openwrt/staging/blogic.git
11 years agodrm/radeon: switch back to using the DMA ring for VM PT updates
Alex Deucher [Thu, 31 Jan 2013 21:26:02 +0000 (16:26 -0500)]
drm/radeon: switch back to using the DMA ring for VM PT updates

Now that we have switched to using IBs for page table updates,
we can switch back the using the DMA ring.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use IBs for VM page table updates v2
Alex Deucher [Fri, 1 Feb 2013 16:32:42 +0000 (17:32 +0100)]
drm/radeon: use IBs for VM page table updates v2

For very large page table updates, we can exceed the
size of the ring.  To avoid this, use an IB to perform
the page table update.

v2(ck): cleanup the IB infrastructure and the use it instead
        of filling the struct ourself.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
11 years agodrm/radeon: don't reset the MC on IGPs/APUs
Alex Deucher [Thu, 24 Jan 2013 20:00:17 +0000 (15:00 -0500)]
drm/radeon: don't reset the MC on IGPs/APUs

The MC isn't part of the GPU per se.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use the reset mask to determine if rings are hung
Alex Deucher [Thu, 24 Jan 2013 16:37:19 +0000 (11:37 -0500)]
drm/radeon: use the reset mask to determine if rings are hung

fetch the reset mask and check if the relevant ring flags
are set to determine whether the ring is hung or not.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: halt engines before disabling MC (si)
Alex Deucher [Thu, 24 Jan 2013 00:00:25 +0000 (19:00 -0500)]
drm/radeon: halt engines before disabling MC (si)

It's better to halt the engines before we disable the MC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: halt engines before disabling MC (cayman/TN)
Alex Deucher [Wed, 23 Jan 2013 23:59:17 +0000 (18:59 -0500)]
drm/radeon: halt engines before disabling MC (cayman/TN)

It's better to halt the engines before we disable the MC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: halt engines before disabling MC (evergreen)
Alex Deucher [Wed, 23 Jan 2013 23:57:56 +0000 (18:57 -0500)]
drm/radeon: halt engines before disabling MC (evergreen)

It's better to halt the engines before we disable the MC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: halt engines before disabling MC (6xx/7xx)
Alex Deucher [Wed, 23 Jan 2013 23:56:08 +0000 (18:56 -0500)]
drm/radeon: halt engines before disabling MC (6xx/7xx)

It's better to halt the engines before we disable the MC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use status regs to determine what to reset (si)
Alex Deucher [Sat, 19 Jan 2013 00:36:20 +0000 (19:36 -0500)]
drm/radeon: use status regs to determine what to reset (si)

When we attempt the reset the GPU, look at the status registers
to determine what blocks need to be reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use status regs to determine what to reset (cayman)
Alex Deucher [Sat, 19 Jan 2013 00:17:22 +0000 (19:17 -0500)]
drm/radeon: use status regs to determine what to reset (cayman)

When we attempt the reset the GPU, look at the status registers
to determine what blocks need to be reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use status regs to determine what to reset (evergreen)
Alex Deucher [Fri, 18 Jan 2013 23:55:54 +0000 (18:55 -0500)]
drm/radeon: use status regs to determine what to reset (evergreen)

When we attempt the reset the GPU, look at the status registers
to determine what blocks need to be reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use status regs to determine what to reset (6xx/7xx)
Alex Deucher [Fri, 18 Jan 2013 23:12:22 +0000 (18:12 -0500)]
drm/radeon: use status regs to determine what to reset (6xx/7xx)

When we attempt the reset the GPU, look at the status registers
to determine what blocks need to be reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: rework GPU reset on cayman/TN
Alex Deucher [Fri, 18 Jan 2013 20:08:38 +0000 (15:08 -0500)]
drm/radeon: rework GPU reset on cayman/TN

Update the code to better match the recommended
programming sequence for soft reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: rework GPU reset on cayman/TN
Alex Deucher [Fri, 18 Jan 2013 19:51:38 +0000 (14:51 -0500)]
drm/radeon: rework GPU reset on cayman/TN

Update the code to better match the recommended
programming sequence for soft reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: rework GPU reset on evergreen
Alex Deucher [Fri, 18 Jan 2013 19:28:41 +0000 (14:28 -0500)]
drm/radeon: rework GPU reset on evergreen

Update the code to better match the recommended
programming sequence for soft reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: rework GPU reset on r6xx/r7xx
Alex Deucher [Fri, 18 Jan 2013 18:53:37 +0000 (13:53 -0500)]
drm/radeon: rework GPU reset on r6xx/r7xx

Update the code to better match the recommended
programming sequence for soft reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add a bios scratch asic hung helper
Alex Deucher [Fri, 18 Jan 2013 18:05:39 +0000 (13:05 -0500)]
drm/radeon: add a bios scratch asic hung helper

Used by all asic families from r600+.
Flag for the vbios and later instances of the driver
that the GPU is hung.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add additional reset flags
Alex Deucher [Fri, 18 Jan 2013 17:18:17 +0000 (12:18 -0500)]
drm/radeon: add additional reset flags

This adds further flags for fine grained reset.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: Deprecate UMS support v2
Christian König [Mon, 21 Jan 2013 12:58:46 +0000 (13:58 +0100)]
drm/radeon: Deprecate UMS support v2

KMS support is out and stable for a couple of years now and
the userspace code has deprecated or abandoned the old UMS interface.

So make the KMS interface the default and deprecate the UMS interface
in the kernel as well.

v2: rebased on alex/drm-next-3.9-wip

Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoradeon/kms: cleanup async dma packet checking
Jerome Glisse [Mon, 14 Jan 2013 16:32:27 +0000 (11:32 -0500)]
radeon/kms: cleanup async dma packet checking

This simplify and cleanup the async dma checking.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: consolidate redundant macros and constants
Ilija Hadzic [Wed, 2 Jan 2013 23:27:48 +0000 (18:27 -0500)]
drm/radeon: consolidate redundant macros and constants

After refactoring the _cs logic, we ended up with many
macros and constants that #define the same thing.
Clean'em up.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use common next_reloc function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:47 +0000 (18:27 -0500)]
drm/radeon: use common next_reloc function

This patch eliminates ASIC-specific ***_cs_packet_next_reloc
functions and hooks up the new common function.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: pull out common next_reloc function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:46 +0000 (18:27 -0500)]
drm/radeon: pull out common next_reloc function

next_reloc function does the same thing in all ASICs with
the exception of R600 which has a special case in legacy mode.
Pull out the common function in preparation for refactoring.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: rename r100_cs_dump_packet to radeon_cs_dump_packet
Ilija Hadzic [Wed, 2 Jan 2013 23:27:45 +0000 (18:27 -0500)]
drm/radeon: rename r100_cs_dump_packet to radeon_cs_dump_packet

This function is not limited to r100, but it can dump a
(raw) packet for any ASIC. Rename it accordingly and move
its declaration to radeon.h

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: add a check to wait_reg_mem command
Ilija Hadzic [Wed, 2 Jan 2013 23:27:44 +0000 (18:27 -0500)]
drm/radeon: add a check to wait_reg_mem command

WAIT_REG_MEM on register does not allow the use of PFP.
Enforce this restriction when checking packets sent from
userland.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: refactor vline packet parsing function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:43 +0000 (18:27 -0500)]
drm/radeon: refactor vline packet parsing function

vline packet parsing function for R600 and Evergreen+ are
the same, except that they use different registers. Factor
out the algorithm into a common function that uses register
table passed from ASIC-specific caller.

This reduces ASIC-specific function to (trivial) setup
of register table and call into the common function.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: factor out cs_next_is_pkt3_nop function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:42 +0000 (18:27 -0500)]
drm/radeon: factor out cs_next_is_pkt3_nop function

Once we factored out radeon_cs_packet_parse function,
evergreen_cs_next_is_pkt3_nop and r600_cs_next_is_pkt3_nop
functions became identical, so they can be factored out
into a common function.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: use common cs packet parse function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:41 +0000 (18:27 -0500)]
drm/radeon: use common cs packet parse function

We now have a common radeon_cs_packet_parse function
that is good for all ASICs. Hook it up and eliminate
ASIC-specific versions.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: implement common cs packet parse function
Ilija Hadzic [Wed, 2 Jan 2013 23:27:40 +0000 (18:27 -0500)]
drm/radeon: implement common cs packet parse function

CS packet parse functions have a lot of in common across
all ASICs. Implement a common function and take care of
small differences between families inside the function.

This patch is a prep for major refactoring and consolidation
of CS parsing code.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix formatting
Ilija Hadzic [Wed, 2 Jan 2013 23:27:39 +0000 (18:27 -0500)]
drm/radeon: fix formatting

Preparatory patch: patches to follow will touch a piece of code
that had broken indentication, so fix it before touching it.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove unused prototype from radeon_cs
Ilija Hadzic [Wed, 2 Jan 2013 23:27:38 +0000 (18:27 -0500)]
drm/radeon: remove unused prototype from radeon_cs

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: remove unecessary assignment
Ilija Hadzic [Wed, 2 Jan 2013 23:27:37 +0000 (18:27 -0500)]
drm/radeon: remove unecessary assignment

length_dw field was assigned twice. While at it, move user_ptr
assignment together with all other assignments to p->chunks[i]
structure to make the code more readable.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: switch back to the CP ring for VM PT updates
Alex Deucher [Thu, 31 Jan 2013 21:19:19 +0000 (16:19 -0500)]
drm/radeon: switch back to the CP ring for VM PT updates

For large VM page table updates, we can sometimes generate
more packets than there is space on the ring.  This happens
more readily with the DMA ring since it is 64K (vs 1M for the
CP).  For now, switch back to the CP.  For the next kernel,
I have a patch to utilize IBs for VM PT updates which
alleviates this problem.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=58354

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: prevent crash in the ring space allocation
Alex Deucher [Wed, 30 Jan 2013 19:24:09 +0000 (14:24 -0500)]
drm/radeon: prevent crash in the ring space allocation

If the requested number of DWs on the ring is larger than
the size of the ring itself, return an error.

In testing with large VM updates, we've seen crashes when we
try and allocate more space on the ring than the total size
of the ring without checking.

This prevents the crash but for large VM updates or bo moves
of very large buffers, we will need to break the transaction
down into multiple batches.  I have patches to use IBs for
the next kernel.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: Calling object_unrefer() when creating fb failure
liu chuansheng [Thu, 31 Jan 2013 14:13:00 +0000 (22:13 +0800)]
drm/radeon: Calling object_unrefer() when creating fb failure

When kzalloc() failed in radeon_user_framebuffer_create(), need to
call object_unreference() to match the object_reference().

Signed-off-by: liu chuansheng <chuansheng.liu@intel.com>
Signed-off-by: xueminsu <xuemin.su@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon/r5xx-r7xx: wait for the MC to settle after MC blackout
Alex Deucher [Thu, 31 Jan 2013 14:01:59 +0000 (09:01 -0500)]
drm/radeon/r5xx-r7xx: wait for the MC to settle after MC blackout

Some chips seem to need a little delay after blacking out
the MC before the requests actually stop. Stop DMAR errors
reported by Shuah Khan.

Reported-by: Shuah Khan <shuahkhan@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon/evergreen+: wait for the MC to settle after MC blackout
Alex Deucher [Thu, 31 Jan 2013 14:00:52 +0000 (09:00 -0500)]
drm/radeon/evergreen+: wait for the MC to settle after MC blackout

Some chips seem to need a little delay after blacking out
the MC before the requests actually stop.

May fix:
https://bugs.freedesktop.org/show_bug.cgi?id=56139
https://bugs.freedesktop.org/show_bug.cgi?id=57567

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: protect against div by 0 in backend setup
Mikko Tiihonen [Wed, 30 Jan 2013 19:10:04 +0000 (14:10 -0500)]
drm/radeon: protect against div by 0 in backend setup

Make sure at least one RB is enabled in
r6xx_remap_render_backend() to avoid an division by
zero in some corner cases.

See:
https://bugzilla.redhat.com/show_bug.cgi?id=892233

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: fix backend map setup on 1 RB sumo boards
Alex Deucher [Wed, 30 Jan 2013 18:57:40 +0000 (13:57 -0500)]
drm/radeon: fix backend map setup on 1 RB sumo boards

Need to adjust the backend map depending on which
RB is enabled.

Fixes:
https://bugzilla.redhat.com/show_bug.cgi?id=892233

Reported-by: Mikko Tiihonen <mikko.tiihonen@iki.fi>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: add quirk for RV100 board
Alex Deucher [Tue, 29 Jan 2013 21:36:47 +0000 (16:36 -0500)]
drm/radeon: add quirk for RV100 board

vbios says external TMDS while the board is actually
internal TMDS.

fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=60037

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: add WAIT_UNTIL to the non-VM safe regs list for cayman/TN
Alex Deucher [Mon, 28 Jan 2013 02:57:51 +0000 (21:57 -0500)]
drm/radeon: add WAIT_UNTIL to the non-VM safe regs list for cayman/TN

Newer versions of mesa emit this.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: fix MC blackout on evergreen+
Christopher Staite [Sat, 26 Jan 2013 16:10:58 +0000 (11:10 -0500)]
drm/radeon: fix MC blackout on evergreen+

Force the crtc mem requests on/off immediately rather
than waiting for the double buffered updates to kick in.
Seems we miss the update in certain conditions.  Also
handle the DCE6 case.

Signed-off-by: Christopher Staite <chris@yourdreamnet.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agoMerge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel
Dave Airlie [Sat, 26 Jan 2013 08:17:39 +0000 (18:17 +1000)]
Merge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel

Just a few small things:
- 2x workaround bits from Chris to fix up the new scanline waits enabled
  in 3.8 on snb. People who've been struck by this on dual-screen also
  need to upgrade the ddx.
- Dump the kernel version into i915_error_state, we've had a few mixups
  there recently.
- Disable gfx DMAR on gen4 devices, acked by David Woodhouse.

* 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel:
  drm/i915: dump UTS_RELEASE into the error_state
  iommu/intel: disable DMAR for g4x integrated gfx
  drm/i915: GFX_MODE Flush TLB Invalidate Mode must be '1' for scanline waits
  drm/i915: Disable AsyncFlip performance optimisations

11 years agoMerge branch 'exynos-drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git...
Dave Airlie [Fri, 25 Jan 2013 07:27:00 +0000 (17:27 +1000)]
Merge branch 'exynos-drm-fixes' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

Inki writes:

"This pull request includes some bug fixes, code cleanups and exception codes.
If there is any problem, please kindly let me know."

* 'exynos-drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos:
  drm/exynos: add check for the device power status
  drm/exynos: Make 'drm_hdmi_get_edid' static
  drm/exynos: fimd and ipp are broken on multiplatform
  drm/exynos: don't include plat/gpio-cfg.h
  drm/exynos: Remove "internal" interrupt handling
  drm/exynos: Add missing static specifiers in exynos_drm_rotator.c
  drm/exynos: Replace mdelay with usleep_range
  drm/exynos: Make ipp_handle_cmd_work static
  drm/exynos: Make g2d_userptr_get_dma_addr static
  drm/exynos: consider DMA_NONE flag to dmabuf import
  drm/exynos: free sg object if dma_map_sg is failed
  drm/exynos: added validation of edid for vidi connection
  drm/exynos: let drm handle edid allocations

11 years agodrm/exynos: add check for the device power status
Shirish S [Thu, 24 Jan 2013 03:03:18 +0000 (22:03 -0500)]
drm/exynos: add check for the device power status

V2: Add mutex protection, while read.

The hdmi and mixer win_commit calls currently are
not checking the status of IP before updating the
respective registers, this patch adds this check.

Signed-off-by: Shirish S <s.shirish@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: Make 'drm_hdmi_get_edid' static
Sachin Kamat [Mon, 21 Jan 2013 09:56:57 +0000 (15:26 +0530)]
drm/exynos: Make 'drm_hdmi_get_edid' static

Fixes the following warning:
drivers/gpu/drm/exynos/exynos_drm_hdmi.c:111:13: warning:
symbol 'drm_hdmi_get_edid' was not declared. Should it be static?

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: fimd and ipp are broken on multiplatform
Arnd Bergmann [Mon, 21 Jan 2013 17:16:01 +0000 (17:16 +0000)]
drm/exynos: fimd and ipp are broken on multiplatform

While the exynos DRM support in principle can work on
multiplatform, the FIMD and IPP sections of it both
include the plat/map-base.h header file, which is
not available on multiplatform. Rather than disabling
the entire driver, we can just conditionally build
these two parts.

Without this patch, building allyesconfig results in:

drivers/gpu/drm/exynos/exynos_drm_fimc.c:19:27: fatal error: plat/map-base.h: No such file or directory
drivers/gpu/drm/exynos/exynos_drm_ipp.c:20:27: fatal error: plat/map-base.h: No such file or directory

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
11 years agodrm/exynos: don't include plat/gpio-cfg.h
Arnd Bergmann [Mon, 21 Jan 2013 17:16:00 +0000 (17:16 +0000)]
drm/exynos: don't include plat/gpio-cfg.h

Patch  9eb3e9e6f3 "drm/exynos: add support for ARCH_MULTIPLATFORM"
allowed building the exynos hdmi driver on non-samsung platforms,
which unfortunately broke compilation in combination with 22c4f42897
"drm: exynos: hdmi: add support for exynos5 hdmi", which added
an inclusion of the samsung-specific plat/gpio-cfg.h header file.

Fortunately, that header file is not required any more here, so
we can simply revert the inclusion in order to build the ARM
allyesconfig again without getting this error:

drivers/gpu/drm/exynos/exynos_hdmi.c:37:27: fatal error: plat/gpio-cfg.h: No such file or directory

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
11 years agodrm/exynos: Remove "internal" interrupt handling
Sean Paul [Wed, 16 Jan 2013 15:17:20 +0000 (10:17 -0500)]
drm/exynos: Remove "internal" interrupt handling

Remove the "internal" interrupt handling since it's never invoked and
remove "external" reference. This patch removes a bunch of dead code
and clarifies how hotplugging is handled in the HDMI driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: Add missing static specifiers in exynos_drm_rotator.c
Sachin Kamat [Mon, 14 Jan 2013 06:59:11 +0000 (12:29 +0530)]
drm/exynos: Add missing static specifiers in exynos_drm_rotator.c

Fixes the following warnings:
drivers/gpu/drm/exynos/exynos_drm_rotator.c:737:24: warning:
symbol 'rot_limit_tbl' was not declared. Should it be static?
drivers/gpu/drm/exynos/exynos_drm_rotator.c:754:27: warning:
symbol 'rotator_driver_ids' was not declared. Should it be static?

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: Replace mdelay with usleep_range
Sean Paul [Mon, 14 Jan 2013 22:03:20 +0000 (17:03 -0500)]
drm/exynos: Replace mdelay with usleep_range

Replace the unnecessary atomic mdelay calls with usleep_range calls.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: Make ipp_handle_cmd_work static
Sachin Kamat [Mon, 14 Jan 2013 06:59:10 +0000 (12:29 +0530)]
drm/exynos: Make ipp_handle_cmd_work static

Fixes the following warning:
drivers/gpu/drm/exynos/exynos_drm_ipp.c:872:6: warning:
symbol 'ipp_handle_cmd_work' was not declared. Should it be static?

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: Make g2d_userptr_get_dma_addr static
Sachin Kamat [Mon, 14 Jan 2013 06:59:09 +0000 (12:29 +0530)]
drm/exynos: Make g2d_userptr_get_dma_addr static

Fixes the following warning:
drivers/gpu/drm/exynos/exynos_drm_g2d.c:327:12: warning:
symbol 'g2d_userptr_get_dma_addr' was not declared. Should it be static?

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: consider DMA_NONE flag to dmabuf import
Inki Dae [Fri, 11 Jan 2013 04:46:58 +0000 (13:46 +0900)]
drm/exynos: consider DMA_NONE flag to dmabuf import

This patch considers DMA_NONE flag for other drivers not using
dma mapping framework with iommu such as 3d gpu driver or others.

For example, there might be 3d gpu driver that has its own iommu
hw unit and iommu table mapping mechnism. So in this case,
the dmabuf buffer imported into this driver needs just only
sg table to map the buffer with its own iommu table itself.

So this patch makes dma_buf_map_attachment ignore dma_map_sg call
and just return sg table containing pages if dma_data_direction
is DMA_NONE.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
11 years agodrm/exynos: free sg object if dma_map_sg is failed
Inki Dae [Fri, 11 Jan 2013 04:38:28 +0000 (13:38 +0900)]
drm/exynos: free sg object if dma_map_sg is failed

This patch releases sgt's sg object allocated by sgt_alloc_table
correctly.

When exynos_gem_map_dma_buf was called by dma_buf_map_attachmemt(),
the sgt's sg object was allocated by sg_alloc_tale() so
if dma_map_sg() is failed, the sg object should be released.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
11 years agodrm/exynos: added validation of edid for vidi connection
Seung-Woo Kim [Thu, 10 Jan 2013 10:35:06 +0000 (19:35 +0900)]
drm/exynos: added validation of edid for vidi connection

If edid of vidi from user is invalid, size calculated from a number
of cea extensions can be wrong. So, validation should be checked.

Changelog v2:
- just code cleanup
  . declare raw_edid only if vidi->connection is enabled.

Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agodrm/exynos: let drm handle edid allocations
Rahul Sharma [Fri, 4 Jan 2013 12:59:11 +0000 (07:59 -0500)]
drm/exynos: let drm handle edid allocations

There's no need to allocate edid twice and do a memcpy when drm helpers
exist to do just that. This patch cleans that interaction up, and
doesn't keep the edid hanging around in the connector.

v4:
- removed error check for drm_mode_connector_update_edid_property
which is expected to fail for Virtual Connectors like VIDI.
Thanks to Seung-Woo Kim.

v3:
- removed MAX_EDID as it is not used anymore.

v2:
- changed vidi_get_edid callback inside vidi driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Rahul Sharma <rahul.sharma@samsung.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
11 years agoMerge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 25 Jan 2013 04:18:32 +0000 (14:18 +1000)]
Merge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm-next

Alex writes:
Just some small misc fixes.

* 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: Enable DMA_IB_SWAP_ENABLE on big endian hosts.
  drm/radeon: fix a rare case of double kfree
  radeon_display: Use pointer return error codes
  drm/radeon: fix cursor corruption on DCE6 and newer

11 years agodrm/radeon: Enable DMA_IB_SWAP_ENABLE on big endian hosts.
Michel Dänzer [Thu, 24 Jan 2013 18:02:01 +0000 (19:02 +0100)]
drm/radeon: Enable DMA_IB_SWAP_ENABLE on big endian hosts.

Fixes GPU hang during DMA ring IB test.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=59672

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix a rare case of double kfree
Ilija Hadzic [Wed, 23 Jan 2013 18:59:05 +0000 (13:59 -0500)]
drm/radeon: fix a rare case of double kfree

If one (but not both) allocations of p->chunks[].kpage[]
in radeon_cs_parser_init fail, the error path will free
the successfully allocated page, but leave a stale pointer
value in the kpage[] field. This will later cause a
double-free when radeon_cs_parser_fini is called.
This patch fixes the issue by forcing both pointers to NULL
after kfree in the error path.

The circumstances under which the problem happens are very
rare. The card must be AGP and the system must run out of
kmalloc area just at the right time so that one allocation
succeeds, while the other fails.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agoradeon_display: Use pointer return error codes
xueminsu [Tue, 22 Jan 2013 14:16:53 +0000 (22:16 +0800)]
radeon_display: Use pointer return error codes

drm_mode_addfb() expects fb_create return error code
instead of NULL.

Signed-off-by: xueminsu <xuemin.su@intel.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: fix cursor corruption on DCE6 and newer
Jerome Glisse [Mon, 21 Jan 2013 20:50:03 +0000 (15:50 -0500)]
drm/radeon: fix cursor corruption on DCE6 and newer

Aruba and newer gpu does not need the avivo cursor work around,
quite the opposite this work around lead to corruption.

agd5f: check DCE6 rather than ARUBA since the issue is DCE
version specific rather than family specific.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/i915: dump UTS_RELEASE into the error_state
Daniel Vetter [Wed, 23 Jan 2013 15:16:35 +0000 (16:16 +0100)]
drm/i915: dump UTS_RELEASE into the error_state

Useful for statistics or on overflowing bug reports to keep things all
lined up.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agoiommu/intel: disable DMAR for g4x integrated gfx
Daniel Vetter [Sun, 20 Jan 2013 22:50:13 +0000 (23:50 +0100)]
iommu/intel: disable DMAR for g4x integrated gfx

DMAR support on g4x/gm45 integrated gpus seems to be totally busted.
So don't bother, but instead disable it by default to allow distros to
unconditionally enable DMAR support.

v2: Actually wire up the right quirk entry, spotted by Adam Jackson.

Note that according to intel marketing materials only g45 and gm45
support DMAR/VT-d. So we have reports for all relevant gen4 pci ids by
now. Still, keep all the other gen4 ids in the quirk table in case the
marketing stuff confused me again, which would not be the first time.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=51921
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=538163
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=538163
Cc: Adam Jackson <ajax@redhat.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: stable@vger.kernel.org
Acked-By: David Woodhouse <David.Woodhouse@intel.com>
Tested-by: stathis <stathis@npcglib.org>
Tested-by: Mihai Moldovan <ionic@ionic.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: GFX_MODE Flush TLB Invalidate Mode must be '1' for scanline waits
Chris Wilson [Sun, 20 Jan 2013 16:33:32 +0000 (16:33 +0000)]
drm/i915: GFX_MODE Flush TLB Invalidate Mode must be '1' for scanline waits

On SNB, if bit 13 of GFX_MODE, Flush TLB Invalidate Mode, is not set to 1,
the hardware can not program the scanline values. Those scanline values
then control when the signal is sent from the display engine to the render
ring for MI_WAIT_FOR_EVENTs. Note setting this bit means that TLB
invalidations must be performed explicitly through the appropriate bits
being set in PIPE_CONTROL.

References: https://bugzilla.kernel.org/show_bug.cgi?id=52311
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Disable AsyncFlip performance optimisations
Chris Wilson [Sun, 20 Jan 2013 16:11:20 +0000 (16:11 +0000)]
drm/i915: Disable AsyncFlip performance optimisations

This is a required workarounds for all products, especially on gen6+
where it causes the command streamer to fail to parse instructions
following a WAIT_FOR_EVENT. We use WAIT_FOR_EVENT for synchronising
between the GPU and the display engines, and so this bit being unset may
cause hangs.

References: https://bugzilla.kernel.org/show_bug.cgi?id=52311
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agottm: on move memory failure don't leave a node dangling
Dave Airlie [Wed, 16 Jan 2013 05:58:34 +0000 (15:58 +1000)]
ttm: on move memory failure don't leave a node dangling

if we have a move notify callback, when moving fails, we call move notify
the opposite way around, however this ends up with *mem containing the mm_node
from the bo, which means we double free it. This is a follow on to the previous
fix.

Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agottm: don't destroy old mm_node on memcpy failure
Dave Airlie [Wed, 16 Jan 2013 04:25:44 +0000 (14:25 +1000)]
ttm: don't destroy old mm_node on memcpy failure

When we are using memcpy to move objects around, and we fail to memcpy
due to lack of memory to populate or failure to finish the copy, we don't
want to destroy the mm_node that has been copied into old_copy.

While working on a new kms driver that uses memcpy, if I overallocated bo's
up to the memory limits, and eviction failed, then machine would oops soon
after due to having an active bo with an already freed drm_mm embedded in it,
freeing it a second time didn't end well.

Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel...
Dave Airlie [Mon, 21 Jan 2013 03:25:30 +0000 (13:25 +1000)]
Merge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

More important fixes for 3.9:
- error_state improvements to help debug the new scanline wait code added
  for gen6+ - bug reports started popping up :( patch from Chris Wilson.
- fix a panel power sequence confusion between the eDP and lvds detection
  code resulting in black screens - regression introduce in 3.8 (Jani
  Nikula)
- Chris fixed the root-cause of the ilk relocation vs. evict bug.
- Another piece of cargo-culted rc6 lore from Jani, fixes up a regression
  where a system refused to go into rc6 after suspend sometimes.

* 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel:
  drm/i915: fix FORCEWAKE posting reads
  drm/i915: Invalidate the relocation presumed_offsets along the slow path
  drm/i915/eDP: do not write power sequence registers for ghost eDP
  drm/i915: Record DERRMR, FORCEWAKE and RING_CTL in error-state

11 years agoMerge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Mon, 21 Jan 2013 03:24:12 +0000 (13:24 +1000)]
Merge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm-next

A number of fixes, and one revert for a patch having some wierd side effects.

* 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux:
  Revert "drm/radeon: do not move bo to different placement at each cs"
  drm/radeon: improve semaphore debugging on lockup
  drm/radeon: allow FP16 color clear registers on r500
  drm/radeon: clear reset flags if engines are idle

11 years agoRevert "drm/radeon: do not move bo to different placement at each cs"
Alex Deucher [Thu, 17 Jan 2013 18:10:50 +0000 (13:10 -0500)]
Revert "drm/radeon: do not move bo to different placement at each cs"

This reverts commit d025e9e2b890db679f1246037bf65bd4be512627.

This causes corruption for a number of users and needs further
investigation in the next cycle.
https://bugzilla.kernel.org/show_bug.cgi?id=52491
https://bugs.freedesktop.org/show_bug.cgi?id=58659
http://lists.freedesktop.org/archives/dri-devel/2013-January/032961.html

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/i915: fix FORCEWAKE posting reads
Jani Nikula [Thu, 17 Jan 2013 08:24:09 +0000 (10:24 +0200)]
drm/i915: fix FORCEWAKE posting reads

We stopped reading FORCEWAKE for posting reads in

commit 8dee3eea3ccd3b6c00a8d3a08dd715d6adf737dd
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Sat Sep 1 22:59:50 2012 -0700

    drm/i915: Never read FORCEWAKE

and started using something from the same cacheline instead. On the
bug reporter's machine this broke entering rc6 states after a
suspend/resume cycle. It turns out reading ECOBUS as posting read
worked fine, while GTFIFODBG did not, preventing RC6 states after
suspend/resume per the bug report referenced below. It's not entirely
clear why, but clearly GTFIFODBG was nowhere near the same cacheline
or address range as FORCEWAKE.

Trying out various registers for posting reads showed that all tested
registers for which NEEDS_FORCE_WAKE() (in i915_drv.c) returns true
work. Conversely, most (but not quite all) registers for which
NEEDS_FORCE_WAKE() returns false do not work. Details in the referenced
bug.

Based on the above, add posting reads on ECOBUS where GTFIFODBG was
previously relied on.

In true cargo cult spirit, add posting reads for FORCEWAKE_VLV writes as
well, but instead of ECOBUS, use FORCEWAKE_ACK_VLV which is in the same
address range as FORCEWAKE_VLV.

v2: Add more details to the commit message. No functional changes.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=52411
Reported-and-tested-by: Alexander Bersenev <bay@hackerdom.ru>
CC: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org
[danvet: add cc: stable and make the commit message a bit clearer that
this is a regression fix and what exactly broke.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915: Invalidate the relocation presumed_offsets along the slow path
Chris Wilson [Tue, 15 Jan 2013 16:17:54 +0000 (16:17 +0000)]
drm/i915: Invalidate the relocation presumed_offsets along the slow path

In the slow path, we are forced to copy the relocations prior to
acquiring the struct mutex in order to handle pagefaults. We forgo
copying the new offsets back into the relocation entries in order to
prevent a recursive locking bug should we trigger a pagefault whilst
holding the mutex for the reservations of the execbuffer. Therefore, we
need to reset the presumed_offsets just in case the objects are rebound
back into their old locations after relocating for this exexbuffer - if
that were to happen we would assume the relocations were valid and leave
the actual pointers to the kernels dangling, instant hang.

Fixes regression from commit bcf50e2775bbc3101932d8e4ab8c7902aa4163b4
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Nov 21 22:07:12 2010 +0000

    drm/i915: Handle pagefaults in execbuffer user relocations

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=55984
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@fwll.ch>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/i915/eDP: do not write power sequence registers for ghost eDP
Jani Nikula [Wed, 16 Jan 2013 08:53:40 +0000 (10:53 +0200)]
drm/i915/eDP: do not write power sequence registers for ghost eDP

Some machines detect an eDP port even if it's not really there, and eDP
initialization has a fail path for this. Typically such machines have an
LVDS display instead. A regression introduced in

commit 82ed61fa1a4e08d5f9e86fb1b715b50ed678b6ac
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sat Oct 20 20:57:41 2012 +0200

    drm/i915: make edp panel power sequence setup more robust

updated the power sequence registers PCH_PP_ON_DELAYS, PCH_PP_OFF_DELAYS,
and PCH_PP_DIVISOR also in the ghost eDP case, messing up the LVDS display.

Split the power sequencer initialization into two, delaying the register
updates until after we know the eDP is real.

Note: Keep the PP_CONTROL unlocking in the first part, even if it does not
update registers, per the commit message of the above mentioned commit.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=52601
Reported-and-tested-by: Ryan Coe <ryan@rycomotorsports.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agodrm/radeon: improve semaphore debugging on lockup
Jerome Glisse [Fri, 11 Jan 2013 20:19:43 +0000 (15:19 -0500)]
drm/radeon: improve semaphore debugging on lockup

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: allow FP16 color clear registers on r500
Marek Olšák [Sat, 12 Jan 2013 03:19:37 +0000 (04:19 +0100)]
drm/radeon: allow FP16 color clear registers on r500

Probably not a candidate for stable kernels because of conflicts
in DRM versioning.

Signed-off-by: Marek Olšák <maraeo@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: clear reset flags if engines are idle
Alex Deucher [Mon, 14 Jan 2013 16:04:39 +0000 (11:04 -0500)]
drm/radeon: clear reset flags if engines are idle

Fixes a hard lock in the gpu reset code after the
rework for DMA support (0ecebb9e0d14e9948e0b1529883a776758117d6f
"drm/radeon: switch to a finer grained reset for evergreen")
due to not bailing before the MC shutdown if the relevant engines
are idle.

Discussion:
http://lists.freedesktop.org/archives/dri-devel/2013-January/032985.html

Reported-by: Eldad Zack <eldad@fogrefinery.com>
Tested-by: Eldad Zack <eldad@fogrefinery.com>
Acked-by: Paul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/i915: Record DERRMR, FORCEWAKE and RING_CTL in error-state
Chris Wilson [Tue, 15 Jan 2013 12:05:55 +0000 (12:05 +0000)]
drm/i915: Record DERRMR, FORCEWAKE and RING_CTL in error-state

These are useful for investigating hangs involving WAIT_FOR_EVENT.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Apply a droplet of Future-Proof in the if-ladder.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
11 years agoudldrmfb: udl_get_edid: drop unneeded i--
Hans de Goede [Fri, 11 Jan 2013 11:08:58 +0000 (12:08 +0100)]
udldrmfb: udl_get_edid: drop unneeded i--

This is a left-over from when udl_get_edid returned the amount of bytes
successfully read, which it no longer does.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoudldrmfb: udl_get_edid: usb_control_msg buffer must not be on the stack
Hans de Goede [Fri, 11 Jan 2013 11:08:57 +0000 (12:08 +0100)]
udldrmfb: udl_get_edid: usb_control_msg buffer must not be on the stack

The buffer passed to usb_control_msg may end up in scatter-gather list, and
may thus not be on the stack. Having it on the stack usually works on x86, but
not on other archs.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoudldrmfb: Fix EDID not working with monitors with EDID extension blocks
Hans de Goede [Fri, 11 Jan 2013 11:08:56 +0000 (12:08 +0100)]
udldrmfb: Fix EDID not working with monitors with EDID extension blocks

udldrmfb only reads the main EDID block, and if that advertises extensions
the drm_edid code expects them to be present, and starts reading beyond the
buffer udldrmfb passes it.

Although it may be possible to read more EDID info with the udl we simpy don't
know how, and even if trial and error gets it working on one device, that is
no guarantee it will work on other revisions. So this patch does a simple fix
in the form of patching the EDID info to report 0 extension blocks, this
fixes udldrmfb only doing 1024x768 on monitors with EDID extension blocks.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
11 years agoMerge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Sun, 13 Jan 2013 22:16:54 +0000 (08:16 +1000)]
Merge branch 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux into drm-next

Fixes for UMS mode which has been broken for a while plus an rn50 fix
and a dma fix.

* 'drm-fixes-3.8' of git://people.freedesktop.org/~agd5f/linux:
  radeon/kms: fix dma relocation checking
  radeon/kms: force rn50 chip to always report connected on analog output
  drm/radeon: fix error path in kpage allocation
  drm/radeon: fix a bogus kfree
  drm/radeon: fix NULL pointer dereference in UMS mode

11 years agoMerge branch 'drm-nouveau-fixes' of git://anongit.freedesktop.org/git/nouveau/linux...
Dave Airlie [Sun, 13 Jan 2013 22:15:36 +0000 (08:15 +1000)]
Merge branch 'drm-nouveau-fixes' of git://anongit.freedesktop.org/git/nouveau/linux-2.6 into drm-next

Regression fixes since rework mostly.

* 'drm-nouveau-fixes' of git://anongit.freedesktop.org/git/nouveau/linux-2.6:
  drm/nvc0/fb: fix crash when different mutex is used to protect same list
  drm/nouveau/clock: fix support for more than 2 monitors on nve0
  drm/nv50/disp: fix selection of bios script for analog outputs
  drm/nv17-50: restore fence buffer on resume
  drm/nouveau: fix blank LVDS screen regression on pre-nv50 cards
  drm/nouveau: fix nouveau_client allocation failure path
  drm/nouveau: don't return freed object from nouveau_handle_create
  drm/nouveau/vm: fix memory corruption when pgt allocation fails
  drm/nouveau: add locking around instobj list operations
  drm/nouveau: do not forcibly power on lvds panels
  drm/nouveau/devinit: ensure legacy vga control is enabled during post

11 years agodrm/nvc0/fb: fix crash when different mutex is used to protect same list
Aleksi Torhamo [Wed, 9 Jan 2013 18:08:48 +0000 (20:08 +0200)]
drm/nvc0/fb: fix crash when different mutex is used to protect same list

Fixes regression introduced in commit 861d2107
"drm/nouveau/fb: merge fb/vram and port to subdev interfaces"

nv50_fb_vram_{new,del} functions were changed to use
nouveau_subdev->mutex instead of the old nouveau_mm->mutex.
nvc0_fb_vram_new still uses the nouveau_mm->mutex, but nvc0 doesn't
have its own fb_vram_del function, using nv50_fb_vram_del instead.
Because of this, on nvc0 a different mutex ends up being used to protect
additions and deletions to the same list.

This patch is a -stable candidate for 3.7.

Signed-off-by: Aleksi Torhamo <aleksi@torhamo.net>
Reported-by: Roy Spliet <r.spliet@student.tudelft.nl>
Tested-by: Roy Spliet <r.spliet@student.tudelft.nl>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
11 years agodrm/nouveau/clock: fix support for more than 2 monitors on nve0
Aleksi Torhamo [Fri, 4 Jan 2013 16:39:13 +0000 (18:39 +0200)]
drm/nouveau/clock: fix support for more than 2 monitors on nve0

Fixes regression introduced in commit 70790f4f
"drm/nouveau/clock: pull in the implementation from all over the place"

When code was moved from nv50_crtc_set_clock to nvc0_clock_pll_set,
the PLLs it is used for got limited to only the first two VPLLs.

nv50_crtc_set_clock was only called to change VPLLs, so it didn't
limit what it was used for in any way. Since nvc0_clock_pll_set is
used for all PLLs, it has to specify which PLLs the code is used for,
and only listed the first two VPLLs.

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=58735
This patch is a -stable candidate for 3.7.

Signed-off-by: Aleksi Torhamo <aleksi@torhamo.net>
Tested-by: Aleksi Torhamo <aleksi@torhamo.net>
Tested-by: Sean Santos <quantheory@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
11 years agodrm/nv50/disp: fix selection of bios script for analog outputs
Marcin Slusarz [Thu, 3 Jan 2013 18:38:45 +0000 (19:38 +0100)]
drm/nv50/disp: fix selection of bios script for analog outputs

Analog output number was overwritten by value from digital output path.
Fix it.

Fixes resume from s2ram: https://bugs.freedesktop.org/show_bug.cgi?id=58729
(as stumbled on by J Binder, Pontus Fuchs and me)
Fixes blank screen on module load (reported by Sune Mølgaard).

Fixes regression from commit 186ecad21c854385823a430b1402053ae7fd59dc
("drm/nv50/disp: move remaining interrupt handling into core").

Reported-by: J Binder <wheel@herr-der-mails.de>
Reported-by: Pontus Fuchs <pontus.fuchs@gmail.com>
Reported-by: Sune Mølgaard <sune@molgaard.org>
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Tested-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Tested-by: Sune Mølgaard <sune@molgaard.org>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agodrm/nv17-50: restore fence buffer on resume
Marcin Slusarz [Tue, 25 Dec 2012 17:13:22 +0000 (18:13 +0100)]
drm/nv17-50: restore fence buffer on resume

Since commit 5e120f6e4b3f35b741c5445dfc755f50128c3c44 "drm/nouveau/fence:
convert to exec engine, and improve channel sync" nouveau fence sync
implementation for nv17-50 and nvc0+ started to rely on state of fence buffer
left by previous sync operation. But as pinned bo's (where fence state is
stored) are not saved+restored across suspend/resume, we need to do it
manually.

nvc0+ was fixed by commit d6ba6d215a538a58f0f0026f0961b0b9125e8042
"drm/nvc0/fence: restore pre-suspend fence buffer context on resume".

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=50121
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
11 years agodrm/nouveau: fix blank LVDS screen regression on pre-nv50 cards
Marcin Slusarz [Tue, 18 Dec 2012 19:30:47 +0000 (20:30 +0100)]
drm/nouveau: fix blank LVDS screen regression on pre-nv50 cards

Commit 2a44e499 ("drm/nouveau/disp: introduce proper init/fini, separate
from create/destroy") started to call display init routines on pre-nv50
hardware on module load. But LVDS init code sets driver state in a way
which prevents modesetting code from operating properly.

nv04_display_init calls nv04_dfp_restore, which sets encoder->last_dpms to
NV_DPMS_CLEARED.

drm_crtc_helper_set_mode
  nv04_dfp_prepare
    nv04_lvds_dpms(DRM_MODE_DPMS_OFF)

nv04_lvds_dpms checks last_dpms mode (which is NV_DPMS_CLEARED) and wrongly
assumes it's a "powersaving mode", the new one (DRM_MODE_DPMS_OFF) is too,
so it skips calling some crucial lvds scripts.

Reported-by: Chris Paulson-Ellis <chris@edesix.com>
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
11 years agodrm/nouveau: fix nouveau_client allocation failure path
Marcin Slusarz [Mon, 10 Dec 2012 20:59:52 +0000 (21:59 +0100)]
drm/nouveau: fix nouveau_client allocation failure path

Depending on the point of failure, freed object would be returned
or memory leak would happen.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agodrm/nouveau: don't return freed object from nouveau_handle_create
Marcin Slusarz [Mon, 10 Dec 2012 20:57:20 +0000 (21:57 +0100)]
drm/nouveau: don't return freed object from nouveau_handle_create

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agodrm/nouveau/vm: fix memory corruption when pgt allocation fails
Marcin Slusarz [Mon, 10 Dec 2012 20:30:51 +0000 (21:30 +0100)]
drm/nouveau/vm: fix memory corruption when pgt allocation fails

If we return freed vm, nouveau_drm_open will happily call nouveau_cli_destroy,
which will try to free it again.

Reported-by: Peter Hurley <peter@hurleysoftware.com>
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agodrm/nouveau: add locking around instobj list operations
Marcin Slusarz [Sun, 2 Dec 2012 11:56:22 +0000 (12:56 +0100)]
drm/nouveau: add locking around instobj list operations

Fixes memory corruptions, oopses, etc. when multiple gpuobjs are
simultaneously created or destroyed.

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: stable@vger.kernel.org
11 years agodrm/nouveau: do not forcibly power on lvds panels
Ben Skeggs [Mon, 10 Dec 2012 08:53:43 +0000 (18:53 +1000)]
drm/nouveau: do not forcibly power on lvds panels

This fix was put in place to fix a bug where the eDP panel on certain
laptops fails to respond over the aux channel after suspend.

It appears that on some systems (Dell M6600, with LVDS panel) there's a
very bad interaction with the eDP init table that causes the SOR to get
very confused and not drive the panel correctly, leading to bleed.

A DPMS off/on cycle is enough to bring it back, but, this will avoid the
problem by not touching the panel GPIOs at times we're not meant to.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agodrm/nouveau/devinit: ensure legacy vga control is enabled during post
Ben Skeggs [Sun, 13 Jan 2013 07:08:01 +0000 (17:08 +1000)]
drm/nouveau/devinit: ensure legacy vga control is enabled during post

Fixes ACPI backlight control after suspend on some systems.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
11 years agoradeon/kms: fix dma relocation checking
Jerome Glisse [Wed, 9 Jan 2013 21:40:42 +0000 (16:40 -0500)]
radeon/kms: fix dma relocation checking

We were checking the index against the size of the relocation buffer
instead of against the last index. This fix kernel segfault when
userspace submit ill formated command stream/relocation buffer pair.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoradeon/kms: force rn50 chip to always report connected on analog output
Jerome Glisse [Tue, 8 Jan 2013 23:41:01 +0000 (18:41 -0500)]
radeon/kms: force rn50 chip to always report connected on analog output

Those rn50 chip are often connected to console remoting hw and load
detection often fails with those. Just don't try to load detect and
report connect.

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
11 years agodrm/radeon: fix error path in kpage allocation
Ilija Hadzic [Mon, 7 Jan 2013 23:21:59 +0000 (18:21 -0500)]
drm/radeon: fix error path in kpage allocation

Index into chunks[] array doesn't look right.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix a bogus kfree
Ilija Hadzic [Mon, 7 Jan 2013 23:21:58 +0000 (18:21 -0500)]
drm/radeon: fix a bogus kfree

parser->chunks[.].kpage[.] is not always kmalloc-ed
by the parser initialization, so parser_fini should
not try to kfree it if it didn't allocate it.

This patch fixes a kernel oops that can be provoked
in UMS mode.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agodrm/radeon: fix NULL pointer dereference in UMS mode
Ilija Hadzic [Mon, 7 Jan 2013 23:21:57 +0000 (18:21 -0500)]
drm/radeon: fix NULL pointer dereference in UMS mode

In UMS mode parser->rdev is NULL, so dereferencing
will cause an oops.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
11 years agoMerge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel
Dave Airlie [Thu, 10 Jan 2013 21:47:25 +0000 (07:47 +1000)]
Merge branch 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel

Daniel writes:
"Pretty much all just major fixes:
- 2 pieces of duct-tape for the ilk bug.
- Sprite regression fixes from Chris.
- OOPS fix for a div-by-zero from Chris, regression due to the modeset
  rework in 3.7, now brought to light by a benign change in 3.8.
- Fix interrupted bo pinning, used to work around CS coherency issues on
  i830/i845 (kernel also has a w/a newly in 3.8, but pinning is more efficient if
  possible)."