Aim the dummy load to the current back buffer instead if the front
buffer. Assuming the idea is to get the next flip to be stuck behind
the dummy load?
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Try to tune the dummy load to ~1 second. The calibration happens the
first time dummy load is generated.
v2: Actually do the number of ops intended and
calibrate to 1 second and not 2
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Make the dummy load independent of the display resolution by using a
two fixed size dummy bos to generate the load. As a final step do
another copy from one of the dummy bos to the fb to make sure there's
a dependency between the dummy load and any subsequent operation on
the fb.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
If we wait for one vblank, we may end up returning almost immediately,
so trying to assert anything but >0 about the minimum duration is
bogus.
Instead wait for two vblanks and then we can assert that we should have
be blocked for at least one frame. And move the upper bound to a little
over two frames to match.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=79050
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
kms_flip/nonblocking-read will block indefinitely if it fails, so
introduce a timeout to indicate test failure.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=85718
Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Make sure DSPSURF will change during the panning operation
in flip-vs-panning-vs-hang.
This will now test agains bugs between the kernel's mmio vs.
CS flip race handling and GPU resets. If the kernel is buggy
if will fail to notice that the panning operation changed the
base address before the GPU reset had a chance to deal with the
pending page flips, and so the flip would never complete due to
DSPSURFLIVE not matching the expected value.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
These tests require DRM master right, so make sure they have it from the
beginning. This gives an early indication if another DRM master is running
and makes the given test skip (with a proper explanation of the reason)
instead of exiting with error.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Since relocations are variable size, depending upon generation, it is
easier to handle the resizing of the batch request inside the
BEGIN_BATCH macro. This still leaves us with having to resize commands
in a few places - which still need adaption for gen8+.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This reveal that quite a few locations were writing relocation offsets
but only allowing for 32 bit addresses. To reveal such places in active
tests, we also now double check that we do not use more batch space than
declared.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The EBUSY checking is very fragile currently in case there's any kind
extra delay in the test loop. At least the flip-vs-rmfb fails reliably
on my IVB.
So to make the test less fragile remove the EBUSY check from all the
current flip tests, and instead add a specific busy-flip test. To
better guarantee that we get the EBUSY add a blocking vblank wait just
before we issue the first flip. This maximizes the time we have to
submit two flips during one frame.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Make sure DSPSURF will change during the panning operation
in flip-vs-panning-vs-hang.
This will now test agains bugs between the kernel's mmio vs.
CS flip race handling and GPU resets. If the kernel is buggy
if will fail to notice that the panning operation changed the
base address before the GPU reset had a chance to deal with the
pending page flips, and so the flip would never complete due to
DSPSURFLIVE not matching the expected value.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
- Some debug logging for the ts continuity checks.
- Add a plain vblank-vs-suspend test where kms_flip doesn't switch off
the displays first.
- Ensure that we do at least 2 rounds, the suspend tests bailed out
after 1 round and so didn't test anything.
- Frob the testnames a bit.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This is a "review by igt test" for a bug located in
i915_gem_object_pin_to_display_plane and fixed by:
commit 392013bdd4b6128795e33c84bd6d6d3fd66ff0a3
Author: Oscar Mateo <oscar.mateo@intel.com>
Date: Fri May 16 11:23:12 2014 +0100
drm/i915: Gracefully handle obj not bound to GGTT in is_pin_display
Otherwise, we do a NULL pointer dereference.
I've seen this happen while handling an error in
i915_gem_object_pin_to_display_plane():
If i915_gem_object_set_cache_level() fails, we call is_pin_display()
to handle the error. At this point, the object is still not pinned
to GGTT and maybe not even bound, so we have to check before we
dereference its GGTT vma.
v2: Chris Wilson says restoring the old value is easier, but that
is_pin_display is useful as a theory of operation. Take the solomonic
decision: at least this way is_pin_display is a little more robust
(until Chris can kill it off).
v2: Avoid code duplication by using igt_create_fb_with_bo_size() as
requested by Ville Syrjälä (original author of the "too big" test idea).
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
These check whether everything is still ok wrt vblank handling after
runtime pm and system suspend-resume.
In addition to the usual checks they also ensure that the vblank frame
counter isn't totally ridiculous, something Keith complained about
aeons ago. With Ville's drm_vblank_on/off rework this should now be
fixed and solid.
v2:
- Ignore seq_step, vblanks completely immediately when the crtc goes off
- Only run system suspend/resume tests once.
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When extracting a raw __gem_set_tiling helper I've fumbled this in
commit 590f6101402b51bca54f69c002380bda967484ea
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Wed Oct 9 20:50:50 2013 +0200
lib/drmtest: extract rawer __gem_set_tiling
Fix things up so that we properly propaget the error again.
Now to make this all properly work we also need to make kms_flip a
notch more robust against such failures ...
This only blows up on gen2/3 with the pan tests which want a too wide
framebuffer for tiling.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Shorter and more in line with our general igt_ prefix for everything
which isn't somehow intel or i915-gem or otherwise hw specific - these
helpers here are all fully generic framebuffer handling functions
based on kms + cairo.
Well, the actual buffer alloc is done with i915 gem, but meh ;-)
Two special cases:
- bpp_depth_to_drm_format and drm_format_to_bpp completely lacked
prefixes, so just add igt_.
- write_fb was a bit misleading given that we have gem_write for
uploading to buffers. Rename that to write_fb_to_png to make it
crystal clear what this thing does even without looking at docs.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use the new-style function using drm fourcc codes instead everywhere.
To easily use thew fourcc based interface also expose
bpp_depth_to_drm_format from the library. Finally include drm_fourcc.h
from the igt_kms.h header since pretty much everyone needs this now.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
And shovel all the various helpers in there.
Also move igt_set_vt_graphics_mode to igt_kms.h since the function is
implemented in igt_kms.c. And it fits better. I kinda missed this in
the prep work.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Step one to properly namespace the rendercpy/mediafill functions. Als
give the buf_height/width helpers a proper igt_ prefix.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
And move the public interfaces into intel_batchbuffer.[hc].
A bit messy since we are fairly inconsistent with our header #include
handling.
Also exclude rendercopy.h from the documentation.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
dpms-vs-vblank-race and modeset-vs-vblank-race try to race wait for
vblank ioctls from one thread against modeset/dpms ioctls from another.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Context banning can prevent the page flip hang tests from actaully
testing anything, so make the relevant subtests fail in that case.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
On a bad day, hanging the GPU may be terminal. Yet even if the GPU is
terminally wedged we expect modesetting (and pageflips) to continue.
That deserves to be a dedicated test, but in the meantime we should
strive to avoid falling over just because the code is not resilient.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As a flip is outstanding, there is an issue that the kernel may not be
able to release one of the fences that userspace requires and erroneous
report EDEADLK (on gen2, gen3).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73696
