We already set this limit for the GGTT.

This is a temporary patch until a full replacement of size_t variables
(inadequate in 32-bit kernel) is in place.

Regression from:
	commit a4e0bedc
	Author: Michel Thierry <michel.thierry@intel.com>
	Date:   Wed Apr 8 12:13:35 2015 +0100

		drm/i915: Use complete address space in true PPGTT

v2: Prettify code and explain why this is needed. (Chris)
v3: Don't hide the compilation warning in 32-bit. (Chris)
Suggested-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NMichel Thierry <michel.thierry@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

With unified modeset and flip paths introduced recently when switching
to fbcon PSR was being disabled on fb_set_par path but re-enabled on
fb_pan_display one, causing missed screen updates and un unusable
console.

Regression introduced with:

commit bb546623
Author: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Date:   Tue Apr 21 17:13:13 2015 +0300

    drm/i915: Unify modeset and flip paths of intel_crtc_set_config()

Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Without this frontbuffer flip when enabling planes PSR got compromised
and wasn't being enabled waiting forever on the flush that never
arrived.

Another solution would to create a enable_cursor function and split this
frontbuffer flip among the different plane enable and disable functions.
But if necessary this can be done in a follow up work. For now let's
just fix the regression.

It was removed by:

commit 87d4300a
Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Date:   Tue Apr 21 17:12:54 2015 +0300

    drm/i915: Move intel_(pre_disable/post_enable)_primary to intel_display.c, and use it there.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This commit is the "sink CRC" version of:

commit 8c740dce
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
Date:   Fri Oct 17 18:42:03 2014 -0300
    drm/i915: disable IPS while getting the pipe CRCs.

For some unknown reason, when IPS gets enabled, the sink CRC changes.
Since hsw_enable_ips() doesn't really guarantee to enable IPS (it
depends on package C-states), we can't really predict if IPS is
enabled or disabled while running our CRC tests, so let's just
completely disable IPS while sink CRCs are being used.

If we find a way to make IPS not change the pipe CRC result, we may
want to fix IPS and then revert this patch (and 8c740dce too). While
this doesn't happen, let's merge this patch, so the IGT tests relying
on sink CRCs can work properly.

This was discovered while developing a new IGT test, which will
probably be called kms_frontbuffer_tracking.

Testcase: igt/kms_frontbuffer_tracking (not on upstream IGT yet)
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

It's totally broken, and since

commit d328c9d7
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Fri Apr 10 16:22:37 2015 +0200

    drm/i915: Select starting pipe bpp irrespective or the primary plane

the kernel will try to use it even for the common rgb888 framebuffers.
Ville has patches to fix it all up properly, but unfortunately they're
stuck in review limbo. And since the 4.2 feature cutoff has passed we
need to somehow  handle this regression.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Paulo Zanoni <przanoni@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

chv_enable_pll() doesn't need to hold sb_lock for the entire duration of
the function. Drop the lock as soon as possible.

valleyview_set_cdclk() does a potential lock+unlock+lock+unlock cycle
with sb_lock. Grab the lock a few lines earlier so we can make do
with a single lock+unlock cycle always.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Rename dpio_lock to sb_lock to inform the reader that its primary
purpose is to protect the sideband mailbox rather than some DPIO
state.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The primary plane frobbing was removed from the sprite code in
 commit ecce87ea
 Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
 Date:   Tue Apr 21 17:12:50 2015 +0300

    drm/i915: Remove implicitly disabling primary plane for now

but the intel_flush_primary_plane() calls were left behind. Replace them
with straight forward POSTING_READ() of the sprite surface address
register.

The other user of intel_flush_primary_plane() is g4x_disable_trickle_feed()
where we can just inline the steps directly.

This allows intel_flush_primary_plane() to be killed off.

Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Expecting CHV power wells to be just an extended versions of the VLV
power wells, a bunch of commented out power wells were added in
anticipation when Punit folks would implement it all. Turns out they
never did, and instead CHV has fewer power wells than VLV. Rip out all
the #if 0'ed junk that's not needed.

v2: Rename the "pipe-a" well to "display" to match VLV
    Clarify the pipe A power well relationship to pipes B and C (Deepak)
Reviewed-by: NDeepak S <deepak.s@linux.intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Not sure which LDO programming sequence delay should be used for the CHV
PHY, but the spec says that 600ns is "Used by default for initial
bringup", and the BIOS seems to use that, so let's do the same.
Reviewed-by: NDeepak S <deepak.s@linux.intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

commit 53292cdb ("drm/i915: Workaround
to avoid lite restore with HEAD==TAIL") added a check for req0 != null
which is unnecessary.

The only way req0 could be null is if the list was empty, and this is
already addressed at the beginning of execlists_context_unqueue().
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NMichel Thierry <michel.thierry@intel.com>
Reviewed-by: NJani Nikula <jani.nikula@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In commit 1854d5ca
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Apr 7 16:20:32 2015 +0100

    drm/i915: Deminish contribution of wait-boosting from clients

we removed an atomic timer based check for allowing waitboosting and
moved it below the mutex taken during RPS. However, that mutex can be
held for long periods of time on Vallyview/Cherryview as communication
with the PCU is slow. As clients may frequently wait for results (e.g.
such as tranform feedback) we introduced contention between the client
and the RPS worker. We can take advantage of the RPS worker, by
switching the wait boost decision to use spin locks and defer the
actual reclocking to the worker.

Fixes a regression of up to 45% on Baytrail and Baswell!

v2 (Daniel):
- Use max_freq_softlimit instead of the not-yet-merged boost
  frequency.
- Don't inject a fake irq into the boost work, instead treat
  client_boost as just another legit waker.

v3: Drop the now unused mask (Chris).

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90112
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

It was reported that this comment was confusing, and indeed it is.

v2: (one year later!) Add the range for the DRM_I915_* iotcl defines
    (Daniel)
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This reverts commit 118182e9.

It's causing too much trouble when compile-testing for non-i915 folks.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jani Nikula <jani.nikula@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@intel.com>

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

As Daniel commented on

commit b7ffe1362c5f468b853223acc9268804aa92afc8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Apr 27 13:41:24 2015 +0100

    drm/i915: Free RPS boosts for all laggards

it is better to be explicit when sharing hardcoded values such as
throttle/boost timeouts. Make it so!
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

After allocating from the slab cache, we then need to free the request
back into the slab cache upon error (and not call kfree as that leads
to eventual memory corruption).

Fixes regression from
commit efab6d8d
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Apr 7 16:20:57 2015 +0100

    drm/i915: Use a separate slab for requests
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There is a mplayer video failure reported with xv.
This is because there is a request to do both plane scaling
and colorkey. Because skl hw doesn't support plane scaling
and colorkey at the same time, request is failed which is expected
behavior.

To make xv operate, this patch allows colorkey continue to work
without using scaler. Then behavior would be similar to platforms
without plane scaler support.
Signed-off-by: NChandra Konduru <chandra.konduru@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90449
[danvet: change can_scale to bool as requested by Ville.]
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

GTT caching was disabled by default on gen8 due to not working with
big pages. Some information suggests that it got fixed, but still
GTT caching has been left disabled by default. Or could be it just
meant that the default was changed to off, and hence the problem
got solved.

Enable GTT caching in the hopes of some performance increase.
Whether or not the big pages issue has been fixed is irrelevant
at this stage since we don't use big pages.

This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried
BDW, but supposedly it has larger TLBs so might not benefit as much.
On HSW GTT caching is enabled by default.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

GEN8_L3SQCREG1 isn't saved in the context (verified by going through
a context dump), and so we shouldn't be using the ring w/a code to
initialize it. Also Bspec explicitly talks about MMIO and writing it
with the CPU.

Additionally there's another w/a WaTempDisableDOPClkGating:bdw which
tells us to disable DOP clock gating around the GEN8_L3SQCREG1 write
to make sure everyone notices the change. So let's do that as well.

Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We're not using ilk_init_lp_watermarks() on BDW for some reason.
Probably due to the BDW patches and the relevant WM patches landing
roughlly at the same time. Fix it up.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Bspec says we should disable the FDI RX/TX before disabling the PCH
ports. Do so.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Follow the BSpec sequence for the CRT port as well on PCH platforms,
ie. disable the pipe before the port.

Didn't bother looking at DDI in detail yet, so leave that one be even
though the CRT is a PCH port there.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

While at it also remove the redundant/unneeded w/a like done for hdmi
already.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Mention that this also removes the unneeded w/a, as suggested
by Jesse.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

BSpec says we should disable all ports after the pipe on PCH
platforms. Do so. Fixes a pipe off timeout on ILK now caused by
the transcoder B workaround.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Currently the IBX transcoder B workarounds are not working correctly.
Well, the HDMI one seems to be working somewhat, but the DP one is
definitely busted.

After a bit of experimentation it looks like the best way to make this
work is first disable the port on transcoder B, and then re-enable it
transcoder A, and immediately disable it again.

We can also clean up the code by noting that we can't be called without
a valid crtc. And also note that port A on ILK does not need the
workaround, so let's check for that one too.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

On IBX the SDVO/HDMI register write may be masked when enabling the
port, so it may need to written twice. The HDMI code does this, but
the SDVO code does not. Add the workaround to the SDVO code as well.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Currently we're always enabling enhanced framing on CPT even if the sink
doesn't support it. Fix this up by actaully looking at what the sink
tells us.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Define a TRANS_DP_PIPE_TO_PORT() to make the CPT DP .get_hw_state()
pipe readout neater.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

intel_dp.c is a mess with all the checks for different
platform/PCH variants and ports. Try to clean it up by recognizing
the following facts:
- IVB port A, and CPT port B/C/D are always the special cases
- VLV/CHV don't have port A
- Using the same kind of logic everywhere makes things much easier to
  parse

So let's move the IVB port A and PCH port B/C/D checks to be done first,
and let the other cases fall through, and always check for these things
using the same logic.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

IBX can have problems with the first write to the port register getting
masked when enabling the port. We are trying to apply the workaround
also when disabling the port where it's not needed, and we also try
to apply it for CPT/PPT as well which don't need it. Just kill it.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Resolve conflict with the remove CHV if block.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The IBX 12bpc port enable toggle is only relevant when enabling
the port, not when disabling it. Also this code doesn't actually
toggle anything, and essentially just writes the port register
one extra time. Furthermore CPT/PPT don't need such workarounds
and yet we include them. Just kill it.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We need to re-init the display hardware when going out of suspend. This
includes:

  - Hooking the PCH to the reset logic
  - Restoring CDCDLK
  - Enabling the DDB power

Among those, only the CDCDLK one is a bit tricky. There's some
complexity in that:

  - DPLL0 (which is the source for CDCLK) has two VCOs, each with a set
    of supported frequencies. As eDP also uses DPLL0 for its link rate,
    once DPLL0 is on, we restrict the possible eDP link rates the chosen
    VCO.
  - CDCLK also limits the bandwidth available to push pixels.

So, as a first step, this commit restore what the BIOS set, until I can
do more testing.

In case that's of interest for the reviewer, I've unit tested the
function that derives the decimal frequency field:

  #include <stdio.h>
  #include <stdint.h>
  #include <assert.h>

  #define ARRAY_SIZE(x) (sizeof(x) / sizeof(*(x)))

  static const struct dpll_freq {
          unsigned int freq;
          unsigned int decimal;
  } freqs[] = {
          { .freq = 308570, .decimal = 0b01001100111},
          { .freq = 337500, .decimal = 0b01010100001},
          { .freq = 432000, .decimal = 0b01101011110},
          { .freq = 450000, .decimal = 0b01110000010},
          { .freq = 540000, .decimal = 0b10000110110},
          { .freq = 617140, .decimal = 0b10011010000},
          { .freq = 675000, .decimal = 0b10101000100},
  };

  static void intbits(unsigned int v)
  {
          int i;

          for(i = 10; i >= 0; i--)
                  putchar('0' + ((v >> i) & 1));
  }

  static unsigned int freq_decimal(unsigned int freq /* in kHz */)
  {
          return (freq - 1000) / 500;
  }

  static void test_freq(const struct dpll_freq *entry)
  {
          unsigned int decimal = freq_decimal(entry->freq);

          printf("freq: %d, expected: ", entry->freq);
          intbits(entry->decimal);
          printf(", got: ");
          intbits(decimal);
          putchar('\n');

          assert(decimal == entry->decimal);
  }

  int main(int argc, char **argv)
  {
          int i;

          for (i = 0; i < ARRAY_SIZE(freqs); i++)
                  test_freq(&freqs[i]);

          return 0;
  }

v2:
  - Rebase on top of -nightly
  - Use (freq - 1000) / 500 for the decimal frequency (Ville)
  - Fix setting the enable bit of HSW_NDE_RSTWRN_OPT (Ville)
  - Rename skl_display_{resume,suspend} to skl_{init,uninit}_cdclk to
    be consistent with the BXT code (Ville)
  - Store boot CDCLK in ddi_pll_init (Ville)
  - Merge dev_priv's skl_boot_cdclk into cdclk_freq
  - Use LCPLL_PLL_LOCK instead of (1 << 30) (Ville)
  - Replace various '0' by SKL_DPLL0 to be a bit more explicit that
    we're programming DPLL0
  - Busy poll the PCU before doing the frequency change. It takes about
    3/4 cycles, each separated by 10us, to get the ACK from the CPU
    (Ville)

v3:
  - Restore dev_priv->skl_boot_cdclk, leaving unification with
    dev_priv->cdclk_freq for a later patch (Daniel, Ville)
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If the client stalls on a congested request, chosen to be 20ms old to
match throttling, allow the client a free RPS boost.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
[danvet: s/0/NULL/ reported by 0-day build]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If we have clients stalled waiting for requests, ignore the GPU if it
signals that it should downclock due to low load. This helps prevent
the automatic timeout from causing extremely long running batches from
taking even longer.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that we have internal clients, rather than faking a whole
drm_i915_file_private just for tracking RPS boosts, create a new struct
intel_rps_client and pass it along when waiting.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Since we will often pageflip to an active surface, we will often have to
wait for the surface to be written before issuing the flip. Also we are
likely to wait on that surface in plenty of time before the vblank.
Since we have a mechanism for boosting when a flip misses the expected
vblank, curtain the number of times we RPS boost when simply waiting for
mmioflip.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Ring switches can occur many times per frame, and are often out of
control, causing frequent RPS boosting for no practical benefit. Treat
the sw semaphore synchronisation as a separate client and only allow it
to boost once per busy/idle cycle.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This trims a little overhead from the common case of not needing to
synchronize between rings.

v2: execlists is special and likes to duplicate code.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Currently, we only track the last request globally across all engines.
This prevents us from issuing concurrent read requests on e.g. the RCS
and BCS engines (or more likely the render and media engines). Without
semaphores, we incur costly stalls as we synchronise between rings -
greatly impacting the current performance of Broadwell versus Haswell in
certain workloads (like video decode). With the introduction of
reference counted requests, it is much easier to track the last request
per ring, as well as the last global write request so that we can
optimise inter-engine read read requests (as well as better optimise
certain CPU waits).

v2: Fix inverted readonly condition for nonblocking waits.
v3: Handle non-continguous engine array after waits
v4: Rebase, tidy, rewrite ring list debugging
v5: Use obj->active as a bitfield, it looks cool
v6: Micro-optimise, mostly involving moving code around
v7: Fix retire-requests-upto for execlists (and multiple rq->ringbuf)
v8: Rebase
v9: Refactor i915_gem_object_sync() to allow the compiler to better
optimise it.

Benchmark: igt/gem_read_read_speed
hsw:gt3e (with semaphores):
Before: Time to read-read 1024k:		275.794µs
After:  Time to read-read 1024k:		123.260µs

hsw:gt3e (w/o semaphores):
Before: Time to read-read 1024k:		230.433µs
After:  Time to read-read 1024k:		124.593µs

bdw-u (w/o semaphores):             Before          After
Time to read-read 1x1:            26.274µs       10.350µs
Time to read-read 128x128:        40.097µs       21.366µs
Time to read-read 256x256:        77.087µs       42.608µs
Time to read-read 512x512:       281.999µs      181.155µs
Time to read-read 1024x1024:    1196.141µs     1118.223µs
Time to read-read 2048x2048:    5639.072µs     5225.837µs
Time to read-read 4096x4096:   22401.662µs    21137.067µs
Time to read-read 8192x8192:   89617.735µs    85637.681µs

Testcase: igt/gem_concurrent_blit (read-read and friends)
Cc: Lionel Landwerlin <lionel.g.landwerlin@linux.intel.com>
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> [v8]
[danvet: s/\<rq\>/req/g]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>