| Age | Commit message (Collapse) | Author |
|---|
|
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Add busy and busy-avg balancers which make balancing decisions by looking
at engine busyness via the i915 PMU.
And thus are able to make decisions on the actual instantaneous load of
the system, and not use metrics that lag behind by a batch or two. In
doing so, each client should be able to greedily maximise their own
usage of the system, leading to improved load balancing even in the face
of other uncooperative clients. On the other hand, we are only using the
instantaneous load without coupling in the predictive factor for dispatch
and execution length.
v2:
* Commit text. (Chris Wilson)
* Rename get_stats to get_pmu_stats. (Chris Wilson)
* Fix PMU readout in VCS remap mode.
v3:
* Integrated Petri's meson build recipe.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Petri Latvala <petri.latvala@intel.com>
|
|
A bunch of tests for the new i915 PMU feature.
Parts of the code were initialy sketched by Dmitry Rogozhkin.
v2: (Most suggestions by Chris Wilson)
* Add new class/instance based engine list.
* Add gem_has_engine/gem_require_engine to work with class/instance.
* Use the above two throughout the test.
* Shorten tests to 100ms busy batches, seems enough.
* Add queued counter sanity checks.
* Use igt_nsec_elapsed.
* Skip on perf -ENODEV in some tests instead of embedding knowledge locally.
* Fix multi ordering for busy accounting.
* Use new guranteed_usleep when sleep time is asserted on.
* Check for no queued when idle/busy.
* Add queued counter init test.
* Add queued tests.
* Consolidate and increase multiple busy engines tests to most-busy and
all-busy tests.
* Guarantte interrupts by using fences.
* Test RC6 via forcewake.
v3:
* Tweak assert in interrupts subtest.
* Sprinkle of comments.
* Fix multi-client test which got broken in v2.
v4:
* Measured instead of guaranteed sleep.
* Missing sync in no_sema.
* Log busyness before asserts for debug.
* access(2) instead of open(2) to determine if cpu0 is hotpluggable.
* Test frequency reporting via min/max setting instead assuming.
^^ All above suggested by Chris Wilson. ^^
* Drop queued subtests to match i915.
* Use long batches with fences to ensure interrupts.
* Test render node as well.
v5:
* Add to meson build. (Petri Latvala)
* Use 1eN constants. (Chris Wilson)
* Add tests for semaphore and event waiting.
v6:
* Fix interrupts subtest by polling the fence from the "outside".
(Chris Wilson)
v7:
* Assert number of initialized engines matches the expectation.
(Chris Wilson)
* Warn instead of skipping if we couldn't restore the initial
frequency. (Chris Wilson)
* Move all asserts to after the test cleanup (just a tidy).
* More 1eN notation for timeouts.
* Bump the tolerance to 5% since I saw a few noisy runs with
sampling counters.
* Always start the PMU before submitting batches to lower
reliance on i915 doing the delayed engine busy stats disable.
v8:
* Update for upstream engine class enum.
v9:
* Add meson build support.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Wire up to the RAPL PMU for GPU energy readings.
The only complication is that we have to add code to parse:
# cat /sys/devices/power/events/energy-gpu.scale
2.3283064365386962890625e-10
v2: Link with -lm.
v3: strtod can handle scientific notation, even though my initial
reading of the man page did not spot that. (Chris Wilson)
v4: Meson fix.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
v2: Update for i915 changes.
v3: Use 1eN for large numbers. (Chris Wilson)
v4: Update for upstream engine class enum.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
v2: Use correct address of. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Configuration and format are uint64_t in the perf API.
Tidy some other details as well.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Various tool modules implement their owm PMU open wrapper which
can be replaced by calling the library one.
v2:
* Remove extra newline. (Chris Wilson)
* Commit msg.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Idea is to avoid duplication across multiple users in
upcoming patches.
v2: Commit message and use a separate library instead of piggy-
backing to libintel_tools. (Chris Wilson)
v3: Add Petri's meson build recipe.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Petri Latvala <petri.latvala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
The rotation property sucks because it may affect whether
drmModeSetPlane succeeds or not. Add some code to handle
this.
First try to set rotation directly, if that succeeds we
return immediately. If it fails we disable the plane, set
the rotation property and run the rest of the code.
This will hopefully make legacy rotation work in more cases when
scaling is not supported.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
This fixes the testlist keeping the out-of-date reference to chamelium on last run.
Changes since v1:
- Remove reference to Makefile.sources, this is already
handled by the dependency on Makefile.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reported-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Cc: Tomi Sarvela <tomi.p.sarvela@intel.com>
Acked-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Reviewed-by: Petri Latvala <petri.latvala@intel.com>
|
|
Code on loop() function shadowed function parameter which led
to inability to try out different engines in the test: we always
loaded RCS0.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103804
Signed-off-by: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
So far this test is basically making sure that we throw appropriate
errors, and don't oops the kernel with silly inputs.
v2: Add to meson build.
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
|
|
This just checks that the appropriate errors get thrown, and that the
modifier can be set/get successfully, and that the modifier doesn't
leak to other BO allocations. Testing of scanout will be done with
the writeback support that Boris is building. The modifier has no
effect on V3D rendering, so no need to test that.
v2: Add to the meson build.
Signed-off-by: Eric Anholt <eric@anholt.net>
|
|
The MMAP_V2 is replaced by just using MMAP, since the official header
has the updated struct. The gem_create_v2 and gem_get_aperture are
left as is, because they seem to not be reflected in the UABI header!
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Petri Latvala <petri.latvala@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Petri Latvala <petri.latvala@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
The autotools build retains the configure.ac option, while meson folds
vc4 into the default build since we don't have any meson_options.txt
to control parts of the build.
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Petri Latvala <petri.latvala@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
Tested by dropping garbage in my libdrm's headers and rebuilding.
v2: Pull in DRM_CFLAGS movement that ended up later in the sequence.
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Petri Latvala <petri.latvala@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
These will be used for tests that want to use a local copy of the
headers instead of depending on libdrm releases. Taken from drm-next
d65d31388a23 ("Merge tag 'drm-misc-next-fixes-2017-11-07' of
git://anongit.freedesktop.org/drm/drm-misc into drm-next")
See README for rules on updating these headers.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Petri Latvala <petri.latvala@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
|
|
Having registers for nonexistent planes in the dumpo might end up being
rather confusing. Try to only include real planes.
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Print hex numbers with "0x" prefix, and make the output a bit more
compact.
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Added support to print SKL watermark and DDB registers.
v2: Printed raw register data, renamed planes and combined two printf()'s
(Ville)
v3: s/drm_fd/-1/ for intel_register_access_init()
Signed-off-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
[vsyrjala: make it build, fix indentation, etc.]
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Let's print the linetime watermarks un usecs. Might make it easier to
spot bogus values.
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Acked-by: Martin Peres <martin.peres@linux.intel.com>
|
|
Instead of first calling kmstest_unset_all_crtcs, and calling
igt_display_init for each test, use igt_display_reset to reset
igt_display between tests, and use atomic commit to disable all
unused crtcs in enable_output().
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
|
|
igt_output_from_connector should be used for disconnected outputs
too, this is useful for chamelium testing, where disconnected outputs
may reappear.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
|
|
A lot of code duplicates this, but it should be handled in the core.
Add it and use it after igt_display_init(), the tests have to be
converted one by one.
Changes since v1:
- Merely reset rotation to HW value.
Changes since v2:
- Compile.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
|
|
[1/316] Compiling C object 'overlay/intel-gpu-overlay@exe/overlay.c.o'.
../overlay/overlay.c: In function ‘show_gem_objects’:
../overlay/overlay.c:798:31: warning: ‘%s’ directive output may be truncated writing up to 255 bytes into a region of size 160 [-Wformat-truncation=]
snprintf(buf, sizeof(buf), "%s %ldMB, %ld objects",
^~
../overlay/overlay.c:798:30: note: using the range [-9223372036854775808, 9223372036854775807] for directive argument
snprintf(buf, sizeof(buf), "%s %ldMB, %ld objects",
^~~~~~~~~~~~~~~~~~~~~~~
../overlay/overlay.c:798:30: note: using the range [-9223372036854775808, 9223372036854775807] for directive argument
../overlay/overlay.c:798:3: note: ‘snprintf’ output between 16 and 309 bytes into a destination of size 160
snprintf(buf, sizeof(buf), "%s %ldMB, %ld objects",
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
comm->name, comm->bytes >> 20, comm->count);
Might as well give it the 309 bytes it wants, plus on more. :-)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
|
Since the tools want to work without the module loaded, remove the
assumption that we want to load the driver to find debugfs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjala <ville.syrjala@linux.intel.com>
|
|
Look at the first 16 characters of the user buffer and if they are all
printable, assume that user buffer contains ASCII data (e.g. a debug log)
that we want to simply print out. Otherwise, we treat it as a binary data
and hexdump as before.
Tvrtko: you know you could just write this as a printf (d'oh)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
|
|
If we can't find the enable_execlists parameter, presume that the switch
is forced by the kernel and enabled for all hw supporting execlists. We
don't have a GETPARAM or ENGINE_INFO to query the internal details.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
|
|
I noticed that dpm was storing some information about which phase of
suspend failed inside suspend_stats. That will be useful to help debug
such failures, so automatically dump it after suspend fails.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
|
|
Since CI runs each subtest individually, we do not get the batching of
tests and execute every fixture around each subtest. The consequence of
this for CI is that we quickly exhaust static allocations like lockdep's
array of lockclasses, or causing hash conflicts with new locks being
reallocation into existing addresses, the result is a warning from
lockdep and ts disabling.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Petri Latvala <petri.latvala@intel.com>
|
|
If we asked to open a particular chipset and we find no matching device,
try again after attempting to load its module. Previously we only did
this for vgem, which is not automatically probed during boot, but if we
want to leave the module unloaded we have to try harder when we need the
device.
v2: DRIVER_* are already masks (and not shifts). Use a common
driver_open for both /dev/dri/cardX and /dev/dri/renderDX.
v3: Beware making local variables accidentally static scoped.
v4: Beware multiple threads trying and failing to open a device
v5: Fixed spelling of render (Petri)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Petri Latvala <petri.latvala@intel.com>
Reviewed-by: Petri Latvala <petri.latvala@intel.com>
|
|
is_mountpoint() asserts rather than report the error. Normally this
isn't a problem, except for atypical selftests.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
|
|
This reverts commit 7296e09ee7f17e6d564e52cf64ee900670849429.
This commit was accidentally applied twice, the first time in
commit e4ba3b75e6de35483b2edea21ceda145ef0b3311.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
CC: Daniel Vetter <daniel.vetter@ffwll.ch> #irc
|
|
Wherever possible we want to use existing ioctls rather than poking the
volatile modparams, in this case we already provide
I915_PARAM_HAS_SEMAPHORES so we have no excuse to peek i915.semaphores.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
|
|
Some debugging information is too voluminous to enable by default, yet
may be invaluable when it comes to post-mortem debugging. trace_printk()
provides the facility for the trace ringbuffer to be dumped on oops,
this way we can cheaply spam the debug log and only present it in case
of emergency.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Petri Latvala <petri.latvala@intel.com>
|
|
Doing modeset on internal panels may have a considerable overhead due to
the panel specific power sequencing delays. To avoid long test runtimes
in CI split out the testing of internal panels from the plane modeset
subtests and test only a reduced number of plane combinations on these:
where only a single plane is enabled, all planes are disabled or all
planes are enabled.
v2:
- Add code comment explaining the need for the separate internal-panels
subtests.
- Calculate the number of enabled planes only once. (Chris)
- Disable the slow internal-panels subtests testing all combinations.
(Petri)
v3: (Maarten)
- Skip the internal-panels subtests if no internal panels are found.
- Also test transitions with all planes disabled.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103334
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Petri Latvala <petri.latvala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
|
|
knowing the assertion triggered on wait_until_enabled() is not that
useful without knowing what exactly caused the failure. It might be an
user error, like too little stolen memory by the bios, or an actual
issue in the kernel. So, let's make life easier, particularly for the
CI, by printing the status before failing out.
Case in point:
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101718
v2: Small typo.
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
|
|
Commit ca20170afc6f ("tests/kms_plane_scaling: Add support for dynamic
number of planes") shifted the tested planes by one after the
refactoring, accidentally ignoring the first plane, which is zero
indexed. A symptom of the issue appears on KBL, where the third plane
is already the shared cursor, causing igt to configure an unsupported
framebuffer format on it, triggering the following error:
[drm:__setplane_internal] Invalid pixel format XR24 little-endian (0x34325258)
With this fixed, we can exposes the pixel clock scaling issue, which is
the actual problem being tracked in Bug 103159, but let's start by
reverting to the old behavior.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103159
Fixes: ca20170afc6f ("tests/kms_plane_scaling: Add support for dynamic number of planes")
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
|
|
Use TEST_NONE instead of TEST_DEVICES to prevent a machine death that
happens on a particular model of SNB (2600 is affected, 2520m is
not). Reset is unreliable, but the exact setup to trigger the death
and how to work around it are not found at this time. There is some
kind of a race lurking, and this commit is a workaround that avoids
it, leaving the test still exercising some of the codepaths.
References: https://intel-gfx-ci.01.org/tree/drm-tip/igt@gem_eio@in-flight-suspend.html
References: https://bugs.freedesktop.org/show_bug.cgi?id=103289
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Petri Latvala <petri.latvala@intel.com>
CC: Daniel Vetter <daniel.vetter@ffwll.ch>
CC: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
CC: "Lofstedt, Marta" <marta.lofstedt@intel.com>
CC: Martin Peres <martin.peres@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
|
|
"other" was misspelled as "otehr". Fix it.
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
The meson build avoids the failure by excluding the igt_x86.c file from
the compilation. autotools being what they are don't support that in an
easy way, so just use the preprocessor to avoid the duplicate function
definitions.
Since igt_x86.c will now be ignored for non-x86 builds, the meson work-
around can be removed.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
GCC will warn about ISO C90 forbidding to mix declarations and code.
Declare the variable at the beginning of the function to suppress the
warning.
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
Data types are defined differently on 32-bit systems, causing gcc to
complain about printf format specifiers not matching the size of the
variables passed in. Use PRIu64 and %zu where appropriate.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
|
|
List the available tests with piglit instead of by hand. This solves
naming inconsistencies (piglit throwing caps away) as seen by
cibuglog, and makes the listing code simpler.
The format of the listing changes from
test-binary/subtest-name
to
igt@test-binary@subtest-name
but so far nothing has been able to directly consume run-tests.sh -l
output. The piglit format is directly consumable by piglit --test-list, and thus
by run-tests.sh -T.
Signed-off-by: Petri Latvala <petri.latvala@intel.com>
Cc: Tomi Sarvela <tomi.p.sarvela@intel.com>
Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Acked-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Acked-by: Tomi Sarvela <tomi.p.sarvela@intel.com>
Acked-by: Martin Peres <martin.peres@linux.intel.com>
|
|
We need to check that the kernel rejects attempts to pass in
unknown flags.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
|
