drm/amdgpu: cleanup unused iterator members for sdma v2.4

Signed-off-by: Huang Rui <ray.huang@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>

drm/amdgpu: cleanup unused iterator members for sdma v3

Signed-off-by: Huang Rui <ray.huang@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>

drm/amdgpu:impl vgt_flush for VI(V5)

when shadowing enabled, tesselation app will trigger
vm fault because below three tesselation registers:

VGT_TF_RING_SIZE__CI__VI,
VGT_HS_OFFCHIP_PARAM__CI__VI,
VGT_TF_MEMORY_BASE__CI__VI,

need to be programed after vgt-flush.

Tesselation picture vm fault disappeared after vgt-flush
introduced.

v2:implement vgt-flush for CI & SI.
v3:move vgt flush inside of cntx_cntrl
v4:count vgt flush in frame_size
v5:squash in typo fix

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable uvd mgcg for Fiji.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: refine cz uvd clock gate logic.

sw clockgate was used on uvd6.0.
when uvd is idle, we gate the uvd clock.
when decode, we ungate the uvd clock.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: change log level to KERN_INFO in ci_dpm.c

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: always un-gate UVD REGS path.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/sdma: fix typo in packet setup

Use the correct define.  No functional change as both
defines are equivalent.

Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu:  Port GMC v6 driver to new SI headers (v2)

Port the GMC v6 driver over to the new SI headers.

Tested with a Tahiti SI ASIC.

(v2) Fixed a couple of typos (in commented code) and moved
     defines to si_enums.h

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add mask bit for fan control mode.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove extra placement for AMDGPU_GEM_CREATE_NO_CPU_ACCESS

This only has the effect of scanning the invisible range twice
since the topdown flag is given anyway.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: JimQu <Jim.Qu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: disable the VRAM manager on special placements v2

This disables the VRAM manager when a special placement is requested, otherwise
we play ping/pong with the buffers on every command submission.

v2: only check lpfn

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: JimQu <Jim.Qu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: refine uvd 5.0 clock gate feature.

1. fix uvd cg status not correct.
2. fix uvd pg can't work on tonga.
3. enable uvd mgcg.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Tom St Denis <tom.stdenis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: partial revert commit 01b0e7fb1.

when uvd is idle, we gate uvd clock.
and uvd is busy, we ungate uvd clock.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: not set bypass mode for uvd5.0/uvd6.0

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: add wave reader to gfx v6

Add support for the debugfs wave reader.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: port gfx6 over to new si headers (v2)

This changes the gfx v6 driver to use the new AMDGPU style
SI headers.  Also fixes a variety of coding style issues, white
space issues, and uses WREG32_FIELD in a few places where
appropriate.

Tested with a Tahiti 0x679A.

v2: Squash in typo fix patch

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: add SI defines/registers

Add missing gca MMIO registers and defines necessary for the
next patch which re-works a lot of gfx v6 to use the new SI
headers.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Introduction of SI registers (v2)

This introduces the SI registers in the amdgpu
driver style.

v2: squash duplicates fix

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu : Fix NULL pointer comparison

Replace direct comparisons to NULL i.e.
'x == NULL' with '!x'. As per coding standard.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Ravikant B Sharma <ravikant.s2@samsung.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Clean up wave gfx8 helper

De-numberify indirect register access for gfx v8.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: Clean up wave gfx7 helper

De-numberify indirect register access for gfx v7.

Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix logic error for checking amdgpu_vram_page_split

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: JimQu <Jim.Qu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay:Tonga not to start SMC if SRIOV

This patch is used for virtualization support. In virtualization case,
Tonga SMC should not be started and SMU firmware should not be loaded
if in SRIOV environment. The same operation will be applied on FIJI in
other patch.

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Trigger Huang <trigger.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: refine uvd_4.2 clock gate sequence.

1. partial revert commit 91db308d6e96.
   not set uvd bypass mode.
2. enable uvd cg before initialize uvd.
3. set uvd clock to default value 100MHz.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: delete duplicate code in smu7_hwmgr.c

Voltage Controller have been enabled (SMC message)
before ULV enablement.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: enable voltage control by default for dgpu.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: fix code style

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: delete duplicate module parameter.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: use mask bit for deep sleep feature on dpm.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: use mask bit for deepsleep/power tune feature.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/powerplay: pp module only enable smu when dpm disabled.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:bypass avfs event manager for sriov

This patch is used for virtualization support. In virtualization case,
the initialization sequences are not totally the same as
non-Virtualization's. The avfs event manager should be bypassed if in
SRIOV virtualization case. At the same, this patch will also bypass
starting SMC within SRIOV in FIJI, so the SMU firmware loading will be
avoid, which is required in SRIOV.

Signed-off-by: Frank Min <Frank.Min@amd.com>
Signed-off-by: Xiangliang Yu <Xiangliang.Yu@amd.com>
Signed-off-by: Trigger Huang <trigger.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu:no gpu scheduler for KIQ

KIQ is used for interaction between driver and
CP, and not exposed to outside client, as such it
doesn't need to be handled by GPU scheduler.

Signed-off-by: Monk Liu <Monk.Liu@amd.com>
Signed-off-by: Xiangliang Yu <Xiangliang.Yu@amd.com>
Signed-off-by: Trigger Huang <trigger.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: Add a ring type KIQ definition

Add a new ring type definition for KIQ. KIQ is used for interaction
between driver and CP.

Signed-off-by: Xiangliang Yu <Xiangliang.Yu@amd.com>
Signed-off-by: Trigger Huang <trigger.huang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable UVD clockgating in Polaris-10/11

UVD clocks are set to be disabled, when not in use.

Signed-off-by: Maruthi Bayyavarapu <maruthi.bayyavarapu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Tom StDenis <Tom.StDenis@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/amdgpu: expose fan rpm though hwmon

Only for cards that are supported by powerplay.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: export a function to read fan rpm

Powerplay hwmgr already has an implementation, all we need to do is to call it.

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: remove amdgpu_cs_handle_lockup

In fence waiting, it never return -EDEADLK yet, so drop this function
here.

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: cleanup amdgpu_cs_ioctl to make code logicality clear

Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: make CAC feature controlled by module parameter.

use same module parameter with powercontainment

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: add new bit mask to ctrl clock stretch feature.

Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Merge tag 'zxdrm-4.10' of git://git.kernel.org/pub/scm/linux/kernel/git/shawnguo/linux into drm-next

ZTE zxdrm driver support for 4.10:

This is the initial ZTE VOU display controller DRM/KMS driver.  There
are still some features to be added, like overlay plane, scaling, and
more output devices support.  But it's already useful with dual CRTCs
and HDMI display working.

[airlied: use drm_format_plane_cpp instead of legacy api]
* tag 'zxdrm-4.10' of git://git.kernel.org/pub/scm/linux/kernel/git/shawnguo/linux:
  MAINTAINERS: add an entry for ZTE ZX DRM driver
  drm: zte: add initial vou drm driver
  dt-bindings: add bindings doc for ZTE VOU display controller

Merge branch 'for-upstream/mali-dp' of git://linux-arm.org/linux-ld into drm-next

Here is the list of fixes that I have for drm/mali-dp. They've been on the mailing
lists for a while and merged into linux-next for a few weeks, but due to holiday and
travel to Linux Plumbers I did not send the pull request earlier. I don't know if
these patches can be pulled into v4.9 still (they will conflict with Ville Syrjälä's
cleanup of DRM_ROTATE series that is already in drm-next), but if you do that would
be great.

* 'for-upstream/mali-dp' of git://linux-arm.org/linux-ld:
  drm: mali-dp: Clear CVAL when leaving config mode
  drm/arm: mark symbols static where possible
  drm: mali-dp: Add support for setting plane's rotation property from userspace.
  drm: mali-dp: Don't set DRM_PLANE_COMMIT_ACTIVE_ONLY
  drm: mali-dp: Store internal format and n_planes in plane state
  drm: mali-dp: Enable alpha blending
  drm: mali-dp: Refactor plane initialisation
  arm: mali-dp: Extract mode_config cleanup into malidp_fini
  drm: mali-dp: Add pitch alignment check for planes
  drm: mali-dp: Add pitch alignment check function
  drm: mali-dp: Set the drm->irq_enabled flag to match driver's state.
  drm: mali-dp: Clear the config_valid flag before using it in wait_event.

Merge tag 'dw-hdmi-next-2016-09-19' of git://git.pengutronix.de/git/pza/linux into drm-next

dw-hdmi i2c master controller

- add support for the HDMI I2C master controller, for boards that
  can have their DDC pins connected only to the HDMI TX directly.

* tag 'dw-hdmi-next-2016-09-19' of git://git.pengutronix.de/git/pza/linux:
  drm: bridge/dw_hdmi: add dw hdmi i2c bus adapter support
  drm: dw_hdmi: use of_get_i2c_adapter_by_node interface

Merge tag 'imx-drm-next-2016-11-10' of git://git.pengutronix.de/git/pza/linux into drm-next

imx-drm plane update cleanup, YUV formats

- request modeset if plane offsets changed, only the plane base
  address can be changed without disabling the plane IDMAC channel.
- cleanup of plane atomic_update
- remove unused ipu_cpmem_set_yuv_planar function
- support YUV 4:4:4, 4:2:2, NV12 and NV16 plane formats
- not only mask interrupts during irq init, also clear them
- remove a legacy check from imx-ldb
- add support to set the CSI downsizing bits
- silence an obnoxious warning during modeset

* tag 'imx-drm-next-2016-11-10' of git://git.pengutronix.de/git/pza/linux:
  gpu: ipu-di: silence videomode logspam
  gpu: ipu-v3: add ipu_csi_set_downsize
  drm/imx: imx-ldb: remove unnecessary double disable check
  gpu: ipu-v3: initially clear all interrupts
  drm/imx: ipuv3-plane: add support for YUV 4:2:2 and 4:4:4, NV12, and NV16 formats
  gpu: ipu-v3: add YUV 4:4:4 support
  gpu: ipu-cpmem: remove unused ipu_cpmem_set_yuv_planar function
  drm/imx: ipuv3-plane: let drm_plane_state_to_ubo/vbo handle chroma subsampling other than 4:2:0
  drm/imx: ipuv3-plane: merge ipu_plane_atomic_set_base into atomic_update
  drm/imx: ipuv3-plane: request modeset if plane offsets changed

Merge tag 'topic/drm-misc-2016-11-10' of git://anongit.freedesktop.org/drm-intel into drm-next

- better atomic state debugging from Rob
- fence prep from gustavo
- sumits flushed out his backlog of pending dma-buf/fence patches from
  various people
- drm_mm leak debugging plus trying to appease Kconfig (Chris)
- a few misc things all over

* tag 'topic/drm-misc-2016-11-10' of git://anongit.freedesktop.org/drm-intel: (35 commits)
  drm: Make DRM_DEBUG_MM depend on STACKTRACE_SUPPORT
  drm/i915: Restrict DRM_DEBUG_MM automatic selection
  drm: Restrict stackdepot usage to builtin drm.ko
  drm/msm: module param to dump state on error irq
  drm/msm/mdp5: add atomic_print_state support
  drm/atomic: add debugfs file to dump out atomic state
  drm/atomic: add new drm_debug bit to dump atomic state
  drm: add helpers to go from plane state to drm_rect
  drm: add helper for printing to log or seq_file
  drm: helper macros to print composite types
  reservation: revert "wait only with non-zero timeout specified (v3)" v2
  drm/ttm: fix ttm_bo_wait
  dma-buf/fence: revert "don't wait when specified timeout is zero" (v2)
  dma-buf/fence: make timeout handling in fence_default_wait consistent (v2)
  drm/amdgpu: add the interface of waiting multiple fences (v4)
  dma-buf: return index of the first signaled fence (v2)
  MAINTAINERS: update Sync File Framework files
  dma-buf/sw_sync: put fence reference from the fence creation
  dma-buf/sw_sync: mark sync_timeline_create() static
  drm: Add stackdepot include for DRM_DEBUG_MM
  ...

Merge tag 'drm-intel-next-2016-11-08' of git://anongit.freedesktop.org/git/drm-intel into drm-next

- gpu idling rework for s/r (Imre)
- vlv mappable scanout fix
- speed up probing in resume (Lyude)
- dp audio workarounds for gen9 (Dhinakaran)
- more conversion to using dev_priv internally (Ville)
- more gen9+ wm fixes and cleanups (Maarten)
- shrinker cleanup&fixes (Chris)
- reorg plane init code (Ville)
- implement support for multiple timelines (prep work for scheduler)
  from Chris and all
- untangle dev->struct_mutex locking as prep for multiple timelines
  (Chris)
- refactor bxt phy code and collect it all in intel_dpio_phy.c (Ander)
- another gvt with bugfixes all over from Zhenyu
- piles of lspcon fixes from Imre
- 90/270 rotation fixes (Ville)
- guc log buffer support (Akash+Sagar)
- fbc fixes from Paulo
- untangle rpm vs. tiling-fences/mmaps (Chris)
- fix atomic commit to wait on the right fences (Daniel Stone)

* tag 'drm-intel-next-2016-11-08' of git://anongit.freedesktop.org/git/drm-intel: (181 commits)
  drm/i915: Update DRIVER_DATE to 20161108
  drm/i915: Mark CPU cache as dirty when used for rendering
  drm/i915: Add assert for no pending GPU requests during suspend/resume in LR mode
  drm/i915: Make sure engines are idle during GPU idling in LR mode
  drm/i915: Avoid early GPU idling due to race with new request
  drm/i915: Avoid early GPU idling due to already pending idle work
  drm/i915: Limit Valleyview and earlier to only using mappable scanout
  drm/i915: Round tile chunks up for constructing partial VMAs
  drm/i915: Remove the vma from the object list upon close
  drm/i915: Reinit polling before hpd when resuming
  drm/i915: Remove redundant reprobe in i915_drm_resume
  drm/i915/dp: Extend BDW DP audio workaround to GEN9 platforms
  drm/i915/dp: BDW cdclk fix for DP audio
  drm/i915: Fix pages pin counting around swizzle quirk
  drm/i915: Fix test on inputs for vma_compare()
  drm/i915/guc: Cache the client mapping
  drm/i915: Tidy slab cache allocations
  drm/i915: Introduce HAS_64BIT_RELOC
  drm/i915: Show the execlist queue in debugfs/i915_engine_info
  drm/i915: Unify global_list into global_link
  ...

drm: Make DRM_DEBUG_MM depend on STACKTRACE_SUPPORT

0day continues to complain about trying to save a stacktrace for the
users of the drm_mm range allocator. This time, it is that m68k has no
save_stack_trace(), which is apparently guarded by STACKTRACE_SUPPORT.
Make it depend so!

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161109143906.11057-1-chris@chris-wilson.co.uk

gpu: ipu-di: silence videomode logspam

Adapting the videomode to the hardware constraints is something that
can and must happen during normal operation and isn't something that
the user can avoid. So printing a warning each time it happens isn't
helpful.

Demote this message to the debug level.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

drm/i915: Restrict DRM_DEBUG_MM automatic selection

DRM_DEBUG_MM is only valid if the DRM.ko is builtin as currently
depot_save_stack is not exported.

Fixes: 5c7fcf2db027 ("drm/i915: Enable drm_mm debug when enabling DRM_I915_DEBUG")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161108131917.6253-1-chris@chris-wilson.co.uk

drm: Restrict stackdepot usage to builtin drm.ko

I misread the kbuild result thinking that we had missed the include
(which we had for completeness anyway), what kbuild was actually warning
me about was that depot_save_stack was not exported.

Temporarily fix this by only selecting STACKDEPOT iff drm.ko is builtin

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Fixes: 5705670d0463 ("drm: Track drm_mm allocators and show leaks on shutdown")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161108131917.6253-2-chris@chris-wilson.co.uk

drm/msm: module param to dump state on error irq

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478358492-30738-8-git-send-email-robdclark@gmail.com

drm/msm/mdp5: add atomic_print_state support

We subclass drm_plane_state, so add mdp5_plane_atomic_print_state() to
dump out our own driver specific plane state.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478358492-30738-7-git-send-email-robdclark@gmail.com

drm/atomic: add debugfs file to dump out atomic state

Useful to dump current state from debugfs, if turning on the drm.debug
bit is too much overhead.

The drm_state_dump() can also be used by drivers, for example to
implement a module param that dumps state on error irqs.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478358492-30738-6-git-send-email-robdclark@gmail.com

drm/atomic: add new drm_debug bit to dump atomic state

The contents of drm_{plane,crtc,connector}_state is dumped before
commit.  If a driver extends any of the state structs, it can implement
the corresponding funcs->atomic_print_state() to add it's own driver
specific state.

Signed-off-by: Rob Clark <robdclark@gmail.com>
[seanpaul resolved conflict in drm_plane.h]
Signed-off-by: Sean Paul <seanpaul@chromium.org>

drm: add helpers to go from plane state to drm_rect

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
[seanpaul resolved conflict in drm_plane.h]
Signed-off-by: Sean Paul <seanpaul@chromium.org>

drm: add helper for printing to log or seq_file

Sometimes it is nice not to duplicate equivalent printk() and
seq_printf() code.

v2: simplify things w/ va_format, and use dev_printk, docs

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478358492-30738-3-git-send-email-robdclark@gmail.com

drm: helper macros to print composite types

I'll want to print things in a similar way in a later patch.  This will
make it easier.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478358492-30738-2-git-send-email-robdclark@gmail.com

reservation: revert "wait only with non-zero timeout specified (v3)" v2

Reverts commit fb8b7d2b9d80
("reservation: wait only with non-zero timeout specified (v3)")

Otherwise signaling might never be activated on the fences. This can
result in infinite waiting with hardware which has unreliable interrupts.

v2: still return one when the timeout is zero and we don't have any fences.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com> (v1)
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
  [sumits: fix checkpatch warnings]
Link: http://patchwork.freedesktop.org/patch/msgid/1478553376-18575-4-git-send-email-alexander.deucher@amd.com

drm/ttm: fix ttm_bo_wait

reservation_object_wait_timeout_rcu() should enable signaling even with a
zero timeout, but ttm_bo_wait() can also be called from atomic context and
then it is not a good idea to do this.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
  [sumits: fix checkpatch warnings]
Link: http://patchwork.freedesktop.org/patch/msgid/1478553376-18575-3-git-send-email-alexander.deucher@amd.com

dma-buf/fence: revert "don't wait when specified timeout is zero" (v2)

Reverts commit 847b19a39e4c
("dma-buf/fence: don't wait when specified timeout is zero")

When we don't call the wait function software signaling might never be
activated. This can cause infinite polling loops with unreliable interrupt
driven hardware.

v2: rebase on drm-next

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
  [sumits: reword commit msg for checkpatch warnings]
Link: http://patchwork.freedesktop.org/patch/msgid/1478553376-18575-2-git-send-email-alexander.deucher@amd.com

dma-buf/fence: make timeout handling in fence_default_wait consistent (v2)

Kernel functions taking a timeout usually return 1 on success even
when they get a zero timeout.

v2: agd: rebase on drm-next

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1478553376-18575-1-git-send-email-alexander.deucher@amd.com

drm/amdgpu: add the interface of waiting multiple fences (v4)

v2: agd: rebase and squash in all the previous optimizations and
changes so everything compiles.
v3: squash in Slava's 32bit build fix
v4: rebase on drm-next (fence -> dma_fence),
    squash in Monk's ioctl update patch

Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Monk Liu <monk.liu@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
 [sumits: fix checkpatch warnings]
Link: http://patchwork.freedesktop.org/patch/msgid/1478290570-30982-2-git-send-email-alexander.deucher@amd.com

dma-buf: return index of the first signaled fence (v2)

Return the index of the first signaled fence.  This information
is useful in some APIs like Vulkan.

v2: rebase on drm-next (fence -> dma_fence)

Signed-off-by: monk.liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
 [sumits: fix warnings]
Link: http://patchwork.freedesktop.org/patch/msgid/1478290570-30982-1-git-send-email-alexander.deucher@amd.com

MAINTAINERS: update Sync File Framework files

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1477515862-7980-1-git-send-email-gustavo@padovan.org

dma-buf/sw_sync: put fence reference from the fence creation

Once sw_sync_ioctl_create_fence() returns we no longer have the
*pt pointer to the fence base object thus we need to put the reference
we have from the fence creation to keep a correct reference accounting.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1477515599-7685-1-git-send-email-gustavo@padovan.org

dma-buf/sw_sync: mark sync_timeline_create() static

We get 1 warning when building kernel with W=1:
drivers/dma-buf/sw_sync.c:87:23: warning: no previous prototype for 'sync_timeline_create' [-Wmissing-prototypes]

In fact, this function is only used in the file in which it is
declared and don't need a declaration, but can be made static.
So this patch marks it 'static'.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
Link: http://patchwork.freedesktop.org/patch/msgid/1474202961-10099-1-git-send-email-baoyou.xie@linaro.org

drm: Add stackdepot include for DRM_DEBUG_MM

0day found that stackdepot.h doesn't get automatically included on all
architectures, so remember to add our #include.

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Fixes: 5705670d0463 ("drm: Track drm_mm allocators and show leaks on shutdown")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Christian König <christian.koenig@amd.com>.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161108115601.22873-1-chris@chris-wilson.co.uk

drm: mali-dp: Clear CVAL when leaving config mode

It's possible for CVAL to get set whilst we are in config mode. If this
happens, afer we leave config mode the HW will latch whatever
configuration is in the registers at the next vsync. Most likely this
will be a partial configuration, as we'll be racing against the ongoing
atomic_commit.

To avoid this, clear CVAL before leaving config mode.

Signed-off-by: Brian Starkey <brian.starkey@arm.com>
Signed-off-by: Liviu Dudau <Liviu.Dudau@arm.com>

drm/plane: add inline doc for struct drm_plane

Some of the members of struct drm_plane had extra comments so for these
add inline kernel comment to consolidate all documentation in one place.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
[danvet: Bikeshed a bit more to have real paragraphs with real
sentences.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478513013-3221-4-git-send-email-gustavo@padovan.org

drm/msm: use drm_atomic_set_fence_for_plane() to set the fence

drm_atomic_set_fence_for_plane() is smart and won't overwrite
plane_state->fence if the user already set an explicit fence there.

Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478513013-3221-3-git-send-email-gustavo@padovan.org

drm/imx: use drm_atomic_set_fence_for_plane() to set the fence

drm_atomic_set_fence_for_plane() is smart and won't overwrite
plane_state->fence if the user already set an explicit fence there.

Cc: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478513013-3221-2-git-send-email-gustavo@padovan.org

drm/atomic: add drm_atomic_set_fence_for_plane()

This new function should be used by drivers when setting a implicit
fence for the plane. It abstracts the fact that the user might have
chosen explicit fencing instead.

Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478513013-3221-1-git-send-email-gustavo@padovan.org

drm: don't override possible_crtcs for primary/cursor planes

It is kind of a pointless restriction.  If userspace does silly things
like using crtcA's cursor plane on crtcB, and then setcursor on crtcA,
it will end up with the overlay disabled on crtcB.  But userspace is
allowed to shoot itself like this.

v2: don't WARN_ON() if caller did not set ->possible_crtcs.  This keeps
    the existing behavior by default, if caller does not set the
    ->possible_crtcs.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478357521-26542-1-git-send-email-robdclark@gmail.com

drm/uapi: Add a warning that mode flags must match the xrandr definitions

Existing userspace expected the mode flags to match the xrandr
definitions 1:1, and even adding new flags in he previously unused
bits is likely to break existing userspace. Add a comment warning
people about this potential trap.

Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: "Lin, Jia" <lin.a.jia@intel.com>
Cc: Akashdeep Sharma <akashdeep.sharma@intel.com>
Cc: Jim Bride <jim.bride@linux.intel.com>
Cc: Jose Abreu <Jose.Abreu@synopsys.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478182201-26086-1-git-send-email-ville.syrjala@linux.intel.com

drm/edid: Add the missing "Hz" to VIC 58,59 comment

All the VICs apart from 58 and 59 have the word "Hz" included in the
comment. Include it for 59 and 59 as well.

Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1478177609-16762-1-git-send-email-ville.syrjala@linux.intel.com

drm/gma500: remove unused ioctl declarations

psb_gem_create_ioctl and psb_gem_mmap_ioctl are not used currently,
their declarations are useless.

Signed-off-by: Jiang Biao <jiang.biao2@zte.com.cn>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1477990879-13139-1-git-send-email-jiang.biao2@zte.com.cn

drm/gma500: make function static to eliminate compiling warning

psb_gtt_remove is only used in this file, and make it static to
eliminate missing-prototypes compiling warning.

Signed-off-by: Jiang Biao <jiang.biao2@zte.com.cn>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1477972185-24826-1-git-send-email-jiang.biao2@zte.com.cn

drm/atomic-helper: fix reference to drm_atomic_helper_commit_planes

The kernel-doc references drm_atomic_commit_planes() which does not
exist. The functions name is drm_atomic_helper_commit_planes().

Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161031173646.19453-1-stefan@agner.ch

drm: update the documentation of drm_framebuffer_unregister_private

Add obvious description to drm_framebuffer_unregister_private()
to explain it is deprecated.

Signed-off-by: Rongrong Zou <zourongrong@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1477915196-100299-1-git-send-email-zourongrong@gmail.com

drm/i915: Enable drm_mm debug when enabling DRM_I915_DEBUG

A frequent issue that arises on shutdown is the drm_mm range manager
complaining of a leak. To aide debugging those, drm can now track the
allocation callsite and print those for the leaks.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161029184214.17329-2-chris@chris-wilson.co.uk

drm: Track drm_mm allocators and show leaks on shutdown

We can use the kernel's stack tracer and depot to record the allocation
site of every drm_mm user. Then on shutdown, as well as warning that
allocated nodes still reside with the drm_mm range manager, we can
display who allocated them to aide tracking down the leak.

v2: Move Kconfig around so it lies underneath the DRM options submenu.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20161031090806.20073-1-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20161108

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Mark CPU cache as dirty when used for rendering

On LLC, or even snooped, machines rendering via the GPU ends up in the CPU
cache. This cacheline dirt also needs to be flushed to main memory when
moving to an incoherent domain, such as the display's scanout engine.
Mostly, this happens because either the object is marked as dirty from
its first use or is avoided by setting the object into the display
domain from the start.

v2: Treat WT as not requiring a clflush prior to use on the display
engine as well.

Fixes: 0f71979ab7fb ("drm/i915: Performed deferred clflush inside set-cache-level")
References: https://bugs.freedesktop.org/show_bug.cgi?id=95414
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: <stable@vger.kernel.org> # v4.0+
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161107165204.7008-1-chris@chris-wilson.co.uk

gpu: ipu-v3: add ipu_csi_set_downsize

Support downsizing to 1/2 width and/or height in the CSI.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

drm/imx: imx-ldb: remove unnecessary double disable check

Since the atomic modeset conversion, this should not be an issue
anymore.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

drm/i915: Add assert for no pending GPU requests during suspend/resume in LR mode

During resume we will reset the SW/HW tracking for each ring head/tail
pointers and so are not prepared to replay any pending requests (as
opposed to GPU reset time). Add an assert for this both to the suspend
and the resume code.

v2:
- Check for ELSP port idle already during suspend and check !gt.awake
  during resume. (Chris)
v3:
- Move the !gt.awake check to i915_gem_resume().
v4:
- s/intel_lr_engines_idle/intel_execlists_idle/ (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1478510405-11799-4-git-send-email-imre.deak@intel.com

drm/i915: Make sure engines are idle during GPU idling in LR mode

We assume that the GPU is idle once receiving the seqno via the last
request's user interrupt. In execlist mode the corresponding context
completed interrupt can be delayed though and until this latter
interrupt arrives we consider the request to be pending on the ELSP
submit port. This can cause a problem during system suspend where this
last request will be seen by the resume code as still pending. Such
pending requests are normally replayed after a GPU reset, but during
resume we reset both SW and HW tracking of the ring head/tail pointers,
so replaying the pending request with its stale tail pointer will leave
the ring in an inconsistent state. A subsequent request submission can
lead then to the GPU executing from uninitialized area in the ring
behind the above stale tail pointer.

Fix this by making sure any pending request on the ELSP port is
completed before suspending. I used a polling wait since the completion
time I measured was <1ms and since normally we only need to wait during
system suspend. GPU idling during runtime suspend is scheduled with a
delay (currently 50-100ms) after the retirement of the last request at
which point the context completed interrupt must have arrived already.

The chance of this bug was increased by

commit 1c777c5d1dcdf8fa0223fcff35fb387b5bb9517a
Author: Imre Deak <imre.deak@intel.com>
Date:   Wed Oct 12 17:46:37 2016 +0300

    drm/i915/hsw: Fix GPU hang during resume from S3-devices state

but it could happen even without the explicit GPU reset, since we
disable interrupts afterwards during the suspend sequence.

v2:
- Do an unlocked poll-wait first. (Chris)
v3-4:
- s/intel_lr_engines_idle/intel_execlists_idle/ and move
  i915.enable_execlists check to the new helper. (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98470
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1478510405-11799-3-git-send-email-imre.deak@intel.com

drm/i915: Avoid early GPU idling due to race with new request

There is a small race where a new request can be submitted and retired
after the idle worker started to run which leads to idling the GPU too
early. Fix this by deferring the idling to the pending instance of the
worker.

This scenario was pointed out by Chris.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1478510405-11799-2-git-send-email-imre.deak@intel.com

drm/i915: Avoid early GPU idling due to already pending idle work

Atm, in case an idle work handler is already pending but haven't yet
started to run, retiring a new request will not extend the active period
as required, rather simply leaves the pending idle work to be scheduled
at the original expiration time. This may lead to idling the GPU too
early. Fix this by using the delayed-work scheduler alternative which
makes sure the handler's expiration time is extended in this case.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Requested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1478510405-11799-1-git-send-email-imre.deak@intel.com

drm/i915: Limit Valleyview and earlier to only using mappable scanout

Valleyview appears to be limited to only scanning out from the first 512MiB
of the Global GTT. Lets presume that this behaviour was inherited from the
display block copied from g4x (not Ironlake) and all earlier generations
are similarly affected, though testing suggests different symptoms. For
simplicity, impose that these platforms must scanout from the mappable
region. (For extra simplicity, use HAS_GMCH_DISPLAY even though this
catches Cherryview which does not appear to be limited to the low
aperture for its scanout.)

v2: Use HAS_GMCH_DISPLAY() to more clearly convey my intent about
limiting this workaround to the old style of display engine.

v3: Update changelog to reflect testing by Ville Syrjälä
v4: Include the changes to the comments as well

Reported-by: Luis Botello <luis.botello.ortega@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98036
Fixes: 2efb813d5388 ("drm/i915: Fallback to using unmappable memory for scanout")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Akash Goel <akash.goel@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.9-rc1+
Link: http://patchwork.freedesktop.org/patch/msgid/20161107110128.28762-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com

drm/i915: Round tile chunks up for constructing partial VMAs

When we split a large object up into chunks for GTT faulting (because we
can't fit the whole object into the aperture) we have to align our cuts
with the fence registers. Each partial VMA must cover a complete set of
tile rows or the offset into each partial VMA is not aligned with the
whole image. Currently we enforce a minimum size on each partial VMA,
but this minimum size itself was not aligned to the tile row causing
distortion.

Reported-by: Andreas Reis <andreas.reis@gmail.com>
Reported-by: Chris Clayton <chris2553@googlemail.com>
Reported-by: Norbert Preining <preining@logic.at>
Tested-by: Norbert Preining <preining@logic.at>
Tested-by: Chris Clayton <chris2553@googlemail.com>
Fixes: 03af84fe7f48 ("drm/i915: Choose partial chunksize based on tile row size")
Fixes: a61007a83a46 ("drm/i915: Fix partial GGTT faulting") # enabling patch
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98402
Testcase: igt/gem_mmap_gtt/medium-copy-odd
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.9-rc1+
Link: http://patchwork.freedesktop.org/patch/msgid/20161107105443.27855-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Remove the vma from the object list upon close

Currently, the vma is being unlink from the object lookup on destroy.
However, we are meant to be decoupling it upon close so that the user
cannot access the closed vma whilst it remains active on the GPU.

[   34.074858] kernel BUG at drivers/gpu/drm/i915/i915_gem_gtt.c:3561!
[   34.074875] invalid opcode: 0000 [#1] PREEMPT SMP
[   34.074888] Modules linked in: snd_hda_intel i915 x86_pkg_temp_thermal coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel lpc_ich mei_me mei snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec_hdmi snd_hda_codec snd_hwdep snd_hda_core i2c_designware_platform i2c_designware_core snd_pcm e1000e ptp pps_core sdhci_acpi sdhci mmc_core i2c_hid [last unloaded: i915]
[   34.075010] CPU: 1 PID: 6224 Comm: gem_close_race Tainted: G     U          4.9.0-rc3-CI-CI_DRM_1800+ #1
[   34.075034] Hardware name:                  /NUC5i7RYB, BIOS RYBDWi35.86A.0355.2016.0224.1501 02/24/2016
[   34.075057] task: ffff8802459a8040 task.stack: ffffc90000524000
[   34.075074] RIP: 0010:[<ffffffffa0392cbc>]  [<ffffffffa0392cbc>] i915_gem_obj_lookup_or_create_vma+0x8c/0xc0 [i915]
[   34.075118] RSP: 0018:ffffc90000527b68  EFLAGS: 00010202
[   34.075135] RAX: ffff8802426c5e40 RBX: 0000000000000000 RCX: ffff8802447fc2a8
[   34.075158] RDX: 0000000000000000 RSI: ffff8802447fc2a8 RDI: ffff880248a4a880
[   34.075181] RBP: ffffc90000527b88 R08: 0000000000000008 R09: 0000000000000000
[   34.075203] R10: 0000000000000001 R11: 0000000000000000 R12: ffff880248a4a880
[   34.075225] R13: ffff8802447fc2a8 R14: ffff880243e9afa8 R15: ffff880248a4a9c8
[   34.075248] FS:  00007f9b43e59740(0000) GS:ffff880256c80000(0000) knlGS:0000000000000000
[   34.075273] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   34.075292] CR2: 00007f9b43419140 CR3: 000000024455d000 CR4: 00000000003406e0
[   34.075314] Stack:
[   34.075323]  0000000000000000 ffffc90000527bd0 ffff880243cb8008 ffff880243e9afa8
[   34.075353]  ffffc90000527c08 ffffffffa03874c7 ffffc90000527bb8 ffff880243e9afa8
[   34.075383]  ffff880243e9afb0 ffffc90000527e10 ffff8802447fc2a8 ffff880243cb8040
[   34.075414] Call Trace:
[   34.075435]  [<ffffffffa03874c7>] eb_lookup_vmas.isra.7+0x247/0x330 [i915]
[   34.075468]  [<ffffffffa0388c34>] i915_gem_do_execbuffer.isra.15+0x604/0x1a10 [i915]
[   34.075507]  [<ffffffffa039c957>] ? i915_gem_object_get_sg+0x347/0x380 [i915]
[   34.075532]  [<ffffffff811a69ce>] ? __might_fault+0x3e/0x90
[   34.075562]  [<ffffffffa038a430>] i915_gem_execbuffer2+0xc0/0x250 [i915]
[   34.075585]  [<ffffffff81552926>] drm_ioctl+0x1f6/0x480
[   34.075604]  [<ffffffff8100107a>] ? trace_hardirqs_on_thunk+0x1a/0x1c
[   34.075635]  [<ffffffffa038a370>] ? i915_gem_execbuffer+0x330/0x330 [i915]
[   34.075658]  [<ffffffff81202d2e>] do_vfs_ioctl+0x8e/0x690
[   34.075677]  [<ffffffff8181582d>] ? _raw_spin_unlock_irqrestore+0x3d/0x60
[   34.075700]  [<ffffffff810fcd51>] ? SyS_timer_settime+0x141/0x1e0
[   34.075721]  [<ffffffff810d6de2>] ? trace_hardirqs_on_caller+0x122/0x1b0
[   34.075742]  [<ffffffff8120336c>] SyS_ioctl+0x3c/0x70
[   34.075760]  [<ffffffff8181602e>] entry_SYSCALL_64_fastpath+0x1c/0xb1
[   34.075781] Code: 44 a0 48 c7 c2 9a 7e 43 a0 be e0 0d 00 00 48 c7 c7 a0 45 44 a0 e8 55 b8 ce e0 48 85 db 74 a3 49 83 bd f8 03 00 00 00 74 99 0f 0b <0f> 0b 48 89 da 4c 89 ee 4c 89 e7 e8 04 a9 ff ff 48 89 da 49 89
[   34.075955] RIP  [<ffffffffa0392cbc>] i915_gem_obj_lookup_or_create_vma+0x8c/0xc0 [i915]
[   34.075994]  RSP <ffffc90000527b68>

Testcase: igt/gem_close_race/basic-threads
Fixes: db6c2b4151f2 ("drm/i915: Store the vma in an rbtree...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161104161241.25871-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

Merge branch 'linux-4.10' of git://github.com/skeggsb/linux into drm-next

- Initial atomic modesetting support.  Used for "legacy" KMS interfaces,
ioctl not exposed by default, but there is a commandline option to
enable it.
- Initial DP 1.2 MST support
- Misc other code cleanups + fixes

* 'linux-4.10' of git://github.com/skeggsb/linux: (64 commits)
  drm/nouveau/fifo/gf100-: protect channel preempt with subdev mutex
  drm/nouveau/gr: fallback to legacy paths during firmware lookup
  drm/nouveau/kms/nv50: initial support for DP 1.2 multi-stream
  drm/nouveau/kms/nv50: allow encoder update to be called from other modules
  drm/nouveau/kms/nv50: rename remaining nv50_crtc to nv50_head
  drm/nouveau/kms/nv50: remove code to create ctxdma for every framebuffer
  drm/nouveau/kms/nv50: remove code to support non-atomic page flips
  drm/nouveau/kms/nv50: remove code to support non-atomic connector properties
  drm/nouveau/kms/nv50: remove code to support non-atomic dpms
  drm/nouveau/kms/nv50: remove code to support non-atomic modesets
  drm/nouveau/kms/nv50: transition to atomic interfaces internally
  drm/nouveau/kms/nv50: turn mode_set_base_atomic() into a stub
  drm/nouveau/kms/nv50: convert encoder mode_fixup into an atomic_check()
  drm/nouveau/kms/nv50: clean-up encoder functions
  drm/nouveau/kms/nv50: ensure encoder normal power state is enabled at startup
  drm/nouveau/kms/nv50: prepare ctxdma interface to be usable with atomic
  drm/nouveau/kms/nv50: separate out cursor channel commit
  drm/nouveau/kms/nv50: separate out base channel commit
  drm/nouveau/kms/nv50: separate out vblank dmi commit
  drm/nouveau/kms/nv50: separate out procamp commit
  ...

drm/nouveau/fifo/gf100-: protect channel preempt with subdev mutex

This avoids an issue that occurs when we're attempting to preempt multiple
channels simultaneously.  HW seems to ignore preempt requests while it's
still processing a previous one, which, well, makes sense.

Fixes random "fifo: SCHED_ERROR 0d []" + GPCCS page faults during parallel
piglit runs on (at least) GM107.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org

drm/nouveau/gr: fallback to legacy paths during firmware lookup

Look for firmware files using the legacy ("nouveau/nvxx_fucxxxx") path
if they cannot be found in the new, "official" path. User setups were
broken by the switch, which is bad.

There are only 4 firmware files we may want to look up that way, so
hardcode them into the lookup function. All new firmware files should
use the standard "nvidia/<chip>/gr/" path.

Fixes: 8539b37acef7 ("drm/nouveau/gr: use NVIDIA-provided external firmwares")
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org

drm/nouveau/kms/nv50: initial support for DP 1.2 multi-stream

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/kms/nv50: allow encoder update to be called from other modules

MST encoders will make use of this to share code with SOR>

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau/kms/nv50: rename remaining nv50_crtc to nv50_head

No code changes, just renames + shuffles.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>