util: fix indent depth for glib compat impls

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: keep glib compat methods in alphabetical order

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: always replace g_fsync usage with our wrapper

g_fsync was introduced in 2.63 which is newer than our minimum
glib version. A future commit will introduce compile time
checking of API versions to prevent accidental usage of APIs
from glib newer than our min declared.

To avoid triggering this warning, however, we need to ensure
that we always use our wrapper function via glibcompat.c,
which will disable the API version warnings.

Reviewed-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

security: improve security driver error message

Currently, when security driver is not available users are informed that
it wasn't found which can be confusing.
1. Update error message
2. Add comment to domain doc

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by Sebastian Mitterle <smitterl@redhat.com>

selinux: Use fd_path instead of /dev/tap* to get context

/dev/tap* is an invalid path but it works with lax policy.
Make it work with more accurate policy as well

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Dominick Grift <dac.override@gmail.com>

qemu: backup: roll-back checkpoint metadata if the checkpoint wasn't taken

We insert the checkpoint metadata into the list of checkpoints prior to
actually creating the on-disk bits. If the 'transaction' or any other
steps done between inserting the checkpoint and creating the on-disk
data fail we'd end up with an unusable checkpoint that would vanish
after libvirtd restart.

Prevent this by rolling back the metadata if we didn't actually take and
record the checkpoint.

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

qemu: checkpoint: Extract and export rollback of checkpoint metadata storing

If we are certain that the checkpoint creation failed we remove the
metadata from the list. To allow reusing this in the backup code add a
new helper and export it.

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

man: virsh: Mention how to disable snapshot of disk in snapshot-create-as

Document that --diskspec DISK,snapshot=no disables snapshot for the
given disk.

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

cpu_map/x86: Add support for BFLOAT16 data type

Introduced in QEMU by commit v4.1.0-266-g80db491da4.

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

vircgroupv2devices: Avoid double close on map FD

When allowing/denying a device in devices CGroupV2 we have to
write a BPF program for it. The program we put there is merely
static and all it does it looks up a device in a hash table (also
known as map in BPF terminology). A map is referenced via an FD
which can be acquired via virBPFCreateMap() and like any other FD
it should be closed when no longer needed. However, we close it
twice: the first time in virCgroupV2DevicesAttachProg() which
closes it unconditionally, and the second time in either
virCgroupV2DevicesCreateProg() or
virCgroupV2DevicesPrepareProg(). Remove the second close.

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

vircgroupv2devices: Unexport virCgroupV2DevicesAttachProg()

This function is not called outside of the source file where it's
defined. There's no need to export it.

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

ci: Fetch list of available container images dynamically

Any static list of images is destined to become outdated eventually,
so let's start generating it dynamically instead.

Unfortunately there doesn't seem to be a straightforward way to get
Podman/Docker to list all repositories under quay.io/libvirt, so we
have to resort to searching and filtering manually; and since the
two tools behave slightly differently in that regard, it's more
sane to have the logic in a separate shell script than it would be
to keep it inline in the Makefile with all the annoying escaping
doing so would entail.

Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

Introduce command 'virsh domstats --memory' for reporting memory BW

Introduce an option '--memory' for showing memory related
information. The memory bandwidth infomatio is listed as:

Domain: 'libvirt-vm'
 memory.bandwidth.monitor.count=4
 memory.bandwidth.monitor.0.name=vcpus_0-4
 memory.bandwidth.monitor.0.vcpus=0-4
 memory.bandwidth.monitor.0.node.count=2
 memory.bandwidth.monitor.0.node.0.id=0
 memory.bandwidth.monitor.0.node.0.bytes.total=10208067584
 memory.bandwidth.monitor.0.node.0.bytes.local=4807114752
 memory.bandwidth.monitor.0.node.1.id=1
 memory.bandwidth.monitor.0.node.1.bytes.total=8693735424
 memory.bandwidth.monitor.0.node.1.bytes.local=5850161152
 memory.bandwidth.monitor.1.name=vcpus_7
 memory.bandwidth.monitor.1.vcpus=7
 memory.bandwidth.monitor.1.node.count=2
 memory.bandwidth.monitor.1.node.0.id=0
 memory.bandwidth.monitor.1.node.0.bytes.total=853811200
 memory.bandwidth.monitor.1.node.0.bytes.local=290701312
 memory.bandwidth.monitor.1.node.1.id=1
 memory.bandwidth.monitor.1.node.1.bytes.total=406044672
 memory.bandwidth.monitor.1.node.1.bytes.local=229425152

Signed-off-by: Wang Huaqiang <huaqiang.wang@intel.com>

util, resctrl: using 64bit interface instead of 32bit for counters

The underlying resctrl monitoring is actually using 64 bit counters,
not the 32bit one. Correct this by using 64bit data type for reading
hardware value.

To keep the interface consistent, the result of CPU last level cache
that occupied by vcpu processors of specific restrl monitor group is
still reported with a truncated 32bit data type. because, in silicon
world, CPU cache size will never exceed 4GB.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Wang Huaqiang <huaqiang.wang@intel.com>

util: time: Fix comment for virTimeFieldsNow

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

schemas: backup: Remove pointless <choice> for 'name' of backup disk

One of the first versions thought of using disk path as the second
option but this was dropped as being a legacy interface. Remove the
leftover pointless <choice> wrapper for the disk name as there's just
one option now.

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

qemu: process: Terminate backup job on VM destroy

Commit d75f865fb989b3e6330c78c28e1c3bf7fa28e6a5 caused a job-deadlock if
a VM is running the backup job and being destroyed as it removed the
cleanup of the async job type and there was nothing to clean up the
backup job.

Add an explicit cleanup of the backup job when destroying a VM.

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

qemu: backup: Properly propagate async job type when cancelling the job

When cancelling the blockjobs as part of failed backup job startup
recover we didn't pass in the correct async job type. Luckily the block
job handler and cancellation code paths use no block job at all
currently so those were correct.

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

qemu: blockjob: Remove infrastructure for remembering to delete image

Now that we delete the images elsewhere it's not required. Additionally
it's safe to do as we never released an upstream version which required
this being in place.

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

qemu: backup: Move deletion of backup images to job termination

While qemu is running both locations are identical in semantics, but the
move will allow us to fix the scenario when the VM is destroyed or
crashes where we'd leak the images.

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

qemu: backup: Configure backup store image with backing file

In contrast to snapshots the backup job does not complain when the
backup job's store file has backing pre-configured. It's actually
required so that the NBD server exposes all the data properly.

Remove our fake termination and use the existing disk source as backing.

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

qemu: Reset the node-name allocator in qemuDomainObjPrivateDataClear

qemuDomainObjPrivateDataClear clears state which become invalid after VM
stopped running and the node name allocator belongs there.

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

qemu: block: Use proper asyncJob when waiting for completion of blockdev-create

The waiting loop used QEMU_ASYNC_JOB_NONE rather than 'asyncJob' passed
from the caller.

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

bootstrap: annotate with info about desired replacement

Add a comment against each gnulib module suggesting strategy
for replacement.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

bootstrap: remove now unused gnulib modules

* canonicalize-lgpl: replaced by realpath()/g_canonicalize_filename()
* clock-gettime: replaced by g_get_(real|monotonic)_time
* dirname-lgpl: replaced by g_path_get_dirname()
* fclose: we aren't affected by any portability problems it fixes
* fdatasync: every platform we call fdatasync on has it present
* fsync: replaced by g_fsync()
* fnmatch: replaced by g_pattern_match()
* getcwd-lgpl: replaced by g_get_current_dir()
* gethostname: replaced by g_get_hostname()
* gettimeofday: replaced by g_get_(real|monotonic)_time
* setenv: replaced by g_setenv()
* strptime: replaced by GDateTime
* timegm: replaced by GDateTime
* unsetenv: replaced by g_unsetenv()

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace strptime()/timegm()/mktime() with GDateTime APIs set

All places where we use strptime/timegm()/mktime() are handling
conversion of dates in a format compatible with ISO 8601, so we
can use the GDateTime APIs to simplify code.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace WSAStartup with g_networking_init()

g_networking_init() does the same as our custom code.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: replace gethostname() with g_get_hostname()

Note the glib function returns a const string because it
caches the hostname using a one time thread initializer
function.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: use realpath/g_canonicalize_filename

The canonicalize_file_name(path) is equivalent to calling
realpath(path, NULL). Passing NULL for the second arg of
realpath is not standardized behaviour, however, Linux,
FreeBSD > 6.4 and macOS > 10.5 all support this critical
extension.

This leaves Windows which doesn't provide realpath at all.
The g_canonicalize_filename() function doesn't expand
symlinks, so is not strictly equivalent to realpath()
but is close enough for our Windows portability needs
right now.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace getcwd() with g_get_current_dir()

commandhelper.c is not converted since this is a standalone
program only run on UNIX, so can rely on getcwd().

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: remove unused imports of dirname.h

A few places were importing dirname.h without actually using it.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace mdir_name() with g_path_get_dirname()

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: replace IS_ABSOLUTE_FILE_NAME with g_path_is_absolute

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace last_component() with g_path_get_basename()

The last_component() method is a GNULIB custom function
that returns a pointer to the base name in the path.
This is similar to g_path_get_basename() but without the
malloc. The extra malloc is no trouble for libvirt's
needs so we can use g_path_get_basename().

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: replace clock_gettime()/gettimeofday() with g_get_real_time()

g_get_real_time() returns the time since epoch in microseconds.
It uses gettimeofday() internally while libvirt used clock_gettime
because it is declared async signal safe. In practice gettimeofday
is also async signal safe *provided* the timezone parameter is
NULL. This is indeed the case in g_get_real_time().

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: switch from fnmatch to g_pattern_match_simple

The g_pattern_match function_simple is an acceptably close
approximation of fnmatch for libvirt's needs.

In contrast to fnmatch(), the '/' character can be matched
by the wildcards, there are no '[...]' character ranges and
'*' and '?' can not be escaped to include them literally in
a pattern.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: use g_lstat() instead of lstat()

The GLib g_lstat() function provides a portable impl for
Win32.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: introduce virFileDataSync

A wrapper that calls g_fsync on Win32/macOS and fdatasync
elsewhere. g_fsync is a stronger flush than we need but it
satisfies the caller's requirements & matches the approach
gnulib takes.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: use g_fsync for portability

The g_fsync() API provides the same Windows portability
as GNULIB does for fsync().

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: add compat wrapper for g_fsync

g_fsync isn't available until 2.63 so we need a compat
wrapper temporarily.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: switch to use g_setenv/g_unsetenv

Eliminate direct use of normal setenv/unsetenv calls in
favour of GLib's wrapper. This eliminates two gnulib
modules

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

src: always pull in glib/gstdio.h header

The gstdio.h header defines some low level wrappers for
things like fsync, stat, lstat, etc.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

util: add note about event file descriptors on Windows

When using GNULIB with Winsock, libvirt will never see the normal HANDLE
objects, instead GNULIB guarantees that libvirt gets a C runtime file
descriptor. The GNULIB poll impl also expects to get C runtime file
descriptors rather than HANDLE objects. Document this behaviour so that
it is clear to applications providing event loop implementations if they
need Windows portability.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

travis: add macOS Xcode 11.3 testing

Ideally we would test macOS 10.15 as the newest release, however, that
is not available in Travis yet. We can at least test newer XCode
versions though to get toolchain validation.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

docs: expand macOS platform support coverage

We initially claimed to only support the most recent macOS
release, which is currently 10.15. Our Travis CI, however,
is validating 10.14.4 / XCode 10.3.

For almost all of our other platforms, we support multiple
releases to some degree. This change brings macOS in line
with other long life distros, covering the most recent &
most recent but one for a 2 year overlap. With this docs
change our CI is now actually testing our minimum version.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

build: set min version for CLang to 3.4 / XCode CLang to 5.1

We have a strong check for GCC >= 4.8, but don't validate any
version number for CLang historically.

This defines the min CLang to be 3.4 which is what is available
for RHEL-7. macOS uses a different versioning scheme for CLang,
based off XCode versions. There is a mapping recorded at

  https://en.wikipedia.org/wiki/Xcode#Toolchain_versions

Here we see upstream CLang 3.4 corresponds to XCode 5.1

XCode 5.1 is available for macOS 10.8.4 or later which
trivially satisfies our platform support matrix requirements.

All these versions match what QEMU declares for its min GCC
and CLang checks.

Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

virsh: Add a completer for `domifaddr` --source parameter.

The command `domifaddr` can use three different sources to grab IP
address of a Virtual Machine: lease, agent and arp. This parameter does
not have a completer function to return source options.

Signed-off-by: Julio Faracco <jcfaracco@gmail.com>
Reviewed-by: Erik Skultety <eskultet@redhat.com>

travis: fix homebrew install of python3

The Python3 package has started failing to install from
HomeBrew with the following:

  Error: The `brew link` step did not complete successfully
  The formula built, but is not symlinked into /usr/local
  Could not symlink Frameworks/Python.framework/Headers
  Target /usr/local/Frameworks/Python.framework/Headers
  is a symlink belonging to python@2. You can unlink it:

    brew unlink python@2

  To force the link and overwrite all conflicting files:

    brew link --overwrite python

The result is that libvirt fails to find python3:

  checking for python3... no
  configure: error: 'python3' binary is required to build libvirt

It is unclear what changed in Travis/HomeBrew to break our
previously working setup, but running the suggested command
fixes it well enough for libvirt's CI needs.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

maint: update to latest gnulib

Update to:

    commit 7d069378921bfa0d7c7198ea177aac0a2440016f
    Author:     Pádraig Brady <P@draigBrady.com>
    CommitDate: 2020-01-01 22:00:28 +0000

       md5, sha1, sha256, sha512: support --with-openssl=auto-gpl-compat

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

schemas: Allow additional qemu cmd line arguments/env variables and qemuCaps to be interleaved

While command line arguments are sort of positional (because you
have to have two entries, one for "-arg" the other for "value"),
it doesn't really matter whether env variables come before or
after command line arguments.

And it matters even less when playing with qemu capabilities.

Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

qemu: hide details of fake reboot

If we use fake reboot then domain goes thru running->shutdown->running
state changes with shutdown state only for short period of time.  At
least this is implementation details leaking into API. And also there is
one real case when this is not convinient. I'm doing a backup with the
help of temporary block snapshot (with the help of qemu's API which is
used in the newly created libvirt's backup API). If guest is shutdowned
I want to continue to backup so I don't kill the process and domain is
in shutdown state. Later when backup is finished I want to destroy qemu
process. So I check if it is in shutdowned state and destroy it if it
is. Now if instead of shutdown domain got fake reboot then I can destroy
process in the middle of fake reboot process.

After shutdown event we also get stop event and now as domain state is
running it will be transitioned to paused state and back to running
later. Though this is not critical for the described case I guess it is
better not to leak these details to user too. So let's leave domain in
running state on stop event if fake reboot is in process.

Reconnection code handles this patch without modification. It detects
that qemu is not running due to shutdown and then calls qemuProcessShutdownOrReboot
which reboots as fake reboot flag is set.

Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

Storage: Use errno parameter in virReportSystemError

Use errno parameter in virReportSystemError.
Remove hold function return values if don't need.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Yi Li <yili@winhong.com>

storage: Fix volStorageBackendRBDRefreshVolInfo function return errors

Fix the return value status comparison checking for call to
volStorageBackendRBDRefreshVolInfo introduced by commit id f46d137e.

we only should fail when the return is < 0. -ENOENT, -ETIMEDOUT will
ignore according commit id f46d137e.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Yi Li <yili@winhong.com>

Storage: Use rc hold intermediate function return values.

most libvirt code uses 'int rc' to hold intermediate
function return values. consistent with the rest of libvirt.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Yi Li <yili@winhong.com>

qemu: store the emulator name in the capabilities XML

We don't need this for any functional purpose, but when debugging hosts
it is useful to know what binary a given capabilities XML document is
associated with.

Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: add qemu caps constructor which takes binary name

Simplify repeated code patterns by providing a new constructor taking
the QEMU binary name.

Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

qemu: add explicit flag to skip qemu caps invalidation

Currently if the binary path is NULL in the qemu capabilities object,
cache invalidation is skipped. A future patch will ensure that the
binary path is always non-NULL, so a way to explicitly skip invalidation
is required.

Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

docs: ensure outputfile is deleted if rst2html/rst2man fail

This avoids leaving a zero length or partially generated output
file on errors.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

docs/auth.html.in: fixed typos

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Andrew Miloradovsky <andrew@interpretmath.pw>

storage: Fix daemon crash on lookup storagepool by targetpath

Causing a crash when storagePoolLookupByTargetPath beacuse of
Some types of storage pool have no target elements.
Use STREQ_NULLABLE instead of STREQ
Avoids segfaults when using NULL arguments.

Core was generated by `/usr/sbin/libvirtd'.
Program terminated with signal 11, Segmentation fault.
(gdb) bt
0  0x0000ffff9e951388 in strcmp () from /lib64/libc.so.6
1  0x0000ffff92103e9c in storagePoolLookupByTargetPathCallback (
    obj=0xffff7009aab0, opaque=0xffff801058b0) at storage/storage_driver.c:1649
2  0x0000ffff9f2c52a4 in virStoragePoolObjListSearchCb (
    payload=0xffff801058b0, name=<optimized out>, opaque=<optimized out>)
    at conf/virstorageobj.c:476
3  0x0000ffff9f1f2f7c in virHashSearch (ctable=0xffff800f4f60,
    iter=iter@entry=0xffff9f2c5278 <virStoragePoolObjListSearchCb>,
    data=data@entry=0xffff95af7488, name=name@entry=0x0) at util/virhash.c:696
4  0x0000ffff9f2c64f0 in virStoragePoolObjListSearch (pools=0xffff800f2ce0,
    searcher=searcher@entry=0xffff92103e68 <storagePoolLookupByTargetPathCallback>,
     opaque=<optimized out>) at conf/virstorageobj.c:505
5  0x0000ffff92101f54 in storagePoolLookupByTargetPath (conn=0xffff5c0009f0,
path=0xffff7009a850 "/vms/images") at storage/storage_driver.c:1672

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Yi Li <yili@winhong.com>

qemu_process.c: remove 'cleanup' label from qemuProcessCreatePretendCmd()

The 'cleanup' flag is doing no cleaup in this function. We can
remove it and return NULL on error or qemuBuildCommandLine().

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

qemu_process.c: remove cleanup labels after g_auto*() changes

The g_auto*() changes made by the previous patches made a lot
of 'cleanup' labels obsolete. Let's remove them.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

qemu_process.c: use g_autoptr()

Change all feasible pointers to use g_autoptr().

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

qemu_domain.h: add G_DEFINE_AUTOPTR_CLEANUP_FUNC for qemuDomainLogContext

This will allow us to g_autoptr qemuDomainLogContext pointers
in the following patch.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

qemu_process.c: use g_autofree

Change all feasible strings and scalar pointers to use g_autofree.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>

vbox: Get rid of "no_memory" labels

As pointed out by Ján Tomko, "no_memory seems suspicious in the times of
abort()".

As libvirt decided to take the path to not report OOM and simply abort
when it happens, let's get rid of the no_memory labels and simplify the
code around them.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>

util: Get rid of "no_memory" labels

As pointed out by Ján Tomko, "no_memory seems suspicious in the times of
abort()".

As libvirt decided to take the path to not report OOM and simply abort
when it happens, let's get rid of the no_memory labels and simplify the
code around them.

Mind that virfirewall.c was not touched and still contains no_memory
labels. The reason those are left behind, at least for now, is because
the conversion seems to be slightly more complicated than the rest, as
some other places are relying on firewall->err being set to ENOMEM.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>

rpc: Get rid of "no_memory" labels

As pointed out by Ján Tomko, "no_memory seems suspicious in the times of
abort()".

As libvirt decided to take the path to not report OOM and simply abort
when it happens, let's get rid of the no_memory labels and simplify the
code around them.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>

openvz: Get rid of "no_memory" labels

As pointed out by Ján Tomko, "no_memory seems suspicious in the times of
abort()".

As libvirt decided to take the path to not report OOM and simply abort
when it happens, let's get rid of the no_memory labels and simplify the
code around them.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>

conf: Get rid of "no_memory" labels

As pointed out by Ján Tomko, "no_memory seems suspicious in the times of
abort()".

As libvirt decided to take the path to not report OOM and simply abort
when it happens, let's get rid of the no_memory labels and simplify the
code around them.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>

news: document phyp removal

Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Cole Robinson <crobinso@redhat.com>

Remove phyp driver

The phyp driver was added in 2009 and does not appear to have had any
real feature change since 2011. There's virtually no evidence online
of users actually using it. IMO it's time to kill it.

This was discussed a bit in April 2016:
https://www.redhat.com/archives/libvir-list/2016-April/msg01060.html

Final discussion is here:
https://www.redhat.com/archives/libvir-list/2019-December/msg01162.html

Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Cole Robinson <crobinso@redhat.com>

docs: rewrite polkit docs generator in Python

As part of a goal to eliminate Perl from libvirt build tools,
rewrite the genaclperms.pl tool in Python.

This was a straight conversion, manually going line-by-line to
change the syntax from Perl to Python. Thus the overall structure
of the file and approach is the same.

Tested-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

docs: rewrite hvsupport.html page generator in python

As part of a goal to eliminate Perl from libvirt build tools,
rewrite the hvsupport.pl tool in Python.

This was a straight conversion, manually going line-by-line to
change the syntax from Perl to Python. Thus the overall structure
of the file and approach is the same.

The new impl generates byte-for-byte identical output to the
old impl.

Tested-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

tests: rewrite file access checker in Python

As part of a goal to eliminate Perl from libvirt build tools,
rewrite the check-file-access.pl tool in Python.

This was a straight conversion, manually going line-by-line to
change the syntax from Perl to Python. Thus the overall structure
of the file and approach is the same.

Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

tests: rewrite qemu capability grouper in Python

As part of a goal to eliminate Perl from libvirt build tools,
rewrite the group-qemu-caps.pl tool in Python.

This was a straight conversion, manually going line-by-line to
change the syntax from Perl to Python. Thus the overall structure
of the file and approach is the same.

Tested-by: Cole Robinson <crobinso@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

esx: set vmfs fs type for vmfs-based datastores

This way they are correctly represented:
  <source>
    <format type='vmfs'/>
  </source>
... instead of 'auto'.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

storage: add vmfs filesystem type

It will be used to represent the type of a filesystem pool in ESXi.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

esx: split virtualswitchToNetwork helper

Move the creation of a virNetworkPtr object from the
esxVI_HostVirtualSwitch object of a virtual switch out of
esxNetworkLookupByName in an own helper. This way it can be used also
in other functions.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>

esx: split targetToStoragePool helper

Move the creation of a virStoragePtr object from the
esxVI_HostInternetScsiHbaStaticTarget object of a target out of
esxStoragePoolLookupByName in an own helper. This way it can be used
also in other functions.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

esx: split datastorePoolType helper

Move the detection of the type of a vmfs pool out of
esxLookupVMFSStoragePoolType in an own helper. This way it can be used
also in other functions.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

esx: split datastoreToStoragePoolPtr helper

Move the creation of a virStoragePtr object from the esxVI_ObjectContent
object of a datastore out of esxStoragePoolLookupByName in an own
helper. This way it can be used also in other functions.

Signed-off-by: Pino Toscano <ptoscano@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

admin: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

admin: Use g_autofree in getSocketPath()

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>

interface: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

locking: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

locking: Use g_autofree in virLockDaemonExecRestartStatePath()

Together with the change, let's also simplify the function and get rid
of the goto.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

locking: Use g_autofree in virLockDaemonUnixSocketPaths()

Together with the change, let's also simplify the function and get rid
of the goto.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

locking: Use g_autofree in virLockManagerLockDaemonPath()

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

logging: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

logging: Use g_autofree in virLogDaemonExecRestartStatePath()

Together with the change, let's also simplify the function and get rid
of the goto.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

logging: Use g_autofree in virLogDaemonUnixSocketPaths()

Together with the change, let's also simplify the function and get rid
of the goto.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

logging: Use g_autofree in virLogManagerDaemonPath()

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

network: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

node_device: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

qemu: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

remote: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

rpc: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

secret: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>

storage: Don't check the output of virGetUserRuntimeDirectory()

virGetUserRuntimeDirectory() *never* *ever* returns NULL, making the
checks for it completely unnecessary.

Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>