While parts of QEMU's CGroup code were moved under hypervisor
agnostic location (src/hypervisor/) a typo sneaked in. The
inspiration for virDomainCgroupConnectCgroup() comes from
qemuConnectCgroup(). The former is called upon reconnecting to a
running domain (after daemon restart). While the latter returned
early if the daemon was running unprivileged, the former returns
early if the daemon runs privileged. This is obviously wrong,
because root can set up CGroups.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2075765
Fixes: 788e2b58cb1896f1c25ebbdbde4bafddc5ed4dc9
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Fixes: db16792aa90cab5c9886fc2990ec13fbb20a3fb5
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Commit d73852c49962 moved the original QEMU-specific helpers to
the utils module, which resulted in build failures on non-Unix
platforms due to the unconditional use of Unix-only symbols such
as _SC_CLK_TCK.
To deal with that situation, commit d7c64453aa0e made the helpers
Linux-only and added stubs for other platforms that, when called,
would always fail with ENOSYS.
However the original helpers had been carefully written so that,
while they would only be able to produce useful output on Linux,
they would still succeed on the other Unix platforms where we
build the QEMU driver.
Restore the original behavior so that calling APIs such as
virDomainGetInfo() can once again work on FreeBSD and macOS.
Resolves: https://gitlab.com/libvirt/libvirt/-/issues/298
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Sometimes it may come handy to learn what address is a NVDIMM
mapped to inside a guest. While users can provide an address they
want to have NVDIMM mapped to, it's optional. Fortunately, when a
domain is being started we issue the 'query-memory-devices'
monitor command and the reply is the same for 'dimm' and 'nvdimm'
types. Therefore, updating NVDIMM address is trivial.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This gets rid of the goto and prepares the function for automatic
mutex management.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This allows a later patch to replace virObjectLock/Unlock
pairs with automatic mutex management code.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This allows a later patch to replace virObjectLock/Unlock
pairs with automatic mutex management code.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Checking for duplicate / NULL keys beforehand will simplify error
handling in a later patch significantly.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
[1] closes gap in virDomainObjListRemove so that concurrent thread can
not step in and obtain the domain while domain is temporary unlocked. But
there is another gap exist:
thread B - executes create API
thread C - executes undefine API
- thread A executes some job on domain
- threads B and C obtains domain from list and wait for job condition
- thread A finishes its job and C grabs job condition, removes domain
from list and finishes
- thread B grabs job condition and start the domain, unfortunately
is not in the list already
[1] commit c7d1c139ca3402e875002753952e80ce8054374e
Author: Martin Kletzander <mkletzan@redhat.com>
Date: Thu Dec 11 11:14:08 2014 +0100
qemu: avoid rare race when undefining domain
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Acquiring job introduced in commit [1] to fix a race described in the
commit. Actually it does not help because we get domain in create API
before acuiring job. Then [2] fixed the race but [1] was not reverted even
it is does not required by [2] to work properly.
[1] commit b629c64e5e0a32ef439b8eeb3a697e2cd76f3248
Author: Martin Kletzander <mkletzan@redhat.com>
Date: Thu Oct 30 14:38:35 2014 +0100
qemu: avoid rare race when undefining domain
[2] commit c7d1c139ca3402e875002753952e80ce8054374e
Author: Martin Kletzander <mkletzan@redhat.com>
Date: Thu Dec 11 11:14:08 2014 +0100
qemu: avoid rare race when undefining domain
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
SPICE ports cleanup looks overly complicated. We can just set *reserved
flags whenever port is reserved (auto or non auto).
Also *Reserved flags are not cleared on stop in case of reconnect with
autoport (flags are set on reconnect in qemuProcessGraphicsReservePorts
call). Yeah config is freed in the end of stopping domain but still.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
VNC websocket port cleanup looks a bit repetetive. Let's set websocketReserved
flag whenever we reserve port (auto or not).
Also websocketReserved flag is not cleared on stop in case of reconnect with
auto port (flags is set on reconnect in qemuProcessGraphicsReservePorts
call). Yeah config is freed in the end of stopping domain but still.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Scenario is with two domains with same VNC websocket port.
- start first domain
- start second, it will fail as port is occupied
As a result port will be released which breaks port reservation logic.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Code to release VNC port looks repetitive. The reason is there were
originally 2 functions to release ports - for auto and non-auto cases.
Also portReserved flag is not cleared on stop in case of reconnect with
auto port (flags is set on reconnect in qemuProcessGraphicsReservePorts call).
Yeah config is freed in the end of stopping domain but still.
Let's use this flag whenever we reserve port (auto or not). This makes
things clearer.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Reviewed-by: Martin Kletzander <mkletzan@redhat.com>
Move validation from virDomainDiskDefIotuneParse into the validation callback.
Signed-off-by: Moteen Shah <codeguy.moteen@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Under certain circumstances nwfilterStateInitialize could leak memory:
If e.g. the call to virNWFilterConfLayerInit fails, the error path
err_techdrivers_shutdown does not free the previously allocated memory
held in driver->stateDir.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Allow nwfilterStateCleanupLocked to be called on a partially constructed
driver object.
This enables the next patch to simplify and fix error handling in
nwfilterStateInitialize.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Allow nwfilterDriverRemoveDBusMatches to be called without
nwfilterDriverInstallDBusMatches being called previously.
This enables a later patch to use nwfilterDriverRemoveDBusMatches
as a cleanup function safely.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
virNWFilterDHCPSnoopShutdown would never destroy the mutexes created
in virNWFilterDHCPSnoopInit. Additionally, if in virNWFilterDHCPSnoopInit
the call to virMutexInitRecursive succeeds and the call to virMutexInit
fails, this would lead to either virNWFilterSnoopState.snoopLock being
initialized twice or virNWFilterSnoopState.activeLock destroyed without
being initialized first.
This enables a later patch to use virNWFilterDHCPSnoopShutdown as a
cleanup function safely, as it is a no-op if virNWFilterSnoopState was
not yet initialized.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Allow virNWFilterObjListFree to be called with a NULL argument.
This enables a later patch to use virNWFilterObjListFree as a
cleanup function safely, as it is a no-op if virNWFilterObj was
not yet initialized.
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
The capability is not used anymore since "-incoming defer" is supported
by all QEMU versions we care about.
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
All QEMU releases currently supported by libvirt already understand
"-incoming defer". We can drop the code handling "-incoming URI".
Signed-off-by: Jiri Denemark <jdenemar@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Don't try to align the output, it's not future-proof and it's for
debugging only.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The new function aggregates the internal working of virXMLFormatElement
and virXMLFormatElementEmpty and also allows skipping the newline
after the opening tag to allow using this helper also in cases where we
don't format any child elements but directly a value.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
All these features are supposed to be handled by the call to
virDriverFeatureIsGlobal() placed right above the switch
statement, so if any of them is actually encountered inside
the switch statement it means there's a bug in the driver and
we should report an error.
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The aim of 'restrictive' numatune mode is to rely solely on
CGroups to have QEMU running on configured NUMA nodes. However,
we were never setting the cpuset controller when a domain was
starting up. We are doing so only when
virDomainSetNumaParameters() is called (aka live pinning).
This is obviously wrong. Fortunately, fix is simple as
'restrictive' is similar to 'strict' - every location where
VIR_DOMAIN_NUMATUNE_MEM_STRICT occurs can be audited and
VIR_DOMAIN_NUMATUNE_MEM_RESTRICTIVE case can be added.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2070380
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This is similar to v7.10.0-354-g06f405c627 except this time it
fixes CH driver.
With strict numatune we can't guarantee that all memory is moved
to new location. Therefore, let's forbid moving memory in that
case. However, allow it for restrictive mode, which is documented
to be best effort.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Since its introduction in v1.3.2-43-gef1fa55e46 there is a dead
code in virDomainCgroupSetupGlobalCpuCgroup() (well,
qemuSetupGlobalCpuCgroup() back then). The code formats NUMA
nodeset but never sets it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The virDomainCgroupSetupVcpuBW() is a NOP if both period and
quota to set are zero. There's no need to check in all the
callers for this special case.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Instead of reporting virReportError(..., g_strerror(), ...) let's
use proper virReportSystemError(). Generated with help of cocci:
@@
expression c;
@@
<...
- virReportError(c,
+ virReportSystemError(errno,
...,
- g_strerror(errno),
...);
...>
But then I had to hand fix format strings, because I'm not sure
if cocci even knows how to do that. And even if it did, I surely
don't.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Older GCC fails to understand that 'char *main' is a variable and
not main() function. Rename the variable to appease old GCC.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Nikolay Shirokovskiy <nshirokovskiy@openvz.org>
Note that we attempt to remove logs only if virtlogd is in use.
Otherwise we do not know the pattern for rotated files.
For example for VM named "foo" we can not use "foo.log*" pattern to
remove rotated logs as we can have VM named "foo.log" with log
"foo.log.log". We can add extra check that filename does not end with
".log" but for VM "foo.log" we can have rotated log "foo.log.log.1". Ok
let's check we don't have "log" in filename part corresponging to * but
what if someone will use logrotate with "%Y.log-%m-%d" 'dateformat'
option. In this case the check will exclude proper rotated files.
Yes, the last example if quite artificial but it shows it is difficult
to find out correctly rotated files when rotated files pattern is not
known. Thus the above decision only to support case with virtlogd when
we know the pattern.
Another reason for not removing log files when logrotate is present is
that due to races some files can escape deletion. For example foo.log.3
will be rotated to foo.log.4 after removing function will read directory
files and thus foo.log.4 will not be deleted.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
If this flag is set on calling virDomainDestroyFlags flags then remove
per domain logs if possible.
This can be used by libguestfs to delete logs for temporary domain.
Otherwise such logs will stay wasting disk resources.
Signed-off-by: Nikolay Shirokovskiy <nshirokovskiy@virtuozzo.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
