non-mem-to-mem transfers require dw aligned byte count.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

This enables the functionality added in the previous
patches.  Userspace acceleration drivers can use the
CS ioctl to submit command buffers to the async DMA
rings.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Allows us to use async DMA from userspace.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Allows us to use the DMA ring from userspace.
DMA doesn't have a good NOP packet in which to embed the
reloc idx, so userspace has to add a reloc for each
buffer used and order them to match the command stream.

v2: fix address bounds checking
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Allows us to use the DMA ring from userspace.
DMA doesn't have a good NOP packet in which to embed the
reloc idx, so userspace has to add a reloc for each
buffer used and order them to match the command stream.

v2: fix address bounds checking, reloc indexing
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Fix the size computation of the htile buffer.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

We need to hold bdev->fence_lock while grabbing a reference to
the fence, to prevent concurrent clearing/changing of the
ttm_bo->sync_obj field.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

With the new per-crtc locking mutliple set-cursor calls could happen
in parallel. Out of sheer paranoia I've opted for an irqsave spinlock.
But if there's indeed an access from interrupt contexts to these regs
it's already broken with the old code, so this can likely just be
reduced to a normal spinlock. Otoh the pageflip completion happens
from the vblank irq handler ...
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Just refactoring to make the next patche simpler. Now all indirect register
access in the new modesetting driver should go through the r100_mm_(w|r)reg
fucntions.

RADEON_READ_MM from the old driver seems to be totally unused, so just kill
it.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

* 'drm-next-3.8' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: fix fence driver for dma ring when wb is disabled
  drm/radeon/si: add VM CS checker support for CP DMA
  drm/radeon/cayman: add VM CS checker support for CP DMA
  drm/radeon: add support for CP DMA packet to evergreen CS checker
  drm/radeon: add support for CP DMA packet to r6xx/r7xx CS checker
  drm/radeon: add register headers for CP DMA on r6xx-SI
  drm/radeon: improve mc_stop/mc_resume on r5xx-r7xx
  drm/radeon: fix amd afusion gpu setup aka sumo v2
  drm/radeon: do not move bo to different placement at each cs

The dma ring can't write to register thus have to write to memory
its fence value. This ensure that it doesn't try to use scratch
register for dma ring fence driver.

Should fix:
https://bugs.freedesktop.org/show_bug.cgi?id=58166Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

Need to verify for copies involving registers.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Need to verify for copies involving registers.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Currently only memory and GDS transfers are allowed.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Currently only memory to memory transfers are allowed.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Along the same lines of what was done for evergreen+
in the last kernel.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Set the proper number of tile pipe that should be a multiple of
pipe depending on the number of se engine.

Fix:
https://bugs.freedesktop.org/show_bug.cgi?id=56405
https://bugs.freedesktop.org/show_bug.cgi?id=56720

v2: Don't change sumo2
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Cc: stable@vger.kernel.org
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

The bo creation placement is where the bo will be. Instead of trying
to move bo at each command stream let this work to another worker
thread that will use more advance heuristic.

agd5f: remove leftover unused variable
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

Alex writes:
"adds support for the
asynchronous DMA engines on r6xx-SI.  These engines are used
for ttm bo moves and VM page table updates currently.  They
could also be exposed via the CS ioctl for userspace use,
but I haven't had a chance to add proper CS checker patches
for them yet.  These patches have been tested extensively
internally for months, so they should be pretty solid."

* 'drm-next-3.8' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: use DMA engine for VM page table updates on SI
  drm/radeon: add dma engine support for vm pt updates on si (v2)
  drm/radeon: use DMA engine for VM page table updates on cayman/TN
  drm/radeon: add dma engine support for vm pt updates on ni (v5)
  drm/radeon: use async dma for ttm buffer moves on 6xx-SI
  drm/radeon/kms: add support for dma rings to radeon_test_moves()
  drm/radeon/kms: Add initial support for async DMA on SI
  drm/radeon/kms: Add initial support for async DMA on cayman/TN
  drm/radeon/kms: Add initial support for async DMA on evergreen
  drm/radeon/kms: Add initial support for async DMA on r6xx/r7xx

DMA engine has special packets to facilitate this and it also keeps
the 3D engine free for other things.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Async DMA has a special packet for contiguous pt updates
which saves overhead.

v2: rebase
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

DMA engine has special packets to facilitate this and it also keeps
the 3D engine free for other things.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Async DMA has a special packet for contiguous pt updates
which saves overhead.

v2: leave the CP method enabled for now as doing the updates
in the DMA rings is not working properly yet.

v3: update for 2 level pts

v4: rebase

v5: drop pte/pde packet.  doesn't seem to work on NI.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Pretty much the same as cayman.  Some changes to the copy
packets.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

There are 2 async DMA engines on cayman, one at 0xd000 and
one at 0xd800.  The programming interface is the same as
evergreen however there are some changes to the commands
for using vmids.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Pretty similar to 6xx/7xx except the count field increased in the
packet header and the max IB size increased.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Uses the new multi-ring infrastucture.  6xx/7xx has a single
async DMA ring.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

All items on the lru list are always reservable, so this is a stupid
thing to keep. Not only that, it is used in a way which would
guarantee deadlocks if it were ever to be set to block on reserve.

This is a lot of churn, but mostly because of the removal of the
argument which can be nested arbitrarily deeply in many places.

No change of code in this patch except removal of the no_wait_reserve
argument, the previous patch removed the use of no_wait_reserve.

v2:
 - Warn if -EBUSY is returned on reservation, all objects on the list
   should be reservable. Adjusted patch slightly due to conflicts.
v3:
 - Focus on no_wait_reserve removal only.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

Replace the goto loop with a simple for each loop, and only run the
delayed destroy cleanup if we can reserve the buffer first.

No race occurs, since lru lock is never dropped any more. An empty list
and a list full of unreservable buffers both cause -EBUSY to be returned,
which is identical to the previous situation, because previously buffers
on the lru list were always guaranteed to be reservable.

This should work since currently ttm guarantees items on the lru are
always reservable, and reserving items blockingly with some bo held
are enough to cause you to run into a deadlock.

Currently this is not a concern since removal off the lru list and
reservations are always done with atomically, but when this guarantee
no longer holds, we have to handle this situation or end up with
possible deadlocks.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

Replace the while loop with a simple for each loop, and only run the
delayed destroy cleanup if we can reserve the buffer first.

No race occurs, since lru lock is never dropped any more. An empty list
and a list full of unreservable buffers both cause -EBUSY to be returned,
which is identical to the previous situation.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

By removing the unlocking of lru and retaking it immediately, a race is
removed where the bo is taken off the swap list or the lru list between
the unlock and relock. As such the cleanup_refs code can be simplified,
it will attempt to call ttm_bo_wait non-blockingly, and if it fails
it will drop the locks and perform a blocking wait, or return an error
if no_wait_gpu was set.

The need for looping is also eliminated, since swapout and evict_mem_first
will always follow the destruction path, no new fence is allowed
to be attached. As far as I can see this may already have been the case,
but the unlocking / relocking required a complicated loop to deal with
re-reservation.

Changes since v1:
 - Simplify no_wait_gpu case by folding it in with empty ddestroy.
 - Hold a reservation while calling ttm_bo_cleanup_memtype_use again.
Changes since v2:
 - Do not remove bo from lru list while waiting
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

The few places that care should have those checks instead.
This allows destruction of bo backed memory without a reservation.
It's required for being able to rework the delayed destroy path,
as it is no longer guaranteed to hold a reservation before unlocking.

However any previous wait is still guaranteed to complete, and it's
one of the last things to be done before the buffer object is freed.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

This requires changing the order in ttm_bo_cleanup_refs_or_queue to
take the reservation first, as there is otherwise no race free way to
take lru lock before fence_lock.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

Alex writes:
Pretty minor -next pull request.  We some additional new bits waiting
internally for release.  Hopefully Monday we can get at least some of
them out.  The others will probably take a few more weeks.

Highlights of the current request:
- ELD registers for passing audio information to the sound hardware
- Handle GPUVM page faults more gracefully
- Misc fixes

Merge radeon test
* 'drm-next-3.8' of git://people.freedesktop.org/~agd5f/linux: (483 commits)
  drm/radeon: bump driver version for new info ioctl requests
  drm/radeon: fix eDP clk and lane setup for scaled modes
  drm/radeon: add new INFO ioctl requests
  drm/radeon/dce32+: use fractional fb dividers for high clocks
  drm/radeon: use cached memory when evicting for vram on non agp
  drm/radeon: add a CS flag END_OF_FRAME
  drm/radeon: stop page faults from hanging the system (v2)
  drm/radeon/dce4/5: add registers for ELD handling
  drm/radeon/dce3.2: add registers for ELD handling
  radeon: fix pll/ctrc mapping on dce2 and dce3 hardware
  Linux 3.7-rc7
  powerpc/eeh: Do not invalidate PE properly
  Revert "drm/i915: enable rc6 on ilk again"
  ALSA: hda - Fix build without CONFIG_PM
  of/address: sparc: Declare of_iomap as an extern function for sparc again
  PM / QoS: fix wrong error-checking condition
  bnx2x: remove redundant warning log
  vxlan: fix command usage in its doc
  8139cp: revert "set ring address before enabling receiver"
  MPI: Fix compilation on MIPS with GCC 4.4 and newer
  ...

Conflicts:
	drivers/gpu/drm/exynos/exynos_drm_encoder.c
	drivers/gpu/drm/exynos/exynos_drm_fbdev.c
	drivers/gpu/drm/nouveau/core/engine/disp/nv50.c

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Need to use the adjusted mode since we are sending native
timing and using the scaler for non-native modes.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
cc: stable@vger.kernel.org