drm/i915: Remove partial attempt to swizzle on pread/pwrite

Our attempt to account for bit17 swizzling of pread/pwrite onto tiled
objects was flawed due to the simple fact that we do not always know the
swizzling for a particular page (due to the swizzling varying based on
location in certain unbalanced configurations). Furthermore, the
pread/pwrite paths are now unbalanced in that we are required to use the
GTT as in some cases we do not have direct CPU access to the backing
physical pages (thus some paths trying to account for the swizzle, but
others neglecting, chaos ensues).

There are no known users who do use pread/pwrite into a tiled object
(you need to manually detile anyway, so why now just use mmap and avoid
the copy?) and no user bug reports to indicate that it is being used in
the wild. As no one is hitting the buggy path, we can just remove the
buggy code.

v2: Just use the fault allowing kmap() + normal copy_(to|from)_user
v3: Avoid int overflow in computing 'length' from 'remain' (Tvrtko)

References: fe115628d567 ("drm/i915: Implement pwrite without struct-mutex")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20190105120758.9237-1-chris@chris-wilson.co.uk

drm/i915: Do not allow unwedging following a failed driver initialisation

If we declare the driver wedged during early initialisation, we leave
the driver in an undefined state (with respect to GEM execution). As
this leads to unexpected behaviour if we allow the user to unwedge the
device (through debugfs, and performed by igt at test start), do not.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190103213340.1669-1-chris@chris-wilson.co.uk

drm/i915: Always try to reset the GPU on takeover

When we first introduced the reset to sanitize the GPU on taking over
from the BIOS and before returning control to third parties (the BIOS!),
we restricted it to only systems utilizing HW contexts as we were
uncertain of how stable our reset mechanism truly was. We now have
reasonable coverage across all machines that expose a GPU reset method,
and so we should be safe to sanitize the GPU state everywhere.

v2: We _have_ to skip the reset if it would clobber the display.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190103112104.19561-1-chris@chris-wilson.co.uk

drm/i915: Show machine type in error state

As the question of 32b/64b kernels became relevant in the light of
certain bugs, include that information in the error state.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190103101245.15100-1-chris@chris-wilson.co.uk

drm/i915/gen6: Flush RING_IMR changes before changing the global GT IMR

On Baytail, notably, we can still detect missed interrupt syndrome
(where we never spot a completed request). In this case, it can be
alleviated by always keeping the interrupt unmasked, implying that the
interrupt is being lost in the window after modifying the IMR. (This is
the reason we still have the posting reads on enable_irq, if we remove
them we miss interrupts!) Having narrowed the issue down to the IMR,
rather than keeping it always enabled, applying the usual posting
read/flush of the RING_IMR before unmasking the GT IMR also seems to
prevent the missed interrupt. So be it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190102163524.19353-1-chris@chris-wilson.co.uk

drm/i915/selftests: Take a breath during check_partial_mappings()

With kasan on a slow machine, it can take an age to check all the
partial mappings in a single iteration, so break it up with a
cond_resched) to avoid RCU stall reports.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190102114431.23022-1-chris@chris-wilson.co.uk

drm/i915: rename dev_priv info to __info to avoid usage

Encourage use of INTEL_INFO() to access dev_priv->info to not accumulate
more direct users of ->info, making further changes easier.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/5f5d81880046331f77624d00278528abc1cf30c6.1546267488.git.jani.nikula@intel.com

drm/i915: drop intel_device_info_dump()

The debugfs, error state and regular dmesg logging dump needs seem to be
different. Remove the generic dump function only used for the welcome
message. This may be added back later when better abstractions are
identified, but at the moment this seems to be the simplest considering
the device info rework in progress. No longer rely on device info being
a substruct of dev_priv.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/70ff0c7c0ec3ef8747af3c78e272b5a82be3d55b.1546267488.git.jani.nikula@intel.com

drm/i915: always use INTEL_INFO() to access device info

Hide the way device info is stored, in preparation of making device info
a pointer to the const rodata in i915_pci.c. No functional changes.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/3cd626f248c0d6638f1288938bbb577a12286050.1546267488.git.jani.nikula@intel.com

drm/i915: pass dev_priv to intel_device_info_runtime_init()

With the static/runtime device info split, this makes more sense.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ad5b448e4e318df0d292d73e6c3378f3e6b9bae5.1546267488.git.jani.nikula@intel.com

drm/i915/reg: abstract display_mmio_offset access

Add a macro wrapper for display_mmio_offset access in register
definitions. Prep work for reducing direct dev_priv->info usage. No
functional changes.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/aa4e8fd85e0445ec5be6c55151239072b4315fda.1546267488.git.jani.nikula@intel.com

drm/i915: start moving runtime device info to a separate struct

First move the low hanging fruit, the fields that are only initialized
runtime. Use RUNTIME_INFO() exclusively to access the fields.

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/c24fe7a4b0492a888690c46814c0ff21ce2f12b1.1546267488.git.jani.nikula@intel.com

drm/i915: Drop unused engine->irq_seqno_barrier w/a

Now that we have eliminated the CPU-side irq_seqno_barrier by moving the
delays on the GPU before emitting the MI_USER_INTERRUPT, we can remove
the engine->irq_seqno_barrier infrastructure. Though intentionally
slowing down the GPU is nasty, so is the code we can now remove!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-6-chris@chris-wilson.co.uk

drm/i915/ringbuffer: Move irq seqno barrier to the GPU for gen5

The irq_seqno_barrier is a tradeoff between doing work on every request
(on the GPU) and doing work after every interrupt (on the CPU). We
presume we have many more requests than interrupts! However, for
Ironlake, the workaround is a pretty hideous usleep() and so even though
it was found we need to repeat the MI_STORE_DWORD_IMM 8 times, or about
1us of GPU time, doing so is preferrable than requiring a sleep of
125-250us on the CPU where we desire to respond immediately (ideally from
within the interrupt handler)!

The additional MI_STORE_DWORD_IMM also have the side-effect of flushing
MI operations from userspace which are not caught by MI_FLUSH!

Testcase: igt/gem_sync
Testcase: igt/gem_exec_whisper
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-5-chris@chris-wilson.co.uk

drm/i915/ringbuffer: Move irq seqno barrier to the GPU for gen7

The irq_seqno_barrier is a tradeoff between doing work on every request
(on the GPU) and doing work after every interrupt (on the CPU). We
presume we have many more requests than interrupts! However, the current
w/a for Ivybridge is an implicit delay that currently fails sporadically
and consistently if we move the w/a into the irq handler itself. This
makes the CPU barrier untenable for upcoming interrupt handler changes
and so we need to replace it with a delay on the GPU before we send the
MI_USER_INTERRUPT. As it turns out that delay is 32x MI_STORE_DWORD_IMM,
or about 0.6us per request! Quite nasty, but the lesser of two evils
looking to the future.

Testcase: igt/gem_sync
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-4-chris@chris-wilson.co.uk

drm/i915/ringbuffer: Remove irq-seqno w/a for gen6 xcs

The MI_FLUSH_DW does appear coherent with the following
MI_USER_INTERRUPT, but only on Sandybridge. Ivybridge requires a heavier
hammer, but on Sandybridge we can stop requiring the irq_seqno barrier.

Testcase: igt/gem_sync
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-3-chris@chris-wilson.co.uk

drm/i915/ringbuffer: Remove irq-seqno w/a for gen6/7 rcs

Having transitioned to using PIPECONTROL to combine the flush with the
breadcrumb write using their post-sync functions, assume that this will
resolve the serialisation with the subsequent MI_USER_INTERRUPT. That is
when inspecting the breadcrumb after an interrupt we can rely on the write
being posted (i.e. the HWSP will be coherent).

Testing using gem_sync shows that the PIPECONTROL + CS stall does
serialise the command streamer sufficient that the breadcrumb lands
before the MI_USER_INTERRUPT. The same is not true for MI_FLUSH_DW.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-2-chris@chris-wilson.co.uk

drm/i915: Remove redundant trailing request flush

Now that we perform the request flushing inline with emitting the
breadcrumb, we can remove the now redundant manual flush. And we can
also remove the infrastructure that remained only for its purpose.

v2: emit_breadcrumb_sz is in dwords, but rq->reserved_space is in bytes

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228171641.16531-1-chris@chris-wilson.co.uk

drm/i915: Update kerneldoc for intel_wm_need_update()

drivers/gpu/drm/i915/intel_display.c:10708: warning: Function parameter or member 'cur' not described in 'intel_wm_need_update'
drivers/gpu/drm/i915/intel_display.c:10708: warning: Function parameter or member 'new' not described in 'intel_wm_need_update'
drivers/gpu/drm/i915/intel_display.c:10708: warning: Excess function parameter 'plane' description in 'intel_wm_need_update'
drivers/gpu/drm/i915/intel_display.c:10708: warning: Excess function parameter 'state' description in 'intel_wm_need_update'

References: cd1d3ee90e5e ("drm/i915: Use intel_ types more consistently for watermark code (v2)")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181231143505.2523-1-chris@chris-wilson.co.uk

drm/i915/params: document I915_PARAMS_FOR_EACH()

Macros with this much magic in them deserve some explanatory text.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/6f012851a54433b23cb4752f9d4ef523165b1e58.1545920737.git.jani.nikula@intel.com

drm/i915: Remove has_pooled_eu static initializer

It is only initialized to zero once so does not need an explicit
initializer.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181231122212.1667-1-jani.nikula@intel.com

drm/i915/params: set i915.enable_hangcheck permissions to 0600

i915.enable_hangcheck has been an outlier since its introduction in
commit 3e0dc6b01f53 ("drm/i915: hangcheck disable parameter") with 0644
permissions, while all the rest are either 0400 or 0600. Follow suit
with 0600.

IGT never reads the value, so there should be no impact.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/5c8f7d1a1654436d38919b7419a209c129db8ad0.1545920737.git.jani.nikula@intel.com

drm/i915/uc: add dev_priv parameter to intel_uc_is_using_* functions

Reveals the build fail fixed in the last hunk. Also prep work.

v2: name it i915 instead of dev_priv (Michal)

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/8e02dcf1b85462d17e96fb183440dd90261b7411.1545920737.git.jani.nikula@intel.com

drm/i915: add a helper to free the members of i915_params

Abstract the one user in anticipation of more. Set the dangling pointers
to NULL while at it.

Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/8637d1e5049dc003718772f19d664aeaf9540856.1545920737.git.jani.nikula@intel.com

drm/i915: add a helper to make a copy of i915_params

Abstract the one user in anticipation of more.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/c6a94b4da8dc723df025b1f602fe46d76d00d53f.1545920737.git.jani.nikula@intel.com

drm/i915/ringbuffer: Pull the render flush into breadcrumb emission

In preparation for removing the manual EMIT_FLUSH prior to emitting the
breadcrumb implement the flush inline with writing the breadcrumb for
ringbuffer emission.

With a combined flush+breadcrumb, we can use a single operation to both
flush and after the flush is complete (post-sync) write the breadcrumb.
This gives us a strongly ordered operation that should be sufficient to
serialise the write before we emit the interrupt; and therefore we may
take the opportunity to remove the irq_seqno_barrier w/a for gen6+.
Although using the PIPECONTROL to write the breadcrumb is slower than
MI_STORE_DWORD_IMM, by combining the operations into one and removing the
extra flush (next patch) it is faster

For gen2-5, we simply combine the MI_FLUSH into the breadcrumb emission,
though maybe we could find a solution here to the seqno-vs-interrupt
issue on Ironlake by mixing up the flush? The answer is no, adding an
MI_FLUSH before the interrupt is insufficient.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228153114.4948-2-chris@chris-wilson.co.uk

drm/i915/execlists: Pull the render flush into breadcrumb emission

In preparation for removing the manual EMIT_FLUSH prior to emitting the
breadcrumb implement the flush inline with writing the breadcrumb for
execlists. Using one command to both flush and write the breadcrumb is
naturally a tiny bit faster than splitting it into two.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228153114.4948-1-chris@chris-wilson.co.uk

drm/i915: Drop debugfs/i915_next_seqno

Having just gutted the implementation as there is no global seqno
tracking, remove the vestigal write-only stub for debugfs/i915_next_seqno.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228140736.32606-3-chris@chris-wilson.co.uk

drm/i915: Remove HW semaphores for gen7 inter-engine synchronisation

The writing is on the wall for the existence of a single execution queue
along each engine, and as a consequence we will not be able to track
dependencies along the HW queue itself, i.e. we will not be able to use
HW semaphores on gen7 as they use a global set of registers (and unlike
gen8+ we can not effectively target memory to keep per-context seqno and
dependencies).

On the positive side, when we implement request reordering for gen7 we
also can not presume a simple execution queue and would also require
removing the current semaphore generation code. So this bring us another
step closer to request reordering for ringbuffer submission!

The negative side is that using interrupts to drive inter-engine
synchronisation is much slower (4us -> 15us to do a nop on each of the 3
engines on ivb). This is much better than it was at the time of introducing
the HW semaphores and equally important userspace weaned itself off
intermixing dependent BLT/RENDER operations (the prime culprit was glyph
rendering in UXA). So while we regress the microbenchmarks, it should not
impact the user.

References: https://bugs.freedesktop.org/show_bug.cgi?id=108888
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228140736.32606-2-chris@chris-wilson.co.uk

drm/i915: Restrict PSMI context load w/a to Haswell GT1

After we found a workaround for a hang on context load, Ben Widawsky
found confirmation that it was for an issue with waking from rc6 and
loading a context image.

The workaround from on high suggests that we should

I915_WRITE(RING_WAIT_FOR_RC6_EXIT(engine->mmio_base),
   _MASKED_FIELD(RING_RC6_SEL_WRITE_ADDR_MASK,
 RING_RC6_SEL_WRITE_ADDR_UPPER_LEFT));

in our rc6 setup for Haswell GT1, but on applying that we find instead
that the machine encounters a GT forcewake error and locks up.

As we are removing HW semaphore usage in the next patch, and the
suggested workaround is no improvement, we need to
decouple the PSMI workaround from HAS_SEMAPHORES to IS_HSW_GT1.

References: 2c550183476d ("drm/i915: Disable PSMI sleep messages on all rings around context switches")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181228140736.32606-1-chris@chris-wilson.co.uk

drm/i915: avoid division by zero on skl_calc_wrpll_link

If for some unexpected reason the registers all read zero it's better
to WARN and return instead of dividing by zero and completely freezing
the machine.

See commit 0e005888b833 ("drm/i915: avoid division by zero on
cnl_calc_wrpll_link") for detail.

Signed-off-by: Young Xiao <YangX92@hotmail.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/SG2PR01MB2169F6E95BC8BB5E29477042ADBC0@SG2PR01MB2169.apcprd01.prod.exchangelabs.com

drm/intel_dsi_vbt: Remove duplicate header

Remove video/mipi_display.h which is included more than once

Signed-off-by: Brajeswar Ghosh <brajeswar.linux@gmail.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181225133055.GA2628@hp-pavilion-15-notebook-pc-brajeswar

drm/i915: Remove debugfs/i915_ppgtt_info

The information presented here is not relevant to current development.
We can either use the context information, but more often we want to
inspect the active gpu state.

The ulterior motive is to eradicate dev->filelist.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181227121549.29139-1-chris@chris-wilson.co.uk

drm/i915: Remove duplicate header from i915_request.h

Remove i915_scheduler.h which is included more than once

Signed-off-by: Brajeswar Ghosh <brajeswar.linux@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181225132340.GA2584@hp-pavilion-15-notebook-pc-brajeswar

drm/i915: DDI: call intel_psr_ and _edp_drrs_enable() on pipe updates (v2)

Call intel_psr_enable() and intel_edp_drrs_enable() on pipe updates to make
sure that we enable PSR / DRRS (when applicable) on fastsets.

Note calling these functions when PSR / DRRS has already been enabled is a
no-op, so it is safe to do this on every encoder->update_pipe callback.

Changes in v2:
-Merge the patches adding the intel_psr_enable() and intel_edp_drrs_enable()
 calls into a single patch

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181220132120.15318-3-hdegoede@redhat.com

drm/i915: Allow calling intel_edp_drrs_enable twice

Do not make it an error to call intel_edp_drrs_enable while drrs has
already been enabled, instead exit silently in this case.

This is a preparation patch for ensuring that DRRS is enabled on fastsets.

Note that the removed WARN_ON could also be triggered from userspace
through the i915_drrs_ctl debugfs entry which was added by
commit 35954e88bc50 ("drm/i915: Runtime disable for eDP DRRS")

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181220132120.15318-2-hdegoede@redhat.com

drm/i915: Add an update_pipe callback to intel_encoder and call this on fastsets (v2)

When we are doing a fastset (needs_modeset=false, update_pipe=true) we
may need to update some encoder-level things such as checking that PSR
is enabled.

This commit adds an update_pipe callback to intel_encoder and a new
intel_encoders_update_pipe helper which calls this for all encoders
connected to a crtc. The new intel_encoders_update_pipe helper is called
from intel_update_crtc when doing a fastset.

Changes in v2:
-Name the new encoder callback update_pipe instead of just update

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181220132120.15318-1-hdegoede@redhat.com

drm/i915: Unwind failure on pinning the gen7 ppgtt

If we fail to pin the ggtt vma slot for the ppgtt page tables, we need
to unwind the locals before reporting the error. Or else on subsequent
attempts to bind the page tables into the ggtt, we will already believe
that the vma has been pinned and continue on blithely. If something else
should happen to be at that location, choas ensues.

Fixes: a2bbf7148342 ("drm/i915/gtt: Only keep gen6 page directories pinned while active")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: <stable@vger.kernel.org> # v4.19+
Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181222030623.21710-1-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20181221

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Update DRIVER_DATE to 20181221

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: don't apply Display WAs 1125 and 1126 to GLK/CNL+

BSpec does not show these WAs as applicable to GLK, and for CNL it
only shows them applicable for a super early pre-production stepping
we shouldn't be caring about anymore. Remove these so we can avoid
them on ICL too.

v2: Change how we check for gen9 display platforms (Ville).

Cc: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181114012432.21809-1-paulo.r.zanoni@intel.com

drm/i915: Disable FBC on fastset if necessary, v2.

Without this, we will get a dmesg-warn when enable_fbc is cleared on a fastset:
WARN_ON(!crtc_state->enable_fbc)
WARNING: CPU: 0 PID: 1090 at drivers/gpu/drm/i915/intel_fbc.c:1091 intel_fbc_enable+0x2ce/0x580 [i915]
RIP: 0010:intel_fbc_enable+0x2ce/0x580 [i915]
Call Trace:
 ? __mutex_unlock_slowpath+0x46/0x2b0
 intel_update_crtc+0x6f/0x2b0 [i915]
 skl_update_crtcs+0x1d1/0x2b0 [i915]
 intel_atomic_commit_tail+0x1ea/0xdb0 [i915]
 intel_atomic_commit+0x244/0x330 [i915]
 drm_mode_atomic_ioctl+0x85d/0x950
 ? drm_atomic_set_property+0x970/0x970
 drm_ioctl_kernel+0x81/0xf0
 drm_ioctl+0x2de/0x390
 ? drm_atomic_set_property+0x970/0x970
 ? __handle_mm_fault+0x81b/0xfc0
 do_vfs_ioctl+0xa0/0x6e0
 ? __do_page_fault+0x2a5/0x550
 ksys_ioctl+0x35/0x60
 __x64_sys_ioctl+0x11/0x20
 do_syscall_64+0x55/0x190
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

Changes since v1:
- Move intel_fbc_disable to intel_update_crtc() (Hans)

Cc: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181220151719.30586-1-maarten.lankhorst@linux.intel.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>

drm/i915/dsc: Fix the deadlock in dsc debugfs node

The DSC debugfs node causes a possible deadlock situation. This patch
resets the try_again at the beginning of loop to fix this.

Fixes: e845f099f1c6 ('drm/i915/dsc: Add Per connector debugfs node for DSC support/enable')
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109097
Cc: Lyude Paul <lyude@redhat.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181219235120.21816-1-manasi.d.navare@intel.com

drm/i915/icl: Add fallback detection method for TypeC legacy ports

Add a fallback detection method for TypeC legacy ports in case the
VBT port information used to detect normally such ports is
incorrect.

For the fallback method we use the TypeC legacy mode specific HPD
interrupt flag which should only be raised for a legacy port.

WARN if the VBT port info is incorrect.

In a case where we'd detect the port in a contradicting way both as a
legacy and also as a USB DP and/or TBT alternate port treat the port
as legacy (by also emitting a WARN from icl_update_tc_port_type).

v2:
- Repurpose the detection as a fallback method instead of using
  it only for the DP legacy case. By now we should normally use VBT to
  detect DP legacy ports as well.

Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> (v1)
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181214182703.18865-5-imre.deak@intel.com

drm/i915/icl: Fix HPD handling for TypeC legacy ports

Atm HPD disconnect events on TypeC ports will break things, since we'll
switch the TypeC mode (between legacy and disconnected modes as well as
among USB DP alternate, Thunderbolt alternate and disconnected modes) on
the fly from the HPD disconnect interrupt work while the port may be
still active.

Even if the port happens to be not active during the disconnect we'd
still have a problem during a subsequent modeset or AUX transfer that
could happen regardless of the port's connected state. For instance the
system resume display mode restore code and userspace could perform a
modeset on the port or userspace could start an AUX transfer even if the
port is in disconnected state.

To fix this keep TypeC legacy ports in legacy mode whenever we're not
suspended. This mode is a static configuration as opposed to the
Thunderbolt and USB DP alternate modes between which we can switch
dynamically.

We determine if a TypeC port is legacy (wired to a legacy HDMI or a
legacy DP connector) via the VBT DDI port specific USB-TypeC and
Thunderbolt flags. If both these flags are cleared then the port is
configured for legacy mode.

On such legacy ports we'll run the TypeC PHY connect sequence explicitly
during driver loading and system resume (vs. running the sequence during
HPD processing). The connect will succeed even if the display is not
connected to begin with (or disappears during the suspended state) since
for legacy ports the PORT_TX_DFLEXDPPMS / DP_PHY_MODE_STATUS_COMPLETED
flag is always set (as opposed to the USB DP alternate mode where it
gets set only when a display is connected).

Correspondingly run the TypeC PHY disconnect sequence during system
suspend and driver unloading. For the unloading case I had to split
up intel_dp_encoder_destroy() to be able to have the 1. flush any
pending encoder work, 2. disconnect TC PHY, 3. call DRM core cleanup and
kfree on the encoder object.

For now run the PHY disconnect during suspend only for TypeC legacy
ports. We will need to disconnect even in USB DP alternate mode in the
future, but atm we don't have a way to reconnect the port in this mode
during resume if the display disappears while being suspended. So for
now punt on this case.

Note that we do not disconnect the port during runtime suspend; in
legacy mode there are no shared HW resources (PHY lanes) with other HW
blocks (USB), so no need to release / reacquire these resources as with
USB DP alternate mode. The only reason to disconnect legacy ports during
system suspend is that the PORT_TX_DFLEXDPPMS /
DP_PHY_MODE_STATUS_COMPLETED flag must be rechecked and the port must be
connected again during system resume. We'll also have to turn the check
for this flag into a poll, after figuring out what's the proper timeout
value for it.

v2:
- Remove the redundant special casing of legacy mode when doing a
  disconnect in icl_tc_port_connected(). It's guaranteed already that we
  won't disconnect legacy ports in that function.
- Add a note about the new intel_ddi_encoder_destroy() hook.
- Reword the commit message after switching to the VBT based detection.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108070
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108924
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181214182703.18865-4-imre.deak@intel.com

drm/i915/bios: Parse the VBT TypeC and Thunderbolt port flags

This is needed by the next patch to determine if a DDI TypeC port is
physically wired to a legacy DP or legacy HDMI connector or if the port
is wired to a USB-C/Thunderbolt connector.

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181214182703.18865-3-imre.deak@intel.com

drm/i915/icl: Add a debug print for TypeC port disconnection

It's useful to see at which point a TypeC port gets disconnected, so add
a debug print for it.

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: José Roberto de Souza <jose.souza@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181214182703.18865-2-imre.deak@intel.com

drm/i915: Apply missed interrupt after reset w/a to all ringbuffer gen

Having completed a test run of gem_eio across all machines in CI we also
observe the phenomenon (of lost interrupts after resetting the GPU) on
gen3 machines as well as the previously sighted gen6/gen7. Let's apply
the same HWSTAM workaround that was effective for gen6+ for all, as
although we haven't seen the same failure on gen4/5 it seems prudent to
keep the code the same.

As a consequence we can remove the extra setting of HWSTAM and apply the
register from a single site.

v2: Delazy and move the HWSTAM into its own function
v3: Mask off all HWSP writes on driver unload and engine cleanup.
v4: And what about the physical hwsp?
v5: No, engine->init_hw() is not called from driver_init_hw(), don't be
daft. Really scrub HWSTAM as early as we can in driver_init_mmio()
v6: Rename set_hwsp as it was setting the mask not the hwsp register.
v7: Ville pointed out that although vcs(bsd) was introduced for g4x/ilk,
per-engine HWSTAM was not introduced until gen6!

References: https://bugs.freedesktop.org/show_bug.cgi?id=108735
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181218102712.11058-1-chris@chris-wilson.co.uk

drm/i915/icl: combo port vswing programming changes per BSPEC

In August 2018 the BSPEC changed the ICL port programming sequence to
closely resemble earlier gen programming sequence. Restrict combo phy to
HBR max rate unless eDP panel is connected to port.

v2: remove debug code that Imre found
v3: simplify translation table if-else
v4: edp translation table now based on link rate and low_swing
v5: Misc review comments + r-b
BSpec: 21257
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1545084827-5776-1-git-send-email-clinton.a.taylor@intel.com

drm/i915: Update crtc scaler settings when update_pipe is set

When the pipe_config's update_pipe flag is set we may need to update the
panel fitting settings. On GEN9+ this means we need to update the crtc's
scaler settings.

This fixes the following WARN_ON, during i915 loading on an Asrock
B150M Pro4S/D3 board with an i5-6500 CPU / graphics:

[drm:pipe_config_err [i915]] *ERROR* mismatch in pch_pfit.enabled
 (expected no, found yes)
pipe state doesn't match!
WARNING: CPU: 3 PID: 305 at drivers/gpu/drm/i915/intel_display.c:12084

With line 12084 being the I915_STATE_WARN call inside the
"if (!intel_pipe_config_compare())" block in verify_crtc_state().

On this board with 2 1920x1080 monitors connected over HDMI the GOP
initializes both monitors at 1920x1080 and despite no scaling being
necessary configures a scaler for one of them.

When booting with fastboot=1 on the initial modeset needs_modeset will
be false while update_pipe is true. Since we were not calling
skl_update_scaler_crtc() in this case we would leave the scaler enabled
causing this error.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181217141903.4182-1-hdegoede@redhat.com

drm/i915/huc: Update the HuC version for BXT

We have an update for HuC for BXT.
Load the latest version.

v2: Change the subject.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Anusha Srivatsa <anusha.srivatsa@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/20181207182840.9292-2-anusha.srivatsa@intel.com

drm/i915/dsc: Add Per connector debugfs node for DSC support/enable

DSC can be supported per DP connector. This patch adds a per connector
debugfs node to expose DSC support capability by the kernel.
The same node can be used from userspace to force DSC enable.

force_dsc_en written through this debugfs node is used to force
DSC even for lower resolutions.

Credits to Ville Syrjala for suggesting the proper locks to be used
and to Lyude Paul for explaining how to use them in this context

v8:
* Add else if (ret) for drm_modeset_lock (Lyude)
v7:
* Get crtc, crtc_state from connector atomic state
and add proper locks and backoff (Ville, Chris Wilson, Lyude)
(Suggested-by: Ville Syrjala <ville.syrjala@linux.intel.com>)
* Use %zu for printing size_t variable (Lyude)
v6:
* Read fec_capable only for non edp (Manasi)
v5:
* Name it dsc sink support and also add
fec support in the same node (Ville)
v4:
* Add missed connector_status check (Manasi)
* Create i915_dsc_support node only for Gen >=10 (manasi)
* Access intel_dp->dsc_dpcd only if its not NULL (Manasi)
v3:
* Combine Force_dsc_en with this patch (Ville)
v2:
* Use kstrtobool_from_user to avoid explicit error checking (Lyude)
* Rebase on drm-tip (Manasi)

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Lyude Paul <lyude@redhat.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181206005407.4698-1-manasi.d.navare@intel.com

drm/i915: Fix Cherryview oops on boot

Do not dereference the LUT blob before checking whether that blob
exists. Or else,

<1>[   13.978684] BUG: unable to handle kernel NULL pointer dereference at 0000000000000048
<6>[   13.978718] PGD 0 P4D 0
<4>[   13.978733] Oops: 0000 [#1] PREEMPT SMP PTI
<4>[   13.978750] CPU: 0 PID: 282 Comm: modprobe Not tainted 4.20.0-rc5-CI-CI_DRM_5294+ #1
<4>[   13.978773] Hardware name:  /NUC5CPYB, BIOS PYBSWCEL.86A.0058.2016.1102.1842 11/02/2016
<4>[   13.978932] RIP: 0010:cherryview_load_csc_matrix+0x1e6/0x210 [i915]
<4>[   13.978953] Code: 41 5c 41 5d e9 7b 83 aa e1 41 c1 e4 0d 48 83 bd 00 02 00 00 00 48 8b 85 10 02 00 00 45 89 e5 74 09 ba 01 00 00 00 31 c9 eb 9d <48> 8b 50 48 48 c1 ea 03 81 fa 00 01 00 00 75 07 31 d2 48 85 c0 75
<4>[   13.979001] RSP: 0018:ffffc9000026f840 EFLAGS: 00010246
<4>[   13.979018] RAX: 0000000000000000 RBX: ffff888165500000 RCX: 7885fe6200000000
<4>[   13.979039] RDX: 0000000000000020 RSI: ffff88816553a008 RDI: ffff888165464a88
<4>[   13.979060] RBP: ffff888165464a88 R08: 000000000ed0e429 R09: 0000000000000001
<4>[   13.979080] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000004000
<4>[   13.979101] R13: 0000000000004000 R14: ffff888165500000 R15: ffff888165464a88
<4>[   13.979122] FS:  00007fb69c4f3540(0000) GS:ffff88817ba00000(0000) knlGS:0000000000000000
<4>[   13.979146] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
<4>[   13.979163] CR2: 0000000000000048 CR3: 000000016d7fa000 CR4: 00000000001006f0
<4>[   13.979184] Call Trace:
<4>[   13.979302]  intel_update_crtc+0x18f/0x2b0 [i915]
<4>[   13.979421]  intel_update_crtcs+0x49/0x60 [i915]
<4>[   13.979538]  intel_atomic_commit_tail+0x1ea/0xd70 [i915]
<4>[   13.979657]  ? intel_atomic_commit_ready+0x3f/0x50 [i915]
<4>[   13.979762]  ? __i915_sw_fence_complete+0x1a0/0x250 [i915]
<4>[   13.979884]  intel_atomic_commit+0x244/0x330 [i915]
<4>[   13.980002]  intel_initial_commit+0xb6/0x140 [i915]
<4>[   13.980127]  intel_modeset_init+0x7a1/0x1880 [i915]
<4>[   13.980235]  i915_driver_load+0xcbb/0x15c0 [i915]
<4>[   13.980257]  ? __pm_runtime_resume+0x4f/0x80
<4>[   13.980277]  ? _raw_spin_unlock_irqrestore+0x4c/0x60
<4>[   13.980296]  ? lockdep_hardirqs_on+0xe0/0x1b0
<4>[   13.980401]  i915_pci_probe+0x29/0xa0 [i915]
<4>[   13.980421]  pci_device_probe+0xa1/0x130
<4>[   13.980440]  really_probe+0xf3/0x3e0
<4>[   13.980456]  driver_probe_device+0x10a/0x120
<4>[   13.980474]  __driver_attach+0xdb/0x100
<4>[   13.980489]  ? driver_probe_device+0x120/0x120
<4>[   13.980505]  ? driver_probe_device+0x120/0x120
<4>[   13.980522]  bus_for_each_dev+0x74/0xc0
<4>[   13.980539]  bus_add_driver+0x15f/0x250
<4>[   13.980554]  ? 0xffffffffa0348000
<4>[   13.980568]  driver_register+0x56/0xe0
<4>[   13.980583]  ? 0xffffffffa0348000
<4>[   13.980597]  do_one_initcall+0x58/0x2e0
<4>[   13.980615]  ? do_init_module+0x1d/0x1ea
<4>[   13.980631]  ? rcu_read_lock_sched_held+0x6f/0x80
<4>[   13.980649]  ? kmem_cache_alloc_trace+0x264/0x290
<4>[   13.980668]  do_init_module+0x56/0x1ea
<4>[   13.980685]  load_module+0x227a/0x29c0
<4>[   13.980715]  ? __se_sys_finit_module+0xd3/0xf0
<4>[   13.980731]  __se_sys_finit_module+0xd3/0xf0
<4>[   13.980756]  do_syscall_64+0x55/0x190
<4>[   13.980772]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
<4>[   13.980789] RIP: 0033:0x7fb69c019839
<4>[   13.980804] Code: 00 f3 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 1f f6 2c 00 f7 d8 64 89 01 48
<4>[   13.980851] RSP: 002b:00007ffdc112e3a8 EFLAGS: 00000246 ORIG_RAX: 0000000000000139
<4>[   13.980875] RAX: ffffffffffffffda RBX: 000055c689fe0b30 RCX: 00007fb69c019839
<4>[   13.980895] RDX: 0000000000000000 RSI: 000055c689a05d2e RDI: 0000000000000000
<4>[   13.980916] RBP: 000055c689a05d2e R08: 0000000000000000 R09: 0000000000000000
<4>[   13.980936] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
<4>[   13.980957] R13: 000055c689fe0c60 R14: 0000000000040000 R15: 000055c689fe0b30
<4>[   13.980986] Modules linked in: i915(+) snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm pinctrl_cherryview prime_numbers
<4>[   13.981027] CR2: 0000000000000048

Fixes: 302da0cdf784 ("drm/i915: Use intel_ types more consistently for color management code (v2)")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109054
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213161241.3461-1-chris@chris-wilson.co.uk

drm/i915/icl: Mind the SFC units when resetting VD or VEBox engines

SFC (Scaler & Format Converter) units are shared between VD and VEBoxes.
They also happen to have separate reset bits. So, whenever we want to reset
one or more of the media engines, we have to make sure the SFCs do not
change owner in the process and, if this owner happens to be one of the
engines being reset, we need to reset the SFC as well.

This happens in 4 steps:

1) Tell the engine that a software reset is going to happen. The engine
will then try to force lock the SFC (if currently locked, it will
remain so; if currently unlocked, it will ignore this and all new lock
requests).

2) Poll the ack bit to make sure the hardware has received the forced
lock from the driver. Once this bit is set, it indicates SFC status
(lock or unlock) will not change anymore (until we tell the engine it
is safe to unlock again).

3) Check the usage bit to see if the SFC has ended up being locked to
the engine we want to reset. If this is the case, we have to reset
the SFC as well.

4) Unlock all the SFCs once the reset sequence is completed.

Obviously, if we are resetting the whole GPU, we don't have to worry
about all of this.

BSpec: 10989
BSpec: 10990
BSpec: 10954
BSpec: 10955
BSpec: 10956
BSpec: 19212

Signed-off-by: Tomasz Lis <tomasz.lis@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213091522.2926-4-chris@chris-wilson.co.uk

drm/i915/icl: Record the valid VDBoxes with SFC capability

In Gen11, only even numbered "logical" VDBoxes are hooked up to an SFC
(Scaler & Format Converter) unit. We will use this information to decide
when the SFC units need to be reset.

BSpec: 20189

Signed-off-by: Tomasz Lis <tomasz.lis@intel.com>
Signed-off-by: Oscar Mateo <oscar.mateo@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213091522.2926-3-chris@chris-wilson.co.uk

drm/i915/selftests: Verify we can perform resets from atomic context

We currently require that our per-engine reset can be called from any
context, even hardirq, and in the future wish to perform the device
reset without holding struct_mutex (which requires some lockless
shenanigans that demand the lowlevel intel_reset_gpu() be able to be
used in atomic context). Test that we meet the current requirements by
calling i915_reset_engine() from under various atomic contexts.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213091522.2926-2-chris@chris-wilson.co.uk

drm/i915/selftests: Check we can recover a wedged device

After declaring a terminally wedged device, we allow ourselves to
recover on the next GPU reset (manually triggered), or resume. Check
that resetting a wedged device does work.

v2: Add rpm (taken explicitly in the subtest in case we remove the outer
wakeref) and early warning to i915_reset() for missed wakerefs

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181213091522.2926-1-chris@chris-wilson.co.uk

drm/i915: merge gen checks to use range

Instead of using IS_GEN() for consecutive gen checks, let's pass the
range to IS_GEN_RANGE(). By code inspection these were the ranges deemed
necessary for spatch:

@@
expression e;
@@
(
- IS_GEN(e, 3) || IS_GEN(e, 2)
+ IS_GEN_RANGE(e, 2, 3)
|
- IS_GEN(e, 3) || IS_GEN(e, 4)
+ IS_GEN_RANGE(e, 3, 4)
|
- IS_GEN(e, 5) || IS_GEN(e, 6)
+ IS_GEN_RANGE(e, 5, 6)
|
- IS_GEN(e, 6) || IS_GEN(e, 7)
+ IS_GEN_RANGE(e, 6, 7)
|
- IS_GEN(e, 7) || IS_GEN(e, 8)
+ IS_GEN_RANGE(e, 7, 8)
|
- IS_GEN(e, 8) || IS_GEN(e, 9)
+ IS_GEN_RANGE(e, 8, 9)
|
- IS_GEN(e, 10) || IS_GEN(e, 9)
+ IS_GEN_RANGE(e, 9, 10)
|
- IS_GEN(e, 9) || IS_GEN(e, 10)
+ IS_GEN_RANGE(e, 9, 10)
)

After conversion, checking we don't have any missing IS_GEN_RANGE() ||
IS_GEN() was also done.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181212181044.15886-3-lucas.demarchi@intel.com

drm/i915: replace IS_GEN<N> with IS_GEN(..., N)

Define IS_GEN() similarly to our IS_GEN_RANGE(). but use gen instead of
gen_mask to do the comparison. Now callers can pass then gen as a parameter,
so we don't require one macro for each gen.

The following spatch was used to convert the users of these macros:

@@
expression e;
@@
(
- IS_GEN2(e)
+ IS_GEN(e, 2)
|
- IS_GEN3(e)
+ IS_GEN(e, 3)
|
- IS_GEN4(e)
+ IS_GEN(e, 4)
|
- IS_GEN5(e)
+ IS_GEN(e, 5)
|
- IS_GEN6(e)
+ IS_GEN(e, 6)
|
- IS_GEN7(e)
+ IS_GEN(e, 7)
|
- IS_GEN8(e)
+ IS_GEN(e, 8)
|
- IS_GEN9(e)
+ IS_GEN(e, 9)
|
- IS_GEN10(e)
+ IS_GEN(e, 10)
|
- IS_GEN11(e)
+ IS_GEN(e, 11)
)

v2: use IS_GEN rather than GT_GEN and compare to info.gen rather than
    using the bitmask

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181212181044.15886-2-lucas.demarchi@intel.com

drm/i915: Rename IS_GEN to IS_GEN_RANGE

RANGE makes it longer, but clearer. We are also going to add a macro to
check an individual gen, so add the _RANGE prefix here.

Diff generated with:

sed 's/IS_GEN(/IS_GEN_RANGE(/g' drivers/gpu/drm/i915/{*/,}*.{c,h} -i

v2: use IS_GEN rather than GT_GEN

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181212181044.15886-1-lucas.demarchi@intel.com

drm/i915: Don't forget to reset blocks when testing lower wm levels

During DDB allocation, we try to distribute enough blocks for each plane
to hit the highest watermark level; if that fails, we retry each lower
level (which should require fewer blocks) until we find one that's
possible (or until the whole commit is rejected as impossible).  We need
to reset our running block count when trying each lower level, otherwise
all lower levels will fail as well.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: d8e8749802 ("drm/i915: Switch to level-based DDB allocation algorithm (v5)")
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181212191720.3706-1-matthew.d.roper@intel.com

drm/i915: DFSM pipe disable is valid from gen9 onwards (v2)

It's not just GEN9 platforms that allow for pipes to be disabled via
the DFSM register, but all later platforms as well.

v2: drop pointless parentheses (Ville)

Signed-off-by: Bob Paauwe <bob.j.paauwe@intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181211192545.140081-1-bob.j.paauwe@intel.com

drm/i915: Switch to level-based DDB allocation algorithm (v5)

The DDB allocation algorithm currently used by the driver grants each
plane a very small minimum allocation of DDB blocks and then divies up
all of the remaining blocks based on the percentage of the total data
rate that the plane makes up.  It turns out that this proportional
allocation approach is overly-generous with the larger planes and can
leave very small planes wthout a big enough allocation to even hit their
level 0 watermark requirements (especially on APL, which has a smaller
DDB in general than other gen9 platforms).  Or there can be situations
where the smallest planes hit a lower watermark level than they should
have been able to hit with a more equitable division of DDB blocks, thus
limiting the overall system sleep state that can be achieved.

The bspec now describes an alternate algorithm that can be used to
overcome these types of issues.  With the new algorithm, we calculate
all plane watermark values for all wm levels first, then go back and
partition a pipe's DDB space second.  The DDB allocation will calculate
what the highest watermark level that can be achieved on *all* active
planes, and then grant the blocks necessary to hit that level to each
plane.  Any remaining blocks are then divided up proportionally
according to data rate, similar to the old algorithm.

There was a previous attempt to implement this algorithm a couple years
ago in bb9d85f6e9d ("drm/i915/skl: New ddb allocation algorithm"), but
some regressions were reported, the patch was reverted, and nobody
ever got around to figuring out exactly where the bug was in that
version.  Our watermark code has evolved significantly in the meantime,
but we're still getting bug reports caused by the unfair proportional
algorithm, so let's give this another shot.

v2:
 - Make sure cursor allocation stays constant and fixed at the end of
   the pipe allocation.
 - Fix some watermark level iterators that weren't handling the max
   level.

v3:
 - Ensure we don't leave any DDB blocks unused by using DIV_ROUND_UP+min
   to calculate the extra blocks for each plane.  (Ville)
 - Replace a while() loop with a for() loop to be more consistent with
   surrounding code.  (Ville)
 - Clean unattainable watermark levels with memset rather than directly
   clearing the member fields.  Also do the same for the transition
   watermark values if they can't be achieved.  (Ville)
 - Drop min_disp_buf_needed calculations in skl_compute_plane_wm() since
   the results are no longer needed or used.  (Ville)
 - Drop skl_latency[0] != 0 sanity check; both watermark methods already
   account for an invalid 0 latency by returning FP_16_16_MAX.  (Ville)

v4:
 - Break DDB allocation loop when total_data_rate=0 rather than
   alloc_size=0.  If total_data_rate has dropped to 0, all remaining
   planes are disabled, which isn't true for alloc_size (we might just
   have not had any remaining blocks to hand out).  Plus
   total_data_rate=0 is the case we need to avoid to a prevent a
   div-by-0.  (Ville)
 - s/DIV_ROUND_UP/DIV64_U64_ROUND_UP/ to prevent 32-bit breakage (Ville)

v5:
 - Don't forget to move 'start' pointer forward for UV surface when
   setting plane DDB boundaries.  (Ville)

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105458
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181211173107.11068-2-matthew.d.roper@intel.com

drm/i915: Don't use DDB allocation when choosing gen9 watermark method

The bspec gives an if/else chain for choosing whether to use "method 1"
or "method 2" for calculating the watermark "Selected Result Blocks"
value for a plane.  One of the branches of the if chain is:

        "Else If ('plane buffer allocation' is known and (plane buffer
        allocation / plane blocks per line) >=1)"

Since our driver currently calculates DDB allocations first and the
actual watermark values second, the plane buffer allocation is known at
this point in our code and we include this test in our driver's logic.
However we plan to soon move to a "watermarks first, ddb allocation
second" sequence where we won't know the DDB allocation at this point.
Let's drop this arm of the if/else statement (effectively considering
the DDB allocation unknown) as an independent patch so that any
regressions can be more accurately bisected to either the different
watermark value (in this patch) or the new DDB allocation (in the next
patch).

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181211173107.11068-1-matthew.d.roper@intel.com

drm/i915/hdmi: SCDC Scrambling enable without CTS mode

Setting the SCDC scrambling CTS mode causes HDMI Link Layer protocol tests
HF1-12 and HF1-13 to fail.

V2: Removed "Source Shall" entries to a new patch
V3: Rebase to drm-tip
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107895
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107896
Signed-off-by: Clint Taylor <clinton.a.taylor@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1544482374-26507-1-git-send-email-clinton.a.taylor@intel.com

drm/i915: Use intel_ types more consistently for color management code (v2)

Try to be more consistent about intel_* types rather than drm_* types
for lower-level driver functions.  While we're at it, let's also be more
consistent with state variable naming (half of the platforms use the
name 'state' whereas the other half used 'crtc_state').

While we're touching these variables, let's also be more consistent
about always naming the intel_crtc_state's "crtc_state" rather than
"state" so that different platform types aren't using different naming
conventions.

v2:
 - s/state/crtc_state/ for consistency between platform types (Ville)
 - Drop the crtc parameter to intel_color_check(); we can just pull that
   out of the state object.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181210215415.19854-2-matthew.d.roper@intel.com

drm/i915: Use intel_ types more consistently for watermark code (v2)

Try to be more consistent about intel_* types rather than drm_* types
for lower-level driver functions.

v2:
 - Also drop the intel_crtc parameter from compute_intermediate_wm()
   since we can just extract it from the crtc_state parameter. (Ville)

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181210215415.19854-1-matthew.d.roper@intel.com

drm/i915: Remove dead update_wm_pre assignment from SKL wm code

SKL+ do not use crtc_state->update_wm_pre, so there is absolutely no
point it setting it. crtc_state->update_wm_pre only exists as a
temporary hack for pre-g4x platforms until we redo their
watermarks to be be atomic.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181113172330.26069-3-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Remove bogus FIXME from SKL wm computation

We do return an error when the watermark calculation fails, so
the FIXME claiming otherwise is outdated. Remove it.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181113172330.26069-2-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Use explicit old crtc state in skl_compute_wm()

skl_compute_wm() wants to compare the old and new watermarks. Currently
it gets at the old watermarks via crtc->state, which is confusing since
it can point at either the old or the new state depending on where
in the sequence we are. In this case it is correct since we have not yet
swapped the states, but let's make it super clear what this is doing
by using the explicit old state.

Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181113172330.26069-1-ville.syrjala@linux.intel.com
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

drm/i915: Flush GPU relocs harder for gen3

Adding an extra MI_STORE_DWORD_IMM to the gpu relocation path for gen3
was good, but still not good enough. To survive 24+ hours under test we
needed to perform not one, not two but three extra store-dw. Doing so
for each GPU relocation was a little unsightly and since we need to
worry about userspace hitting the same issues, we should apply the dummy
store-dw into the EMIT_FLUSH.

Fixes: 7dd4f6729f92 ("drm/i915: Async GPU relocation processing")
References: 7fa28e146994 ("drm/i915: Write GPU relocs harder with gen3")
Testcase: igt/gem_tiled_fence_blits # blb/pnv
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181207134037.11848-1-chris@chris-wilson.co.uk

drm/i915: Skip the ERR_PTR error state

Although commit fb6f0b64e455 ("drm/i915: Prevent machine hang from
Broxton's vtd w/a and error capture") applied cleanly after a 24 month
hiatus, the code had moved on with new methods for peeking and fetching
the captured gpu info. Make sure we catch all uses of the stashed error
state and avoid dereferencing the error pointer.

v2: Move error pointer determination into i915_gpu_capture_state
v3: Restore early check to avoid capturing and then throwing away
subsequent GPU error states.

Fixes: fb6f0b64e455 ("drm/i915: Prevent machine hang from Broxton's vtd w/a and error capture")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181207110554.19897-1-chris@chris-wilson.co.uk

drm/i915: Pipeline PDP updates for Braswell

Currently we face a severe problem on Braswell that manifests as invalid
ppGTT accesses. The code tries to maintain the PDP (page directory
pointers) inside the context in two ways, direct write into the context
and a pipelined LRI update. The direct write into the context is
fundamentally racy as it is unserialised with any access (read or write)
the GPU is doing. By asserting that Braswell is not used with vGPU
(currently an unsupported platform) we can eliminate the dangerous
direct write into the context image and solely use the pipelined update.

However, the LRI of the PDP fouls up the GPU, causing it to freeze and
take out the machine with "forcewake ack timeouts". This seems possible
to workaround by preventing the GPU from sleeping (via means of
disabling the power-state management interface, i.e. forcing each ring
to remain awake) around the update. Equally, it seems an EMIT_INVALIDATE
before the LRI is sufficient to prevent the forcewake errors.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108656
References: https://bugs.freedesktop.org/show_bug.cgi?id=108714
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181207090213.14352-3-chris@chris-wilson.co.uk

drm/i915/ringbuffer: EMIT_INVALIDATE after switch context

The recommend procedure was to switch contexts (and mm) then invalidate
the TLBs. Make it so.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181207090213.14352-2-chris@chris-wilson.co.uk

drm/i915: Push EMIT_INVALIDATE at request start to backends

Move the common engine->emit_flush(EMIT_INVALIDATE) back to the backends
(where it was once previously) as we seek to specialise it in future
patches.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181207090213.14352-1-chris@chris-wilson.co.uk

drm/i915/icl: Forcibly evict stale csb entries

Gen11 fails to deliver wrt global observation point on
tail/entry updates and we sometimes see old entry.

Use clflush to forcibly evict our possibly stale copy
of the cacheline in hopes that we get fresh one from gpu.
Obviously there is something amiss in the coherency protocol so
this can be consired as a workaround until real cause
is found.

The working hardware will do the evict without our cue anyways,
so the cost in there should be ameliorated by that fact.

v2: for next pass, s/flush/evict, add reset (Chris)

References: https://bugzilla.freedesktop.org/show_bug.cgi?id=108315
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181205134612.24822-1-mika.kuoppala@linux.intel.com

drm/i915/execlists: Apply a full mb before execution for Braswell

Braswell is really picky about having our writes posted to memory before
we execute or else the GPU may see stale values. A wmb() is insufficient
as it only ensures the writes are visible to other cores, we need a full
mb() to ensure the writes are in memory and visible to the GPU.

The most frequent failure in flushing before execution is that we see
stale PTE values and execute the wrong pages.

References: 987abd5c62f9 ("drm/i915/execlists: Force write serialisation into context image vs execution")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: stable@vger.kernel.org
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181206084431.9805-3-chris@chris-wilson.co.uk

drm/i915/execlists: Move RCS mmio workaround to new common wa_list

We can move the remaining RCS workarounds applied to only gen8 to the
engine->wa_list, and then reduce all engine->init_hw callbacks to common
code. The benefit of using the new wa_list is that we verify that the
registers are indeed restored and keep their magic values.

v2: INSTPM_FORCE_ORDERING is already part of gen8_ctx_workarounds, and
as confirmed by the mmio verification is a part of the context image!
v3: MI_MODE is already part of gen8_ctx_workarounds...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181206180713.6827-2-chris@chris-wilson.co.uk

drm/i915/selftests: verify_gt_engine_wa() needs rpm wakeref

The mmio readback for verify_gt_engine_wa() also needs a runtime-pm
wakeref, so effectively do the entirety of both engine workarounds
tests. As such simplify the rpm behaviour here by acquiring the wakeref
for the whole of each subtest. It would be still useful to later verify
the registers retain their magic values across rpm suspend/resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181206180713.6827-1-chris@chris-wilson.co.uk

drm/i915: Increase timeout for Encrypt status change

At enable/disable of the HDCP encryption, for encryption status change
we need minimum one frame duration. And we might program this bit any
point(start/End) in the previous frame.

With 20mSec, observed the timeout for change in encryption status.
Since this is not time critical operation and we need to hold on
until the status is changed, fixing the timeout to 50mSec. (Based on
trial and error method!)

v2:
  %s/TIME_FOR_ENCRYPT_STATUS_CHANGE/ENCRYPT_STATUS_CHANGE_TIMEOUT_MS
[Sean Paul]

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Sean Paul <sean@poorly.run>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1544010283-20223-5-git-send-email-ramalingam.c@intel.com

drm/i915: debug log for REPLY_ACK missing

Adding a debug log when the DP_AUX_NATIVE_REPLY_ACK is missing
for aksv write. This helps to locate the possible non responding
DP HDCP sinks.

v2:
  Rewritten for readability [Sean Paul]

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Sean Paul <sean@poorly.run>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1544010283-20223-4-git-send-email-ramalingam.c@intel.com

drm/i915: Fix platform coverage for HDCP1.4

HDCP1.4 is enabled and validated only on GEN9+ platforms.

v2:
  Removed the unnecessary parens [Ville]

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Sean Paul <sean@poorly.run>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1544010283-20223-3-git-send-email-ramalingam.c@intel.com

drm/i915: Fix GEN9 HDCP1.4 key load process

HDCP1.4 key load process varies between Intel platform to platform.

For Gen9 platforms except BXT and GLK, HDCP1.4 key is loaded using
the GT Driver Mailbox interface. So all GEN9_BC platforms will use
the GT Driver Mailbox interface for HDCP1.4 key load.

v2:
  Using the IS_GEN9_BC for filtering the platforms [Ville]

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Sean Paul <sean@poorly.run>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/1544010283-20223-2-git-send-email-ramalingam.c@intel.com

drm/i915/fbc/cnl: Add GLK and CNL+ hardware tracking area

GLK and CNL+ supports a bigger FBC tracking area.

Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181205004823.30823-1-jose.souza@intel.com

drm: Fix documentation generation for DP_DPCD_QUIRK_NO_PSR

The DP_DPCD_QUIRK_NO_PSR comment is missing colon causing this
warning when generating kernel documentation.

./include/drm/drm_dp_helper.h:1374: warning: Incorrect use of kernel-doc format:          * @DP_DPCD_QUIRK_NO_PSR

Fixes: 7c5c641a930e (drm/i915: Disable PSR in Apple panels)
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181205184850.13218-1-jose.souza@intel.com

drm/i915: implement EXTENDED_RECEIVER_CAPABILITY_FIELD_PRESENT

According to DP spec (2.9.3.1 of DP 1.4) if
EXTENDED_RECEIVER_CAPABILITY_FIELD_PRESENT is set the addresses in DPCD
02200h through 0220Fh shall contain the DPRX's true capability. These
values will match 00000h through 0000Fh, except for DPCD_REV,
MAX_LINK_RATE, DOWN_STREAM_PORT_PRESENT.

Read from DPCD once for all 3 values as this is an expensive operation.
Spec mentions that all of address space 02200h through 0220Fh should
contain the right information however currently only 3 values can
differ.

There is no address space in the intel_dp->dpcd struct for addresses
02200h through 0220Fh, and since so much of the data is a identical,
simply overwrite the values stored in 00000h through 0000Fh with the
values that can be overwritten from addresses 02200h through 0220Fh.

This patch helps with backward compatibility for devices pre DP1.3.

v2: read only dpcd values which can be affected, remove incorrect check,
split into drm include changes into separate patch, commit message,
verbose debugging statements during overwrite.
v3: white space fixes
v4: make path dependent on DPCD revision > 1.2
v5: split into function, removed DPCD rev check
v6: add debugging prints for early exit conditions
v7 (From Manasi):
* Memcpy, memcmp and debig logging based on sizeof(dpcd_ext) (Jani N)
* Exit early (Jani N)
v8 (From Manasi):
* Get rid of superfluous debug prints (Jani N)
* Print entire base DPCD before memcpy (Jani N)
v9 (From Manasi):
* Add uniform newlines (Rodrigo)

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Signed-off-by: Matt Atwood <matthew.s.atwood@intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Tested-by: Manasi Navare <manasi.d.navare@intel.com>
Acked-by: Manasi Navare <manasi.d.navare@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181129220058.19636-1-manasi.d.navare@intel.com

Makefile: fix mixed tab and spaces

Fixes a comestic issue where spaces were being used
instead of mixed tab and spaces.

Signed-off-by: Pedro Tammela <pctammela@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181205110608.5513-1-pctammela@gmail.com

drm/i915/psr: Check if source supports sink specific SU granularity

According to eDP spec, sink can required specific selective update
granularity that source must comply.
Here caching the value if required and checking if source supports
it.

v3:
- Returning the default granularity in case DPCD read fails(Dhinakaran)
- Changed DPCD error message level(Dhinakaran)

v4:
- Setting granularity to defaul when granularity read is equal to
0(Dhinakaran)

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-9-jose.souza@intel.com

drm/i915/psr: Check if resolution is supported by default SU granularity

Selective updates have a default granularity requirements as stated
by eDP spec(PSR2 SELECTIVE UPDATE X GRANULARITY CAPABILITY register
definition), so check if HW can match those requirements before
enabling PSR2.

v3:
- Changes in the comments and commit message(Dhinakaran)
- Printing the hdisplay that do not match with default granularity

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-8-jose.souza@intel.com

drm: Add the PSR SU granularity registers offsets

Source is required to comply to sink SU granularity when
DP_PSR2_SU_GRANULARITY_REQUIRED is set in DP_PSR_CAPS,
so adding the registers offsets.

v2: Also adding DP_PSR2_SU_Y_GRANULARITY(Rodrigo)

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-7-jose.souza@intel.com

drm/i915: Remove old PSR2 FIXME about frontbuffer tracking

Our frontbuffer tracking improved over the years + the WA #0884
helped us keep PSR2 enabled while triggering screen updates when
necessary so this FIXME is not valid anymore.

Acked-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-6-jose.souza@intel.com

drm/i915/icl: Do not change reserved registers related to PSR2

For ICL the bit 12 of CHICKEN_TRANS is reserved so we should not
touch it and as by default VSC_DATA_SEL_SOFTWARE_CONTROL is already
unset in gen10 + GLK we can just drop it and fix for both gens.

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-5-jose.souza@intel.com

drm/i915/psr: Enable sink to trigger a interruption on PSR2 CRC mismatch

eDP spec states 2 different bits to enable sink to trigger a
interruption when there is a CRC mismatch.
DP_PSR_CRC_VERIFICATION is for PSR only and
DP_PSR_IRQ_HPD_WITH_CRC_ERRORS is for PSR2 only.

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-4-jose.souza@intel.com

drm/i915/psr: Set PSR CRC verification bit in sink inside PSR1 block

As we have a else block for the 'if (dev_priv->psr.psr2_enabled) {'
and this bit is only set for PSR1 move it to that block to make it
more easy to read.

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-3-jose.souza@intel.com

drm/i915/psr: Don't tell sink that main link will be active while is active PSR2

For PSR2 there is no register to tell HW to keep main link enabled
while PSR2 is active, so don't configure sink DPCD with a
misleading value.

v2: Moving the set of DP_PSR_CRC_VERIFICATION to the else block
of 'if (dev_priv->psr.psr2_enabled)' to another patch. (Rodrigo)

Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-2-jose.souza@intel.com

drm/i915: Disable PSR in Apple panels

i915 yet don't support PSR in Apple panels, so lets keep it disabled
while we work on that.

v2: Renamed DP_DPCD_QUIRK_PSR_NOT_CURRENTLY_SUPPORTED to
DP_DPCD_QUIRK_NO_PSR (Ville)

v3:
Adding documentation to DP_DPCD_QUIRK_NO_PSR(Dhinakaran and Jani)
Fixed typo in comment of the new quirk entry(Jani)

Fixes: 598c6cfe0690 (drm/i915/psr: Enable PSR1 on gen-9+ HW)
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204003403.23361-1-jose.souza@intel.com

drm/i915: Dial down workaround debug messages

For selftests like live_hangcheck, which does thousands of reset per
second, it is too verbose to log every time we re-apply the workarounds.

Remove those messages since, coupled with the dedicated selftest, it is
sufficient to only log the number of recorded workarounds at init time.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204171530.28583-1-tvrtko.ursulin@linux.intel.com

drm/i915/selftests: Reorder request allocation vs vma pinning

Impose a restraint that we have all vma pinned for a request prior to
its allocation. This is to simplify request construction, and should
facilitate unravelling the lock interdependencies later.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204141522.13640-3-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20181204

Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/icl: fix transcoder state readout

Commit 2ca711caeca2 ("drm/i915/icl: Consider DSI for getting transcoder
state") clobbers the previously read TRANS_DDI_FUNC_CTL_EDP register
contents with TRANS_DDI_FUNC_CTL_DSI0 contents. Fix the state readout,
and handle DSI 1 while at it.

Use a bitmask for iterating and logging transcoders, because the allowed
combinations are a bit funky.

Fixes: 2ca711caeca2 ("drm/i915/icl: Consider DSI for getting transcoder state")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=108928
Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Cc: Madhav Chauhan <madhav.chauhan@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181204101926.17174-1-jani.nikula@intel.com