Pavel Hrdina [Fri, 24 Jul 2020 12:02:39 +0000 (14:02 +0200)]
meson: tests: virmockstathelpers: replace check for _FILE_OFFSET_BITS
Meson always defines _FILE_OFFSET_BITS=64 which effectively makes
mocking of non 64-bit stat functions dead code.
On linux it was not an issue because we use the 64-bit versions but
on FreeBSD there are not 64-bit versions, there is only stat & lstat.
We cannot simply drop the check as that would resolve to compilation
error on 64-bit linux:
{standard input}: Assembler messages:
{standard input}:11468: Error: symbol `__xstat64' is already defined
{standard input}:11679: Error: symbol `__xstat64.cold' is already defined
{standard input}:12034: Error: symbol `__lxstat64' is already defined
{standard input}:12245: Error: symbol `__lxstat64.cold' is already defined
So we have to replace the _FILE_OFFSET_BITS with a check if the
corresponding 64-bit version of the stat function exists.
Replicate the meson behavior by always defining _FILE_OFFSET_BITS
instead of using AC_SYS_LARGEFILE otherwise this change would break
our tests.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Neal Gompa <ngompa13@gmail.com>
With meson we no longer have .libs directory with the actual binary so
we have to take a different approach to detect if running from build
directory.
This is not as robust as for autotools because if you select --prefix
in the build directory it will incorrectly enable the override as well
but nobody should do that.
We have to modify some of the tests to not add current build path into
PATH variable and use the full path for virsh instead. Otherwise it
would be impossible to figure out that we are running virsh from build
directory.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Neal Gompa <ngompa13@gmail.com>
Pavel Hrdina [Fri, 24 Jul 2020 14:04:19 +0000 (16:04 +0200)]
meson: drop test_suite configure argument
With autoconf this option controlled if the test suite is compiled by
default or not with the fact that it will be compiled later when
running `make check`.
With meson it is not possible to compile it later when running
`ninja test` as it will be always compiled if referenced by `test()`
function in meson.build files.
Since we cannot postpone compilation of the test suite drop this option
as it will not be converted to meson.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Neal Gompa <ngompa13@gmail.com>
Pavel Hrdina [Tue, 14 Jul 2020 14:29:26 +0000 (16:29 +0200)]
meson: ci: increase git clone depth to 1000
Having limit set to 100 is in most cases perfect but sometimes there can
be a larger series that will have more than 100 patches and it will make
the check-dco job fail.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Neal Gompa <ngompa13@gmail.com>
Ján Tomko [Fri, 31 Jul 2020 12:51:00 +0000 (14:51 +0200)]
ci: run container jobs right away
The jobs building containers do not depend on anything.
Make this explicit by including the 'needs' keyword in case
we need to shuffle around the stages in the future.
Signed-off-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Andrea Bolognani <abologna@redhat.com> Reviewed-by: Neal Gompa <ngompa13@gmail.com>
Ján Tomko [Sat, 25 Jul 2020 07:42:14 +0000 (09:42 +0200)]
conf: pass elementName to virDomainThreadSchedParseHelper
Pass the scheduler element name instead of trying to reconstructing
it from the attribute name.
This has the benefit of not mixing '%s' with regular text in
translatable strings as well as preventing the confusion when
the 's' marking the plural in the element name ('vcpus') is taken
as a first letter of the 'sched' suffix.
Signed-off-by: Ján Tomko <jtomko@redhat.com> Fixes: 7ea55a481dd45f09b54425005362533ecc800cd2 Fixes: 99c5fe0e7c26c08103415248ffef1f5acb81ddc7 Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Ján Tomko [Sat, 25 Jul 2020 07:29:36 +0000 (09:29 +0200)]
conf: rename 'name' in scheduler parser
virDomainThreadSchedParseHelper is used for parsing both iothread
and vcpu scheduling settings. Rename its 'name' attribute to
make it obvious this refers to the attribute name, not the name of
the element (which is currently constructed from the attribute name).
Signed-off-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Ján Tomko [Tue, 28 Jul 2020 09:05:16 +0000 (11:05 +0200)]
ci: move sanity_checks after other stages
Be less hostile to developers and run the build jobs even
if the dco-check failed. That way they can test their own
"private" branches without a sign-off.
Also specify empty `needs` for it, since it does not depend
on any of the jobs in the container phase and can be run
right away.
Signed-off-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Peter Krempa [Fri, 31 Jul 2020 13:23:56 +0000 (15:23 +0200)]
qemu: snapshot: Collect 'query-named-block-nodes' prior to memory migration
When doing an external snapshot we migrate memory to a file as a form of
taking the memory state. This creates a problem as qemu deactivates all
active bitmaps after a successful migration. This means that calling
'query-named-block-nodes' will return an empty list of bitmaps for
devices. We use the bitmap list to propagate the active bitmaps into the
overlay files being created which is required for backups to work after
a snapshot. Since we wouldn't propagate anything a subsequent backup
will fail with:
invalid argument: missing or broken bitmap 'testchck' for disk 'vda'
To fix this, we can simply collect the bitmap list prior to the
migration.
Michal Privoznik [Fri, 31 Jul 2020 14:48:54 +0000 (16:48 +0200)]
NEWS: Document changes for upcoming release
In contrast with other releases, this time NEWS is pretty much
updated. I've identified only a couple of features/bug fixes
worth mentioning that were not mentioned yet.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Peter Krempa [Thu, 30 Jul 2020 15:41:43 +0000 (17:41 +0200)]
tests: qemuxml2argv: Use only modern versions of 'disk-network-tlsx509' test
We already test with real caps so there's no real need for this special
case. While it technically tested the state without TLS encryption key
secrets, it doesn't really matter that much.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Use g_autoptr to free the temporary virDomainDef object created by
qemuDomainSaveInternal() when xmlin is non-NULL. Leak was added in
commit 0ea479f8f6, first appearing in libvirt 0.9.4 in August 2011.
Signed-off-by: Zheng Chuan <zhengchuan@huawei.com> Reviewed-by: Laine Stump <laine@redhat.com>
docs: fix name of file containing max number of VFs
Signed-off-by: Paulo de Rezende Pinatti <ppinatti@linux.ibm.com> Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com> Reviewed-by: Boris Fiuczynski <fiuczy@linux.ibm.com> Signed-off-by: Ján Tomko <jtomko@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
lib: clarify docs for hugetlb in virDomainMemoryStatTags
The term number is used for other stats and even for hugetlb
stats in virsh man page. The term number is also more clear.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com> Signed-off-by: Michal Privoznik <mprivozn@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
docs: virConnectGetCapabilities do not provide pool types
Remove the paragraph in the storage pool page that mentions
virConnectGetCapabilities, as virConnectGetCapabilities does not return
any information about pools.
Signed-off-by: Pino Toscano <ptoscano@redhat.com> Reviewed-by: John Ferlan <jferlan@redhat.com>
Peter Krempa [Tue, 21 Jul 2020 10:32:10 +0000 (12:32 +0200)]
tests: commandtest: Make 'test4' checking daemonization more reliable
The 'commandhelper' checks effectively whether the parent process is
still around to report whether it was daemonized or not.
This creates a unlikely race condition in cases when we do actually
daemonize the process as the intermediate process used for the
daemonization might not have terminated yet which would report wrong
result leading to test failure.
For now there's just 'test4' which actually daemonizes the process.
Add an argument '--check-daemonize' which asks for retries of the
daemonization check in cases where we expect that the commandhelper is
going to be daemonized and use it in 'test4' to make the test more
reliable.
I've observed the test failure sporadically when my box is under load
e.g. while building two trees at once.
Signed-off-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
Michal Privoznik [Thu, 23 Jul 2020 14:02:00 +0000 (16:02 +0200)]
virdevmapper: Don't use libdevmapper to obtain dependencies
CVE-2020-14339
When building domain's private /dev in a namespace, libdevmapper
is consulted for getting full dependency tree of domain's disks.
The reason is that for a multipath devices all dependent devices
must be created in the namespace and allowed in CGroups.
However, this approach is very fragile as building of namespace
happens in the forked off child process, after mass close of FDs
and just before dropping privileges and execing QEMU. And it so
happens that when calling libdevmapper APIs, one of them opens
/dev/mapper/control and saves the FD into a global variable. The
FD is kept open until the lib is unlinked or dm_lib_release() is
called explicitly. We are doing neither.
However, the virDevMapperGetTargets() function is called also
from libvirtd (when setting up CGroups) and thus has to be thread
safe. Unfortunately, libdevmapper APIs are not thread safe (nor
async signal safe) and thus we can't use them. Reimplement what
libdevmapper would do using plain C (ioctl()-s, /proc/devices
parsing, /dev/mapper dirwalking, and so on).
conf: add control over COW for storage pool directories
The storage pool code now attempts to disable COW by default on btrfs,
but management applications may wish to override this behaviour. Thus we
introduce a concept of storage pool features:
<features>
<cow state='yes|no'/>
</features>
If the <cow> feature policy is set, it will be enforced. It will always
return an hard error if COW cannot be explicitly set or unset.
Reviewed-by: Neal Gompa <ngompa13@gmail.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This calls virFileSetCOW when building a pool with a request to attempt,
but not require, COW to be disabled. The effect is that nothing changes
on non-btrfs filesystems, but btrfs will get COW disabled on the
directory. This setting is then inherited by all newly created files in
the pool, avoiding the need for mgmt app to set "nocow" on a per-volume
basis.
Reviewed-by: Neal Gompa <ngompa13@gmail.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
When disabling COW on individual files, we now use the virFileSetCOW
method. Note that this change has a slight semantic difference to the
old implementation.
The original code reported errors but returned success when disabling
COW failed.
With this new code, we will always report an error if the user requested
disabling of COW and we could not honour it, either because btrfs
returned an error, or because the filesystem is not btrfs.
Reviewed-by: Neal Gompa <ngompa13@gmail.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
util: add a helper method for controlling the COW flag on btrfs
btrfs defaults to performing copy-on-write for files. This is often
undesirable for VM images, so we need to be able to control whether this
behaviour is used.
The virFileSetCOW() will allow for this. We use a tristate, since out of
the box, we want the default behaviour attempt to disable cow, but only
on btrfs, silently do nothing on non-btrfs. If someone explicitly asks
to disable/enable cow, then we want to raise a hard error on non-btrfs.
Reviewed-by: Neal Gompa <ngompa13@gmail.com> Reviewed-by: Peter Krempa <pkrempa@redhat.com> Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
This is only used in the ESX driver where, when set to "no", it will
ignore all the checks libvirt does about the origin of the MAC address
(whether or not it's in a VMWare OUI) and forward the original one to
the ESX server telling it not to check it either.
This allows keeping a deterministic MAC address which can be useful for
licensed software which might dislike changes.
Signed-off-by: Bastien Orivel <bastien.orivel@diateam.net>
VMX conversion parts rewritten to apply on top of previously merged
support for type='generated|static'
Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
the VMX -> XML parser was not updated. As a result while we
accept the 'type' attribute on input, we never show it again
on 'output', so we loose information during the roundtrip.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
which intentionally added the 'checkMACAddress' side effect based on
the 'type' attribute.
With this change, we're reverting the handling of checkMACAddress
to match what existed historically. The 'type' attribute now directly
maps to the addressType attribute, so the above config becomes:
The mingw header define time() as a static inline function and this
causes a duplicate definition build failure. Since we're not using the
LD_PRELOAD at all on Mingw, we ideally wouldn't compile any of the
mock libraries. Rather than change the build system now though, this
just stubs out the offending function.
Reviewed-by: Laine Stump <laine@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Historically we avoided -fstack-protector* since it resulted in a broken
build on Mingw. In GCC 10 in Fedora though, we have the opposite problem,
getting a broken build if we don't enable one of the -fstack-protector*
options. This also works in GCC 9, so we don't need to worry about the
old brokeness which evidentally got fixed at some time without noticing.
Reviewed-by: Laine Stump <laine@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
gcc 10.1.0 on Debian sid has a bug where the bounds checking gets
confused beteen two branches:
In file included from /usr/include/string.h:495,
from ../../src/internal.h:28,
from ../../src/util/virsocket.h:21,
from ../../src/util/virsocketaddr.h:21,
from ../../src/util/virnetdevip.h:21,
from ../../src/util/virnetdevip.c:21:
In function 'memcpy',
inlined from 'virNetDevGetifaddrsAddress' at ../../src/util/virnetdevip.c:914:13,
inlined from 'virNetDevIPAddrGet' at ../../src/util/virnetdevip.c:962:16:
/usr/include/arm-linux-gnueabihf/bits/string_fortified.h:34:10: error: '__builtin_memcpy' offset [16, 27] from the object at 'addr' is out of the bounds of referenced subobject 'inet4' with type 'struct sockaddr_in' at offset 0 [-Werror=array-bounds]
34 | return __builtin___memcpy_chk (__dest, __src, __len, __bos0 (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from ../../src/util/virnetdevip.h:21,
from ../../src/util/virnetdevip.c:21:
../../src/util/virnetdevip.c: In function 'virNetDevIPAddrGet':
../../src/util/virsocketaddr.h:29:28: note: subobject 'inet4' declared here
29 | struct sockaddr_in inet4;
| ^~~~~
cc1: all warnings being treated as errors
Note the source location is pointing to the "inet6" / AF_INET6 branch of
the "if", but is complaining about bounds of the "inet4" field. Changing
the code into a switch() is sufficient to avoid triggering the bug and
is arguably better code too.
Reviewed-by: Laine Stump <laine@redhat.com> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Andrea Bolognani [Thu, 23 Jul 2020 08:29:35 +0000 (10:29 +0200)]
ci: Run 'make distcheck' on FreeBSD
The Cirrus CI integration was modeled after the Travis CI jobs,
but those were limited to macOS where the test suite is currently
still broken. FreeBSD can run the full distcheck just fine, so
let's do that.
Fixes: 6190c14151c3e2cf5c30b9df9131697f5c3b64b9 Signed-off-by: Andrea Bolognani <abologna@redhat.com> Reviewed-by: Ján Tomko <jtomko@redhat.com>
The reason why we align down the guest area (total-size - label-size) is
explained in the body of qemuDomainNVDimmAlignSizePseries(). This
behavior must also be documented in the user docs.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com> Reviewed-by: Andrea Bolognani <abologna@redhat.com>
projects / thirdparty / libvirt.git / log
