Apparently the pthread shim on Android doesn't have pthread cancellation,
so use the plain old volatile to terminate the CPU hogs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
One scenario under recent discussion is that of having a thundering herd
in i915_wait_request - where the overhead of waking up every waiter for
every batchbuffer was significantly impacting customer throughput. This
benchmark tries to replicate something to that effect by having a large
number of consumers generating a busy load (a large copy followed by
lots of small copies to generate lots of interrupts) and tries to wait
upon all the consumers concurrenctly (to reproduce the thundering herd
effect). To measure the overhead, we have a bunch of cpu hogs - less
kernel overhead in waiting should allow more CPU throughput.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The main goal of this subtest is to trigger the following warning in
the function i915_gem_object_get_fence():
if (WARN_ON(!obj->map_and_fenceable))
To trigger this warning, the subtest first creates a Y-tiled object and
an associated framebuffer with the Y-fb modifier. Furthermore, to
prevent the map_and_fenceable from being set, we make sure that
the object does not have a normal VMA by refraining from rendering to the
object and by setting the rotation property upfront before calling commit.
v2: Do not call paint_squares and just use one output.
v3: Convert an if condition to igt_require and move the plane rotation
requirement further up before the fb allocation.
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
That option is not needed anymore since:
commit 982934625ac67234c6d85c6cf29a5a487e54d4f0
Author: Thomas Wood <thomas.wood@intel.com>
Date: Wed Sep 16 14:36:24 2015 +0100
lib: allow wildcard matching when specifying subtests
In fact, using "--run-subtest 'fbc-*'" is better than using --fbc-only
due to how SKIPs are handled. In the former, only the tests matching
the expression are tried, so the number of SKIPs only contains the
number of tests on the specified pattern that were skipped. If you
used --fbc-only, all the non-fbc tests would count as SKIPs, so it
would be harder to know which of the tests marked as skipped were
actual FBC tests.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
This is a corner case not exercised by the other subtests. The test is
expected to pass both with the current Kernel tree and with the
patches that are on the mailing list.
The patches currently on the mailing list change how the CFB is
allocated, and this subtest is designed to make sure everything still
works as expected.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
This fixes the failures for cases where you use --run-subtest to run
single subtests that don't use any drawing patterns.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Don't use the rounded vrefresh info to predict the frame duration.
Instead calculate if from the clock.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Do a dry run with rtcwake first to determine if the system even supports
the intended suspend state. If not, skip the test.
Fixes a bunch of stuff on my BYT FFRD8 that doesn't support S3.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Don't try to test invaliud pipe/port combos. Fixes the test on
VLV w/ DSI since the pipe<->DSI port mapping is fixed. Should also
fix other platforms with similar restrictions.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
We need a new ioctl to find the correct GTT size to use when submitting
execbuffers (as opposed to wishing to know the global GTT size).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We've had bugs in the past that caused cursor updates to be synced to
vblank, resulting in sluggish cursor movement. Add a test to try to
make sure we don't regress and reintroduce these bugs.
Cc: kalyan.kondapally@intel.com
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Apparently GCC treats the #pragma directive as a code block and so we
returned unconditionally from memchr_inv() on the first byte fubaring
the test.
commit b04691b3e8e62bd105f0f4d8d3b9e7da187f0c90
Author: Thomas Wood <thomas.wood@intel.com>
Date: Tue Sep 8 11:16:46 2015 +0100
tests/gem_pwrite_snooped: disable const cast warning
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92227
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
core_prop_blob was using ioctls not in the android kernel. Added a
igt_require_propblob() function and local defines/structures so the
test will compile and skip on kernels where the feature is unsupported.
v2: moved igt_require_propblob() to core_prop_blob.c (Daniel Vetter)
Moved gem_blt.c to a seperate patch (Thomas Wood)
Signed-off-by: Derek Morton <derek.j.morton@intel.com>
[Thomas: fix compiler warning]
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Do some page flipping on the rotated plane just to exercise
that code path.
v2: Actually render to flip fb and fixed flip counter.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Sonika Jindal <sonika.jindal@intel.com>
Cc: Arun R Murthy <arun.r.murthy@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Gen9 adds some new capabilities not present on previous platforms
(primary plane windowing, 90/270 rotation, etc.). Add a new subtest to
check how these new features interact with the use of the universal
plane API.
For now we just check whether pageflips work as expected in a windowed
setting. We may want to add some rotation testing in future patches.
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Also, following kernel definition Kabylake is Skylake.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
This patch adds support for dumping audio registers of Broxton.
Signed-off-by: Lu, Han <han.lu@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Rename the current gem_mmap__{cpu,gtt,wc}() functions into
__gem_mmap__{cpu,gtt,wc}(), and add back wrappers with the original name
that assert that the pointer is valid. Most callers will expect a valid
pointer and shouldn't have to bother with failures.
To avoid changing anything (yet), sed 's/gem_mmap__/__gem_mmap__/g'
over the entire codebase.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Stochastically-reviwewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cairo helpfully allocates a new buffer for us when
cairo_image_surface_create_for_data() is called with a NULL ptr. That
means if gem_mmap__gtt() fails, we get a totally silent failure and
nothing ever drawn into the framebuffer. Very confusing.
Put in an igt_assert() to make sure we managed to mmap something.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Stochastically-reviwewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Do the following
ptr = gem_mmap__{cpu,gtt,wc}()
+igt_assert(ptr);
whenever the code doesn't handle the NULL ptr in any kind of
specific way.
Makes it easier to move the assert into gem_mmap__{cpu,gtt,wc}() itself.
Mostly done with coccinelle, with some manual cleanups:
@@
identifier I;
@@
<... when != igt_assert(I)
when != igt_require(I)
when != igt_require_f(I, ...)
when != I != NULL
when != I == NULL
(
I = gem_mmap__gtt(...);
+ igt_assert(I);
|
I = gem_mmap__cpu(...);
+ igt_assert(I);
|
I = gem_mmap__wc(...);
+ igt_assert(I);
)
...>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Stochastically-reviwewed-by: Chris Wilson <chris@chris-wilson.co.uk>
gem_mmap__{cpu,gtt,wc} never return MAP_FAILED, it gets converted to
NULL internally. So don't go asserting that the returned value is
not MAP_FAILED.
Done with coccinelle:
@@
type T;
identifier I;
@@
(
I = gem_mmap__gtt(...);
|
I = gem_mmap__cpu(...);
|
I = gem_mmap__wc(...);
)
...
(
- igt_assert(I != MAP_FAILED);
+ igt_assert(I);
|
- igt_assert(I && I != MAP_FAILED);
+ igt_assert(I);
|
- igt_assert(I != (T *) MAP_FAILED);
+ igt_assert(I);
|
- igt_assert(I != NULL);
+ igt_assert(I);
)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Stochastically-reviwewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Get rid of the gem_mmap() alias of gem_mmap__gtt(). I don't see any
point in having it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Stochastically-reviwewed-by: Chris Wilson <chris@chris-wilson.co.uk>
The current upstream Kernel code stops and then restarts FBC at every
page flip. I am working on patches to keep FBC enabled all the time,
so this brings the possibility of having the hardware tracking
mechanism looking at the old buffer instead of the new one. This test
should catch this problem.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
There's no reason to do a mode unset since we don't care about whether
FBC or PSR are enabled or disabled. This should save about a second
for each time you invoke the program.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
It's not testing a lot since we lack the Kernel Stolen Memory Checker,
but some day this will be useful.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
... instead of just setting it for the fullscreen_plane subtest and
then unsetting it. We want to use it more, so just make sure all the
code knows how to deal with multiple planes per CRTC.
This also allows us to do some upfront caching of the plane
information, so the code for the tests gets simpler.
There's a new test that's going to use all these changes.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Exercises the new blob-creation ioctl, testing lifetimes and behaviour
of user-created blobs, as well as exercising all the invariant
conditions we guarantee from modes exposed as blob properties.
v2: Renamed to core_prop_blob, skip test if blob not available.
v3: No changes.
v4: Consistently return 0/errno.
v5: Use do_ioctl_err and igt_assert_fd.
Use igt_assert_*() helper macros rather than direct igt_assert().
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Daniel Vetter <daniel.vetter@intel.com>
do_ioctl demands that the ioctl returns success; add a variant named
do_ioctl_err, which expects the ioctl to fail, and demands a particular
result.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Logical negation is hard.
v2: The second integer isn't plural (Daniel).
Cc: "Morton, Derek J" <derek.j.morton@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Execute N blits and time how long they complete to measure both GPU
limited bandwidth and submission overhead.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Android M-Dessert treats implicit declaration of function warnings
as errors resulting in igt failing to build.
This patch fixes the errors by including missing header files as
required. Mostly this involved including igt.h in the benchmarks.
Signed-off-by: Derek Morton <derek.j.morton@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The code is confused about the units of CODE_TIME. The comment
says 50 microsseconds, but the actual code makes it 50
milliseconds. Avoid the whole mess by measuring the sleep
duration ourselves. Since the time measurement is taken around
the whole operation it obviously includes a bit of extra, but
at least it's much less than the fixed 50 ms.
For instance on one VLV board I now get something like this:
- Residency in rc6 or deeper state: 3002 ms (ratio to expected duration: 0.98)
+ Residency in rc6 or deeper state: 3001 ms (sleep duration 3003 ms) (ratio to expected duration: 1.00)
so the reported ratio is now much closer to reality.
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
When testing all modes on a connector with a single mode, if the modeset
fails, the code attempts to remove fb_info[-1], because old_fb still has
the inital value of -1.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90625
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Test that VMAs associated with a context are cleaned up when
contexts are destroyed.
In practice this emulates the leak seen between fbcon and X server.
Every time the X server exits we gain one VMA on the fbcon frame
buffer object as externally visible via for example
/sys/kernel/debug/dri/0/i915_gem_gtt.
v2: Use igt_debugfs_open, getline and strstr instead of home-brewed
string matching. (Thomas Wood)
v3: Rebase for drm_open_driver.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Thomas Wood <thomas.wood@intel.com>
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
