By recording the location of every request in the ringbuffer, we know
that in order to retire the request the GPU must have finished reading
it and so the GPU head is now beyond the tail of the request. We can
therefore provide a conservative estimate of where the GPU is reading
from in order to avoid having to read back the ring buffer registers
when polling for space upon starting a new write into the ringbuffer.

A secondary effect is that this allows us to convert
intel_ring_buffer_wait() to use i915_wait_request() and so consolidate
upon the single function to handle the complicated task of waiting upon
the GPU. A necessary precaution is that we need to make that wait
uninterruptible to match the existing conditions as all the callers of
intel_ring_begin() have not been audited to handle ERESTARTSYS
correctly.

By using a conservative estimate for the head, and always processing all
outstanding requests first, we prevent a race condition between using
the estimate and direct reads of I915_RING_HEAD which could result in
the value of the head going backwards, and the tail overflowing once
again. We are also careful to mark any request that we skip over in
order to free space in ring as consumed which provides a
self-consistency check.

Given sufficient abuse, such as a set of unthrottled GPU bound
cairo-traces, avoiding the use of I915_RING_HEAD gives a 10-20% boost on
Sandy Bridge (i5-2520m):
  firefox-paintball  18927ms -> 15646ms: 1.21x speedup
  firefox-fishtank   12563ms -> 11278ms: 1.11x speedup
which is a mild consolation for the performance those traces achieved from
exploiting the buggy autoreported head.

v2: Add a few more comments and make request->tail a conservative
estimate as suggested by Daniel Vetter.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
[danvet: resolve conflicts with retirement defering and the lack of
the autoreport head removal (that will go in through -fixes).]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This was pointed by Jesse Barnes. The code now seems to follow the
specification but I don't have an SDVO device to really test this.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

GMBUS has several ports and each has it's own corresponding
I2C adpater. When multiple I2C adapters call gmbus_xfer() at
the same time there is a race condition in using the underlying
GMBUS controller. Fixing this by adding a mutex lock when calling
gmbus_xfer().

v2: Moved gmbus_mutex below intel_gmbus and added comments.
Rebased to drm-intel-next-queued.
Signed-off-by: NYufeng Shen <miletus@chromium.org>
[danvet: Shortened the gmbus_mutex comment a bit and add the patch
revision comment to the commit message.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

When HDMI-DVI converter is used, it's not only necessary to turn off
audio, but also to disable HDMI_MODE_SELECT and video infoframe. Since
the DVI mode is mainly tied to audio functionality from end user POV,
add a new "force-dvi" audio mode:

	xrandr --output HDMI1 --set audio force-dvi

Note that most users won't need to set this and happily rely on the EDID
based DVI auto detection.
Reported-by: NAndrea Arcangeli <aarcange@redhat.com>
Signed-off-by: NWu Fengguang <fengguang.wu@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This patch replaces the locking from the downclock routines with an assert
to ensure the registers are indeed unlocked. Without this patch, pre-SNB
devices would lock the registers when downclocking which would cause a
WARNING on suspend/resume with downclocking enabled.

Note: To hit this bug, you need to have lvds downclocking enabled.
Signed-off-by: NSean Paul <seanpaul@chromium.org>
Acked-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The locking in our setup and teardown paths is rather arbitrary, but
generally we try to protect gem stuff with dev->struct_mutex. Further,
the ums/gem ioctl to setup gem _does_ take the look. So fix up this
benign inconsistency.

Notice while reading through code.

v2: Rebased on top of the ppgtt code.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We still have reports of missed irqs even on Sandybridge with the
HWSTAM workaround in place. Testing by the bug reporter gets rid of
them with the forcewake voodoo and no HWSTAM writes.

Because I've slightly botched the rebasing I've left out the ACTHD
readback which is also required to get IVB working. Seems to still
work on the tester's machine, so I think we should go with the more
minmal approach on SNB. Especially since I've only found weak evidence
for holding forcewake while waiting for an interrupt to arrive, but
none for the ACTHD readback.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45181
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45332
Tested-by: Nicolas Kalkhof nkalkhof()at()web.de
Acked-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Currently we reserve seqnos only when we emit the request to the ring
(by bumping dev_priv->next_seqno), but start using it much earlier for
ring->oustanding_lazy_request. When 2 threads compete for the gpu and
run on two different rings (e.g. ddx on blitter vs. compositor)
hilarity ensued, especially when we get constantly interrupted while
reserving buffers.

Breakage seems to have been introduced in

commit 6f392d54
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sat Aug 7 11:01:22 2010 +0100

    drm/i915: Use a common seqno for all rings.

This patch fixes up the seqno reservation logic by moving it into
i915_gem_next_request_seqno. The ring->add_request functions now
superflously still return the new seqno through a pointer, that will
be refactored in the next patch.

Note that with this change we now unconditionally allocate a seqno,
even when ->add_request might fail because the rings are full and the
gpu died. But this does not open up a new can of worms because we can
already leave behind an outstanding_request_seqno if e.g. the caller
gets interrupted with a signal while stalling for the gpu in the
eviciton paths. And with the bugfix we only ever have one seqno
allocated per ring (and only that ring), so there are no ordering
issues with multiple outstanding seqnos on the same ring.

v2: Keep i915_gem_get_seqno (but move it to i915_gem.c) to make it
clear that we only have one seqno counter for all rings. Suggested by
Chris Wilson.

v3: As suggested by Chris Wilson use i915_gem_next_request_seqno
instead of ring->oustanding_lazy_request to make the follow-up
refactoring more clearly correct. Also improve the commit message
with issues discussed on irc.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45181
Tested-by: Nicolas Kalkhof nkalkhof()at()web.de
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

So don't assign it false, that's just confusing ... No functional
change here.
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

If we don't have a sufficient number of free entries in the FIFO, we
proceed to do a write anyway. With this check we should have a clue if
that write actually failed or not.

After some discussion with Daniel Vetter regarding his original
complaint, we agreed upon this.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This is similar to a patch I wrote several months ago. It's been updated
for the new FORCEWAKE_MT. As recommended by Chris Wilson, use WARN()
instead of DRM_ERROR, so we can get a backtrace.

This shouldn't impact performance too much as the extra register read
can replace the POSTING_READ we had previously.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Add register definitions for GTFIFODBG, and clear it during init time to
make sure state is correct.

This register tells us if either a read, or a write occurred while the
fifo was full. It seems like bit 2 is an OR of bit 0 and bit 1, so we
check that as well, but the documents are not quite clear.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by (v1): Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I'm not sure why they are needed (I didn't notice any difference in my
tests), but these bits are in our documentation and they are also set by
the Windows driver.
Signed-off-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The drm core _really_ likes to frob around with the crtc timings and
put halfed vertical timings (in fields) in there. Which confuses the
overlay code, resulting in it's refusal to display anything at the
lower half of an interlaced pipe.
Tested-by: NChristopher Egert <cme3000@gmail.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NPeter Ross <pross@xvid.org>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NChristopher Egert <cme3000@gmail.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NPeter Ross <pross@xvid.org>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NChristopher Egert <cme3000@gmail.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The hw seems to use this to correctly insert the required delay
before/after an even/odd interlaced field. This might also explain
why we need to substract 1 half-line from vtotal - if the hw just
adds the delay programmend in VSYNCSHIFT the total frame time would be
about that too long.

These registers seems to only exist on gen4 and later. For paranoia
also program it to 0 for progressive modes, but according to
documentation the hw should just ignore it in this case.
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

gen2 doesn't support it, so be a bit more paranoid and add a check to
ensure that we never ever set an unsupported interlaced bit.

Ensure that userspace can't set an interlaced mode by resetting
interlace_allowed for the crt on gen2. dvo and lvds are the only other
encoders that gen2 supports and these already disallow interlaced
modes.
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

According to Paulo Zanoni, this is what windows does.
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

According to bspec, we need to subtract an additional line from vtotal
for interlaced modes and vblank_end needs to equal vtotal. All other
timing fields do not need this special treatment, so kill it.

Bspec says that this is irrespective of whether the interlaced mode
has an odd or even vtotal, both modes are supported.
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NChristopher Egert <cme3000@gmail.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have a pretty decent confusion about vertical timings of interlaced
modes. Peter Ross has written a patch that makes interlace modes work
on a lot more platforms/output combinations by doubling the vertical
timings.

The issue with that patch is that core drm _does_ support specifying
whether we want these vertical timings in fields or frames, we just
haven't managed to consistently use this facility. The relavant
function is drm_mode_set_crtcinfo, which fills in the crtc timing
information.

The first thing to note is that the drm core keeps interlaced modes in
frames, but displays modelines in fields. So when the crtc modeset
helper copies over the mode into adjusted_mode it will already contain
vertical timings in half-frames. The result is that the fixup code in
intel_crtc_mode_fixup doesn't actually do anything (in most cases at
least).

Now gen3+ natively supports interlaced modes and wants the vertical
timings in frames. Which is what sdvo already fixes up, at least under
some conditions.

There are a few other place that demand vertical timings in fields
but never actually deal with interlaced modes, so use frame timings
for consistency, too. These are:
- lvds panel,
- dvo encoders - dvo is the only way gen2 could support interlaced
  mode, but currently we don't support any encoders that do.
- tv out - despite that the tv dac sends out an interlaced signal it
  expects a progressive mode pipe configuration.
All these encoders enforce progressive modes by resetting
interlace_allowed.

Hence we always want crtc vertical timings in frames. Enforce this in
our crtc mode_fixup function and rip out any redudant timing
computations from the encoders' mode_fixup function.

v2-4: Adjust the vertical timings a bit.

v5: Split out the 'subtract-one for interlaced' fixes.

v6: Clarify issues around tv-out and gen2.
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NChristopher Egert <cme3000@gmail.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

- Clarify which bits are for which chips.
- Note that gen2 can't do interlaced directly (only via dvo tv chips).
- Move the mask to the top to make it clearer how wide this field is.
- Add defintions for all possible values.

This patch doesn't change any code.

v2: Paulo Zanoni pointed out that the pixel doubling modes do no
longer exist on ivb.

Cc: Peter Ross <pross@xvid.org>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: NChristopher Egert <cme3000@gmail.com>
Tested-by: NAlfonso Fiore <alfonso.fiore@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We want to unconditionally enable ppgtt for two reasons:
- Windows uses this on snb and later.
- We need the basic hw support to work before we can think about real
  per-process address spaces and other cool features we want.

But Chris Wilson was complaining all over irc and intel-gfx that this
will blow up if we don't have a module option to disable it. Hence add
one, to prevent this.

ppgtt support seems to slightly change the timings and make crashy
things slightly more or less crashy. Now in my testing and the testing
this got on troublesome snb machines, it seems to have improved things
only. But on ivb it makes quite a few crashes happen much more often,
see

https://bugs.freedesktop.org/show_bug.cgi?id=41353

Luckily Eugeni Dodonov seems to have a set of workarounds that fix
this issue.

v2: Don't try to enable ppgtt on pre-snb.

v3: Pimp commit message and make Chris Wilson less grumpy by adding a
module option.

v4: New try at making Chris Wilson happy.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Acked-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This was pretty usefull for debugging, might be useful for diagnosing
issues.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Split out for easier cross-checking of the boring pieces with bspec.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This adds support to bind/unbind objects and wires it up. Objects are
only put into the ppgtt when necessary, i.e. at execbuf time.

Objects are still unconditionally put into the global gtt.

v2: Kill the quick hack and explicitly pass cache_level to ppgtt_bind
like for the global gtt function. Noticed by Chris Wilson.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This just adds the setup and teardown code for the ppgtt PDE and the
last-level pagetables, which are fixed for the entire lifetime, at
least for the moment.

v2: Kill the stray debug printk noted by and improve the pte
definitions as suggested by Chris Wilson.

v3: Clean up the aperture stealing code as noted by Ben Widawsky.

v4: Paint the init code in a more pleasing colour as suggest by Chris
Wilson.

v5: Explain the magic numbers noticed by Ben Widawsky.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Tested-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Chris Wilson and me have again stared at funny error states and it's
been pretty clear from the start that something was seriously amiss.
The seqnos last seen by the cpu were a few hundred behind those that
the gpu could have possibly emitted last before it died ...

Chris now tracked it down (hopefully, definit verdict's still out),
but in hindsight we'd have found the bug by simply dumping the cpu
side tracking of the ring head and tail registers.

Fix this and prevent an identical time-waster in the future.

Because the hangs always involved semaphores in one way or another,
we've tried to dump the mbox registers, but couldn't find any
inconsistencies. Still, dump them too.
Reviewed-and-wanted-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

These are all user-trigerable, so tune down their loudness a notch.
For some of these we have i-g-t tests (because they prevent
newly-discovered bugs), without this patches running the test suite
leaves behind a dirty dmesg.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

On gen5 we also need to correctly set up swizzling in the display
scanout engine, but only there. Consolidate this into the same
function.

This has a small effect on ums setups - the kernel now also sets this
bit in addition to userspace setting it. Given that this code only
runs when userspace either can't (resume, gpu reset) or explicitly
won't(gem_init) touch the hw this shouldn't have an adverse effect.
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have to do this manually. Somebody had a Great Idea.

I've measured speed-ups just a few percent above the noise level
(below 5% for the best case), but no slowdows. Chris Wilson measured
quite a bit more (10-20% above the usual snb variance) on a more
recent and better tuned version of sna, but also recorded a few
slow-downs on benchmarks know for uglier amounts of snb-induced
variance.

v2: Incorporate Ben Widawsky's preliminary review comments and
elaborate a bit about the performance impact in the changelog.

v3: Add a comment as to why we don't need to check the 3rd memory
channel.

v4: Fixup whitespace.
Acked-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NEric Anholt <eric@anholt.net>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

An identical patch has been merged for i9xx_crtc_mode_set:

Commit 59df7b17
Author: Christian Schmidt <schmidt@digadd.de>
Date:   Mon Dec 19 20:03:33 2011 +0100

    drm/intel: Fix initialization if startup happens in interlaced mode [v2]

But that one neglected to fix up the ironlake+ path.

This should fix the issue reported by Alfonso Fiore where booting with
only a HDMI cable connected to his TV failed to display anything. The
issue is that the bios set up things for 1080i and used the pannel
fitter to scale up the lower progressive resolutions. We failed to
clear the interlace bit in the PIPEACONF register, resulting in havoc.

v2: Be more paranoid and just unconditionally clear the field before
setting new values.

Cc: Peter Ross <pross@xvid.org>
Cc: Alfonso Fiore <alfonso.fiore@gmail.com>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NKeith Packard <keithp@keithp.com>

According to a bug report, it doesn't have one.

Cc: stable@kernel.org
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=44263Acked-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NKeith Packard <keithp@keithp.com>

It is never correct to use intel_crtc->bpp in intel_dp_link_required,
so instead pass an explicit bpp in to this function. This patch
only supports 18bpp and 24bpp modes, which means that 10bpc modes will
be computed incorrectly. Fixing that will require more extensive
changes, and so must be addressed separately from this bugfix.

intel_dp_link_required is called from intel_dp_mode_valid and
intel_dp_mode_fixup.

* intel_dp_mode_valid is called to list supported modes; in this case,
  the current crtc values cannot be relevant as the modes in question
  may never be selected. Thus, using intel_crtc->bpp is never right.

* intel_dp_mode_fixup is called during mode setting, but it is run
  well before ironlake_crtc_mode_set is called to set intel_crtc->bpp,
  so using intel_crtc-bpp in this path can only ever get a stale
  value.

Cc: Lubos Kolouch <lubos.kolouch@gmail.com>
Cc: Adam Jackson <ajax@redhat.com>
Cc: stable@vger.kernel.org
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=42263
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=44881Tested-by: NDave Airlie <airlied@redhat.com>
Tested-by: camalot@picnicpark.org (Dell Latitude 6510)
Tested-by: NRoland Dreier <roland@digitalvampire.org>
Signed-off-by: NKeith Packard <keithp@keithp.com>

As the buffer is not necessarily accessible through the GTT at the time
of a GPU hang, and capturing some of its contents is far more valuable
than skipping it, provide a clflushed fallback read path. We still
prefer to read through the GTT as that is more consistent with the GPU
access of the same buffer. So example it will demonstrate any errorneous
tiling or swizzling of the command buffer as seen by the GPU.

This becomes necessary with use of CPU relocations and lazy GTT binding,
but could potentially happen anyway as a result of a pathological error.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NEugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Like for shmem_pwrite_slow. The only difference is that because we
read data, we can leave the fetched cachelines in the cpu: In the case
that the object isn't in the cpu read domain anymore, the clflush for
the next cpu read domain invalidation will simply drop these
cachelines.

slow_shmem_bit17_copy is now ununsed, so kill it.

With this patch tests/gem_mmap_gtt now actually works.

v2: add __ to copy_to_user_swizzled as suggested by Chris Wilson.

v3: Fixup the swizzling logic, it swizzled the wrong pages.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=38115Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

... instead of get_user_pages, because that fails on non page-backed
user addresses like e.g. a gtt mapping of a bo.

To get there essentially copy the vfs read path into pagecache. We
can't call that right away because we have to take care of bit17
swizzling. To not deadlock with our own pagefault handler we need
to completely drop struct_mutex, reducing the atomicty-guarantees
of our userspace abi. Implications for racing with other gem ioctl:

- execbuf, pwrite, pread: Due to -EFAULT fallback to slow paths there's
  already the risk of the pwrite call not being atomic, no degration.
- read/write access to mmaps: already fully racy, no degration.
- set_tiling: Calling set_tiling while reading/writing is already
  pretty much undefined, now it just got a bit worse. set_tiling is
  only called by libdrm on unused/new bos, so no problem.
- set_domain: When changing to the gtt domain while copying (without any
  read/write access, e.g. for synchronization), we might leave unflushed
  data in the cpu caches. The clflush_object at the end of pwrite_slow
  takes care of this problem.
- truncating of purgeable objects: the shmem_read_mapping_page call could
  reinstate backing storage for truncated objects. The check at the end
  of pwrite_slow takes care of this.

v2:
- add missing intel_gtt_chipset_flush
- add __ to copy_from_user_swizzled as suggest by Chris Wilson.

v3: Fixup bit17 swizzling, it swizzled the wrong pages.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The gtt_pwrite slowpath grabs the userspace memory with
get_user_pages. This will not work for non-page backed memory, like a
gtt mmapped gem object. Hence fall throuh to the shmem paths if we hit
-EFAULT in the gtt paths.

Now the shmem paths have exactly the same problem, but this way we
only need to rearrange the code in one write path.

v2: v1 accidentaly falls back to shmem pwrite for phys objects. Fixed.

v3: Make the codeflow around phys_pwrite cleara as suggested by Chris
Wilson.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This will also come handy for the gen6+ swizzling support, where the
driver is supposed to control swizzling depending upon dram
configuration.

v2: CxDRB3 are 16 bit regs! Noticed by Chris Wilson.
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>