In drm_release(), we currently call drm_remove_magic() if the drm_file
has a drm-magic attached. Therefore, once drm_master_release() is called,
the magic-list _must_ be empty.

By dropping the no-op cleanup, we can move "struct drm_magic_entry" to
drm_auth.c and avoid exposing it to all of DRM.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NThierry Reding <treding@nvidia.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Make all the drm_vma_entry handling local to drm_vm.c and hide it from
global headers. This requires to extract the inlined legacy drm_vma_entry
cleanup into a small helper and also move a weirdly placed drm_vma_info
helper into drm_vm.c.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NThierry Reding <treding@nvidia.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Move internal declarations to drm_legacy.h and add drm_legacy_*() prefix
to all legacy functions.

[airlied: add a bit of an explaination to drm_legacy.h]
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NThierry Reding <treding@nvidia.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Radeon UMS is the last user of drm_buffer. Move it out of sight so radeon
can drop it together with UMS.
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NThierry Reding <treding@nvidia.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

This crash was already here before the conversion, but qxl never leaked
hard enough to hit this.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

This is how you implement a memory sieve in a driver. ;-)
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

The locking of release_lock was stupid; t should have been be called with
fence_lock_irq if it was legitimately used. Unfortunately it never protected
anything except the fence implementation correctly.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

As there should not be any other virtual device that might share buffers,
the callbacks remain empty stubs. Still prime can be used to transfer buffers
between processes that use qxl.
Signed-off-by: NAndreas Pokorny <andreas.pokorny@canonical.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Signed-off-by: NAndreas Pokorny <andreas.pokorny@canonical.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

Merge the move to generic fences for TTM using drivers.

* 'for-airlied-next' of git://people.freedesktop.org/~mlankhorst/linux:
  drm/nouveau: use shared fences for readable objects
  drm/nouveau: Keep only a single list for validation.
  drm/ttm: use rcu in core ttm
  drm/vmwgfx: use rcu in vmw_user_dmabuf_synccpu_grab
  drm/radeon: use rcu waits in some ioctls
  drm/nouveau: use rcu in nouveau_gem_ioctl_cpu_prep
  drm/ttm: flip the switch, and convert to dma_fence
  drm/qxl: rework to new fence interface
  drm/nouveau: rework to new fence interface
  drm/vmwgfx: rework to new fence interface, v2
  drm/vmwgfx: get rid of different types of fence_flags entirely
  drm/radeon: use common fence implementation for fences, v4
  drm/ttm: kill off some members to ttm_validate_buffer
  drm/ttm: add interruptible parameter to ttm_eu_reserve_buffers
  drm/ttm: kill fence_lock
  drm/ttm: call ttm_bo_wait while inside a reservation
  drm/nouveau: require reservations for nouveau_fence_sync and nouveau_bo_fence
  drm/nouveau: add reservation to nouveau_gem_ioctl_cpu_prep

drm-intel-next-2014-08-22:
- basic code for execlist, which is the fancy new cmd submission on gen8. Still
  disabled by default (Ben, Oscar Mateo, Thomas Daniel et al)
- remove the useless usage of console_lock for I915_FBDEV=n (Chris)
- clean up relations between ctx and ppgtt
- clean up ppgtt lifetime handling (Michel Thierry)
- various cursor code improvements from Ville
- execbuffer code cleanups and secure batch fixes (Chris)
- prep work for dev -> dev_priv transition (Chris)
- some of the prep patches for the seqno -> request object transition (Chris)
- various small improvements all over

* tag 'drm-intel-next-2014-09-01' of git://anongit.freedesktop.org/drm-intel: (86 commits)
  drm/i915: fix suspend/resume for GENs w/o runtime PM support
  drm/i915: Update DRIVER_DATE to 20140822
  drm: fix plane rotation when restoring fbdev configuration
  drm/i915/bdw: Disable execlists by default
  drm/i915/bdw: Enable Logical Ring Contexts (hence, Execlists)
  drm/i915/bdw: Document Logical Rings, LR contexts and Execlists
  drm/i915/bdw: Print context state in debugfs
  drm/i915/bdw: Display context backing obj & ringbuffer info in debugfs
  drm/i915/bdw: Display execlists info in debugfs
  drm/i915/bdw: Disable semaphores for Execlists
  drm/i915/bdw: Make sure gpu reset still works with Execlists
  drm/i915/bdw: Don't write PDP in the legacy way when using LRCs
  drm/i915: Track cursor changes as frontbuffer tracking flushes
  drm/i915/bdw: Help out the ctx switch interrupt handler
  drm/i915/bdw: Avoid non-lite-restore preemptions
  drm/i915/bdw: Handle context switch events
  drm/i915/bdw: Two-stage execlist submit process
  drm/i915/bdw: Write the tail pointer, LRC style
  drm/i915/bdw: Implement context switching (somewhat)
  drm/i915/bdw: Emission of requests with logical rings
  ...

Conflicts:
	drivers/gpu/drm/i915/i915_drv.c

nouveau keeps track in userspace whether a buffer is being
written to or being read, but it doesn't use that information.

Change this to allow multiple readers on the same bo.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Maintain the original order to handle VRAM/GART/mixed correctly for <nv50,
it's likely not as important on newer cards.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

With the conversion to the reservation api this should be safe.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

Final driver! \o/

This is not a proper dma_fence because the hardware may never signal
anything, so don't use dma-buf with qxl, ever.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Use the new fence interface on vmwgfx too.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

---
Changes since v1:
Fix a sleeping function called from invalid context in enable_signaling.

Only one type was ever used. This is needed to simplify the fence
support in the next commit.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

Changes since v1:
- Kill the sw interrupt dance, add and use
  radeon_irq_kms_sw_irq_get_delayed instead.
- Change custom wait function, lockdep complained about it.
  Holding exclusive_lock in the wait function might cause deadlocks.
  Instead do all the processing in .enable_signaling, and wait
  on the global fence_queue to pick up gpu resets.
- Process all fences in radeon_gpu_reset after reset to close a race
  with the trylock in enable_signaling.
Changes since v2:
- Small changes to work with the rewritten lockup recovery patches.
Changes since v3:
- Call radeon_fence_schedule_check when exclusive_lock cannot be
  acquired to always cause a wake up.
- Reset irqs from hangup check.
- Drop reading seqno in the callback, use cached value.
- Fix indentation in radeon_fence_default_wait
- Add a radeon_test_signaled function, drop a few test_bit calls.
- Make to_radeon_fence global.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: NChristian König <christian.koenig@amd.com>

This reorders the list to keep track of what buffers are reserved,
so previous members are always unreserved.

This gets rid of some bookkeeping that's no longer needed,
while simplifying the code some.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

It seems some drivers really want this as a parameter,
like vmwgfx.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

No users are left, kill it off! :D
Conversion to the reservation api is next on the list, after
that the functionality can be restored with rcu.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

This is the last remaining function that doesn't use the reservation
lock completely to fence off access to a buffer.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>

This will ensure we always hold the required lock when calling those functions.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Apart from some code inside ttm itself and nouveau_bo_vma_del,
this is the only place where ttm_bo_wait is used without a reservation.
Fix this so we can remove the fence_lock later on.

After the switch to rcu the reservation lock will be
removed again.
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Acked-by: NBen Skeggs <bskeggs@redhat.com>

Just clearing out my -next queue before I go on vacation.  Two UVD
improvements that depend on the ttm change you just merged.

* 'drm-next-3.18' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: preallocate mem for UVD create/destroy msg
  drm/radeon: allow UVD to use a second 256MB segment

llocating memory for UVD create and destroy messages can fail, which is
rather annoying when this happens in the middle of a GPU reset. Try to
avoid this condition by preallocating a page for those dummy messages.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This improves concurrent stream decoding.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

More radeon changes for drm-next.  Highlights:
- UVD support for older asics
- Reset rework in preparation for Maarten's fence patches
I have a few more patches which depend on Christian's ttm changes,
I'll send them out separately once you've merged the ttm changes.

* 'drm-next-3.18' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: drop doing resets in a work item
  drm/radeon: drop RADEON_FENCE_SIGNALED_SEQ v2
  drm/radeon: add timeout argument to radeon_fence_wait_seq v2
  drm/radeon: handle lockup in delayed work, v5
  drm/radeon: take exclusive_lock in read mode during ring tests, v5
  drm/radeon: force fence completion only on problematic rings (v2)
  drm/radeon: wake up all fences on manual reset
  drm/radeon: add UVD fw names for older asic
  drm/radeon: enable RB_ARB before resetting the VCPU
  drm/radeon: 760G/780V/880V don't have UVD
  drm/radeon: implement UVD hw workarounds for R6xx v3
  drm/radeon: add UVD support for older asics v4
  drm/radeon: add set_uvd_clocks callback for r6xx v4
  drm/radeon: properly init UVD MC bits on R600
  drm/radeon: force UVD buffers into VRAM on RS[78]80 v2
  drm/radeon: move the IB test after the AGP fallback

pull in placement changes radeon requires.

* 'ttm_pfn' of git://people.freedesktop.org/~deathsimple/linux:
  drm/ttm: move fpfn and lpfn into each placement v2

Blocking completely innocent processes with a GPU reset is
a pretty bad idea. Just set needs_reset and let the next
command submission or fence wait do the job.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

It's causing issues with VMID handling and comparing the
fence value two times actually doesn't make handling faster.

v2: rebased on reset changes
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This makes it possible to wait for a specific amount of time,
rather than wait until infinity.

v2 (chk): rebased on other changes
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

v5 (chk): complete rework, start when the first fence is emitted,
          stop when the last fence is signalled, make it work
          correctly with GPU resets, cleanup radeon_fence_wait_seq
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

This is needed for the next commit, because the lockup detection
will need the read lock to run.

v4 (chk): split out forced fence completion, remove unrelated changes,
          add and handle in_reset flag
v5 (agd5f): rebase fix
Signed-off-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NChristian König <christian.koenig@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

Instead of resetting all fence numbers, only reset the
number of the problematic ring. Split out from a patch
from Maarten Lankhorst <maarten.lankhorst@canonical.com>

v2 (agd5f): rebase build fix
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NMaarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>