Instead of caching the value during amdgpu_device_init, just call the
function directly. This avoids issues where the unique_id hasn't been
saved by the time that KFD's topology snapshot is done (e.g. Arcturus).
KFD's topology information from the amdgpu_device was initially cached
at KFD initialization due to amdkfd and amdgpu being separate modules.
Now that they are combined together, we can directly call the functions
that we need and avoid this unnecessary duplication and complexity.
As a side-effect of this change, we also remove unique_id=0 for CPUs,
which is obviously not unique.
v2: Drop previous patch printing unique_id in hex
Signed-off-by: Kent Russell <kent.russell@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
The power limit and clock ragne are different in AC mode and DC mode.
Firmware does the setting after this feature is enabled.
Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Enable gfx wave limiting for gfx jobs before pushing high priority
compute jobs so that high priority compute jobs gets more resources
to finish early.
v2: use ring priority instead of job priority.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Wave limiting can be use to load balance high priority
compute jobs along with gfx jobs. When enabled, this will reserve
~75% of waves for compute jobs.
We do not need this from gfx10 onwards because >=gfx10 has
asynchronous compute tunneling to replace wave limit requirement.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
For high priority compute to work properly we need to enable
wave limiting on gfx pipe. Wave limiting is done through writing
into mmSPI_WCL_PIPE_PERCENT_GFX register. Enable only one high
priority compute queue to avoid race condition between multiple
high priority compute queues writing that register simultaneously.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch consist of below related changes:
1 Rename ring->priority to ring->hw_prio.
2 Assign correct hardware ring priority.
3 Remove ring->priority_mutex as ring priority remains unchanged
after initialization.
4 Remove unused ring->num_jobs.
v3: remove ring->num_jobs.
v2: remove ring->priority_mutex.
Fixes: 33abcb1f5a ("drm/amdgpu: set compute queue priority at mqd_init")
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
The dm struct is only being used if DCN config is defined and this
causes a unused variable warning if DCN option is not set.
[How]
Remove the compile flag so the variable is used (there also seems to be
a duplicate guard due to a bad rebase) so remove the outer guard to fix
the warning.
Fixes this warning
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c: In function 'dm_set_vblank':
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c:5380:33: warning: unused variable 'dm' [-Wunused-variable]
Fixes: 98ab5f3513 ("drm/amd/display: Fix deadlock during gpu reset v3")
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This reverts commit 4c3a329273.
Bhawan's fix is better (less ifdefs), so revert this and apply Bhawan's.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
amd-drm-next-5.12-2021-02-03:
amdgpu:
- Display fixes and cleanups
- Vangogh fixes
- Fix possible race when there are timeouts on two rings
- SR-IOV fixes
- Add missing license
- DCE 10/12 bpc fixes
- Display MALL fixes
- Fix SMU user preference settings persistence
- Fix retry in gem allocate
- Add new PCI DID
- Fix for manual fan speed control on cards where it was problematic
- Fix regression in pinning GTT
- Misc display fixes
- Misc code cleanups
amdkfd:
- Fix config handling
- Fix regression in buffer free
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210204045717.3823-1-alexander.deucher@amd.com
Signed-off-by: Dave Airlie <airlied@redhat.com>
Current implementation of mtk_dither_set() cast dev data to
struct mtk_ddp_comp_dev. But other devices with different dev data
would also call this function.
Separate necessary parameters out so other device components (dither,
gamma) can call this function.
Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org>
Reviewed-by: CK Hu <ck.hu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Moving mutex resource management from client driver to mutex driver
could prevent client drivers negotiating for resource management.
Signed-off-by: CK Hu <ck.hu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
mtk mutex is used by both drm and mdp driver, so change disp/ddp term to
mutex to show that it's a common driver for drm and mdp.
Signed-off-by: CK Hu <ck.hu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
After mmsys routing function is moved out of mtk_drm_ddp.c, mtk_drm_ddp.c
has only mtk mutex function, so rename it to match the function in it.
Signed-off-by: CK Hu <ck.hu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Get the fifo size from device tree
because each rdma in the same SoC may have different fifo size
Signed-off-by: Yongqiang Niu <yongqiang.niu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
The orginal setting is not correct, fix it to follow hardware data sheet.
If keep this error setting, mt8173/mt8183 display ok
but mt8192 display abnormal.
Fixes: 0664d1392c ("drm/mediatek: Add AAL engine basic function")
Signed-off-by: Yongqiang Niu <yongqiang.niu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Similar to commit a9d9fea74b
("drm/mediatek: mtk_dsi: Create connector for bridges"):
Use the drm_bridge_connector helper to create a connector for pipelines
that use drm_bridge. This allows splitting connector operations across
multiple bridges when necessary, instead of having the last bridge in
the chain creating the connector and handling all connector operations
internally.
Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
It's possible that state->base.fb is null. Add a check before access its
format.
Fixes: b6b1bb980e ("drm/mediatek: Turn off Alpha bit when plane format has no alpha")
Signed-off-by: Yongqiang Niu <yongqiang.niu@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
If the user passes a "level" value which is higher than 31 then that
leads to shift wrapping. The undefined behavior will lead to a
syzkaller stack dump.
Fixes: 5632708f44 ("drm/amd/powerplay: add dpm force multiple levels on cz/tonga/fiji/polaris (v2)")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This version brings along following fixes:
- Better handling of dummy p-state table
- Workaround for some legacy DP-VGA dongles
- Add Freesync HDMI support to DMCU
- Enable "trigger_hotplug" debugfs on all outputs
- fix initial bounding box values for dcn3.02
- implement support for DID2.0 dsc passthrough
- fix calculation for the pwl backlight curve
- Fix multiple memory leaks
Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[why]
Overlay won't move to a new positon if viewport size is smaller than
what can be handled. It'd either disappear or stay at the old
position. This condition is for example hit if overlay is moved too
much outside of left or top edge of the screen, but it applies to
any non-cursor plane type.
[how]
Reject this contidion at validation time. This gives the calling
level a chance to handle this gracefully and avoid inconsistent
behaivor.
Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Some scenarios where we use a UCLK frequency in between dummy p-state table
entries result in a p-state hang, due to the table not having a close
enough match, so the default DPM0 latency is used, which can be too long to
support dummy p-state switching in these scenarios.
[How]
- old: match if current freq is within +- margin of table entry
- new: find largest table entry that is lower than current freq + margin
- lower than DPM0 will still use DPM0
Signed-off-by: Joshua Aberback <joshua.aberback@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Maximum resolution is 1440*900 when connecting to FHD monitor via some DP-VGA
dongles. The display EDID reading fails over AUX/I2C via DP->VGA dongle, and
this leads to the maximum resolution 1920*1080 cannot be obtained from EDID.
[How]
Provide a workaround for some legacy DP-VGA dongles with a longer aux delay.
Signed-off-by: Taimur Hassan <syed.hassan@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[why]
An old dc_sink state is causing a memory leak because it is missing a
dc_sink_release before a new dc_sink is assigned back to
aconnector->dc_sink.
[how]
Decrement the dc_sink refcount before reassigning it to a new dc_sink.
Signed-off-by: Victor Lu <victorchengchi.lu@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[why]
drm_atomic_commit was changed so that the caller must free their
drm_atomic_state reference on successes.
[how]
Add drm_atomic_commit_put after drm_atomic_commit call in
dm_force_atomic_commit.
Signed-off-by: Victor Lu <victorchengchi.lu@amd.com>
Reviewed-by: Roman Li <Roman.Li@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Per-connector debugfs entry "trigger_hotplug" is available on DP/eDP only.
New IGT tests need this entry to test other outputs.
[How]
Enable this debugfs entry on all types of connectors
Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Mikita Lipski <Mikita.Lipski@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
The conditions for whether we used cached vs non-cached inbox1 depend
on a version check that mismatches what the shared helpers in dmub20
implement.
[How]
Use the dmub_dcn20_use_cached_inbox check for dmub_dcn30 as well.
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[Why]
Initial bounding box values are updated in dcn30_update_bw_bounding_box
but they use dcn3_0_soc and dcn3_0_ip instead of dcn3_02_soc and
dcn3_02_ip
[How]
Add dcn302_update_bw_bounding_box and
dcn302_get_optimal_dcfclk_fclk_for_uclk so it uses
dcn3_02_soc and dcn3_02_ip.
Use sr_exit_time_us, sr_enter_plus_exit_time_us,
from dcn30 on dcn302 to fix flicker on eDP.
Also use dram_clock_change_latency_us from dcn30.
Signed-off-by: Samson Tam <Samson.Tam@amd.com>
Reviewed-by: Joshua Aberback <Joshua.Aberback@amd.com>
Acked-by: Anson Jacob <Anson.Jacob@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
