: zfs: Don't split string if we need only first/last component

Use str(r)chr to find the correct bit rather than fully splitting the
string.

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

virStorageSourceParseBackingJSONUriCookies: Use g_strsplit instead of virStringSplitCount

Count the elements after splitting the string.

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

virshParseRateStr: Use g_strsplit instead of virStringSplitCount

Count the elements after splitting the string.

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

virshParseRateStr: Refactor cleanup

Use g_auto for the string list and remove 'ret' and 'cleanup'.

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

virHostValidateGetCPUFlags: Use g_strsplit instead of virStringSplitCount

We don't need the count of elements to iterate the list.

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

bhyveParsePCIFbuf: Use g_strsplit instead of virStringSplitCount

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

virshParseEventStr: Use g_strsplit and automatic memory freeing

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

virLogParseOutput: Replace virStringSplitCount by g_strsplit

Unfortunately here we do need the count of elements. Use g_strv_length
to calculate it so that virStringSplitCount can be removed later.

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

virLogParseFilter: Replace virStringSplitCount by g_strsplit

We don't really need the count.

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

virLogParseFilters: Refactor string list handling

Rewrite the code to remove the need to calculate the string list count.

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

virLogParseOutputs: Refactor string list handling

Rewrite the code to remove the need to calculate the string list count.

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

util: virlog: Remove pointless 'cleanup' labels

Previous refactors left empty cleanup labels. Remove them.

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

util: virlog: Use g_auto(GStrv) instead of g_strfreev in cleanup section

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

virStorageFileParseBackingStoreStr: use g_strsplit instead of virStringSplitCount

The presence of the second element can be checked by looking at it
directly.

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

virJSONValueObjectDeflattenWorker: use g_strsplit instead of virStringSplitCount

The presence of the second element can be checked by looking at it
directly.

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

virDomainDiskAddISCSIPoolSourceHost: use g_strsplit instead of virStringSplitCount

Count the elements directly using g_strv_length.

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

Remove virStorageFileCanonicalizePath

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

tests: Remove testing of virStorageFileCanonicalizePath

Remove the last code using the function.

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

virStorageFileBackendGlusterPriv: Remove 'cannonpath'

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

storage_file: Remove virStorageFileBackendFsPriv

The private data structure is no longer used.

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

Remove virStorageSourceGetUniqueIdentifier file backend API

The API isn't used any more.

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

virStorageSourceGetMetadata: Use depth limit instead of unique path checking

Prevent unbounded chains by limiting the recursion depth of
virStorageSourceGetMetadataRecurse to the maximum number of image layers
we limit anyways.

This removes the last use of virStorageSourceGetUniqueIdentifier which
will allow us to delete some crusty old infrastructure which isn't
really needed.

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

qemuDomainStorageSourceValidateDepth: Define chain depth as macro

The magic constant will be used in one more place.

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

util: json: Remove virJSONValueNewArrayFromBitmap

The function is used only inside of the file. We can open-code it and
remove it as it's not very useful.

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

util: json: Remove virJSONValueGetArrayAsBitmap

The function is not used.

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

virQEMUBuildCommandLineJSONArrayBitmap: Open code bitmap conversion

Add a simpler algorithm converting the JSON array to bitmap so that
virJSONValueGetArrayAsBitmap can be removed in next step.

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

Translated using Weblate (Korean)

Currently translated at 32.9% (3485 of 10581 strings)

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

Translated using Weblate (Korean)

Currently translated at 32.8% (3481 of 10581 strings)

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

Co-authored-by: simmon <simmon@nplob.com>
Signed-off-by: simmon <simmon@nplob.com>

meson: Drop readline kludge

Both FreeBSD ports and Homebrew on macOS have readline 8.1 now,
and that version contains a correct pkg-config file so the kludge
is no longer necessary.

Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

api: Add 'flags' param to virNodeDeviceCreate/Undefine()

Follow best practices and add a unsigned int flags parameter to these
new APIs that have not been in a release yet.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

nodedev: fix release version in comments for new API

The comments mistakenly say 7.2.0, when they were actually merged during
the 7.3 development cycle.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>

downloads.html: Add a link to GPG key used signing releases

While the key is available on public GPG key servers, having it locally
at https://libvirt.org/sources/gpg_key.asc is even better.

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

vsh: fix memory leak in vshCommandParse

One of the error branches used a plain free where vshCommandFree
was required.

https://bugzilla.redhat.com/show_bug.cgi?id=1943415

Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: avoid delay when defining a new mdev

When calling virNodeDeviceDefineXML() to define a new mediated device,
we call virMdevctlDefine() and then wait for the new device to appear in
the driver's device list before returning. This caused long delays due
to the behavior of nodeDeviceFindNewMediatedDevice(). This function
checks to see if the device is in the list and then waits for 5s before
checking again.

Because mdevctl is relatively slow to query the list of defined
devices[0], the newly-defined device was generally not in the device
list when we first checked. This results in libvirt almost always taking
at least 5s to complete this API call for mediated devices, which is
unacceptable.

In order to avoid this long delay, we resort to a workaround. If the
call to virMdevctlDefine() was successful, we can assume that this new
device will exist the next time we query mdevctl for new devices. So we
simply add this provisional device definition directly to the nodedev
driver's device list and return from the function. At some point in the
future, the mdevctl handler will run and the "official" device will be
processed, which will update the provisional device if any new details
need to be added.

The reason that this is not necessary for virNodeDeviceCreateXML() is
because detecting newly-created (not defined) mdevs happens through
udev instead of mdevctl. And nodeDeviceFindNewMediatedDevice() always
calls 'udevadm settle' before checking to see whether the device is in
the list. This allows us to wait just long enough for all udev events to
be processed, so the device is almost always in the list the first time
we check and so we almost never end up hitting the 5s sleep.

[0] on my machine, 'mdevctl list --defined' took around 0.8s to
complete for only 3 defined mdevs.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: factor out function to add mediated devices

To accomodate re-use of this functionality in a following patch, split
out the processing of an individual mdev definition into a separate
function.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: add docs about mdev attribute order

Mention that mdev attribute order is significant.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: fix hang when destroying an mdev in use

Calling `mdevctl stop` for a mediated device that is in use by an active
domain will block until that vm exits (or the vm closes the device).
Since the nodedev driver cannot query the hypervisor driver to see
whether any active domains are using the device, we resort to a
workaround that relies on the fact that a vfio group can only be opened
by one user at a time. If we get an EBUSY error when attempting to open
the group file, we assume the device is in use and refuse to try to
destroy that device.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: add ability to specify UUID for new mdevs

Use the new <uuid> element in the mdev caps to define and start devices
with a specific UUID.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

nodedev: add <uuid> element to mdev caps

It will be useful to be able to specify a particular UUID for a mediated
device when defining the node device. To accomodate that, allow this to
be specified in the xml schema. This patch also parses and formats that
value to the xml, but does not yet use it.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

virsh: add "nodedev-start" command

This virsh command maps to virNodeDeviceCreate(), which starts a node
device that has been previously defined by virNodeDeviceDefineXML().
This is only supported for mediated devices.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

api: add virNodeDeviceCreate()

This new API function provides a way to start a persistently-defined
mediate device that was defined by virNodeDeviceDefineXML() (or one that
was defined externally via mdevctl)

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

virsh: add nodedev-undefine command

Add a virsh command that maps to virNodeDeviceUndefine().

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

virsh: Factor out function to find node device

Several functions accept providing a node device by name or by wwnn,wwpn
pair. Extract the logic to do this into a function that can be used by
both callers.

Signed-off-by: Jonathon Jongsma <jjongsma@redhat.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>