drm/i915/gen9: Drop re-allocation of DDB at atomic commit (v2)

Now that we're properly pre-allocating the DDB during the atomic check
phase and we trust that the allocation is appropriate, let's actually
use the allocation computed and not duplicate that work during the
commit phase.

v2:
 - Significant rebasing now that we can use cached data rates and
   minimum block allocations to avoid grabbing additional plane states.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-11-git-send-email-matthew.d.roper@intel.com

drm/i915/gen9: Compute DDB allocation at atomic check time (v4)

Calculate the DDB blocks needed to satisfy the current atomic
transaction at atomic check time.  This is a prerequisite to calculating
SKL watermarks during the 'check' phase and rejecting any configurations
that we can't find valid watermarks for.

Due to the nature of DDB allocation, it's possible for the addition of a
new CRTC to make the watermark configuration already in use on another,
unchanged CRTC become invalid.  A change in which CRTC's are active
triggers a recompute of the entire DDB, which unfortunately means we
need to disallow any other atomic commits from racing with such an
update.  If the active CRTC's change, we need to grab the lock on all
CRTC's and run all CRTC's through their 'check' handler to recompute and
re-check their per-CRTC DDB allocations.

Note that with this patch we only compute the DDB allocation but we
don't actually use the computed values during watermark programming yet.
For ease of review/testing/bisecting, we still recompute the DDB at
watermark programming time and just WARN() if it doesn't match the
precomputed values.  A future patch will switch over to using the
precomputed values once we're sure they're being properly computed.

Another clarifying note:  DDB allocation itself shouldn't ever fail with
the algorithm we use today (i.e., we have enough DDB blocks on BXT to
support the minimum needs of the worst-case scenario of every pipe/plane
enabled at full size).  However the watermarks calculations based on the
DDB may fail and we'll be moving those to the atomic check as well in
future patches.

v2:
 - Skip DDB calculations in the rare case where our transaction doesn't
   actually touch any CRTC's at all.  Assuming at least one CRTC state
   is present in our transaction, then it means we can't race with any
   transactions that would update dev_priv->active_crtcs (which requires
   _all_ CRTC locks).

v3:
 - Also calculate DDB during initial hw readout, to prevent using
   incorrect bios values. (Maarten)

v4:
 - Use new distrust_bios_wm flag instead of skip_initial_wm (which was
   never actually set).
 - Set intel_state->active_pipe_changes instead of just realloc_pipes

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Lyude Paul <cpaul@redhat.com>
Cc: Radhakrishna Sripada <radhakrishna.sripada@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-10-git-send-email-matthew.d.roper@intel.com

drm/i915: Add distrust_bios_wm flag to dev_priv (v2)

SKL-style platforms can't fully trust the watermark/DDB settings
programmed by the BIOS and need to do extra sanitization on their first
atomic update.  Add a flag to dev_priv that is set during hardware
readout and cleared at the end of the first commit.

Note that for the somewhat common case where everything is turned off
when the driver starts up, we don't need to bother with a recompute...we
know exactly what the DDB should be (all zero's) so just setup the DDB
directly in that case.

v2:
 - Move clearing of distrust_bios_wm up below the swap_state call since
   it's a more natural / self-explanatory location.  (Maarten)
 - Use dev_priv->active_crtcs to test whether any CRTC's are turned on
   during HW WM readout rather than trying to count the active CRTC's
   again ourselves.  (Maarten)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-9-git-send-email-matthew.d.roper@intel.com

drm/i915/gen9: Allow skl_allocate_pipe_ddb() to operate on in-flight state (v3)

We eventually want to calculate watermark values at atomic 'check' time
instead of atomic 'commit' time so that any requested configurations
that result in impossible watermark requirements are properly rejected.
The first step along this path is to allocate the DDB at atomic 'check'
time.  As we perform this transition, allow the main allocation function
to operate successfully on either an in-flight state or an
already-commited state.  Once we complete the transition in a future
patch, we'll come back and remove the unnecessary logic for the
already-committed case.

v2: Rebase/refactor; we should no longer need to grab extra plane states
    while allocating the DDB since we can pull cached data rates and
    minimum block counts from the CRTC state for any planes that aren't
    being modified by this transaction.

v3:
 - Simplify memsets to clear DDB plane entries.  (Maarten)
 - Drop a redundant memset of plane[pipe][PLANE_CURSOR] that was added
   by an earlier Coccinelle patch.  (Maarten)
 - Assign *num_active at the top of skl_ddb_get_pipe_allocation_limits()
   so that no code paths return without setting it.  (kbuild robot)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-8-git-send-email-matthew.d.roper@intel.com

drm/i915: Track whether an atomic transaction changes the active CRTC's

For the purposes of DDB re-allocation we need to know whether a
transaction changes the list of CRTC's that are active.  While
state->modeset could be used for this purpose, that would be slightly
too aggressive since it would lead us to re-allocate the DDB when a
CRTC's mode changes, but not its final active state.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-7-git-send-email-matthew.d.roper@intel.com

drm/i915/gen9: Store plane minimum blocks in CRTC wm state (v2)

This will eventually allow us to re-use old values without
re-calculating them for unchanged planes (which also helps us avoid
re-grabbing extra plane states).

v2:
 -  Drop unnecessary memset's; they were meant for a later patch (which
    got reworked anyway to not need them, but were mis-rebased into this
    one.  (Maarten)

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-6-git-send-email-matthew.d.roper@intel.com

drm/i915/gen9: Allow calculation of data rate for in-flight state (v2)

Our skl_get_total_relative_data_rate() function gets passed a crtc state
object to calculate the data rate for, but it currently always looks
up the committed plane states that correspond to that CRTC.  Let's
check whether the CRTC state is an in-flight state (meaning
cstate->state is non-NULL) and if so, use the corresponding in-flight
plane states.

We'll soon be using this function exclusively for in-flight states; at
that time we'll be able to simplify the function a bit, but for now we
allow it to be used in either mode.

v2:
 - Rebase on top of changes to cache plane data rates.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-5-git-send-email-matthew.d.roper@intel.com

drm/i915/gen9: Cache plane data rates in CRTC state

This will be important when we start calculating CRTC data rates for
in-flight CRTC states since it will allow us to calculate the total data
rate without needing to grab the plane state for any planes that aren't
updated by the transaction.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-4-git-send-email-matthew.d.roper@intel.com

drm/i915: Rename s/skl_compute_pipe_wm/skl_build_pipe_wm/

When we added atomic watermarks, we added a new display vfunc
'compute_pipe_wm' that is used to compute any pipe-specific watermark
information that we can at atomic check time.  This was a somewhat poor
naming choice since we already had a 'skl_compute_pipe_wm' function that
doesn't quite fit this model --- the existing SKL function is something
that gets used at atomic commit time, after the DDB allocation has been
determined.  Let's rename the existing SKL function to avoid confusion.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-3-git-send-email-matthew.d.roper@intel.com

drm/i915: Reorganize WM structs/unions in CRTC state

Reorganize the nested structures and unions we have for pipe watermark
data in intel_crtc_state so that platform-specific data can be added in
a more sensible manner (and save a bit of memory at the same time).

The change basically changes the organization from:

        union {
                struct intel_pipe_wm ilk;
                struct intel_pipe_wm skl;
        } optimal;

        struct intel_pipe_wm intermediate /* ILK-only */

to

        union {
                struct {
                        struct intel_pipe_wm intermediate;
                        struct intel_pipe_wm optimal;
                } ilk;

                struct {
                        struct intel_pipe_wm optimal;
                } skl;
        }

There should be no functional change here, but it will allow us to add
more platform-specific fields going forward (and more easily extend to
other platform types like VLV).

While we're at it, let's move the entire watermark substructure out to
its own structure definition to make the code slightly more readable.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463061971-19638-2-git-send-email-matthew.d.roper@intel.com

drm/i915: Remove redundant const from function return type

Marking function return types as const is redundant, as these are
rvalues and as such constant by definition. Code checkers and GCC will
warn about this so remove the modifier.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463059132-1720-5-git-send-email-imre.deak@intel.com

drm/i915: Handle error return from dma_set_coherent_mask()

A failure from these functions can lead to obscure bugs later, so it's
better not to suppress them and just fail module loading.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463059132-1720-4-git-send-email-imre.deak@intel.com

drm/i915: Add comments about fixed pipe->transcoder/PLL mapping

Code checkers may complain about the explicit casts between different
enum types, so add comments for known-valid cases to help future
triaging of such complaints.

v2:
- Make the comments more logical (Ville).

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463059132-1720-3-git-send-email-imre.deak@intel.com

drm/i915/gen9: Avoid using negative array index in skl_update_plane()

scaler_id may be negative as shown by conditions later in the function,
so don't use it as an array index in that case.

v2:
- Remove ps_ctrl while at it (Ville).

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463059132-1720-2-git-send-email-imre.deak@intel.com

drm/i915: Stop automatically retiring requests after a GPU hang

Following a GPU hang, we break out of the request loop in order to
unlock the struct_mutex for use by the GPU reset. However, if we retire
all the requests at that moment, we cannot identify the guilty request
after performing the reset.

v2: Not automatically retiring requests forces us to recheck for
available ringspace.

Fixes: f4457ae71fd6 ("drm/i915: Prevent leaking of -EIO from i915_wait_request()")
Testcase: igt/gem_reset_stats/ban-*
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Tested-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463137042-9669-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Stop retiring requests from busy/wait ioctls

In order to reduce the workload of the caller, we do not want to
actually have to retire requests of others when checking the busy status
of this object. This applies to both busy/wait ioctls as the wait ioctl
has a semantically equivalent mode to the busy ioctl.

At the present time, this is only a minor improvement to reduce the
workload of the busy ioctl under the struct_mutex which helps to reduce
its impact upon contention of struct_mutex. However, since it is mostly
a victim in highly contended scenarios, the impact is very minor until
we can eliminate the struct_mutex requirement for busy-ioctl in the near
future.

v2: Mention the patches intended limited impact. It is just paving the
way for greater changes whilst reducing the impact of a bugfix in the
next patch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463137042-9669-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Complete pending resets before get-reset-stats ioctl

The get-reset-stats ioctls wasn't waiting for a pending reset before
reporting its statistics, and so was ignoring a hang generated by the
context that should have been reported against said context.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1463137042-9669-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move get-reset-stats ioctl from intel_uncore.c to i915_gem_context.c

The get-reset-stats ioctl reports upon the statistics (number of hangs,
be it as a victim or the guilty party) of a particular context. It is
semantically better as being part of i915_gem_context.c user interface,
as opposed to the hardware level access of intel_uncore.c

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/1463137042-9669-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Remove intel_limit_t typedef

The coding style documentation says the following about typedefs:

"In general, a pointer, or a struct that has elements that can
reasonably be directly accessed should _never_ be a typedef."

intel_limit_t falls in that category, so just use "struct intel_limit"
instead.

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/1462353119-9738-3-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Remove intel_range_t and intel_p2_t typedefs

Those are only used for defining struct intel_limit, so use anonymous
structs instead.

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/1462353119-9738-2-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Remove intel_clock_t typedef

Just use "struct dpll" everywhere. That's actually shorter than
intel_clock_t.

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/1462353119-9738-1-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915/guc: Add Broxton GuC firmware loading support

Issue: VIZ-7772
Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: peter.antoine@intel.com
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462531373-34819-2-git-send-email-nicholas.hoath@intel.com

drm/i915/guc: Use major_minor version for filename

Load guc firmware from file with major_minor number
in filename instead of using symolic link with only
major number.

This change is so that new firmwares can only be used
with a kernel change. This in case there is a regression
with a new firmware, it won't be used by default without
some testing.

Issue: VIZ-7713
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Acked-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: peter.antoine@intel.com
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: Convert intel_overlay.c to use native drm_i915_private pointers

Another day, another long overdue conversion. Not much to update inside
intel_overlay.c, but still

   text    data     bss     dec     hex filename
6309547 3578778  696320 10584645  a18245 vmlinux
6309291 3578778  696320 10584389  a18145 vmlinux

a couple of hundred bytes of pointer misdirection.

Whilst here, rename the ioctl entry points to include the _ioctl suffix
so that the user entry points are clear (following the idiom).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1463053403-25086-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>

drm/i915: s/DPPL/DPLL/ for SKL DPLLs

SKL DPLLs shouldn't be called DPPLs.

Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Fixes: 2edd6443e3d0 ("drm/i915: Use a table to initilize shared dplls")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462993473-8254-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Ander Conselvan de Oliveira <conselvan2@gmail.com>

drm/i915: Introduce IS_GEN macro

To be used for more efficient Gen range checking.

v2: Remove spurious chunk. (Chris Wilson)
v3: Rebase.
v4: Renamed from INTEL_GEN_RANGE and added GEN_FOREVER.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v3)
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462874228-6601-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Do not use a bitfield for INTEL_INFO->num_pipes

It just makes more work for the compiler and generates more code.

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

drm/i915: Replace "INTEL_INFO->gen == x" checks with IS_GENx

This way optimization from a previous patch works even better.

v2: Rebase.

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

drm/i915: Promote IS_BROADWELL to a simple macro

If we allow it a dedicated flag in dev_priv we enable the
compiler to nicely optimize conditions like IS_HASSWELL ||
IS_BROADWELL.

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

drm/i915: Make IS_GENx macros work on a mask

If instead of numerical comparison me make these test a
bitmask, we enable the compiler to optimize all instances
of IS_GENx || IS_GENy.

v2: Make bit zero of gen mask mean gen 1.

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

drm/i915: Use drm_i915_private as the native pointer for intel_uncore.c

Pass drm_i915_private to the uncore init/fini routines and their
subservients as it is their native type.

   text    data     bss     dec     hex filename
6309978 3578778  696320 10585076         a183f4 vmlinux
6309530 3578778  696320 10584628         a18234 vmlinux

a modest 400 bytes of saving, but 60 lines of code deleted!

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/1462885804-26750-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Re-enable GGTT earlier during resume on pre-gen6 platforms

Move the intel_enable_gtt() call to happen before we touch the GTT
during resume. Right now it's done way too late. Before
commit ebb7c78d358b ("agp/intel-gtt: Only register fake agp driver for gen1")
it was actually done earlier on account of also getting called from
the resume hook of the fake agp driver. With the fake agp driver
no longer getting registered we must move the call up.

The symptoms I've seen on my 830 machine include lowmem corruption,
other kinds of memory corruption, and straight up hung machine during
or just after resume. Not really sure what causes the memory corruption,
but so far I've not seen any with this fix.

I think we shouldn't really need to call this during init, but we have
been doing that so I've decided to keep the call. However moving that
call earlier could be prudent as well. Doing it right after the
intel-gtt probe seems appropriate.

Also tested this on 946gz,elk,ilk and all seemed quite happy with
this change.

v2: Reorder init_hw vs. enable_hw functions (Chris)

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: drm-intel-fixes@lists.freedesktop.org
Fixes: ebb7c78d358b ("agp/intel-gtt: Only register fake agp driver for gen1")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462559755-353-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>

x86: Silence 32bit compiler warning in intel_graphics_stolen()

arch/x86/kernel/early-quirks.c: In function ‘intel_graphics_stolen’:
arch/x86/kernel/early-quirks.c:539:9: warning: format ‘%llx’ expects
argument of type ‘long long unsigned int’, but argument 2 has type ‘phys_addr_t’ [-Wformat=]
         "0x%llx-0x%llx\n", base, base + size - 1);
         ^
arch/x86/kernel/early-quirks.c:539:9: warning: format ‘%llx’ expects
argument of type ‘long long unsigned int’, but argument 3 has type ‘phys_addr_t’ [-Wformat=]

v2: Use %pa for addresses

Fixes: ee0629cfd3c16 (drm/i915: Function per early graphics quirk)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> #v1
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462811982-1567-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Add a FIXME about crtc !active vs. watermarks

When the crtc is enabled but !active, we should still compute the
watermarks as if the planes were visible. That would make it more
likely that the we can later transition to active without errors.

Add a FIXME to remind people that we're doing the wrong thing now.
We should perhaps just move the wm computation for each individual plane
into the .check_plane hook, and later we'd just combine the results from
all active planes.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461940278-17122-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Calculate IPS linetime watermark based on future cdclk

Use the cdclk we're going to be using when the pipe gets enabled to
compute the IPS linetime watermark. The current cdclk frequency is
irrelevant at this point since it can still change.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461940278-17122-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915: Store a i915 backpointer from engine, and use it

   text    data     bss     dec     hex filename
6309351 3578714  696320 10584385  a18141 vmlinux
6308391 3578714  696320 10583425  a17d81 vmlinux

Almost 1KiB of code reduction.

v2: More s/INTEL_INFO()->gen/INTEL_GEN()/ and IS_GENx() conversions

   text    data     bss     dec     hex filename
6304579 3578778  696320 10579677  a16edd vmlinux
6303427 3578778  696320 10578525  a16a5d vmlinux

Now over 1KiB!

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/1462545621-30125-3-git-send-email-chris@chris-wilson.co.uk

drm/i915/execlists: Refactor common engine setup

Move all of the constant assignments up front and into a common
function. This is primarily to ensure the backpointers are set as early
as possible for later use during initialisation.

v2: Use a constant struct so that all the similar values are set
together.
v3: Sanitize the engine's IMR to disable any potential interrupt before
we are ready (enabled in init_hw).
v4: Ignore the engine's IMR, to be resolved later

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462545621-30125-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Small display interrupt handlers tidy

I have noticed some of our interrupt handlers use both dev and
dev_priv while they could get away with only dev_priv in the
huge majority of cases.

Tidying that up had a cascading effect on changing functions
prototypes, so relatively big churn factor, but I think it is
for the better.

For example even where changes cascade out of i915_irq.c, for
functions prefixed with intel_, genX_ or <plat>_, it makes more
sense to take dev_priv directly anyway.

This allows us to eliminate local variables and intermixed usage
of dev and dev_priv where only one is good enough.

End result is shrinkage of both source and the resulting binary.

i915.ko:

 - .text         000b0899
 + .text         000b0619

Or if we look at the Gen8 display irq chain:

 -00000000000006ad t gen8_irq_handler
 +0000000000000663 t gen8_irq_handler
   -0000000000000028 T intel_opregion_asle_intr
   +0000000000000024 T intel_opregion_asle_intr
   -000000000000008c t ilk_hpd_irq_handler
   +000000000000007f t ilk_hpd_irq_handler
   -0000000000000116 T intel_check_page_flip
   +0000000000000112 T intel_check_page_flip
   -000000000000011a T intel_prepare_page_flip
   +0000000000000119 T intel_prepare_page_flip
   -0000000000000014 T intel_finish_page_flip_plane
   +0000000000000013 T intel_finish_page_flip_plane
   -0000000000000053 t hsw_pipe_crc_irq_handler
   +000000000000004c t hsw_pipe_crc_irq_handler
   -000000000000022e t cpt_irq_handler
   +0000000000000213 t cpt_irq_handler

So small shrinkage but it is all fast paths so doesn't harm.

Situation is similar in other interrupt handlers as well.

v2: Tidy intel_queue_rps_boost_for_request as well. (Chris Wilson)

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

drm/i915: Determine DP++ type 1 DVI adaptor presence based on VBT

DP dual mode type 1 DVI adaptors aren't required to implement any
registers, so it's a bit hard to detect them. The best way would
be to check the state of the CONFIG1 pin, but we have no way to
do that. So as a last resort, check the VBT to see if the HDMI
port is in fact a dual mode capable DP port.

v2: Deal with VBT code reorganization
    Deal with DRM_DP_DUAL_MODE_UNKNOWN
    Reduce DEVICE_TYPE_DP_DUAL_MODE_BITS a bit
    Accept both DP and HDMI dvo_port in VBT as my BSW
    at least declare its DP port as HDMI :(
v3: Ignore DEVICE_TYPE_NOT_HDMI_OUTPUT (Shashank)

Cc: stable@vger.kernel.org
Cc: Tore Anderson <tore@fud.no>
Reported-by: Tore Anderson <tore@fud.no>
Fixes: 7a0baa623446 ("Revert "drm/i915: Disable 12bpc hdmi for now"")
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462362322-31278-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Shashank Sharma <shashank.sharma@intel.com>

drm/i915: Enable/disable TMDS output buffers in DP++ adaptor as needed

To save a bit of power, let's try to turn off the TMDS output buffers
in DP++ adaptors when we're not driving the port.

v2: Let's not forget DDI, toss in a debug message while at it
v3: Just do the TMDS output control based on adaptor type. With the
    helper getting passed the type, we wouldn't actually have to
    check at all in the driver, but the check eliminates the debug
    output more honest

Cc: stable@vger.kernel.org
Cc: Tore Anderson <tore@fud.no>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462216105-20881-4-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Shashank Sharma <shashank.sharma@intel.com>

drm/i915: Respect DP++ adaptor TMDS clock limit

Try to detect the max TMDS clock limit for the DP++ adaptor (if any)
and take it into account when checking the port clock.

Note that as with the sink (HDMI vs. DVI) TMDS clock limit we'll ignore
the adaptor TMDS clock limit in the modeset path, in case users are
already "overclocking" their TMDS links. One subtle change here is that
we'll have to respect the adaptor TMDS clock limit when we decide whether
to do 12bpc or 8bpc, otherwise we might end up picking 12bpc and
accidentally driving the TMDS link out of spec even when the user chose
a mode that fits wihting the limits at 8bpc. This means you can't
"overclock" your DP++ dongle at 12bpc anymore, but you can continue to
do so at 8bpc.

Note that for simplicity we'll use the I2C access method for all dual
mode adaptors including type 2. Otherwise we'd have to start mixing
DP AUX and HDMI together. In the future we may need to do that if we
come across any board designs that don't hook up the DDC pins to the
DP++ connectors. Such boards would obviously only work with type 2
dual mode adaptors, and not type 1.

v2: Store adaptor type under indel_hdmi->dp_dual_mode
    Deal with DRM_DP_DUAL_MODE_UNKNOWN
    Pass adaptor type to drm_dp_dual_mode_max_tmds_clock(),
    and use it for type1 adaptors as well

Cc: stable@vger.kernel.org
Reported-by: Tore Anderson <tore@fud.no>
Fixes: 7a0baa623446 ("Revert "drm/i915: Disable 12bpc hdmi for now"")
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462216105-20881-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Shashank Sharma <shashank.sharma@intel.com>

drm: Add helper for DP++ adaptors

Add a helper which aids in the identification of DP dual mode
(aka. DP++) adaptors. There are several types of adaptors
specified: type 1 DVI, type 1 HDMI, type 2 DVI, type 2 HDMI

Type 1 adaptors have a max TMDS clock limit of 165MHz, type 2 adaptors
may go as high as 300MHz and they provide a register informing the
source device what the actual limit is. Supposedly also type 1 adaptors
may optionally implement this register. This TMDS clock limit is the
main reason why we need to identify these adaptors.

Type 1 adaptors provide access to their internal registers and the sink
DDC bus through I2C. Type 2 adaptors provide this access both via I2C
and I2C-over-AUX. A type 2 source device may choose to implement either
of these methods. If a source device implements the I2C-over-AUX
method, then the driver will obviously need specific support for such
adaptors since the port is driven like an HDMI port, but DDC
communication happes over the AUX channel.

This helper should be enough to identify the adaptor type (some
type 1 DVI adaptors may be a slight exception) and the maximum TMDS
clock limit. Another feature that may be available is control over
the TMDS output buffers on the adaptor, possibly allowing for some
power saving when the TMDS link is down.

Other user controllable features that may be available in the adaptors
are downstream i2c bus speed control when using i2c-over-aux, and
some control over the CEC pin. I chose not to provide any helper
functions for those since I have no use for them in i915 at this time.
The rest of the registers in the adaptor are mostly just information,
eg. IEEE OUI, hardware and firmware revision, etc.

v2: Pass adaptor type to helper functions to ease driver implementation
    Fix a bunch of typoes (Paulo)
    Add DRM_DP_DUAL_MODE_UNKNOWN for the case where we don't (yet) know
    the type (Paulo)
    Reject 0x00 and 0xff DP_DUAL_MODE_MAX_TMDS_CLOCK values (Paulo)
    Adjust drm_dp_dual_mode_detect() type2 vs. type1 detection to
    ease future LSPCON enabling
    Remove the unused DP_DUAL_MODE_LAST_RESERVED define
v3: Fix kernel doc function argument descriptions (Jani)
    s/NONE/UNKNOWN/ in drm_dp_dual_mode_detect() docs
    Add kernel doc for enum drm_dp_dual_mode_type
    Actually build the docs
    Fix more typoes
v4: Adjust code indentation of type2 adaptor detection (Shashank)
    Add debug messages for failurs cases (Shashank)
v5: EXPORT_SYMBOL(drm_dp_dual_mode_read) (Paulo)

Cc: stable@vger.kernel.org
Cc: Tore Anderson <tore@fud.no>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Shashank Sharma <shashank.sharma@intel.com> (v4)
Link: http://patchwork.freedesktop.org/patch/msgid/1462542412-25533-1-git-send-email-ville.syrjala@linux.intel.com

drm/i915: Dump pipe config when intel_modeset_pipe_config fails.

This makes it easier to debug issues like https://bugs.freedesktop.org/show_bug.cgi?id=93477

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Chris Wilson <|chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/843f4327-1574-cf8e-0776-adbb0d58c2c0@mblankhorst.nl
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Revert "mfd: intel_soc_pmic_core: Terminate panel control GPIO lookup table correctly"

This reverts commit 3543995a71b9150621bf30f4b45820d9cf65adb4.

I mixed up maintainers and thought Linus' ack was for the mfd tree.
But Lee Jones (the real maintainer) wants to merge this through the
mfd tree, so revert here.

Cc: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Allow MI_LOAD_REGISTER_REG between whitelisted registers.

Allowing register copies where the source and destination are both
whitelisted should be safe, and is useful.  For example, Mesa uses
this to load the command streamer math registers with data from the
pipeline statistics counters.

v2: Reject writes to OACONTROL (and reads as well :(

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> # v1
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1462521014-13595-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Update DRIVER_DATE to 20160508

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/kbl: Introduce the first official DMC for Kabylake.

Version 1.01.

This firmware is made for Kabylake platform so it doesn't
need the stepping workaround that we had before.

v2: Rebased on top of latest nightly with min version
    required change.
v3: With right CSR_VERSION (Patrik).

Cc: Christophe Prigent <christophe.prigent@intel.com>
Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Reviewed-by: Ben Widawsky <benjamin.widawsky@intel.com> (v1)
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461707991-15336-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Unexport i915_ppgtt_init()

As i915_ppgtt_init() is not used outside of i915_gem_gtt.c we can make
it static.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1462443767-5194-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Matthew Auld <matthew.auld@intel.com>

drm/i915: Report command parser version 0 if disabled

If the command parser is not active, then it is appropriate to report it
as operating at version 0 as no higher mode is supported. This greatly
simplifies userspace querying for the command parser as we then do not
need to second guess when it will be active (a mixture of module
parameters and generational support, which may change over time).

v2: s/comand/command/ misspelling in comment

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1462368336-21230-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

drm/i915: Bail out of pipe config compute loop on LPT

LPT is pch, so might run into the fdi bandwidth constraint (especially
since it has only 2 lanes). But right now we just force pipe_bpp back
to 24, resulting in a nice loop (which we bail out with a loud
WARN_ON). Fix this.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=93477
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462264381-7573-1-git-send-email-daniel.vetter@ffwll.ch

drm/i915: remove i915_gem_object_ggtt_unbind

Only has one user and is nothing more than a shim on top of
i915_vma_unbind, so let's just get rid of it.

Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Suggested-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461842691-27575-1-git-send-email-matthew.auld@intel.com

drm/i915: Simplify intel_mark_busy/idle

They use dev_priv exclusively so pass it in instead of dev
for smaller source and binary.

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/1461844620-35360-1-git-send-email-tvrtko.ursulin@linux.intel.com

Revert "drm/i915: start adding dp mst audio"

Right now MST audio is causing too many kernel panics to really keep
around in the kernel. On top of that, even after fixing said panics it's
still basically non-functional (at least on all the setups I've tested
it on). Revert until we have a proper solution for this.

This reverts commit 3d52ccf52f2c51f613e42e65be0f06e4e6788093.

Signed-off-by: Lyude <cpaul@redhat.com>
Fixes: 3d52ccf52f2c ("drm/i915: start adding dp mst audio")
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1462287692-28570-1-git-send-email-cpaul@redhat.com

mfd: intel_soc_pmic_core: Terminate panel control GPIO lookup table correctly

GPIO lookup tables are supposed to be zero terminated. Let's do that
and avoid accidentally walking off the end.

Cc: Shobhit Kumar <shobhit.kumar@intel.com>
Cc: Samuel Ortiz <sameo@linux.intel.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Alexandre Courbot <gnurou@gmail.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Lee Jones <lee.jones@linaro.org>
Cc: stable@vger.kernel.org
Fixes: 61dd2ca2d44e ("mfd: intel_soc_pmic_core: Add lookup table for Panel Control as GPIO signal")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1461353935-8078-1-git-send-email-ville.syrjala@linux.intel.com

drm/i915/chv: Tune L3 SQC credits based on actual latencies

While browsing BSpec I bumped into a note saying we need to tune these
values based on actual measurements done after initial enabling. I've
checked that it indeed improves things on BXT. I haven't checked this on
CHV, but here it is if someone wants to give it a go.

v2:
- Add note about the discrepancy wrt. to the spec in the formula
  calculating the credit encodings. (Mika, Ville)
- Move the WA comment to the new function. (Ville)
v3:
- Keep the comment about the SQC WA in the caller. (Ville)

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
CC: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462280061-1457-4-git-send-email-imre.deak@intel.com

drm/i915: Clean up L3 SQC register field definitions

No need for hard-coding the register value, the corresponding fields are
defined properly in BSpec.

No functional change.

v2:
- Rebased on BXT L3 SQC tuning patch merged meanwhile.

CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v1)
Link: http://patchwork.freedesktop.org/patch/msgid/1462280061-1457-3-git-send-email-imre.deak@intel.com

drm/i915/bdw: Add missing delay during L3 SQC credit programming

BSpec requires us to wait ~100 clocks before re-enabling clock gating,
so make sure we do this.

CC: stable@vger.kernel.org
CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1462280061-1457-2-git-send-email-imre.deak@intel.com

drm/i915: Parse LFP brightness control field in VBT

These fields in VBT indicates the PWM source which
is used and also the controller number.

v2 by Jani: check for out of bounds access, some renames, change default
type, etc.

v3 by Jani: s/INTEL_BACKLIGHT_CABC/INTEL_BACKLIGHT_DSI_DCS/

Signed-off-by: Deepak M <m.deepak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/eee2f7b683a081f006a7df1ddad9b20fbf53c48c.1461676337.git.jani.nikula@intel.com

drm/i915: Add rpm get/put in oom and vmap notifier

i915_gem_shrink() will scan the bound list only if device is not
suspended but in OOM failure scenario it becomes absolutely necessary
to release as much memory as possible. Also in allocation failure from
vmap address space, it is incumbent on the Driver to reap all its
vmaps. So, adding rpm get/put in i915_gem_shrinker_oom() and
i915_gem_shrinker_vmap() to ensure shrinking of bound objects as well.

Signed-off-by: Praveen Paneri <praveen.paneri@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1462178429-13449-2-git-send-email-praveen.paneri@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915: Unbind objects in shrinker only if device is runtime active

When the system is running low on memory, gem shrinker is invoked.
In this process objects will be unbounded from GTT and unbinding process
will require access to GTT(GTTADR) and also to fence register potentially.
That requires a resume of gfx device, if suspended, in the shrinker path.
Considering the power leakage due to intermediate resume, perform unbinding
operation only if device is already runtime active.

v2: Use newly implemented intel_runtime_pm_get_if_in_use (Chris)

Signed-off-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Praveen Paneri <praveen.paneri@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1462178429-13449-1-git-send-email-praveen.paneri@intel.com
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

drm/i915/lvds: separate border enable readout from panel fitter

The LVDS border enable is independent from the panel fitter. Move the
readout of the "border bits" from i9xx_get_pfit_config() to
intel_lvds_get_config(), where it will be read if LVDS is enabled even
if the panel fitter is not.

This fixes the state checker warning:

[drm:intel_pipe_config_compare [i915]] *ERROR* mismatch in
gmch_pfit.lvds_border_bits (expected 0x00008000, found 0x00000000)

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: drm-intel-fixes@lists.freedesktop.org
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87632
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Tested-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461933243-2140-1-git-send-email-jani.nikula@intel.com

drm/i915: Enable semaphores for legacy submission on gen8

We have sufficient evidence from igt to support that semaphores are in
a working state. Enabling semaphores now for legacy provides a better
comparison of execlists against legacy ring submission.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1461928823-10298-6-git-send-email-chris@chris-wilson.co.uk

drm/i915: Reload PD tables after semaphore wait on gen8

When the engine idles waiting upon a semaphore, it loses its
pagetables and we must reload them before executing the batch.

v2: Restrict w/a to non-RCS rings (RCS works correctly apparently).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461932305-14637-5-git-send-email-chris@chris-wilson.co.uk

drm/i915: Fix serialisation of pipecontrol write vs semaphore signal

In order for the MI_SEMAPHORE_SIGNAL command to wait until after the
pipecontrol writing the signal value is complete, we have to pause the
CS inside the PIPE_CONTROL with the CS_STALL bit.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461932305-14637-4-git-send-email-chris@chris-wilson.co.uk

drm/i915: Fix gen8 semaphores id for legacy mode

With the introduction of a distinct engine->id vs the hardware id, we need
to fix up the value we use for selecting the target engine when signaling
a semaphore. Note that these values can be merged with engine->guc_id.

Fixes: de1add360522c876c25ef2bbbbab1c94bdb509ab
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461932305-14637-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Fix ordering of sanitize ppgtt and sanitize execlists

The i915.enable_ppgtt option depends upon the state of
i915.enable_execlists option - so we need to sanitize execlists first.

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/1461932305-14637-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Apply strongly ordered RCS breadcrumb to gen8/legacy

For legacy ringbuffer mode, we need the new ordered breadcrumb emission
tried and tested on execlists in order to avoid the dreaded "missed
interrupt" syndrome. A secondary advantage of the execlists method is
that it writes to an arbitrary address, useful if one wants to write a
breadcrumb elsewhere.

This fix is taken from commit 7c17d377374dd (drm/i915: Use ordered seqno
write interrupt generation on gen8+ execlists).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461932305-14637-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Trim the flush for the execlists request emission

At the start of request emission, we flush some space for the request,
estimating the typical size for the request body. The common tail is now
much larger than the typical body, so we can shrink the flush
substantially.

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/1461917226-9132-3-git-send-email-chris@chris-wilson.co.uk

drm/i915: Trim the flush for the legacy request emission

At the start of request emission, we flush some space for the request,
estimating the typical size for the request body. The tail is now much
larger than the typical body, so we can shrink the flush slightly.

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/1461917226-9132-2-git-send-email-chris@chris-wilson.co.uk

drm/i915: Bump reserved size for legacy gen8 semaphore emission

With 5 rings and a flush, we need 192 bytes of space to emit the
breadcrumb and semaphores. However, we need some spare room the size of
the single largest packet (36 dwords, 144 bytes) to accommodate
wraparound giving a grand total of 336 bytes

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/1461917226-9132-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move VLV HDMI lane reset work around logic to intel_dpio_phy.c

This moves the last phy specific code from the encoders to the phy
specific file.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-11-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate pre encoder enabling phy code

The phy code in vlv_pre_enable_dp() and vlv_hdmi_pre_enable() is
exectly the same, so extract it to intel_dpio_phy.c.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-10-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate VLV phy pre pll enabling code

The code used by the DP and HDMI paths was very similar, so make them
share it. Note that this removes the write to signal level registers
from the HDMI pre pll enable path, but that's OK since those are set
in vlv_hdmi_pre_enable() function.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-9-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate VLV signal level code

The logic for setting signal levels is used for both HDMI and DP with
small variations. But it is similar enough to put behind a function
called from the encoders.

v2: Remove unrelated MST changes due to rebase fumble. (Jim Bride)
    Fix typo in the commit message. (Jim Bride)

v3: Really fix the typo. (Jim)

Cc: Jim Bride <jim.bride@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-8-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate CHV encoders' post pll disable code

The exact same code was used by HDMI and DP encoders, so move it to
intel_dpio_phy.c.

v2: Fix typo in the commit message. (Jim Bride)
v3: Call the new function chv_phy_post_pll_disable() instead of
    chv_phy_post_disable(), as it should be called after the pll
    is disabled. (Ville)

Cc: Jim Bride <jim.bride@linux.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: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-7-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate CHV pre-encoder enabling phy logic

The only difference between the DP and HDMI versions was the lane count.
Since lane_count is now set appropriately for HDMI too, get rid of the
duplication and move this to intel_dpio_phy.c

v2: Don't move comments about 2nd common lane staying alive. (Ville)
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-6-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate CHV phy-releated pre pll enabling code

The same logic is used for DP and HDMI so move it to intel_dpio_phy.c.

v2: Rebase

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-5-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate chv_data_lane_soft_reset()

The function chv_data_lane_soft_reset() was duplicated in DP and HDMI
code. Move it to intel_dpio_phy.c.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-4-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Unduplicate CHV signal level code

The code for programming voltage swing and emphasis was duplicated
between DP and HDMI code. Move that to a new file, intel_dpio_phy.c.

v2: Keep the "Use 800mV-0dB" comment in the HDMI code. (Ville)
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-3-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Set crtc_state->lane_count for HDMI

Set the lane count for HDMI to 4. This will make it easier to
unduplicate CHV phy code.

This also fixes the the soft reset programming for HDMI with CHV. After
commit a8f327fb8464 ("drm/i915: Clean up CHV lane soft reset
programming"), it wouldn't set the right bits for PCS23 since it relied
on a lane count that was never set.

v2: Set lane_count in *_get_config() to please state checker. (0day)
v3: Set lane_count for DDI in DVI mode too. (CI)
v4: Add note about CHV soft lane reset. (Ander)

Fixes: a8f327fb8464 ("drm/i915: Clean up CHV lane soft reset programming")
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Jim Bride <jim.bride@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461761065-21195-2-git-send-email-ander.conselvan.de.oliveira@intel.com

drm/i915: Fix comments about GMBUSFREQ register

The comment about GMBUSFREQ is confused. The spec actually explains
the 4MHz thing perfectly by noting that the 4MHz divider values is
actually just bits [9:2] not [9:0], hence the divide by 1000 correct.
Replace the confused note with a quote from the spec, and eliminate
the duplicated comment that snuck in.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461689194-6079-4-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915: Use cached cdclk value in i915_audio_component_get_cdclk_freq()

No point in reading the cdclk out from the hardware every single time
since we have it cached already. Just return the cached value to the
audio driver.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461689194-6079-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915: Update CDCLK_FREQ register on BDW after changing cdclk frequency

Update CDCLK_FREQ on BDW after changing the cdclk frequency. Not sure
if this is a late addition to the spec, or if I simply overlooked this
step when writing the original code.

This is what Bspec has to say about CDCLK_FREQ:
"Program this field to the CD clock frequency minus one. This is used to
 generate a divided down clock for miscellaneous timers in display."

And the "Broadwell Sequences for Changing CD Clock Frequency" section
clarifies this further:
"For CD clock 337.5 MHz, program 337 decimal.
 For CD clock 450 MHz, program 449 decimal.
 For CD clock 540 MHz, program 539 decimal.
 For CD clock 675 MHz, program 674 decimal."

Cc: stable@vger.kernel.org
Cc: Mika Kahola <mika.kahola@intel.com>
Fixes: b432e5cfd5e9 ("drm/i915: BDW clock change support")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461689194-6079-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Mika Kahola <mika.kahola@intel.com>

drm/i915/bxt: Adjusting the error in horizontal timings retrieval

In BXT DSI there is no regs programmed with few horizontal timings
in Pixels but txbyteclkhs.. So retrieval process adds some
ROUND_UP ERRORS in the process of PIXELS<==>txbyteclkhs.

Actually here for the given adjusted_mode, we are calculating the
value programmed to the port and then back to the horizontal timing
param in pixels. This is the expected value at the end of get_config,
including roundup errors. And if that is same as retrieved value
from port, then retrieved (HW state) adjusted_mode's horizontal
timings are corrected to match with SW state to nullify the errors.

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461053894-5058-2-git-send-email-ramalingam.c@intel.com

drm/i915/BXT: Retrieving the horizontal timing for DSI

Retriving the horizontal timings from the port registers as part of
get_config().

This fixes a division by zero:

[   56.916557] divide error: 0000 [#1] PREEMPT SMP
[   56.921741] Modules linked in: i915(+) drm_kms_helper syscopyarea
sysfillrect sysimgblt fb_sys_fops drm intel_gtt agpgart cf
g80211 rfkill binfmt_misc ax88179_178a kvm_intel kvm irqbypass crc32c_intel
efivars tpm_tis tpm fuse
[   56.944106] CPU: 3 PID: 1097 Comm: modprobe Not tainted 4.6.0-rc4+ #433
[   56.951501] Hardware name: Intel Corp. Broxton M/RVP, BIOS
BXT1RVPA.X64.0131.B30.1604142217 04/14/2016
[   56.961908] task: ffff88007a854d00 ti: ffff88007aea0000 task.ti:
ffff88007aea0000
[   56.970273] RIP: 0010:[<ffffffffa01235b2>]  [<ffffffffa01235b2>]
drm_mode_hsync+0x22/0x40 [drm]
[   56.980043] RSP: 0018:ffff88007aea3788  EFLAGS: 00010206
[   56.985982] RAX: 000000000788b600 RBX: ffff880073c22108 RCX:
0000000000000000
[   56.993957] RDX: 0000000000000000 RSI: ffff88007ab06800 RDI:
ffff880073c22108
[   57.001935] RBP: ffff88007aea3788 R08: 0000000000000001 R09:
ffff880073c221e8
[   57.009903] R10: ffff880073c22108 R11: 0000000000000001 R12:
ffff88007a300000
[   57.017872] R13: ffff880073c22000 R14: ffff880175f78000 R15:
ffff880175f78798
[   57.025849] FS:  00007f105d3e6700(0000) GS:ffff88017fd80000(0000)
knlGS:0000000000000000
[   57.034894] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   57.041317] CR2: 00007f4d485101d0 CR3: 000000007a820000 CR4:
00000000003406e0
[   57.049292] Stack:
[   57.051539]  ffff88007aea37a0 ffffffffa043b632 ffff880175f787c8
ffff88007aea3810
[   57.059825]  ffffffffa043d59e ffff880175f787b0 ffff88007ab68c00
ffff88007aea37f0
[   57.068128]  ffff880073c221e8 ffff880073c22108 ffff880175f78780
ffff880100000000
[   57.076430] Call Trace:
[   57.079254]  [<ffffffffa043b632>] intel_mode_from_pipe_config+0x82/0xb0
[i915]
[   57.087405]  [<ffffffffa043d59e>] intel_modeset_setup_hw_state+0x55e/0xd60
[i915]
[   57.095847]  [<ffffffffa043ff94>] intel_modeset_init+0x8e4/0x1630 [i915]
[   57.103415]  [<ffffffffa047bcf0>] i915_driver_load+0xbe0/0x1980 [i915]
[   57.110745]  [<ffffffffa0116c19>] drm_dev_register+0xa9/0xc0 [drm]
[   57.117681]  [<ffffffffa011921d>] drm_get_pci_dev+0x8d/0x1e0 [drm]
[   57.124600]  [<ffffffff8195f942>] ? _raw_spin_unlock_irqrestore+0x42/0x70
[   57.132253]  [<ffffffffa03b0384>] i915_pci_probe+0x34/0x50 [i915]
[   57.139070]  [<ffffffff8149c375>] local_pci_probe+0x45/0xa0
[   57.145303]  [<ffffffff8149d300>] ? pci_match_device+0xe0/0x110
[   57.151924]  [<ffffffff8149d6cb>] pci_device_probe+0xdb/0x130
[   57.158355]  [<ffffffff81579b93>] driver_probe_device+0x223/0x440
[   57.165169]  [<ffffffff81579e85>] __driver_attach+0xd5/0x100
[   57.171500]  [<ffffffff81579db0>] ? driver_probe_device+0x440/0x440
[   57.178510]  [<ffffffff81577736>] bus_for_each_dev+0x66/0xa0
[   57.184841]  [<ffffffff815793de>] driver_attach+0x1e/0x20
[   57.190881]  [<ffffffff81578d6e>] bus_add_driver+0x1ee/0x280
[   57.197212]  [<ffffffff8157abc0>] driver_register+0x60/0xe0
[   57.203447]  [<ffffffff8149bc50>] __pci_register_driver+0x60/0x70
[   57.210285]  [<ffffffffa0119450>] drm_pci_init+0xe0/0x110 [drm]
[   57.216911]  [<ffffffff810dcd8d>] ? trace_hardirqs_on+0xd/0x10
[   57.223434]  [<ffffffffa023a000>] ? 0xffffffffa023a000
[   57.229237]  [<ffffffffa023a092>] i915_init+0x92/0x99 [i915]
[   57.235570]  [<ffffffff810003db>] do_one_initcall+0xab/0x1d0
[   57.241900]  [<ffffffff810f9eef>] ? rcu_read_lock_sched_held+0x7f/0x90
[   57.249205]  [<ffffffff81204f18>] ? kmem_cache_alloc_trace+0x248/0x2b0
[   57.256509]  [<ffffffff811a5eee>] ? do_init_module+0x27/0x1d9
[   57.262934]  [<ffffffff811a5f26>] do_init_module+0x5f/0x1d9
[   57.269167]  [<ffffffff8112392f>] load_module+0x20ef/0x27b0
[   57.275401]  [<ffffffff8111f8e0>] ? store_uevent+0x40/0x40
[   57.281541]  [<ffffffff81124243>] SYSC_finit_module+0xc3/0xf0
[   57.287969]  [<ffffffff8112428e>] SyS_finit_module+0xe/0x10
[   57.294203]  [<ffffffff81960069>] entry_SYSCALL_64_fastpath+0x1c/0xac
[   57.301406] Code: ff 5d c3 66 0f 1f 44 00 00 0f 1f 44 00 00 8b 87 d8 00 00
00 55 48 89 e5 85 c0 75 22 8b 4f 68 85 c9 78 1b 69 47 58 e8 03 00 00 99 <f7> f9
b9 d3 4d 62 10 05 f4 01 00 00 f7 e1 89 d0 c1 e8 06 5d c3
[   57.322964] RIP  [<ffffffffa01235b2>] drm_mode_hsync+0x22/0x40 [drm]
[   57.330103]  RSP <ffff88007aea3788>
[   57.334276] ---[ end trace d414224cb2e2a4cf ]---
[   57.339861] modprobe (1097) used greatest stack depth: 12048 bytes left

Fixes: 6f0e7535e7e1 ("drm/i915/BXT: Get pipe conf from the port registers")
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Acked-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461053894-5058-1-git-send-email-ramalingam.c@intel.com

drm/i915: Unify GPU resets upon shutdown

Both execlists and legacy need to reset the context (and mode) of the
GPU before we lose control of the system. By resetting the GPU, we
revert back to default settings. This simplifies the life of any
subsequent driver (in particular for virtualized setups) as it does not
then have to try and recover from an unknown condition. As both paths
need to reset for the same reason, move the reset to a common point.

This unifies the resets added in a647828afc (drm/i915: Also perform gpu
reset under execlist mode) and 8e96d9c4d9 (drm/i915: reset the GPU on
context fini).

v2: Restrict the reset to "modern" gen (where we enable HW contexts) to
try and avoid leaving the machine in an unusable state with a risky
reset on older GPU. This should keep the status quo as to who performs
resets (i.e. currently only GPUs with HW contexts perform a reset on
shutdown).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
CC: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: "Niu, Bing" <bing.niu@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-25-git-send-email-chris@chris-wilson.co.uk

drm/i915: Stop tracking execlists retired requests

With the previous patch having extended the pinned lifetime of
contexts by referencing the previous context from the current
request until the latter is retired (completed by the GPU),
we can now remove usage of execlist retired queue entirely.

This is because the above now guarantees that all execlist
object access requirements are satisfied by this new tracking,
and we can stop taking additional references and stop keeping
request on the execlists retired queue.

The latter was a source of significant scalability issues in
the driver causing performance hits on some tests. Most
dramatical of which was igt/gem_close_race which had run time
in tens of minutes which is now reduced to tens of seconds.

Signed-off-by: Tvrtko Ursulin <tvrtko@ursulin.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-24-git-send-email-chris@chris-wilson.co.uk

drm/i915: Store LRC hardware id in the request

This way in the following patch we can disconnect requests
from contexts.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-23-git-send-email-chris@chris-wilson.co.uk

drm/i915: Track the previous pinned context inside the request

As the contexts are accessed by the hardware until the switch is completed
to a new context, the hardware may still be writing to the context object
after the breadcrumb is visible. We must not unpin/unbind/prune that
object whilst still active and so we keep the previous context pinned until
the following request. We can generalise the tracking we already do via
the engine->last_context and move it to the request so that it works
equally for execlists and GuC.

v2: Drop the execlists double pin as that exposes a race inside the lrc
irq handler as it tries to access the context after it may be retired.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-22-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move releasing of the GEM request from free to retire/cancel

If we move the release of the GEM request (i.e. decoupling it from the
various lists used for client and context tracking) after it is complete
(either by the GPU retiring the request, or by the caller cancelling the
request), we can remove the requirement that the final unreference of
the GEM request need to be under the struct_mutex.

The careful reader may notice that one or two impossible NULL pointer
tests are dropped for readability. These pointers cannot be NULL since
they are assigned during request construction and never unset.

v2,v3: Rebalance execlists by moving the context unpinning.
v4: Rebase onto -nightly
v5: Avoid trying to rebalance execlist/GuC context pinning, leave that
to the next step

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-21-git-send-email-chris@chris-wilson.co.uk

drm/i915: Move the magical deferred context allocation into the request

We can hide more details of execlists from higher level code by removing
the explicit call to create an execlist context from execbuffer and
into its first use by execlists.

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/1461833819-3991-20-git-send-email-chris@chris-wilson.co.uk

drm/i915: Refactor execlists default context pinning

Refactor pinning and unpinning of contexts, such that the default
context for an engine is pinned during initialisation and unpinned
during teardown (pinning of the context handles the reference counting).
Thus we can eliminate the special case handling of the default context
that was required to mask that it was not being pinned normally.

v2: Rebalance context_queue after rebasing.
v3: Rebase to -nightly (not 40 patches in)
v4: Rebase onto request_alloc unwinding

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>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-19-git-send-email-chris@chris-wilson.co.uk

drm/i915: Replace the pinned context address with its unique ID

Rather than reuse the current location of the context in the global GTT
for its hardware identifier, use the context's unique ID assigned to it
for its whole lifetime.

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/1461833819-3991-18-git-send-email-chris@chris-wilson.co.uk

drm/i915: Assign every HW context a unique ID

The hardware tracks contexts and expects all live contexts (those active
on the hardware) to have a unique identifier. This is used by the
hardware to assign pagefaults and the like to a particular context.

v2: Reorder to make sure ctx->link is not left dangling if the
assignment of a hw_id fails (Mika).

v3: We have 21bits of context space, not 20.

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/1461833819-3991-17-git-send-email-chris@chris-wilson.co.uk

drm/i915: Update execlists context descriptor format commentary

The comments describing the Context Descriptor Format are off by a bit
for the size of the context ID.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-16-git-send-email-chris@chris-wilson.co.uk

drm/i915: Preallocate enough space for the average request

Rather than being interrupted when we run out of space halfway through
the request, and having to restart from the beginning (and returning to
userspace), flush a little more free space when we prepare the request.

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/1461833819-3991-15-git-send-email-chris@chris-wilson.co.uk

drm/i915: Manually unwind after a failed request allocation

In the next patches, we want to move the work out of freeing the request
and into its retirement (so that we can free the request without
requiring the struct_mutex). This means that we cannot rely on
unreferencing the request to completely teardown the request any more
and so we need to manually unwind the failed allocation. In doing so, we
reorder the allocation in order to make the unwind simple (and ensure
that we don't try to unwind a partial request that may have modified
global state) and so we end up pushing the initial preallocation down
into the engine request initialisation functions where we have the
requisite control over the state of the request.

Moving the initial preallocation into the engine is less than ideal: it
moves logic to handle a specific problem with request handling out of
the common code. On the other hand, it does allow those backends
significantly more flexibility in performing its allocations.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-14-git-send-email-chris@chris-wilson.co.uk

drm/i915: Remove the identical implementations of request space reservation

Now that we share intel_ring_begin(), reserving space for the tail of
the request is identical between legacy/execlists and so the tautology
can be removed. In the process, we move the reserved space tracking
from the ringbuffer on to the request. This is to enable us to reorder
the reserved space allocation in the next patch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-13-git-send-email-chris@chris-wilson.co.uk

drm/i915: Unify intel_ring_begin()

Combine the near identical implementations of intel_logical_ring_begin()
and intel_ring_begin() - the only difference is that the logical wait
has to check for a matching ring (which is assumed by legacy).

In the process some debug messages are culled as there were following a
WARN if we hit an actual error.

v2: Updated commentary

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-12-git-send-email-chris@chris-wilson.co.uk

drm/i915: Rearrange switch_context to load the aliasing ppgtt on first use

The code to switch_mm() is already handled by i915_switch_context(), the
only difference required to setup the aliasing ppgtt is that we need to
emit te switch_mm() on the first context, i.e. when transitioning from
engine->last_context == NULL. This allows us to defer the
initialisation of the GPU from early device initialisation to first use,
which should marginally speed up both. The caveat is that we then defer
the context initialisation until first use - i.e. we cannot assume that
the GPU engines are initialised. For example, this means that power
contexts for rc6 (Ironlake) need to explicitly loaded, as they are.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1461833819-3991-11-git-send-email-chris@chris-wilson.co.uk