drm/i915: Remove the preempted request from the execution queue

After the request is cancelled, we then need to remove it from the
global execution timeline and return it to the context timeline, the
inverse of submit_request().

v2: Move manipulation of struct intel_wait to helpers

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-12-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Allow a request to be cancelled

If we preempt a request and remove it from the execution queue, we need
to undo its global seqno and restart any waiters.

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/20170223074422.4125-11-chris@chris-wilson.co.uk

drm/i915: Take a reference whilst processing the signaler request

The plan in the near-future is to allow requests to be removed from the
signaler. We can no longer then rely on holding a reference to the
request for the duration it is in the signaling tree, and instead must
obtain a reference to the request for the current operation using RCU.

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/20170223074422.4125-10-chris@chris-wilson.co.uk

drm/i915: Protect the request->global_seqno with the engine->timeline lock

A request is assigned a global seqno only when it is on the hardware
execution queue. The global seqno can be used to maintain a list of
requests on the same engine in retirement order, for example for
constructing a priority queue for waiting. Prior to its execution, or
if it is subsequently removed in the event of preemption, its global
seqno is zero. As both insertion and removal from the execution queue
may operate in IRQ context, it is not guarded by the usual struct_mutex
BKL. Instead those relying on the global seqno must be prepared for its
value to change between reads. Only when the request is complete can
the global seqno be stable (due to the memory barriers on submitting
the commands to the hardware to write the breadcrumb, if the HWS shows
that it has passed the global seqno and the global seqno is unchanged
after the read, it is indeed 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/20170223074422.4125-9-chris@chris-wilson.co.uk

drm/i915: Deconstruct execute fence

On reflection, we are only using the execute fence as a waitqueue on the
global_seqno and not using it for dependency tracking between fences
(unlike the submit and dma fences). By only treating it as a waitqueue,
we can then treat it similar to the other waitqueues during submit,
making the code simpler.

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/20170223074422.4125-8-chris@chris-wilson.co.uk

drm/i915: Inline __i915_gem_request_wait_for_execute()

It had only one callsite and existed to keep the code clearer. Now
having shared the wait-on-error between phases and with plans to change
the wait-for-execute in the next few patches, remove the out of line
wait loop and move it into the main body of i915_wait_request.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-7-chris@chris-wilson.co.uk

drm/i915: Add ourselves to the gpu error waitqueue for the entire wait

Add ourselves to the gpu error waitqueue earlier on, even before we
determine we have to wait on the seqno. This is so that we can then
share the waitqueue between stages in subsequent patches.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-6-chris@chris-wilson.co.uk

drm/i915: Use a local to shorten req->i915->gpu_error.wait_queue

Use a local variable to avoid having to type out the full name of the
gpu_error wait_queue.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-5-chris@chris-wilson.co.uk

drm/i915: Move reserve_seqno() next to unreserve_seqno()

Move the companion functions next to each other.

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/20170223074422.4125-4-chris@chris-wilson.co.uk

drm/i915: Keep a global seqno per-engine

Replace the global device seqno with one for each engine, and account
for in-flight seqno on each separately. This is consistent with
dma-fence as each timeline has separate fence-contexts for each engine
and a seqno is only ordered within a fence-context (i.e.  seqno do not
need to be ordered wrt to other engines, just ordered within a single
engine). This is required to enable request rewinding for preemption on
individual engines (we have to rewind the global seqno to avoid
overflow, and we do not have to rewind all engines just to preempt one.)

v2: Rename active_seqno to inflight_seqnos to more clearly indicate that
it is a counter and not equivalent to the existing seqno. Update
functions that operated on active_seqno similarly.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-3-chris@chris-wilson.co.uk

drm/i915: Check against the signaled bit for fences/requests

When dma_fence_signal() is called, it sets a flag to indicate the fence
is complete. Before the dma_fence is signaled, the seqno check will
first be passed. During an unlocked check (such as inside a waiter), it
is possible for the fence to be signaled even though the seqno has been
reset (by engine wraparound). In this case the waiter will be kicked,
but for an extra layer of protection we can check the persistent
signaled bit from the fence.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223074422.4125-2-chris@chris-wilson.co.uk

Revert "drm/i915/dp: Ratelimit DP aux timeout messages"

This reverts commit 7ee686034b8b "drm/i915/dp: Ratelimit DP aux timeout
messages" as although it successfully squelches the debug messages, when
it does so it generates a warning instead. CI lights up orange with all
the warnings!

In its current incarnation DRM_DEBUG_RATELIMITED is not usable for us,
and we need to first teach lib/ratelimit.c not to warn when used for
debug messages.

Fixes: 7ee686034b8b ("drm/i915/dp: Ratelimit DP aux timeout messages")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lyude <lyude@redhat.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: intel-gfx@lists.freedesktop.org
Link: http://patchwork.freedesktop.org/patch/msgid/20170223115102.7059-1-chris@chris-wilson.co.uk
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/glk: Fix Geminilake scalers mode programming

Geminilake scalers can do 7x7 filtering for all supported input sizes,
so it doesn't need the "high quality" mode programming, which was
actually removed from that platform.

v2: Split dev_priv parameter change out. (Ville)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>,
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223071600.14356-5-ander.conselvan.de.oliveira@intel.com

drm/i915/glk: Pass dev_priv to intel_atomic_setup_scalers()

Pass dev_priv to intel_atomic_setup_scalers(). The next patch will need
a dev_priv pointer.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223071600.14356-4-ander.conselvan.de.oliveira@intel.com

drm/i915/glk: Fix maximum scaling factor for Geminilake scalers

Geminilake can output two pixels per clock, and that affects the maximum
scaling factor for its scalers. Take that into account and avoid the
following warning:

WARNING: CPU: 1 PID: 593 at drivers/gpu/drm/i915/intel_display.c:13223 skl_max_scale.part.129+0x78/0x80 [i915]
WARN_ON_ONCE(!crtc_clock || cdclk < crtc_clock)
Modules linked in: x86_pkg_temp_thermal i915 coretemp kvm_intel kvm i2c_algo_bit drm_kms_helper irqbypass crct10dif_pclmul prime_numbers crc32_pclmul drm ghash_clmulni_intel shpchp tpm_tis tpm_tis_core tpm nfsd authw
CPU: 1 PID: 593 Comm: kworker/u8:3 Tainted: G        W       4.10.0-rc8ander+ #330
Hardware name: Intel Corp. Geminilake/GLK RVP1 DDR4 (05), BIOS GELKRVPA.X64.0035.B33.1702150552 02/15/2017
Workqueue: events_unbound async_run_entry_fn
Call Trace:
 dump_stack+0x86/0xc3
 __warn+0xcb/0xf0
 warn_slowpath_fmt+0x5f/0x80
 skl_max_scale.part.129+0x78/0x80 [i915]
 intel_check_primary_plane+0xa6/0xc0 [i915]
 intel_plane_atomic_check_with_state+0xd1/0x1a0 [i915]
 ? drm_printk+0xb5/0xc0 [drm]
 intel_plane_atomic_check+0x3d/0x80 [i915]
 drm_atomic_helper_check_planes+0x7c/0x200 [drm_kms_helper]
 intel_atomic_check+0xa5b/0x11a0 [i915]
 drm_atomic_check_only+0x353/0x600 [drm]
 ? drm_atomic_add_affected_connectors+0x10c/0x120 [drm]
 drm_atomic_commit+0x18/0x50 [drm]
 restore_fbdev_mode+0x14c/0x2a0 [drm_kms_helper]
 drm_fb_helper_restore_fbdev_mode_unlocked+0x34/0x80 [drm_kms_helper]
 drm_fb_helper_set_par+0x2d/0x60 [drm_kms_helper]
 intel_fbdev_set_par+0x1a/0x70 [i915]
 fbcon_init+0x582/0x610
 visual_init+0xd6/0x130
 do_bind_con_driver+0x1da/0x3c0
 do_take_over_console+0x116/0x180
 do_fbcon_takeover+0x5c/0xb0
 fbcon_event_notify+0x772/0x8a0
 ? __blocking_notifier_call_chain+0x35/0x70
 notifier_call_chain+0x4a/0x70
 __blocking_notifier_call_chain+0x4d/0x70
 blocking_notifier_call_chain+0x16/0x20
 fb_notifier_call_chain+0x1b/0x20
 register_framebuffer+0x278/0x360
 drm_fb_helper_initial_config+0x253/0x440 [drm_kms_helper]
 intel_fbdev_initial_config+0x18/0x30 [i915]
 async_run_entry_fn+0x39/0x170
 process_one_work+0x212/0x670
 ? process_one_work+0x197/0x670
 worker_thread+0x4e/0x490
 kthread+0x101/0x140
 ? process_one_work+0x670/0x670
 ? kthread_create_on_node+0x60/0x60
 ret_from_fork+0x31/0x40

v2: s/max_pixclk/max_dotclk/ (Ville)
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223071600.14356-3-ander.conselvan.de.oliveira@intel.com

drm/i915/glk: Fix watermark computations for third sprite plane

Geminilake has a third sprite plane (or fourth universal plane) that is
independent from the cursor. Make sure that for_each_plane_id_on_crtc()
is aware of that extra plane so that the watermark code takes it into
account.

Fixes: e9c9882556fc ("drm/i915/glk: Configure number of sprite planes properly")
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@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
Cc: <drm-intel-fixes@lists.freedesktop.org>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170223071600.14356-2-ander.conselvan.de.oliveira@intel.com

drm/i915: Remove Braswell GGTT update w/a

Testing with concurrent GGTT accesses no longer show the coherency
problems from yonder, commit 5bab6f60cb4d ("drm/i915: Serialise updates
to GGTT with access through GGTT on Braswell"). My presumption is that
the root cause was more likely fixed by commit 3b5724d702ef ("drm/i915:
Wait for writes through the GTT to land before reading back"), along
with the use of WC updates to the global gTT in commit 8448661d65f6
("drm/i915: Convert clflushed pagetables over to WC maps". Given
that the original symptoms can no longer be reproduced, time to remove
the workaround.

Testcase: igt/gem_concurrent_blit
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/20170220124718.14796-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/dp: Ratelimit DP aux timeout messages

Right now this is just leaving a lot of spam in dmesg that makes real
issues more difficult to debug. As well (as noted by the comment right
above the DRM_DEBUG_KMS() call) this is normal behavior when there's
nothing connected to the DisplayPort connector.

Signed-off-by: Lyude <lyude@redhat.com>

drm/i915: Remove 'retire' parameter from intel_fb_obj_flush

Setting retire=true is identical to using origin=ORIGIN_CS, so make the
same simplification to intel_fb_obj_flush() as already employed for
intel_fb_obj_invalidate().

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/20170222114049.28456-6-chris@chris-wilson.co.uk

drm/i915: Perform object clflushing asynchronously

Flushing the cachelines for an object is slow, can be as much as 100ms
for a large framebuffer. We currently do this under the struct_mutex BKL
on execution or on pageflip. But now with the ability to add fences to
obj->resv for both flips and execbuf (and we naturally wait on the fence
before CPU access), we can move the clflush operation to a workqueue and
signal a fence for completion, thereby doing the work asynchronously and
not blocking the driver or its clients.

v2: Introduce i915_gem_clflush.h and use a new name, split out some
extras into separate patches.

Suggested-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170222114049.28456-5-chris@chris-wilson.co.uk

drm/i915: Skip clflushes for all non-page backed objects

Generalise the skip for physical and stolen objects by skipping anything
we do not have a valid address for inside the sg.

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/20170222114049.28456-4-chris@chris-wilson.co.uk

drm/i915: Amalgamate flushing of display objects

We have three different paths by which userspace wants to flush the
display plane (i.e. objects with obj->pin_display). Use a common helper
to identify those paths and to simplify a later change.

v2: Include the conditional in the name, i915_gem_object_flush_if_display

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/20170222114049.28456-3-chris@chris-wilson.co.uk

drm/i915: Move cpu_cache_is_coherent() to header

For use in the next patch, take the current is-coherent helper and add
it to i915_gem_object.h

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/20170222114049.28456-2-chris@chris-wilson.co.uk

drm/i915: Remove change_domain tracepoint

The change_domain tracepoint has been inaccurate for a few years - it
doesn't fully capture the domains, especially with userspace bypassing
them. It is defunct, misleading and time to be removed.

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/20170222114049.28456-1-chris@chris-wilson.co.uk

drm/i915: Add i915_param charp macro magic

Handling the dynamic charp module parameter requires us to copy it for
the error state, or remember to lock it when reading (in case it used
with 0600).

v2: Use __always_inline and __builtin_strcmp

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/20170221162619.15954-1-chris@chris-wilson.co.uk
Reviewed-by: Jani Nikula <jani.nikula@intel.com>

drm/i915/gvt: set ring buffer size to default for guc submission

When not using GuC submission, the ring buffer size for GVT context is
512KB which is the max size. When switching to GuC submission, the ring
buffer size is required to be less than 16KB. So use the GVT context
default ring buffer size if GuC submission is enabled.

Signed-off-by: Chuanxiao Dong <chuanxiao.dong@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170216063639.GA17107@intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Add hw_id to context tracepoints

It is useful to provide this info to match the one provided
in the request tracepoints.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170221091350.14605-1-tvrtko.ursulin@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Add backend level request in and out tracepoints

Two new tracepoints placed at the call sites where requests are
actually passed to the GPU enable userspace to track engine
utilisation.

These tracepoints are only enabled when the
DRM_I915_LOW_LEVEL_TRACEPOINTS Kconfig option is enabled.

v2: Fix compilation with !CONFIG_DRM_I915_LOW_LEVEL_TRACEPOINTS.

v3: Name global seqno consistently across tracepoints.

v4: Remove port info from request out tracepoint. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Rename i915_gem_request_notify

i915_gem_ring_notify is more appropriate since we do not have
the request information at this point, but it is simply a
signal from the engine that some request has been completed.

v2:
  * Always trace and log if there were any waiters.
  * Rename to intel_engine_notify. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Add request submit and execute tracepoints

These new tracepoints are emitted once the request is ready to
be submitted to the GPU and once the request is about to
be submitted to the GPU, respectively.

Former condition triggers as soon as all the fences and
dependencies have been resolved, and the latter once the
backend is about to submit it to the GPU.

New tracepoint are enabled via the new
DRM_I915_LOW_LEVEL_TRACEPOINTS Kconfig option which is disabled
by default to alleviate the performance impact concerns.

v2: Move execute tracepoint to __i915_gem_request_submit.
    (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Remove unused i915_gem_request_complete

Tracepoint is not used and won't be suitable for its replacement.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Tidy i915_gem_request_wait_begin

Provide the same information as the other request event classes.

v2: Pass in flags so we can properly report the blocking status.
    (Chris Wilson)

v3: Log hex with 0x prefix for clarity.

v4: Derive blocking status from flags. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Adjust i915_gem_ring_dispatch

Rename it to i915_gem_request_queue and fix the logged info
equivalent to the i915_gem_request even class. Also moved it
a bit further apart from the i915_gem_request_add tracepoint
since they otherwise provide similar information too close in
time.

v2: Remove sw fence singalling. We will rely on the soon to
    come GuC scheduling backend to enable that. (Chris Wilson)

v3: Log hex with 0x prefix for clarity.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/tracepoints: Tidy request event class

At the moment only the global seqno is logged which is not set
until the request is ready for submission.

Add the per-contex seqno and the context hardware id which are
both interesting data points.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Tidy execlists_init_reg_state

Compact the name of the macro and reg_state variable, and cache
some data in local variables to make the function more compact
and more readable.

v2: Fixup some checkpatch warnings.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170221095839.30525-1-tvrtko.ursulin@linux.intel.com

drm/i915: Use reservation_object_lock()

Replace the calls to ww_mutex_lock(&resv->lock) with the helper
reservation_object_lock(resv) and similarly for unlock.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170221091723.6219-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Assert that the request->tail is always qword aligned

The hardware requires that the tail pointer only advance in qword units,
so assert that the value we write is aligned to qwords, and similarly
enforce this restriction onto the request->tail.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217163833.731-1-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>

drm/i915: Add power well SW/HW state verification

Verify that the refcount of all power wells match their HW enabled
state at the end of modeset HW state readout.

Also add documentation on how the reference count for each power well is
supposed to be acquired during initialization and HW state readout.

Suggested by Ander.

Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487345986-26511-6-git-send-email-imre.deak@intel.com

drm/i915: Preserve the state of power wells not explicitly enabled

Atm, power wells that BIOS has enabled, but which we don't explicitly
enable during power domain initialization would get disabled as we clear
the BIOS request bit in the given power well sync_hw hook. To prevent
this copy over any set request bits in the BIOS request register to the
driver request register and clear the BIOS request bit only afterwards.

This doesn't make a difference now, since we enable all power wells
during power domain initialization. A follow-up patchset will add power
wells for which this isn't true, so fix up the inconsistency.

Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487345986-26511-5-git-send-email-imre.deak@intel.com

drm/i915/gen9: Fix clearing of the BIOS power well request register

Atm, in the power well sync_hw hook we are clearing all BIOS request
bits, not just the one corresponding to the given power well. This could
turn off an unrelated power well inadvertently if it didn't have a
request bit set in the driver request register.

This didn't cause a problem so far, since we enabled all power wells
explicitly before clearing the BIOS request register. A follow-up
patchset will add power wells that won't get enabled this way, so fix up
the inconsistency.

Note that this patch only makes the clearing of the BIOS req register
more logical. Power wells without a reference would still get disabled
by the end of power domain initialization, that is fixed by the next
patch.

v2:
- Clarify in the commit log that this patch doesn't address the case of
  power wells without a reference. (Ander)

Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487345986-26511-4-git-send-email-imre.deak@intel.com

drm/i915: Call the sync_hw hook for power wells without a domain

So far the sync_hw hook wasn't called for power wells not belonging to
any power domain, that is the GEN9 PW1 and MISC_IO power wells. This
wasn't a problem so far since the goal of the sync_hw hook - to clear
the corresponding BIOS request bit - was guaranteed by clearing the
whole BIOS request register elsewhere. This will change with the next
patch, so fix up the inconsistency.

While at it clean up the power well iterator helpers and move them to
the rest of iterators.

v2:
- Clean up the power well iterator helpers. (Ander)
- Move the helpers to i915_drv.h.

Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487345986-26511-3-git-send-email-imre.deak@intel.com

drm/i915: Remove redundant toggling from the power well sync_hw hooks

Doing an explicit enable/disable in the power well sync_hw hook based on
the power well's reference count is redundant, since by the time these
hooks are called all the power wells are enabled and have a reference.
So remove the redundant toggling.

This is needed by a follow-up patchset that adds power wells which we
can't enable/disable during power domain initialization and so want to
preserve their state until modeset init time.

Cc: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Cc: David Weinehall <david.weinehall@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487345986-26511-2-git-send-email-imre.deak@intel.com

drm/i915: Remove unrequired POSTING_READ from gen6_set_rps()

The uncached mmio is sufficient to queue the mmio writes without raising
forcewake. The forced flush along with acquiring forcewake from the
posting read is not required for adjusting the RPS frequency.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170220094713.22874-3-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: Store the requested frequency whilst RPS is disabled

If intel_set_rps() is called whilst the hw is disabled, just store the
requested frequency (from the user) for application when we wake the hw
up.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170220094713.22874-2-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: Move the common RPS warnings to intel_set_rps()

Instead of having each back-end provide identical guards, just have a
singular set in intel_set_rps() to verify that the caller is obeying the
rules.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170220094713.22874-1-chris@chris-wilson.co.uk

drm/i915: Fix typo in semaphore debug message

Pronounces less rude when fixed.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170220104657.5237-1-tvrtko.ursulin@linux.intel.com

drm/i915: Prevent divide-by-zero in debugfs/i915_rps_boost_info

Either by chance, or by misread, the current evaluation interval may be
zero. If that is the case, don't divide by it!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170218112708.24504-1-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: Remove unneeded struct_mutex around rpm

We don't need struct_mutex for acquiring an rpm wakeref, and do not need
to serialise those register read (it's the wrong mutex for those
registers in any case). Begone!

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170218150050.10414-1-chris@chris-wilson.co.uk
Reviewed-by: Radoslaw Szwichtenberg <radoslaw.szwichtenberg@intel.com>

drm/i915: Only apply legacy PDE overflow detection to 3lvl machines

Prevent the overflow check from firing on machines with the full 4lvl
page tables, that are not restricted to GEN8_LEGACY_PDES.

v2: Also fix the off-by-one in the compare

Fixes: 894ccebee2b0 ("drm/i915: Micro-optimise gen8_ppgtt_insert_entries()")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217141455.19877-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: DMC 1.03 for Geminilake

There is a new version of DMC available for Geminilake.

It's release notes only mention:
- Enhancement in the FW to restore the PG2 state

v2: Fixed the platform name on commit message.
    Noticed by Jani S.
v3: cook on top of drm-tip without depending on kbl
    one so CI can check.
v4: make v3 on top of v2.

Cc: David Weinehall <tao@kernel.org>
Cc: Jani Saarinen <jani.saarinen@intel.com>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487295515-15396-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Remove completed fences after a wait

If we wait upon the full (i.e. all shared fences, or upon an exclusive
fence) reservation object successfully, we know that all fences beneath
it have been signaled, so long as no new fences were added whilst we
slept. If the reservation_object remains the same, as detected by its
seqcount, we can then reap all the fences upon completion.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217151304.16665-6-chris@chris-wilson.co.uk

drm/i915: Only start with the fake-irq timer if interrupts are dead

As a backup to waiting on a user-interrupt from the GPU, we use a heavy
and frequent timer to wake up the waiting process should we detect an
inconsistency whilst waiting. After seeing a "missed interrupt", the
next time we wait, we restart the heavy timer. This patch is more
reluctant to restart the timer and will only do so if we have not see any
interrupts since when we started the fake irq timer. If we are seeing
interrupts, then the waiters are being woken normally and we had an
incoherency that caused to miss last time - that is unlikely to reoccur
and so taking the risk of stalling again seems pragmatic.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217151304.16665-5-chris@chris-wilson.co.uk

drm/i915: Defer declaration of missed-interrupt until the waiter is asleep

If the waiter was currently running, assume it hasn't had a chance
to process the pending interrupt (e.g, low priority task on a loaded
system) and wait until it sleeps before declaring a missed interrupt.

References: https://bugs.freedesktop.org/show_bug.cgi?id=99816
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217151304.16665-4-chris@chris-wilson.co.uk

drm/i915: Break i915_spin_request() if we see an interrupt

If an interrupt has been posted, and we were spinning on the active
seqno waiting for it to advance but it did not, then we can expect that
it will not see its advance in the immediate future and should call into
the irq-seqno barrier. We can stop spinning at this point, and leave the
difficulty of handling the coherency to the caller.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217151304.16665-3-chris@chris-wilson.co.uk

drm/i915: Postpone fake breadcrumb interrupt until real interrupts cease

When the timer expires for checking on interrupt processing, check to
see if any interrupts arrived within the last time period. If real
interrupts are still being delivered, we can be reassured that we
haven't missed the final interrupt as the waiter will still be woken.
Only once all activity ceases, do we have to worry about the waiter
never being woken and so need to install a timer to kick the waiter for
a slow arrival of a seqno.

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/20170217151304.16665-2-chris@chris-wilson.co.uk

drm/i915/glk: Enable pipe CSC

Now that the pre-csc degamma table is set up correctly in Geminilake,
pipe CSC can be enabled without causing a black screen.

v2: Rebase.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217120630.6143-3-ander.conselvan.de.oliveira@intel.com

drm/i915/glk: Load the degamma LUT even in legacy gamma mode

In Geminilake, the degamma table is enabled or disabled by the pipe CSC
enable bit, so its active even when running in the legacy gamma mode.
So always set sane values for that table, since the default value is all
zeroes.

This fixes blank screens after a suspend/resume cycle while legacy gamma
is in use.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170217120630.6143-2-ander.conselvan.de.oliveira@intel.com

drm/i915: Consolidate gen8_emit_pipe_control

We have a few open coded instances in the execlists code and an
almost suitable helper in intel_ringbuf.c

We can consolidate to a single helper if we change the existing
helper to emit directly to ring buffer memory and move the space
reservation outside it.

v2: Drop memcpy for memset. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170216122325.31391-2-tvrtko.ursulin@linux.intel.com

drm/i915: Tidy workaround batch buffer emission

Use the "*batch++ = " style as in the ring emission for better
readability and also simplify the logic a bit by consolidating
the offset and size calculations and overflow checking. The
latter is a programming error so it is not required to check
for it after each write to the object, but rather do it once the
whole state has been written and fail the driver if something
went wrong.

v2: Rebase.

v3: Keep track of offsets and sizes in bytes for simplicity
    and rename function pointer variable to _fn suffix.
    (Chris Wilson)

v4: Fix size calc broken in v3 and add alignment warning. (Chris Wilson)

v5: Fix return code.

v6: I added an exit from loop in v5 but forgot to put back
    the object teardown.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v5)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Move common workaround code to intel_engine_cs

It is used by all submission backends.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Simplify cleanup path in intel_engines_init

We can call the engine cleanup vfunc instead of duplicating the
decision making here.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Make int __intel_ring_space static

It is only used within intel_ringbuffer.c

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.oc.uk>

drm/i915: Drop struct_mutex around frontbuffer flushes

Since the frontbuffer has self-contained locking, it does not require us
to hold the BKL struct_mutex as we send invalidate and flush messages.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215105919.7347-3-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: struct_mutex is not required for allocating the framebuffer

We do not need the BKL struct_mutex in order to allocate a GEM object,
nor to create the framebuffer, so resist the temptation to take the BKL
willy nilly. As this changes the locking contract around internal API
calls, the patch is a little larger than a plain removal of a pair of
mutex_lock/unlock.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215105919.7347-2-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Remove struct_mutex for destroying framebuffers

We do not need to hold struct_mutex for destroying drm_i915_gem_objects
any longer, and with a little care taken over tracking
obj->framebuffer_references, we can relinquish BKL locking around the
destroy of intel_framebuffer.

v2: Use atomic check for WARN_ON framebuffer miscounting

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/20170216094621.3426-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Unwind conversion to i915_gem_phys_ops on failure

The physical object is treated as permanently pinned. If we fail to take
this initial pin during i915_gem_object_attach_phys() we need to revert
it back to an ordinary shmemfs object before reporting the failure.

v2: git-add

Reported-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215163900.11606-1-chris@chris-wilson.co.uk
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Squelch any ktime/jiffie rounding errors for wait-ioctl

We wait upon jiffies, but report the time elapsed using a
high-resolution timer. This discrepancy can lead to us timing out the
wait prior to us reporting the elapsed time as complete.

This restores the squelching lost in commit e95433c73a11 ("drm/i915:
Rearrange i915_wait_request() accounting with callers").

Fixes: e95433c73a11 ("drm/i915: Rearrange i915_wait_request() accounting with callers")
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/20170216125441.30923-1-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915/bxt: Disable device ready before shutdown command

Disable device ready before MIPI port shutdown command.
This helps to avoid mipi split screen issues.

Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Vidya Srinivas <vidya.srinivas@intel.com>
Reviewed-by: Bob Paauwe <bob.j.paauwe@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486551058-22596-8-git-send-email-vidya.srinivas@intel.com

drm/i915/bxt: Fix BXT DSI ULPS sequence

Fix the Sequence to program BXT DSI Latch and ULPS.

Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Vidya Srinivas <vidya.srinivas@intel.com>
Reviewed-by: Bob Paauwe <bob.j.paauwe@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486551058-22596-6-git-send-email-vidya.srinivas@intel.com

drm/i915: Set the Z inversion overlap field

Dual link Z-inversion overlap field is present
in MIPI_CTRL register unlike the older platforms,
hence setting the same in this patch.

Signed-off-by: Deepak M <m.deepak@intel.com>
Signed-off-by: Madhav Chauhan <madhav.chauhan@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487078180-15147-5-git-send-email-madhav.chauhan@intel.com

drm/i915: Do RPM Wake during GuC/HuC status read

HUC_STATUS, GUC_STATUS, SOFT_SCRATCH registers are read in debugfs
and getparam ioctl. This patch covers those accesses by RPM get/put.

v2: Covering access in i915_getparam(I915_PARAM_HUC_STATUS) (ChrisW)

Cc: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Fiedorowicz, Lukasz <lukasz.fiedorowicz@intel.com>
Signed-off-by: Sagar Arun Kamble <sagar.a.kamble@intel.com>
Reviewed-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486110513-12130-1-git-send-email-sagar.a.kamble@intel.com

drm/i915: Only enable hotplug interrupts if the display interrupts are enabled

In order to prevent accessing the hpd registers outside of the display
power wells, we should refrain from writing to the registers before the
display interrupts are enabled.

[    4.740136] WARNING: CPU: 1 PID: 221 at drivers/gpu/drm/i915/intel_uncore.c:795 __unclaimed_reg_debug+0x44/0x50 [i915]
[    4.740155] Unclaimed read from register 0x1e1110
[    4.740168] Modules linked in: i915(+) intel_gtt drm_kms_helper prime_numbers
[    4.740190] CPU: 1 PID: 221 Comm: systemd-udevd Not tainted 4.10.0-rc6+ #384
[    4.740203] Hardware name:                  /        , BIOS PYBSWCEL.86A.0027.2015.0507.1758 05/07/2015
[    4.740220] Call Trace:
[    4.740236]  dump_stack+0x4d/0x6f
[    4.740251]  __warn+0xc1/0xe0
[    4.740265]  warn_slowpath_fmt+0x4a/0x50
[    4.740281]  ? insert_work+0x77/0xc0
[    4.740355]  ? fwtable_write32+0x90/0x130 [i915]
[    4.740431]  __unclaimed_reg_debug+0x44/0x50 [i915]
[    4.740507]  fwtable_read32+0xd8/0x130 [i915]
[    4.740575]  i915_hpd_irq_setup+0xa5/0x100 [i915]
[    4.740649]  intel_hpd_init+0x68/0x80 [i915]
[    4.740716]  i915_driver_load+0xe19/0x1380 [i915]
[    4.740784]  i915_pci_probe+0x32/0x90 [i915]
[    4.740799]  pci_device_probe+0x8b/0xf0
[    4.740815]  driver_probe_device+0x2b6/0x450
[    4.740828]  __driver_attach+0xda/0xe0
[    4.740841]  ? driver_probe_device+0x450/0x450
[    4.740853]  bus_for_each_dev+0x5b/0x90
[    4.740865]  driver_attach+0x19/0x20
[    4.740878]  bus_add_driver+0x166/0x260
[    4.740892]  driver_register+0x5b/0xd0
[    4.740906]  ? 0xffffffffa0166000
[    4.740920]  __pci_register_driver+0x47/0x50
[    4.740985]  i915_init+0x5c/0x5e [i915]
[    4.740999]  do_one_initcall+0x3e/0x160
[    4.741015]  ? __vunmap+0x7c/0xc0
[    4.741029]  ? kmem_cache_alloc+0xcf/0x120
[    4.741045]  do_init_module+0x55/0x1c4
[    4.741060]  load_module+0x1f3f/0x25b0
[    4.741073]  ? __symbol_put+0x40/0x40
[    4.741086]  ? kernel_read_file+0x100/0x190
[    4.741100]  SYSC_finit_module+0xbc/0xf0
[    4.741112]  SyS_finit_module+0x9/0x10
[    4.741125]  entry_SYSCALL_64_fastpath+0x17/0x98
[    4.741135] RIP: 0033:0x7f8559a140f9
[    4.741145] RSP: 002b:00007fff7509a3e8 EFLAGS: 00000246 ORIG_RAX: 0000000000000139
[    4.741161] RAX: ffffffffffffffda RBX: 00007f855aba02d1 RCX: 00007f8559a140f9
[    4.741172] RDX: 0000000000000000 RSI: 000055b6db0914f0 RDI: 0000000000000011
[    4.741183] RBP: 0000000000020000 R08: 0000000000000000 R09: 000000000000000e
[    4.741193] R10: 0000000000000011 R11: 0000000000000246 R12: 000055b6db0854d0
[    4.741204] R13: 000055b6db091150 R14: 0000000000000000 R15: 000055b6db035924

v2: Set dev_priv->display_irqs_enabled to true for all platforms other
than vlv/chv that manually control the display power domain.

Fixes: 19625e85c6ec ("drm/i915: Enable polling when we don't have hpd")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97798
Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Lyude <cpaul@redhat.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Hans de Goede <jwrdegoede@fedoraproject.org>
Cc: stable@vger.kernel.org
Link: http://patchwork.freedesktop.org/patch/msgid/20170215131547.5064-1-chris@chris-wilson.co.uk
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/bxt: remove WaRsDisableCoarsePowerGating for early BXT

No need to cater for old A revisions.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487172099-24873-4-git-send-email-jani.nikula@intel.com

drm/i915/bxt: remove snooping workaround on old A revisions

No need to cater for old A revisions.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487172099-24873-3-git-send-email-jani.nikula@intel.com

drm/i915/bxt: apply clock gating workaround to all revisions

No need to cater for old A revisions.

Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1487172099-24873-2-git-send-email-jani.nikula@intel.com

drm/i915: Fix PLL 8x/3 divider for MIPI video mode

MIPI Video Mode for high res panels (requiring dual link), need a
8X/3 divider to be programmed as 0x2. Modifying the same
in this patch.

Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Vidya Srinivas <vidya.srinivas@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486551058-22596-3-git-send-email-vidya.srinivas@intel.com

drm/i915: Check for platform specific GPIO config

Panel GPIO control should be done based on platform. Add a check
to restrict VLV and CHT specific GPIO confirguration, so that
they dont apply to other platforms.

The VBT spec fails to mention the PMIC backlight control option is valid
only for VLV/CHT, and the field may be set to "PMIC" for BXT even if
PMIC is not desired or possible.

Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Vidya Srinivas <vidya.srinivas@intel.com>
[Jani: amended commit message a bit and fixed indentation.]
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486551058-22596-2-git-send-email-vidya.srinivas@intel.com

drm/i915/dp: Reset the link params on HPD/connected boot/resume

The max link parameters should be set/reset only on HPD or
connected boot case or on system resume.

Add a flag reset_link_params to intel_dp to decide when
to reset the max link parameters. This prevents the parameters
from getting reset/overwritten through all other
connector->funcs->detect() calls. This is important when link
training fails and the max link params are modified to the
lower fallback values.

Cc: Ville Syrjala <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Manasi Navare <manasi.d.navare@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1486515251-23469-1-git-send-email-manasi.d.navare@intel.com

drm/i915: Use preferred kernel types in i915_gem_gtt.c

Make checkpatch happy and make the use of u32/u64 consistent throughout
i915_gem_gtt.[ch]

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-23-chris@chris-wilson.co.uk

drm/i915: Differentiate the aliasing_ppgtt with an invalid filp

Use an invalid filp so that the aliasing_ppgtt can be clearly
identified.

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/20170215084357.19977-22-chris@chris-wilson.co.uk

drm/i915: Only preallocate the aliasing GTT to the extents of the global GTT

As the aliasing GTT is only accessed via the global GTT, we will never
use more of it than we expose via the Global GTT and so we only need to
preallocate sufficient space within the ppgtt for the full GTT. Equally,
if the aliasing GTT is smaller than the global GTT, we have a serious
issue and must bail.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-21-chris@chris-wilson.co.uk

drm/i915: Remove i915_address_space.start

Once upon a time, back in the UMS days, we supported userspace
initialising the GTT and sharing portions of the GTT with other users.
Now, we own the GTT (both global and per-process) and the tables always
start at 0 - so we can remove i915_address_space.start and forget about
this old complication.

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/20170215084357.19977-20-chris@chris-wilson.co.uk

drm/i915: Remove unused ppgtt->enable()

We never assign or use the ppgtt->enable() callback, so remove 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/20170215084357.19977-19-chris@chris-wilson.co.uk

drm/i915: Remove defunct GTT tracepoints

The tracepoints are now entirely synonymous with binding and unbinding the
VMA (and the tracepoints there).

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/20170215084357.19977-18-chris@chris-wilson.co.uk

drm/i915: Always mark the PDP as dirty when altered

We want to reload the PDP (and flush the TLB) when the addresses are
changed.

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/20170215084357.19977-17-chris@chris-wilson.co.uk

drm/i915: Remove superfluous posting reads after clear GGTT

The barrier here is not required - we apply the barrier before the range
is ever reused by the GPU instead.

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/20170215084357.19977-16-chris@chris-wilson.co.uk

drm/i915: Remove bitmap tracking for used-pml4

We only operate on known extents (both for alloc/clear) and so we can use
both the knowledge of the bind/unbind range along with the knowledge of
the existing pagetable to avoid having to allocate temporary and
auxiliary bitmaps.

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/20170215084357.19977-15-chris@chris-wilson.co.uk

drm/i915: Remove bitmap tracking for used-pdpes

We only operate on known extents (both for alloc/clear) and so we can use
both the knowledge of the bind/unbind range along with the knowledge of
the existing pagetable to avoid having to allocate temporary and
auxiliary bitmaps.

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/20170215084357.19977-14-chris@chris-wilson.co.uk

drm/i915: Remove bitmap tracking for used-pdes

We only operate on known extents (both for alloc/clear) and so we can use
both the knowledge of the bind/unbind range along with the knowledge of
the existing pagetable to avoid having to allocate temporary and
auxiliary bitmaps.

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/20170215084357.19977-13-chris@chris-wilson.co.uk

drm/i915: Remove bitmap tracking for used-ptes

We only operate on known extents (both for alloc/clear) and so we can use
both the knowledge of the bind/unbind range along with the knowledge of
the existing pagetable to avoid having to allocate temporary and
auxiliary bitmaps.

Bugzilla: 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>
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20170215084357.19977-12-chris@chris-wilson.co.uk

drm/i915: Tidy gen6_write_pde()

Stop passing around unused parameters makes the code more compact.

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/20170215084357.19977-11-chris@chris-wilson.co.uk

drm/i915: Remove redundant clear of appgtt

Upon creation of the va range, it is initialised to point at scratch.

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/20170215084357.19977-10-chris@chris-wilson.co.uk

drm/i915: Always preallocate gen6/7 ppgtt

The hardware does not cope very well with us changing the PD within an
active context (the context must be idle for it to re-read the PD). As
we only check whether the page is idle before changing the entry (and on
through the PD tree), we cannot reliably replace PD entries on
gen6/gen7. To fully avoid changing the tree at runtime, preallocate it
on init.

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/20170215084357.19977-9-chris@chris-wilson.co.uk

drm/i915: Move allocate_va_range to GTT

In the future, we need to call allocate_va_range on the aliasing-ppgtt
which means moving the call down from the vma into the vm (which is
more appropriate for calling the vm function).

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/20170215084357.19977-8-chris@chris-wilson.co.uk

drm/i915: Remove kmap/kunmap wrappers

As these are now both plain and simple kmap_atomic/kunmap_atomic pairs,
we can remove the wrappers for a small gain of clarity (in particular,
not hiding the atomic critical sections!).

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/20170215084357.19977-7-chris@chris-wilson.co.uk

drm/i915: Convert clflushed pagetables over to WC maps

We flush the entire page every time we update a few bytes, making the
update of a page table many, many times slower than is required. If we
create a WC map of the page for our updates, we can avoid the clflush
but incur additional cost for creating the pagetable. We amoritize that
cost by reusing page vmappings, and only changing the page protection in
batches.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>

drm/i915: Split ggtt/alasing_gtt unbind_vma

Similar to how we already split the bind_vma for ggtt/aliasing_gtt, also
split up the unbind for symmetry.

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/20170215084357.19977-5-chris@chris-wilson.co.uk

drm/i915: Don't special case teardown of aliasing_ppgtt

The aliasing_ppgtt is a regular ppgtt, and we can use the regular
i915_ppgtt_put() to properly tear it down.

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/20170215084357.19977-4-chris@chris-wilson.co.uk

drm/i915: Micro-optimise gen8_ppgtt_insert_entries()

Improve the sg iteration and in hte process eliminate a bug in
miscomputing the pml4 length as orig_nents<<PAGE_SHIFT is no longer the
full length of the sg table.

v2: Check for the end of the fourth level page table (the final pdpe)
and move onto the next.
v3: Assert that 3lvl insert_pte_entries doesn't overflow its smaller set
of PDP.

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/20170215084357.19977-3-chris@chris-wilson.co.uk

drm/i915: Micro-optimise gen6_ppgtt_insert_entries()

Inline the address computation to avoid the vfunc call for every page.
We still have to pay the high overhead of sg_page_iter_next(), but now
at least GCC can optimise the inner most loop, giving a significant
boost to some thrashing Unreal Engine workloads.

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/20170215084357.19977-2-chris@chris-wilson.co.uk