Merge tag 'drm-intel-gt-next-2021-12-09' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

Core Changes:

- Fix PENDING_ERROR leak in dma_fence_array_signaled() (Thomas Hellström)

Driver Changes:

- Fix runtime PM handling during PXP suspend (Tejas Upadhyay)
- Improve eviction performance on discrete by implementing async TTM moves (Thomas Hellström, Maarten Lankhorst)
- Improve robustness of error capture under memory pressure (Thomas Hellström)
- Fix GuC PMU versus GPU reset handling (Umesh Nerlige Ramappa)
- Use per device iommu check (Tvrtko Ursulin)
- Make error capture work with async migration (Thomas Hellström)
- Revert incorrect implementation of Wa_1508744258 causing hangs (José Roberto de Souza)
- Disable coarse power gating on some DG2 steppings workaround (Matt Roper)
- Add IC cache invalidation workaround on DG2 (Ramalingam C)
- Move two Icelake workarounds to the right place (Raviteja Goud Talla)
- Fix error pointer dereference in i915_gem_do_execbuffer() (Dan Carpenter)
- Fixup a couple of generic and DG2 specific issues in migration code (Matthew Auld)

- Fix kernel-doc warnings in i915_drm_object.c (Randy Dunlap)
- Drop stealing of bits from i915_sw_fence function pointer (Matthew Brost)
- Introduce new macros for i915 PTE (Michael Cheng)
- Prep work for engine reset by reset domain lookup (Tejas Upadhyay)

- Fixup drm-intel-gt-next build failure (Matthew Auld)
- Fix live_engine_busy_stats selftests in GuC mode (Umesh Nerlige Ramappa)
- Remove dma_resv_prune (Maarten Lankhorst)
- Preserve huge pages enablement after driver reload (Matthew Auld)
- Fix a NULL pointer dereference in igt_request_rewind() (selftests) (Zhou Qingyang)
- Add workaround numbers to GEN7_COMMON_SLICE_CHICKEN1 whitelisting (José Roberto de Souza)
- Increase timeouts in i915_gem_contexts selftests to handle GuC being slower (Bruce Chang)

Signed-off-by: Dave Airlie <airlied@redhat.com>
# Conflicts:
# drivers/gpu/drm/i915/display/intel_fbc.c
From: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/YbIBOeqhn+nPzaYD@tursulin-mobl2

Merge tag 'drm-misc-next-2021-12-09' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

drm-misc-next for 5.17:

UAPI Changes:

Cross-subsystem Changes:

 * dma-buf: Make fences mandatory in dma_resv_add_excl_fence

Core Changes:

 * Move hashtable to legacy code
 * Return error pointers from struct drm_driver.gem_create_object

 * cma-helper: Improve public interfaces; Remove CONFIG_DRM_KMS_CMA_HELPER option
 * mipi-dbi: Don't depend on CMA helpers
 * ttm: Don't include DRM hashtable; Stop prunning fences after wait; Documentation

Driver Changes:

 * aspeed: Select CONFIG_DRM_GEM_CMA_HELPER

 * bridge/lontium-lt9611: Fix HDMI sensing
 * bridge/parade-ps8640: Fixes
 * bridge/sn65dsi86: Defer probe is no dsi host found

 * fsl-dcu: Select CONFIG_DRM_GEM_CMA_HELPER

 * i915: Remove dma_resv_prune

 * omapdrm: Fix scatterlist export; Support virtual planes; Fixes

 * panel: Boe-tv110c9m,Inx-hj110iz: Update init code

 * qxl: Use dma-resv iterator

 * rockchip: Use generic fbdev emulation

 * tidss: Fixes

 * vmwgfx: Fix leak on probe errors; Fail probing on broken hosts; New
   placement for MOB page tables; Hide internal BOs from userspace; Cleanups

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/YbHskHZc9HoAYuPZ@linux-uq9g.fritz.box

Merge tag 'du-next-20211206' of git://linuxtv.org/pinchartl/media into drm-next

R-Car DU updates:

- DSI output support
- Misc fixes

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/Ya4/MCeuNf601tL4@pendragon.ideasonboard.com

Merge tag 'amd-drm-next-5.17-2021-12-02' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-5.17-2021-12-02:

amdgpu:
- Use generic drm fb helpers
- PSR fixes
- Rework DCN3.1 clkmgr
- DPCD 1.3 fixes
- Misc display fixes can cleanups
- Clock query fixes for APUs
- LTTPR fixes
- DSC fixes
- Misc PM fixes
- RAS fixes
- OLED backlight fix
- SRIOV fixes
- Add STB (Smart Trace Buffer) for supported dGPUs
- IH rework
- Enable seamless boot for DCN3.01

amdkfd:
- Rework more stuff around IP discovery enumeration
- Further clean up of interfaces with amdgpu
- SVM fixes

radeon:
- Indentation fixes

UAPI:
- Add a new KFD header that defines some of the sysfs bitfields and enums that userspace has been using for a while
  The corresponding bit-fields and enums in user mode are defined in
  https://github.com/RadeonOpenCompute/ROCT-Thunk-Interface/blob/master/include/hsakmttypes.h

Signed-off-by: Dave Airlie <airlied@redhat.com>
# Conflicts:
# drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211202191643.5970-1-alexander.deucher@amd.com

Merge tag 'drm-misc-next-2021-11-29' of git://anongit.freedesktop.org/drm/drm-misc into drm-next

drm-misc-next for 5.17:

UAPI Changes:

Cross-subsystem Changes:

 * Move 'nomodeset' kernel boot option into DRM subsystem

Core Changes:

 * Replace several DRM_*() logging macros with drm_*() equivalents
 * panel: Add quirk for Lenovo Yoga Book X91F/L
 * ttm: Documentation fixes

Driver Changes:

 * Cleanup nomodeset handling in drivers
 * Fixes
 * bridge/anx7625: Fix reading EDID; Fix error code
 * bridge/megachips: Probe both bridges before registering
 * vboxvideo: Fix ERR_PTR usage

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
From: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/YaSVz15Q7dAlEevU@linux-uq9g.fritz.box

drm/i915/gt: Use hw_engine_masks as reset_domains

We need a way to reset engines by their reset domains.
This change sets up way to fetch reset domains of each
engine globally.

Changes since V1:
- Use static reset domain array - Ville and Tvrtko
- Use BUG_ON at appropriate place - Tvrtko

Signed-off-by: Tejas Upadhyay <tejaskumarx.surendrakumar.upadhyay@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206081026.4024401-1-tejaskumarx.surendrakumar.upadhyay@intel.com

drm/bridge: sn65dsi86: defer if there is no dsi host

Otherwise we don't get another shot at it if the bridge probes before
the dsi host is registered.  It seems like this is what *most* (but not
all) of the other bridges do.

It looks like this was missed in the conversion to attach dsi host at
probe time.

Fixes: c3b75d4734cb ("drm/bridge: sn65dsi86: Register and attach our DSI device at probe")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Doug Anderson <dianders@chromium.org>
Tested-by: Doug Anderson <dianders@chromium.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
[dianders: squashed in Stephen's simplification]
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211207215753.635841-1-robdclark@gmail.com

drm/i915/selftests: handle object rounding

Ensure we account for any object rounding due to min_page_size
restrictions.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-4-matthew.auld@intel.com

drm/i915/migrate: fix length calculation

No need to insert PTEs for the PTE window itself, also foreach expects a
length not an end offset, which could be gigantic here with a second
engine.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-3-matthew.auld@intel.com

drm/i915/migrate: fix offset calculation

Ensure we add the engine base only after we calculate the qword offset
into the PTE window.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-2-matthew.auld@intel.com

drm/i915/migrate: don't check the scratch page

The scratch page might not be allocated in LMEM(like on DG2), so instead
of using that as the deciding factor for where the paging structures
live, let's just query the pt before mapping it.

Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206112539.3149779-1-matthew.auld@intel.com

omapdrm: dss: mark runtime PM functions __maybe_unused

Using the SET_RUNTIME_PM_OPS() macro causes a warning about the
referenced functions when they are marked static but not __maybe_unused:

drivers/gpu/drm/omapdrm/dss/dss.c:1572:12: error: unused function 'dss_runtime_suspend' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dss.c:1584:12: error: unused function 'dss_runtime_resume' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dispc.c:4845:12: error: unused function 'dispc_runtime_suspend' [-Werror,-Wunused-function]
drivers/gpu/drm/omapdrm/dss/dispc.c:4860:12: error: unused function 'dispc_runtime_resume' [-Werror,-Wunused-function]

Fixes: b92f7ea556f8 ("drm/omap: dss: Make use of the helper macro SET_RUNTIME_PM_OPS()")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211205131612.3192652-1-arnd@kernel.org

drm/omap: Add a 'right overlay' to plane state

If the drm_plane has a source width that's greater than the max width
supported by a single hw overlay, then we assign a 'r_overlay' to it in
omap_plane_atomic_check().

Both overlays should have the capabilities required to handle the source
framebuffer. The only parameters that vary between the left and right
hwoverlays are the src_w, crtc_w, src_x and crtc_x as we just even chop
the fb into left and right halves.

We also take care of not creating odd width size when dealing with YUV
formats.

Since both halves need to be 'appear' side by side the zpos is
recalculated when dealing with dual overlay cases so that the other
planes zpos is consistent.

Depending on user space usage it is possible that on occasion the number
of requested planes exceeds the numbers of overlays required to display
them. In that case a failure would be returned for the plane that cannot
be handled at that time. It is up to user space to make sure the H/W
resource are not over-subscribed.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-10-narmstrong@baylibre.com

drm/omap: add plane_atomic_print_state support

Now that we added specific item to our subclassed drm_plane_state
we can add omap_plane_atomic_print_state() helper to dump out our own
driver specific plane state.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-9-narmstrong@baylibre.com

drm/omap: dynamically assign hw overlays to planes

(re)assign the hw overlays to planes based on required caps, and to
handle situations where we could not modify an in-use plane.

This means all planes advertise the superset of formats and properties.
Userspace must (as always) use atomic TEST_ONLY step for atomic updates,
as not all planes may be available for use on every frame.

The mapping of hwoverlays to plane is stored in omap_global_state, so
that state updates are atomically committed in the same way that
plane/etc state updates are managed.  This is needed because the
omap_plane_state keeps a pointer to the hwoverlay, and we don't want
global state to become out of sync with the plane state if an atomic
update fails, we hit deadlock/ backoff scenario, etc.  The use of
global_state_lock keeps multiple parallel updates which both re-assign
hwoverlays properly serialized.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-8-narmstrong@baylibre.com

drm/omap: Add global state as a private atomic object

Global shared resources (like hw overlays) for omapdrm are implemented
as a part of atomic state using the drm_private_obj infrastructure
available in the atomic core.

omap_global_state is introduced as a drm atomic private object. The two
funcs omap_get_global_state() and omap_get_existing_global_state() are
the two variants that will be used to access omap_global_state.

drm_mode_config_init() needs to be called earlier because it
creates/initializes the private_obj link list maintained by the atomic
framework. The private_obj link list has to exist prior to calling
drm_atomic_private_obj_init(). Similarly the cleanup handler are
reordered appropriately.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-7-narmstrong@baylibre.com

drm/omap: omap_plane: subclass drm_plane_state

In preparation to add omap plane state specific extensions we need to
subclass drm_plane_state and add the relevant helpers.

The addition of specific extension will be done separately.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-6-narmstrong@baylibre.com

drm/omap: introduce omap_hw_overlay

Split out the hardware overlay specifics from omap_plane.
To start, the hw overlays are statically assigned to planes.

The goal is to eventually assign hw overlays dynamically to planes
during plane->atomic_check() based on requested caps (scaling, YUV,
etc). And then perform hw overlay re-assignment if required.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-5-narmstrong@baylibre.com

drm/omap: Add ovl checking funcs to dispc_ops

In order to be able to dynamically assign overlays to planes we need to
be able to asses the overlay capabilities.

Add a helper function to be able to retrieve the supported capabilities
of an overlay.

And export the function to check if a fourcc is supported on a given
overlay.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-4-narmstrong@baylibre.com

drm/omap: Add ability to check if requested plane modes can be supported

We currently assume that an overlay has the same maximum width and
maximum height as the overlay manager. This assumption is incorrect. On
some variants the overlay manager maximum width is twice the maximum
width that the overlay can handle. We need to add the appropriate data
per variant as well as export a helper function to retrieve the data so
check can be made dynamically in omap_plane_atomic_check().

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-3-narmstrong@baylibre.com

drm/omap: add sanity plane state check

Call drm_atomic_helper_check_plane_state() from the plane
atomic_check() callback in order to add plane state sanity
checking.

It will permit filtering out totally bad scaling factors, even
if the real check are done later in the atomic commit.

Calling drm_atomic_helper_check_plane_state() also sets
plane_state->visible which will be useful when dynamically
assigning hw overlays to planes.

Signed-off-by: Benoit Parrot <bparrot@ti.com>
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117141928.771082-2-narmstrong@baylibre.com

drm: omapdrm: Export correct scatterlist for TILER backed BOs

Memory of BOs backed by TILER is not contiguous, but omap_gem_map_dma_buf()
exports it like it is. This leads to (possibly) invalid memory accesses if
another device imports such a BO.

Fix that by providing sg that correctly describes TILER memory layout.
Align TILER allocations to page, so importer to be able to correctly set
its MMU if have one. Set export size accounting for the alignment. Also,
make sure to destroy sg on unpin, as it is no longer valid.

Tested on Motorola Droid4 by using GPU (sgx540) to render.

Suggested-by: Matthijs van Duin <matthijsvanduin@gmail.com>
Signed-off-by: Ivaylo Dimitrov <ivo.g.dimitrov.75@gmail.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1637309180-31032-1-git-send-email-ivo.g.dimitrov.75@gmail.com

drm/bridge: parade-ps8640: Add backpointer to drm_device in drm_dp_aux

When we added the support for the AUX channel in commit 13afcdd7277e
("drm/bridge: parade-ps8640: Add support for AUX channel") we forgot
to set "drm_dev" to avoid the warning splat at the beginning of
drm_dp_aux_register(). Since everything was working I guess I never
noticed the splat when testing against mainline. In any case, it's
easy to fix. This is basically just like commit 6cba3fe43341 ("drm/dp:
Add backpointer to drm_device in drm_dp_aux") but just for the
parade-ps8640.

Fixes: 13afcdd7277e ("drm/bridge: parade-ps8640: Add support for AUX channel")
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Philip Chen <philipchen@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206162907.1.I1f5d1eba741e4663050ec1b8e39a753f6e42e38b@changeid

drm/panel: Update Boe-tv110c9m and Inx-hj110iz initial code

At present, we have enough panel to confirm the effect,
update the initial code to achieve the best effect.
Such as gamma, Gop timing. They are all minor modifications
and doesn't affect the lighting of the panel.

a)Boe-tv110c9m panel Optimized touch horizontal grain.
b)Inx-hj110iz panel Optimized GOP timing and gamma.

Signed-off-by: yangcong <yangcong5@huaqin.corp-partner.google.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211201023230.344976-1-yangcong5@huaqin.corp-partner.google.com

drm/i915/selftests: Follow up on increase timeout in i915_gem_contexts selftests

Follow up on below commit, to increase the timeout further on new
platforms, to accomodate the additional time required for the completion
of guc submissions for numerous requests created in loop.

commit 5e076529e2652244ec20a86d8f99ba634a16c4f4
Author: Matthew Brost <matthew.brost@intel.com>
Date:   Mon Jul 26 20:17:03 2021 -0700

    drm/i915/selftests: Increase timeout in i915_gem_contexts selftests

Signed-off-by: Bruce Chang <yu.bruce.chang@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Cc: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211207003845.12419-1-yu.bruce.chang@intel.com

drm/rockchip: pass 0 to drm_fbdev_generic_setup()

Allow drm_fbdev_generic_setup() to pick the default bpp value for the
framebuffer.

This has no functional impact because the default is 32, given that
mode_config.preferred_depth is not set for Rockchip.

Suggested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211101114622.813536-1-john@metanate.com

drm/rockchip: use generic fbdev setup

The Rockchip fbdev code does not add anything compared to
drm_fbdev_generic_setup(); the one custom function for .fb_mmap does the
same thing as gem_prime_mmap which is called by the helper.

Signed-off-by: John Keeping <john@metanate.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211029115014.264084-1-john@metanate.com

drm/i915: Introduce new macros for i915 PTE

Certain functions within i915 uses macros that are defined for
specific architectures by the mmu, such as _PAGE_RW and _PAGE_PRESENT
(Some architectures don't even have these macros defined, like ARM64).

Instead of re-using bits defined for the CPU, we should use bits
defined for i915. This patch introduces two new 64 bit macros,
GEN8_PAGE_PRESENT and GEN8_PAGE_RW, to check for bits 0 and 1 and, to
replace all occurrences of _PAGE_RW and _PAGE_PRESENT within i915.

v2(Michael Cheng): Use GEN8_ instead of I915_

Signed-off-by: Michael Cheng <michael.cheng@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
[ Move defines together with other GEN8 defines ]
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211206215245.513677-2-michael.cheng@intel.com

drm: rcar-du: Add R-Car DSI driver

The driver supports the MIPI DSI/CSI-2 TX encoder found in the R-Car V3U
SoC. It currently supports DSI mode only.

Signed-off-by: LUU HOAI <hoai.luu.ub@renesas.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Tested-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Jagan Teki <jagan@amarulasolutions.com>

dt-bindings: display: bridge: Add binding for R-Car MIPI DSI/CSI-2 TX

The R-Car MIPI DSI/CSI-2 TX is embedded in the Renesas R-Car V3U SoC. It
can operate in either DSI or CSI-2 mode, with up to four data lanes.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Rob Herring <robh@kernel.org>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>

drm: rcar-du: crtc: Support external DSI dot clock

On platforms with an external clock, both the group and crtc must be
handled accordingly to correctly pass through the external clock and
configure the DU to use the external rate.

The CRTC support was missed while adding the DSI support on the r8a779a0
which led to the output clocks being incorrectly determined.

Ensure that when a CRTC is routed through the DSI encoder, the external
clock is used without any further divider being applied.

Fixes: b291fdcf5114 ("drm: rcar-du: Add r8a779a0 device support")
Signed-off-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

drm: rcar-du: Add DSI support to rcar_du_output_name

The DSI output names were not added when the DSI pipeline support was
introduced.

Add the correct labels for these outputs, and fix the sort order to
match 'enum rcar_du_output' while we are here.

Fixes: b291fdcf5114 ("drm: rcar-du: Add r8a779a0 device support")
Suggested-by: Biju Das <biju.das.jz@bp.renesas.com>
Signed-off-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

drm: rcar-du: Fix CRTC timings when CMM is used

When the CMM is enabled, an offset of 25 pixels must be subtracted from
the HDS (horizontal display start) and HDE (horizontal display end)
registers. Fix the timings calculation, and take this into account in
the mode validation.

This fixes a visible horizontal offset in the image with VGA monitors.
HDMI monitors seem to be generally more tolerant to incorrect timings,
but may be affected too.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

drm/rcar: stop using 'imply' for dependencies

The meaning of the 'imply' keyword has changed recently, and neither the
old meaning (select the symbol if its dependencies are met) nor the new
meaning (enable it by default, but let the user set any other setting)
is what we want here.

Work around this by adding two more Kconfig options that lead to
the correct behavior: if DRM_RCAR_USE_CMM and DRM_RCAR_USE_LVDS
are enabled, that portion of the driver becomes usable, and no
configuration results in a link error.

This avoids a link failure:

arm-linux-gnueabi-ld: drivers/gpu/drm/rcar-du/rcar_du_crtc.o: in function `rcar_du_crtc_atomic_begin':
rcar_du_crtc.c:(.text+0x1444): undefined reference to `rcar_cmm_setup'
arm-linux-gnueabi-ld: drivers/gpu/drm/rcar-du/rcar_du_crtc.o: in function `rcar_du_crtc_atomic_enable':
rcar_du_crtc.c:(.text+0x14d4): undefined reference to `rcar_cmm_enable'
arm-linux-gnueabi-ld: rcar_du_crtc.c:(.text+0x1548): undefined reference to `rcar_cmm_setup'
arm-linux-gnueabi-ld: drivers/gpu/drm/rcar-du/rcar_du_crtc.o: in function `rcar_du_crtc_atomic_disable':
rcar_du_crtc.c:(.text+0x18b8): undefined reference to `rcar_cmm_disable'
arm-linux-gnueabi-ld: drivers/gpu/drm/rcar-du/rcar_du_kms.o: in function `rcar_du_modeset_init':

Link: https://lore.kernel.org/all/20200417155553.675905-5-arnd@arndb.de/
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>

drm: aspeed: select CONFIG_DRM_GEM_CMA_HELPER

The aspeed driver uses the gem_cma_helper code, but does
noto enforce enabling this through Kconfig:

x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_drv.o:(.rodata+0x2c8): undefined reference to `drm_gem_cma_prime_import_sg_table'
x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_drv.o:(.rodata+0x2d8): undefined reference to `drm_gem_cma_dumb_create'
x86_64-linux-ld: drivers/gpu/drm/aspeed/aspeed_gfx_crtc.o: in function `aspeed_gfx_pipe_update':
aspeed_gfx_crtc.c:(.text+0xe5): undefined reference to `drm_fb_cma_get_gem_obj'

Add the same 'select' that is used in other such drivers.

Fixes: 09717af7d13d ("drm: Remove CONFIG_DRM_KMS_CMA_HELPER option")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211204174637.1160725-2-arnd@kernel.org

drm: fsl-dcu: select CONFIG_DRM_GEM_CMA_HELPER

The fsl-dcu driver uses the gem_cma_helper code, but does
noto enforce enabling this through Kconfig:

arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.o:(.rodata+0x32c): undefined reference to `drm_gem_cma_prime_import_sg_table'
arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_drv.o:(.rodata+0x334): undefined reference to `drm_gem_cma_dumb_create'
arm-linux-gnueabi-ld: drivers/gpu/drm/fsl-dcu/fsl_dcu_drm_plane.o: in function `fsl_dcu_drm_plane_atomic_update':
fsl_dcu_drm_plane.c:(.text+0x144): undefined reference to `drm_fb_cma_get_gem_obj'

Add the same 'select' that is used in other such drivers.

Fixes: 09717af7d13d ("drm: Remove CONFIG_DRM_KMS_CMA_HELPER option")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20211204174637.1160725-1-arnd@kernel.org

drm/i915: Fix error pointer dereference in i915_gem_do_execbuffer()

Originally "out_fence" was set using out_fence = sync_file_create() but
which returns NULL, but now it is set with out_fence = eb_requests_create()
which returns error pointers.  The error path needs to be modified to
avoid an Oops in the "goto err_request;" path.

Fixes: 544460c33821 ("drm/i915: Multi-BB execbuf")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211202044831.29583-1-matthew.brost@intel.com

drm/i915/gen11: Moving WAs to icl_gt_workarounds_init()

Bspec page says "Reset: BUS", Accordingly moving w/a's:
Wa_1407352427,Wa_1406680159 to proper function icl_gt_workarounds_init()
Which will resolve guc enabling error

v2:
  - Previous patch rev2 was created by email client which caused the
    Build failure, This v2 is to resolve the previous broken series

Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Raviteja Goud Talla <ravitejax.goud.talla@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211203145603.4006937-1-ravitejax.goud.talla@intel.com

drm/i915/dg2: Add Wa_16013000631

Invalidate IC cache through pipe control command as part of the ctx
restore flow through indirect ctx pointer.

v2:
 - Move pipe control from xcs indirect context to the rcs indirect
   context.  We'll eventually need this on the CCS engines too, but
   support for those hasn't landed yet.

Cc: Chris Wilson <chris.p.wilson@intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116174818.2128062-5-matthew.d.roper@intel.com

drm/i915/dg2: Add Wa_16011777198

Coarse power gating for render should not be enabled on some DG2
steppings.

Bspec: 52698
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116174818.2128062-4-matthew.d.roper@intel.com

drm: Return error codes from struct drm_driver.gem_create_object

GEM helper libraries use struct drm_driver.gem_create_object to let
drivers override GEM object allocation. On failure, the call returns
NULL.

Change the semantics to make the calls return a pointer-encoded error.
This aligns the callback with its callers. Fixes the ingenic driver,
which already returns an error pointer.

Also update the callers to handle the involved types more strictly.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Steven Price <steven.price@arm.com>
Acked-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20211130095255.26710-1-tzimmermann@suse.de

Merge tag 'drm-intel-next-2021-11-30' of git://anongit.freedesktop.org/drm/drm-intel into drm-next

drm/i915 feature pull for v5.17:

Features and functionality:
- Implement per-lane DP drive settings for ICL+ (Ville)
- Enable runtime pm autosuspend by default (Tilak Tangudu)
- ADL-P DSI support (Vandita)
- Add support for pipe C and D DMC firmware (Anusha)
- Implement (near)atomic gamma LUT updates via vblank workers (Ville)
- Split plane updates to noarm+arm phases (Ville)
- Remove the CCS FB stride restrictions on ADL-P (Imre)
- Add PSR selective fetch support for biplanar formats (Jouni)
- Add support for display audio codec keepalive (Kai)
- VRR platform support for display 11 (Manasi)

Refactoring and cleanups:
- FBC refactoring and cleanups preparing for multiple FBC instances (Ville)
- PCH modeset refactoring, move to its own file (Ville)
- Refactor and simplify handling of modifiers (Imre)
- PXP cleanups (Ville)
- Display header and include refactoring (Jani)
- Some register macro cleanups (Ville)
- Refactor DP HDMI DFP limit code (Ville)

Fixes:
- Disable DSB usage for now due to incorrect gamma LUT updates (Ville)
- Check async flip state of every crtc and plane only once (José)
- Fix DPT FB suspend/resume (Imre)
- Fix black screen on reboot due to disabled DP++ TMDS output buffers (Ville)
- Don't request GMBUS to generate irqs when called while irqs are off (Ville)
- Fix type1 DVI DP dual mode adapter heuristics for modern platforms (Ville)
- Fix fix integer overflow in 128b/132b data rate calculation (Jani)
- Fix bigjoiner state readout (Ville)
- Build fix for non-x86 (Siva)
- PSR fixes (José, Jouni, Ville)
- Disable ADL-P underrun recovery (José)
- Fix DP link parameter usage before valid DPCD (Imre)
- VRR vblank and frame counter fixes (Ville)
- Fix fastsets on TypeC ports following a non-blocking modeset (Imre)
- Compiler warning fixes (Nathan Chancellor)
- Fix DSI HS mode commands (William Tseng)
- Error return fixes (Dan Carpenter)
- Update memory bandwidth calculations (Radhakrishna)
- Implement WM0 cursor WA for DG2 (Stan)
- Fix DSI Double pixelclock on read-back for dual-link panels (Hans de Goede)
- HDMI 2.1 PCON FRL configuration fixes (Ankit)

Merges:
- DP link training delay helpers, via topic branch (Jani)
- Backmerge drm-next (Jani)

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/87v909it0t.fsf@intel.com

drm/amdgpu/display: Only set vblank_disable_immediate when PSR is not enabled

[Why]
PSR currently relies on the kernel's delayed vblank on/off mechanism
as an implicit bufferring mechanism to prevent excessive entry/exit.

Without this delay the user experience is impacted since it can take
a few frames to enter/exit.

[How]
Only allow vblank disable immediate for DC when psr is not supported.

Leave a TODO indicating that this support should be extended in the
future to delay independent of the vblank interrupt.

Fixes: 92020e81ddbeac ("drm/amdgpu/display: set vblank_disable_immediate for DC")
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/i915: Add workaround numbers to GEN7_COMMON_SLICE_CHICKEN1 whitelisting

Those two workarounds needs to be implemented in UMD, KMD only needs
to whitelist the registers, so here only adding the workaround number
to facilitate future workaroud table checks.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Matt Atwood <matthew.s.atwood@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211119140931.32791-2-jose.souza@intel.com

Revert "drm/i915: Implement Wa_1508744258"

This workarounds are causing hangs, because I missed the fact that it
needs to be enabled for all cases and disabled when doing a resolve
pass.

So KMD only needs to whitelist it and UMD will be the one setting it
on per case.

This reverts commit 28ec02c9cbebf3feeaf21a59df9dfbc02bda3362.

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/4145
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Fixes: 28ec02c9cbeb ("drm/i915: Implement Wa_1508744258")
Reviewed-by: Matt Atwood <matthew.s.atwood@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211119140931.32791-1-jose.souza@intel.com

drm/ttm: Clarify that the TTM_PL_SYSTEM is under TTMs control

TTM takes full control over TTM_PL_SYSTEM placed buffers. This makes
driver internal usage of TTM_PL_SYSTEM prone to errors because it
requires the drivers to manually handle all interactions between TTM
which can swap out those buffers whenever it thinks it's the right
thing to do and driver.

CPU buffers which need to be fenced and shared with accelerators should
be placed in driver specific placements that can explicitly handle
CPU/accelerator buffer fencing.
Currently, apart, from things silently failing nothing is enforcing
that requirement which means that it's easy for drivers and new
developers to get this wrong. To avoid the confusion we can document
this requirement and clarify the solution.

This came up during a discussion on dri-devel:
https://lore.kernel.org/dri-devel/232f45e9-8748-1243-09bf-56763e6668b3@amd.com

Signed-off-by: Zack Rusin <zackr@vmware.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211110145034.487512-1-zackr@vmware.com

drm/vmwgfx: Switch the internal BO's to ttm_bo_type_kernel

There's never a need to access our internal kernel bo's from
user-space. Those objects are used exclusively for internal
support to guest backed surfaces (in otable setup and mob
page tables) and there's no need to have them be of device
type, i.e. mmappable from user-space.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-6-zackr@vmware.com

drm/vmwgfx: Introduce a new placement for MOB page tables

For larger (bigger than a page) and noncontiguous mobs we have
to create page tables that allow the host to find the memory.
Those page tables just used regular system memory. Unfortunately
in TTM those BO's are not allowed to be busy thus can't be
fenced and we have to fence those bo's  because we don't want
to destroy the page tables while the host is still executing
the command buffers which might be accessing them.

To solve it we introduce a new placement VMW_PL_SYSTEM which
is very similar to TTM_PL_SYSTEM except that it allows
fencing. This fixes kernel oops'es during unloading of the driver
(and pci hot remove/add) which were caused by busy BO's in
TTM_PL_SYSTEM being present in the delayed deletion list in
TTM (TTM_PL_SYSTEM manager is destroyed before the delayed
deletions are executed)

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-5-zackr@vmware.com

drm/vmwgfx: Fail to initialize on broken configs

Some of our hosts have a bug where rescaning a pci bus results in stale
fifo memory being mapped on the host. This makes any fifo communication
impossible resulting in various kernel crashes.

Instead of unexpectedly crashing, predictably fail to load the driver
which will preserve the system.

Fixes: fb1d9738ca05 ("drm/vmwgfx: Add DRM driver for VMware Virtual GPU")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-4-zackr@vmware.com

drm/vmwgfx: Release ttm memory if probe fails

The ttm mem global state was leaking if the vmwgfx driver load failed.

In case of a driver load failure we have to make sure we also release
the ttm mem global state.

Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-3-zackr@vmware.com

drm/vmwgfx: Remove the deprecated lower mem limit

TTM during the transition to the new page allocator lost the ability
to constrain the allocations via the lower_mem_limit. The code has
been unused since the change:
256dd44bd897 ("drm/ttm: nuke old page allocator")
and there's no reason to keep it.

Fixes: 256dd44bd897 ("drm/ttm: nuke old page allocator")
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Martin Krastev <krastevm@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211105193845.258816-2-zackr@vmware.com

drm/i915: Update error capture code to avoid using the current vma state

With asynchronous migrations, the vma state may be several migrations
ahead of the state that matches the request we're capturing.
Address that by introducing an i915_vma_snapshot structure that
can be used to snapshot relevant state at request submission.
In order to make sure we access the correct memory, the snapshots take
references on relevant sg-tables and memory regions.

Also move the capture list allocation out of the fence signaling
critical path and use the CONFIG_DRM_I915_CAPTURE_ERROR define to
avoid compiling in members and functions used for error capture
when they're not used.

Finally, Introduce lockdep annotation.

v4:
- Break out the capture allocation mode change to a separate patch.
v5:
- Fix compilation error in the !CONFIG_DRM_I915_CAPTURE_ERROR case
  (kernel test robot)
v6:
- Use #if IS_ENABLED() instead of #ifdef to match driver style.
- Move yet another change of allocation mode to the separate patch.
- Commit message rework due to patch reordering.
v7:
- Adjust for removal of region refcounting.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129202245.472043-1-thomas.hellstrom@linux.intel.com

drm/i915/gem: Fix a NULL pointer dereference in igt_request_rewind()

In igt_request_rewind(), mock_context(i915, "A") is assigned to ctx[0]
and used in i915_gem_context_get_engine(). There is a dereference
of ctx[0] in i915_gem_context_get_engine(), which could lead to a NULL
pointer dereference on failure of mock_context(i915, "A") .

So as mock_context(i915, "B").

Although this bug is not serious for it belongs to testing code, it is
better to be fixed to avoid unexpected failure in testing.

Fix this bugs by adding checks about ctx[0] and ctx[1].

This bug was found by a static analyzer. The analysis employs
differential checking to identify inconsistent security operations
(e.g., checks or kfrees) between two code paths and confirms that the
inconsistent operations are not recovered in the current function or
the callers, so they constitute bugs.

Note that, as a bug found by static analysis, it can be a false
positive or hard to trigger. Multiple researchers have cross-reviewed
the bug.

Builds with CONFIG_DRM_I915_SELFTEST=y show no new warnings,
and our static analyzer no longer warns about this code.

References: 591c0fb85d1c ("drm/i915: Exercise request cancellation using a mock selftest")
[tursulin: Replaced fixes with references to avoid.]
Signed-off-by: Zhou Qingyang <zhou1615@umn.edu>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211130141545.153899-1-zhou1615@umn.edu

drm/i915: Use per device iommu check

With both integrated and discrete Intel GPUs in a system, the current
global check of intel_iommu_gfx_mapped, as done from intel_vtd_active()
may not be completely accurate.

In this patch we add i915 parameter to intel_vtd_active() in order to
prepare it for multiple GPUs and we also change the check away from Intel
specific intel_iommu_gfx_mapped (global exported by the Intel IOMMU
driver) to probing the presence of IOMMU on a specific device using
device_iommu_mapped().

This will return true both for IOMMU pass-through and address translation
modes which matches the current behaviour. If in the future we wanted to
distinguish between these two modes we could either use
iommu_get_domain_for_dev() and check for __IOMMU_DOMAIN_PAGING bit
indicating address translation, or ask for a new API to be exported from
the IOMMU core code.

v2:
  * Check for dmar translation specifically, not just iommu domain. (Baolu)

v3:
 * Go back to plain "any domain" check for now, rewrite commit message.

v4:
 * Use device_iommu_mapped. (Robin, Baolu)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211126141424.493753-1-tvrtko.ursulin@linux.intel.com

drm/i915: Drop stealing of bits from i915_sw_fence function pointer

Rather than stealing bits from i915_sw_fence function pointer use
separate fields for function pointer and flags. If using two different
fields, the 4 byte alignment for the i915_sw_fence function pointer can
also be dropped.

v2:
 (CI)
  - Set new function field rather than flags in __i915_sw_fence_init
v3:
 (Tvrtko)
  - Remove BUG_ON(!fence->flags) in reinit as that will now blow up
  - Only define fence->flags if CONFIG_DRM_I915_SW_FENCE_CHECK_DAG is
    defined
v4:
  - Rebase, resend for CI

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Alan Previn <alan.previn.teres.alexis@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211116194929.10211-1-matthew.brost@intel.com

drm/i915/pmu: Fix synchronization of PMU callback with reset

Since the PMU callback runs in irq context, it synchronizes with gt
reset using the reset count. We could run into a case where the PMU
callback could read the reset count before it is updated. This has a
potential of corrupting the busyness stats.

In addition to the reset count, check if the reset bit is set before
capturing busyness.

In addition save the previous stats only if you intend to update them.

v2:
- The 2 reset counts captured in the PMU callback can end up being the
  same if they were captured right after the count is incremented in the
  reset flow. This can lead to a bad busyness state. Ensure that reset
  is not in progress when the initial reset count is captured.

Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211108211057.68783-1-umesh.nerlige.ramappa@intel.com

dma-buf: make fence mandatory for dma_resv_add_excl_fence v2

Calling dma_resv_add_excl_fence() with the fence as NULL and expecting
that that this frees up the fences is simply abuse of the internals of
the dma_resv object.

v2: drop the fence pruning completely.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129120659.1815-4-christian.koenig@amd.com

drm/ttm: stop pruning fences after wait

This is just abusing internals of the dma_resv object.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129120659.1815-3-christian.koenig@amd.com

drm/i915: Remove dma_resv_prune

The signaled bit is already used for quick testing if a fence is signaled.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/460722/
Signed-off-by: Christian König <christian.koenig@amd.com>

drm/qxl: use iterator instead of dma_resv_shared_list

I'm not sure why it is useful to know the number of fences
in the reservation object, but we try to avoid exposing the
dma_resv_shared_list() function.

So use the iterator instead. If more information is desired
we could use dma_resv_describe() as well.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129120659.1815-5-christian.koenig@amd.com

lontium-lt9611: check a different register bit for HDMI sensing

It has been observed that with certain monitors such as the HP Z27n,
the register 0x825e reads a value of 0x79 when the HDMI cable is
connected and 0x78 when it is disconnected, i.e. bit 0 appears
to correspond to the HDMI connection status and bit 2 is never
set. Therefore, change the driver to check bit 0 instead of bit 2.

Signed-off-by: Peter Collingbourne <pcc@google.com>
Link: https://linux-review.googlesource.com/id/I7e76411127e1ce4988a3f6d0c8ba5f1c3d880c23
Reviewed-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Robert Foss <robert.foss@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211117020724.2647769-1-pcc@google.com

drm: Remove CONFIG_DRM_KMS_CMA_HELPER option

Link drm_fb_cma_helper.o into drm_cma_helper.ko if CONFIG_DRM_KMS_HELPER
has been set. Remove CONFIG_DRM_KMS_CMA_HELPER config option. Selecting
KMS helpers and CMA will now automatically enable CMA KMS helpers.

Some drivers' Kconfig files did not correctly select KMS or CMA helpers.
Fix this as part of the change.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211106193509.17472-3-tzimmermann@suse.de

drm/mipi-dbi: Remove dependency on GEM CMA helper library

The MIPI DBI helpers access struct drm_gem_cma_object.vaddr in a
few places. Replace all instances with the correct generic GEM
functions. Use drm_gem_fb_vmap() for mapping a framebuffer's GEM
objects and drm_gem_fb_vunmap() for unmapping them. This removes
the dependency on CMA helpers within MIPI DBI.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211106193509.17472-2-tzimmermann@suse.de

drm/cma-helper: Pass GEM CMA object in public interfaces

Change all GEM CMA object functions that receive a GEM object
of type struct drm_gem_object to expect an object of type
struct drm_gem_cma_object instead.

This change reduces the number of upcasts from struct drm_gem_object
by moving them into callers. The C compiler can now verify that the
GEM CMA functions are called with the correct type.

For consistency, the patch also renames drm_gem_cma_free_object to
drm_gem_cma_free. It further updates documentation for a number of
functions.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115120148.21766-4-tzimmermann@suse.de

drm/cma-helper: Export dedicated wrappers for GEM object functions

Wrap GEM CMA functions for struct drm_gem_object_funcs and update
all callers. This will allow for an update of the public interfaces
of the GEM CMA helper library.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115120148.21766-3-tzimmermann@suse.de

drm/cma-helper: Move driver and file ops to the end of header

Restructure the header file for CMA helpers by moving declarations
for driver and file operations to the end of the file. No functional
changes.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Maxime Ripard <maxime@cerno.tech>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211115120148.21766-2-tzimmermann@suse.de

drm: Declare hashtable as legacy

The DRM hashtable code is only used by internal functions for legacy
UMS drivers. Move the implementation behind CONFIG_DRM_LEGACY and the
declarations into legacy header files. Unexport the symbols.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129094841.22499-4-tzimmermann@suse.de

drm/vmwgfx: Copy DRM hash-table code into driver

Besides some legacy code, vmwgfx is the only user of DRM's hash-
table implementation. Copy the code into the driver, so that the
core code can be retired.

No functional changes. However, the real solution for vmwgfx is to
use Linux' generic hash-table functions.

v2:
* add TODO item for updating vmwgfx (Sam)

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129094841.22499-3-tzimmermann@suse.de

drm/ttm: Don't include drm_hashtab.h

Remove the include statement for drm_hashtab.h. It's not required
by TTM.

Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Christian König <christian.koenig@amd.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129094841.22499-2-tzimmermann@suse.de

dma_fence_array: Fix PENDING_ERROR leak in dma_fence_array_signaled()

If a dma_fence_array is reported signaled by a call to
dma_fence_is_signaled(), it may leak the PENDING_ERROR status.

Fix this by clearing the PENDING_ERROR status if we return true in
dma_fence_array_signaled().

v2:
- Update Cc list, and add R-b.

Fixes: 1f70b8b812f3 ("dma-fence: Propagate errors to dma-fence-array container")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: Gustavo Padovan <gustavo@padovan.org>
Cc: Christian König <christian.koenig@amd.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: linux-media@vger.kernel.org
Cc: dri-devel@lists.freedesktop.org
Cc: linaro-mm-sig@lists.linaro.org
Cc: <stable@vger.kernel.org> # v5.4+
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129152727.448908-1-thomas.hellstrom@linux.intel.com

drm/i915: Fix DPT suspend/resume on !HAS_DISPLAY platforms

The drm.mode_config state is not initialized in case of !HAS_DISPLAY
so taking the fb_lock and iterating the fb list won't work on those
platforms. Skip the suspend/resume with an explicit check for this.

Fixes: 9755f055f512 ("drm/i915: Restore memory mapping for DPT FBs across system suspend/resume")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211125171603.1775179-1-imre.deak@intel.com

drm/tidss: Fix warning: unused variable 'tidss_pm_ops'

Mark 'tidss_pm_ops' as __maybe_unused to avoid
the warning: unused variable 'tidss_pm_ops'

Fixes: 6e120594631f ("drm/tidss: Make use of the helper macro SET_RUNTIME_PM_OPS()")
Signed-off-by: Cai Huoqing <caihuoqing@baidu.com>
Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211129063347.404-1-caihuoqing@baidu.com

drm: Fix build error caused by missing drm_nomodeset.o

The patch for commit 6a2d2ddf2c34 ("drm: Move nomodeset kernel parameter
to the DRM subsystem") was generated with config 'diff.noprefix true'.

But later was applied using 'cat nomodeset.mbox | dim apply-branch' on a
machine with 'diff.noprefix false'. And command 'git am --scissors -3' as
used by the dim tool doesn't handle that case well, since the 3-way merge
wrongly resolves the path for new file drivers/gpu/drm/drm_nomodeset.c as
gpu/drm/drm_nomodeset.c instead.

It led to the following build error as reported by the kernel test robot:

  make[4]: *** No rule to make target 'drivers/gpu/drm/drm_nomodeset.o', needed by 'drivers/gpu/drm/built-in.a'.

Fixes: 6a2d2ddf2c34 ("drm: Move nomodeset kernel parameter to the DRM subsystem")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20211127191910.709356-1-javierm@redhat.com

drm: Make the nomodeset message less sensational

The message printed when nomodeset is present in the kernel command line
makes it look as if the parameter must never be used and it's a bad idea.

But there are valid reasons to use this parameter and the message should
not imply otherwise. Change the text to be more accurate and restrained.

Suggested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-7-javierm@redhat.com

Documentation/admin-guide: Document nomodeset kernel parameter

The nomodeset kernel command line parameter is not documented. Its name
is quite vague and is not intuitive what's the behaviour when it is set.

Document in kernel-parameters.txt what actually happens when nomodeset
is used. That way, users could know if they want to enable this option.

Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-6-javierm@redhat.com

drm: Decouple nomodeset from CONFIG_VGA_CONSOLE

This relationship was only for historical reasons and the nomodeset option
should be available even on platforms that don't enable CONFIG_VGA_CONSOLE.

Suggested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-5-javierm@redhat.com

drm: Move nomodeset kernel parameter to the DRM subsystem

The "nomodeset" kernel cmdline parameter is handled by the vgacon driver
but the exported vgacon_text_force() symbol is only used by DRM drivers.

It makes much more sense for the parameter logic to be in the subsystem
of the drivers that are making use of it.

Let's move the vgacon_text_force() function and related logic to the DRM
subsystem. While doing that, rename it to drm_firmware_drivers_only() and
make it return true if "nomodeset" was used and false otherwise. This is
a better description of the condition that the drivers are testing for.

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-4-javierm@redhat.com

drm/vboxvideo: Drop CONFIG_VGA_CONSOLE guard to call vgacon_text_force()

It is already handled by the console.h macro since a stub inline function
is defined for vgacon_text_force() if CONFIG_VGA_CONSOLE is not set.

There's no need to have ifdefery in the driver when calling the function.

Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-3-javierm@redhat.com

drm: Don't print messages if drivers are disabled due nomodeset

The nomodeset kernel parameter handler already prints a message that the
DRM drivers will be disabled, so there's no need for drivers to do that.

Suggested-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211112133230.1595307-2-javierm@redhat.com

drm: get rid of DRM_DEBUG_* log calls in drm core, files drm_a*.c

DRM_DEBUG_* and DRM_* log calls are deprecated.
Change them to drm_dbg_* / drm_{err,info,...} calls in drm core
files.

To avoid making a very big patch, this change is split in
smaller patches. This one includes drm_a*.c

Signed-off-by: Claudio Suarez <cssk@net-c.es>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/YaC7zXW119tlzfVh@gineta.localdomain

drm/i915/gemfs: don't mark huge_opt as static

vfs_kernel_mount() modifies the passed in mount options, leaving us with
"huge", instead of "huge=within_size". Normally this shouldn't matter
with the usual module load/unload flow, however with the core_hotunplug
IGT we are hitting the following, when re-probing the memory regions:

i915 0000:00:02.0: [drm] Transparent Hugepage mode 'huge'
tmpfs: Bad value for 'huge'
[drm] Unable to create a private tmpfs mount, hugepage support will be disabled(-22).

References: https://gitlab.freedesktop.org/drm/intel/-/issues/4651
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211126110843.2028582-1-matthew.auld@intel.com

drm/i915: Use __GFP_KSWAPD_RECLAIM in the capture code

The capture code is typically run entirely in the fence signalling
critical path. We're about to add lockdep annotation in an upcoming patch
which reveals a lockdep splat similar to the below one.

Fix the associated potential deadlocks using __GFP_KSWAPD_RECLAIM
(which is the same as GFP_WAIT, but open-coded for clarity) rather than
GFP_KERNEL for memory allocation in the capture path. This has the
potential drawback that capture might fail in situations with memory
pressure.

[  234.842048] WARNING: possible circular locking dependency detected
[  234.842050] 5.15.0-rc7+ #20 Tainted: G     U  W
[  234.842052] ------------------------------------------------------
[  234.842054] gem_exec_captur/1180 is trying to acquire lock:
[  234.842056] ffffffffa3e51c00 (fs_reclaim){+.+.}-{0:0}, at: __kmalloc+0x4d/0x330
[  234.842063]
               but task is already holding lock:
[  234.842064] ffffffffa3f57620 (dma_fence_map){++++}-{0:0}, at: i915_vma_snapshot_resource_pin+0x27/0x30 [i915]
[  234.842138]
               which lock already depends on the new lock.

[  234.842140]
               the existing dependency chain (in reverse order) is:
[  234.842142]
               -> #2 (dma_fence_map){++++}-{0:0}:
[  234.842145]        __dma_fence_might_wait+0x41/0xa0
[  234.842149]        dma_resv_lockdep+0x1dc/0x28f
[  234.842151]        do_one_initcall+0x58/0x2d0
[  234.842154]        kernel_init_freeable+0x273/0x2bf
[  234.842157]        kernel_init+0x16/0x120
[  234.842160]        ret_from_fork+0x1f/0x30
[  234.842163]
               -> #1 (mmu_notifier_invalidate_range_start){+.+.}-{0:0}:
[  234.842166]        fs_reclaim_acquire+0x6d/0xd0
[  234.842168]        __kmalloc_node+0x51/0x3a0
[  234.842171]        alloc_cpumask_var_node+0x1b/0x30
[  234.842174]        native_smp_prepare_cpus+0xc7/0x292
[  234.842177]        kernel_init_freeable+0x160/0x2bf
[  234.842179]        kernel_init+0x16/0x120
[  234.842181]        ret_from_fork+0x1f/0x30
[  234.842184]
               -> #0 (fs_reclaim){+.+.}-{0:0}:
[  234.842186]        __lock_acquire+0x1161/0x1dc0
[  234.842189]        lock_acquire+0xb5/0x2b0
[  234.842192]        fs_reclaim_acquire+0xa1/0xd0
[  234.842193]        __kmalloc+0x4d/0x330
[  234.842196]        i915_vma_coredump_create+0x78/0x5b0 [i915]
[  234.842253]        intel_engine_coredump_add_vma+0x36/0xe0 [i915]
[  234.842307]        __i915_gpu_coredump+0x290/0x5e0 [i915]
[  234.842365]        i915_capture_error_state+0x57/0xa0 [i915]
[  234.842415]        intel_gt_handle_error+0x348/0x3e0 [i915]
[  234.842462]        intel_gt_debugfs_reset_store+0x3c/0x90 [i915]
[  234.842504]        simple_attr_write+0xc1/0xe0
[  234.842507]        full_proxy_write+0x53/0x80
[  234.842509]        vfs_write+0xbc/0x350
[  234.842513]        ksys_write+0x58/0xd0
[  234.842514]        do_syscall_64+0x38/0x90
[  234.842516]        entry_SYSCALL_64_after_hwframe+0x44/0xae
[  234.842519]
               other info that might help us debug this:

[  234.842521] Chain exists of:
                 fs_reclaim --> mmu_notifier_invalidate_range_start --> dma_fence_map

[  234.842526]  Possible unsafe locking scenario:

[  234.842528]        CPU0                    CPU1
[  234.842529]        ----                    ----
[  234.842531]   lock(dma_fence_map);
[  234.842532]                                lock(mmu_notifier_invalidate_range_start);
[  234.842535]                                lock(dma_fence_map);
[  234.842537]   lock(fs_reclaim);
[  234.842539]
                *** DEADLOCK ***

[  234.842540] 4 locks held by gem_exec_captur/1180:
[  234.842543]  #0: ffff9007812d9460 (sb_writers#17){.+.+}-{0:0}, at: ksys_write+0x58/0xd0
[  234.842547]  #1: ffff900781d9ecb8 (&attr->mutex){+.+.}-{3:3}, at: simple_attr_write+0x3a/0xe0
[  234.842552]  #2: ffffffffc11913a8 (capture_mutex){+.+.}-{3:3}, at: i915_capture_error_state+0x1a/0xa0 [i915]
[  234.842602]  #3: ffffffffa3f57620 (dma_fence_map){++++}-{0:0}, at: i915_vma_snapshot_resource_pin+0x27/0x30 [i915]
[  234.842656]
               stack backtrace:
[  234.842658] CPU: 0 PID: 1180 Comm: gem_exec_captur Tainted: G     U  W         5.15.0-rc7+ #20
[  234.842661] Hardware name: ASUS System Product Name/PRIME B560M-A AC, BIOS 0403 01/26/2021
[  234.842664] Call Trace:
[  234.842666]  dump_stack_lvl+0x57/0x72
[  234.842669]  check_noncircular+0xde/0x100
[  234.842672]  ? __lock_acquire+0x3bf/0x1dc0
[  234.842675]  __lock_acquire+0x1161/0x1dc0
[  234.842678]  lock_acquire+0xb5/0x2b0
[  234.842680]  ? __kmalloc+0x4d/0x330
[  234.842683]  ? finish_task_switch.isra.0+0xf2/0x360
[  234.842686]  ? i915_vma_coredump_create+0x78/0x5b0 [i915]
[  234.842734]  fs_reclaim_acquire+0xa1/0xd0
[  234.842737]  ? __kmalloc+0x4d/0x330
[  234.842739]  __kmalloc+0x4d/0x330
[  234.842742]  i915_vma_coredump_create+0x78/0x5b0 [i915]
[  234.842793]  ? capture_vma+0xbe/0x110 [i915]
[  234.842844]  intel_engine_coredump_add_vma+0x36/0xe0 [i915]
[  234.842892]  __i915_gpu_coredump+0x290/0x5e0 [i915]
[  234.842939]  i915_capture_error_state+0x57/0xa0 [i915]
[  234.842985]  intel_gt_handle_error+0x348/0x3e0 [i915]
[  234.843032]  ? __mutex_lock+0x81/0x830
[  234.843035]  ? simple_attr_write+0x3a/0xe0
[  234.843038]  ? __lock_acquire+0x3bf/0x1dc0
[  234.843041]  intel_gt_debugfs_reset_store+0x3c/0x90 [i915]
[  234.843083]  ? _copy_from_user+0x45/0x80
[  234.843086]  simple_attr_write+0xc1/0xe0
[  234.843089]  full_proxy_write+0x53/0x80
[  234.843091]  vfs_write+0xbc/0x350
[  234.843094]  ksys_write+0x58/0xd0
[  234.843096]  do_syscall_64+0x38/0x90
[  234.843098]  entry_SYSCALL_64_after_hwframe+0x44/0xae
[  234.843101] RIP: 0033:0x7fa467480877
[  234.843103] Code: 75 05 48 83 c4 58 c3 e8 37 4e ff ff 0f 1f 80 00 00 00 00 f3 0f 1e fa 64 8b 04 25 18 00 00 00 85 c0 75 10 b8 01 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 51 c3 48 83 ec 28 48 89 54 24 18 48 89 74 24
[  234.843108] RSP: 002b:00007ffd14d79b08 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
[  234.843112] RAX: ffffffffffffffda RBX: 00007ffd14d79b60 RCX: 00007fa467480877
[  234.843114] RDX: 0000000000000014 RSI: 00007ffd14d79b60 RDI: 0000000000000007
[  234.843116] RBP: 0000000000000007 R08: 0000000000000000 R09: 00007ffd14d79ab0
[  234.843119] R10: ffffffffffffffff R11: 0000000000000246 R12: 0000000000000014
[  234.843121] R13: 0000000000000000 R14: 00007ffd14d79b60 R15: 0000000000000005

v5:
- Use __GFP_KSWAPD_RECLAIM rather than __GFP_NOWAIT for clarity.
  (Daniel Vetter)
v6:
- Include an instance in execlists_capture_work().
- Rework the commit message due to patch reordering.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211108174547.979714-3-thomas.hellstrom@linux.intel.com

drm/i915: Avoid allocating a page array for the gpu coredump

The gpu coredump typically takes place in a dma_fence signalling
critical path, and hence can't use GFP_KERNEL allocations, as that
means we might hit deadlocks under memory pressure. However
changing to __GFP_KSWAPD_RECLAIM which will be done in an upcoming
patch will instead mean a lower chance of the allocation succeeding.
In particular large contigous allocations like the coredump page
vector.
Remove the page vector in favor of a linked list of single pages.
Use the page lru list head as the list link, as the page owner is
allowed to do that.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Ramalingam C <ramalingam.c@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211108174547.979714-2-thomas.hellstrom@linux.intel.com

drm/i915: Remove dma_resv_prune

The signaled bit is already used for quick testing if a fence is signaled.
On top of that, it's a terrible abuse of dma-fence api, and in the common
case where the object is already locked by the caller, the trylock will fail.

If it were useful, the core dma-api would have exposed the same functionality.

The fact that i915 has a dma_resv_utils.c file should be a warning that the
functionality either belongs in core, or is not very useful at all.
In this case the latter.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
[mlankhorst: Improve commit message]
Acked-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211021103605.735002-3-maarten.lankhorst@linux.intel.com
Reviewed-by: Matthew Auld <matthew.auld@intel.com> #irc

Merge drm/drm-next into drm-intel-gt-next

Maarten requested a backmerge due his work depending on subtle semantic
changes introduced by:

  7e2e69ed4678 ("drm/i915: Fix i915_request fence wait semantics")
  2cbb8d4d6770 ("drm/i915: use new iterator in i915_gem_object_wait_reservation")

Both should probably have been merged to drm-intel-gt-next anyway.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/bridge: megachips: Ensure both bridges are probed before registration

In the configuration used by the b850v3, the STDP2690 is used to read EDID
data whilst it's the STDP4028 which can detect when monitors are connected.

This can result in problems at boot with monitors connected when the
STDP4028 is probed first, a monitor is detected and an attempt is made to
read the EDID data before the STDP2690 has probed:

[    3.795721] Unable to handle kernel NULL pointer dereference at virtual address 00000018
[    3.803845] pgd = (ptrval)
[    3.806581] [00000018] *pgd=00000000
[    3.810180] Internal error: Oops: 5 [#1] SMP ARM
[    3.814813] Modules linked in:
[    3.817879] CPU: 0 PID: 64 Comm: kworker/u4:1 Not tainted 5.15.0 #1
[    3.824161] Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
[    3.830705] Workqueue: events_unbound deferred_probe_work_func
[    3.836565] PC is at stdp2690_get_edid+0x44/0x19c
[    3.841286] LR is at ge_b850v3_lvds_get_modes+0x2c/0x5c
[    3.846526] pc : [<805eae10>]    lr : [<805eb138>]    psr: 80000013
[    3.852802] sp : 81c359d0  ip : 7dbb550b  fp : 81c35a1c
[    3.858037] r10: 81c73840  r9 : 81c73894  r8 : 816d9800
[    3.863270] r7 : 00000000  r6 : 81c34000  r5 : 00000000  r4 : 810c35f0
[    3.869808] r3 : 80e3e294  r2 : 00000080  r1 : 00000cc0  r0 : 81401180
[    3.876349] Flags: Nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment none
[    3.883499] Control: 10c5387d  Table: 1000404a  DAC: 00000051
[    3.889254] Register r0 information: slab kmem_cache start 81401180 pointer offset 0
[    3.897034] Register r1 information: non-paged memory
[    3.902097] Register r2 information: non-paged memory
[    3.907160] Register r3 information: non-slab/vmalloc memory
[    3.912832] Register r4 information: non-slab/vmalloc memory
[    3.918503] Register r5 information: NULL pointer
[    3.923217] Register r6 information: non-slab/vmalloc memory
[    3.928887] Register r7 information: NULL pointer
[    3.933601] Register r8 information: slab kmalloc-1k start 816d9800 pointer offset 0 size 1024
[    3.942244] Register r9 information: slab kmalloc-2k start 81c73800 pointer offset 148 size 2048
[    3.951058] Register r10 information: slab kmalloc-2k start 81c73800 pointer offset 64 size 2048
[    3.959873] Register r11 information: non-slab/vmalloc memory
[    3.965632] Register r12 information: non-paged memory
[    3.970781] Process kworker/u4:1 (pid: 64, stack limit = 0x(ptrval))
[    3.977148] Stack: (0x81c359d0 to 0x81c36000)
[    3.981517] 59c0:                                     80b2b668 80b2b5bc 000002e2 0000034e
[    3.989712] 59e0: 81c35a8c 816d98e8 81c35a14 7dbb550b 805bfcd0 810c35f0 81c73840 824addc0
[    3.997906] 5a00: 00001000 816d9800 81c73894 81c73840 81c35a34 81c35a20 805eb138 805eadd8
[    4.006099] 5a20: 810c35f0 00000045 81c35adc 81c35a38 80594188 805eb118 80d7c788 80dd1848
[    4.014292] 5a40: 00000000 81c35a50 80dca950 811194d3 80dca7c4 80dca944 80dca91c 816d9800
[    4.022485] 5a60: 81c34000 81c760a8 816d9800 80c58c98 810c35f0 816d98e8 00001000 00001000
[    4.030678] 5a80: 00000000 00000000 8017712c 81c60000 00000002 00000001 00000000 00000000
[    4.038870] 5aa0: 816d9900 816d9900 00000000 7dbb550b 805c700c 00000008 826282c8 826282c8
[    4.047062] 5ac0: 00001000 81e1ce40 00001000 00000002 81c35bf4 81c35ae0 805d9694 80593fc0
[    4.055255] 5ae0: 8017a970 80179ad8 00000179 00000000 81c35bcc 81c35b00 80177108 8017a950
[    4.063447] 5b00: 00000000 81c35b10 81c34000 00000000 81004fd8 81010a38 00000000 00000059
[    4.071639] 5b20: 816d98d4 81fbb718 00000013 826282c8 8017a940 81c35b40 81134448 00000400
[    4.079831] 5b40: 00000178 00000000 e063b9c1 00000000 c2000049 00000040 00000000 00000008
[    4.088024] 5b60: 82628300 82628380 00000000 00000000 81c34000 00000000 81fbb700 82628340
[    4.096216] 5b80: 826283c0 00001000 00000000 00000010 816d9800 826282c0 801766f8 00000000
[    4.104408] 5ba0: 00000000 81004fd8 00000049 00000000 00000000 00000001 80dcf940 80178de4
[    4.112601] 5bc0: 81c35c0c 7dbb550b 80178de4 81fbb700 00000010 00000010 810c35f4 81e1ce40
[    4.120793] 5be0: 81c40908 0000000c 81c35c64 81c35bf8 805a7f18 805d94a0 81c35c3c 816d9800
[    4.128985] 5c00: 00000010 81c34000 81c35c2c 81c35c18 8012fce0 805be90c 81c35c3c 81c35c28
[    4.137178] 5c20: 805be90c 80173210 81fbb600 81fbb6b4 81c35c5c 7dbb550b 81c35c64 81fbb700
[    4.145370] 5c40: 816d9800 00000010 810c35f4 81e1ce40 81c40908 0000000c 81c35c84 81c35c68
[    4.153565] 5c60: 805a8c78 805a7ed0 816d9800 81fbb700 00000010 00000000 81c35cac 81c35c88
[    4.161758] 5c80: 805a8dc4 805a8b68 816d9800 00000000 816d9800 00000000 8179f810 810c42d0
[    4.169950] 5ca0: 81c35ccc 81c35cb0 805e47b0 805a8d18 824aa240 81e1ea80 81c40908 81126b60
[    4.178144] 5cc0: 81c35d14 81c35cd0 8060db1c 805e46cc 81c35d14 81c35ce0 80dd90f8 810c4d58
[    4.186338] 5ce0: 80dd90dc 81fe9740 fffffffe 81fe9740 81e1ea80 00000000 810c4d6c 80c4b95c
[    4.194531] 5d00: 80dd9a3c 815c6810 81c35d34 81c35d18 8060dc9c 8060d8fc 8246b440 815c6800
[    4.202724] 5d20: 815c6810 eefd8e00 81c35d44 81c35d38 8060dd80 8060dbec 81c35d6c 81c35d48
[    4.210918] 5d40: 805e98a4 8060dd70 00000000 815c6810 810c45b0 81126e90 81126e90 80dd9a3c
[    4.219112] 5d60: 81c35d8c 81c35d70 80619574 805e9808 815c6810 00000000 810c45b0 81126e90
[    4.227305] 5d80: 81c35db4 81c35d90 806168dc 80619514 80625df0 80623c80 815c6810 810c45b0
[    4.235498] 5da0: 81c35e6c 815c6810 81c35dec 81c35db8 80616d04 80616800 81c35de4 81c35dc8
[    4.243691] 5dc0: 808382b0 80b2f444 8116e310 8116e314 81c35e6c 815c6810 00000003 80dd9a3c
[    4.251884] 5de0: 81c35e14 81c35df0 80616ec8 80616c60 00000001 810c45b0 81c35e6c 815c6810
[    4.260076] 5e00: 00000001 80dd9a3c 81c35e34 81c35e18 80617338 80616e90 00000000 81c35e6c
[    4.268269] 5e20: 80617284 81c34000 81c35e64 81c35e38 80614730 80617290 81c35e64 8171a06c
[    4.276461] 5e40: 81e220b8 7dbb550b 815c6810 81c34000 815c6854 81126e90 81c35e9c 81c35e68
[    4.284654] 5e60: 8061673c 806146a8 8060f5e0 815c6810 00000001 7dbb550b 00000000 810c5080
[    4.292847] 5e80: 810c5320 815c6810 81126e90 00000000 81c35eac 81c35ea0 80617554 80616650
[    4.301040] 5ea0: 81c35ecc 81c35eb0 80615694 80617544 810c5080 810c5080 810c5094 81126e90
[    4.309233] 5ec0: 81c35efc 81c35ed0 80615c6c 8061560c 80615bc0 810c50c0 817eeb00 81412800
[    4.317425] 5ee0: 814c3000 00000000 814c300d 81119a60 81c35f3c 81c35f00 80141488 80615bcc
[    4.325618] 5f00: 81c60000 81c34000 81c35f24 81c35f18 80143078 817eeb00 81412800 817eeb18
[    4.333811] 5f20: 81412818 81003d00 00000088 81412800 81c35f74 81c35f40 80141a48 80141298
[    4.342005] 5f40: 81c35f74 81c34000 801481ac 817efa40 817efc00 801417d8 817eeb00 00000000
[    4.350199] 5f60: 815a7e7c 81c34000 81c35fac 81c35f78 80149b1c 801417e4 817efc20 817efc20
[    4.358391] 5f80: ffffe000 817efa40 801499a8 00000000 00000000 00000000 00000000 00000000
[    4.366583] 5fa0: 00000000 81c35fb0 80100130 801499b4 00000000 00000000 00000000 00000000
[    4.374774] 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
[    4.382966] 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
[    4.391155] Backtrace:
[    4.393613] [<805eadcc>] (stdp2690_get_edid) from [<805eb138>] (ge_b850v3_lvds_get_modes+0x2c/0x5c)
[    4.402691]  r10:81c73840 r9:81c73894 r8:816d9800 r7:00001000 r6:824addc0 r5:81c73840
[    4.410534]  r4:810c35f0
[    4.413073] [<805eb10c>] (ge_b850v3_lvds_get_modes) from [<80594188>] (drm_helper_probe_single_connector_modes+0x1d4/0x84c)
[    4.424240]  r5:00000045 r4:810c35f0
[    4.427822] [<80593fb4>] (drm_helper_probe_single_connector_modes) from [<805d9694>] (drm_client_modeset_probe+0x200/0x1384)
[    4.439074]  r10:00000002 r9:00001000 r8:81e1ce40 r7:00001000 r6:826282c8 r5:826282c8
[    4.446917]  r4:00000008
[    4.449455] [<805d9494>] (drm_client_modeset_probe) from [<805a7f18>] (__drm_fb_helper_initial_config_and_unlock+0x54/0x5b4)
[    4.460713]  r10:0000000c r9:81c40908 r8:81e1ce40 r7:810c35f4 r6:00000010 r5:00000010
[    4.468556]  r4:81fbb700
[    4.471095] [<805a7ec4>] (__drm_fb_helper_initial_config_and_unlock) from [<805a8c78>] (drm_fbdev_client_hotplug+0x11c/0x1b0)
[    4.482434]  r10:0000000c r9:81c40908 r8:81e1ce40 r7:810c35f4 r6:00000010 r5:816d9800
[    4.490276]  r4:81fbb700
[    4.492814] [<805a8b5c>] (drm_fbdev_client_hotplug) from [<805a8dc4>] (drm_fbdev_generic_setup+0xb8/0x1a4)
[    4.502494]  r7:00000000 r6:00000010 r5:81fbb700 r4:816d9800
[    4.508160] [<805a8d0c>] (drm_fbdev_generic_setup) from [<805e47b0>] (imx_drm_bind+0xf0/0x130)
[    4.516805]  r7:810c42d0 r6:8179f810 r5:00000000 r4:816d9800
[    4.522474] [<805e46c0>] (imx_drm_bind) from [<8060db1c>] (try_to_bring_up_master+0x22c/0x2f0)
[    4.531116]  r7:81126b60 r6:81c40908 r5:81e1ea80 r4:824aa240
[    4.536783] [<8060d8f0>] (try_to_bring_up_master) from [<8060dc9c>] (__component_add+0xbc/0x184)
[    4.545597]  r10:815c6810 r9:80dd9a3c r8:80c4b95c r7:810c4d6c r6:00000000 r5:81e1ea80
[    4.553440]  r4:81fe9740
[    4.555980] [<8060dbe0>] (__component_add) from [<8060dd80>] (component_add+0x1c/0x20)
[    4.563921]  r7:eefd8e00 r6:815c6810 r5:815c6800 r4:8246b440
[    4.569589] [<8060dd64>] (component_add) from [<805e98a4>] (dw_hdmi_imx_probe+0xa8/0xe8)
[    4.577702] [<805e97fc>] (dw_hdmi_imx_probe) from [<80619574>] (platform_probe+0x6c/0xc8)
[    4.585908]  r9:80dd9a3c r8:81126e90 r7:81126e90 r6:810c45b0 r5:815c6810 r4:00000000
[    4.593662] [<80619508>] (platform_probe) from [<806168dc>] (really_probe+0xe8/0x460)
[    4.601524]  r7:81126e90 r6:810c45b0 r5:00000000 r4:815c6810
[    4.607191] [<806167f4>] (really_probe) from [<80616d04>] (__driver_probe_device+0xb0/0x230)
[    4.615658]  r7:815c6810 r6:81c35e6c r5:810c45b0 r4:815c6810
[    4.621326] [<80616c54>] (__driver_probe_device) from [<80616ec8>] (driver_probe_device+0x44/0xe0)
[    4.630313]  r9:80dd9a3c r8:00000003 r7:815c6810 r6:81c35e6c r5:8116e314 r4:8116e310
[    4.638068] [<80616e84>] (driver_probe_device) from [<80617338>] (__device_attach_driver+0xb4/0x12c)
[    4.647227]  r9:80dd9a3c r8:00000001 r7:815c6810 r6:81c35e6c r5:810c45b0 r4:00000001
[    4.654981] [<80617284>] (__device_attach_driver) from [<80614730>] (bus_for_each_drv+0x94/0xd8)
[    4.663794]  r7:81c34000 r6:80617284 r5:81c35e6c r4:00000000
[    4.669461] [<8061469c>] (bus_for_each_drv) from [<8061673c>] (__device_attach+0xf8/0x190)
[    4.677753]  r7:81126e90 r6:815c6854 r5:81c34000 r4:815c6810
[    4.683419] [<80616644>] (__device_attach) from [<80617554>] (device_initial_probe+0x1c/0x20)
[    4.691971]  r8:00000000 r7:81126e90 r6:815c6810 r5:810c5320 r4:810c5080
[    4.698681] [<80617538>] (device_initial_probe) from [<80615694>] (bus_probe_device+0x94/0x9c)
[    4.707318] [<80615600>] (bus_probe_device) from [<80615c6c>] (deferred_probe_work_func+0xac/0xf0)
[    4.716305]  r7:81126e90 r6:810c5094 r5:810c5080 r4:810c5080
[    4.721973] [<80615bc0>] (deferred_probe_work_func) from [<80141488>] (process_one_work+0x1fc/0x54c)
[    4.731139]  r10:81119a60 r9:814c300d r8:00000000 r7:814c3000 r6:81412800 r5:817eeb00
[    4.738981]  r4:810c50c0 r3:80615bc0
[    4.742563] [<8014128c>] (process_one_work) from [<80141a48>] (worker_thread+0x270/0x570)
[    4.750765]  r10:81412800 r9:00000088 r8:81003d00 r7:81412818 r6:817eeb18 r5:81412800
[    4.758608]  r4:817eeb00
[    4.761147] [<801417d8>] (worker_thread) from [<80149b1c>] (kthread+0x174/0x190)
[    4.768574]  r10:81c34000 r9:815a7e7c r8:00000000 r7:817eeb00 r6:801417d8 r5:817efc00
[    4.776417]  r4:817efa40
[    4.778955] [<801499a8>] (kthread) from [<80100130>] (ret_from_fork+0x14/0x24)
[    4.786201] Exception stack(0x81c35fb0 to 0x81c35ff8)
[    4.791266] 5fa0:                                     00000000 00000000 00000000 00000000
[    4.799459] 5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
[    4.807651] 5fe0: 00000000 00000000 00000000 00000000 00000013 00000000
[    4.814279]  r10:00000000 r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:801499a8
[    4.822120]  r4:817efa40
[    4.824664] Code: e3a02080 e593001c e3a01d33 e3a05000 (e5979018)

Split the registration from the STDP4028 probe routine and only perform
registration once both the STDP4028 and STDP2690 have probed.

Signed-off-by: Martyn Welch <martyn.welch@collabora.com>
CC: Peter Senna Tschudin <peter.senna@gmail.com>
CC: Martyn Welch <martyn.welch@collabora.co.uk>
CC: Neil Armstrong <narmstrong@baylibre.com>
CC: Robert Foss <robert.foss@linaro.org>
CC: Laurent Pinchart <Laurent.pinchart@ideasonboard.com>
CC: Jonas Karlman <jonas@kwiboo.se>
CC: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Robert Foss <robert.foss@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/43552c3404e8fdf92d8bc5658fac24e9f03c2c57.1637836606.git.martyn.welch@collabora.com

drm/bridge: anx7625: fix an error code in anx7625_register_audio()

This code accidentally returns IS_ERR(), which is 1, instead of
propagating the negative error code.  The caller doesn't check for
errors so it doesn't affect run time at all.

Fixes: 566fef1226c1 ("drm/bridge: anx7625: add HDMI audio function")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Tzung-Bi Shih <tzungbi@google.com>
Signed-off-by: Robert Foss <robert.foss@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20211124145219.GD13656@kili

drm/i915/ttm: Update i915_gem_obj_copy_ttm() to be asynchronous

Update the copy function i915_gem_obj_copy_ttm() to be asynchronous for
future users and update the only current user to sync the objects
as needed after this function.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-7-thomas.hellstrom@linux.intel.com

drm/i915/ttm: Implement asynchronous TTM moves

Don't wait sync while migrating, but rather make the GPU blit await the
dependencies and add a moving fence to the object.

This also enables asynchronous VRAM management in that on eviction,
rather than waiting for the moving fence to expire before freeing VRAM,
it is freed immediately and the fence is stored with the VRAM manager and
handed out to newly allocated objects to await before clears and swapins,
or for kernel objects before setting up gpu vmas or mapping.

To collect dependencies before migrating, add a set of utilities that
coalesce these to a single dma_fence.

What is still missing for fully asynchronous operation is asynchronous vma
unbinding, which is still to be implemented.

This commit substantially reduces execution time in the gem_lmem_swapping
test.

v2:
- Make a couple of functions static.
v4:
- Fix some style issues (Matthew Auld)
- Audit and add more checks for ghost objects (Matthew Auld)
- Add more documentation for the i915_deps utility (Mattew Auld)
- Simplify the i915_deps_sync() function
v6:
- Re-check for fence signaled before returning -EBUSY (Matthew Auld)
- Use dma_resv_iter_is_exclusive() (Matthew Auld)
- Await all dma-resv fences before a migration blit (Matthew Auld)

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-6-thomas.hellstrom@linux.intel.com

drm/i915/ttm: Correctly handle waiting for gpu when shrinking

With async migration, the shrinker may end up wanting to release the
pages of an object while the migration blit is still running, since
the GT migration code doesn't set up VMAs and the shrinker is thus
oblivious to the fact that the GPU is still using the pages.

Add waiting for gpu in the shrinker_release_pages() op and an
argument to that function indicating whether the shrinker expects it
to not wait for gpu. In the latter case the shrinker_release_pages()
op will return -EBUSY if the object is not idle.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-5-thomas.hellstrom@linux.intel.com

drm/i915/ttm: Drop region reference counting

There is an interesting refcounting loop:
struct intel_memory_region has a struct ttm_resource_manager,
ttm_resource_manager->move may hold a reference to i915_request,
i915_request may hold a reference to intel_context,
intel_context may hold a reference to drm_i915_gem_object,
drm_i915_gem_object may hold a reference to intel_memory_region.

Break this loop by dropping region reference counting.

In addition, Have regions with a manager moving fence make sure
that all region objects are released before freeing the region.

v6:
- Fix a code comment.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-4-thomas.hellstrom@linux.intel.com

drm/i915/ttm: Move the i915_gem_obj_copy_ttm() function

Move the i915_gem_obj_copy_ttm() function to i915_gem_ttm_move.h.
This will help keep a number of functions static when introducing
async moves.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-3-thomas.hellstrom@linux.intel.com

drm/i915: Add support for moving fence waiting

For now, we will only allow async migration when TTM is used,
so the paths we care about are related to TTM.

The mmap path is handled by having the fence in ttm_bo->moving,
when pinning, the binding only becomes available after the moving
fence is signaled, and pinning a cpu map will only work after
the moving fence signals.

This should close all holes where userspace can read a buffer
before it's fully migrated.

v2:
- Fix a couple of SPARSE warnings
v3:
- Fix a NULL pointer dereference
v4:
- Ditch the moving fence waiting for i915_vma_pin_iomap() and
  replace with a verification that the vma is already bound.
  (Matthew Auld)
- Squash with a previous patch introducing moving fence waiting and
  accessing interfaces (Matthew Auld)
- Rename to indicated that we also add support for sync waiting.
v5:
- Fix check for NULL and unreferencing i915_vma_verify_bind_complete()
  (Matthew Auld)
- Fix compilation failure if !CONFIG_DRM_I915_DEBUG_GEM
- Fix include ordering. (Matthew Auld)
v7:
- Fix yet another compilation failure with clang if
  !CONFIG_DRM_I915_DEBUG_GEM

Co-developed-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211122214554.371864-2-thomas.hellstrom@linux.intel.com

drm/amd/display: update bios scratch when setting backlight

Update the bios scratch register when updating the backlight
level.  Some platforms apparently read this scratch register
and do additional operations in their hotkey handlers.

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1518
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/pm: fix powerplay OD interface

The overclocking interface currently appends data to a
string.  Revert back to using sprintf().

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1774
Fixes: 6db0c87a0a8ee1 ("amdgpu/pm: Replace hwmgr smu usage of sprintf with sysfs_emit")
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>

drm/amdgpu: Skip ASPM programming on aldebaran

There is no need for additional programming, keep the default settings.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix byteorder error in amdgpu discovery

fix some byteorder issues about amdgpu discovery.
This will result in running errors on the big end system. (e.g:MIPS)

Signed-off-by: Yang Wang <KevinYang.Wang@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: enable Navi retry fault wptr overflow

If xnack is on, VM retry fault interrupt send to IH ring1, and ring1
will be full quickly. IH cannot receive other interrupts, this causes
deadlock if migrating buffer using sdma and waiting for sdma done
while handling retry fault.

Remove VMC from IH storm client, enable ring1 write pointer
overflow, then IH will drop retry fault interrupts and be able to receive
other interrupts while driver is handling retry fault.

IH ring1 write pointer doesn't writeback to memory by IH, and ring1
write pointer recorded by self-irq is not updated, so always read
the latest ring1 write pointer from register.

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>

drm/amdgpu: enable Navi 48-bit IH timestamp counter

By default this timestamp is 32 bit counter. It gets overflowed in
around 10 minutes.

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>

drm/amdkfd: simplify drain retry fault

unmap range always increase atomic svms->drain_pagefaults to simplify
both parent range and child range unmap, page fault handle ignores the
retry fault if svms->drain_pagefaults is set to speed up interrupt
handling. svm_range_drain_retry_fault restart draining if another
range unmap from cpu.

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>