sna: Reduce assertion when using asynchronous CPU access

If we are not actually accessing the memory through the pointer, we do
not care if it not currently coherent.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Silence snarky compliers

0 != (void *)0

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Mollify assert to be consistent with the scanout flush

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

test/lowlevel-blt-bench: Exercise masks

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

test/lowlevel-blt-bench: Exercise SHM pixmaps

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel-virtual-output: Remove one redundant cleanup on error

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel-virtual-output: Cleanup singleton on failure

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel-virtual-output: Convert debugging messages to verbosity

Original patch by main.haarp:
"A verbose switch is quite useful for debugging, it shouldn't require
editing and recompilation to gain more useful output. I added it."

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80664
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Ensure CPU bo is synchronized before writing

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Update allocation of CPU bo to avoid creating active buffers

Since we now prefer CPU detiling, exactly when we want active/inactive
buffers is a little more complex - and we also need to take into account
when we want to use the CPU bo as a render target.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Sync CPU bo before writes

Fixes regression from

commit 961139f5878572ebea268a0bbf47caf05af9093f [2.99.912]
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri May 30 09:45:15 2014 +0100

    sna: Use manual detiling for downloads

Reported-by: Harald Judt <h.judt@gmx.at>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80560
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Only preferentially upload through the GTT for large transfers

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Prefer linear if below tile_width

Be stricter in order to allow greater use of CPU bo.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/io: Prefer CPU copies on LLC

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Don't discard damage for SHM pixmaps

We don't really want to rendering into SHM pixmaps except for copying
back due to the strict serialisation requirements.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Check for a mappable GPU bo before migrating damage

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Skip adding damage if it is already contained

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Tidy calling memcpy_from_tiled

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Micro-optimise unswizzling tiling/detiling

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/trapezoids: Handle mono traps just in case

I disabled a few paths and ended up in an assert that mono trapezoids
shouldn't get that far...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

uxa: Update Screen Pixmap width/height first

Since commit dd6db82680b05cde4a47116b7096c054f3837e20 [2.99.912]
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri May 9 20:26:19 2014 +0100

    uxa: Add DRI3 and miSyncShm support

we verify that the attaching bo meets the constraints required for the
Pixmap. However, when updating the ScreenPixmap following a resize, we
did not update the Pixmap size until after we tried to update the bo,
resulting in a validation failure when shrinking the screen.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

test: Expand number of sources for basic benchmarking

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: DBG compile fixes

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/glyphs: Add DBG option for glyph tolerance

And bump it to 3 so that glyph filtering doesn't force us to use the
mask too often.

References: https://bugs.freedesktop.org/show_bug.cgi?id=77436
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/glyphs: Add DBG option for forcing glyphs-to-dst

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Check source bo is suitable for BLT before doing so

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Only reuse pinned batches for 830/845

Trying to fly too close to the sun. :(

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80558
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Slightly improved DBG messages

Missing git add from

commit 2e8c09f3fe468abba9c307ba3d7b2d22908e1172
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 25 22:19:23 2014 +0100

    sna/dri2: Hook into ClientGone callback to clear dangling references

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Tidy selection of a near-miss active buffer.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80560
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix userptr defines to match i915_drm.h

Now that we are starting to see userptr in the headers, we see that the
definitions slightly changed.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Remove scare quotes from hotplug detection "enabled"

Tidy up the log message by improving its formatting, and making it more
accurate.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Hook into ClientGone callback to clear dangling references

As the Window may exist for multiple Clients, we cannot rely on the
destruction of the Window decoupling all outstanding events and
preventing chasing a stale Client pointer.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80157
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Make output destroy more defensive

We may never create the properties as it may be ignored or the
allocation may fail, so we need to be careful when freeing we do not
stumble over invalid pointers.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80355#c28
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Compile fixes for ancient Xorg

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix uninitialisable variable

sna_render.c: In function 'sna_render_picture_downsample':
sna_render.c:822: warning: 'priv' may be used uninitialized in this function

introduced in

commit ded05e8abb248664124d2b86f77c27497a252c4e
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Jun 23 22:15:56 2014 +0100

    sna: Allow scratch pixmap to allocate linear GPU bo

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Improve throttling during bo allocation

By controlling how long we may block during buffer allocation, we can
relax the throttle elsewhere to prevent render lag buildup.

References: https://bugs.freedesktop.org/show_bug.cgi?id=77436
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Only mark throttle as done after success

References: https://bugs.freedesktop.org/show_bug.cgi?id=77436
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Track desired flushes better after starting a new batch

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel: Fix weblinks (again)

Julien pointed out that I fail at checking links.

Reported-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel: Fix weblinks

Matti reported a few outdated links to intellinuxgraphics.org, now
superseded by 01.org.

Reported-by: Matti Hämäläinen <ccr@tnsp.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Tweak scanout flush

We now distinguish gpu_dirty whilst the bo is not flushed within a batch
and needs_flush for when it is dirty and still busy.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Throttle before doing a TearFree buffer replacement

If the client is rendering fast, he may be rendering too fast and
starting to build up a backlog.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: And fix the typo...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Disable userptr for bdw

Something fishy is afoot. But let's kill this particularly worrisome
regression so that we can do a full round of testing.

References: https://bugs.freedesktop.org/show_bug.cgi?id=79053
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Check all connectors for reuse

Do not rely on the MST path being present to indicate a connector that
may be reused.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80355
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Add marketing names for Broadwell

Even the unknown/reserved ones will stay with HD Graphics.

v2: Add missing names to intel.man and README files as well. (Chris)

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>

Add reserved PCD IDs for Broadwell

Updating using kernel commit fb7023e0e248a33cb00d0a9cdce0bcedaa1ad284
Author: Rodrigo Vivi <rodrigo.vivi@intel.com>
Date:   Tue Jun 10 10:09:52 2014 -0700

    drm/i915: BDW: Adding Reserved PCI IDs.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Add yet more DBG messages to MST discovery

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

configure: Display our README before the summary

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel: Add a note about the myriad places we have identifier strings

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel: Wait for our expected device node to appear

If the path we want to open for the KMS device is not yet available,
wait a short while and try again.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80397
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Add some more DBG around connector discovery

References: https://bugs.freedesktop.org/show_bug.cgi?id=80355
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Ensure serial is not 0 for runtime MST discovery

We reserve serial==0 for the initial setup that is run before RandR is
configured. After the server is running, we have to take extra steps to
make the outputs visible to userspace.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix the usual "x = y" typos in creating boxes

Fixes regression from

commit 1d74b2e07d125ad95b9db6c9c032e90faf2bfa60
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Jun 24 08:58:51 2014 +0100

    sna: Decompose self-copy into overlapping/non-overlapping regions

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/glyphs: Clip damage to dst

We only check if the hardware will automatically discard writes outside
of the target pixmap, but we must manually clip the damage before adding
it to our tracker.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
Reported-by: Zdenek Kabelac <zkabelac@redhat.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=70461#c58
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Trim composite region against dst clip extents

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

intel: Wait for the DRM device to load

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80397
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Decompose self-copy into overlapping/non-overlapping regions

We only need to stage the copy for the overlapping portion of the
self-copy, for the rest we can do in a single pass.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Add a dash more asserts

References: https://bugs.freedesktop.org/attachment.cgi?id=101618
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Disable CRTC after MST topology change

In order to disable CRTC after marking outputs as removed after probing
MST topology updates, we must manually mark the CRTCs as no longer
enabled.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Silence compiler warning for shadowed variable

sna_display.c: In function 'sna_mode_redisplay':
sna_display.c:6322:12: warning: declaration of 'fb_id' shadows a global declaration [-Wshadow]
sna_display.c:223:24: warning: shadowed declaration is here [-Wshadow]
  CC     sna_display_fake.lo

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Allow scratch pixmap to allocate linear GPU bo

When allocating a scratch pixmap, we do so in the expectation that
rendering on the GPU is always preferrable, so even allocate a small
linear bo.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Promote to full GPU bo after forced migration away from CPU bo

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Prefer rendering on the CPU to avoid damage migration

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Force creation of GPU bo for drawing if need be

Now that we are more eager to transfer to the GPU bo if we expect
rendering to the CPU bo to be either slow or risk potential corruption,
we can find ourselves in a recursive loop when we also decide not to
create the GPU bo. Prevent that by explicitly creating the GPU bo if
desired.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/gen6+: Initialise src offset before use

Fixes regression from

commit 30348efd57135edee41fccb87133c572b6473aa8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Jun 23 11:14:25 2014 +0100

    sna/gen6+: Avoid adjusting copy coordinates until commited to using them

which assumed that the render state was being zeroed.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix flushing empty batches

A logic inversion in

commit 1909910fdf89216d18703e50728f4604f75d5d66
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Jun 22 20:19:22 2014 +0100

    sna: Inject a batch flush before adding a fresh bo

also applied the injection to when the batch was empty.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
References: https://bugs.freedesktop.org/show_bug.cgi?id=70461#c48
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Handle old busy_ioctl not reporting the active ring

If we have an old kernel, just presume that the bo was last active on the
render ring.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/gen6+: Avoid adjusting copy coordinates until commited to using them

If we need to fallback to the BLT unit, we need to pass it the original
source/dest coordinates and not our transformed render coordinates. So
keep the original values intact until we start emitting the render
operation.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/gen4+: Add some more asserts around recreating vertex buffers

References: https://bugs.freedesktop.org/show_bug.cgi?id=70461#c46
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Allow the CPU bo to be created if GPU rendering is forced

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/gen6+: Encourage the compiler to inline a couple of key functions

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Prefer to render over damage into the GPU bo

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Do not set CAN_CREATE_GPU flag for untiled allocations

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Rename IGNORE_CPU to IGNORE_DAMAGE to better reflect its purpose

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Add DBG breadcrumbs to sna_output_discover()

References: https://bugs.freedesktop.org/show_bug.cgi?id=80355
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/gen4+: Add box emitters for the generic vertex paths

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Inject a batch flush before adding a fresh bo

Fresh bo (those without a reservation already defined, ala
presumed_offset) will cause the kernel to do a full relocation pass. So,
if possible flush the already correct batch in the hope of trimming the
amount of checking the kernel has to perform on this new batch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Add missing string for DBG

References: https://bugs.freedesktop.org/show_bug.cgi?id=80355#c12
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Disable unused CRTC after undock events

If undocking removes output (as the MST topology changes and some
outputs then become unreachable), we may leave dangling CRTC. This
confuses us, so disable any enabled but unconnected CRTC after the
undock event.

Reported-by: Sree Harsha Totakura <freedesktop@h.totakura.in>
References: https://bugs.freedesktop.org/show_bug.cgi?id=80355#c5
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Track CRTC mode changes and discard old flips

If we have a pending flip across a modeset we must be careful to drop
the stale flip_bo after the modeset.

Reported-by: Sree Harsha Totakura <freedesktop@h.totakura.in>
References: https://bugs.freedesktop.org/show_bug.cgi?id=80355#c8
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Clean up CRTC on CloseScreen

We need to do this to stop any stray references escaping through kernel
events now that we are passing the CRTC through the event.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Discard write hint from a couple more move-to-gpu

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Clear the transform flag on disabling the CRTC

In places, we assume that if the flag is set, we must have an active
CRTC. However, we were not clearing when disabling the CRTC.

Reported-by: Sree Harsha Totakura <freedesktop@h.totakura.in>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80325
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Avoid blending with the render engine on snoopable bo

This seems slightly weird, as there is nothing to indicate that this
would generate incorrect results only for snoopable bo. But no many of
flushes make my pnv box happy, so adjust the placement logic to move the
blending operation away from the snoopable target. This is sensible for
reasons of avoiding snoops from the render pipeline anyway.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80253
Reported-and-tested-by: Matti Hämäläinen <ccr@tnsp.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Be more flexible in converting random subsamples to 8x8 tiles

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Mark up const boxes to keep the compiler happy

Pass around const to suppress a few compiler warnings.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Feed more operations into the 8x8 BLT tiler

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Color patterns for BLT are required to be aligned to 256 byte boundaries

This so far appears to be the most restrictive alignment required, so
simply increase the upload alignment to 256 bytes.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=80033
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

test: Increase number of tiled sources

Significantly improve the stress impose upon the tiled BLT operations.
Also start dumping pngs of the failures.

References: https://bugs.freedesktop.org/show_bug.cgi?id=80033
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Don't decouple front buffer tracking for invalid SwapBuffer calls

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Mark upload from SHM segment as read-only

As this may be mapped by the Xserver using a read-only SHM segment, we
are forced to treat it always as read-only. And this being X, that it is
using a SHM segment is opaque to the driver. Fantastic middlelayer.

This was incorrectly removed in
commit e680e54eab6ffa72e5e1eb6cc0e3fe4b235b06a1
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 11 10:48:26 2014 +0100

    sna: Ignore setting read-only for temporary userptr maps

Also let's not forget the ShmPutImage -> CopyArea path.

Reported-by: Christoph Haag <haagch.christoph@googlemail.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=78411
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Mark upload from SHM segment as read-only

As this may be mapped by the Xserver using a read-only SHM segment, we
are forced to treat it always as read-only. And this being X, that it is
using a SHM segment is opaque to the driver. Fantastic middlelayer.

This was incorrectly removed in
commit e680e54eab6ffa72e5e1eb6cc0e3fe4b235b06a1
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 11 10:48:26 2014 +0100

    sna: Ignore setting read-only for temporary userptr maps

Reported-by: Christoph Haag <haagch.christoph@googlemail.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=78411
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Missed fixing y tile offset in last commit

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Use the right pattern origin for tiled 8x8 extraction

References: https://bugs.freedesktop.org/show_bug.cgi?id=80033
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix up small extents for 8x8 tile construction

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Wrap pointer access for 8x8 tile construction

The kernel can quite happily send a SIGBUS whenever we read from an bo
(due to oom), so catch it and handle it.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Always readback shadow damage after replacing the buffer

Instead track what we explicitly want to discard for the next operation,
so that we don't have to copy back the whole buffer if we have to
replace the shadow and intend to overwrite it all.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna/dri2: Replace assertion with code to skip updating the back buffer

After a window resize or redirection, the DRI2BufferPtr become stale but
it may take a while for the Client to notice and reconfigure. Just
skip touching stale backbuffers and let the Client catch up.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

sna: Fix modulus for partial conversion of tiles to 8x8 patterns

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>