linux-2.6-microblaze.git
4 years agoMerge tag 'amd-drm-fixes-5.8-2020-06-04' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Mon, 8 Jun 2020 01:55:33 +0000 (11:55 +1000)]
Merge tag 'amd-drm-fixes-5.8-2020-06-04' of git://people.freedesktop.org/~agd5f/linux into drm-next

amd-drm-fixes-5.8-2020-06-04

amdgpu:
- Prevent hwmon accesses while GPU is in reset
- CTF interrupt fix
- Backlight fix for renoir
- Fix for display sync groups
- Display bandwidth validation workaround

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200604181900.4609-1-alexander.deucher@amd.com
4 years agoMerge branch 'linux-5.8' of git://github.com/skeggsb/linux into drm-next
Dave Airlie [Mon, 8 Jun 2020 00:06:09 +0000 (10:06 +1000)]
Merge branch 'linux-5.8' of git://github.com/skeggsb/linux into drm-next

- HDMI/DP audio HDA fixes
- display hang fix for Volta/Turing
- GK20A regression fix.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Ben Skeggs <skeggsb@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/
4 years agodrm/nouveau/kms/nv50-: clear SW state of disabled windows harder
Ben Skeggs [Thu, 4 Jun 2020 01:00:01 +0000 (11:00 +1000)]
drm/nouveau/kms/nv50-: clear SW state of disabled windows harder

The most innocuous result of not having done this is that we end up
sending unnecessary methods when we next enable the window.

However, interactions with the code handling skipping disables when
an update immediately follows, and window ownership assignment, can
lead to upsetting the display hardware on Volta and newer.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: gr/gk20a: Use firmware version 0
Thierry Reding [Wed, 3 Jun 2020 14:20:02 +0000 (16:20 +0200)]
drm/nouveau: gr/gk20a: Use firmware version 0

Tegra firmware doesn't actually use any version numbers and passing -1
causes the existing firmware binaries not to be found. Use version 0 to
find the correct files.

Fixes: ef16dc278ec2 ("drm/nouveau/gr/gf100-: select implementation based on available FW")
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/gm200-: detect and potentially disable HDA support on some SORs
Ben Skeggs [Wed, 3 Jun 2020 02:51:03 +0000 (12:51 +1000)]
drm/nouveau/disp/gm200-: detect and potentially disable HDA support on some SORs

Some HDA pin widgets may be disabled by BIOS, and unavailable from a
SOR.  Our SOR allocation policy uses this information to allocate an
appropriate SOR when HDA is supported by a display.

Thank you to NVIDIA for providing the information to determine this.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/gp100: split SOR implementation from gm200
Ben Skeggs [Thu, 4 Jun 2020 04:22:32 +0000 (14:22 +1000)]
drm/nouveau/disp/gp100: split SOR implementation from gm200

GP100 needs different HDA detection.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp: modify OR allocation policy to account for HDA requirements
Ben Skeggs [Wed, 3 Jun 2020 01:43:23 +0000 (11:43 +1000)]
drm/nouveau/disp: modify OR allocation policy to account for HDA requirements

Since GM200, SORs are no longer tied to a specific connector, and we
allocate them instead, with the assumption that all SORs are equally
capable.

However, there's a 1<->1 mapping between SOR and HDA pin widget, and
it turns out that it's possible for some widgets to be disabled...

In order to avoid picking a SOR without a valid pin widget, some new
rules need to be added.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp: split part of OR allocation logic into a function
Ben Skeggs [Wed, 3 Jun 2020 01:40:47 +0000 (11:40 +1000)]
drm/nouveau/disp: split part of OR allocation logic into a function

No logical changes here, this is just moving the code to make the
changes in the next commit more obvious.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp: provide hint to OR allocation about HDA requirements
Ben Skeggs [Wed, 3 Jun 2020 01:37:56 +0000 (11:37 +1000)]
drm/nouveau/disp: provide hint to OR allocation about HDA requirements

Will be used by a subsequent commit to influence SOR allocation policy.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/amd/display: Revalidate bandwidth before commiting DC updates
Nicholas Kazlauskas [Wed, 3 Jun 2020 00:42:33 +0000 (20:42 -0400)]
drm/amd/display: Revalidate bandwidth before commiting DC updates

[Why]
Whenever we switch between tiled formats without also switching pixel
formats or doing anything else that recreates the DC plane state we
can run into underflow or hangs since we're not updating the
DML parameters before committing to the hardware.

[How]
If the update type is FULL then call validate_bandwidth again to update
the DML parmeters before committing the state.

This is basically just a workaround and protective measure against
update types being added DC where we could run into this issue in
the future.

We can only fully validate the state in advance before applying it to
the hardware if we recreate all the plane and stream states since
we can't modify what's currently in use.

The next step is to update DM to ensure that we're creating the plane
and stream states for whatever could potentially be a full update in
DC to pre-emptively recreate the state for DC global validation.

The workaround can stay until this has been fixed in DM.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/display: use blanked rather than plane state for sync groups
Alex Deucher [Tue, 2 Jun 2020 21:22:48 +0000 (17:22 -0400)]
drm/amdgpu/display: use blanked rather than plane state for sync groups

We may end up with no planes set yet, depending on the ordering, but we
should have the proper blanking state which is either handled by either
DPG or TG depending on the hardware generation.  Check both to determine
the proper blanked state.

Bug: https://gitlab.freedesktop.org/drm/amd/issues/781
Fixes: 5fc0cbfad45648 ("drm/amd/display: determine if a pipe is synced by plane state")
Cc: nicholas.kazlauskas@amd.com
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
4 years agodrm/nouveau/kms/gt215-: fix race with audio driver runpm
Ben Skeggs [Fri, 29 May 2020 07:57:29 +0000 (17:57 +1000)]
drm/nouveau/kms/gt215-: fix race with audio driver runpm

The audio driver can call into nouveau right while we're in the middle
of re-fetching the EDID, and decide it no longer needs to be awake.

Stop depending on EDID in the audio component get_eld() callback, and
instead cache whether audio support is present from the prior modeset.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/gm200-: fix NV_PDISP_SOR_HDMI2_CTRL(n) selection
Ben Skeggs [Fri, 29 May 2020 05:18:47 +0000 (15:18 +1000)]
drm/nouveau/disp/gm200-: fix NV_PDISP_SOR_HDMI2_CTRL(n) selection

This is a SOR register, and not indexed by the bound head.

Fixes display not coming up on high-bandwidth HDMI displays under a
number of configurations.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agoRevert "drm/amd/display: disable dcn20 abm feature for bring up"
Harry Wentland [Thu, 28 May 2020 13:44:44 +0000 (09:44 -0400)]
Revert "drm/amd/display: disable dcn20 abm feature for bring up"

This reverts commit 96cb7cf13d8530099c256c053648ad576588c387.

This change was used for DCN2 bringup and is no longer desired.
In fact it breaks backlight on DCN2 systems.

Cc: Alexander Monakov <amonakov@ispras.ru>
Cc: Hersen Wu <hersenxs.wu@amd.com>
Cc: Anthony Koo <Anthony.Koo@amd.com>
Cc: Michael Chiu <Michael.Chiu@amd.com>
Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reported-and-tested-by: Alexander Monakov <amonakov@ispras.ru>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
4 years agodrm/amd/powerplay: ack the SMUToHost interrupt on receive V2
Evan Quan [Tue, 12 May 2020 11:06:37 +0000 (19:06 +0800)]
drm/amd/powerplay: ack the SMUToHost interrupt on receive V2

There will be no further interrupt without proper ack
for current one.

V2: fix typo to really set ACK bit only

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/pm: return an error during GPU reset or suspend (v2)
Alex Deucher [Sun, 24 May 2020 06:46:53 +0000 (02:46 -0400)]
drm/amdgpu/pm: return an error during GPU reset or suspend (v2)

Return an error for sysfs and debugfs power interfaces during
gpu reset and suspend.  Prevents access to the hw while it may
be in an unusable state.

v2: squash in fix to drop suspend check

Acked-by: Nirmoy Das <nirmoy.das@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agoMerge tag 'amd-drm-next-5.8-2020-05-27' of git://people.freedesktop.org/~agd5f/linux...
Dave Airlie [Thu, 28 May 2020 06:10:16 +0000 (16:10 +1000)]
Merge tag 'amd-drm-next-5.8-2020-05-27' of git://people.freedesktop.org/~agd5f/linux into drm-next

amd-drm-next-5.8-2020-05-27:

amdgpu:
- SRIOV fixes
- RAS fixes
- VCN 2.5 DPG (Dynamic PowerGating) fixes
- FP16 updates for display
- CTF cleanups
- Display fixes
- Fix pcie bw sysfs handling
- Enable resizeable BAR support for gmc 10.x
- GFXOFF fixes for Raven
- PM sysfs handling fixes

amdkfd:
- Fix a race condition
- Warning fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200527231219.3930-1-alexander.deucher@amd.com
4 years agoMerge tag 'drm-misc-next-fixes-2020-05-27' of git://anongit.freedesktop.org/drm/drm...
Dave Airlie [Thu, 28 May 2020 05:38:26 +0000 (15:38 +1000)]
Merge tag 'drm-misc-next-fixes-2020-05-27' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

Short summary of fixes pull (less than what git shortlog provides):

There's a fix for panel brighness on Lenovo X13 Yoga devices and a fix for
-Wformat warnings on architectures where atomic-64 counters are not of
type unsigned long long.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20200527080123.GA8186@linux-uq9g
4 years agodrm/amd/display: Fix potential integer wraparound resulting in a hang
Aric Cyr [Tue, 12 May 2020 15:48:48 +0000 (11:48 -0400)]
drm/amd/display: Fix potential integer wraparound resulting in a hang

[Why]
If VUPDATE_END is before VUPDATE_START the delay calculated can become
very large, causing a soft hang.

[How]
Take the absolute value of the difference between START and END.

Signed-off-by: Aric Cyr <aric.cyr@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: drop cursor position check in atomic test
Simon Ser [Sat, 23 May 2020 11:53:41 +0000 (11:53 +0000)]
drm/amd/display: drop cursor position check in atomic test

get_cursor_position already handles the case where the cursor has
negative off-screen coordinates by not setting
dc_cursor_position.enabled.

Signed-off-by: Simon Ser <contact@emersion.fr>
Fixes: 626bf90fe03f ("drm/amd/display: add basic atomic check for cursor plane")
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: fix device attribute node create failed with multi gpu
Kevin Wang [Fri, 22 May 2020 14:06:17 +0000 (22:06 +0800)]
drm/amdgpu: fix device attribute node create failed with multi gpu

the origin design will use varible of "attr->states" to save node
supported states on current gpu device, but for multi gpu device, when
probe second gpu device, the driver will check attribute node states
from previous gpu device wthether to create attribute node.
it will cause other gpu device create attribute node faild.

1. add member attr_list into amdgpu_device to link supported device attribute node.
2. add new structure "struct amdgpu_device_attr_entry{}" to track device attribute state.
3. drop member "states" from amdgpu_device_attr.

v2:
1. move "attr_list" into amdgpu_pm and rename to "pm_attr_list".
2. refine create & remove device node functions parameter.

fix:
drm/amdgpu: optimize amdgpu device attribute code

Signed-off-by: Kevin Wang <kevin1.wang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/nouveau: use correct conflicting framebuffer API
Dave Airlie [Tue, 26 May 2020 04:18:27 +0000 (14:18 +1000)]
drm/nouveau: use correct conflicting framebuffer API

nouveau was calling the fbdev API which has issues with modules
and built-ins. Call the correct API.

Reported-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Fixes: 2dd4d163cd9c ("drm/nouveau: remove open-coded version of remove_conflicting_pci_framebuffers()")
Link: https://lore.kernel.org/lkml/21b52c28-3ace-cd13-d8ce-f38f2c6b2a96@infradead.org/T/#u
Signed-off-by: Dave Airlie <airlied@redhat.com>
4 years agodrm/vblank: Fix -Wformat compile warnings on some arches
Lyude Paul [Thu, 21 May 2020 20:46:47 +0000 (16:46 -0400)]
drm/vblank: Fix -Wformat compile warnings on some arches

On some architectures like ppc64le and aarch64, compiling with
-Wformat=1 will throw the following warnings:

  In file included from drivers/gpu/drm/drm_vblank.c:33:
  drivers/gpu/drm/drm_vblank.c: In function 'drm_update_vblank_count':
  drivers/gpu/drm/drm_vblank.c:273:16: warning: format '%llu' expects
  argument of type 'long long unsigned int', but argument 4 has type
  'long int' [-Wformat=]
    DRM_DEBUG_VBL("updating vblank count on crtc %u:"
                  ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  ./include/drm/drm_print.h:407:22: note: in definition of macro
  'DRM_DEBUG_VBL'
    drm_dbg(DRM_UT_VBL, fmt, ##__VA_ARGS__)
                        ^~~
  drivers/gpu/drm/drm_vblank.c:274:22: note: format string is defined here
           " current=%llu, diff=%u, hw=%u hw_last=%u\n",
                     ~~~^
                     %lu

So, fix that with a typecast.

Co-developed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20200521204647.2578479-1-lyude@redhat.com
4 years agodrm/amdgpu: Sync with VM root BO when switching VM to CPU update mode
Felix Kuehling [Wed, 20 May 2020 01:02:45 +0000 (21:02 -0400)]
drm/amdgpu: Sync with VM root BO when switching VM to CPU update mode

This fixes an intermittent bug where a root PD clear operation still in
progress could overwrite a PDE update done by the CPU, resulting in a
VM fault.

Fixes: 108b4d928c03 ("drm/amd/amdgpu: Update VM function pointer")
Reported-by: Jay Cornwall <Jay.Cornwall@amd.com>
Tested-by: Jay Cornwall <Jay.Cornwall@amd.com>
Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Handle GPU reset for DC block
Bhawanpreet Lakha [Mon, 11 May 2020 18:21:17 +0000 (14:21 -0400)]
drm/amd/display: Handle GPU reset for DC block

[Why]
Previously we used the s3 codepath for gpu reset. This can lead to issues in
certain case where we end of waiting for fences which will never come (because
parts of the hw are off due to gpu reset) and we end up waiting forever causing
a deadlock.

[How]
Handle GPU reset separately from normal s3 case. We essentially need to redo
everything we do in s3, but avoid any drm calls.

For GPU reset case

suspend:
-Acquire DC lock
-Cache current dc_state
-Commit 0 stream/planes to dc (this puts dc into a state where it can be
 powered off)
-Disable interrupts
resume
-Edit cached state to force full update
-Commit cached state from suspend
-Build stream and plane updates from the cached state
-Commit stream/plane updates
-Enable interrupts
-Release DC lock

v2:
-Formatting
-Release dc_state

Signed-off-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add apu flags (v2)
Alex Deucher [Fri, 15 May 2020 18:18:29 +0000 (14:18 -0400)]
drm/amdgpu: add apu flags (v2)

Add some APU flags to simplify handling of different APU
variants.  It's easier to understand the special cases
if we use names flags rather than checking device ids and
silicon revisions.

v2: rebase on latest code

Acked-by: Evan Quan <evan.quan@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/powerpay: Disable gfxoff when setting manual mode on picasso and raven
chen gong [Thu, 21 May 2020 09:15:34 +0000 (17:15 +0800)]
drm/amd/powerpay: Disable gfxoff when setting manual mode on picasso and raven

[Problem description]
1. Boot up picasso platform, launches desktop, Don't do anything (APU enter into "gfxoff" state)
2. Remote login to platform using SSH, then type the command line:
sudo su -c "echo manual > /sys/class/drm/card0/device/power_dpm_force_performance_level"
sudo su -c "echo 2 > /sys/class/drm/card0/device/pp_dpm_sclk" (fix SCLK to 1400MHz)
3. Move the mouse around in Window
4. Phenomenon :  The screen frozen

Tester will switch sclk level during glmark2 run time.
APU will enter "gfxoff" state intermittently during glmark2 run time.
The system got hanged if fix GFXCLK to 1400MHz when APU is in "gfxoff"
state.

[Debug]
1. Fix SCLK to X MHz
1400: screen frozen, screen black, then OS will reboot.
1300: screen frozen.
1200: screen frozen, screen black.
1100: screen frozen, screen black, then OS will reboot.
1000: screen frozen, screen black.
900:  screen frozen, screen black, then OS will reboot.
800:  Situation Nomal, issue disappear.
700:  Situation Nomal, issue disappear.
2. SBIOS setting: AMD CBS --> SMU Debug Options -->SMU Debug --> "GFX DLDO Psm Margin Control":
50 : Situation Nomal, issue disappear.
45 : Situation Nomal, issue disappear.
40 : Situation Nomal, issue disappear.
35 : Situation Nomal, issue disappear.
30 : screen black.
25 : screen frozen, then blurred screen.
20 : screen frozen.
15 : screen black.
10 : screen frozen.
5  : screen frozen, then blurred screen.
3. Disable GFXOFF feature
Situation Nomal, issue disappear.

[Why]
Through a period of time debugging with Sys Eng team and SMU team, Sys
Eng team said this is voltage/frequency marginal issue not a F/W or H/W
bug. This experiment proves that default targetPsm [for f=1400MHz] is
not sufficient when GFXOFF is enabled on Picasso.

SMU team think it is an odd test conditions to force sclk="1400MHz" when
GPU is in "gfxoff" state,then wake up the GFX. SCLK should be in the
"lowest frequency" when gfxoff.

[How]
Disable gfxoff when setting manual mode.
Enable gfxoff when setting other mode(exiting manual mode) again.

By the way, from the user point of view, now that user switch to manual
mode and force SCLK Frequency, he don't want SCLK be controlled by
workload.It becomes meaningless to "switch to manual mode" if APU enter "gfxoff"
due to lack of workload at this point.

Tips: Same issue observed on Raven.

Signed-off-by: chen gong <curry.gong@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: fix pm sysfs node handling (v2)
Alex Deucher [Thu, 21 May 2020 14:08:11 +0000 (10:08 -0400)]
drm/amdgpu: fix pm sysfs node handling (v2)

Fix typos that prevented them from showing up.

v2: switch other files in addition to pp_clk_voltage

Fixes: 4e01847c38f7a5 ("drm/amdgpu: optimize amdgpu device attribute code")
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1150
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Evan Quan <evan.quan@amd.com>
4 years agodrm/amdgpu: move gpu_info parsing after common early init
Alex Deucher [Fri, 15 May 2020 18:04:17 +0000 (14:04 -0400)]
drm/amdgpu: move gpu_info parsing after common early init

We need to get the silicon revision id before we parse
the firmware in order to load the correct gpu info firmware
for raven2 variants.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1103
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: move discovery gfx config fetching
Alex Deucher [Fri, 15 May 2020 18:00:11 +0000 (14:00 -0400)]
drm/amdgpu: move discovery gfx config fetching

Move it into the fw_info function since it's logically part
of the same functionality.

Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agoMerge branch 'linux-5.8' of git://github.com/skeggsb/linux into drm-next
Dave Airlie [Fri, 22 May 2020 03:17:05 +0000 (13:17 +1000)]
Merge branch 'linux-5.8' of git://github.com/skeggsb/linux into drm-next

- HD audio fixes on recent systems
- vGPU detection (fail probe if we're on one, for now)
- Interlaced mode fixes (mostly avoidance on Turing, which doesn't support it)
- SVM improvements/fixes
- NVIDIA format modifier support
- Misc other fixes.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Ben Skeggs <skeggsb@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/
4 years agoMerge tag 'mediatek-drm-next-5.8' of https://git.kernel.org/pub/scm/linux/kernel...
Dave Airlie [Fri, 22 May 2020 02:20:07 +0000 (12:20 +1000)]
Merge tag 'mediatek-drm-next-5.8' of https://git./linux/kernel/git/chunkuang.hu/linux into drm-next

Mediatek DRM Next for Linux 5.8

This include dpi pin mode swap, config mipi_tx current and impedance,
and some fixup.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20200521014612.17175-1-chunkuang.hu@kernel.org
4 years agodrm/nouveau/dispnv50: fix runtime pm imbalance on error
Dinghao Liu [Wed, 20 May 2020 10:47:48 +0000 (18:47 +0800)]
drm/nouveau/dispnv50: fix runtime pm imbalance on error

pm_runtime_get_sync() increments the runtime PM usage counter even
the call returns an error code. Thus a pairing decrement is needed
on the error handling path to keep the counter balanced.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: fix runtime pm imbalance on error
Dinghao Liu [Wed, 20 May 2020 10:36:04 +0000 (18:36 +0800)]
drm/nouveau: fix runtime pm imbalance on error

pm_runtime_get_sync() increments the runtime PM usage counter even
the call returns an error code. Thus a pairing decrement is needed
on the error handling path to keep the counter balanced.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: fix runtime pm imbalance on error
Dinghao Liu [Wed, 20 May 2020 10:25:49 +0000 (18:25 +0800)]
drm/nouveau: fix runtime pm imbalance on error

pm_runtime_get_sync() increments the runtime PM usage counter even
the call returns an error code. Thus a pairing decrement is needed
on the error handling path to keep the counter balanced.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/debugfs: fix runtime pm imbalance on error
Dinghao Liu [Wed, 20 May 2020 10:14:53 +0000 (18:14 +0800)]
drm/nouveau/debugfs: fix runtime pm imbalance on error

pm_runtime_get_sync() increments the runtime PM usage counter even
the call returns an error code. Thus a pairing decrement is needed
on the error handling path to keep the counter balanced.

Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/nouveau/hmm: fix migrate zero page to GPU
Ralph Campbell [Wed, 20 May 2020 18:36:52 +0000 (11:36 -0700)]
drm/nouveau/nouveau/hmm: fix migrate zero page to GPU

When calling OpenCL clEnqueueSVMMigrateMem() on a region of memory that
is backed by pte_none() or zero pages, migrate_vma_setup() will fill the
source PFN array with an entry indicating the source page is zero.
Use this to optimize migration to device private memory by allocating
GPU memory and zero filling it instead of failing to migrate the page.

Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/nouveau/hmm: fix nouveau_dmem_chunk allocations
Ralph Campbell [Tue, 21 Apr 2020 23:11:07 +0000 (16:11 -0700)]
drm/nouveau/nouveau/hmm: fix nouveau_dmem_chunk allocations

In nouveau_dmem_init(), a number of struct nouveau_dmem_chunk are allocated
and put on the dmem->chunk_empty list. Then in nouveau_dmem_pages_alloc(),
a nouveau_dmem_chunk is removed from the list and GPU memory is allocated.
However, the nouveau_dmem_chunk is never removed from the chunk_empty
list nor placed on the chunk_free or chunk_full lists. This results
in only one chunk ever being actually used (2MB) and quickly leads to
migration to device private memory failures.

Fix this by having just one list of free device private pages and if no
pages are free, allocate a chunk of device private pages and GPU memory.

Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms/nv50-: Share DP SST mode_valid() handling with MST
Lyude Paul [Mon, 11 May 2020 22:41:27 +0000 (18:41 -0400)]
drm/nouveau/kms/nv50-: Share DP SST mode_valid() handling with MST

Currently, the nv50_mstc_mode_valid() function is happy to take any and
all modes, even the ones we can't actually support sometimes like
interlaced modes.

Luckily, the only difference between the mode validation that needs to
be performed for MST vs. SST is that eventually we'll need to check the
minimum PBN against the MSTB's full PBN capabilities (remember-we don't
care about the current bw state here). Otherwise, all of the other code
can be shared.

So, we move all of the common mode validation in
nouveau_connector_mode_valid() into a separate helper,
nv50_dp_mode_valid(), and use that from both nv50_mstc_mode_valid() and
nouveau_connector_mode_valid(). Note that we allow for returning the
calculated clock that nv50_dp_mode_valid() came up with, since we'll
eventually want to use that for PBN calculation in
nv50_mstc_mode_valid().

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms/nv50-: Move 8BPC limit for MST into nv50_mstc_get_modes()
Lyude Paul [Mon, 11 May 2020 22:41:26 +0000 (18:41 -0400)]
drm/nouveau/kms/nv50-: Move 8BPC limit for MST into nv50_mstc_get_modes()

This just limits the BPC for MST connectors to a maximum of 8 from
nv50_mstc_get_modes(), instead of doing so during
nv50_msto_atomic_check(). This doesn't introduce any functional changes
yet (other then userspace now lying about the max bpc, but we can't
support that yet anyway so meh). But, we'll need this in a moment so
that we can share mode validation between SST and MST which will fix
some real world issues.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms/gv100-: Add support for interlaced modes
Lyude Paul [Mon, 11 May 2020 22:41:25 +0000 (18:41 -0400)]
drm/nouveau/kms/gv100-: Add support for interlaced modes

We advertise being able to set interlaced modes, so let's actually make
sure to do that. Otherwise, we'll end up hanging the display engine due
to trying to set a mode with timings adjusted for interlacing without
telling the hardware it's actually an interlaced mode.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms/nv50-: Probe SOR and PIOR caps for DP interlacing support
Lyude Paul [Mon, 11 May 2020 22:41:24 +0000 (18:41 -0400)]
drm/nouveau/kms/nv50-: Probe SOR and PIOR caps for DP interlacing support

Right now, we make the mistake of allowing interlacing on all
connectors. Nvidia hardware does not always support interlacing with DP
though, so we need to make sure that we don't allow interlaced modes to
be set in such situations as otherwise we'll end up accidentally hanging
the display HW.

This fixes some hangs with Turing, which would be caused by attempting
to set an interlaced mode on hardware that doesn't support it. This
patch likely fixes other hardware hanging in the same way as well.

Note that we say we probe PIOR caps, but they don't actually have any
interlacing caps. So, the get_caps() function for PIORs just sets
interlacing support to true.

Changes since v1:
* Actually probe caps correctly this time, both on EVO and NVDisplay.
Changes since v2:
* Fix probing for < GF119
* Use vfunc table, in prep for adding more caps in the future.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms/nv50-: Initialize core channel in nouveau_display_create()
Lyude Paul [Mon, 11 May 2020 22:41:23 +0000 (18:41 -0400)]
drm/nouveau/kms/nv50-: Initialize core channel in nouveau_display_create()

We'll need the core channel initialized and ready by the time that we
start creating modesetting objects, so that we can call the
NV507D_GET_CAPABILITIES method to make the hardware expose it's
modesetting capabilities for later probing.

So, when loading the driver prepare the core channel from within
nouveau_display_create(). Everywhere else, we initialize the core
channel during resume.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/hda/gv100-: NV_PDISP_SF_AUDIO_CNTRL0 register moved
Ben Skeggs [Wed, 6 May 2020 04:41:01 +0000 (14:41 +1000)]
drm/nouveau/disp/hda/gv100-: NV_PDISP_SF_AUDIO_CNTRL0 register moved

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/hda/gf119-: select HDA device entry based on bound head
Ben Skeggs [Wed, 6 May 2020 04:40:58 +0000 (14:40 +1000)]
drm/nouveau/disp/hda/gf119-: select HDA device entry based on bound head

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/hda/gf119-: add HAL for programming device entry in SF
Ben Skeggs [Wed, 6 May 2020 04:40:56 +0000 (14:40 +1000)]
drm/nouveau/disp/hda/gf119-: add HAL for programming device entry in SF

Register has moved on GV100.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/hda/gt215-: pass head to nvkm_ior.hda.eld()
Ben Skeggs [Wed, 6 May 2020 04:40:52 +0000 (14:40 +1000)]
drm/nouveau/disp/hda/gt215-: pass head to nvkm_ior.hda.eld()

We're going to use the bound head to select HDA device entry.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/nv50-: increase timeout on pio channel free() polling
Ben Skeggs [Wed, 6 May 2020 04:40:45 +0000 (14:40 +1000)]
drm/nouveau/disp/nv50-: increase timeout on pio channel free() polling

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Fix regression by audio component transition
Takashi Iwai [Thu, 16 Apr 2020 07:54:28 +0000 (09:54 +0200)]
drm/nouveau/kms: Fix regression by audio component transition

Since the commit 742db30c4ee6 ("drm/nouveau: Add HD-audio component
notifier support"), the nouveau driver notifies and pokes the HD-audio
HPD and ELD via audio component, but this seems broken.  The culprit
is the naive assumption that crtc->index corresponds to the HDA pin.
Actually this rather corresponds to the MST dev_id (alias "pipe" in
the audio component framework) while the actual port number is given
from the output ior id number.

This patch corrects the assignment of port and dev_id arguments in the
audio component ops to recover from the HDMI/DP audio regression.

Fixes: 742db30c4ee6 ("drm/nouveau: Add HD-audio component notifier support")
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=207223
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/device: use regular PRI accessors in chipset detection
Ben Skeggs [Thu, 30 Apr 2020 04:08:53 +0000 (14:08 +1000)]
drm/nouveau/device: use regular PRI accessors in chipset detection

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/device: detect vGPUs
Karol Herbst [Tue, 28 Apr 2020 16:54:04 +0000 (18:54 +0200)]
drm/nouveau/device: detect vGPUs

Using ENODEV as this prevents probe failed errors in dmesg.

v2: move check further down

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/device: detect if changing endianness failed
Karol Herbst [Tue, 28 Apr 2020 16:54:03 +0000 (18:54 +0200)]
drm/nouveau/device: detect if changing endianness failed

v2: relax the checks a little

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/device: rework mmio mapping code to get rid of second map
Karol Herbst [Tue, 28 Apr 2020 16:54:02 +0000 (18:54 +0200)]
drm/nouveau/device: rework mmio mapping code to get rid of second map

Fixes warnings on GPUs with smaller a smaller mmio region like vGPUs.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/mmu: Remove unneeded semicolon
Zheng Bin [Fri, 24 Apr 2020 07:36:01 +0000 (15:36 +0800)]
drm/nouveau/mmu: Remove unneeded semicolon

Fixes coccicheck warning:

drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.h:307:2-3: Unneeded semicolon
drivers/gpu/drm/nouveau/nvkm/subdev/mmu/vmm.c:583:2-3: Unneeded semicolon

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: Use generic helper to check _PR3 presence
Kai-Heng Feng [Thu, 23 Apr 2020 06:23:58 +0000 (14:23 +0800)]
drm/nouveau: Use generic helper to check _PR3 presence

Replace nouveau_pr3_present() in favor of a more generic one,
pci_pr3_present().

Also the presence of upstream bridge _PR3 doesn't need to go hand in
hand with device's _DSM, so check _PR3 before _DSM.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/acr: Use kmemdup instead of kmalloc and memcpy
Zou Wei [Tue, 21 Apr 2020 12:37:31 +0000 (20:37 +0800)]
drm/nouveau/acr: Use kmemdup instead of kmalloc and memcpy

Fixes coccicheck warning:

drivers/gpu/drm/nouveau/nvkm/subdev/acr/hsfw.c:103:23-30: WARNING opportunity for kmemdup
drivers/gpu/drm/nouveau/nvkm/subdev/acr/hsfw.c:113:22-29: WARNING opportunity for kmemdup

Fixes: 22dcda45a3d1 ("drm/nouveau/acr: implement new subdev to replace "secure boot"")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zou Wei <zou_wei@huawei.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/core/memory: remove redundant assignments to variable ret
Colin Ian King [Sat, 29 Feb 2020 00:53:07 +0000 (00:53 +0000)]
drm/nouveau/core/memory: remove redundant assignments to variable ret

The variable ret is being initialized with a value that is never
read and it is being updated later with a new value. The initialization
is redundant and can be removed.

Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/svm: map pages after migration
Ralph Campbell [Wed, 4 Mar 2020 00:13:39 +0000 (16:13 -0800)]
drm/nouveau/svm: map pages after migration

When memory is migrated to the GPU, it is likely to be accessed by GPU
code soon afterwards. Instead of waiting for a GPU fault, map the
migrated memory into the GPU page tables with the same access permissions
as the source CPU page table entries. This preserves copy on write
semantics.

Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/disp/gv100-: expose capabilities class
Ben Skeggs [Wed, 12 Feb 2020 23:39:34 +0000 (09:39 +1000)]
drm/nouveau/disp/gv100-: expose capabilities class

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/bios: move ACPI _ROM handling
Ben Skeggs [Wed, 29 Jan 2020 08:27:39 +0000 (18:27 +1000)]
drm/nouveau/bios: move ACPI _ROM handling

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: remove open-coded version of remove_conflicting_pci_framebuffers()
Ben Skeggs [Sun, 2 Feb 2020 03:55:23 +0000 (13:55 +1000)]
drm/nouveau: remove open-coded version of remove_conflicting_pci_framebuffers()

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/gr/gk20a: move MODULE_FIRMWARE firmware definitions
Ben Skeggs [Fri, 7 Feb 2020 02:39:25 +0000 (12:39 +1000)]
drm/nouveau/gr/gk20a: move MODULE_FIRMWARE firmware definitions

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/ibus: use nvkm_subdev_new_()
Ben Skeggs [Tue, 11 Feb 2020 07:36:49 +0000 (17:36 +1000)]
drm/nouveau/ibus: use nvkm_subdev_new_()

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/core: add nvkm_subdev_new_() for bare subdevs
Ben Skeggs [Tue, 11 Feb 2020 07:34:58 +0000 (17:34 +1000)]
drm/nouveau/core: add nvkm_subdev_new_() for bare subdevs

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Support NVIDIA format modifiers
James Jones [Mon, 10 Feb 2020 23:15:55 +0000 (15:15 -0800)]
drm/nouveau/kms: Support NVIDIA format modifiers

Allow setting the block layout of a nouveau FB
object using DRM format modifiers.  When
specified, the format modifier block layout and
kind overrides the GEM buffer's implicit layout
and kind.  The specified format modifier is
validated against the list of modifiers supported
by the target display hardware.

v2: Used Tesla family instead of NV50 chipset compare
v4: Do not cache kind, tile_mode in nouveau_framebuffer
v5: Resolved against nouveau_framebuffer cleanup

Signed-off-by: James Jones <jajones@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Check framebuffer size against bo
James Jones [Mon, 10 Feb 2020 23:15:54 +0000 (15:15 -0800)]
drm/nouveau/kms: Check framebuffer size against bo

Make sure framebuffer dimensions and tiling
parameters will not result in accesses beyond the
end of the GEM buffer they are bound to.

v3: Return EINVAL when creating FB against BO with
    unsupported tiling
v5: Resolved against nouveau_framebuffer cleanup

Signed-off-by: James Jones <jajones@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Add format mod prop to base/ovly/nvdisp
James Jones [Mon, 10 Feb 2020 23:15:53 +0000 (15:15 -0800)]
drm/nouveau/kms: Add format mod prop to base/ovly/nvdisp

Advertise support for the full list of format
modifiers supported by each class of NVIDIA
desktop GPU display hardware.  Stash the array
of modifiers in the nouveau_display struct for
use when validating userspace framebuffer
creation requests, which will be supportd in
a subsequent change.

Signed-off-by: James Jones <jajones@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/acr: ensure falcon providing acr functions is bootstrapped first
Ben Skeggs [Mon, 10 Feb 2020 22:45:04 +0000 (08:45 +1000)]
drm/nouveau/acr: ensure falcon providing acr functions is bootstrapped first

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Remove struct nouveau_framebuffer
Thomas Zimmermann [Thu, 6 Feb 2020 10:19:42 +0000 (11:19 +0100)]
drm/nouveau/kms: Remove struct nouveau_framebuffer

After its cleanup, struct nouveau_framebuffer is only a wrapper around
struct drm_framebuffer. Use the latter directly.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Remove field nvbo from struct nouveau_framebuffer
Thomas Zimmermann [Thu, 6 Feb 2020 10:19:41 +0000 (11:19 +0100)]
drm/nouveau/kms: Remove field nvbo from struct nouveau_framebuffer

The buffer object stored in nvbo is also available GEM object in obj[0]
of struct drm_framebuffer. Therefore remove nvbo in favor obj[0] and
replace all references accordingly. This may require an additional cast.

With this change we can already replace nouveau_user_framebuffer_destroy()
and nouveau_user_framebuffer_create_handle() with generic GEM helpers.
Calls to nouveau_framebuffer_new() receive a GEM object.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Move struct nouveau_framebuffer.vma to struct nouveau_fbdev
Thomas Zimmermann [Thu, 6 Feb 2020 10:19:40 +0000 (11:19 +0100)]
drm/nouveau/kms: Move struct nouveau_framebuffer.vma to struct nouveau_fbdev

The vma field of struct nouveau_framebuffer is a special field for the
the accelerated fbdev console. Hence there's at most one single instance
for the active console. Moving it into struct nouveau_fbdev makes struct
nouveau_framebuffer slightly smaller and brings it closer to struct
drm_framebuffer.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau/kms: Remove unused fields from struct nouveau_framebuffer
Thomas Zimmermann [Thu, 6 Feb 2020 10:19:39 +0000 (11:19 +0100)]
drm/nouveau/kms: Remove unused fields from struct nouveau_framebuffer

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/nouveau: fix out-of-tree module build
Ben Skeggs [Wed, 29 Jan 2020 07:32:22 +0000 (17:32 +1000)]
drm/nouveau: fix out-of-tree module build

The $(srctree) addition a while back busted building the out-of-tree
version of the module, and I've been hacking it up ever since.

This allows us to work around the issue.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm: Generalized NV Block Linear DRM format mod
James Jones [Wed, 11 Dec 2019 20:55:47 +0000 (12:55 -0800)]
drm: Generalized NV Block Linear DRM format mod

Builds upon the existing NVIDIA 16Bx2 block linear
format modifiers by adding more "fields" to the
existing parameterized
DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK format modifier
macro that allow fully defining a unique-across-
all-NVIDIA-hardware bit layout using a minimal
set of fields and values.  The new modifier macro
DRM_FORMAT_MOD_NVIDIA_BLOCK_LINEAR_2D is
effectively backwards compatible with the existing
macro, introducing a superset of the previously
definable format modifiers.

Backwards compatibility has two quirks.  First,
the zero value for the "kind" field, which is
implied by the DRM_FORMAT_MOD_NVIDIA_16BX2_BLOCK
macro, must be special cased in drivers and
assumed to map to the pre-Turing generic kind of
0xfe, since a kind of "zero" is reserved for
linear buffer layouts on all GPUs.

Second, it is assumed backwards compatibility
is only needed when running on Tegra GPUs, and
specifically Tegra GPUs prior to Xavier.  This
is based on two assertions:

-Tegra GPUs prior to Xavier used a slightly
 different raw bit layout than desktop GPUs,
 making it impossible to directly share block
 linear buffers between the two.

-Support for the existing block linear modifiers
 was incomplete, making them useful only for
 exporting buffers created by nouveau and
 importing them to Tegra DRM as framebuffers for
 scan out.  There was no support for adding
 framebuffers using format modifiers in nouveau,
 nor importing dma-buf/PRIME GEM objects into
 nouveau userspace drivers with modifiers in Mesa.

Hence it is assumed the prior modifiers were not
intended for use on desktop GPUs, and as a
corollary, were not intended to support sharing
block linear buffers across two different NVIDIA
GPUs.

v2:
  - Added canonicalize helper function

v3:
  - Added additional bit to compression field to
    support Tesla (NV5x,G8x,G9x,GT1xx,GT2xx) class
    chips.

Signed-off-by: James Jones <jajones@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
4 years agodrm/amdgpu: resize VRAM BAR for CPU access on gfx10
Alan Swanson [Thu, 21 May 2020 20:29:30 +0000 (21:29 +0100)]
drm/amdgpu: resize VRAM BAR for CPU access on gfx10

Try to resize BAR0 to let CPU access all of VRAM on Navi. Syncs
code with previous gfx generations from commit d6895ad39f3b39
("drm/amdgpu: resize VRAM BAR for CPU access v6").

Signed-off-by: Alan Swanson <reiver@improbability.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: drop navi pcie bw callback
Alex Deucher [Tue, 19 May 2020 21:06:20 +0000 (17:06 -0400)]
drm/amdgpu: drop navi pcie bw callback

It's not implemented yet so just drop it so the sysfs
pcie bw file returns an appropriate error instead of
garbage.

Reviewed-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-By: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: improve error handling in pcie_bw
Alex Deucher [Tue, 19 May 2020 20:54:55 +0000 (16:54 -0400)]
drm/amdgpu: improve error handling in pcie_bw

1. Initialize the counters to 0 in case the callback
   fails to initialize them.
2. The counters don't exist on APUs so return an error
   for them.
3. Return an error if the callback doesn't exist.

Reviewed-by: Yong Zhao <Yong.Zhao@amd.com>
Reviewed-By: Kent Russell <kent.russell@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdkfd: fix restore worker race condition
Philip Yang [Thu, 21 May 2020 13:56:58 +0000 (09:56 -0400)]
drm/amdkfd: fix restore worker race condition

In free memory of gpu path, remove bo from validate_list to make sure
restore worker don't access the BO any more, then unregister bo MMU
interval notifier. Otherwise, the restore worker will crash in the
middle of validating BO user pages if MMU interval notifer is gone.

Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: off by one in amdgpu_device_attr_create_groups() error handling
Dan Carpenter [Wed, 20 May 2020 15:25:56 +0000 (18:25 +0300)]
drm/amdgpu: off by one in amdgpu_device_attr_create_groups() error handling

This loop in the error handling code should start a "i - 1" and end at
"i == 0".  Currently it starts a "i" and ends at "i == 1".  The result
is that it removes one attribute that wasn't created yet, and leaks the
zeroeth attribute.

Fixes: 4e01847c38f7 ("drm/amdgpu: optimize amdgpu device attribute code")
Acked-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Kevin Wang <kevin1.wang@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/smu10: Replace one-element array and use struct_size() helper
Gustavo A. R. Silva [Tue, 19 May 2020 22:55:45 +0000 (17:55 -0500)]
drm/amdgpu/smu10: Replace one-element array and use struct_size() helper

The current codebase makes use of one-element arrays in the following
form:

struct something {
    int length;
    u8 data[1];
};

struct something *instance;

instance = kmalloc(sizeof(*instance) + size, GFP_KERNEL);
instance->length = size;
memcpy(instance->data, source, size);

but the preferred mechanism to declare variable-length types such as
these ones is a flexible array member[1][2], introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on. So, replace
the one-element array with a flexible-array member.

Also, make use of the new struct_size() helper to properly calculate the
size of struct smu10_voltage_dependency_table.

This issue was found with the help of Coccinelle and, audited and fixed
_manually_.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Set/Reset avmute when disable/enable stream
Jinze Xu [Thu, 7 May 2020 07:19:37 +0000 (15:19 +0800)]
drm/amd/display: Set/Reset avmute when disable/enable stream

[Why]
When disconnect fe from be, something such as unstable clock may cause
garbage occurs.

[How]
Send set avmute at the beginning of disable stream and send reset avmute
at the end of enable stream.

Signed-off-by: Jinze Xu <jinze.xu@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Acked-by: Tony Cheng <Tony.Cheng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fix incorrectly pruned modes with deep color
Stylon Wang [Thu, 30 Apr 2020 08:40:09 +0000 (16:40 +0800)]
drm/amd/display: Fix incorrectly pruned modes with deep color

[Why]
When "max bpc" is set to enable deep color, some modes are removed from
the list if they fail validation on max bpc. These modes should be kept
if they validates fine with lower bpc.

[How]
- Retry with lower bpc in mode validation.
- Same in atomic commit to apply working bpc, not necessarily max bpc.

Signed-off-by: Stylon Wang <stylon.wang@amd.com>
Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: correct rn NUM_VMID
Dmytro Laktyushkin [Mon, 4 May 2020 19:34:30 +0000 (15:34 -0400)]
drm/amd/display: correct rn NUM_VMID

Save the correct num vmid during resource creation and fix RN gpuvm
level from 1 to 16 vmid entries.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Avoid pipe split when plane is too small
Nicholas Kazlauskas [Wed, 6 May 2020 18:21:35 +0000 (14:21 -0400)]
drm/amd/display: Avoid pipe split when plane is too small

[Why]
The minimum plane size we can support in DML is 16x16. If we try to pass
a 16x16 plane with dynamic pipe split then validation will fail since it
tries to split it into two pipes, each 8x8.

Some userspace doesn't check that the commit fails and because the
commit fails the old state is retained, resulting in corruption.

[How]
Add a workaround to avoid pipe split if any plane is 16x16 or smaller.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Defer cursor lock until after VUPDATE
Nicholas Kazlauskas [Mon, 4 May 2020 20:49:28 +0000 (16:49 -0400)]
drm/amd/display: Defer cursor lock until after VUPDATE

[Why]
We dropped the delay after changed the cursor functions locking the
entire pipe to locking just the CURSOR registers to fix page flip
stuttering - this introduced cursor stuttering instead, and an underflow
issue.

The cursor update can be delayed indefinitely if the cursor update
repeatedly happens right around VUPDATE.

The underflow issue can happen if we do a viewport update on a pipe
on the same frame where a cursor update happens around VUPDATE - the
old cursor registers are retained which can be in an invalid position.

This can cause a pipe hang and indefinite underflow.

[How]
The complex, ideal solution to the problem would be a software
triple buffering mechanism from the DM layer to program only one cursor
update per frame just before VUPDATE.

The simple workaround until we have that infrastructure in place is
this change - bring back the delay until VUPDATE before locking, but
with some corrections to the calculations.

This didn't work for all timings before because the calculation for
VUPDATE was wrong - it was using the offset from VSTARTUP instead and
didn't correctly handle the case where VUPDATE could be in the back
porch.

Add a new hardware sequencer function to use the existing helper to
calculate the real VUPDATE start and VUPDATE end - VUPDATE can last
multiple lines after all.

Change the udelay to incorporate the width of VUPDATE as well.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: DP training to set properly SCRAMBLING_DISABLE
Vladimir Stempen [Tue, 28 Apr 2020 17:04:35 +0000 (13:04 -0400)]
drm/amd/display: DP training to set properly SCRAMBLING_DISABLE

[Why]
DP training sequence to set SCRAMBLING_DISABLE bit properly based on
training pattern - per DP Spec.

[How]
Update dpcd_pattern.v1_4.SCRAMBLING_DISABLE with 1 for TPS1, TPS2, TPS3,
but not for TPS4.

Signed-off-by: Vladimir Stempen <vladimir.stempen@amd.com>
Reviewed-by: Wenjing Liu <Wenjing.Liu@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Remove dml_common_def file
Rodrigo Siqueira [Fri, 24 Apr 2020 13:53:07 +0000 (09:53 -0400)]
drm/amd/display: Remove dml_common_def file

During the rework for removing the FPU issues, I found the following
warning:

 [..] dml_common_defs.o: warning: objtool: dml_round()+0x9: FPU
      instruction outside of kernel_fpu_{begin,end}()

This file has a single function that does not need to be in a specific
file. This commit drop dml_common_defs file, and move dml_round function
to dml_inline_defs.

CC: Christian König <christian.koenig@amd.com>
CC: Alexander Deucher <Alexander.Deucher@amd.com>
CC: Peter Zijlstra <peterz@infradead.org>
CC: Tony Cheng <tony.cheng@amd.com>
CC: Harry Wentland <hwentlan@amd.com>
Signed-off-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: fix and simplify pipe split logic
Dmytro Laktyushkin [Mon, 27 Apr 2020 18:55:08 +0000 (14:55 -0400)]
drm/amd/display: fix and simplify pipe split logic

Current odm/mpc combine logic to detect which pipes need to split
logically is flawed leading to incorrect pipe merge/split operations
being taken.

This change cleans up the logic and fixes the logical errors.

Signed-off-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Eric Bernstein <Eric.Bernstein@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Minimize DSC resource re-assignment
Nikola Cornij [Thu, 28 Nov 2019 19:14:56 +0000 (14:14 -0500)]
drm/amd/display: Minimize DSC resource re-assignment

[why]
Assigning a different DSC resource than the one previosly used is
currently not handled. This causes black screen on mode change when more
than one monitor is connected on some ASICs.

[how]
- Acquire the previously used DSC if available
- Make sure re-program is triggered if new DSC is used

Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Nikola Cornij <nikola.cornij@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: add condition to set MP1 state on gpu reset
Likun Gao [Mon, 18 May 2020 10:19:30 +0000 (18:19 +0800)]
drm/amdgpu: add condition to set MP1 state on gpu reset

Only ras supportted need to set MP1 state to prepare for unload before
reloading SMU FW.

Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdkfd: report the real PCI bus number
Evan Quan [Thu, 21 May 2020 03:30:28 +0000 (11:30 +0800)]
drm/amdkfd: report the real PCI bus number

Since the PCI bus number retrieved by PCI_BUS_NUM(pdev->devfn)
is wrong.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu fix incorrect sysfs remove behavior for xgmi
Jack Zhang [Mon, 18 May 2020 03:15:52 +0000 (11:15 +0800)]
drm/amdgpu fix incorrect sysfs remove behavior for xgmi

Under xgmi setup,some sysfs fail to create for the second time of kmd
driver loading. It's due to sysfs nodes are not removed appropriately
in the last unlod time.

Changes of this patch:
1. remove sysfs for dev_attr_xgmi_error
2. remove sysfs_link adev->dev->kobj with target name.
   And it only needs to be removed once for a xgmi setup
3. remove sysfs_link hive->kobj with target name

In amdgpu_xgmi_remove_device:
1. amdgpu_xgmi_sysfs_rem_dev_info needs to be run per device
2. amdgpu_xgmi_sysfs_destroy needs to be run on the last node of
device.

v2: initialize array with memset

Signed-off-by: Jack Zhang <Jack.Zhang1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/powerplay: unify the prompts on thermal interrupts
Evan Quan [Wed, 20 May 2020 10:13:50 +0000 (18:13 +0800)]
drm/amd/powerplay: unify the prompts on thermal interrupts

The prompts will contain pci address(segment/bus/port/function),
severity(warn or error) and some keywords(GPU, amdgpu). Also this
address the issue that pci bus retrieved by PCI_BUS_NUM(adev->pdev->devfn)
is wrong.

Signed-off-by: Evan Quan <evan.quan@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Enable fp16 also on DCE-11.0 - DCE-12. (v2)
Mario Kleiner [Fri, 15 May 2020 05:19:24 +0000 (07:19 +0200)]
drm/amd/display: Enable fp16 also on DCE-11.0 - DCE-12. (v2)

Testing on a Polaris11 gpu with DCE-11.2 suggests that it
seems to work fine there, so optimistically enable it for
DCE-11 and later.

v2: drop DCE 11.0 hunk.  Carrizo (DCE 11.0) has a HW bug where FP16
scaling doesn't work.  The upscale and downscale factors were
intended to block those FP16 cases and reject the commit but
nobody ever added those to atomic check.  Once those are added
to atomic check, this can be re-enabled.

Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Expose support for xBGR ordered fp16 formats.
Mario Kleiner [Fri, 15 May 2020 05:19:23 +0000 (07:19 +0200)]
drm/amd/display: Expose support for xBGR ordered fp16 formats.

Expose support for DRM_FORMAT_ABGR16161616F and
DRM_FORMAT_XBGR16161616F to the DRM core, complementing
the already existing xRGB ordered fp16 formats.

These are especially useful for creating presentable
swapchains in Vulkan for VK_FORMAT_R16G16B16A16_SFLOAT.

Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/vcn2.5: Remove old DPG workaround
James Zhu [Tue, 19 May 2020 14:25:01 +0000 (10:25 -0400)]
drm/amdgpu/vcn2.5: Remove old DPG workaround

SCRATCH2 is used to keep decode wptr as a workaround
which fix a hardware DPG decode wptr update bug for
vcn2.5 beforehand.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/jpeg2.5: Remove JPEG_ENC_MASK from clock ungating
James Zhu [Tue, 19 May 2020 14:11:05 +0000 (10:11 -0400)]
drm/amdgpu/jpeg2.5: Remove JPEG_ENC_MASK from clock ungating

Remove JPEG_ENC_MASK from clock ungating since MJPEG encoder
hasn't been support yet.

Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: resolve ras recovery vs smi race condition
John Clements [Wed, 20 May 2020 02:28:51 +0000 (10:28 +0800)]
drm/amdgpu: resolve ras recovery vs smi race condition

during ras recovery block smu access via smi

Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: John Clements <john.clements@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdkfd: Fix boolreturn.cocci warnings
Aishwarya Ramakrishnan [Mon, 18 May 2020 16:23:12 +0000 (21:53 +0530)]
drm/amdkfd: Fix boolreturn.cocci warnings

Return statements in functions returning bool should use
true/false instead of 1/0.

drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c:40:9-10:
WARNING: return of 0/1 in function 'event_interrupt_isr_v9' with return type bool

Generated by: scripts/coccinelle/misc/boolreturn.cocci

Signed-off-by: Aishwarya Ramakrishnan <aishwaryarj100@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Respect PP_STUTTER_MODE but don't override DC_DISABLE_STUTTER
Harry Wentland [Thu, 7 May 2020 15:34:08 +0000 (11:34 -0400)]
drm/amd/display: Respect PP_STUTTER_MODE but don't override DC_DISABLE_STUTTER

Signed-off-by: Harry Wentland <harry.wentland@amd.com>
Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>