| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
PTR_RET is deprecated, use PTR_ERR_OR_ZERO instead.
Signed-off-by: Gustavo A. R. Silva <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
The suspend/resume functions are not referenced when power
management is disabled:
drivers/gpu/drm/msm/disp/dpu1/dpu_kms.c:1288:12: error: 'dpu_runtime_resume' defined but not used [-Werror=unused-function]
drivers/gpu/drm/msm/disp/dpu1/dpu_kms.c:1261:12: error: 'dpu_runtime_suspend' defined but not used [-Werror=unused-function]
This marks them as __maybe_unused to let the compiler
drop the functions without complaining.
Fixes: 591225291ca2 ("drm/msm: Add SDM845 DPU support")
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Fix the sparse error. the dpu_rm_init declaration is not consistent
with the implement.
Signed-off-by: zhong jiang <[email protected]>
[robclark un-typo'd subject line]
Signed-off-by: Rob Clark <[email protected]>
|
|
All users of do_gettimeofday() have been removed, but this one recently
crept in, along with an incorrect printing of the microseconds portion.
This converts it to using ktime_get_real_timespec64() as a direct
replacement, and adds the leading zeroes. I considered using monotonic
times (ktime_get()) instead, but as this timestamp appears to only
be used for humans rather than compared with other timestamps, the
real time domain is probably good enough.
Fixes: e43b045e2c82 ("drm/msm/gpu: Capture the state of the GPU")
Signed-off-by: Arnd Bergmann <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
This fixes up a collision between introducing dual-channel support and
the dsi refactors. This patch applies the same dual-channel
considerations and pclk calculations to both v2 and 6G, with a bit of
abstracting for good measure.
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
In these cases, we want to enumerate _all_ clocks, not just the ones
that are assigned a rate.
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Should work with the legacy handling in of, but we shouldn't rely on
that.
Signed-off-by: Sean Paul <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
For hangs, dump copy out the contents of the buffer objects attached to the
guilty submission and print them in the crash dump report.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
HLSQ, SP and TP registers are only accessible from a special
aperture and to make matters worse the aperture is blocked from
the CPU on targets that can support secure rendering. Luckily the
GPU hardware has its own purpose built register dumper that can
access the registers from the aperture. Add a5xx specific code
to program the crashdumper and retrieve the wayward registers
and dump them for the crash state.
Also, remove a block of registers the regular CPU accessible
list that aren't useful for debug which helps reduce the size
of the crash state file by a goodly amount.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Add the contents of each ringbuffer to the GPU state and dump the
data in the crash file encoded with ascii85. To save space only
the used portions of the ringbuffer are dumped.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Convert the format of the 'show' debugfs file and the crash
dump to a format resembling YAML. This should be easier to
parse and be more flexible for future changes and expansions.
v2: Use a standard .rst for the msm crashdump documentation
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Capture the GPU state on a GPU hang and store it for later playback
via the devcoredump facility. Only one crash state is stored at a
time on the assumption that the first hang is usually the most
interesting. The existing crash state can be cleared after capturing
it and then a new one will be captured on the next hang.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Do a bit of cleanup to prepare for upcoming changes to pass the
hanging task comm and cmdline to the crash dump function.
v2: Use GFP_ATOMIC while holding the rcu lock per Chris Wilson
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Convert the existing GPU show function to use the GPU state to
dump the information rather than reading it directly from the hardware.
This will require an additional step to capture the state before
dumping it for the existing nodes but it will greatly facilitate reusing
the same code for dumping a previously captured state from a GPU hang.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Add the infrastructure to capture the current state of the GPU and
store it in memory so that it can be dumped later.
For now grab the same basic ringbuffer information and registers
that are provided by the debugfs 'gpu' node but obviously this should
be extended to capture a much larger set of GPU information.
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Add a puts function for the coredump printer to bypass printf()
for constant strings for a speed boost. Reorganize the
coredump printf callback to share as much code as possible.
v2: Try to reuse code between print and puts as suggested by
Chris Wilson
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Add a puts() function to use seq_puts() to help speed up
up print time for constant strings.
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Add drm_puts() for a much faster path to print constant strings
into a drm_printer object with memcpy and friends. This can
have seconds off of really large outputs such as GPU dumps.
If the drm_printer object supports a custom puts function then
use that otherwise fall back to the slower legacy printf call.
v2: Add documentation for drm_puts() per Daniel Vetter
Reviewed-by: Daniel Vetter <[email protected]>
Signed-off-by: Jordan Crouse <[email protected]>
[robclark fix minor htmldocs warning]
Signed-off-by: Rob Clark <[email protected]>
|
|
Add a drm printer suitable for use with the read callback for
devcoredump or other suitable buffer based output format that
isn't otherwise covered by seq_file.
v2: Add improved documentation per Daniel Vetter
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
The i915 DRM driver very cleverly used ascii85 encoding for their
GPU state file. Move the encode functions to a general header file to
support other drivers that might be interested in the same
functionality.
v4: Make the return value const char * as suggested by Chris Wilson
v3: Fix error_puts -> err_puts pointed out by the 01.org bot
v2: Update API to be cleaner for the caller as suggested by Chris Wilson
Reviewed-by: Chris Wilson <[email protected]>
Signed-off-by: Jordan Crouse <[email protected]>
Signed-off-by: Rob Clark <[email protected]>
|
|
Remove the obsolete fb unreferencing system that is no longer used
since we've transitioned to atomic modeset.
Signed-off-by: Russell King <[email protected]>
|
|
We no longer require a private armada_plane structure, so eliminate
it, and use the drm_plane structure directly.
Signed-off-by: Russell King <[email protected]>
|
|
We no longer need a private plane structure, so get rid of it. Use the
drm_plane structure directly.
Signed-off-by: Russell King <[email protected]>
|
|
The framebuffer base address and toggling mode needs to be updated
when the interlaced flag for mode changes is updated. Arrange to
reprogram these parameters when only the mode has changed.
Signed-off-by: Russell King <[email protected]>
|
|
Write out the plane updates after the dumb frame has completed, but
just before the blank period. This allows all the plane updates to
be performed in a flicker-free non-tearing manner.
Signed-off-by: Russell King <[email protected]>
|
|
Switch the overlay plane away from the transitional helpers and legacy
methods, and use atomic helpers instead to implement the legacy
set_plane ioctl methods.
Signed-off-by: Russell King <[email protected]>
|
|
Switch the primary plane away from the transitional helpers, and
use the atomic helpers instead to implement the legacy set_plane
ioctl call for this plane.
Signed-off-by: Russell King <[email protected]>
|
|
Switch the legacy set_config() method to use the atomic modeset
helper, which allows us to get rid of the legacy dpms, prepare,
commit, mode_set, mode_set_base and disable helper methods.
Signed-off-by: Russell King <[email protected]>
|
|
Enable atomic modeset helpers, and internal DRM use of atomic modeset
with armada-drm.
Signed-off-by: Russell King <[email protected]>
|
|
Implement the atomic_enable()/atomic_disable() methods used by the
atomic modeset helpers. atomic_disable() will need some transitional
code during conversion to ensure proper ordering is maintained.
Signed-off-by: Russell King <[email protected]>
|
|
Explicitly pass in the desired enable/disable state into
armada_drm_crtc_update() rather than having it use the DPMS state
stored in our crtc structure.
Signed-off-by: Russell King <[email protected]>
|
|
Push responsibility for managing the clock during DPMS down into the
variant backend, rather than the CRTC layer having knowledge of its
state.
Signed-off-by: Russell King <[email protected]>
|
|
Prepare handling for atomic modeset CRTC events. Currently, using the
transition helpers, CRTC events do not exist, but once we switch to
proper atomic modeset, they have to be handled.
We queue an event for the next vblank in two places:
- armada_drm_crtc_atomic_flush() provided we aren't doing an
atomic modeset.
- armada_drm_crtc_commit() if we are committing a modeset.
This ensures that the event is sent at the correct time (after all
updates have been written to the hardware and after the following
vblank.)
Signed-off-by: Russell King <[email protected]>
|
|
Rather than writing all bits of SPU_ADV_REG on modeset, only write
what we need to change, and initialise the register in the variant
initialisation.
Signed-off-by: Russell King <[email protected]>
|
|
Update debug to use KMS level, and print the mode using the standard
format for mode lines, but print the adjusted CRTC parameters as
that's what we will be programming for.
Signed-off-by: Russell King <[email protected]>
|
|
For atomic modeset, we need to set the sync signal polarities from the
CRTC state structure rather than the legacy mode structure stored in
CRTC. In any case, we should update this from our mode_set_nofb()
method, rather than the commit() method. Move it there, and ensure
that armada_drm_crtc_update() will not overwrite these bits.
Signed-off-by: Russell King <[email protected]>
|
|
Push the interlaced frame calculation down into armada_drm_plane_calc()
which needs to apply the same correction for both the overlay and
primary planes.
Signed-off-by: Russell King <[email protected]>
|
|
Provide the framebuffer pitches from armada_drm_plane_calc_addrs() as
well as the base addresses for each plane. Since this is now about
more than just addresses, rename to armada_drm_plane_calc().
Signed-off-by: Russell King <[email protected]>
|
|
armada_drm_plane_calc_addrs() gets all its information from the plane
state, so it makes sense to pass the plane state pointer down into this
function, rather than extracting the information in identical ways,
sometimes a couple of layers up.
Signed-off-by: Russell King <[email protected]>
|
|
Move the armada_drm_mode_config_funcs to armada_drv.c, since this now
has less to do with FBs than it does with general mode configuration.
In doing so, we need to make armada_fb_create() visible to armada_drv.c,
which reveals a function name clash with armada_fbdev.c. Rename the
version in armada_fbdev.c.
Signed-off-by: Russell King <[email protected]>
|
|
Use the DRM standard plane properties for specifying the YUV
colour encoding parameter. Our colour range is fixed at limited
range.
Since we are transitioning to atomic modeset, we need to explicitly
add handling of these properties to our atomic_set_property() method,
but once the transition is complete, these will be removed.
Signed-off-by: Russell King <[email protected]>
|
|
Remove the unused CRTC colourspace properties - userspace does not make
use of these. In any case, these are not a property of the CRTC, since
they demonstrably only affect the video (overlay) plane, irrespective
of the format of the graphics (primary) plane.
Signed-off-by: Russell King <[email protected]>
|
|
Move the overlay plane colorkey properties into the plane state,
keeping the existing driver behaviour to avoid breaking userspace.
Signed-off-by: Russell King <[email protected]>
|
|
Move the contrast, brightness, and saturation properties to the overlay
plane state structure, and call our overlay commit function to update
the hardware via the planes atomic_update() method.
Signed-off-by: Russell King <[email protected]>
|
|
Only overlay makes use of these now, so move these to the overlay code.
Signed-off-by: Russell King <[email protected]>
|
|
Split out the primary plane support; this is now entirely separate from
the CRTC support.
Signed-off-by: Russell King <[email protected]>
|
|
Rather than tracking the register state, we can now check the previous
state and decide which registers need updating from that since the old
plane state indicates the previous state which was programmed into the
hardware.
Signed-off-by: Russell King <[email protected]>
|
|
Now that we have the CRTC using the atomic modeset transitional helper,
there is no need to build a temporary crtc state anymore - we can use
the CRTC atomic state directly.
Signed-off-by: Russell King <[email protected]>
|
|
The overlay plane support updates asynchronously to the request, but the
drm_plane_helper_update() transitional helper waits for a vblank event
before releasing the framebuffer. Using the transitional helper would
make the call block, which would introduce a performance regression.
Convert the overlay plane update to use the atomic state structures and
methods for the plane, but implement our own legacy update method
rather than the transitional helper.
Signed-off-by: Russell King <[email protected]>
|
|
page_flip requests happen asynchronously, so we can't wait on the
vblank event before returning to userspace, as the transitional plane
update helper would do. Craft our own implementation that keeps the
asynchronous behaviour of this request, while making use of the atomic
infrastructure for the primary plane update.
Signed-off-by: Russell King <[email protected]>
|