The information about sockets having different label than the one on the file
and the way it needs to be set is very difficult to find for those who did not
come across it before. Let's describe what needs to happen in order for the
migration to go through rather than rely on general knowledge of others.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Even though it is technically possible, when running the migrations QEMU's
nbd-server-start errors out with:
"TLS is only supported with IPv4/IPv6"
We can always enable it when QEMU adds this feature, but for now it is safer to
show our error message rather than rely on QEMU to error out properly.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
This reverts commit b3710e9a2a.
That check is very valuable for our code, but it causes issue with glib >=
2.67.0 when building with clang.
The reason is a combination of two commits in glib, firstly fdda405b6b1b which
adds a g_atomic_pointer_{set,get} variants that enforce stricter type
checking (by removing an extra cast) for compilers that support __typeof__, and
commit dce24dc4492d which effectively enabled the new variant of glib's atomic
code for clang. This will not be necessary when glib's issue #600 [0] (8 years
old) is fixed. Thankfully, MR #1719 [1], which is supposed to deal with this
issue was opened 3 weeks ago, so there is a slight sliver of hope.
[0] https://gitlab.gnome.org/GNOME/glib/-/issues/600
[1] https://gitlab.gnome.org/GNOME/glib/-/merge_requests/1719
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The cpu mask was free()'d immediately on any error and at the end of the
function, where it was expected that it would either error out and return or
goto another allocation if the code was to fail. However since commit
9514e24984 the error path did not return in one new case which caused
double-free in such situation. In order to make the code more straightforward
just free the mask after it's been used even before checking the return code of
the call.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1819801
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
This is just a clean-up of commit 3791f29b08 using the new parameter of
virProcessSetAffinity() introduced in commit 9514e24984 so that there is
no error reported in the logs.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This allows:
a) migration without access to network
b) complete control of the migration stream
c) easy migration between containerised libvirt daemons on the same host
Resolves: https://bugzilla.redhat.com/1638889
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Consider a host with 8 CPUs. There are the following possible scenarios
1. Bare metal; libvirtd has affinity of 8 CPUs; QEMU should get 8 CPUs
2. Bare metal; libvirtd has affinity of 2 CPUs; QEMU should get 8 CPUs
3. Container has affinity of 8 CPUs; libvirtd has affinity of 8 CPus;
QEMU should get 8 CPUs
4. Container has affinity of 8 CPUs; libvirtd has affinity of 2 CPus;
QEMU should get 8 CPUs
5. Container has affinity of 4 CPUs; libvirtd has affinity of 4 CPus;
QEMU should get 4 CPUs
6. Container has affinity of 4 CPUs; libvirtd has affinity of 2 CPus;
QEMU should get 4 CPUs
Scenarios 1 & 2 always work unless systemd restricted libvirtd privs.
Scenario 3 works because libvirt checks current affinity first and
skips the sched_setaffinity call, avoiding the SYS_NICE issue
Scenario 4 works only if CAP_SYS_NICE is availalbe
Scenarios 5 & 6 works only if CAP_SYS_NICE is present *AND* the cgroups
cpuset is not set on the container.
If libvirt blindly ignores the sched_setaffinity failure, then scenarios
4, 5 and 6 should all work, but with caveat in case 4 and 6, that
QEMU will only get 2 CPUs instead of the possible 8 and 4 respectively.
This is still better than failing.
Therefore libvirt can blindly ignore the setaffinity failure, but *ONLY*
ignore it when there was no affinity specified in the XML config.
If user specified affinity explicitly, libvirt must report an error if
it can't be honoured.
Resolves: https://bugzilla.redhat.com/1819801
Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Local socket connections were outright disabled because there was no "server"
part in the URI. However, given how requirements and usage scenarios are
evolving, some management apps might need the source libvirt daemon to connect
to the destination daemon over a UNIX socket for peer2peer migration. Since we
cannot know where the socket leads (whether the same daemon or not) let's decide
that based on whether the socket path is non-standard, or rather explicitly
specified in the URI. Checking non-standard path would require to ask the
daemon for configuration and the only misuse that it would prevent would be a
pretty weird one. And that's not worth it. The assumption is that whenever
someone uses explicit UNIX socket paths in the URI for migration they better
know what they are doing.
Partially resolves: https://bugzilla.redhat.com/1638889
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Adds new typed param for migration and uses this as a UNIX socket path that
should be used for the NBD part of migration. And also adds virsh support.
Partially resolves: https://bugzilla.redhat.com/1638889
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
For this we need to make the function accessible (at least privately). The
behaviour will change in following patches and the test helps explaining the
change.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Clean up the semantics by using one extra self-describing variable.
This also fixes the port allocation when the port is specified.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
Instead of saving some data from a union up front and changing an overlayed
struct before using said data, let's just set the new values after they are
decided. This will increase the readability of future commit(s).
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Jiri Denemark <jdenemar@redhat.com>
There is no distinction between Read/Write locks for resctrl from libvirt's
point of view any more.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
It was created to get rid of conditional compilation in the resctrl code and
make it usable anywhere else. However this is not something that is going to be
used in other places because it is not portable and resctrl is just very
specific in this regard. And there is no reason why there could not be a
preprocessor conditional in the resctrl code. Also the interface of
virFileFlock() was very ambiguous which lead to some issues.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
That's the way it should've been all the time. It was originally the case, but
then the rework to virFileFlock() made the function ambiguous when it was
created in commit 5a0a5f7fb5, and due to that it was misused in commit
657ddeff23 and since then the lock being taken was shared rather than
exclusive.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
When preparing for the removal of GNULIB commit 18dca21a32 removed the
unneeded O_DIRECTORY, but unfortunately started opening the directory for
writing which fails every time for a directory. There is also no need for that
as flock() works on O_RDONLY file descriptor as well, even for LOCK_EX.
https://bugzilla.redhat.com/1852741
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
I guess the reason for that was the automatic interpretation/stringification of
setfilecon_errno, but the code was not nice to read and it was a bit confusing.
Also, the logs and error states get cleaner this way.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
It will not be executed when the page is loaded locally. It needs
planet.virt-tools.org to supply the right headers (which it does now).
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Don't use the global namespace, unify quotes and semicolons at the end of lines
and "use strict".
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
If the scheduler is set before vCPU0 cannot be moved into its cpu,cpuacct
cgroup. While it is not yet known whether this is a bug or not, it makes sense
for us to do that later as otherwise the scheduler would be inherited by vCPU
and I/O Threads even when they do not have any such setting specified.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
This way more users can run our CI builds locally.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
The way it works now the Makefile needs to be both make valid and automake
valid. That is fine for now, but if we want to use anything more advanced, like
conditionals, we cannot have it like that any more.
So instead forward all ci-* rules to that file.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
This helps in a scenarios where vCPUs run with a priority that is so high they
might starve the emulator thread. And it also fits with the rest of the
settings.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Just one missing occurrence of iothreadsched fixed plus some rewording for this
to make more sense for the readers.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This will be used later when we want to format emulator scheduler parameters
which don't apply for multiple threads.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This will become useful later when parsing emulatorsched parameters which don't
need the rest of the current function.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
This is a zero-cost workaround for a bug in GCC 8.3.0 which causes the
compilation to fail, because the compiler thinks that the value might be used
uninitialized even though it clearly cannot be.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
This does not cause a problem in usual scenarios thanks to us allowing
CAP_DAC_OVERRIDE for the qemu process, however in some scenarios this might be
an issue because the directory is created with mkdtemp(3) which explicitly
creates that with 0700 permissions and qemu running as non-root cannot access
that.
The scenarios include:
- Builds without CAPNG
- Running libvirtd in certain container configurations [1]
- and possibly others.
[1] https://github.com/kubevirt/kubevirt/pull/2181#issuecomment-481840304
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Similarly to CAT, when you set some values in an group, remove the group and
recreate it, the previous values will be kept there. In order to not get values
from a previous setting (a previous VM, for example), we need to set them to
sensible defaults. The same way we do that for CAT, just set the same values as
the default group has.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
For CAT we calculate unallocated parts of the cache, however with MBA this does
not make sense as the purpose of that is to limit the bandwidth and the setting
is only proportional relative to bandwidth settings for other groups.
This means it makes sense to set the values to 100% even if there are other
groups with some allocations and that we don't need to find the available
(unallocated) bandwidth in all the groups.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Since the functions only return 0 or 1, they should return bool. I missed the
change when "refactoring" the first commit.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
We already have that in the code (commit c1bc9c662b), we just forgot to
mention that in the docs.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
It is not a problem at all if the `tss` user/group does not exist, the code
fallbacks to the `root` user/group. However we report a warning for no reason
on every start-up. Fix this by checking if the user/group actually exists.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Instead of duplicating the code from virGet{User,Group}IDByName(), which are
static anyway, extend those functions to accept NULL pointers for the result and
a boolean for controlling the error reporting.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
The proper file that should be included is `sys/xattr.h` as that comes from
`glibc` and not `attr/xattr.h` which ships with the `attr` utility.
We're most probably not the only ones because `attr/xattr.h` added a #warning to
their include resulting in the following compilation errors:
In file included from securityselinuxlabeltest.c:31:0:
/usr/include/attr/xattr.h:5:2: error: #warning "Please change your <attr/xattr.h> includes to <sys/xattr.h>" [-Werror=cpp]
#warning "Please change your <attr/xattr.h> includes to <sys/xattr.h>"
^~~~~~~
In file included from securityselinuxhelper.c:37:0:
/usr/include/attr/xattr.h:5:2: error: #warning "Please change your <attr/xattr.h> includes to <sys/xattr.h>" [-Werror=cpp]
#warning "Please change your <attr/xattr.h> includes to <sys/xattr.h>"
^~~~~~~
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
It was used just temporarily to do a calculation, no need to keep that around.
Also use virBitmap in the code instead of reimplementing two of its existing
functions. And move the counting part next to where the value is read.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
It will be used in that file later on, plus it makes sense for all the
implementations to be in same place. Also comment each one of them nicely and
add a comment explaining why they all need to end with the same _LAST value.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
There is no need to have virResctrlGetInfo() when it must be called after
virResctrlInfoNew() anyway, otherwise it's just an unusable object. When we
wrap the logic inside the New() function we'll save some calls later as well.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Move description of the purpose of the file before any definition.
One empty line between related enum definitions.
All typedefs before all structs. This is exception from the usual, but not the
only one, we already have something similar for some other structs. This way we
can move contents between structs and reorder some parts nicely without moving
all definitions of one type before another one just so it's defined.
Define all classes in one place.
Have one initialization function for all classes in the file.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
That way we get rid of the last preprocessor conditional so the code compiles on
all platforms.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
We already have virFileLock(), but we are now using flock() in the code as
well (due to requirements for mutual exclusion between libvirt and other
programs using flock() as well), so let's have a function for that as well so we
don't need to have stubs for unsupported platforms in other files.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
For getting the reply I queried the newest and oldest QEMU using
test/qemucapsprobe. From the differences I only extracted the reply to the new
QMP command and discarded the rest. For all the versions below the one which
added support for the new option I used the output from the oldest QEMU release
and for those that support it I used the output from the newest one.
In order to make doubly sure the reply is where it is supposed to be (the
replies files are very forgiving) I added the property to all the replies files,
reran the tests again and fixed the order in replies files so that all the
versions are reporting the new capability. Then removed that one property.
After that I used test/qemucapsfixreplies to fix the reply IDs.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
TSEG (Top of Memory Segment) is one of many regions that SMM (System Management
Mode) can occupy. This one, however is special, because a) most of the SMM code
lives in TSEG nowadays and b) QEMU just (well, some time ago) added support for
so called 'extended' TSEG. The difference to the TSEG implemented in real q35's
MCH (Memory Controller Hub) is that it can offer one extra size to the guest OS
apart from the standard TSEG's 1, 2, and 8 MiB and that size can be selected in
1 MiB increments. Maximum may vary based on QEMU and is way too big, so we
don't need to check for the maximum here. Similarly to the memory size we'll
leave it to the hypervisor to try satisfying that and giving us an error message
in case it is not possible.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Acked-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
One of the things that this is improving is the fact that instead of error
message (that was wrong) you get when starting a domain with SMM and i440fx we
allow the setting to go through. SMM option exists and makes sense on i440fx as
well (basically whenever that _SMM_OPT capability is set).
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
We are still hoping all of such checks will be moved there and this is one small
step in that direction.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The default is actually `on` when `<smm/>` is specified.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: John Ferlan <jferlan@redhat.com>
These two functions were duplicating some cleanup paths, so let's just merge
both cleanup and error paths together. To distinguish whether we need to
clean-up the return value let's keep it in @tmp until the function is successful
in which case we set @ret to the value of @tmp and set @tmp to NULL.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Since we started using `make distcheck` in travis, the part that's printing the
tests/test-suite.log file is rather pointless. Let's check for both known
locations (keeping the previous one there on purpose) so that the output is
usable again.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
That is a job of libvirtd and virtlogd has a dependency on it, so that will
prevent it properly. Doing it one extra time in virtlogd might also cause AVC
denials because it is not allowed to call that dbus method.
Caused by commit df34363d58.
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1547250
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Due to conditional execution of virTestRun(), the testCounter was incremented
only if all the cases were run. When using VIR_TEST_RANGE=x-y, first x/2 of the
increments were skipped and that made figuring out a precise case a PITA.
Moving the condition into the test function makes it way nicer to find out the
test numbers to use in VIR_TEST_RANGE.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Since one of the things in capabilities (info from resctrl updated with data
about caches) can be change on the system by remounting the /sys/fs/resctrl with
different options, the capabilities need to be refreshed. There is a better fix
in the works, but it's going to be way bigger than this (hence the XXX note
there), so for the time being let's workaround this. And in order not to slow
down the domain starting, only get the capabilities if there are any cachetunes.
Relates-to: https://bugzilla.redhat.com/show_bug.cgi?id=1540780
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Cachetune for unavailable vCPUs should be cleared the same way vcpupin and other
things do, so let's add tests for it.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Some of the other functions depend on the fact that unused bits and longs are
always zero and it's less error-prone to clear it than fix the other functions.
It's enough to zero out one piece of the map since we're calling realloc() to
get rid of the rest (and updating map_len).
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1540817
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
During reconnect we need to reconstruct the paths of all cachetunes so that they
get cleaned up when the domain is stopped.
Signed-off-by: Martin Kletzander <mkletzan@redhat.com>
Martin Kletzander