drm/i915: don't deactivate FBC at skylake_disable_primary_plane

FBC is already deactivated at this point.

Besides, nothing should be calling these lower-level function
pointers. A few months ago, the only caller of
dev_priv->fbc.deactivate was intel_pipe_set_base_atomic(), which was
the kgdboc function. But the following commit added it to the SKL
function:

    commit a8d201af68506b375b701d0d8dbe8487034256f2
    Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
    Date:   Thu Jan 7 11:54:11 2016 +0100
        drm/i915: Use plane state for primary plane updates.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454101060-23198-1-git-send-email-paulo.r.zanoni@intel.com

drm/i915: implement WaIncreaseDefaultTLBEntries

WaIncreaseDefaultTLBEntries increases the number of TLB
entries available for GPGPU workloads and gives significant
( > 10% ) performance gain for some OCL benchmarks.
Put this in a new function that can be a place for
workarounds that are GT related but not required per ring.
This function is called on driver load and also after a
reset and on resume, so it is safe for workarounds that get
clobbered in these situations. This function currently has
just this one workaround.

v2: This was originally split into 3 patches but following
  review feedback was squashed into 1.
  I have not incorporated some style comments from Chris
  Wilson as I felt that after defining and intialising a
  temporary variable and then adding an additional if block
  to only write the register if the temporary variable had
  been set, this didn't really give a net gain.

v3: Resending in the hope that BAT will run

v4: Change subject line to trigger BAT (please!)

Signed-off-by: Tim Gore <tim.gore@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454586574-2343-1-git-send-email-tim.gore@intel.com

drm/i915/dsi: Configure DSI after enabling DSI pll

We need to enable DSI PLL before configuring the DSI registers.

This has worked before on BYT/CHV, but BXT is more fussy.

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Mika Kahola <mika.kahola@intel.com>
Tested-by: Mika Kahola <mika.kahola@intel.com> # BXT
Tested-by: Jani Nikula <jani.nikula@intel.com> # BYT
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454503846-12103-1-git-send-email-ramalingam.c@intel.com

drm/i915: Allow i915_gem_object_get_page() on userptr as well

commit 033908aed5a596f6202c848c6bbc8a40fb1a8490
Author: Dave Gordon <david.s.gordon@intel.com>
Date:   Thu Dec 10 18:51:23 2015 +0000

    drm/i915: mark GEM object pages dirty when mapped & written by the CPU

introduced a check into i915_gem_object_get_dirty_pages() that returned
a NULL pointer when called with a bad object, one that was not backed by
shmemfs. This WARN was too strict as we can work on all struct page
backed objects, and resulted in a WARN + GPF for existing userspace. In
order to differentiate the various types of objects, add a new flags field
to the i915_gem_object_ops struct to describe their capabilities, with
the first flag being whether the object has struct pages.

v2: Drop silly const before an integer in the structure declaration.

Testcase: igt/gem_userptr_blits/relocations
Reported-and-tested-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Tested-by: Michal Winiarski <michal.winiarski@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453487551-16799-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Pretend cursor is always on for ILK-style WM calculations (v2)

Due to our lack of two-step watermark programming, our driver has
historically pretended that the cursor plane is always on for the
purpose of watermark calculations; this helps avoid serious flickering
when the cursor turns off/on (e.g., when the user moves the mouse
pointer to a different screen).  That workaround was accidentally
dropped as we started working toward atomic watermark updates.  Since we
still aren't quite there yet with two-stage updates, we need to
resurrect the workaround and treat the cursor as always active.

v2: Tweak cursor width calculations slightly to more closely match the
    logic we used before the atomic overhaul began.  (Ville)

Cc: simdev11@outlook.com
Cc: manfred.kitzbichler@gmail.com
Cc: drm-intel-fixes@lists.freedesktop.org
Reported-by: simdev11@outlook.com
Reported-by: manfred.kitzbichler@gmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93892
Fixes: 43d59eda1 ("drm/i915: Eliminate usage of plane_wm_parameters from ILK-style WM code (v2)")
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454479611-6804-1-git-send-email-matthew.d.roper@intel.com

drm/i915: Check DDI max lanes after applying BXT workaround

In commit bfb9faab8 we added a workaround for some BXT BIOS that fail to
properly initialize the DDI_A_4_LANES bit of the control register (4
lanes is the only valid configuration on BXT since there is no DDI E to
share with).  A recent patch added some additional checks on this
register bit before the workaround gets applied; this breaks eDP on BXT
in some settings.  Some minor code shuffling is all we need to restore
the workaround.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Fixes: 7cd87cb80 ("drm/i915: Check max number of lanes when registering DDI ports")
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454022577-834-1-git-send-email-matthew.d.roper@intel.com
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

drm/i915/bxt: Don't save/restore eDP panel power during suspend (v3)

Our attempts save/restore panel power state in i915_suspend.c are
causing unclaimed register warnings on BXT since the registers for this
platform differ from older platforms.

The big hammer suspend/resume shouldn't be necessary for PP since the
connector/encoder hooks should already handle this.  In theory we could
remove this for all platforms, but in practice it's likely that would
cause some regressions since older platforms with LVDS may have
incomplete PP handling.  For now we'll leave the PCH save/restore alone
and change the non-PCH branch to only operate on gen <= 4 so that BXT
and future platforms aren't included.

v2: Typo fix: s/||/&&/

v3: Change non-PCH condition to a gen <= 4 test rather than listing
    VLV/CHV/BXT as specific platforms to exclude; should be more
    future-proof as we add new platforms.  (Daniel)

Cc: Vandana Kannan <vandana.kannan@intel.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452102821-17190-1-git-send-email-matthew.d.roper@intel.com

drm/i915: Properly terminate KMS mode name string during tv init

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454071949-24677-4-git-send-email-imre.deak@intel.com

drm/i915: Add debug info for failed MSI enabling

While not being able to enable MSI interrupts may be a normal
circumstance, for debugging it may still be a useful information, so
emit an info about this.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454071949-24677-3-git-send-email-imre.deak@intel.com

drm/i915/chv: Fix error path in GPU freq helpers

Atm we wouldn't catch these errors or on the error path we would end up
with a division-by-zero, fix this up.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454071949-24677-2-git-send-email-imre.deak@intel.com

drm/i915: Sanity check DP AUX message buffer and size

While we are calling intel_dp_aux_transfer() with msg->size=0 whenever
msg->buffer is NULL, passing NULL to memcpy() is undefined according to
the ISO C standard. I haven't found any notes about this in the GNU C's
or the kernel's documentation of the function and can't imagine what it
would do with the NULL ptr. To better document this use of the
parameters it still make sense to add an explicit check for this to the
code.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454071949-24677-1-git-send-email-imre.deak@intel.com

drm/i915/bxt: update list of PCIIDs

Add PCIIDs for new versions of the SOC, based on BSpec. Also add the
name of the versions as code comment where this is available. The new
versions don't have any changes visible to the kernel driver.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453989852-13569-1-git-send-email-imre.deak@intel.com

drm/i915/skl/kbl: Add support for pipe fusing

On SKL and KBL we can have pipe A/B/C disabled by fuse settings. The
pipes must be fused in descending order (e.g. C, B+C, A+B+C). We simply
decrease info->num_pipes if we find a valid fused out config.

v2: Don't store the pipe disabled mask in device info (Damien)

v3: Don't check FUSE_STRAP register for pipe c disabled

Cc: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Patrik Jakobsson <patrik.jakobsson@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[Jani: fixed some checkpatch indentation complaints]
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453300280-10661-1-git-send-email-patrik.jakobsson@linux.intel.com

drm/i915: Instrument PSR parameter for debuging with link standby x link off.

Unfortunately we don't know all panels and platforms out there and we
found internal prototypes without VBT proper set but where only
link in standby worked well.

So, before enable PSR by default let's instrument the PSR parameter
in a way that we can identify different panels out there that might
require or work better with link standby mode.

It is also useful to say that for backward compatibility I'm not
changing the meaning of this flag. So "0" still means disabled
and "1" means enabled with full support and maximum power savings.

v2: Use positive value instead of negative for different operation mode
    as suggested by Daniel.

v3: As Paulo suggested use 2 to force link standby and 3 to force link
    fully on. Also split the link_standby introduction in a separated patch.

v4: Use DRM_ERROR for link off request on platforms that don't support and
    Remove the quirk promise.

Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454356928-19779-1-git-send-email-rodrigo.vivi@intel.com

drm/i915: Add PSR main link standby support back

Link standby support has been deprecated with 'commit 89251b177
("drm/i915: PSR: deprecate link_standby support for core platforms.")'

The reason for that is that main link in full off offers more power
savings and on HSW and BDW implementations on source side had known
bugs with link standby.

However that same HSD report only mentions BDW and HSW and tells that
a fix was going to new platforms. Since on Skylake link standby
didn't cause the bad blank flickering screens seen on HSW and BDW
let's respect VBT again for this and future platforms.

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>

drm/i915: PSR simplify port and link standby checks.

Current code not just block link_standby for non DDI platforms but also
block PSR from work on other ports B/C/D/E.

So, besides change any behaviour let's just fix the mess a bit here and
reuse HSW check to block the other ports and reduce the second if only to
link stadnby request.

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>

drm/i915/skl: Fix DMC load on Skylake J0 and K0

The driver does not load firmware for unknown steppings, so these new
steppings must be added to the list.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454023163-25469-1-git-send-email-mathew.j.martineau@linux.intel.com

drm/i915: Sink CRC: tune down error message at stop to debug_kms.

When we stop the sink CRC calculation we wait a while until the counter
is reset to zero and return -ETIMEDOUT. However the sink crc was
calculated already by this point so we just ignore this return at
the main function.

So, let's also ignore the message and put it as a debug message instead
of an error one. The message might still be useful when debuging
test failures so we could be able to know something was not going so
well with sink crc stop.

v2: Improve log message.

Reference: https://bugs.freedesktop.org/show_bug.cgi?id=93694
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1454107499-29678-1-git-send-email-rodrigo.vivi@intel.com

drm/i915/fbc: refactor some small functions called only once

The FBC fixes we've been doing in the last months required a lot of
refactor, so functions that were once big and called from different
spots are now small and called only once. IMHO now it's better to just
move the contents of these functions to their only callers since this
reduces the number of indirections while reading the code.

While at it, also improve the related comments a little bit.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-26-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't store/check a pointer to the FB

We already make sure we run intel_fbc_update_update during modesets
and page flips, and this function takes care of deactivating FBC, so
it shouldn't be possible for us to reach the condition we check at
intel_fbc_work_fn. So instead of grabbing framebuffer references and
adding a lot of code to track when we need to free them, just don't
track anything at all since we shouldn't need to.

v2: Rebase.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-25-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: call intel_fbc_pre_update earlier during page flips

Make sure we do the pre_update - which also deactivates FBC - before
we actually schedule the page flip, just to make sure we don't
flip to the new FB with FBC still activated for the previous FB.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-24-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't store the fb_id on reg_params

We don't actually use fb_id anywhere. We already compare all
parameters that matter to the hardware: pixel format, stride,
fence_reg and ggtt_offset. The ID shouldn't make a difference.

Besides, we already update the FBC data at every modeset/flip, so this
can't change behind our backs.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-23-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't print no_fbc_reason to dmesg

Our dmesg messages started being misleading after we converted to the
enable+activate model: we always print "Disabling FBC", even when
we're just deactivating it. So, for example, when I boot my machine
and do "dmesg | grep -i fbc", I see:
  [drm:intel_fbc_enable] Enabling FBC on pipe A
  [drm:set_no_fbc_reason] Disabling FBC: framebuffer not tiled or fenced
but then, if I read the debugfs file, I will see:
  $ sudo cat i915_fbc_status
  FBC enabled
  Compressing: yes
so we can conclude that dmesg is misleading, since FBC is actually
enabled. What happened is that we deactivated FBC due to fbcon not
being tiled, but when we silently reactivated it when the display
manager started. We don't print activation messages since there may be
way too many of these operations per second during normal desktop
usage.

One possible solution would be to change set_no_fbc_reason to
correctly differentiate between disable and deactivation, but we
removed support from printing activation/deactivation messages in the
past because they were too frequent. So instead of doing this, let's
just not print anything on dmesg, and leave the debugfs file if the
user needs to investigate something. We already print when we enable
and disable FBC anyway on a given pipe, so this should already help
triaging bugs.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-22-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't try to deactivate FBC if it's not enabled

During FBC invalidation, don't call intel_fbc_deactivate if it's not
enabled. This doesn't fix any bug, but helps making the interface
saner.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-21-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: make FBC work with fastboot

Move intel_fbc_enable to a place where it is called regardless of the
"modeset" variable, and make sure intel_fbc_enable can be called
multiple times without intel_fbc_disable being called.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-20-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: move intel_fbc_{enable, disable} call one level up

Instead of duplicating the calls for every platform, let's just put
them in the correct places inside intel_atomic_commit. This will also
make it easier for us to move the enable call in order to support
fasbtoot.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-19-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: choose the new FBC CRTC during atomic check

This opens the possibility of implementing nicer schemes to choose the
CRTC, such as checking the amount of stolen memory available, or
choosing the best pipe on platforms that don't die FBC to pipe or
plane A.

This code was written for another refactor that I ended up discarding,
so I don't actually need it, but I figured this patch would be an
improvement on its own so I kept it on the series.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-18-git-send-email-paulo.r.zanoni@intel.com

drm/i915: simplify struct drm_device access at intel_atomic_check()

We already have a dev variable, there's no need to access state->dev.
Also, I plan to add another dev_priv user here, so declare one for the
current user.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-17-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: rewrite the multiple_pipes_ok() code for locking

Older FBC platforms have this restriction where FBC can't be enabled
if multiple pipes are enabled. In the current code, we disable FBC
before the second pipe becomes visible.

One of the problems with this code is that the current
multiple_pipes_ok() implementation just iterates through all CRTCs
looking at their states, but it doesn't make sure that the state
locks are grabbed. It also can't just grab the locks for every CRTC
since this would kill one of the biggest advantages of atomic
modesetting.

After the recent FBC changes, we now have the appropriate locks for
the given CRTC, so we can just try to maintain the state of each CRTC
and update it once intel_fbc_pre_update is called.

As a last note, I don't have gen 2/3 machines to test this code. My
current plan is to enable FBC on just the newer platforms, so this
patch is just an attempt to get the gen 2/3 code at least looking
sane, so if one day someone decide to fix FBC on these platforms, they
may have less work to do.

Not-tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com> (only on HSW+)
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-16-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: make sure we cancel the work function at fbc_disable

Just to be sure nothing will survive a module unload. We need to do
this after the unlock in order to make sure the function won't get
stuck trying to grab the lock we already own while we wait for it to
finish.

Reported-by: Reported-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-15-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: rename the FBC disable functions

Instead of:
 - intel_fbc_disable_crtc(crtc)
 - intel_fbc_disable(dev_priv)
we now have:
 - intel_fbc_disable(crtc)
 - intel_fbc_global_disable(dev_priv)

This is because all the other functions that take a CRTC are called
 - intel_fbc_something(crtc)
Instead of:
 - intel_fbc_something_crtc(crtc)

And I also hope that the word "global" is going to help make it more
explicit that "global" is the unusual case, not the opposite.

Reported-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-14-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: unexport intel_fbc_deactivate

With the addition and usage of intel_fbc_pre_update,
intel_fbc_deactivate is not used anymore outside intel_fbc.c, so kill
the exported function and rename __intel_fbc_deactivate.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-13-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: fix the FBC state checking code

We'll now call intel_fbc_pre_update instead of intel_fbc_deactivate
during atomic commits. This will continue to guarantee that we
deactivate FBC and it will also update the state checking structures
at the correct time. Then, later, at the point where we were calling
intel_fbc_update, we'll only need to call intel_fbc_post_update.

Also add the proper warnings in case we don't have the appropriate
locks. Daniel mentioned the warnings will have to be removed for async
commits, but let's keep them here while we can.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-12-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: split intel_fbc_update into pre and post update

So now pre_update will be responsible for unconditionally deactivating
FBC and updating the state cache, while post_update will be
responsible for checking if it can be enabled, then enabling it.

This is one more step into proper locking.

Notice that intel_fbc_flush now calls post_update directly. The FBC
flush can only happen for drawing operations - since we explicitly
ignore the flips -, so the FBC state is not expected to have changed
at this point. With this we can just run post_update, which will make
sure we won't deactivate+reactivate FBC as would be the case now if we
called pre_update + post_update.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-11-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: introduce struct intel_fbc_state_cache

Per the new atomic locking rules, we need to cache the CRTC, plane and
FB state structures we use so we can access them later without needing
more locks. So do this.

Notice that there are some pieces of the FBC code that look at things
that are only computed during the modeset, so we can't just can't
precompute whether FBC can be activated during the update_state_cache
stage. We may be able to do this later.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-10-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: unconditionally update FBC during atomic commits

We unconditionally disable/update FBC even during the page flip
IOCTLs, and an unconditional disable/update at every atomic commit
touching the primary plane shouldn't impact PC state residency
noticeably. Besides, the code that checks for rotation is a good hint
that we may be forgetting something else, so let's leave all the
decisions to intel_fbc.c, making the code much safer.

Once we have the code to properly make FBC enable/update decisions
based on atomic states, with proper locking, then we'll be able to
evaluate whether it will be worth trying to optimize the cases where a
disable isn't needed.

v2: Upstream moved and now our patch needs to remove dev_priv.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453406837-10511-1-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't flush for operations on the wrong frontbuffer

If frontbuffer_bits doesn't match the current frontbuffer, there's no
reason to recompress or update FBC.

There was a plan to make the FBC test suite catch this type of
problem, but it never got implemented due to being low priority.

While at it, also implement Ville's suggestion and use
plane->frontbuffer_bit instead of INTEL_FRONTBUFFER_PRIMARY.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-8-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: don't use the frontbuffer tracking subsystem for flips

Before this patch, page flips would call intel_frontbuffer_flip() and
intel_frontbuffer_flip_complete(), which would call intel_fbc_flush(),
which would call intel_fbc_update(). The problem is that drawing
operations also trigger intel_fbc_flush() calls, so it's not
guaranteed that we have the CRTC and FB locks grabbed when
intel_fbc_flush() happens, since the call trace may come from the
rendering path.

We're trying to make the FBC code grab the appropriate CRTC/FB locks,
so split the drawing and the flipping logic in order to achieve that
in later patches. So now the frontbuffer tracking code is just going
to be used for frontbuffer drawing, and intel_fbc_update() is going to
be used directly for actual page flips.

As a note, we don't need to call intel_fbc_flip() during the two
places where we call intel_frontbuffer_flip() since in one of them we
already have an intel_fbc_update() call, and in the other we have the
planes disabled.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-7-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: replace frequent dev_priv->fbc.x with fbc->x

We say "dev_priv->fbc.something" way too many times in our code while
we could be saying just "fbc->something" with a previous declaration
of fbc. This has been bothering me for a while but I didn't want to
patch it since I wanted to fix the real problems first. But as I add
more code I keep thinking about it, especially since it makes the code
easier to read and it can make us fit 80 columns easier, so let's just
do the change now.

While at it, also rename from i915_fbc to intel_fbc because the whole
FBC code uses intel_fbc.

v2: Rebase after the work_fn changes.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453406763-10400-1-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: introduce struct intel_fbc_reg_params

The early return inside __intel_fbc_update does not completely check
all the parameters that affect the FBC register values. For example,
we currently lack looking at crtc->adjusted_y (for the fence Y offset)
and all the parameters that affect the CFB size (for i8xx).

Instead of just adding the missing parameters to the check and hoping
that any changes to the fbc_activate functions also come with a
matching change to the __intel_fbc_update check, introduce a new
structure where we store these parameters and use the structure at the
fbc_activate function. Of course, it's still possible to access
everything from dev_priv in those functions, but IMHO the new code
will be harder to break.

v2: Rebase.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-5-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: extract intel_fbc_can_enable()

Make our enable/activate checking model more explicit, especially
since we now have intel_fbc_can_activate().

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-4-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: extract intel_fbc_can_activate()

Extract all the code that checks if the FBC configuration is valid to
its own function, making __intel_fbc_update() much simpler.

Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453210558-7875-3-git-send-email-paulo.r.zanoni@intel.com

drm/i915/fbc: wait for a vblank instead of 50ms when enabling

Instead of waiting for 50ms, just wait until the next vblank, since
it's the minimum requirement. The whole infrastructure of FBC is based
on vblanks, so waiting for X vblanks instead of X milliseconds sounds
like the correct way to go. Besides, 50ms may be less than a vblank on
super slow modes that may or may not exist.

There are some small improvements in PC state residency (due to the
fact that we're now using 16ms for the common modes instead of 50ms),
but the biggest advantage is still the correctness of being
vblank-based instead of time-based.

v2:
  - Rebase after changing the patch order.
  - Update the commit message.
v3:
  - Fix bogus vblank_get() instead of vblank_count() (Ville).
  - Don't forget to call drm_crtc_vblank_{get,put} (Chris, Ville)
  - Adjust the performance details on the commit message.
v4:
  - Don't grab the FBC mutex just to grab the vblank (Maarten)

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453406585-10233-1-git-send-email-paulo.r.zanoni@intel.com

drm/i915: Fix file permissions

No functional change

Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Fixes: f8d03ea0053b ("drm/i915: increase the tries for HDMI hotplug live status checking")
Link: http://patchwork.freedesktop.org/patch/msgid/1454023325-26265-1-git-send-email-mathew.j.martineau@linux.intel.com
Signed-off-by: Jani Nikula <jani.nikula@intel.com>

drm/i915: refine qemu south bridge detection

The test for the qemu q35 south bridge added by commit
"39bfcd52 drm/i915: more virtual south bridge detection"
also matches on real hardware.  Having the check for
virtual systems last in the list is not enough to avoid
that ...

Refine the check by additionally verifying the pci
subsystem id to see whenever it *really* is qemu.

[ v2: fix subvendor tyops ]

Reported-and-tested-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Tested-by: Bruno Wolff III <bruno@wolff.to>
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453719748-10944-1-git-send-email-kraxel@redhat.com

drm/i915: Fix intel_tile_width() parameters

The fb_modifiers and cpp arguments passed to intel_tile_width() in
intel_fill_fb_ggtt_view() got accidentally swapped around. I'm pretty
sure I fixed this already, but could be I lost the fix accidentally
during some rebases or something. Anyway, fix it up for real.

Fixes: d9b3288ecf2f ("drm/i915: change intel_fill_fb_ggtt_view() to use the real tile size")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-8-git-send-email-ville.syrjala@linux.intel.com
Testcase: igt/kms_rotation_crc/primary-rotation-90
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: drm-intel-fixes@lists.freedesktop.org

drm/i915: Standardize on 'cpp' for bytes per pixel

We more or less randomly call the "bytes per pixel" value
'cpp', 'bytes_per_pixel', 'pixel_size', or even 'bpp'. Let's just pick
one and stick to it. I've chosen 'cpp'.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-6-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Make display gtt offsets u32

Using 'unsigned long' for ggtt offsets doesn't make much sense. Use
'u32' instead since we've not yet seen a >4GiB ggtt.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-5-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Pass rotation_info to intel_rotate_fb_obj_pages()

intel_rotate_fb_obj_pages() doens't need the entire gtt view, just the
rotation info suffices.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-4-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Pass stride to rotate_pages()

Pass stride in addition to width and height to rotate_pages(). For now
width and stride are the same, but once framebuffer offsets enter the
scene that may no longer be the case.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-3-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Rename the rotated gtt view member to 'rotated'

Also rename 'rotation_info' to 'rotated' to match the view type exactly,
this should avoid confusion which union members is valid for each view
type.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453316739-13296-2-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Fix pointer tests in error-handling paths

In the error-handling paths of i915_gem_do_execbuffer() and
intel_crtc_page_flip(), the local pointer-to-request variables
were expected to be either valid pointers or NULL. Since

  2682708 drm/i915: simplify allocation of driver-internal requests

they could also be ERR_PTR() values, so the tests need to be
updated to accommodate this case.

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453978089-29127-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Fix premature LRC unpin in GuC mode

In GuC mode LRC pinning lifetime depends exclusively on the
request liftime. Since that is terminated by the seqno update
that opens up a race condition between GPU finishing writing
out the context image and the driver unpinning the LRC.

To extend the LRC lifetime we will employ a similar approach
to what legacy ringbuffer submission does.

We will start tracking the last submitted context per engine
and keep it pinned until it is replaced by another one.

Note that the driver unload path is a bit fragile and could
benefit greatly from efforts to unify the legacy and exec
list submission code paths.

At the moment i915_gem_context_fini has special casing for the
two which are potentialy not needed, and also depends on
i915_gem_cleanup_ringbuffer running before itself.

v2:
 * Move pinning into engine->emit_request and actually fix
   the reference/unreference logic. (Chris Wilson)

 * ring->dev can be NULL on driver unload so use a different
   route towards it.

v3:
 * Rebase.
 * Handle the reset path. (Chris Wilson)
 * Exclude default context from the pinning - it is impossible
   to get it right before default context special casing in
   general is eliminated.

v4:
 * Rebased & moved context tracking to
   intel_logical_ring_advance_and_submit.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Issue: VIZ-4277
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Nick Hoath <nicholas.hoath@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453976997-25424-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Extract context unpinning to its own function

Will enable cleaner implementation of a following fix and
easier code unification in the future.

Idea and code by Chris Wilson.

v2: Do not return before last_contexts on engines are unpinned.

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

drm/i915: Make LRC pinning own a reference to the context

Will simplify the following fix and sounds logical.

v2: Add some whitespace to separate logic better. (Chris Wilson)

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

drm/i915: Make LRC (un)pinning work on context and engine

Previously intel_lr_context_(un)pin were operating on requests
which is in conflict with their names.

If we make them take a context and an engine, it makes the names
make more sense and it also makes future fixes possible.

v2: Rebase for default_context/kernel_context change.

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

drm/i915: Fix VCS ring selection after uapi decoupling

This got broken in:

   commit de1add360522c876c25ef2bbbbab1c94bdb509ab
   Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
   Date:   Fri Jan 15 15:12:50 2016 +0000

       drm/i915: Decouple execbuf uAPI from internal implementation

BSD ring flags need to be shifted before they can be considered
indices into the ring array.

Reported by Zhipeng Gong.

v2: Simplify the code. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Zhipeng Gong <zhipeng.gong@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1453902069-31353-1-git-send-email-tvrtko.ursulin@linux.intel.com
Testcase: igt/gem_exec_basic # bdw-gt3

drm/i915: Move stolen memory initialization earlier during loading

The only device specific dependency of the stolen memory setup is the
MMIO mapping and the stolen memory size. Both are already available in
i915_gtt_init(), so move the stolen initialization to there. The
clean-up code for i915_gtt_init() is in i915_global_gtt_cleanup(), so
move the stolen memory clean-up code there too.

This will be needed by an upcoming patch that needs the details of the
memory we reserve, but the change is also part of our generic goal to
move the initialization of resources with no or little dependencies on
other device specific resources towards the beginning of the init
sequence.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-8-git-send-email-imre.deak@intel.com

drm/i915: Move MCHBAR setup earlier during init

Move the MCHBAR setup right after the MMIO setup, since the two things
are logically related and the MCHBAR setup code doesn't depend on any
other device specific resource. We'll also need MCHBAR to be ready
earlier in an upcoming patch, so this is also a preparation for that.

Factor out the init/clean-up code to separate functions to make things
clearer in the i915_driver_load()/unload() functions.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-7-git-send-email-imre.deak@intel.com

drm/i915: Move allocation of various workqueues earlier during init

Workqueue initalization doesn't depend on any other device specific
resource, so move it close to the beginning, so we don't need to
consider them when thinking about dependencies for other resources.

Also factor out things to separate init/cleanup functions to make
i915_driver_load()/unload() clearer, atm it's somewhat difficult to
follow there in what order resources are inited/cleaned-up.

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-6-git-send-email-imre.deak@intel.com

drm/i915: Sanitize i915_gem_load() init and clean-up

Factor out common clean-up code for the GEM load time init function.
Also rename i915_gem_load() to i915_gem_load_init() to have a better
match with its new clean-up function.

No functional change.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-5-git-send-email-imre.deak@intel.com

drm/i915: Sanitize GEM shrinker init and clean-up

Factor out the common GEM shrinker clean-up code and call the shrinker
init function from the same function from where the corresponding
shrinker clean-up function is called. Also add sanity checking to the
shrinker and OOM registration calls.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-4-git-send-email-imre.deak@intel.com

drm/i915: Sanitize i915_get_bridge_dev() error path

Clarify the name of the label on the error path, making it clear what's
being cleaned up. The kmem_cache_destroy() calls are NOPs on the
corresponding error path.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-3-git-send-email-imre.deak@intel.com

drm/i915: Sanitize DMC/CSR ucode cleanup code

commit ebae38d061df3deffa7c17b030ea14a5216ee55f
Author: Animesh Manna <animesh.manna@intel.com>
Date:   Wed Oct 28 23:58:55 2015 +0200

    drm/i915/gen9: csr_init after runtime pm enable

moved the DMC/CSR initialization later during driver loading, but didn't
move the cleanup earlier correspondingly during unloading. Fix this up.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: David Weinehall <david.weinehall@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453209992-25995-2-git-send-email-imre.deak@intel.com

Revert "drm/i915: Fix context/engine cleanup order"

This reverts commit 1803c035efb88afb9d3e7feb279ac29a83216382.

It seems to blow up on module unload due to a use-after free hitting a
BUG_ON with CONFIG_DEBUG_SG. Quoting from Tvrtko's mail:

"I've decoded the instructions and it pointed to SG_MAGIC checking:

488b8098010000  mov 0x198(%rax),%rax
ba21436587      mov $0x87654321,%edx
488b00          mov (%rax),%rax       *** CRASH

"Grep showed 0x87654321 is SG_MAGIC, so likely candidate for this code
pattern is:

static inline struct page *sg_page(struct scatterlist *sg)
{
    BUG_ON(sg->sg_magic != SG_MAGIC);
    BUG_ON(sg_is_chain(sg));
    return (struct page *)((sg)->page_link & ~0x3);
}

"Which would mean the offender is in intel_logical_ring_cleanup is most
likely:

...
    if (ring->status_page.obj) {
        kunmap(sg_page(ring->status_page.obj->pages->sgl));
        ring->status_page.obj = NULL;
    }
...

"I think that the i915_gem_context_fini will do a final unref on
dev_priv->kernel_context and then the ring buff has a copy which is
left dangling because:

    lrc_setup_hardware_status_page(ring,
        dev_priv->kernel_context->engine[ring->id].state);

and:

ring->status_page.obj = default_ctx_obj;

"Where default_ctx_obj == dev_priv->kernel_context->engine[ring->id].state
So indeed looks like the unload ordering is the trigger.  In fact it
is almost the same fragility wrt/ kernel_context hidden dependency I
expressed my worry about in an e-mail yesterday or so. It only shows
if CONFIG_DEBUG_SG is set, otherwise it accesses freed memory and
probably just survives."

This causes serious trouble in our CI system since it took out all
gen8+ machines. Not yet clear why this wasn't caught in pre-merge
testing.

Backtrace from CI, for posterity:

[  163.737836] general protection fault: 0000 [#1] PREEMPT SMP
[  163.737849] Modules linked in: ax88179_178a usbnet mii snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic i915(-) x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel snd_hda_codec snd_hwdep snd_hda_core snd_pcm mei_me mei i2c_hid e1000e ptp pps_core [last unloaded: snd_hda_intel]
[  163.737902] CPU: 0 PID: 5812 Comm: rmmod Tainted: G     U  W       4.5.0-rc1-gfxbench+ #1
[  163.737911] Hardware name: System manufacturer System Product Name/Z170M-PLUS, BIOS 0505 11/16/2015
[  163.737920] task: ffff8800bb99cf80 ti: ffff88022ff2c000 task.ti: ffff88022ff2c000
[  163.737928] RIP: 0010:[<ffffffffa018f723>]  [<ffffffffa018f723>] intel_logical_ring_cleanup+0x83/0x100 [i915]
[  163.737969] RSP: 0018:ffff88022ff2fd30  EFLAGS: 00010282
[  163.737975] RAX: 6b6b6b6b6b6b6b6b RBX: ffff8800bb2f31b8 RCX: 0000000000000002
[  163.737982] RDX: 0000000087654321 RSI: 000000000000000d RDI: ffff8800bb2f31f0
[  163.737989] RBP: ffff88022ff2fd40 R08: 0000000000000000 R09: 0000000000000001
[  163.737996] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800bb2f0000
[  163.738003] R13: ffff8800bb2f8fc8 R14: ffff8800bb285668 R15: 000055af1ae55210
[  163.738010] FS:  00007f187014b700(0000) GS:ffff88023bc00000(0000) knlGS:0000000000000000
[  163.738021] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  163.738030] CR2: 0000558f84e4cbc8 CR3: 000000022cd55000 CR4: 00000000003406f0
[  163.738039] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  163.738048] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  163.738057] Stack:
[  163.738062]  ffff8800bb2f31b8 ffff8800bb2f0000 ffff88022ff2fd70 ffffffffa0180414
[  163.738079]  ffff8800bb2f0000 ffff8800bb285668 ffff8800bb2856c8 ffffffffa0242460
[  163.738094]  ffff88022ff2fd98 ffffffffa0202d30 ffff8800bb285668 ffff8800bb285668
[  163.738109] Call Trace:
[  163.738140]  [<ffffffffa0180414>] i915_gem_cleanup_engines+0x34/0x60 [i915]
[  163.738185]  [<ffffffffa0202d30>] i915_driver_unload+0x150/0x270 [i915]
[  163.738198]  [<ffffffff815100f4>] drm_dev_unregister+0x24/0xa0
[  163.738208]  [<ffffffff815106ce>] drm_put_dev+0x1e/0x60
[  163.738225]  [<ffffffffa01412a0>] i915_pci_remove+0x10/0x20 [i915]
[  163.738237]  [<ffffffff8143d9b4>] pci_device_remove+0x34/0xb0
[  163.738249]  [<ffffffff81533d15>] __device_release_driver+0x95/0x140
[  163.738259]  [<ffffffff81533eb6>] driver_detach+0xb6/0xc0
[  163.738268]  [<ffffffff81532de3>] bus_remove_driver+0x53/0xd0
[  163.738278]  [<ffffffff815348d7>] driver_unregister+0x27/0x50
[  163.738289]  [<ffffffff8143ca15>] pci_unregister_driver+0x25/0x70
[  163.738299]  [<ffffffff81511de4>] drm_pci_exit+0x74/0x90
[  163.738337]  [<ffffffffa02034a9>] i915_exit+0x20/0x1a5 [i915]
[  163.738349]  [<ffffffff8110400f>] SyS_delete_module+0x18f/0x1f0
[  163.738361]  [<ffffffff817b8a9b>] entry_SYSCALL_64_fastpath+0x16/0x73
[  163.738370] Code: ff d0 48 89 df e8 de a1 fd ff 48 8d 7b 38 e8 25 ab fd ff 48 8b 83 90 00 00 00 48 85 c0 74 25 48 8b 80 98 01 00 00 ba 21 43 65 87 <48> 8b 00 48 39 10 75 3c f6 40 08 01 75 38 48 c7 83 90 00 00 00
[  163.738459] RIP  [<ffffffffa018f723>] intel_logical_ring_cleanup+0x83/0x100 [i915]
[  163.738498]  RSP <ffff88022ff2fd30>
[  163.738507] ---[ end trace 68f69ce4740fa44f ]---

Cc: Nick Hoath <nicholas.hoath@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Tested-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>

drm/i915: Fix context/engine cleanup order

Swap the order of context & engine cleanup, so that contexts are cleaned
up first, and *then* engines. This is a more sensible order anyway, but
in particular has become necessary since the 'intel_ring_initialized()
must be simple and inline' patch, which now uses ring->dev as an
'initialised' flag, so it can now be NULL after engine teardown. This
in turn can cause a problem in the context code, which (used to) check
the ring->dev->struct_mutex -- causing a fault if ring->dev was NULL.

Also rename the cleanup function to reflect what it actually does
(cleanup engines, not a ringbuffer), and fix an annoying whitespace issue.

v2: Also make the fix in i915_load_modeset_init, not just in
    i915_driver_unload (Chris Wilson)
v3: Had extra stuff in it.
v4: Reverted extra stuff (so we're back to v2).
    Rebased and updated commentary above (Dave Gordon).

Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> (v2)
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1453405067-32890-3-git-send-email-david.s.gordon@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Improve handling of overlapping objects

The generic interval tree we use to speed up range invalidation is an
augmented rbtree that can report all overlapping intervals for a given
range. Therefore we do not need to degrade to a linear list if we find
overlapping objects. Oops.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453397563-2848-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Add WaOCLCoherentLineFlush

This is mainly required for future enabling of pre-emptive
command execution.

v2: explain purpose of change (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-9-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Enable Per context Preemption granularity control

Per context preemption granularity control is only available from SKL:E0+

Actual WA is to disable percontext preemption granularity control until D0
which is the default case so this is equivalent to the inverse of
WaDisablePerCtxtPreemptionGranularityControl:skl

v2: add some detail to commit msg (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-8-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/skl: Add GEN8_L3SQCREG4 to HW whitelist

Required for WaDisableLSQCROPERFforOCL:skl

This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.

v2: explain purpose of changes (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-7-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bxt: Add GEN8_L3SQCREG4 to HW whitelist

Required for WaDisableLSQCROPERFforOCL:bxt

According to WA database these are only applicable for BXT:A0 but since
A0 and A1 shares the same GT these are extended for A1 as well.

This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.

v2: explain purpose of changes (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-6-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/bxt: Add GEN9_CS_DEBUG_MODE1 to HW whitelist

Required for,
WaDisableObjectLevelPreemptionForTrifanOrPolygon:bxt
WaDisableObjectLevelPreemptionForInstancedDraw:bxt
WaDisableObjectLevelPreemtionForInstanceId:bxt

According to WA database these are only applicable for BXT:A0 but since
A0 and A1 shares the same GT these are extended for A1 as well.

These are also required for SKL until B0 but not adding them because they
are pre-production steppings.

This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.

v2: use lower case in register defines (Nick)
v3: explain purpose of changes (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-5-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Add HDC_CHICKEN1 to HW whitelist

Required for WaAllowUMDToModifyHDCChicken1:skl,bxt

This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.

v2: explain purpose of changes (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-4-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Add GEN8_CS_CHICKEN1 to HW whitelist

Required for WaEnablePreemptionGranularityControlByUMD:skl,bxt

This register is added to HW whitelist to support WA required for future
enabling of pre-emptive command execution, WA implementation will be in
userspace and it cannot program this register if it is not on HW whitelist.

v2: explain purpose of WA (Chris)

Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-3-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/gen9: Add framework to whitelist specific GPU registers

Some of the HW registers are privileged and cannot be written to from
non-privileged batch buffers coming from userspace unless they are added to
the HW whitelist. This whitelist is maintained by HW and it is different from
SW whitelist. Userspace need write access to them to implement preemption
related WA.

The reason for using this approach is, the register bits that control
preemption granularity at the HW level are not context save/restored; so even
if we set these bits always in kernel they are going to change once the
context is switched out.  We can consider making them non-privileged by
default but these registers also contain other chicken bits which should not
be allowed to be modified.

In the later revisions controlling bits are save/restored at context level but
in the existing revisions these are exported via other debug registers and
should be on the whitelist. This patch adds changes to provide HW with a list
of registers to be whitelisted. HW checks this list during execution and
provides access accordingly.

HW imposes a limit on the number of registers on whitelist and it is
per-engine.  At this point we are only enabling whitelist for RCS and we don't
foresee any requirement for other engines.

The registers to be whitelisted are added using generic workaround list
mechanism, even these are only enablers for userspace workarounds. But by
sharing this mechanism we get some test assets without additional cost (Mika).

v2: rebase

v3: parameterize RING_FORCE_TO_NONPRIV() as _MMIO() should be limited to
i915_reg.h (Ville), drop inline for wa_ring_whitelist_reg (Mika).

v4: improvements suggested by Chris Wilson.
Clarify that this is HW whitelist and different from the one maintained in
driver. This list is engine specific but it gets initialized along with other
WA which is RCS specific thing, so make it clear that we are not doing any
cross engine setup during initialization.
Make HW whitelist count of each engine available in debugfs.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453412634-29238-2-git-send-email-arun.siluvery@linux.intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Remove select to deleted STOP_MACHINE from Kconfig

Commit 5bab6f60cb4d ("drm/i915: Serialise updates to GGTT with access
through GGTT on Braswell") depended upon a working stop_machine() and
so forced the selection of STOP_MACHINE. However, commit 86fffe4a61dd
("kernel: remove stop_machine() Kconfig dependency") removed the option
STOP_MACHINE from init/Kconfig and ensured that stop_machine()
universally works. Due to the order in which the patches were applied,
removing the select from DRM_I915 got lost during merging.

Remove the now obsolete select statement.

Signed-off-by: Andreas Ziegler <andreas.ziegler@fau.de>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1453722079-2604-1-git-send-email-andreas.ziegler@fau.de
Cc: drm-intel-fixes@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915/guc: Decouple GuC engine id from ring id

Previously GuC uses ring id as engine id because of same definition.
But this is not true since this commit:

commit de1add360522c876c25ef2bbbbab1c94bdb509ab
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date:   Fri Jan 15 15:12:50 2016 +0000

    drm/i915: Decouple execbuf uAPI from internal implementation

Added GuC engine id into GuC interface to decouple it from ring id used
by driver.

v2: Keep ring name print out in debugfs; using for_each_ring() where
    possible to keep driver consistent. (Chris W.)

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453579094-29860-1-git-send-email-yu.dai@intel.com

drm/i915: More use of the cached LRC state

Since:

commit 82352e908acd36d7244c75a008c9f27a2ced44d5
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date:   Fri Jan 15 17:12:45 2016 +0000

    drm/i915: Cache LRC state page in the context

and:

commit 0eb973d31d0aadb6bc801fd6d796afecbbfc3d5b
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date:   Fri Jan 15 15:10:28 2016 +0000

    drm/i915: Cache ringbuffer GTT VMA

We can also remove the ring buffer start updates on every
context update since the address will not change for the
duration of the LRC pin.

For GuC we can remove the update altogether because it
only cares about the ring buffer start.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Alex Dai <yu.dai@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1453466567-33369-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Update DRIVER_DATE to 20160124

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

drm/i915: Seal busy-ioctl uABI and prevent leaking of internal ids

Tvrtko was looking through the execbuffer-ioctl and noticed that the
uABI was tightly coupled to our internal engine identifiers. Close
inspection also revealed that we leak those internal engine identifiers
through the busy-ioctl, and those internal identifiers already do not
match the user identifiers. Fortuitiously, there is only one user of the
set of busy rings from the busy-ioctl, and they only wish to choose
between the RENDER and the BLT engines.

Let's fix the userspace ABI while we still can.

v2: Update the uAPI documentation to explain the identifiers.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Testcase: igt/gem_busy
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1452876706-21620-1-git-send-email-chris@chris-wilson.co.uk

drm/i915: Decouple execbuf uAPI from internal implementation

At the moment execbuf ring selection is fully coupled to
internal ring ids which is not a good thing on its own.

This dependency is also spread between two source files and
not spelled out at either side which makes it hidden and
fragile.

This patch decouples this dependency by introducing an explicit
translation table of execbuf uAPI to ring id close to the only
call site (i915_gem_do_execbuffer).

This way we are free to change driver internal implementation
details without breaking userspace. All state relating to the
uAPI is now contained in, or next to, i915_gem_do_execbuffer.

As a side benefit, this patch decreases the compiled size
of i915_gem_do_execbuffer.

v2: Extract ring selection into eb_select_ring. (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1452870770-13981-1-git-send-email-tvrtko.ursulin@linux.intel.com

drm/i915: Use ordered seqno write interrupt generation on gen8+ execlists

Broadwell and later currently use the same unordered command sequence to
update the seqno in the HWS status page and then assert the user
interrupt. We should apply the w/a from legacy (where we do an mmio
read to delay the seqno read after the interrupt), but this is not
enough to enforce coherent seqno visibilty on Skylake. Rather than
search for the proper post-interrupt seqno barrier, use a strongly
ordered command sequence to write the seqno, then assert the user
interrupt from the ring.

v2: Move around the wa tail dwords to avoid adding duplicate code.

v3: Add references, comments on workarounds and bit5 check.

References: https://bugs.freedesktop.org/show_bug.cgi?id=93693
Testcase: igt/gem_ring_sync_loop #skl
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453297415-17793-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Limit the auto arming of mmio debugs on vlv/chv

The capability to detect unclaimed register access was
recently introduced for vlv/chv platforms. Apparently
there are plenty of unclaimed access on these platforms,
resulting in new dmesg warns. But as we are trying to form
a beachhead for CI/Bat, all new warns are adding to the
noise and thus not desirable at this point in time.

Make it so that if in these platforms the automatic arming
was responsible for mmio_debug enabling, ignore the warns.

If user/dev wants to fix these, he can still do so by
i915.mmio_debug=1234.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1453285943-24614-1-git-send-email-mika.kuoppala@intel.com

drm/i915: Tune down "GT register while GT waking disabled" message

We've had this since forever, and's randomly reporting issues and as
such causing piles&piles of CI noise. Mika is working on proper debug
infrastructure for this, and on fixing this properly.

Meanwhile make CI more useful for everyone else.

Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93121
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453233656-12955-1-git-send-email-daniel.vetter@ffwll.ch

drm/i915: tidy up a few leftovers

There are a few bits of code which the transformations implemented by
the previous patch reveal to be suboptimal, once the notion of a per-
ring default context has gone away. So this tidies up the leftovers.

It could have been squashed into the previous patch, but that would have
made that patch less clearly a simple transformation. In particular, any
change which alters the code block structure or indentation has been
deferred into this separate patch, because such things tend to make
diffs more difficult to read.

v4: Rebased

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453230175-19330-4-git-send-email-david.s.gordon@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: abolish separate per-ring default_context pointers

Now that we've eliminated a lot of uses of ring->default_context,
we can eliminate the pointer itself.

All the engines share the same default intel_context, so we can just
keep a single reference to it in the dev_priv structure rather than one
in each of the engine[] elements. This make refcounting more sensible
too, as we now have a refcount of one for the one pointer, rather than
a refcount of one but multiple pointers.

From an idea by Chris Wilson.

v2: transform an extra instance of ring->default_context introduced by
    42f1cae8c drm/i915: Restore inhibiting the load of the default context
    That patch's commentary includes:
v2: Mark the global default context as uninitialized on GPU reset so
    that the context-local workarounds are reloaded upon re-enabling
    The code implementing that now also benefits from the replacement of
    the multiple (per-ring) pointers to the default context with a single
    pointer to the unique kernel context.

v4: Rebased, remove underused local (Nick Hoath)

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1453230175-19330-3-git-send-email-david.s.gordon@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: simplify allocation of driver-internal requests

There are a number of places where the driver needs a request, but isn't
working on behalf of any specific user or in a specific context. At
present, we associate them with the per-engine default context. A future
patch will abolish those per-engine context pointers; but we can already
eliminate a lot of the references to them, just by making the allocator
allow NULL as a shorthand for "an appropriate context for this ring",
which will mean that the callers don't need to know anything about how
the "appropriate context" is found (e.g. per-ring vs per-device, etc).

So this patch renames the existing i915_gem_request_alloc(), and makes
it local (static inline), and replaces it with a wrapper that provides
a default if the context is NULL, and also has a nicer calling
convention (doesn't require a pointer to an output parameter). Then we
change all callers to use the new convention:
OLD:
err = i915_gem_request_alloc(ring, user_ctx, &req);
if (err) ...
NEW:
req = i915_gem_request_alloc(ring, user_ctx);
if (IS_ERR(req)) ...
OLD:
err = i915_gem_request_alloc(ring, ring->default_context, &req);
if (err) ...
NEW:
req = i915_gem_request_alloc(ring, NULL);
if (IS_ERR(req)) ...

v4: Rebased

Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453230175-19330-2-git-send-email-david.s.gordon@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Fix NULL plane->fb oops on SKL

In this atomic age, we can't trust the plane->fb pointer anymore.
It might get update too late. Instead we are supposed to use the
plane_state->fb pointer instead. Let's do that in
intel_plane_obj_offset() and avoid problems from dereferencing the
potentially stale plane->fb pointer.

Paulo found this with 'kms_frontbuffer_tracking --show-hidden --run-subtest nop-1p-rte'
but it can be reproduced with just plain old kms_setplane.

I was too lazy to bisect this, so not sure exactly when it broke. The
most obvious candidate
commit ce7f17285639 ("drm/i915: Fix i915_ggtt_view_equal to handle rotation correctly")
was actually still fine, so it must have broken some time after that.

Here's the resulting fireworks:
BUG: unable to handle kernel NULL pointer dereference at           (null)
IP: [<ffffffffa02d2d9a>] intel_fill_fb_ggtt_view+0x1b/0x15a [i915]
PGD 8a5f6067 PUD 8a5f5067 PMD 0
Oops: 0000 [#1] PREEMPT SMP
Modules linked in: i915 i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops drm intel_gtt agpgart netconsole mousedev hid_generic psmouse usbhid atkbd libps2 coretemp hwmon efi_pstore intel_rapl iosf_mbi x86_pkg_temp_thermal efivars pcspkr e1000e sdhci_pci ptp pps_core sdhci i2c_i801 mmc_core i2c_hid hid i8042 serio evdev sch_fq_codel ip_tables x_tables ipv6 autofs4
CPU: 1 PID: 260 Comm: kms_plane Not tainted 4.4.0-skl+ #171
Hardware name: Intel Corporation Skylake Client platform/Skylake Y LPDDR3 RVP3, BIOS SKLSE2R1.R00.B104.B00.1511030553 11/03/2015
task: ffff88008bde2d80 ti: ffff88008a6ec000 task.ti: ffff88008a6ec000
RIP: 0010:[<ffffffffa02d2d9a>]  [<ffffffffa02d2d9a>] intel_fill_fb_ggtt_view+0x1b/0x15a [i915]
RSP: 0018:ffff88008a6efa10  EFLAGS: 00010086
RAX: 0000000000000001 RBX: ffff8801674f4240 RCX: 0000000000000014
RDX: ffff88008a7440c0 RSI: 0000000000000000 RDI: ffff88008a6efa40
RBP: ffff88008a6efa30 R08: ffff88008bde3598 R09: 0000000000000001
R10: ffff88008b782000 R11: 0000000000000000 R12: 0000000000000000
R13: ffff88008a7440c0 R14: 0000000000000000 R15: ffff88008a7449c0
FS:  00007fa0c07a28c0(0000) GS:ffff88016ec40000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000000 CR3: 000000008a6ff000 CR4: 00000000003406e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Stack:
 ffff8801674f4240 0000000000000000 ffff88008a7440c0 0000000000000000
 ffff88008a6efaa0 ffffffffa02daf25 ffffffff814ec80e 0000000000070298
 ffff8800850d0000 ffff88008a6efaa0 ffffffffa02c49c2 0000000000000002
Call Trace:
 [<ffffffffa02daf25>] intel_plane_obj_offset+0x2d/0xa9 [i915]
 [<ffffffff814ec80e>] ? _raw_spin_unlock_irqrestore+0x4b/0x60
 [<ffffffffa02c49c2>] ? gen9_write32+0x2e8/0x3b8 [i915]
 [<ffffffffa02eecfc>] skl_update_plane+0x203/0x4c5 [i915]
 [<ffffffffa02ca1ab>] intel_plane_atomic_update+0x53/0x6a [i915]
 [<ffffffffa02494a4>] drm_atomic_helper_commit_planes_on_crtc+0x142/0x1d5 [drm_kms_helper]
 [<ffffffffa02de44b>] intel_atomic_commit+0x1262/0x1350 [i915]
 [<ffffffffa024a0ee>] ? __drm_atomic_helper_crtc_duplicate_state+0x2f/0x41 [drm_kms_helper]
 [<ffffffffa01ef089>] ? drm_atomic_check_only+0x3e3/0x552 [drm]
 [<ffffffffa01ef245>] drm_atomic_commit+0x4d/0x52 [drm]
 [<ffffffffa024996b>] drm_atomic_helper_update_plane+0xcb/0x118 [drm_kms_helper]
 [<ffffffffa01e42e8>] __setplane_internal+0x1c8/0x224 [drm]
 [<ffffffffa01e477f>] drm_mode_setplane+0x14e/0x172 [drm]
 [<ffffffffa01d8117>] drm_ioctl+0x265/0x3ad [drm]
 [<ffffffffa01e4631>] ? drm_mode_cursor_common+0x158/0x158 [drm]
 [<ffffffff810d00ab>] ? current_kernel_time64+0x5e/0x98
 [<ffffffff810a76ea>] ? trace_hardirqs_on_caller+0x17a/0x196
 [<ffffffff8119880f>] do_vfs_ioctl+0x42b/0x4ea
 [<ffffffff811a2b72>] ? __fget_light+0x4d/0x71
 [<ffffffff81198911>] SyS_ioctl+0x43/0x61
 [<ffffffff814ed057>] entry_SYSCALL_64_fastpath+0x12/0x6f

Cc: drm-intel-fixes@lists.freedesktop.org
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Testcase: igt/kms_plane
Reported-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453220597-28973-1-git-send-email-ville.syrjala@linux.intel.com
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Do not put big intel_crtc_state on the stack

Having this on stack triggers the -Wframe-larger-than=1024 and
is not nice to put such big things on the kernel stack anyway.

This required a little bit of refactoring to handle the new
failure path from vlv_force_pll_on.

v2: Corrected some whitespace.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: John Harrison <john.c.harrison@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1453217117-26125-1-git-send-email-tvrtko.ursulin@linux.intel.com

Revert "drm/i915: Add two-stage ILK-style watermark programming (v10)"

This reverts commit 396e33ae204f52abebec9e578f44c749305500f4.

This commit was triggering some FIFO underrun warnings on ILK-IVB
platforms (but surprisingly not on HSW/BDW that share more or less the
same codepaths).  These underruns were caught by the continuous
integration (CI) system and could be reproduced consistently when
running the basic acceptance tests (BAT) on the affected platforms.

Note that this revert will cause a visible regression for some
end-users; the "flicker when mouse moves between monitors in X" issue
that was reported before this patch was merged will now return.  However
regressions that are visible to CI have higher priority since they
prevent proper testing of future patches on those platforms.  Hopefully
we'll be able to figure out the cause of the underruns quickly and
remerge an improved version of this patch to fix the regression.

Cc: Daniel Vetter <daniel@ffwll.ch>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93640
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Daniel Vetter <daniel@ffwll.ch>
Link: http://patchwork.freedesktop.org/patch/msgid/1453232584-8543-1-git-send-email-matthew.d.roper@intel.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: add DOC: headline to RC6 kernel-doc

Without the DOC:, kernel-doc confuses the documentation block for
something else.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453101588-18008-3-git-send-email-jani.nikula@intel.com

drm/i915: turn some bogus kernel-doc comments to normal comments

Apparently accidental or misplaced /** kernel-doc comments, confusing
the tool. Turn them to normal comments.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453101588-18008-2-git-send-email-jani.nikula@intel.com

drm/i915/sdvo: revert bogus kernel-doc comments to normal comments

The comments were never proper kernel-doc, but with SDVO it's not worth
the trouble to make them kernel-doc. Just turn them into normal
comments.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453101588-18008-1-git-send-email-jani.nikula@intel.com

drm/i915/gen9: Correct max save/restore register count during gpu reset with GuC

In GuC submission mode, driver has to provide a list of registers to be
save/restored during gpu reset, make the max no. of registers value consistent
with that of the value defined in FW. If they are not in sync then register
save/restore during gpu reset won't work as expected.

Cc: Alex Dai <yu.dai@intel.com>
Cc: Dave Gordon <david.s.gordon@intel.com>
Signed-off-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1453132776-22229-1-git-send-email-arun.siluvery@linux.intel.com
Reviewed-by: Alex Dai <yu.dai@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: Demote user facing DMC firmware load failure message

This is an expected error given the lack of the firmware so emit it at
KERN_NOTICE and not KERN_ERROR. Also include the firmware URL in the
user facing message so that the user can investigate and fix the issue
on their own, and also explain the consequence in plain language.

The complete failure message, including the first line from the firmware
loader, becomes

i915 0000:00:02.0: Direct firmware load for i915/skl_dmc_ver1.bin failed with error -2
i915 0000:00:02.0: Failed to load DMC firmware [https://01.org/linuxgraphics/intel-linux-graphics-firmwares], disabling runtime power management.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Sunil Kamath <sunil.kamath@intel.com>
Cc: Daniel Vetter <daniel.vetter@intel.com>
Cc: Animesh Manna <animesh.manna@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452706695-13518-1-git-send-email-chris@chris-wilson.co.uk
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: use hlist_for_each_entry

Use hlist_for_each_entry() instead of hlist_for_each() to simplify
the code.

Signed-off-by: Geliang Tang <geliangtang@163.com>
Link: http://patchwork.freedesktop.org/patch/msgid/b80568b2990ebcc145229a132f045e852ac51ad6.1453126187.git.geliangtang@163.com
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

drm/i915: skl_update_scaler() wants a rotation bitmask instead of bit number

Pass BIT(DRM_ROTATE_0) instead of DRM_ROTATE_0 to skl_update_scaler().
The former is a mask, the latter just the bit number.

Fortunately the only thing skl_update_scaler() does with the rotation
is check if it's 90/270 degrees or not, and so in this case it would
still do the right thing.

Cc: Chandra Konduru <chandra.konduru@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1444917718-28495-1-git-send-email-ville.syrjala@linux.intel.com
Fixes: 6156a45602f9 ("drm/i915: skylake primary plane scaling using shared scalers")
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>

drm/i915: Don't reject primary plane windowing with color keying enabled on SKL+

On SKL+ plane scaling is mutually exclusive with color keying. The code
check for this, but during some refactoring the code got changed to
also reject primary plane windowing when color keying is used. There is
no such restriction in the hardware, so restore the original logic.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: 061e4b8d650a ("drm/i915: clean up atomic plane check functions, v2.")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452883613-28549-1-git-send-email-ville.syrjala@linux.intel.com
Cc: stable@vger.kernel.org
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

drm/i915/guc: Fix a memory leak where guc->execbuf_client is not freed

During driver unloading, the guc_client created for command submission
needs to be released to avoid memory leak.

The struct_mutex needs to be held before tearing down GuC.

v1: Move i915_guc_submission_disable out of i915_guc_submission_fini and
    take struct_mutex lock before release GuC client. (Dave Gordon)
v2: Add the locking for failure case in guc_fw_fetch. (Dave Gordon)
    Add i915_guc_submission_fini for failure case in intel_guc_ucode_load.

Signed-off-by: Alex Dai <yu.dai@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/1452711710-4505-1-git-send-email-yu.dai@intel.com

drm/i915: Cache LRC state page in the context

LRC lifetime is well defined so we can cache the page pointing
to the object backing store in the context in order to avoid
walking over the object SG page list from the interrupt context
without the big lock held.

v2: Also cache the mapping. (Chris Wilson)
v3: Unmap on the error path.
v4: No need to cache the page. (Chris Wilson)
v5: No need to dirty the page on unpin. (Chris Wilson)
v6: kmap() cannot fail and use kmap_to_page to simplify unpin.
    (Chris Wilson)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/1452877965-32042-1-git-send-email-tvrtko.ursulin@linux.intel.com