hersen wu [Thu, 4 Oct 2018 13:28:20 +0000 (09:28 -0400)]
drm/amdgpu/display: dm/amdgpu: make dp phy debugfs for eDP
[WHY] dp debugfs file does not exist for eDP under
/sys/kernel/debug/dri/0/eDP-1. the root is phy debugfs
is created for dp connector only.
[HOW] for eDP connector, create phy debugfs too.
Emily Deng [Fri, 12 Oct 2018 10:14:32 +0000 (18:14 +0800)]
drm/amdgpu: Set the default value about gds vmid0 size
For sriov, when first run windows guest, then run linux guest, the gds
vmid0 size will be reset to 0 by windows guest. So if the value has been
reset to 0, then set the value to the default value in linux guest.
v2:
Fixed value instead of reading mmGDS_VMID0_SIZE.
Dave Airlie [Thu, 11 Oct 2018 04:53:40 +0000 (14:53 +1000)]
Merge branch 'drm-next-4.20' of git://people.freedesktop.org/~agd5f/linux into drm-next
Add a new list.h helper for doing bulk updates. Used by ttm.
- Fixes for display underflow on VI APUs at 4K with UVD running
- Endian fixes for powerplay on vega
- DC fixes for interlaced video
- Vega20 powerplay fixes
- RV/RV2/PCO powerplay fixes
- Fix for spurious ACPI events on HG laptops
- Fix a memory leak in DC on driver unload
- Fixes for manual fan control mode switching
- Suspend/resume robustness fixes
- Fix display handling on RV2
- VCN fixes for DPG on PCO
- Misc code cleanups and warning fixes
Lyude Paul [Thu, 23 Aug 2018 01:40:07 +0000 (21:40 -0400)]
drm/nouveau: Start using new drm_dev initialization helpers
Per the documentation in drm_get_pci_dev(), this function is deprecated
and shouldn't be used anymore. As it turns out, we're going to need to
stop using drm_get_pci_dev() anyway in order to allow us to turn off the
card before full system shutdowns, otherwise we'll hit race conditions
with userspace while trying to tear down the card on shutdown.
So, start using drm_dev_get() and drm_dev_put(), and just turn our
load/unload callbacks into open coded init/fini() functions.
Lyude Paul [Thu, 6 Sep 2018 21:43:26 +0000 (17:43 -0400)]
drm/nouveau: Refactor nvXX_backlight_init()
There's literally no difference between any of the backlight init
functions besides the backlight properties they set and the backlight
callbacks that they set, so move all of the duplicated backlight init
code out of there and into nouveau_backlight_init().
This gets rid of a lot of copy pasta!
Changes since v1:
- Some of the pre-refactor callbacks were storing nv_encoder in callback
data for the backlight devices that they registered, as opposed to
nouveau_drm. This got missed and caused some bugs that didn't
originally appear on my setup (NULL kernel derefs) for some reason.
So, fix this by finding the nouveau_encoder in
nouveau_backlight_init(), and using that as the callback data for all
gens instead even if they don't care about the encoder.
Lyude Paul [Thu, 6 Sep 2018 21:43:23 +0000 (17:43 -0400)]
drm/nouveau: Move backlight device into nouveau_connector
Currently module unloading is broken in nouveau due to a rather annoying
race condition resulting from nouveau_backlight.c having gone a bit
stale over time:
While I haven't actually gone into the details of what's causing this to
happen (maybe the kernel removes the backlight device in the device core
before we get to it?), it doesn't really matter anyway because the way
nouveau handles backlights has long since been deprecated.
According to the documentation on the drm_connector->late_register()
hook, the ->late_register() hook should be used for adding extra
connector-related devices. Vice versa, the ->early_unregister() hook is
meant to be used for removing those devices.
So: gut nouveau_drm->bl_list and nouveau_drm->backlight, and replace
them with per-connector backlight structures. Additionally, move
backlight registration/teardown into the ->late_register() and
->early_unregister() hooks so that DRM can give us a chance to remove
the backlight before the connector is even removed. This appears to fix
the problem once and for all.
Changes since v2:
- Use NV_INFO_ONCE for printing GMUX information, since otherwise this
will end up printing that message for as many times as we have
connectors
drm/amdgpu: remove set but not used variable 'header'
Fixes gcc '-Wunused-but-set-variable' warning:
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c: In function 'amdgpu_ucode_init_bo':
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c:431:39: warning:
variable 'header' set but not used [-Wunused-but-set-variable]
drm/amdgpu: remove set but not used variable 'ring' in psp_v11_0_ring_stop
Fixes gcc '-Wunused-but-set-variable' warning:
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c: In function 'psp_v11_0_ring_stop':
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c:309:19: warning:
variable 'ring' set but not used [-Wunused-but-set-variable]
YueHaibing [Tue, 9 Oct 2018 01:53:00 +0000 (01:53 +0000)]
drm/amdkfd: Remove set but not used variable 'preempt_all_queues'
Fixes gcc '-Wunused-but-set-variable' warning:
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c: In function 'destroy_queue_cpsch':
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c:1366:7: warning:
variable 'preempt_all_queues' set but not used [-Wunused-but-set-variable]
It never used since introduct in
commit 992839ad64f2 ("drm/amdkfd: Add static user-mode queues support")
Colin Ian King [Mon, 8 Oct 2018 16:22:28 +0000 (17:22 +0100)]
drm/amdgpu/powerplay: fix missing break in switch statements
There are several switch statements that are missing break statements.
Add missing breaks to handle any fall-throughs corner cases.
Detected by CoverityScan, CID#1457175 ("Missing break in switch")
Fixes: 18aafc59b106 ("drm/amd/powerplay: implement fw related smu interface for iceland.") Acked-by: Huang Rui <[email protected]> Signed-off-by: Colin Ian King <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Lyude Paul [Sat, 22 Sep 2018 00:43:44 +0000 (20:43 -0400)]
drm/amdgpu: Suppress keypresses from ACPI_VIDEO events
Currently we return NOTIFY_DONE for any event which we don't think is
ours. However, many laptops will send more then just an ATIF event and
will also send an ACPI_VIDEO_NOTIFY_PROBE event as well. Since we don't
check for this, we return NOTIFY_DONE which causes a keypress for the
ACPI event to be propogated to userspace. This is the equivalent of
someone pressing the display key on a laptop every time there's a
hotplug event.
So, check for ACPI_VIDEO_NOTIFY_PROBE events and suppress keypresses
from them.
Rex Zhu [Wed, 3 Oct 2018 08:10:45 +0000 (16:10 +0800)]
drm/amdgpu: Refine function amdgpu_device_ip_late_init
1. only call late_init when hw_init successful,
so check status.hw instand of status.valid in late_init.
2. set status.late_initialized true if late_init was not implemented.
Dave Airlie [Mon, 8 Oct 2018 06:45:56 +0000 (16:45 +1000)]
Merge tag 'drm-msm-next-2018-10-07' of git://people.freedesktop.org/~robclark/linux into drm-next
This time mostly further refinement of dpu1+a6xx for sdm845 and
beyond.. and hurray for more negative diffstat :-)
- Misc cleanups and fixes
- GPU preemption optimization
- a6xx perf improvements and clock fixes (ie. lets actually not run at
minimum clks)
- a6xx devfreq/DCVS
- Lots of code cleanup across dpu (Bruce, Jeykumar, Sean)
- Fixed a few crashes on startup relating to dsi (Sean)
- Add cursor support (Sravanthi, Sean)
- Properly free mdss irq on destroy (Jordan)
- Use correct encoder_type when initializing, fixes crash on boot (Stephen)
Rex Zhu [Sat, 29 Sep 2018 07:42:52 +0000 (15:42 +0800)]
drm/amd/pp: Allocate ucode bo in request_smu_load_fw
ucode bo is needed by request_smu_load_fw,
the request_smu_load_fw maybe called by gfx/sdma
before smu hw init.
so move amdgpu_ucode_bo_init to request_smu_lowd_fw
from smu hw init.
Nick Alcock [Thu, 4 Oct 2018 19:58:09 +0000 (20:58 +0100)]
drm/radeon: ratelimit bo warnings
So a few days ago I started getting sprays of these warnings --
sorry, but because it was a few days ago I'm not sure what I was
running at the time (but it was probably either Stellaris or Chromium).
followed by a massive stream of "vbo resource seems too big for the bo".
The most extreme flood ran from 23:01:58 to 23:02:47 and emitted 91,000
lines of log in that time. This... seems excessive, given that each log
message after the first contains more or less no information.
So ratelimit these messages. (We probably want to see at least *some* so
that the underlying bug can be fixed -- always assuming the bug isn't in
unfixable closed-source game code somewhere.)
Shirish S [Fri, 5 Oct 2018 05:24:21 +0000 (10:54 +0530)]
drm/amdgpu: remove the intterupt handling for the KIQ events
[Why]
1. we never submit IBs to the KIQ
2. there seems to be ~500ms delay during amdgpu resume spent in KIQ,
hence pointing toward interrupts are not working correctly.
Felix Kuehling [Wed, 22 Aug 2018 19:28:44 +0000 (15:28 -0400)]
drm/amdkfd: Fix incorrect use of process->mm
This mm_struct pointer should never be dereferenced. If running in
a user thread, just use current->mm. If running in a kernel worker
use get_task_mm to get a safe reference to the mm_struct.
[Why]
1. We never submit IBs to KIQ.
2. Ring test pass without KIQ's ring also.
3. By skipping we see an improvement of around 500ms
in the amdgpu's resume time.
drm/scheduler: Simplify spsc_queue_count check in drm_sched_entity_select_rq
Clang generates a warning when it sees a logical not followed by a
conditional operator like ==, >, or <.
drivers/gpu/drm/scheduler/sched_entity.c:470:6: warning: logical not is
only applied to the left hand side of this comparison
[-Wlogical-not-parentheses]
if (!spsc_queue_count(&entity->job_queue) == 0 ||
^ ~~
drivers/gpu/drm/scheduler/sched_entity.c:470:6: note: add parentheses
after the '!' to evaluate the comparison first
if (!spsc_queue_count(&entity->job_queue) == 0 ||
^
( )
drivers/gpu/drm/scheduler/sched_entity.c:470:6: note: add parentheses
around left hand side expression to silence this warning
if (!spsc_queue_count(&entity->job_queue) == 0 ||
^
( )
1 warning generated.
It assumes the author might have made a mistake in their logic:
if (!a == b) -> if (!(a == b))
Sometimes that is the case; other times, it's just a super convoluted
way of saying 'if (a)' when b = 0:
if (!1 == 0) -> if (0 == 0) -> if (true)
Alternatively:
if (!1 == 0) -> if (!!1) -> if (1)
Simplify this comparison so that Clang doesn't complain.
Fixes: 35e160e781a0 ("drm/scheduler: change entities rq even earlier") Signed-off-by: Nathan Chancellor <[email protected]> Reviewed-by: Christian König <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Rex Zhu [Sat, 29 Sep 2018 05:54:33 +0000 (13:54 +0800)]
drm/amd/pp: Refine smu7/8 request_smu_load_fw callback function
The request_smu_load_fw of VI is used to load gfx/sdma
ip's firmware.
Check whether the gfx/sdma firmware have been loaded successfully
in this callback function.
if failed, driver can exit to avoid gpu hard hung.
if successful, clean the flag reload_fw to avoid duplicated fw load.
when suspend/resume, driver need to reload fw.
so in suspend, reset the reload_fw flag to true to enable load fw when
resume.
Rex Zhu [Thu, 20 Sep 2018 06:30:55 +0000 (14:30 +0800)]
drm/amdgpu: Add fan RPM setting via sysfs
Add fan1_target for get/set fan speed in RPM unit
Add fan1_min/fan1_max for get min, max fan speed in RPM unit
Add fan1_enable to enable/disable the fan1 sensor
v3: drop the hardcode value of min/max rpm in comments pointed
out by Alex.
v2: query the min/max rpm gpu support instand of hardcode value.
Shirish S [Mon, 24 Sep 2018 13:31:47 +0000 (19:01 +0530)]
drm/amd/display: Signal hw_done() after waiting for flip_done()
In amdgpu_dm_commit_tail(), wait until flip_done() is signaled before
we signal hw_done().
[Why]
This is to temporarily address a paging error that occurs when a
nonblocking commit contends with another commit, particularly in a
mirrored display configuration where at least 2 CRTCs are updated.
The error occurs in drm_atomic_helper_wait_for_flip_done(), when we
attempt to access the contents of new_crtc_state->commit.
Here's the sequence for a mirrored 2 display setup (irrelevant steps
left out for clarity):
**THREAD 1** | **THREAD 2**
|
Initialize atomic state for flip |
|
Queue worker |
...
| Do work for flip
|
| Signal hw_done() on CRTC 1
| Signal hw_done() on CRTC 2
|
| Wait for flip_done() on CRTC 1
<---- **PREEMPTED BY THREAD 1**
Initialize atomic state for cursor |
update (1) |
|
Do cursor update work on both CRTCs |
|
Clear atomic state (2) |
**DONE** |
...
|
| Wait for flip_done() on CRTC 2
| *ERROR*
|
The issue starts with (1). When the atomic state is initialized, the
current CRTC states are duplicated to be the new_crtc_states, and
referenced to be the old_crtc_states. (The new_crtc_states are to be
filled with update data.)
Some things to note:
* Due to the mirrored configuration, the cursor updates on both CRTCs.
* At this point, the pflip IRQ has already been handled, and flip_done
signaled on all CRTCs. The cursor commit can therefore continue.
* The old_crtc_states used by the cursor update are the **same states**
as the new_crtc_states used by the flip worker.
At (2), the old_crtc_state is freed (*), and the cursor commit
completes. We then context switch back to the flip worker, where we
attempt to access the new_crtc_state->commit object. This is
problematic, as this state has already been freed.
(*) Technically, 'state->crtcs[i].state' is freed, which was made to
reference old_crtc_state in drm_atomic_helper_swap_state()
[How]
By moving hw_done() after wait_for_flip_done(), we're guaranteed that
the new_crtc_state (from the flip worker's perspective) still exists.
This is because any other commit will be blocked, waiting for the
hw_done() signal.
Note that both the i915 and imx drivers have this sequence flipped
already, masking this problem.
drm/amd/display: Use proper enums in process_channel_reply
Clang warns when one enumerated type is implicitly converted to another.
drivers/gpu/drm/amd/amdgpu/../display/dc/dce/dce_aux.c:315:19: warning:
implicit conversion from enumeration type 'enum
aux_channel_operation_result' to different enumeration type 'enum
aux_transaction_reply' [-Wenum-conversion]
reply->status = AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON;
~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/amd/amdgpu/../display/dc/i2caux/dce110/aux_engine_dce110.c:349:19:
warning: implicit conversion from enumeration type 'enum
aux_channel_operation_result' to different enumeration type 'enum
aux_transaction_reply' [-Wenum-conversion]
reply->status = AUX_CHANNEL_OPERATION_FAILED_HPD_DISCON;
~ ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The current enum is incorrect, it should be from aux_transaction_reply,
so use AUX_TRANSACTION_REPLY_HPD_DISCON.
drm/amd/display: Raise dispclk value for dce_update_clocks
[Why]
The DISPCLK value was previously requested to be 15% higher for all
ASICS that went through the dce110 bandwidth code path. As part of a
refactoring of dce_clocks and dce110 set_bandwidth this was removed
for power saving considerations.
This changed caused corruption under certain display configurations.
Originally thought to be Vega specific, it was also observed on Polaris.
[How]
The 15% is brought back but its placement differs from the original
patch. This boost should only be enable while DFS bypass is inactive.
This (like the Vega patch) is also a workaround that should be
removed after the root cause is identified.
Charlene Liu [Tue, 18 Sep 2018 17:23:42 +0000 (13:23 -0400)]
drm/amd/display: fix Interlace video timing.
[Description] interlace mode shows wrong vertical timing.
Interface timing in Edid is half vertical timing as progressive timing.
driver doubled the vertical timing in edid_paser,
no need to double in optc again.
Jun Lei [Tue, 18 Sep 2018 13:38:20 +0000 (09:38 -0400)]
drm/amd/display: fix memory leak in resource pools
[why]
ddc engines were recently changed to be independently tracked
from pipe count. the change was reflected in resource constructor
but not in destructor. this manifests as a memory leak when
pipe harvesting is enabled, since not all constructed ddc engines
are freed
[how]
make destructor symmetric with constructor for all dcX_resource
projects / linux.git / log