drm/i915: Don't accidentally increase the frequency in handling DOWN rps

If we receive a DOWN_TIMEOUT rps interrupt, we respond by reducing the
GPU clocks significantly. Before we do, double check that the frequency
we pick is actually a decrease.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210150348.22146-2-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: Enable fine-tuned RPS for cherryview

When the RPS tuning was applied to Baytrail, in commit 8fb55197e64d
("drm/i915: Agressive downclocking on Baytrail"), concern was given that
it might cause Cherryview excess wakeups of the common power well.
However, the static thresholds perform poorly for Kodi, and the GPU is
unable to deliver the video frames on time. Enabling the dynamic, finer
thresholds used on all other platforms (including Skylake and Broxton
that also have the same multiple powerwell concerns) allows the GPU to
pick a more appropriate frequency and not drop frames.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210150348.22146-1-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: The return of i915_gpu_info to debugfs

Once upon a time before we had automated GPU state capture upon hangs,
we had intel_gpu_dump. Now we come almost full circle and reinstate that
view of the current GPU queues and registers by using the error capture
facility to snapshot the GPU state when debugfs/.../i915_gpu_info is
opened - which should provided useful debugging to both the error
capture routines (without having to cause a hang and avoid the error
state being eaten by igt) and generally.

v2: Rename drm_i915_error_state to i915_gpu_state to alleviate some name
collisions between the error state dump and inspecting the gpu state.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214164611.11381-1-chris@chris-wilson.co.uk

drm/i915/guc: Don't take struct_mutex for object unreference

We no longer need to take the struct_mutex for freeing objects, and on
the finalisation paths here the mutex is not been used for serialisation
of the pointer access, so remove the BKL wart.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214133420.7977-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Silence compiler warning for seltests/i915_gem_coherency

In general, the compiler should not be able to detect if we do any
passes through the test loops:

In file included from drivers/gpu/drm/i915/i915_gem.c:5029:
drivers/gpu/drm/i915/selftests/i915_gem_coherency.c: In function 'igt_gem_coherency':
drivers/gpu/drm/i915/selftests/i915_gem_coherency.c:274: error: 'err' may be used uninitialized in this function

Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214143509.15719-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Silence compiler for GTT selftests

gcc-4.7 spotted that

In file included from drivers/gpu/drm/i915/i915_gem_gtt.c:3791:0:
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c: In function ‘pot_hole’:
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c:594:6: error: ‘err’ may be used uninitialized in this function [-Werror=maybe-uninitialized]

So set it to 0 should we ever skip over a hole smaller than a few pages.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214113756.27834-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Avoid overflow in computing pot_hole loop termination

When using the mock_ppgtt selftest, the GTT is large enough to cause an
overflow in pot_hole() when adding 2 pages to the address. Avoid the
overflow by computing the final valid address and iterating up to that
address.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214092344.12330-1-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915/chv: Set min freq to RPn on CHV.

With latest Punit FW, vgg input voltag drop falling to minimum is fixed.
So reverting the WA patch & moving to turbo freq opreation range to [RPn -> RP0]

This is not a 1:1 revert of the commit 5b7c91b78b1ce6663e0f1f037f6cb4d7c9537d44.
You can refer to commit 5b5929cbe3f7 ("drm/i915/chv: remove
pre-production hardware workarounds") as the reason for the discrepancy

commit 5b7c91b78b1ce6663e0f1f037f6cb4d7c9537d44
Author: Deepak S <deepak.s@linux.intel.com>
Date:   Sat May 9 18:15:46 2015 +0530

    drm/i915/chv: Set min freq to efficient frequency on chv

v2: Fix inconsistent return type. (Chris)
v3: drop pre-production hw case (Ville)

Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1471007801-86075-1-git-send-email-deepak.s@linux.intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Dump more configuration information for DSI

Dump out more of the DSI configuration details during init.
This includes pclk, burst_mode_ratio, lane_count, pixel_overlap,
video_mode_format and reset_timer_val.

v2: Dump more info (Chris)
v3: Use the VIDEO_MODE_ defines for consistency (Chris)
    Dump dphy_reg too (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20161221143114.23530-1-ville.syrjala@linux.intel.com

drm/i915: fix for WaDisableDopClockGating:bdw

This workaround for BDW was incomplete as it also requires EUTC clock
gating to be disabled via UCGCTL1.

v2: read modify write UCGTL1 in broadwell_init_clock_gating (Ville)

Signed-off-by: Robert Bragg <robert@sixbynine.org>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212133252.20990-1-robert@sixbynine.org
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915: Fix uninitialized return from mi_set_context

For some reason my compiler (and CI as well) failed to spot the
uninitialized ret in mi_set_context.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 73dec95e6ba3 ("drm/i915: Emit to ringbuffer directly")
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214152901.20361-1-tvrtko.ursulin@linux.intel.com

drm/i915: Emit to ringbuffer directly

This removes the usage of intel_ring_emit in favour of
directly writing to the ring buffer.

intel_ring_emit was preventing the compiler for optimising
fetch and increment of the current ring buffer pointer and
therefore generating very verbose code for every write.

It had no useful purpose since all ringbuffer operations
are started and ended with intel_ring_begin and
intel_ring_advance respectively, with no bail out in the
middle possible, so it is fine to increment the tail in
intel_ring_begin and let the code manage the pointer
itself.

Useless instruction removal amounts to approximately
two and half kilobytes of saved text on my build.

Not sure if this has any measurable performance
implications but executing a ton of useless instructions
on fast paths cannot be good.

v2:
 * Change return from intel_ring_begin to error pointer by
   popular demand.
 * Move tail increment to intel_ring_advance to enable some
   error checking.

v3:
 * Move tail advance back into intel_ring_begin.
 * Rebase and tidy.

v4:
 * Complete rebase after a few months since v3.

v5:
 * Remove unecessary cast and fix !debug compile. (Chris Wilson)

v6:
 * Make intel_ring_offset take request as well.
 * Fix recording of request postfix plus a sprinkle of asserts.
   (Chris Wilson)

v7:
 * Use intel_ring_offset to get the postfix. (Chris Wilson)
 * Convert GVT code as well.

v8:
 * Rename *out++ to *cs++.

v9:
 * Fix GVT out to cs conversion in GVT.

v10:
 * Rebase for new intel_ring_begin in selftests.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Zhi Wang <zhi.a.wang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170214113242.29241-1-tvrtko.ursulin@linux.intel.com

drm/i915: Convert remaining users of 32bit power domain masks

I screwed up the rebase of commit d8fc70b7367b ("drm/i915: Make power
domain masks 64 bit long") before sending v2, causing a couple of
conversions from 32 to 64 bit masks to be lost.

Fixes: d8fc70b7367b ("drm/i915: Make power domain masks 64 bit long")
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213145733.8779-1-ander.conselvan.de.oliveira@intel.com

drm/i915: Pass timeout==0 on to i915_gem_object_wait_fence()

The i915_gem_object_wait_fence() uses an incoming timeout=0 to query
whether the current fence is busy or idle, without waiting. This can be
used by the wait-ioctl to implement a busy query.

Fixes: e95433c73a11 ("drm/i915: Rearrange i915_wait_request() accounting with callers")
Testcase: igt/gem_wait/basic-busy-write-all
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.10-rc1+
Cc: stable@vger.kernel.org
Link: http://patchwork.freedesktop.org/patch/msgid/20170212215344.16600-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/gvt: Disable access to stolen memory as a guest

Explicitly disable stolen memory when running as a guest in a virtual
machine, since the memory is not mediated between clients and reserved
entirely for the host. The actual size should be reported as zero, but
like every other quirk we want to tell the user what is happening.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99028
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161109103905.17860-1-chris@chris-wilson.co.uk
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: stable@vger.kernel.org

drm/i915: Add initial selftests for hang detection and resets

Check that we can reset the GPU and continue executing from the next
request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-47-chris@chris-wilson.co.uk

drm/i915: Exercise crossing pot boundaries in the GTT

As the page-table trees within the GTT are naturally aligned to
power-of-two boundaries, by inserting an object that crosses a
power-of-two (and the power-of-two intervals) we can quickly check the
code for errors in switching between levels in the tree.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-46-chris@chris-wilson.co.uk

drm/i915: Exercise manipulate of single pages in the GGTT

Move a single page of an object around within the GGTT and check
coherency of writes and reads.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-45-chris@chris-wilson.co.uk

drm/i915: Add mock tests for GTT/VMA handling

Use the live tests against the mock ppgtt for quick testing on all
platforms of the VMA layer.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-44-chris@chris-wilson.co.uk

drm/i915: Add mock exercise for i915_gem_gtt_insert

i915_gem_gtt_insert should allocate from the available free space in the
GTT, evicting as necessary to create space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-43-chris@chris-wilson.co.uk

drm/i915: Add mock exercise for i915_gem_gtt_reserve

i915_gem_gtt_reserve should put the node exactly as requested in the
GTT, evicting as required.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-42-chris@chris-wilson.co.uk

drm/i915: Initial selftests for exercising eviction

Very simple tests to just ask eviction to find some free space in a full
GTT and one with some available space.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-41-chris@chris-wilson.co.uk

drm/i915: Force an aliasing_ppgtt test for context execution

Ensure that we minimally exercise the aliasing_ppgtt, even on a
full-ppgtt, by allocating one and similarly creating a context to use
it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-40-chris@chris-wilson.co.uk

drm/i915: Extract aliasing ppgtt setup

In order to force testing of the aliasing ppgtt, extract its
initialisation function.

v2: Also extract the cleanup function for symmetry.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-39-chris@chris-wilson.co.uk

drm/i915: Live testing for context execution

Check we can create and execution within a context.

v2: Write one set of dwords through each context/engine to exercise more
contexts within the same time period.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-38-chris@chris-wilson.co.uk

drm/i915: Test creation of partial VMA

Mock testing to ensure we can create and lookup partial VMA.

v2: Named phases

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-37-chris@chris-wilson.co.uk

drm/i915: Verify page layout for rotated VMA

Exercise creating rotated VMA and checking the page order within.

v2..v3: Be more creative in rotated params

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-36-chris@chris-wilson.co.uk

drm/i915: Exercise i915_vma_pin/i915_vma_insert

High-level testing of the struct drm_mm by verifying our handling of
weird requests to i915_vma_pin.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-35-chris@chris-wilson.co.uk

drm/i915: Test creation of VMA

Simple test to exercise creation and lookup of VMA within an object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-34-chris@chris-wilson.co.uk

drm/i915: Use fault-injection to force the shrinker to run in live GTT tests

It is possible whilst allocating the page-directory tree for a ppgtt
bind that the shrinker may run and reap unused parts of the tree. If the
shrinker happens to remove a chunk of the tree that the
allocate_va_range has already processed, we may then try to insert into
the dangling tree. This test uses the fault-injection framework to force
the shrinker to be invoked before we allocate new pages, i.e. new chunks
of the PD tree.

References: https://bugs.freedesktop.org/show_bug.cgi?id=99295
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Live testing of lowlevel GTT operations

Directly test allocating the va range and clearing it, this bypasses the
use of i915_vma_bind() and inserting the pages to focus on testing of
the pagetables.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-32-chris@chris-wilson.co.uk

drm/i915: Exercise filling and removing random ranges from the live GTT

Test the low-level i915_address_space interfaces to sanity check the
live insertion/removal of address ranges.

v2: Split out the low-level operations to a new test

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-31-chris@chris-wilson.co.uk

drm/i915: Fill different pages of the GTT

Exercise filling different pages of the GTT

v2: Walk all holes until we timeout

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-30-chris@chris-wilson.co.uk

drm/i915: Exercise filling the top/bottom portions of the global GTT

Same test as previously for the per-process GTT instead applied to the
global GTT.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-29-chris@chris-wilson.co.uk

drm/i915: Exercise filling the top/bottom portions of the ppgtt

Allocate objects with varying number of pages (which should hopefully
consist of a mixture of contiguous page chunks and so coalesced sg
lists) and check that the sg walkers in insert_pages cope.

v2: Check both small <-> large

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-28-chris@chris-wilson.co.uk

drm/i915: Add initial selftests for i915_gem_gtt

Simple starting point for adding selftests for i915_gem_gtt, first
try creating a ppGTT and filling it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-27-chris@chris-wilson.co.uk

drm/i915: Add a live dmabuf selftest

Though we have good coverage of our dmabuf interface through the mock
tests, we also want to check the heavy module unload paths of the live
i915 driver.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-26-chris@chris-wilson.co.uk

drm/i915: Add some mock tests for dmabuf interop

Check that we can create both dmabuf and objects from dmabuf.

v2: Cleanups, correct include, fix unpin on dead path and prevent
explosion on dmabuf init failure

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-25-chris@chris-wilson.co.uk

drm/i915: Sanity check all registers for matching fw domains

Add a late selftest that walks over all forcewake registers (those below
0x40000) and uses the mmio debug register to check to see if any are
unclaimed. This is possible if we fail to wake the appropriate
powerwells for the register.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-24-chris@chris-wilson.co.uk

drm/i915: Test all fw tables during mock selftests

In addition to just testing the fw table we load, during the initial
mock testing we can test that all tables are valid (so the testing is
not limited to just the platforms that load that particular table).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-23-chris@chris-wilson.co.uk

drm/i915: Move uncore selfchecks to live selftest infrastructure

Now that the kselftest infrastructure exists, put it to use and add to
it the existing consistency checks on the fw register lookup tables.

v2: s/tabke/table/

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-22-chris@chris-wilson.co.uk

drm/i915: Test coherency of and barriers between cache domains

Write into an object using WB, WC, GTT, and GPU paths and make sure that
our internal API is sufficient to ensure coherent reads and writes.

v2: Avoid invalid free upon allocation error

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-21-chris@chris-wilson.co.uk

drm/i915: Test exhaustion of the mmap space

An unlikely error condition that we can simulate by stealing most of
the range before trying to insert new objects.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-20-chris@chris-wilson.co.uk

drm/i915: Test partial mappings

Create partial mappings to cover a large object, investigating tiling
(fenced regions) and VMA reuse.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-19-chris@chris-wilson.co.uk

drm/i915: Add a live seftest for GEM objects

Starting with a placeholder test just to reassure that we can create a
test object,

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-18-chris@chris-wilson.co.uk

drm/i915: Add selftests for object allocation, phys

The phys object is a rarely used device (only very old machines require
a chunk of physically contiguous pages for a few hardware interactions).
As such, it is not exercised by CI and to combat that we want to add a
test that exercises the phys object on all platforms.

v2: Always set err on error paths and not rely on inheriting the err.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-17-chris@chris-wilson.co.uk

drm/i915: Live testing of empty requests

Primarily to emphasize the difference between just advancing the
breadcrumb using a bare request and the overhead of dispatching an
execbuffer.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-16-chris@chris-wilson.co.uk

drm/i915: Test request ordering between engines

A request on one engine with a dependency on a request on another engine
must wait for completion of the first request before starting.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-15-chris@chris-wilson.co.uk

drm/i915: Test simultaneously submitting requests to all engines

Use a recursive-batch to busy spin on each to ensure that each is being
run simultaneously.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-14-chris@chris-wilson.co.uk

drm/i915: Simple selftest to exercise live requests

Just create several batches of requests and expect it to not fall over!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-13-chris@chris-wilson.co.uk

drm/i915: Add a simple fence selftest to i915_gem_request

Do a quick selftest on in the interoperability of dma_fence_wait on a
i915_gem_request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-12-chris@chris-wilson.co.uk

drm/i915: Add a simple request selftest for waiting

A trivial kselftest to submit a request and wait upon it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-11-chris@chris-wilson.co.uk

drm/i915: Add selftests for i915_gem_request

Simple starting point for adding seltests for i915_gem_request, first
mock a device (with engines and contexts) that allows us to construct
and execute a request, along with waiting for the request to complete.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-10-chris@chris-wilson.co.uk

drm/i915: Create a fake object for testing huge allocations

We would like to be able to exercise huge allocations even on memory
constrained devices. To do this we create an object that allocates only
a few pages and remaps them across its whole range - each page is reused
multiple times. We can therefore pretend we are rendering into a much
larger object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-9-chris@chris-wilson.co.uk

drm/i915: Mock infrastructure for request emission

Create a fake engine that runs requests using a timer to simulate hw.

v2: Prevent leaks of ctx->name along error paths

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-8-chris@chris-wilson.co.uk

drm/i915: Mock a GGTT for self-testing

A very simple mockery, just a random manager and timeline. Useful for
inserting objects and ordering retirement; and not much else.

v2: mock_fini_ggtt() to complement mock_init_ggtt().

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-7-chris@chris-wilson.co.uk

drm/i915: Mock the GEM device for self-testing

A simulacrum of drm_i915_private to let us pretend interactions with the
device.

v2: Tidy init error paths

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-6-chris@chris-wilson.co.uk

drm/i915: Add unit tests for the breadcrumb rbtree, wakeups

Third retroactive test, make sure that the seqno waiters are woken.

v2: Smattering of comments, rearrange code
v3: Fix IDLE assert to avoid startup/sleep races

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-5-chris@chris-wilson.co.uk

drm/i915: Add unit tests for the breadcrumb rbtree, completion

Second retroactive test, make sure that the waiters are removed from the
global wait-tree when their seqno completes.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-4-chris@chris-wilson.co.uk

drm/i915: Add unit tests for the breadcrumb rbtree, insert/remove

First retroactive test, make sure that the waiters are in global seqno
order after random inserts and removals.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-3-chris@chris-wilson.co.uk

drm/i915: Add some selftests for sg_table manipulation

Start exercising the scattergather lists, especially looking at
iteration after coalescing.

v2: Comment on the peculiarity of table construction (i.e. why this
sg_table might be interesting).
v3: Added one __func__ to identify expect_pfn_sg()
v4: Loop until we have crossed the chain boundary (forcing sg_table to
do multiple allocations) before squelching a potential ENOMEM from oom.

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: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-2-chris@chris-wilson.co.uk

drm/i915: Provide a hook for selftests

Some pieces of code are independent of hardware but are very tricky to
exercise through the normal userspace ABI or via debugfs hooks. Being
able to create mock unit tests and execute them through CI is vital.
Start by adding a central point where we can execute unit tests and
a parameter to enable them. This is disabled by default as the
expectation is that these tests will occasionally explode.

To facilitate integration with igt, any parameter beginning with
i915.igt__ is interpreted as a subtest executable independently via
igt/drv_selftest.

Two classes of selftests are recognised: mock unit tests and integration
tests. Mock unit tests are run as soon as the module is loaded, before
the device is probed. At that point there is no driver instantiated and
all hw interactions must be "mocked". This is very useful for writing
universal tests to exercise code not typically run on a broad range of
architectures. Alternatively, you can hook into the live selftests and
run when the device has been instantiated - hw interactions are real.

v2: Add a macro for compiling conditional code for mock objects inside
real objects.
v3: Differentiate between mock unit tests and late integration test.
v4: List the tests in natural order, use igt to sort after modparam.
v5: s/late/live/
v6: s/unsigned long/unsigned int/
v7: Use igt_ prefixes for long helpers.
v8: Deobfuscate macros overriding functions, stop using -I$(src)

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170213171558.20942-1-chris@chris-wilson.co.uk

drm/i915: Clear the last_retired_context following a hang/reset

Following a hang and reset, we know that the engine is idle and all
context state has been saved or lost. Consequently, we know that the
engine is no longer referencing the last context and we can relinquish
our tracking.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tursulin@ursulin.net>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-5-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Park the breadcrumbs signaler across a GPU reset

The signal threads may be running concurrently with the GPU reset. The
completion from the GPU run asynchronous with the reset and two threads
may see different snapshots of the state, and the signaler may mark a
request as complete as we try to reset it. We don't tolerate 2 different
views of the same state and complain if we try to mark a request as
failed if it is already complete. Disable the signal threads during
reset to prevent this conflict (even though the conflict implies that
the state we resetting to is invalid, we have already made our
decision!).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99733
References: https://bugs.freedesktop.org/show_bug.cgi?id=99671
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-4-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Kill the tasklet then disable

Disabling the tasklet leaves it if scheduled on the ready to run list
until it is re-enabled. This will leave the ksoftird thread spinning
until satisfied. To prevent this situation on starting the GPU reset, we
want to kill the tasklet first and then disable. The same problem will
arise when a tasklet is scheduled from another device, so a better
solution is required for the general case.

Reported-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 1f7b847d72c3 ("drm/i915: Disable engine->irq_tasklet around resets")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-3-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Always call i915_gem_reset_finish() following i915_gem_reset_prepare()

As i915_gem_reset_finish() undoes the steps from
i915_gem_reset_prepare() to leave the system in a fully-working state,
e.g. to be able to free the breadcrumb signal threads, make sure that we
always call it even on the error path.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-2-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Assert that the active request hasn't been signaled

As the request is not complete, it should not be signaled. Assert that
this is true before we process the request for a reset.

References: https://bugs.freedesktop.org/show_bug.cgi?id=99671
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170212172002.23072-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Rename conditional GEM execution macros

After a brief discussion, we settled on a naming convention for the
conditional GEM debugging data that should be clearer to the casual
user: GEM_DEBUG

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>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170207102319.10910-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Move the irq_barrier for reset earlier into reset_prepare

When updating the bookkeeping following the reset, we need the seqno to
be coherent on the CPU prior to trusting its result for deciding whether
any request is completed. We need the irq_barrier before we start making
these decisions, i.e. in reset_prepare.

References: https://bugs.freedesktop.org/show_bug.cgi?id=99733
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210185214.23463-1-chris@chris-wilson.co.uk
Reviewed-by: Michel Thierry <michel.thierry@intel.com>

drm/i915: Flush the freed object queue on device release

As dmabufs may live beyond the PCI device removal, we need to flush the
freed object worker on device release, and include a warning in case
there is a leak.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210163523.17533-3-chris@chris-wilson.co.uk

drm/i915: Unbind any residual objects/vma from the Global GTT on shutdown

We may unload the PCI device before all users (such as dma-buf) are
completely shutdown. This may leave VMA in the global GTT which we want
to revoke, whilst keeping the objects themselves around to service the
dma-buf.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210163523.17533-2-chris@chris-wilson.co.uk

drm/i915: Split device release from unload

We may need to keep our memory management alive after we have unloaded
the physical pci device. For example, if we have exported an object via
dmabuf, that will keep the device around but the pci device may be
removed before the dmabuf itself is released, use of the pci hardware
will be revoked, but the memory and object management needs to persist
for the dmabuf.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210163523.17533-1-chris@chris-wilson.co.uk

drm/i915/debugfs: Add i915_hpd_storm_ctl

This adds a file in i915's debugfs directory that allows userspace to
manually control HPD storm detection. This is mainly for hotplugging
tests, where we might want to test HPD storm functionality or disable
storm detection to speed up hotplugging tests without breaking anything.

Changes since v1:
- Make HPD storm interval configurable
- Misc code cleanup

Signed-off-by: Lyude <lyude@redhat.com>
Acked-by: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Tomeu Vizoso <tomeu@tomeuvizoso.net>

drm/i915: Simplify platform checks in intel_fb_pitch_limit()

Replace the VLV/CHV check with a HAS_GMCH_DISPLAY check in
intel_fb_pitch_limit(), because it's shorter.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170208175328.11064-1-ville.syrjala@linux.intel.com
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>

drm/i915: Avoid spurious WARNs about the wrong pipe in the PPS code

Until recently vlv_steal_power_sequencer() wasn't being called for
normal DP ports, and hence it could assert that it should only be
called for pipe A and B (since pipe C doesn't support eDP). However
that changed when we started to consider normal DP ports as well when
choosing a PPS. So we will now get spurious warnings when
vlv_steal_power_sequencer() does get called for pipe C. Avoid this by
moving the WARN down into vlv_detach_power_sequencer() where this
assertion should still hold.

Cc: Imre Deak <imre.deak@intel.com>
Cc: stable@vger.kernel.org
Fixes: 9f2bdb006a7e ("drm/i915: Prevent PPS stealing from a normal DP port on VLV/CHV")
References: https://bugs.freedesktop.org/show_bug.cgi?id=95287
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170208175254.10958-1-ville.syrjala@linux.intel.com
Reviewed-by: Imre Deak <imre.deak@intel.com>

Merge remote-tracking branch 'airlied/drm-next' into drm-intel-next-queued

Chris Wilson needs the new drm_driver->release callback to make sure
the shiny new dma-buf testcases don't oops the driver on unload.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915/byt: Take powerwell for reading PIPESTAT in debugfs

[12493.693827] WARNING: CPU: 1 PID: 14860 at drivers/gpu/drm/i915/intel_uncore.c:795 __unclaimed_reg_debug+0x5d/0x80 [i915]
[12493.693868] Unclaimed read from register 0x1f0024
[12493.693905] Modules linked in: vgem i915 drm_kms_helper drm intel_gtt i2c_algo_bit syscopyarea sysfillrect sysimgblt fb_sys_fops prime_numbers intel_powerclamp crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel cryptd lpc_ich i2c_i801 mfd_core video i2c_designware_platform i2c_designware_core i2c_core button autofs4 sd_mod ahci libahci libata scsi_mod [last unloaded: i915]
[12493.694039] CPU: 1 PID: 14860 Comm: intel-gpu-overl Tainted: G     U          4.10.0-rc7+ #11
[12493.694079] Hardware name: GIGABYTE GB-BXBT-1900/MZBAYAB-00, BIOS F8 03/02/2016
[12493.694121] Call Trace:
[12493.694169]  dump_stack+0x67/0x9d
[12493.694235]  __warn+0x117/0x140
[12493.694288]  warn_slowpath_fmt+0x4f/0x60
[12493.694344]  ? do_raw_spin_lock+0x116/0x180
[12493.694533]  ? check_for_unclaimed_mmio+0x98/0xe0 [i915]
[12493.694727]  __unclaimed_reg_debug+0x5d/0x80 [i915]
[12493.694923]  fwtable_read32+0x2c5/0x330 [i915]
[12493.695108]  i915_interrupt_info+0xd52/0xf80 [i915]
[12493.695302]  ? gen6_write16+0x310/0x310 [i915]
[12493.695357]  seq_read+0x187/0x710
[12493.695412]  full_proxy_read+0x75/0xc0
[12493.695472]  __vfs_read+0x5a/0x220
[12493.695524]  ? kmem_cache_free+0x6c/0x260
[12493.695577]  ? putname+0x97/0xa0
[12493.695629]  ? putname+0x97/0xa0
[12493.695682]  ? rcu_read_lock_sched_held+0xb8/0xd0
[12493.695735]  ? rw_verify_area+0x65/0x140
[12493.695787]  vfs_read+0xd1/0x1f0
[12493.695840]  SyS_read+0x62/0xc0
[12493.695893]  entry_SYSCALL_64_fastpath+0x1c/0xb1
[12493.695943] RIP: 0033:0x7f82dca99ba0
[12493.695985] RSP: 002b:00007ffc0bdfd4f8 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
[12493.696031] RAX: ffffffffffffffda RBX: 00007ffc0be005a0 RCX: 00007f82dca99ba0
[12493.696073] RDX: 0000000000001fff RSI: 00007ffc0bdfd500 RDI: 000000000000001a
[12493.696115] RBP: ffffffff810fb639 R08: 302f6972642f6775 R09: 00007f82dca0999a
[12493.696157] R10: 00007f82dcd62760 R11: 0000000000000246 R12: ffff880069a17f98
[12493.696199] R13: 00007ffc0bdfd428 R14: 0000000000000003 R15: 00007ffc0bdfd428
[12493.696250]  ? trace_hardirqs_off_caller+0xd9/0x130
[12493.696300] ---[ end trace 52ccf4d39793cc59 ]---

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99761
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210133632.16946-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Always pin contexts into the high GGTT

Now that we have fast top-down insertion into the drm_mm, we can use it
for frequent runtime operations like insertion of the context object,
whereas before we limited it to the one-off insertion of the pinned
kernel context. Keeping the active context objects out of the mappable
region of the global GTT (except under memory pressure) improves our
ability to allocate mappable aperture region without triggering a GPU
stall.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170210101422.1598-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Remove unused function intel_ddi_get_link_dpll()

The function intel_ddi_get_link_dpll() was added in f169660ed4e5
("drm/i915/dp: Add a standalone function to obtain shared dpll for
HSW/BDW/SKL/BXT") to "allow for the implementation of a platform
neutral upfront link training function", but such implementation
never landed.

So remove that function and clean up the exported shared DPLL interface.

Fixes: f169660ed4e5 ("drm/i915/dp: Add a standalone function to obtain shared dpll for HSW/BDW/SKL/BXT")
Cc: Durgadoss R <durgadoss.r@intel.com>
Cc: Manasi Navare <manasi.d.navare@intel.com>
Cc: Jim Bride <jim.bride@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1484310032-1863-1-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Make power domain masks 64 bit long

There are currently 30 power domains, which puts us pretty close to the
limit with 32 bit masks. Prepare for the future and increase the limit
to 64 bit.

v2: Rebase
v3: s/unsigned long long/u64/ (Joonas)
    Allow the 64th bit of the mask to be used. (Joonas)

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170209093121.24410-1-ander.conselvan.de.oliveira@intel.com

Merge branch 'drm-next-4.11' of git://people.freedesktop.org/~agd5f/linux into drm-next

Some additional fixes for 4.11.  Delayed a bit due to Chinese New Year. Highlights:
- Powerplay fixes
- VCE and UVD powergating fixes
- Clean up amdgpu SI gfx code to match CI and VI
- Misc bug fixes

* 'drm-next-4.11' of git://people.freedesktop.org/~agd5f/linux: (30 commits)
  drm/amdgpu: report the number of bytes moved at buffer creation
  drm/amdgpu: fix a potential deadlock in amdgpu_bo_create_restricted()
  drm/amdgpu: add support for new smc firmware on polaris
  drm/amd/powerplay: refine code to avoid potential bug that the memory not cleared.
  drm/amdgpu: shut up #warning for compile testing
  drm/amdgpu/virt: fix double kfree on bo_va
  drm/radeon: remove some dead code
  drm/radeon: avoid kernel segfault in vce when gpu fails to resume
  drm/amd/powerplay: set fan speed to max in profile peak mode only.
  drm/amd/gfx6: update gb_addr_config
  drm/amdgpu: update HAINAN_GB_ADDR_CONFIG_GOLDEN
  drm/amdgpu: update VERDE_GB_ADDR_CONFIG_GOLDEN
  drm/amdgpu: refine si_read_register
  drm/amdgpu/gfx6: clean up spi configuration
  drm/amdgpu/gfx6: clean up cu configuration
  drm/amdgpu/gfx6: clean up rb configuration
  drm/amdgpu: refine vce3.0 code and related powerplay pg code.
  drm/amdgpu: move subfunctions to the front of vce_v2_0.c.
  drm/amdgpu: enable vce pg feature on Kv.
  drm/amdgpu: refine code for VCE2.0 and related dpm code.
  ...

Merge tag 'drm-fsl-dcu-for-v4.11' of git.agner.ch/git/linux-drm-fsl-dcu into drm-next

two minor fixes.

* tag 'drm-fsl-dcu-for-v4.11' of http://git.agner.ch/git/linux-drm-fsl-dcu:
  drm/fsl-dcu: check for clk_prepare_enable() error
  drm/fsl-dcu: remove unneeded 'ret' assignment

Merge tag 'drm-misc-next-fixes-2017-02-09' of git://anongit.freedesktop.org/git/drm-misc into drm-next

Just 3 bugfixes for 4.11 merge window:

- fbdev module unload oops fix from Chris
- patch from Dan that look really dangers, better safe than sorry

* tag 'drm-misc-next-fixes-2017-02-09' of git://anongit.freedesktop.org/git/drm-misc:
  drm/atomic: fix an error code in mode_fixup()
  drm: Cancel drm_fb_helper_resume_work on unload
  drm: Cancel drm_fb_helper_dirty_work on unload

drm/i915: Use the size/type of address space to make decisions

Once the address space has been created (using 3 or 4 levels of page
tables), we should use that to program the appropriate type into the
contexts. This gives us the flexibility to handle different types of
address spaces at runtime.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170209144036.23664-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/amdgpu: report the number of bytes moved at buffer creation

Like ttm_bo_validate(), ttm_bo_init() might need to move BO and
the number of bytes moved by TTM should be reported. This can help
the throttle buffer migration mechanism to make a better decision.

v2: fix computation

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: fix a potential deadlock in amdgpu_bo_create_restricted()

When ttm_bo_init() fails, the reservation mutex should be unlocked.

In debug build, the kernel reported "possible recursive locking
detected" in this codepath. For debugging purposes, I also added
a "WARN_ON(ww_mutex_is_locked())" when ttm_bo_init() fails and the
mutex was locked as expected.

This should fix (random) GPU hangs. The easy way to reproduce the
issue is to change the "Super Sampling" option from 1.0 to 2.0 in
Hitman. It will create a huge buffer, evict a bunch of buffers
(around ~5k) and deadlock.

This regression has been introduced pretty recently.

v2: only release the mutex if resv is NULL

Fixes: 12a852219583 ("drm/amdgpu: improve AMDGPU_GEM_CREATE_VRAM_CLEARED handling (v2)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu: add support for new smc firmware on polaris

Some polaris variants require new smc firmware.

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

drm/amd/powerplay: refine code to avoid potential bug that the memory not cleared.

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

drm/i915: Enable atomic support by default on supported platforms.

i915 is pretty much feature complete. Support for atomic i915-specific
connector properties is still missing; those properties can (for now)
only be set through the legacy ioctl.

ILK style watermarks and gen9+ watermarks are handled atomically,
and nonblocking modesets work. FBC has also been made to work with
atomic.

gen4x- and vlv/chv watermarks still need to be fixed, so disable atomic
by default there for now.

Flip the switch!!

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Lyude <cpaul@redhat.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486021302-24910-1-git-send-email-maarten.lankhorst@linux.intel.com
[mlankhorst: Fix checkpatch warning about extra space in match_info]
Acked-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Lyude <lyude@redhat.com>

drm/i915: Add __destroy_hw_context

__create_hw_context can use a good counterpart.

Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1486640065-13695-1-git-send-email-joonas.lahtinen@linux.intel.com

drm/i915: Assert that we never create a vma for the aliasing_ppgtt

The aliasing_ppgtt is just a container for the HW context that mirrors
the global gtt. It should never be used directly, so assert if we make
the mistake of trying to allocate a VMA for it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170209111933.12420-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Check for timeout completion when waiting for the rq to submitted

We first wait for a request to be submitted to hw and assigned a seqno,
before we can wait for the hw to signal completion (otherwise we don't
know the hw id we need to wait upon). Whilst waiting for the request to
be submitted, we may exceed the user's timeout and need to propagate the
error back.

v2: Make ETIME into an error from wait_for_execute for consistent exit
handling.

Reported-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Fixes: 4680816be336 ("drm/i915: Wait first for submission, before waiting for request completion")
Testcase: igt/gem_wait/basic-await
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>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.10-rc1+
Cc: stable@vger.kernel.org
Link: http://patchwork.freedesktop.org/patch/msgid/20170208181238.7232-1-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: A hotfix for making aliasing PPGTT work for GVT-g

This patch makes PPGTT page table non-shrinkable when using aliasing PPGTT
mode. It's just a temporary solution for making GVT-g work.

Fixes: 2ce5179fe826 ("drm/i915/gtt: Free unused lower-level page tables")
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Michal Winiarski <michal.winiarski@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Zhenyu Wang <zhenyuw@linux.intel.com>
Cc: Zhiyuan Lv <zhiyuan.lv@intel.com>
Signed-off-by: Zhi Wang <zhi.a.wang@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486559013-25251-2-git-send-email-zhi.a.wang@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: <drm-intel-fixes@lists.freedesktop.org> # v4.10-rc1+
Cc: stable@vger.kernel.org
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/dp: Move initialization of hpd_pin to a new function

This shaves a few lines from intel_dp_init_connector() and will serve as
a good place to add other port specific information in a follow up
patch.

While at it, convert BUG() to MISSING_CASE() in the default case.

v2: s/BUG/MISSING_CASE. (Chris)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170203140316.20792-2-ander.conselvan.de.oliveira@intel.com

drm/i915: Remove WA for swapped HPD pins in broxton A stepping

Remove workaround for swapped HPD pins in broxton A stepping, which is
pre-production hardware.

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170203140316.20792-1-ander.conselvan.de.oliveira@intel.com

drm/amdgpu: shut up #warning for compile testing

My randconfig tests on linux-next showed a newly introduced warning:

drivers/gpu/drm/amd/amdgpu/amdgpu_object.c: In function 'amdgpu_bo_create_restricted':
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c:377:2: error: #warning Please enable CONFIG_MTRR and CONFIG_X86_PAT for better performance thanks to write-combining [-Werror=cpp]

Generally speaking, warnings about bad kernel configuration are not particularly
helpful. We could enforce the selection of X86_PAT through Kconfig, so the driver
cannot even be used unless it is enabled, or we could just rely on the runtime
warning that is also there.

In this version, I'm making the warning conditional on CONFIG_COMPILE_TEST, which
shuts it up for me, but not people that may actually want to run the kernel
as a compromize.

Fixes: a2e2f29970aa ("drm/amdgpu: Bring bo creation in line with radeon driver (v2)")
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amdgpu/virt: fix double kfree on bo_va

bo_va is being kfree'd twice, once in the call to amdgpu_vm_bo_rmv
and then a short while later. Fix this double free by removing
the 2nd kfree.

Detected by CoverityScan, CID#1399524 ("Double Free")

Reviewed-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: remove some dead code

If "rdev->bios" is NULL then we don't need to free it.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/radeon: avoid kernel segfault in vce when gpu fails to resume

When GPU fails to resume we can not trust that value we write to GPU
memory will post and we might get garbage (more like 0xffffffff on
x86) when reading them back. This trigger out of range memory access
in the kernel inside the vce resume code path.

This patch use canonical value to compute offset instead of reading
back value from GPU memory.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drm/amd/powerplay: set fan speed to max in profile peak mode only.

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