Commit graph

126,670 commits

Author SHA1 Message Date
Ville Syrjälä
b6bf160102 drm/i915/pps: Extract msecs_to_pps_units()
Replace all the hand rolled *10 stuff with something a bit
more descriptive (msecs_to_pps_units()).

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-7-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 18:03:47 +02:00
Ville Syrjälä
87545d7584 drm/i915/pps: Spell out the eDP spec power sequencing delays a bit more clearly
We determine the "spec" eDP power sequencing delays
by referencing some max values from the eDP spec.
Write out each number from the spec explicitly instead
of precomputing the final number (that's the job of
the computer). Makes it a bit easier to see what the
supposed spec defined numbers actually are.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-6-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 18:03:13 +02:00
Ville Syrjälä
ccae47c64b drm/i915/lvds: Use struct intel_pps_delays for LVDS power sequencing
Reuse struct intel_pps_delays for the LVDS power
sequencing delays instead of hand rolling it all.
Perhaps in the future we could reuse some of the
same PPS code for both LVDS and eDP (assuming we
can decouple the PPS code from intel_dp...).

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-5-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 18:02:41 +02:00
Ville Syrjälä
c55bc70309 drm/i915/pps: Rename intel_pps_delay members
Stop using the semi-random eDP spec T1,T3,... names for the
power sequencing delays, and instead call them by their human
readable names. Much easier to keep track what delay goes
where when you don't have to constantly cross reference against
the eDP spec.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-4-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 18:01:17 +02:00
Ville Syrjälä
dac2ec8d3b drm/i915/pps: Decouple pps delays from VBT struct definition
We currently lack a proper struct definition for the VBT power
squencing delays, and instead we use the same struct definition
(in intel_bios.h) for both the VBT layout and our driver side
state. Decouple those two things by moving the current struct
into intel_vbt_defs.h and adding a new one for the driver's use.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-3-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 18:00:48 +02:00
Ville Syrjälä
57ecdc5521 drm/i915/pps: Store the power cycle delay without the +1
The code initializing the power sequencing delays is a bit
hard to follow. One confusing thing is that we keep doing the
+/-1 adjustment for the hardware register value in several places.
Simplify this a bit by doing the adjustment only when reading or
writing the actual register.

This also matches how the LVDS code does things.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241106215859.25446-2-ville.syrjala@linux.intel.com
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
2024-11-28 17:59:50 +02:00
Ville Syrjälä
da5bb8974c drm/i915/dsb: Nuke the MMIO->indexed register write logic
We've determined that indexed DSB writes are only faster
than MMIO writes when writing the same register ~5 or more
times. That seems very unlikely to happen in any other case
than when using indexed LUT registers. Simplify the code
by removing the MMIO->indexed write conversion logic and
just emit the instruction as an indexed write from the get go.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241120164123.12706-4-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
2024-11-28 17:44:39 +02:00
Ville Syrjälä
2504a316b3 drm/i915/color: Stop using non-posted DSB writes for legacy LUT
DSB LUT register writes vs. palette anti-collision logic
appear to interact in interesting ways:
- posted DSB writes simply vanish into thin air while
  anti-collision is active
- non-posted DSB writes actually get blocked by the anti-collision
  logic, but unfortunately this ends up hogging the bus for
  long enough that unrelated parallel CPU MMIO accesses start
  to disappear instead

Even though we are updating the LUT during vblank we aren't
immune to the anti-collision logic because it kicks in briefly
for pipe prefill (initiated at frame start). The safe time
window for performing the LUT update is thus between the
undelayed vblank and frame start. Turns out that with low
enough CDCLK frequency (DSB execution speed depends on CDCLK)
we can exceed that.

As we are currently using non-posted writes for the legacy LUT
updates, in which case we can hit the far more severe failure
mode. The problem is exacerbated by the fact that non-posted
writes are much slower than posted writes (~4x it seems).

To mititage the problem let's switch to using posted DSB
writes for legacy LUT updates (which will involve using the
double write approach to avoid other problems with DSB
vs. legacy LUT writes). Despite writing each register twice
this will in fact make the legacy LUT update faster when
compared to the non-posted write approach, making the
problem less likely to appear. The failure mode is also
less severe.

This isn't the 100% solution we need though. That will involve
estimating how long the LUT update will take, and pushing
frame start and/or delayed vblank forward to guarantee that
the update will have finished by the time the pipe prefill
starts...

Cc: stable@vger.kernel.org
Fixes: 34d8311f4a ("drm/i915/dsb: Re-instate DSB for LUT updates")
Fixes: 25ea3411bd ("drm/i915/dsb: Use non-posted register writes for legacy LUT")
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12494
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241120164123.12706-3-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
2024-11-28 17:42:22 +02:00
Ville Syrjälä
ecba559a88 drm/i915/dsb: Don't use indexed register writes needlessly
Turns out the DSB indexed register write command has
rather significant initial overhead compared to the normal
MMIO write command. Based on some quick experiments on TGL
you have to write the register at least ~5 times for the
indexed write command to come out ahead. If you write the
register less times than that the MMIO write is faster.

So it seems my automagic indexed write logic was a bit
misguided. Go back to the original approach only use
indexed writes for the cases we know will benefit from
it (indexed LUT register updates).

Currently we shouldn't have any cases where this truly
matters (just some rare double writes to the precision
LUT index registers), but we will need to switch the
legacy LUT updates to write each LUT register twice (to
avoid some palette anti-collision logic troubles).
This would be close to the worst case for using indexed
writes (two writes per register, and 256 separate registers).
Using the MMIO write command should shave off around 30%
of the execution time compared to using the indexed write
command.

Cc: stable@vger.kernel.org
Fixes: 34d8311f4a ("drm/i915/dsb: Re-instate DSB for LUT updates")
Fixes: 25ea3411bd ("drm/i915/dsb: Use non-posted register writes for legacy LUT")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241120164123.12706-2-ville.syrjala@linux.intel.com
Reviewed-by: Uma Shankar <uma.shankar@intel.com>
2024-11-28 17:41:55 +02:00
Ville Syrjälä
24387a21dc Revert "drm/i915/dpt: Make DPT object unshrinkable"
This reverts commit 51064d471c.

Now that we forcefully evict all DPT VMAs during suspend
there should be no problem allowing the shrinker to eat
the DPT objects.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12965
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-5-ville.syrjala@linux.intel.com
Acked-by: Brian Geffon <bgeffon@google.com>
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>
2024-11-28 17:33:48 +02:00
Ville Syrjälä
106216c220 drm/i915/dpt: Evict all DPT VMAs on suspend
Currently intel_dpt_resume() tries to blindly rewrite all the
PTEs for currently bound DPT VMAs. That is problematic because
the CPU mapping for the DPT is only really guaranteed to exist
while the DPT object has been pinned. In the past we worked
around this issue by making DPT objects unshrinkable, but that
is undesirable as it'll waste physical RAM.

Let's instead forcefully evict all the DPT VMAs on suspend,
thus guaranteeing that intel_dpt_resume() has nothing to do.
To guarantee that all the DPT VMAs are evictable by
intel_dpt_suspend() we need to flush the cleanup workqueue
after the display output has been shut down.

And for good measure throw in a few extra WARNs to catch
any mistakes.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-4-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>
2024-11-28 17:33:44 +02:00
Ville Syrjälä
37ab41e11f drm/i915: Intruduce display.wq.cleanup
Introduce a dedicated workqueue for the commit cleanup work.
In the future we'll need this to guarantee all the cleanup
works have finished at a specific point during suspend.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-3-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>
2024-11-28 17:33:40 +02:00
Ville Syrjälä
0768530b7c drm/i915: Don't reuse commit_work for the cleanup
Currently we reuse the commit_work for a later cleanup step.
Let's not do that so that atomic ioctl handler won't accidentally
wait for the cleanup work when it really wants to just wait on the
commit_tail() part. We'll just add another work struct for the
cleanup.

Cc: Brian Geffon <bgeffon@google.com>
Cc: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241127061117.25622-2-ville.syrjala@linux.intel.com
Reviewed-by: Vidya Srinivas <vidya.srinivas@intel.com>
Tested-by: Vidya Srinivas <vidya.srinivas@intel.com>
2024-11-28 17:33:33 +02:00
Matthew Brost
aef0b4a072 drm/xe: Take PM ref in delayed snapshot capture worker
The delayed snapshot capture worker can access the GPU or VRAM both of
which require a PM reference. Take a reference in this worker.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: 4f04d07c0a ("drm/xe: Faster devcoredump")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-5-matthew.brost@intel.com
(cherry picked from commit 1c6878af11)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:22:36 +01:00
Matthew Auld
c78f439918 drm/xe/migrate: use XE_BO_FLAG_PAGETABLE
On some HW we want to avoid the host caching PTEs, since access from GPU
side can be incoherent. However here the special migrate object is
mapping PTEs which are written from the host and potentially cached. Use
XE_BO_FLAG_PAGETABLE to ensure that non-cached mapping is used, on
platforms where this matters.

Fixes: 7a060d786c ("drm/xe/mtl: Map PPGTT as CPU:WC")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Nirmoy Das <nirmoy.das@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126181259.159713-4-matthew.auld@intel.com
(cherry picked from commit febc689b27)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:22:36 +01:00
Matthew Auld
23346f8516 drm/xe/migrate: fix pat index usage
XE_CACHE_WB must be converted into the per-platform pat index for that
particular caching mode, otherwise we are just encoding whatever happens
to be the value of that enum.

Fixes: e8babb280b ("drm/xe: Convert multiple bind ops into single job")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: Nirmoy Das <nirmoy.das@intel.com>
Cc: <stable@vger.kernel.org> # v6.12+
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126181259.159713-3-matthew.auld@intel.com
(cherry picked from commit f3dc9246f9)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:22:36 +01:00
Matthew Auld
87651f31ae drm/xe/guc_submit: fix race around suspend_pending
Currently in some testcases we can trigger:

xe 0000:03:00.0: [drm] Assertion `exec_queue_destroyed(q)` failed!
....
WARNING: CPU: 18 PID: 2640 at drivers/gpu/drm/xe/xe_guc_submit.c:1826 xe_guc_sched_done_handler+0xa54/0xef0 [xe]
xe 0000:03:00.0: [drm] *ERROR* GT1: DEREGISTER_DONE: Unexpected engine state 0x00a1, guc_id=57

Looking at a snippet of corresponding ftrace for this GuC id we can see:

162.673311: xe_sched_msg_add:     dev=0000:03:00.0, gt=1 guc_id=57, opcode=3
162.673317: xe_sched_msg_recv:    dev=0000:03:00.0, gt=1 guc_id=57, opcode=3
162.673319: xe_exec_queue_scheduling_disable: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0x29, flags=0x0
162.674089: xe_exec_queue_kill:   dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0x29, flags=0x0
162.674108: xe_exec_queue_close:  dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa9, flags=0x0
162.674488: xe_exec_queue_scheduling_done: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa9, flags=0x0
162.678452: xe_exec_queue_deregister: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa1, flags=0x0

It looks like we try to suspend the queue (opcode=3), setting
suspend_pending and triggering a disable_scheduling. The user then
closes the queue. However the close will also forcefully signal the
suspend fence after killing the queue, later when the G2H response for
disable_scheduling comes back we have now cleared suspend_pending when
signalling the suspend fence, so the disable_scheduling now incorrectly
tries to also deregister the queue. This leads to warnings since the queue
has yet to even be marked for destruction. We also seem to trigger
errors later with trying to double unregister the same queue.

To fix this tweak the ordering when handling the response to ensure we
don't race with a disable_scheduling that didn't actually intend to
perform an unregister.  The destruction path should now also correctly
wait for any pending_disable before marking as destroyed.

Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3371
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241122161914.321263-6-matthew.auld@intel.com
(cherry picked from commit f161809b36)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:22:36 +01:00
Matthew Auld
6965f91a00 drm/xe/guc_submit: fix race around pending_disable
Currently in some testcases we can trigger:

[drm] *ERROR* GT0: SCHED_DONE: Unexpected engine state 0x02b1, guc_id=8, runnable_state=0
[drm] *ERROR* GT0: G2H action 0x1002 failed (-EPROTO) len 3 msg 02 10 00 90 08 00 00 00 00 00 00 00

Looking at a snippet of corresponding ftrace for this GuC id we can see:

498.852891: xe_sched_msg_add:     dev=0000:03:00.0, gt=0 guc_id=8, opcode=3
498.854083: xe_sched_msg_recv:    dev=0000:03:00.0, gt=0 guc_id=8, opcode=3
498.855389: xe_exec_queue_kill:   dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x3, flags=0x0
498.855436: xe_exec_queue_lr_cleanup: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x83, flags=0x0
498.856767: xe_exec_queue_close:  dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x83, flags=0x0
498.862889: xe_exec_queue_scheduling_disable: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0xa9, flags=0x0
498.863032: xe_exec_queue_scheduling_disable: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x2b9, flags=0x0
498.875596: xe_exec_queue_scheduling_done: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x2b9, flags=0x0
498.875604: xe_exec_queue_deregister: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x2b1, flags=0x0
499.074483: xe_exec_queue_deregister_done: dev=0000:03:00.0, 5:0x1, gt=0, width=1, guc_id=8, guc_state=0x2b1, flags=0x0

This looks to be the two scheduling_disable racing with each other, one
from the suspend (opcode=3) and then again during lr cleanup. While
those two operations are serialized, the G2H portion is not, therefore
when marking the queue as pending_disabled and then firing off the first
request, we proceed do the same again, however the first disable
response only fires after this which then clears the pending_disabled.
At this point the second comes back and is processed, however the
pending_disabled is no longer set, hence triggering the warning.

To fix this wait for pending_disabled when doing the lr cleanup and
calling disable_scheduling_deregister. Also do the same for all other
disable_scheduling callers.

Fixes: dd08ebf6c3 ("drm/xe: Introduce a new DRM driver for Intel GPUs")
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3515
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <stable@vger.kernel.org> # v6.8+
Reviewed-by: Matthew Brost <mattheq.brost@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241122161914.321263-5-matthew.auld@intel.com
(cherry picked from commit ddb106d212)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:22:20 +01:00
Matt Roper
ece45026b0 drm/xe: Update xe2_graphics name string
Since both Xe2 and Xe3 platforms currently use the same set of graphics
IP feature flags, we associate the "graphics_xe2" structure with both IPs.
Update the name string on that IP structure to clarify this and avoid
confusion as Xe3 platforms start going into public CI.

Fixes: 800d75bf20 ("drm/xe/xe3: Define Xe3 feature flags")
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241125194838.1190599-2-matthew.d.roper@intel.com
(cherry picked from commit 4fe70f664a)
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
2024-11-28 15:07:37 +01:00
AngeloGioacchino Del Regno
822ba16048 drm/panfrost: Add GPU ID for MT8188 Mali-G57 MC3
The MediaTek MT8188 SoC has a Mali-G57 MC3 GPU and, similarly to
MT8192, it has yet another special GPU ID.

Add the GPU ID to the list and treat it as a standard Mali-G57.

Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113112622.123044-1-angelogioacchino.delregno@collabora.com
2024-11-28 11:25:20 +00:00
Boris Brezillon
34f4e94c83 drm/panthor: Fix a typo in the FW iface flag definitions
Drop the _RD_ in the flag names.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Acked-by: Liviu Dudau <liviu.dudau@arm.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241113160257.2002333-1-boris.brezillon@collabora.com
2024-11-28 10:58:17 +01:00
Matthew Brost
5425472783 drm/xe: Only allow contiguous BOs to use xe_bo_vmap
xe_bo_vmap only works on contiguous BOs, disallow xe_bo_vmap on BO
unless we are certain the BO is contiguous.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-9-matthew.brost@intel.com
2024-11-27 16:38:56 -08:00
Matthew Brost
e03b0aa67a drm/xe: Set XE_BO_FLAG_PINNED in migrate selftest BOs
We only allow continguous BOs to be vmapped, set XE_BO_FLAG_PINNED on
BOs in migrate selftest as this forces continguous BOs and selftest uses
vmaps.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-8-matthew.brost@intel.com
2024-11-27 16:38:55 -08:00
Matthew Brost
5f7bec831f drm/xe: Use ttm_bo_access in xe_vm_snapshot_capture_delayed
Non-contiguous mapping of BO in VRAM doesn't work, use ttm_bo_access
instead.

v2:
 - Fix error handling

Fixes: 0eb2a18a8f ("drm/xe: Implement VM snapshot support for BO's and userptr")
Suggested-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-7-matthew.brost@intel.com
2024-11-27 16:38:54 -08:00
Matthew Brost
b6308aaa24 drm/xe/display: Update intel_bo_read_from_page to use ttm_bo_access
Don't open code vmap of a BO, use ttm_bo_access helper which is safe for
non-contiguous BOs and non-visible BOs.

Suggested-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-6-matthew.brost@intel.com
2024-11-27 16:38:53 -08:00
Matthew Brost
1c6878af11 drm/xe: Take PM ref in delayed snapshot capture worker
The delayed snapshot capture worker can access the GPU or VRAM both of
which require a PM reference. Take a reference in this worker.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: 4f04d07c0a ("drm/xe: Faster devcoredump")
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-5-matthew.brost@intel.com
2024-11-27 16:38:52 -08:00
Matthew Brost
97e8cecb6a drm/xe: Add xe_ttm_access_memory
Non-contiguous VRAM cannot easily be mapped in TTM nor can non-visible
VRAM easily be accessed. Add xe_ttm_access_memory which hooks into
ttm_bo_access to access such memory.

v4:
 - Assert memory access rather than taking RPM ref (Thomas / Auld)
 - Fix warning on xe_res_cursor.h for non-zero offset (Mika)

Reported-by: Christoph Manszewski <christoph.manszewski@intel.com>
Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-4-matthew.brost@intel.com
2024-11-27 16:38:51 -08:00
Matthew Brost
7d08df5d0b drm/ttm: Add ttm_bo_access
Non-contiguous VRAM cannot easily be mapped in TTM nor can non-visible
VRAM easily be accessed. Add ttm_bo_access, which is similar to
ttm_bo_vm_access, to access such memory.

v4:
 - Fix checkpatch warnings (CI)
v5:
 - Fix checkpatch warnings (CI)
v6:
 - Fix kernel doc (Auld)
v7:
 - Move ttm_bo_access to ttm_bo_vm.c (Christian)

Cc: Christian König <christian.koenig@amd.com>
Reported-by: Christoph Manszewski <christoph.manszewski@intel.com>
Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Tested-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-3-matthew.brost@intel.com
2024-11-27 16:38:50 -08:00
Matthew Brost
f85dc3c5ab drm/xe: Add xe_bo_vm_access
Add xe_bo_vm_access which is wrapper around ttm_bo_vm_access which takes
rpm refs for device access.

Suggested-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20241126174615.2665852-2-matthew.brost@intel.com
2024-11-27 16:38:49 -08:00
Dave Stevenson
e511f8fc3b drm/vc4: Drop planes that are completely off-screen or 0 crtc size
It is permitted for a plane to be configured such that none
of it is on-screen via either negative dest rectangle X,Y
offset, or an offset that is greater than the crtc dimensions.

These planes were resized via drm_atomic_helper_check_plane_state
such that the source rectangle had a zero width or height, but
they still created a dlist entry even though they contributed
no pixels. In the case of vc6_plane_mode_set, that it could result
in negative values being written into registers, which caused
incorrect behaviour.

Drop planes that result in a source width or height of 0 pixels
or an on-screen size of 0 pixels to avoid the incorrect rendering.

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-28-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:21 +00:00
Dave Stevenson
52b0a07316 drm/vc4: Enable bg_fill if there are no planes enabled
The default was to have enable_bg_fill disabled and the first
plane set it if it wasn't opaque and covering the whole screen.
However that meant that if no planes were enabled, then the
background fill wasn't enabled, and would give a striped
output from the uninitialised output buffer.

Initialise it to enabled to avoid this.

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-27-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:21 +00:00
Maxime Ripard
3e048861c5 drm/vc4: Add additional warn_on for incorrect revisions
Some code path in vc4 are conditional to a generation and cannot be
executed on others. Let's put a WARN_ON if that ever happens.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-26-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:21 +00:00
Dave Stevenson
88c065c739 drm/vc4: hdmi: Support 2712 D-step register map
The D-step has increased FIFO sizes of the MAI_THR blocks,
resulting in changes to the register masking. Add support for
it.

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-25-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:21 +00:00
Dave Stevenson
b7b14b31c8 drm/vc4: plane: Add support for 2712 D-step.
There are a few minor changes in the display list generation
for the D-step of the chip, so add them.

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-24-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:21 +00:00
Dave Stevenson
b460e0072c drm/vc4: hvs: Add in support for 2712 D-step.
The registers have been moved around, and a couple of minor changes
made, so adapt for this.

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-23-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Dave Stevenson
83c4f5947c drm/vc4: drv: Add support for 2712 D-step
Add in the compatible string and VC4_GEN_ enum for the D-step

Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-22-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
fa9ccf0102 drm/vc4: txp: Add BCM2712 MOPLET support
The BCM2712 features a simpler TXP called MOPLET. Let's add support for
it.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-21-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
5aab428cfe drm/vc4: txp: Add support for BCM2712 MOP
The BCM2712 has an evolution of what used to be called TXP in the
earlier SoCs, but is now called MOP.

There's a few differences still, so we can add a new compatible to deal
with them easily.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-20-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
7dc41ea6c0 drm/vc4: txp: Add a new TXP encoder type
Starting with BCM2712, we'll have a two TXP. Let's follow the HDMI
example and add two encoder types for TXP: TXP0 and TXP1.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-19-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
99f34ea894 drm/vc4: txp: Move the encoder type in the variant structure
We'll have multiple TXP instances in the BCM2712, so we can't use a
single encoder type anymore. Let's tie the encoder type to the
compatible.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-18-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
cf1c8f63f3 drm/vc4: txp: Handle 40-bits DMA Addresses
The BCM2712 MOP and MOPLET can handle addresses larger than 32bits
through an extra register. We can easily support it and make it
conditional based on the compatible through a boolean in our variant
structure.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-17-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
d60205bb6a drm/vc4: txp: Add horizontal and vertical size offset toggle bit
The new writeback controllers that can be found on the BCM2712 require
to have their horizontal and vertical size reduced by one.

Let's tie that behaviour to the compatible so we can support both the
new and old controllers.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-16-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
fcd1465d22 drm/vc4: txp: Add byte enable toggle bit
The MOPLET doesn't have the BYTE_ENABLE field to set, but the TXP and
MOP do, so let's add a boolean to control whether or not we need to set
it.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-15-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
4013716de7 drm/vc4: txp: Rename TXP data structure
The TXP data structure has a name too generic for the multiple variants
we'll have to support. Let's rename it to mention the SoC it applies to.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-14-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
66fbec8367 drm/vc4: txp: Introduce structure to deal with revision differences
The BCM2712 will have several TXP with small differences. Let's add a
structure tied to the compatible to deal with those differences.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-13-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
0b1dbe82e0 drm/vc4: hdmi: Add support for BCM2712 HDMI controllers
The HDMI controllers found in the BCM2712 are largely the ones found in
the BCM2711 with a different PHY.

There's some difference with how timings are split between registers,
and HDMI1 is now able to run at 4k/60Hz.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-12-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
9a9ef0da86 drm/vc4: crtc: Add support for BCM2712 PixelValves
The PixelValves found on the BCM2712 are similar to the ones found in
the previous generation.

Compared to BCM2711:
- the pixelvalves only drive one HDMI controller each
- HDMI1 PixelValve has a FIFO long enough to support 4k at 60Hz
- support has been added for odd horizontal timings whilst at 2pixels/clock

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-11-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:20 +00:00
Maxime Ripard
7687a12153 drm/vc4: hvs: Add support for BCM2712 HVS
The HVS found in the BCM2712, while having a similar role, is very
different from the one found in the previous SoCs. Indeed, the register
layout is fairly different, and the DLIST format is new as well.

Let's introduce the needed functions to support the new HVS.

This commit adds the C-step register layout. The D-step will be
added later.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-10-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:19 +00:00
Maxime Ripard
626ffc5f87 drm/vc4: drv: Support BCM2712
The BCM2712 has an improved display pipeline, most notably with a
different HVS and only HDMI and writeback outputs.

Let's introduce it as a new VideoCore generation and compatible.

Signed-off-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-9-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:19 +00:00
Dave Stevenson
8f2fc64773 drm/vc4: Fix reading of frame count on GEN5 / Pi4
The frame count values moved within registers DISPSTAT1 and
DISPSTAT2 with GEN5, so update the accessor function to
accommodate that.

Fixes: b51cd7ad14 ("drm/vc4: hvs: Fix frame count register readout")
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20241025-drm-vc4-2712-support-v2-2-35efa83c8fc0@raspberrypi.com
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
2024-11-27 14:37:19 +00:00