Currently meson registers a single test that invokes an entire group of
I/O tests, hiding the test granularity from meson. There are various
downsides of doing this
* You cannot ask 'meson test' to invoke a single I/O test
* The meson test timeout can't be applied to the individual
tests
* Meson only gets a pass/fail for the overall I/O test group
not individual tests
* If a CI job gets killed by the GitLab timeout, we don't
get visibility into how far through the I/O tests
execution got.
This switches meson to perform test discovery by invoking 'check' in
dry-run mode. It then registers one meson test case for each I/O
test. Parallel execution remains disabled since the I/O tests do not
use self contained execution environments and thus conflict with
each other.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-8-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-25-alex.bennee@linaro.org>
The current test runner is only safe against parallel execution within
a single instance of the 'check' process, and only if -j is given a
value greater than 2. This prevents running multiple copies of the
'check' process for different test scenarios.
This change switches the output / socket directories to always include
the test name, image format and image protocol. This should allow full
parallelism of all distinct test scenarios. eg running both qcow2 and
raw tests at the same time, or both file and nbd tests at the same
time.
It would be possible to allow for parallelism of the same test scenario
by including the pid, but that would potentially let many directories
accumulate over time on failures, so is not done.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-7-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-24-alex.bennee@linaro.org>
Currently the tests have their stdin inherited from the test harness,
meaning they are connected to a TTY. The QEMU processes spawned by
certain tests, however, modify TTY settings and if the test exits
abnormally the settings might not be restored.
The python test harness thus has some logic which will capture the
initial TTY settings and restore them once all tests are finished.
This does not, however, take into account the possibility of many
copies of the 'check' program running in parallel. With parallel
execution, a later invokation may save the TTY state that QEMU has
already modified, and thus restore bad state leaving the TTY
non-functional.
None of the I/O tests shnould actually be interactive requiring
user input and so they should not require a TTY at all. To avoid
this while TTY save/restore complexity we can connect the test
stdin to /dev/null instead.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-6-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-23-alex.bennee@linaro.org>
When asking 'check' to list individual tests by invoking it in dry run
mode, it prints the paths to the tests relative to the base of the
I/O test directory.
When asking 'check' to run an individual test, however, it mandates that
only the unqualified test name is given, without any path prefix. This
inconsistency makes it harder to ask for a list of tests and then invoke
each one.
Thus the test listing code is change to flatten the test names, by
printing only the base name, which can be directly invoked.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-4-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-21-alex.bennee@linaro.org>
The 'check' script can be invoked in "dry run" mode, in which case it
merely does test discovery and prints out all their names. Despite only
doing test discovery it still validates that the various QEMU binaries
can be found. This makes it impossible todo test discovery prior to
building QEMU. This is a desirable feature to support, because it will
let meson discover tests.
Fortunately the code in the TestEnv constructor is ordered in a way
that makes this fairly trivial to achieve. We can just short circuit
the constructor after the basic directory paths have been set.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-3-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-20-alex.bennee@linaro.org>
The 'check' script has some rather dubious logic whereby it assumes
that if invoked as a symlink, then it is running from a separate
source tree and build tree, otherwise it assumes the current working
directory is a combined source and build tree.
This doesn't work if you want to invoke the 'check' script using
its full source tree path while still using a split source and build
tree layout. This would be a typical situation with meson if you ask
it to find the 'check' script path using files('check').
Rather than trying to make the logic more magical, add support for
explicitly passing the dirs using --source-dir and --build-dir. If
either is omitted the current logic is maintained.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Hanna Czenczek <hreitz@redhat.com>
Tested-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20230303160727.3977246-2-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-19-alex.bennee@linaro.org>
The openbsd image is 20GB in size, but the automatic partitioning
done by the installer leaves /home with a mere ~3.5 GB of space,
wasting free space across many other partitions that are not
used by our build process:
openbsd$ df
Filesystem 512-blocks Used Avail Capacity Mounted on
/dev/sd0a 1229692 213592 954616 18% /
/dev/sd0k 7672220 40 7288572 0% /home
/dev/sd0d 1736604 24 1649752 0% /tmp
/dev/sd0f 4847676 2505124 2100172 54% /usr
/dev/sd0g 1326684 555656 704696 44% /usr/X11R6
/dev/sd0h 4845436 1445932 3157236 31% /usr/local
/dev/sd0j 10898972 4 10354020 0% /usr/obj
/dev/sd0i 3343644 4 3176460 0% /usr/src
/dev/sd0e 2601212 19840 2451312 1% /var
This change tells the installer todo custom partitioning with
4 GB on /, 256 MB swap, and the remaining ~15GB for /home
openbsd$ df
Filesystem 512-blocks Used Avail Capacity Mounted on
/dev/sd0a 7932412 4740204 2795588 63% /
/dev/sd0d 32164636 40 30556368 0% /home
This will avoid ENOSPC failures when tests that need to create
big files (disk images) run in parallel.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230322123639.836104-3-berrange@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
We are abusing the avocado tags which are intended to provide test
selection metadata to provide parameters to our test. This works OK up
until the point you need to have ,'s in the field as this is the tag
separator character which is the case for a number of the drive
parameters. Fix this by making drive a parameter to the common helper
function.
Fixes: 267fe57c23 (tests: add tuxrun baseline test to avocado)
Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-11-alex.bennee@linaro.org>
This was broken when we moved to using the pre-built packages as we
didn't take care to ensure we used RPMs where required.
NB: I could never get this to complete on my test setup but I suspect
this was down to network connectivity and timeouts while downloading.
Fixes: 69c4befba1 (scripts/ci: update gitlab-runner playbook to use latest runner)
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230315174331.2959-5-alex.bennee@linaro.org>
The core of the test was utilising "ethtool -t eth1 offline" to run
through a test sequence. For reasons unknown the test hangs under some
configurations of the build on centos8-stream. Fundamentally running
the old fedora-31 cloud-init is just too much for something that is
directed at testing one device. So we:
- replace fedora with a custom kernel + buildroot rootfs
- rename the test from IGB to NetDevEthtool
- re-factor the common code, add (currently skipped) tests for other
devices which support ethtool
- remove the KVM limitation as its fast enough to run in KVM or TCG
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Cc: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-Id: <20230322145529.4079753-1-alex.bennee@linaro.org>
A recent attempt to let avocado run more tests on the CentOS stream
build failed because there was no gating on the multiprocess feature.
Like missing accelerators avocado should gracefully skip when the
feature is not enabled.
In this case we use the existence of the proxy device as a proxy for
multi-process support.
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Cc: Elena Ufimtseva <elena.ufimtseva@oracle.com>
Cc: Jagannathan Raman <jag.raman@oracle.com>
Cc: John G Johnson <john.g.johnson@oracle.com>
Message-Id: <20230321111752.2681128-1-alex.bennee@linaro.org>
Unfortunately a bug in older versions of gdb means that they will
crash if QEMU sends them the aarch64-pauth.xml. This bug is fixed in
gdb commit 1ba3a3222039eb25, and there are plans to backport that to
affected gdb release branches, but since the bug affects gdb 9
through 12 it is very widely deployed (for instance by distros).
It is not currently clear what the best way to deal with this is; it
has been proposed to define a new XML feature name that old gdb will
ignore but newer gdb can handle. Since QEMU's 8.0 release is
imminent and at least one of our CI runners is now falling over this,
disable the pauth XML for the moment. We can follow up with a more
considered fix either in time for 8.0 or else for the 8.1 release.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Since its inception elf2dmp has checked MZ signatures within an
address space above IDT[0] interrupt vector and took first PE image
found as Windows Kernel.
But in Windows Server 2022 memory dump this address space range is
full of invalid PE fragments and the tool must check that PE image
is 'ntoskrnl.exe' actually.
So, introduce additional validation by checking image name from
Export Directory against 'ntoskrnl.exe'.
Signed-off-by: Viktor Prutyanov <viktor@daynix.com>
Tested-by: Yuri Benditovich <yuri.benditovich@daynix.com>
Reviewed-by: Annie Li <annie.li@oracle.com>
Message-id: 20230222211246.883679-4-viktor@daynix.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This workaround was put in place in the original implementation almost
10 years ago, considering a very old SDL2 version. Currently it prevents
users to run in a wayland-only environment without manually forcing the
backend.
The SDL2 wayland backend has been supported by distributions for a very
long time (e.g. in Fedora, first available 8 years ago), and is now
considered stable and becoming the default for new SDL2 releases.
Instead of requiring the x11 backend to exist by default, let new qemu
releases run with the default chosen by the installed SDL2 version.
Signed-off-by: Erico Nunes <ernunes@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20230301141205.514338-1-ernunes@redhat.com>
Do not attempt to move the pointer if the widget is not yet realized.
The mouse cursor is placed to the corner of the screen, on X11 at least,
as x_root and y_root are then miscalculated. (this is not reproducible
on Wayland, because Gtk doesn't implement device warping there)
This also fixes the following warning at start:
qemu: Gdk: gdk_window_get_root_coords: assertion 'GDK_IS_WINDOW (window)' failed
Fixes: 6effaa16ac ("ui: set cursor position upon listener
registration")
Reported-by: Bernhard Beschow <shentey@gmail.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Tested-by: Bernhard Beschow <shentey@gmail.com>
Message-Id: <20230320132624.1612464-1-marcandre.lureau@redhat.com>
