Hi Linus,
Since we just introduced GEM in this kernel, and this patch series only
touches GEM and fixes some problems discovered in it since it went
upstream, Intel would like for these fixes to go in ASAP. They fix a host
of general bugs founds by running OpenGL conformance tests and using GEM
from the userspace 2D driver. If you don't want them now, they can wait
until stable.
Please pull the 'drm-gem-update' branch from
ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6.git drm-gem-update
Dave.
drivers/gpu/drm/i915/i915_drv.h | 15 +-
drivers/gpu/drm/i915/i915_gem.c | 637 ++++++++++++++++++++------------
drivers/gpu/drm/i915/i915_gem_proc.c | 5 +-
drivers/gpu/drm/i915/i915_gem_tiling.c | 7 +-
drivers/gpu/drm/i915/i915_reg.h | 1 +
5 files changed, 420 insertions(+), 245 deletions(-)
commit 0235439232cb6f8a54f8976aa8330c1c98ebad0b
Author: Eric Anholt <[email protected]>
Date: Wed Nov 26 13:58:13 2008 -0800
drm/i915: Return error in i915_gem_set_to_gtt_domain if we're not in the GTT.
It's only for flushing caches appropriately for GTT access, not for actually
getting it there. Prevents potential smashing of cpu read/write domains on
unbound objects.
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit ac94a962b24a88ea5d00f4697550d9982f300751
Author: Keith Packard <[email protected]>
Date: Thu Nov 20 23:30:27 2008 -0800
drm/i915: Retry execbuffer pinning after clearing the GTT
If we fail to pin all of the buffers in an execbuffer request, go through
and clear the GTT and try again to see if its just a matter of fragmentation
Signed-off-by: Keith Packard <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit 646f0f6e43bf6628b1f0f8ca6c0227ce72e8ef3c
Author: Keith Packard <[email protected]>
Date: Thu Nov 20 23:23:03 2008 -0800
drm/i915: Move the execbuffer domain computations together
This eliminates the dev_set_domain function and just in-lines it
where its used, with the goal of moving the manipulation and use of
invalidate_domains and flush_domains closer together. This also
avoids calling add_request unless some domain has been flushed.
Signed-off-by: Keith Packard <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit c0d90829288942fa06d7483f9e84059a64605da5
Author: Keith Packard <[email protected]>
Date: Thu Nov 20 23:11:08 2008 -0800
drm/i915: Rename object_set_domain to object_set_to_gpu_domain
Now that the CPU and GTT domain operations are isolated to their own
functions, the previously general-purpose set_domain function is now used
only to set GPU domains. It also has no failure cases, which is important as
this eliminates any possible interruption of the computation of new object
domains and subsequent emmission of the flushing instructions into the ring.
Signed-off-by: Keith Packard <[email protected]>
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit e47c68e9c5d71e2faab8c2b82f57c6c73e6456fd
Author: Eric Anholt <[email protected]>
Date: Fri Nov 14 13:35:19 2008 -0800
drm/i915: Make a single set-to-cpu-domain path and use it wherever needed.
This fixes several domain management bugs, including potential lack of cache
invalidation for pread, potential failure to wait for set_domain(CPU, 0),
and more, along with producing more intelligible code.
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit 2ef7eeaa553d88e78d9a4520271f26a7bc0e2968
Author: Eric Anholt <[email protected]>
Date: Mon Nov 10 10:53:25 2008 -0800
drm/i915: Make a single set-to-gtt-domain path.
This fixes failure to flush caches in the relocation update path, and
failure to wait in the set_domain ioctl, each of which could lead to incorrect
rendering.
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit b670d8158283c35842ae1c650f75c375d8710607
Author: Eric Anholt <[email protected]>
Date: Fri Nov 14 16:27:47 2008 -0800
drm/i915: If interrupted while setting object domains, still emit the flush.
Otherwise, we would leave the objects in an inconsistent state, such as
write_domain == 0 but on the flushing list.
Signed-off-by: Dave Airlie <[email protected]>
commit ce44b0ea3dc04236d852d78a06f850d1f7b03f3d
Author: Eric Anholt <[email protected]>
Date: Thu Nov 6 16:00:31 2008 -0800
drm/i915: Move flushing list cleanup from flush request retire to request emit.
obj_priv->write_domain is "write domain if the GPU went idle now", not
"write domain at this moment." By postponing the clear, we confused the
concept, required more storage, and potentially emitted more flushes than
are required.
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
commit a7f014f2de04893f95cfe40fe35f15c8dae4b36e
Author: Eric Anholt <[email protected]>
Date: Tue Nov 25 14:02:05 2008 -0800
drm/i915: Respect GM965/GM45 bit-17-instead-of-bit-11 option for swizzling.
This fixes readpixels and buffer corruption when swapped out and in by
disabling tiling on them.
Now that we know that the bit 17 mode isn't just a mistake of older chipsets,
we'll need to work on a clever fix so that we can get the performance of
tiling on these chipsets, but that will require intrusive changes targeted
at the next kernel release, not this one.
Signed-off-by: Eric Anholt <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>