Before queuing the flip but crucially after attaching the unpin-work to
the crtc, we continue to setup the unpin-work. However, should the
hardware fire early, we see the connected unpin-work and queue the task.
The task then promptly runs and unpins the fb before we finish taking
the required references or even pinning it... Havoc.

To close the race, we use the flip-pending atomic to indicate when the
flip is finally setup and enqueued. So during the flip-done processing,
we can check more accurately whether the flip was expected.

v2: Add the appropriate mb() to ensure that the writes to the page-flip
worker are complete prior to marking it active and emitting the MI_FLIP.
On the read side, the mb should be enforced by the spinlocks.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
[danvet: Review the barriers a bit, we need a write barrier both
before and after updating ->pending. Similarly we need a read barrier
in the interrupt handler both before and after reading ->pending. With
well-ordered irqs only one barrier in each place should be required,
but since this patch explicitly sets out to combat spurious interrupts
with is staged activation of the unpin work we need to go full-bore on
the barriers, too. Discussed with Chris Wilson on irc and changes
acked by him.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

When l3 parity support for Haswell was enabled in

commit f27b9265
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Tue Jul 24 20:47:32 2012 -0700

    drm/i915: Expand DPF support to Haswell

no one noticed that the patch which introduced this macro

commit e1ef7cc2


Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Tue Jul 24 20:47:31 2012 -0700

    drm/i915: Macro to determine DPF support

missed one spot. Fix this.

Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57441
Reviewed-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

In a couple of places we attempt to adjust the existing watermark
registers to update them for the new cursor watermarks. This goes
horribly wrong as instead of clearing the cursor bits prior to or'ing in
the new values, we clear the rest of the register with the result that
the watermark registers contain bogus values.

References: https://bugs.freedesktop.org/show_bug.cgi?id=47034
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

The BLC_PWM_CTL2 register does not exist before gen4. While at it, do a
slight drive by cleanup of the code.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

From BSpec:
"If the Ring Buffer Head Pointer and the Tail Pointer are on the same
cacheline, the Head Pointer must not be greater than the Tail
Pointer."

The easiest way to enforce this is to reduce the reported ring space.

References:
Gen2 BSpec "1. Programming Environment" / 1.4.4.6 "Ring Buffer Use"
Gen3 BSpec "vol1c Memory Interface Functions" / 2.3.4.5 "Ring Buffer Use"
Gen4+ BSpec "vol1c Memory Interface and Command Stream" / 5.3.4.5 "Ring Buffer Use"

v2: Include the exact BSpec references in the description

v3: s/64/I915_RING_FREE_SPACE, and add the BSpec information to the code

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

As we may actually allocate in order to save the physical swizzling bits
during the free, we have to be careful not to trigger the shrinker on
the same object.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Added a small comment in the code to really drive the
scariness of this patch home.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

i915_gem_handle_seqno_wrap() will zero all sync_seqnos but as the
wrap can happen inside ring->sync_to(), pre wrap seqno was
carried over and overwrote the zeroed sync_seqno.

When wrap is handled, all outstanding requests will be retired and
objects moved to inactive queue, causing their last_read_seqno to be zero.
Use this to update the sync_seqno correctly.

RING_SYNC registers after wrap will contain pre wrap values which
are >= seqno. So injecting the semaphore wait into ring completes
immediately.

Original idea for using last_read_seqno from Chris Wilson.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Should be useful to know what the driver thought the other ring's seqno
was when it last used a semaphore.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Replace the wait for the ring to be clear with the more common wait for
the ring to be idle. The principle advantage is one less exported
intel_ring_wait function, and the removal of a hardcoded value.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

As we now always preallocate the seqno before writing to the ring, we
can trivially test if we have any pending activity on the ring by
inspecting the olr. This makes it then possible to flush operations that
are not normally associated with a request, like power-management.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Based on the work by Mika Kuoppala, we realised that we need to handle
seqno wraparound prior to committing our changes to the ring. The most
obvious point then is to grab the seqno inside intel_ring_begin(), and
then to reuse that seqno for all ring operations until the next request.
As intel_ring_begin() can fail, the callers must already be prepared to
handle such failure and so we can safely add further checks.

This patch looks like it should be split up into the interface
changes and the tweaks to move seqno wrapping from the execbuffer into
the core seqno increment. However, I found no easy way to break it into
incremental steps without introducing further broken behaviour.

v2: Mika found a silly mistake and a subtle error in the existing code;
inside i915_gem_retire_requests() we were resetting the sync_seqno of
the target ring based on the seqno from this ring - which are only
related by the order of their allocation, not retirement. Hence we were
applying the optimisation that the rings were synchronised too early,
fortunately the only real casualty there is the handling of seqno
wrapping.

v3: Do not forget to reset the sync_seqno upon module reinitialisation,
ala resume.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> [v2]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

There seem to be indeed some awkwards machines around, mostly those
without OpRegion support, where the firmware changes the display hw
state behind our backs when closing the lid.

This force-restore logic has been originally introduced in

commit c1c7af60
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Thu Sep 10 15:28:03 2009 -0700

    drm/i915: force mode set at lid open time

but after the modeset-rework we've disabled it in the vain hope that
it's no longer required:

commit 3b7a89fc


Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Mon Sep 17 22:27:21 2012 +0200

    drm/i915: fix OOPS in lid_notify

Alas, no.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54677
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57434
Tested-by: Krzysztof Mazur <krzysiek@podlesie.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

In commit 69c2fc89


Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Jul 20 12:41:03 2012 +0100

    drm/i915: Remove the per-ring write list

the explicit flush was removed from i915_ring_idle(). However, we
continued to wait upon the next seqno which now did not correspond to
any request (except for the unusual condition of a failure to queue a
request after execbuffer) and so would wait indefinitely.

This has an important side-effect that i915_gpu_idle() does not cause
the seqno to be incremented. This is vital if we are to be able to idle
the GPU to handle seqno wraparound, as in subsequent patches.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Dereference dev_priv only after we know it is valid.
Found with smatch.

Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Some devices may respond very slowly and only flag that the reply is
pending within the first 15us response window. Be kind to such devices
and wait a further 15ms, before checking for the pending reply. This
moves the existing special case delay of 30ms down from the detection
routine into the common path and pretends to explain it...

v2: Simplify the loop constructs as suggested by Jani Nikula.

References: https://bugs.freedesktop.org/show_bug.cgi?id=36997
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

We currently set "0" as the VIC value of the AVI InfoFrames. According
to the specs this should be fine and work for every mode, so to my
point of view we can't consider the current behavior as a bug. The
problem is that  we recently received a bug report (Kernel bug #50371)
from a user that has an AV receiver that gives a black screen for any
mode with VIC set to 0.

So in order to make at least some modes work for him, this patch sets
the correct VIC number when sending AVI InfoFrames.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=50371
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This function returns the VIC of the mode. This value can be used when
creating AVI InfoFrames.

Cc: Thierry Reding <thierry.reding@avionic-design.de>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=50371
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Thierry Reding <thierry.reding@avionic-design.de>
Acked-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This bug was introduced by me:
commit e76e9aeb


Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Sun Nov 4 09:21:27 2012 -0800

    drm/i915: Stop using AGP layer for GEN6+

The existing code uses memset_io which follows memset semantics in only
guaranteeing a write of individual bytes. Since a PTE entry is 4 bytes,
this can only be correct if the scratch page address is 0.

This caused unsightly errors when we clear the range at load time,
though I'm not really sure what the heck is referencing that memory
anyway. I caught this is because I believe we have some other bug where
the display is doing reads of memory we feel should be cleared (or we
are relying on scratch pages to be a specific value).

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Since it should be working a little bit better now.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
[danvet: resolve conflict around the call to intel_crtc_mode_get. And
add the missing NULL check Chris spotted while at it.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Use the recorded panel fixed-mode to populate the get_modes() request in
the absence of an EDID.

Fixes regression from
commit 9cd300e0


Author: Jani Nikula <jani.nikula@intel.com>
Date:   Fri Oct 19 14:51:52 2012 +0300

    drm/i915: Move cached EDID to intel_connector

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
[danvet: Drop the retval-changing hunk, as suggested by Jani in his
review and acked by Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Since the base fields in both struct intel_connector and struct
intel_sdvo_connector are at the beginning of the enclosing struct, the
pointers are essentially the same, but there is no requirement or guarantee
that this is always the case. Kfree the enclosing intel_sdvo_connector
pointer that was originally allocated, not the enclosed drm_connector, in
case someone ever rearranges the structs.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

v2: Rebased.

Signed-off-by: Rob Clark <rob@ti.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com> (v1)
[danvet: Pimp commit message a bit.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This was leftover crap from kill-agp. The current code is theoretically
broken for 64b bars. (I resist removing theoretically because I am too
lazy to test).

We still need to ioremap things ourselves because we want to ioremap_wc
the PTEs.

v2: Forgot to kill the tmp variable in v1

CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

If we have hit oom whilst holding our struct_mutex, then currently we
cannot reap our own GPU buffers which likely pin most of memory, making
an outright OOM more likely. So if we are running in direct reclaim and
already hold the mutex, attempt to free buffers knowing that the
original function can not continue until we return.

v2: Add a note explaining that the mutex may be stolen due to
pre-emption, and that is bad.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

As we may invoke the shrinker whilst trying to allocate memory to hold
the gtt_space for this object, we need to be careful not to mark the
drm_mm_node as activated (by assigning it to this object) before we
have finished our sequence of allocations.

Note: We also need to move the binding of the object into the actual
pagetables down a bit. The best way seems to be to move it out into
the callsites.

Reported-by: Imre Deak <imre.deak@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Added small note to commit message to summarize review
discussion.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

As the SDVO/HDMI registers are multiplex, it is safe to assume that the
w/a required for HDMI on IbexPoint, namely that the SDVO register cannot
both be disabled and have selected transcoder B, is also required for
SDVO. At least the modeset state checker detects that the transcoder
selection is left in the undefined state, and so it appears sensible to
apply the w/a:

[ 1814.480052] WARNING: at drivers/gpu/drm/i915/intel_display.c:1487 assert_pch_hdmi_disabled+0xad/0xb5()
[ 1814.480053] Hardware name: Libretto W100
[ 1814.480054] IBX PCH hdmi port still using transcoder B

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=57066
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Also document the WA name for the previous gens that implement it.

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

For now, this code is just used by the eDP AUX channel frequency.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This goes on a separate patch since it won't apply on the stable
trees and there's nothing using panel fitter on HSW on the older
Kernels.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

I actually found this problem on Haswell, but then discovered Ivy
Bridge also has it by reading the spec.

I don't have the hardware to test this.

Cc: stable@vger.kernel.org
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

DDI A and E have 4 lanes to share, so if DDI A is using 4 lanes,
there's nothing left for DDI E, which means there's no CRT port on the
machine.

The bit we're checking here is programmed at system boot and it cannot
be changed afterwards, so we cannot change the amount of lanes
reserved for each DDI port.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

We need to enable a special bit, otherwise none of the DP functions
requiring the PCH will work.

Version 2: store the PCH ID inside dev_priv, as suggested by Daniel
Vetter.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

We don't check if the "unclaimed register" bit is set before we call
writel, so if it was already set before, we might print a misleading
message about "unclaimed write" on the wrong register.

This patch makes us check the unclaimed bit before the writel, so we
can print a new "Unknown unclaimed register before writing to %x"
message.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This function runs on Haswell, so set the correct pch_transcoder and
cpu_transcoder variables. This fixes an assertion failure on Haswell
VGA.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This is a full revert of 59c859d6

:
    drm/i915: account for only one PCH receiver on Haswell

Now that the PCH code is fixed to be able use the only PCH transcoder
independently of the pipe and CPU transcoder, we can revert this.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Resolve conflict due to the rebasing of dinq on top of
drm-next.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

If we accumulate unpin tasks because we are pageflipping faster than the
system can schedule its workers, we can effectively create a
pin-leak. The solution taken here is to limit the number of unpin tasks
we have per-crtc and to flush those outstanding tasks if we accumulate
too many. This should prevent any jitter in the normal case, and also
prevent the hang if we should run too fast.

Note: It is important that we switch from the system workqueue to our
own dev_priv->wq since all work items on that queue are guaranteed to
only need the dev->struct_mutex and not any modeset resources. For
otherwise if we have a work item ahead in the queue which needs the
modeset lock (like the output detect work used by both polling or
hpd), this work and so the unpin work will never execute since the
pageflip code already holds that lock. Unfortunately there's no
lockdep support for this scenario in the workqueue code.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=46991
Reported-and-tested-by: Tvrtko Ursulin <tvrtko.ursulin@onelan.co.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Added note about workqueu deadlock.]
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=56337
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

But disabled by default. This essentially reverts

commit bcd5023c
Author: Dave Airlie <airlied@redhat.com>
Date:   Mon Mar 14 14:17:55 2011 +1000

    drm/i915: disable opregion lid detection for now

but leaves the autodetect mode disabled. There's also the explicit lid
status option added in

commit fca87409


Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Feb 17 13:44:48 2011 +0000

    drm/i915: Add a module parameter to ignore lid status

Which overloaded the meaning for the panel_ignore_lid parameter even
more. To fix up this mess, give the non-negative numbers 0,1 the
original meaning back and use negative numbers to force a given state.
So now we have

1  - disable autodetect, return unknown
0  - enable autodetect
-1 - force to disconnected/lid closed
-2 - force to connected/lid open

v2: My C programmer license has been revoked ...

v3: Beautify the code a bit, as suggested by Chris Wilson.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=27622
Tested-by: Andreas Sturmlechner <andreas.sturmlechner@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

This patch adds the missing code to send ELD for Haswell DisplayPort,
based on Xingchao's original patch.

A test was performed with HSW-D machine and NEC EA232Wmi DP monitor.

Cc: Xingchao Wang <xingchao.wang@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>