drm/i915/gt: Drop stale commentary for timeline density

We no longer allocate a contiguous set of timeline ids for all engines
upon creation, so we no longer should assume that the timelines are
densely allocated within a context. Hopefully, the set of fences used
within a workload are still dense enough for us to take advantage of
the compressed radix tree used for the syncmap.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625233349.32371-1-chris@chris-wilson.co.uk

drm/i915/selftests: Hold ref on request across waits

As we wait upon the request, we should be sure to hold our own reference
for our checks.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625130128.11009-14-chris@chris-wilson.co.uk

drm/i915: Rename intel_wakeref_[is]_active

Our general rule is to use is/has as the verb for boolean functions,
rename intel_wakeref_active to intel_wakeref_is_active so the question
being asked is clear.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625130128.11009-6-chris@chris-wilson.co.uk

drm/i915/gt: Pass intel_gt to pm routines

Switch from passing the i915 container to newly named struct intel_gt.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625130128.11009-2-chris@chris-wilson.co.uk

drm/i915/guc: Add debug capture of GuC exception

Detect GuC firmware load failure due to an exception during execution
in GuC firmware. Output the GuC EIP where exception occurred to dmesg
for GuC debug information.

v2: correct typos, change debug message and error code returned for
GuC exception (Michal)

Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625164107.21512-1-robert.m.fosha@intel.com

drm/i915/execlists: Convert recursive defer_request() into iterative

As this engine owns the lock around rq->sched.link (for those waiters
submitted to this engine), we can use that link as an element in a local
list. We can thus replace the recursive algorithm with an iterative walk
over the ordered list of waiters.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625130128.11009-1-chris@chris-wilson.co.uk

drm/i915/ehl: Add missing VECS engine

EHL can have up to one VECS(video enhancement) engine, so add it to
the device_info.

BSpec: 29152
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Bob Paauwe <bob.j.paauwe@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190614213749.15870-1-jose.souza@intel.com

drm/i915/ehl: Add one additional PCH ID to MCC

There's one additional ID that we should treat as Mule Creek Canyon.

Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621151847.31302-1-matthew.d.roper@intel.com
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>

drm/i915/perf: fix ICL perf register offsets

We got the wrong offsets (could they have changed?). New values were
computed off an error state by looking up the register offset in the
context image as written by the HW.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: 1de401c08fa805 ("drm/i915/perf: enable perf support on ICL")
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20190610081914.25428-1-lionel.g.landwerlin@intel.com

drm/i915: Disable SAMPLER_STATE prefetching on all Gen11 steppings.

The Demand Prefetch workaround (binding table prefetching) only applies
to Icelake A0/B0.  But the Sampler Prefetch workaround needs to be
applied to all Gen11 steppings, according to a programming note in the
SARCHKMD documentation.

Using the Intel Gallium driver, I have seen intermittent failures in
the dEQP-GLES31.functional.copy_image.non_compressed.* tests.  After
applying this workaround, the tests reliably pass.

v2: Remove the overlap with a pre-production w/a

BSpec: 9663
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190625090655.19220-1-chris@chris-wilson.co.uk

drm/i915/gem: Clear read/write domains for GPU clear

Update the domains for the write via the GPU so that we do not
shortcircuit any set-domain clflush afterwards.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110978
Fixes: b2dbf8d982a4 ("drm/i915/blt: Remove recursive vma->lock")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190624141630.11015-1-chris@chris-wilson.co.uk

drm/i915/execlists: Always clear ring_pause if we do not submit

In the unlikely case (thank you CI!), we may find ourselves wanting to
issue a preemption but having no runnable requests left. In this case,
we set the semaphore before computing the preemption and so must unset
it before forgetting (or else we leave the machine busywaiting until the
next request comes along and so likely hang).

v2: Replace readback with only a wmb after asserting the semaphore

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/20190624092009.30189-1-chris@chris-wilson.co.uk

drm/i915/blt: Remove recursive vma->lock

As we have already plugged the w->dma into the reservation_object, and
have set ourselves up to automatically signal the request and w->dma on
completion, we do not need to export the rq->fence directly and just use
the w->dma fence.

This avoids having to take the reservation_lock inside the worker which
cross-release lockdep would complain about. :)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621215733.12070-1-chris@chris-wilson.co.uk

drm/i915: Local debug BUG_ON for intel_wakeref

Avoid pulling in i915_gem.h just so that we can use a conditional BUG_ON
for debugging.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-5-chris@chris-wilson.co.uk

drm/i915: Provide an i915_active.acquire callback

If we introduce a callback for i915_active that is only called the first
time we use the i915_active and is symmetrically paired with the
i915_active.retire callback, we can replace the open-coded and
non-atomic implementations -- which will be very fragile (i.e. broken)
upon removing the struct_mutex serialisation.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-4-chris@chris-wilson.co.uk

drm/i915: Throw away the active object retirement complexity

Remove the accumulated optimisations that we have for i915_vma_retire
and reduce it to the bare essential of tracking the active object
reference. This allows us to only use atomic operations, and so will be
able to avoid the struct_mutex requirement.

The principal loss here is the shrinker MRU bumping, so now if we have
to shrink, we will do so in much more random order and more likely to
try and shrink recently used objects. That is a nuisance, but shrinking
active objects is a second step we try to avoid and will always be a
system-wide performance issue.

The other loss is here is in the automatic pruning of the
reservation_object when idling. This is not as large an issue as upon
reservation_object introduction as now adding new fences into the object
replaces already signaled fences, keeping the array compact. But we do
lose the auto-expiration of stale fences and unused arrays. That may be
a noticeable problem for which we need to re-implement autopruning.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-3-chris@chris-wilson.co.uk

drm/i915: Track i915_active using debugobjects

Provide runtime asserts and tracking of i915_active via debugobjects.
For example, this should allow us to check that the i915_active is only
active when we expect it to be and is never freed too early.

One consequence is that, for simplicity, we no longer allow i915_active
to be on-stack which only affected the selftests.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-2-chris@chris-wilson.co.uk

drm/i915: Remove waiting & retiring from shrinker paths

i915_gem_wait_for_idle() and i915_retire_requests() introduce a
dependency on the timeline->mutex. This is problematic as we want to
later perform allocations underneath i915_active.mutex, forming a link
between the shrinker, the timeline and active mutexes. Nip this cycle in
the bud by removing the acquisition of the timeline mutex (i.e.
retiring) from inside the shrinker.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621183801.23252-1-chris@chris-wilson.co.uk

drm/i915/guc: handle GuC messages received with CTB disabled

There is a very small chance of triggering a log flush event when
enabling or disabling CT buffers. Events triggered while CT buffers
are disabled are logged in the SCRATCH_15 register using the same bits
used in the CT message payload. Since our communication channel with
GuC is turned off, we can save the message and handle it after we turn
it back on.
GuC should be idle and not generate more events in the meantime because
we're not talking to it.

v2: clear the mmio register on stop_communication as well (Chris)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: 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/20190621182123.31368-2-daniele.ceraolospurio@intel.com

drm/i915/guc: reorder enable/disable communication steps

Make sure we always have CT buffers enabled when the interrupts are
enabled, so we can always handle interrupts from GuC. Also move the
setting of the guc->send and guc->handler functions to the GuC
communication control functions for consistency.

The reorder also fixes the onion unwinding of intel_uc_init_hw, because
guc_enable_communication would've left interrupts enabled when failing
to enable CTB.

v2: always retunr the result of ctch_enable() in
    intel_guc_ct_enable() (Michal)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110943
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Michal Wajdeczko <michal.wajdeczko@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621182123.31368-1-daniele.ceraolospurio@intel.com

drm/i915/gt: Fixup kerneldoc parameters

drivers/gpu/drm/i915/gt/intel_mocs.c:513: warning: Function parameter or member 'gt' not described in 'intel_mocs_init_l3cc_table'
 drivers/gpu/drm/i915/gt/intel_mocs.c:513: warning: Excess function parameter 'dev_priv' description in 'intel_mocs_init_l3cc_table'

intel_vgt_balloon/deballoon, i915_ggtt_probe_hw intel_wopcm_init_hw need
similar treatment

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/20190621131640.28864-2-chris@chris-wilson.co.uk

drm/i915/gt: Rename i915_gt_timelines

Since the anonymous i915_gt became struct intel_gt and encloses
struct i915_gt_timelines, rename i915_gt_timelines to intel_gt_timelines
to match its parentage.

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/20190621131640.28864-1-chris@chris-wilson.co.uk

drm/i915: Prevent dereference of engine before NULL check in error capture

smatch caught,
drivers/gpu/drm/i915/i915_gpu_error.c:1418 gem_record_rings() warn: variable dereferenced before check 'engine' (see line 1413)

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/20190621135246.20683-1-chris@chris-wilson.co.uk

drm/i915: Eliminate dual personality of i915_scratch_offset

Scratch vma lives under gt but the API used to work on i915. Make this
consistent by renaming the function to intel_gt_scratch_offset and make
it take struct intel_gt.

v2:
 * Move to intel_gt. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-33-tvrtko.ursulin@linux.intel.com

drm/i915: Rename i915_timeline to intel_timeline and move under gt

Move all timeline code under gt and rename to intel_gt prefix.

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>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-32-tvrtko.ursulin@linux.intel.com

drm/i915: Make timelines gt centric

Our timelines are stored inside intel_gt so we can convert the interface
to take exactly that and not i915.

At the same time re-order the params to our more typical layout and
replace the backpointer to the new containing structure.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-31-tvrtko.ursulin@linux.intel.com

drm/i915: Save trip via top-level i915 in a few more places

For gt related operations it makes more logical sense to stay in the realm
of gt instead of dereferencing via driver i915.

This patch handles a few of the easy ones with work requiring more
refactoring still outstanding.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-30-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize ring buffer creation

Continuing on the theme of compartmentalizing the code better to make
future split between gt and display in global i915 clearer.

v2:
 * Pass in ggtt instead of gt. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-29-tvrtko.ursulin@linux.intel.com

drm/i915: Store ggtt pointer in intel_gt

This will become useful in the following patch.

v2:
 * Assign the pointer through a helper on the top level to work around
   the layering violation. (Chris)

v3:
 * Handle selftests.

v4:
 * Move call to intel_gt_init_hw into mock_init_ggtt. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-28-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize i915_gem_init_ggtt

Continuing on the theme of better logical organization of our code, make
the first step towards making the ggtt code better isolated from wider
struct drm_i915_private.

v2:
 * Bring the ickle onion unwind back. (Chris)
 * Rename to i915_init_ggtt. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-27-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize i915_ggtt_cleanup_hw

Continuing on the theme of better logical organization of our code, make
the first step towards making the ggtt code better isolated from wider
struct drm_i915_private.

v2:
 * Cleanup of mm.wc_stash does not need struct_mutex. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-26-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize timeline_init/park/fini

Continuing on the theme of better logical organization of our code, make
the first step towards making the timeline code better isolated from wider
struct drm_i915_private.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-25-tvrtko.ursulin@linux.intel.com

drm/i915: Move i915_gem_chipset_flush to intel_gt

This aligns better with the rest of restructuring.

v2:
 * Move call out of line. (Chris)

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>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-24-tvrtko.ursulin@linux.intel.com

drm/i915: Convert i915_gem_flush_ggtt_writes to intel_gt

Having introduced struct intel_gt (named the anonymous structure in i915)
we can start using it to compartmentalize our code better. It makes more
sense logically to have the code internally like this and it will also
help with future split between gt and display in i915.

v2:
 * Keep ggtt flush before fb obj flush. (Chris)

v3:
 * Fix refactoring fail.
 * Always flush ggtt writes. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-23-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize i915_gem_suspend/restore_gtt_mappings

Having made start to better code compartmentalization by introducing
struct intel_gt, continue the theme elsewhere in code by making functions
take parameters take what logically makes most sense for them instead of
the global struct drm_i915_private.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-22-tvrtko.ursulin@linux.intel.com

drm/i915: Store intel_gt backpointer in vm

This will come useful in the following patch.

v2:
 * Handle mock ggtt.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-21-tvrtko.ursulin@linux.intel.com

drm/i915: Make ggtt invalidation work on ggtt

It is more logical for ggtt invalidation to take ggtt as input parameter.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-20-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize i915_ggtt_init_hw

Having made start to better code compartmentalization by introducing
struct intel_gt, continue the theme elsewhere in code by making functions
take parameters take what logically makes most sense for them instead of
the global struct drm_i915_private.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-19-tvrtko.ursulin@linux.intel.com

drm/i915: Compartmentalize i915_ggtt_probe_hw

Having made start to better code compartmentalization by introducing
struct intel_gt, continue the theme elsewhere in code by making functions
take parameters take what logically makes most sense for them instead of
the global struct drm_i915_private.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-18-tvrtko.ursulin@linux.intel.com

drm/i915: Stop using I915_READ/WRITE in intel_wopcm_init_hw

More legacy mmio accessor removal. We pass in intel_gt explicitly allowing
code to use new intel_uncore_read/write helpers.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-17-tvrtko.ursulin@linux.intel.com

drm/i915: Move intel_engines_resume into common init

Since this part still operates on i915 and not intel_gt, move it to the
common (top-level) function.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-16-tvrtko.ursulin@linux.intel.com

drm/i915: Convert i915_gem_init_hw to intel_gt

More removal of implicit dev_priv from using old mmio accessors.

Actually the top level function remains but is split into a part which
writes to i915 and part which operates on intel_gt in order to initialize
the hardware.

GuC and engines are the only odd ones out remaining.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-15-tvrtko.ursulin@linux.intel.com

drm/i915: Consolidate some open coded mmio rmw

Replace some gen6/7 open coded rmw with intel_uncore_rmw.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-14-tvrtko.ursulin@linux.intel.com

drm/i915: Convert i915_ppgtt_init_hw to intel_gt

More removal of implicit dev_priv from using old mmio accessors.

v2:
 * Rebase for uncore_to_i915 removal.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-13-tvrtko.ursulin@linux.intel.com

drm/i915: Convert intel_mocs_init_l3cc_table to intel_gt

More removal of implicit dev_priv from using old mmio accessors.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-12-tvrtko.ursulin@linux.intel.com

drm/i915: Store backpointer to intel_gt in the engine

It will come useful in the next patch.

v2:
 * Do mock_engine as well.

v3:
 * And the virtual engine...

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-11-tvrtko.ursulin@linux.intel.com

drm/i915: Convert gt workarounds to intel_gt

More conversion of i915_gem_init_hw to uncore.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-10-tvrtko.ursulin@linux.intel.com

drm/i915: Convert init_unused_rings to intel_gt

More removal of implicit dev_priv from using old mmio accessors.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-9-tvrtko.ursulin@linux.intel.com

drm/i915: Use intel_uncore_rmw in intel_gt_init_swizzling

Two easy opportunities to compact the code by using the existing helper.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Suggested-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-8-tvrtko.ursulin@linux.intel.com

drm/i915: Convert i915_gem_init_swizzling to intel_gt

Start using the newly introduced struct intel_gt to fuse together correct
logical init flow with uncore for more removal of implicit dev_priv in
mmio access.

v2:
 * Move code to i915_gem_fence_reg. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-7-tvrtko.ursulin@linux.intel.com

drm/i915: Make i915_check_and_clear_faults take intel_gt

Continuing the conversion and elimination of implicit dev_priv.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Suggested-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-6-tvrtko.ursulin@linux.intel.com

drm/i915: Move intel_gt_pm_init under intel_gt_init_early

And also rename to intel_gt_pm_init_early and make it operate on gt.

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>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-5-tvrtko.ursulin@linux.intel.com

drm/i915: Store some backpointers in struct intel_gt

We need an easy way to get back to i915 and uncore.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-4-tvrtko.ursulin@linux.intel.com

drm/i915: Move intel_gt initialization to a separate file

As it will grow in a following patch make a new home for it.

v2:
 * Convert mock_gem_device as well. (Chris)

v3:
 * Rename to intel_gt_init_early and move call site to i915_drv.c. (Chris)

v4:
 * Adjust SPDX tags.
 * No need to gt/ path when including intel_gt_types.h. (Chris)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-3-tvrtko.ursulin@linux.intel.com

drm/i915: Introduce struct intel_gt as replacement for anonymous i915->gt

We have long been slighlty annoyed by the anonymous i915->gt.

Promote it to a separate structure and give it its own header.

This is a first step towards cleaning up the separation between
i915 and gt.

v2:
 * Adjust SPDX header.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-2-tvrtko.ursulin@linux.intel.com

drm/i915: Convert intel_vgt_(de)balloon to uncore

More removal of implicit dev_priv from using old mmio accessors.

Furthermore these calls really operate on ggtt so it logically makes sense
if they take it as parameter.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621070811.7006-1-tvrtko.ursulin@linux.intel.com

drm/i915/execlists: Keep virtual context alive until after we kick

The call to kick_siblings() dereferences the rq->context, so we should
not drop our local reference until afterwards!

v2: Stick to setting ce.inflight=NULL before kicking as this is what the
other threads will check to see if the context is ready for takeover.

Fixes: 22b7a426bbe1 ("drm/i915/execlists: Preempt-to-busy")
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>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190621080729.2652-1-chris@chris-wilson.co.uk

drm/i915/gtt: Defer address space cleanup to an RCU worker

Enable RCU protection of i915_address_space and its ppgtt superclasses,
and defer its cleanup into a worker executed after an RCU grace period.

In the future we will be able to use the RCU protection to reduce the
locking around VM lookups, but the immediate benefit is being able to
defer the release into a kworker (process context). This is required as
we may need to sleep to reap the WC pages stashed away inside the ppgtt.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110934
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/20190620183705.31006-1-chris@chris-wilson.co.uk

drm/i915/ehl/dsi: Enable AFE over PPI strap

The other additional step in the DSI sequence for EHL.

v2:
- Using REG_BIT()(Matt)
- Fixed commit message typo(Vandita)

BSpec: 20597
Cc: Uma Shankar <uma.shankar@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190619233134.20009-2-jose.souza@intel.com

drm/i915/ehl/dsi: Set lane latency optimization for DW1

EHL has 2 additional steps in the DSI sequence, this is one of then
the lane latency optimization for DW1.

BSpec: 20597
Cc: Uma Shankar <uma.shankar@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Vandita Kulkarni <vandita.kulkarni@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190619233134.20009-1-jose.souza@intel.com

drm/i915/selftests: Use request managed wakerefs

Since commit 79ffac8599c4 ("drm/i915: Invert the GEM wakeref
hierarchy"), the request creation itself took responsibility for
managing the engine/GT wakerefs and so we can remove the redundant grabs
in our selftests.

References: 79ffac8599c4 ("drm/i915: Invert the GEM wakeref hierarchy")
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/20190620102432.31580-1-chris@chris-wilson.co.uk

drm/i915: Rings are always flushed

Our intel_rings are always flushed as they are continually used to submit
commands to the GPU, and so do not need to be flushed on unpinning. This
avoids pulling in the flush_ggtt_writes locking into our context
unpin, which we want to allow from atomic context (for simplicity).

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/20190619203504.4220-1-chris@chris-wilson.co.uk

drm/i915/execlists: Minimalistic timeslicing

If we have multiple contexts of equal priority pending execution,
activate a timer to demote the currently executing context in favour of
the next in the queue when that timeslice expires. This enforces
fairness between contexts (so long as they allow preemption -- forced
preemption, in the future, will kick those who do not obey) and allows
us to avoid userspace blocking forward progress with e.g. unbounded
MI_SEMAPHORE_WAIT.

For the starting point here, we use the jiffie as our timeslice so that
we should be reasonably efficient wrt frequent CPU wakeups.

Testcase: igt/gem_exec_scheduler/semaphore-resolve
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/20190620142052.19311-2-chris@chris-wilson.co.uk

drm/i915/execlists: Preempt-to-busy

When using a global seqno, we required a precise stop-the-workd event to
handle preemption and unwind the global seqno counter. To accomplish
this, we would preempt to a special out-of-band context and wait for the
machine to report that it was idle. Given an idle machine, we could very
precisely see which requests had completed and which we needed to feed
back into the run queue.

However, now that we have scrapped the global seqno, we no longer need
to precisely unwind the global counter and only track requests by their
per-context seqno. This allows us to loosely unwind inflight requests
while scheduling a preemption, with the enormous caveat that the
requests we put back on the run queue are still _inflight_ (until the
preemption request is complete). This makes request tracking much more
messy, as at any point then we can see a completed request that we
believe is not currently scheduled for execution. We also have to be
careful not to rewind RING_TAIL past RING_HEAD on preempting to the
running context, and for this we use a semaphore to prevent completion
of the request before continuing.

To accomplish this feat, we change how we track requests scheduled to
the HW. Instead of appending our requests onto a single list as we
submit, we track each submission to ELSP as its own block. Then upon
receiving the CS preemption event, we promote the pending block to the
inflight block (discarding what was previously being tracked). As normal
CS completion events arrive, we then remove stale entries from the
inflight tracker.

v2: Be a tinge paranoid and ensure we flush the write into the HWS page
for the GPU semaphore to pick in a timely fashion.

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/20190620142052.19311-1-chris@chris-wilson.co.uk

drm/i915/gvt: decouple check_vgpu() from uncore_init()

With multiple uncore to initialize (GT vs Display), it makes little
sense to have the vgpu_check inside uncore_init(). We also have
a catch-22 scenario where the uncore is required to read the vgpu
capabilities while the vgpu capabilities are required to decide if
we need to initialize forcewake support. To remove this circular
dependency, we can perform the required MMIO access by mmapping just
the vgtif shared page in mmio space and use raw accessors.

v2: rename check_vgpu to detect_vgpu (Chris)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-7-daniele.ceraolospurio@intel.com

drm/i915: dynamically allocate forcewake domains

We'd like to introduce a display uncore with no forcewake domains, so
let's avoid wasting memory and be ready to allocate only what we need.
Even without multiple uncore, we still don't need all the domains on all
gens.

v2: avoid hidden control flow, improve checks (Tvrtko), fix IVB special
case, add failure injection point

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-6-daniele.ceraolospurio@intel.com

drm/i915: skip forcewake actions on forcewake-less uncore

We always call some of the setup/cleanup functions for forcewake, even
if the feature is not actually available. Skipping these operations if
forcewake is not available saves us some operations on older gens and
prepares us for having a forcewake-less display uncore.

v2: do not make suspend/resume functions forcewake-specific (Chris,
Tvrtko), use GEM_BUG_ON in internal forcewake-only functions (Tvrtko)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-5-daniele.ceraolospurio@intel.com

drm/i915: kill uncore_to_i915

Let's get rid of it before it proliferates, since with split GT/Display
uncores the container_of won't work anymore.

I've kept the rpm pointer as well to minimize the pointer chasing in the
MMIO accessors.

v2: swap parameter order for intel_uncore_init_early (Tvrtko)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-4-daniele.ceraolospurio@intel.com

drm/i915: kill uncore_sanitize

uncore_sanitize performs no action on the uncore structure and just
calls intel_sanitize_gt_powersave, so we can just call the latter
directly.

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-3-daniele.ceraolospurio@intel.com

drm/i915: use vfuncs for reg_read/write_fw_domains

Instead of going through the if-else chain every time, let's save the
function in the uncore structure. Note that the new functions are
purposely not used from the reg read/write functions to keep the
inlining there.

While at it, use the new macro to call the old ones to clean the code a
bit.

v2: Rename macros for no-forcewake function assignment (Tvrtko)

Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190620010021.20637-2-daniele.ceraolospurio@intel.com

drm/i915: Keep rings pinned while the context is active

Remember to keep the rings pinned as well as the context image until the
GPU is no longer active.

v2: Introduce a ring->pin_count primarily to hide the
mock_ring that doesn't fit into the normal GGTT vma picture.

v3: Order is important in teardown, ringbuffer submission needs to drop
the pin count on the engine->kernel_context before it can gleefully free
its ring.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110946
Fixes: ce476c80b8bf ("drm/i915: Keep contexts pinned until after the next kernel context switch")
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>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190619170135.15281-1-chris@chris-wilson.co.uk

drm/i915/ehl: Allow combo PHY A to drive a third external display

EHL has a mux on combo PHY A that allows it to be driven either by an
internal display (DDI-A or DSI DPHY) or by an external display (DDI-D).
This is a motherboard design decision that can not be changed on the
fly.  Unfortunately there are no strap registers that allow us to detect
the board configuration directly, so let's use the VBT to try to figure
it out and program the mux accordingly.

For now if we run across a broken VBT that tries to claim that PHY A
is attached to both internal and external displays at the same time,
we'll resolve the conflict in favor of the internal display.  To help
debug these kind of bad VBT's, let's also add a quick DRM_DEBUG message
during child device parsing so that it's easier to understand these
cases if they show up in bug reports.

v2:
 - Confirmed that VBT's dvo port refers to the DDI and not the PHY.
   Thus we can check more explicitly for (ddi_d && !(ddi_a || dsi)).  If
   a bad VBT contradicts itself, let internal display win.  (Ville)

v3:
 - Switch condition from !IS_ICELAKE to IS_ELKHARTLAKE.  Although the
   convention is usually to assume that future platforms will inherit
   all current platform behavior, this feels more like a one-platform
   quirk.  (Ville)
 - Update commit message to describe what we do if/when we encounter
   broken VBT's, and note that the new debug print during child device
   parsing is intentional.

Cc: Clint Taylor <Clinton.A.Taylor@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
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/20190618175131.9139-1-matthew.d.roper@intel.com

drm/i915: Flush the execution-callbacks on retiring

In the unlikely case the request completes while we regard it as not even
executing on the GPU (see the next patch!), we have to flush any pending
execution callbacks at retirement and ensure that we do not add any
more.

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/20190618074153.16055-4-chris@chris-wilson.co.uk

drm/i915: Signal fence completion from i915_request_wait

With the upcoming change to automanaged i915_active, the intent is that
whenever we wait on the set of active fences, they are signaled and
collected.  The requirement is that all successful returns from
i915_request_wait() signal the fence, so fixup the one remaining path
where we may return before the interrupt has been run.

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/20190619112341.9082-3-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20190619

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

drm/i915: Stop passing I915_WAIT_LOCKED to i915_request_wait()

Since commit eb8d0f5af4ec ("drm/i915: Remove GPU reset dependence on
struct_mutex"), the I915_WAIT_LOCKED flags passed to i915_request_wait()
has been defunct. Now go ahead and remove it from all callers.

References: eb8d0f5af4ec ("drm/i915: Remove GPU reset dependence on struct_mutex")
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/20190618074153.16055-3-chris@chris-wilson.co.uk

drm/i915/execlists: Detect cross-contamination with GuC

The process_csb routine from execlists_submission is incompatible with
the GuC backend. Add a warning to detect if we accidentally end up in
the wrong spot.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618110736.31155-1-chris@chris-wilson.co.uk

drm/i915: Make the semaphore saturation mask global

The idea behind keeping the saturation mask local to a context backfired
spectacularly. The premise with the local mask was that we would be more
proactive in attempting to use semaphores after each time the context
idled, and that all new contexts would attempt to use semaphores
ignoring the current state of the system. This turns out to be horribly
optimistic. If the system state is still oversaturated and the existing
workloads have all stopped using semaphores, the new workloads would
attempt to use semaphores and be deprioritised behind real work. The
new contexts would not switch off using semaphores until their initial
batch of low priority work had completed. Given sufficient backload load
of equal user priority, this would completely starve the new work of any
GPU time.

To compensate, remove the local tracking in favour of keeping it as
global state on the engine -- once the system is saturated and
semaphores are disabled, everyone stops attempting to use semaphores
until the system is idle again. One of the reason for preferring local
context tracking was that it worked with virtual engines, so for
switching to global state we could either do a complete check of all the
virtual siblings or simply disable semaphores for those requests. This
takes the simpler approach of disabling semaphores on virtual engines.

The downside is that the decision that the engine is saturated is a
local measure -- we are only checking whether or not this context was
scheduled in a timely fashion, it may be legitimately delayed due to user
priorities. We still have the same dilemma though, that we do not want
to employ the semaphore poll unless it will be used.

v2: Explain why we need to assume the worst wrt virtual engines.

Fixes: ca6e56f654e7 ("drm/i915: Disable semaphore busywaits on saturated systems")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
Cc: Dmitry Ermilov <dmitry.ermilov@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618074153.16055-8-chris@chris-wilson.co.uk

drm/i915/psr: Force manual PSR exit in older gens

To do frontbuffer tracking we are depending on Display WA #0884 to
exit PSR when there is a frontbuffer modification but according to
user reports a write to CURSURFLIVE do not cause PSR to exit in older
gens so lets force a PSR exit.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110799
Cc: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Tested-by: Thomas Rohwer <trohwer85@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190617195154.30292-1-jose.souza@intel.com

drm/i915: Don't dereference request if it may have been retired when printing

This has caught me out on countless occasions, when we retrieve a pointer
from the submission/execlists backend, it does not carry a reference to
the context or ring. Those are only pinned while the request is active,
so if we see the request is already completed, it may be in the process
of being retired and those pointers defunct.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110938
Fixes: 3a068721a973 ("drm/i915: Show ring->start for the ELSP context/request queue")
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/20190618161951.28820-2-chris@chris-wilson.co.uk

drm/i915/selftests: Flush live_evict

Be sure to cleanup after live_evict by flushing any residual state off
the GPU using igt_flush_test.

Tvrtko mentioned that it is probably wise to stop repeating this ad hoc
around the tests and implement a live test runner.

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/20190618161951.28820-1-chris@chris-wilson.co.uk

drm/i915: Skip shrinking already freed pages

Previously, we wanted to shrink the pages of freed objects before they
were finally RCU collected. However, by removing the struct_mutex
serialisation around the active reference, we need to acquire an extra
reference around the wait. Unfortunately this means that we have to skip
objects that are waiting RCU collection.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110937
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/20190618074153.16055-2-chris@chris-wilson.co.uk

drm/i915: Update workarounds selftest for read only regs

Updates the live_workarounds selftest to handle whitelisted
registers that are flagged as read only.

Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com>
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618010108.27499-5-John.C.Harrison@Intel.com

drm/i915: Add whitelist workarounds for ICL

Updated whitelist table for ICL.

v2: Reduce changes to just those required for media driver until
the selftest can be updated to support the new features of the
other entries.

Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618010108.27499-4-John.C.Harrison@Intel.com

drm/i915: Support whitelist workarounds on all engines

Newer hardware requires setting up whitelists on engines other than
render. So, extend the whitelist code to support all engines.

Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618010108.27499-3-John.C.Harrison@Intel.com

drm/i915: Support flags in whitlist WAs

Newer hardware adds flags to the whitelist work-around register. These
allow per access direction privileges and ranges.

Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Robert M. Fosha <robert.m.fosha@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618010108.27499-2-John.C.Harrison@Intel.com

drm/i915: Drop the _INCOMPLETE for has_infoframe

We have full infoframe readout now so we can replace the
PIPE_CONF_CHECK_BOOL_INCOMPLETE(has_infoframe) with the normal
PIPE_CONF_CHECK_BOOL(has_infoframe).

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190612130801.2085-4-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Make pipe_config_err() vs. fastset less confusing

Rename pipe_config_err() to pipe_config_mismatch(), and also print
whether we're doing the fastset check or the sw vs. hw state readout
check. Should make the logs a bit less confusing when they're not
filled with what looks like a real error.

Also rename the 'adjust' variable to 'fastset' to make it clear what
it means.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190612130801.2085-3-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Constify intel_pipe_config_compare()

Now that intel_pipe_config_compare() no longer clobbers the passed
in state we can make both crtc states const. And while at we simplify
the calling convention, and clean up intel_compare_link_m_n() a bit.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190612130801.2085-2-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Don't clobber M/N values during fastset check

We're now calling intel_pipe_config_compare(..., true) uncoditionally
which means we're always going clobber the calculated M/N values with
the old values if the fuzzy M/N check passes. That causes problems
because the fuzzy check allows for a huge difference in the values.

I'm actually tempted to just make the M/N checks exact, but that might
prevent fastboot from kicking in when people want it. So for now let's
overwrite the computed values with the old values only if decide to skip
the modeset.

v2: Copy has_drrs along with M/N M2/N2 values

Cc: stable@vger.kernel.org
Cc: Blubberbub@protonmail.com
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Hans de Goede <hdegoede@redhat.com>
Tested-by: Blubberbub@protonmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110782
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110675
Fixes: d19f958db23c ("drm/i915: Enable fastset for non-boot modesets.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190612172423.25231-1-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

drm/i915: Use drm_gem_object.resv

Since commit 1ba627148ef5 ("drm: Add reservation_object to
drm_gem_object"), struct drm_gem_object grew its own builtin
reservation_object rendering our own private one bloat. Remove our
redundant reservation_object and point into obj->base.resv instead.

References: 1ba627148ef5 ("drm: Add reservation_object to drm_gem_object")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618125858.7295-1-chris@chris-wilson.co.uk

drm/i915: Keep engine alive as we retire the context

Though we pin the context first before taking the pm wakeref, during
retire we need to unpin before dropping the pm wakeref (breaking the
"natural" onion). During the unpin, we may need to attach a cleanup
operation on to the engine wakeref, ergo we want to keep the engine
awake until after the unpin.

v2: Push the engine wakeref into the barrier so we keep the onion unwind
ordering in the request itself

Fixes: ce476c80b8bf ("drm/i915: Keep contexts pinned until after the next kernel context switch")
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/20190618074153.16055-1-chris@chris-wilson.co.uk

drm/i915/guc: Reduce verbosity on log overflows

If the user is clearing the log buffer too slowly, we overflow. As this
is an expected condition, and the driver tries to handle it, reduce the
error message down to a notice.

Michal mentioned that another cause would be incorrect reset handling,
so we don't want to lose the notification entirely.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=110817
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190617100917.13110-1-chris@chris-wilson.co.uk

Documentation/i915: fix file references after display/ subdir renames

Fix the plethora of Sphinx build errors after moving the display files
under a subdirectory.

Fixes: 379bc100232a ("drm/i915: move modesetting output/encoder code under display/")
Fixes: df0566a641f9 ("drm/i915: move modesetting core code under display/")
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190617102944.25129-1-jani.nikula@intel.com

drm/i915/ehl: Introduce Mule Creek Canyon PCH

Although EHL introduces a new PCH, the South Display part of the PCH
that we care about is nearly identical to ICP, just with some pins
remapped.  Most notably, Port C is mapped to the pins that ICP uses for
TC Port 1.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190615004210.16656-1-matthew.d.roper@intel.com

drm/i915/gtt: Serialise both updates to PDE and our shadow

Currently, we perform a locked update of the shadow entry when
allocating a page directory entry such that if two clients are
concurrently allocating neighbouring ranges we only insert one new entry
for the pair of them. However, we also need to serialise both clients
wrt to the actual entry in the HW table, or else we may allow one client
or even a third client to proceed ahead of the HW write. My handwave
before was that under the _pathological_ condition we would see the
scratch entry instead of the expected entry, causing a temporary
glitch. That starvation condition will eventually show up in practice, so
fix it.

The reason for the previous cheat was to avoid having to free the extra
allocation while under the spinlock. Now, we keep the extra entry
allocated until the end instead.

v2: Fix error paths for gen6

Fixes: 1d1b5490b91c ("drm/i915/gtt: Replace struct_mutex serialisation for allocation")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190617140426.7203-1-chris@chris-wilson.co.uk

drm/i915/dmc: protect against loading wrong firmware

In intel_package_header version 2 there's a new field in the
fw_info table that must be 0, otherwise it's not the correct DMC
firmware. Add a check for version 2 or later.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190607091230.1489-10-lucas.demarchi@intel.com

drm/i915/dmc: remove redundant return in parse_csr_fw()

parse_csr_fw() is responsible to set up several fields in struct intel_csr,
including the payload. We don't need to assign it again.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190607091230.1489-9-lucas.demarchi@intel.com

drm/i915/dmc: add support to load dmc_header version 3

Main difference is that now there are up to 20 MMIOs that can be set and
a lot of noise due to the struct changing the fields in the middle.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190607091230.1489-8-lucas.demarchi@intel.com

drm/i915/dmc: extract function to parse dmc_header

Complete the extraction of functions to parse specific parts of the
firmware. The return of the function parse_csr_fw() is now redundant
since it already sets the dmc_payload field. Changing it is left for
later to avoid noise in the commit.

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190607091230.1489-7-lucas.demarchi@intel.com