GitHub/LineageOS/G12/android_kernel_amlogic_linux-4.9.git
10 years agodrm/msm: add support for non-IOMMU systems
Rob Clark [Sat, 16 Nov 2013 17:56:06 +0000 (12:56 -0500)]
drm/msm: add support for non-IOMMU systems

Add a VRAM carveout that is used for systems which do not have an IOMMU.

The VRAM carveout uses CMA.  The arch code must setup a CMA pool for the
device (preferrably in highmem.. a 256m-512m VRAM pool in lowmem is not
cool).  The user can configure the VRAM pool size using msm.vram module
param.

Technically, the abstraction of IOMMU behind msm_mmu is not strictly
needed, but it simplifies the GEM code a bit, and will be useful later
when I add support for a2xx devices with GPUMMU, so I decided to keep
this part.

It appears to be possible to configure the GPU to restrict access to
addresses within the VRAM pool, but this is not done yet.  So for now
the GPU will refuse to load if there is no sort of mmu.  Once address
based limits are supported and tested to confirm that we aren't giving
the GPU access to arbitrary memory, this restriction can be lifted

Signed-off-by: Rob Clark <robdclark@gmail.com>
10 years agodrm/msm: fix bus scaling
Rob Clark [Fri, 15 Nov 2013 14:03:15 +0000 (09:03 -0500)]
drm/msm: fix bus scaling

This got a bit broken with original patches when re-arranging things to
move dependencies on mach-msm inside #ifndef OF.

Signed-off-by: Rob Clark <robdclark@gmail.com>
10 years agodrm/msm: add missing MODULE_FIRMWARE()s
Rob Clark [Fri, 15 Nov 2013 14:03:48 +0000 (09:03 -0500)]
drm/msm: add missing MODULE_FIRMWARE()s

Signed-off-by: Rob Clark <robdclark@gmail.com>
10 years agodrm/msm: COMPILE_TEST support
Rob Clark [Wed, 27 Nov 2013 21:29:59 +0000 (16:29 -0500)]
drm/msm: COMPILE_TEST support

With a simple stub, we can get COMPILE_TEST support.

Signed-off-by: Rob Clark <robdclark@gmail.com>
10 years agoMerge tag 'drm-intel-next-2013-12-13' of git://people.freedesktop.org/~danvet/drm...
Dave Airlie [Mon, 23 Dec 2013 00:46:07 +0000 (10:46 +1000)]
Merge tag 'drm-intel-next-2013-12-13' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

- fbc1 improvements from Ville (pre-gm45).
- vlv forcewake improvements from Deepak S.
- Some corner-cases fixes from Mika for the context hang stat code.
- pc8 improvements and prep work for runtime D3 from Paulo, almost ready for
  primetime.
- gen2 dpll fixes from Ville.
- DSI improvements from Shobhit Kumar.
- A few smaller fixes and improvements all over.

[airlied: intel_ddi.c conflict fixed up]

* tag 'drm-intel-next-2013-12-13' of git://people.freedesktop.org/~danvet/drm-intel: (61 commits)
  drm/i915/bdw: Implement ff workarounds
  drm/i915/bdw: Force all Data Cache Data Port access to be Non-Coherent
  drm/i915/bdw: Don't use forcewake needlessly
  drm/i915: Clear out old GT FIFO errors in intel_uncore_early_sanitize()
  drm/i915: dont call irq_put when irq test is on
  drm/i915: Rework the FBC interval/stall stuff a bit
  drm/i915: Enable FBC for all mobile gen2 and gen3 platforms
  drm/i915: FBC_CONTROL2 is gen4 only
  drm/i915: Gen2 FBC1 CFB pitch wants 32B units
  drm/i915: split intel_ddi_pll_mode_set in 2 pieces
  drm/i915: Fix timeout with missed interrupts in __wait_seqno
  drm/i915: touch VGA MSR after we enable the power well
  drm/i915: extract hsw_power_well_post_{enable, disable}
  drm/i915: remove i915_disable_vga_mem declaration
  drm/i915: Parametrize the dphy and other spec specific parameters
  drm/i915: Remove redundant DSI PLL enabling
  drm/i915: Reorganize the DSI enable/disable sequence
  drm/i915: Try harder to get best m, n, p values with minimal error
  drm/i915: Compute dsi_clk from pixel clock
  drm/i915: Use FLISDSI interface for band gap reset
  ...

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

10 years agoMerge tag 'drm/for-3.14-rc1' of git://anongit.freedesktop.org/tegra/linux into drm...
Dave Airlie [Mon, 23 Dec 2013 00:43:42 +0000 (10:43 +1000)]
Merge tag 'drm/for-3.14-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/tegra: Changes for v3.14-rc1

This series of changes brings DRM panel support as well as initial code
to register DSI hosts and peripherals and bind them to DSI drivers. The
panel and DSI code are both used by the simple panel driver.

The Tegra-specific changes build on top of this work to add support for
various panels found on Tegra boards. New drivers enable the DSI host
found on Tegra114 and a special hardware block that calibrates the pads
used for DSI and CSI. The host1x and the display controller drivers gain
basic Tegra124 support. To round of the new features, the DRM driver now
sports a very simple PRIME implementation.

In addition there are various improvements such as the host1x API being
exported so that client drivers (like the Tegra DRM driver) can be built
as modules. HDMI now does better power management and legacy FBDEV can
now be disabled via Kconfig (though it's still enabled by default). A
few sparse warnings have been squashed and various parts of the code
have become more robust.

* tag 'drm/for-3.14-rc1' of git://anongit.freedesktop.org/tegra/linux: (121 commits)
  drm/tegra: fix compile w/ CONFIG_DYNAMIC_DEBUG
  drm/tegra: Add PRIME support
  drm/tegra: Relocate some output-specific code
  drm/tegra: Add Tegra124 DC support
  drm/tegra: Fix small leak on error in tegra_fb_alloc()
  drm/tegra: Make legacy fbdev support optional
  drm/tegra: Sort reverse-dependencies alphabetically
  drm/tegra: Fix return value check
  drm/tegra: Add DSI support
  drm/tegra: Disable outputs for power-saving
  drm/tegra: Track HDMI enable state
  drm/tegra: Fix HDMI audio frequency typo
  drm/tegra: Do not export tegra_bo_ops
  drm/tegra: Remove spurious blank line
  drm/tegra: Increase compile test coverage
  drm/tegra: Allow the driver to be built as a module
  gpu: host1x: Add Tegra124 support
  gpu: host1x: clk_round_rate() can return a zero upon error
  gpu: host1x: Fix build warnings
  gpu: host1x: Increase compile test coverage
  ...

10 years agoMerge branch 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev into drm-next
Dave Airlie [Mon, 23 Dec 2013 00:42:36 +0000 (10:42 +1000)]
Merge branch 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev into drm-next

rcar misc changes.

* 'drm/next/du' of git://linuxtv.org/pinchartl/fbdev:
  drm/rcar-du: Add support for the r8a7791 DU
  drm/rcar-du: Add LVDS_LANES quirk
  drm/rcar-du: Split features and quirks
  drm/rcar-du: Update plane pitch in .mode_set_base() operation
  drm/rcar-du: Don't cast crtc to rcrtc twice in the same function
  drm/rcar-du: fix return value check in rcar_du_lvdsenc_get_resources()

10 years agodrm/tegra: fix compile w/ CONFIG_DYNAMIC_DEBUG
Stephen Warren [Wed, 18 Dec 2013 20:53:33 +0000 (13:53 -0700)]
drm/tegra: fix compile w/ CONFIG_DYNAMIC_DEBUG

With CONFIG_DYNAMIC_DEBUG=y, the following compile error occurs:

drivers/gpu/drm/tegra/mipi-phy.c: In function ‘mipi_dphy_timing_validate’:
drivers/gpu/drm/tegra/mipi-phy.c:69:11: error: ‘EINVAL’ undeclared (first use in this function)
drivers/gpu/drm/tegra/mipi-phy.c:69:11: note: each undeclared identifier is reported only once for each function it appears in

Fix this by directly including the header that defines EINVAL.

Fixes: dec727399a4b ("drm/tegra: Add DSI support")
Signed-off-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Add PRIME support
Thierry Reding [Thu, 12 Dec 2013 09:00:43 +0000 (10:00 +0100)]
drm/tegra: Add PRIME support

Implement very basic PRIME support. This currently only works with
buffers that are contiguous in memory and will refuse to import any
physically non-contiguous buffers.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Relocate some output-specific code
Thierry Reding [Thu, 12 Dec 2013 10:06:55 +0000 (11:06 +0100)]
drm/tegra: Relocate some output-specific code

Some of the code in the CRTC's mode setting code is specific to the RGB
output or needs to be called slightly differently depending on the type
of output. Push that code down into the output drivers.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Add Tegra124 DC support
Thierry Reding [Thu, 12 Dec 2013 10:03:59 +0000 (11:03 +0100)]
drm/tegra: Add Tegra124 DC support

Tegra124 and later support interlacing, but the driver doesn't support
it yet. Make sure interlacing stays disabled on hardware that supports
it.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Fix small leak on error in tegra_fb_alloc()
Dan Carpenter [Tue, 12 Nov 2013 10:25:26 +0000 (13:25 +0300)]
drm/tegra: Fix small leak on error in tegra_fb_alloc()

If we don't have enough memory for ->planes then we leak "fb".

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Make legacy fbdev support optional
Thierry Reding [Thu, 31 Oct 2013 12:28:50 +0000 (13:28 +0100)]
drm/tegra: Make legacy fbdev support optional

A lot of the modern userspace is capable of working without the legacy
fbdev support. kmscon can be used as a replacement for the framebuffer
console, and KMS X drivers create their own framebuffers.

Most people don't have a system where all of this works yet, though, so
leave support enabled by default.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Sort reverse-dependencies alphabetically
Thierry Reding [Fri, 8 Nov 2013 10:56:59 +0000 (11:56 +0100)]
drm/tegra: Sort reverse-dependencies alphabetically

Move the TEGRA_HOST1X and DRM_KMS_HELPER entries around to keep the list
sorted.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Fix return value check
Wei Yongjun [Mon, 21 Oct 2013 03:34:07 +0000 (11:34 +0800)]
drm/tegra: Fix return value check

In case of error, the devm_ioremap_resource() function returns ERR_PTR()
and never NULL. The NULL test in the return value check should therefore
be replaced with IS_ERR().

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Add DSI support
Thierry Reding [Tue, 3 Sep 2013 06:45:46 +0000 (08:45 +0200)]
drm/tegra: Add DSI support

This commit adds support for both DSI outputs found on Tegra. Only very
minimal functionality is implemented, so advanced features like ganged
mode won't work.

Due to the lack of other test hardware, some sections of the driver are
hardcoded to work with Dalmore.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Disable outputs for power-saving
Thierry Reding [Tue, 29 Oct 2013 15:03:03 +0000 (16:03 +0100)]
drm/tegra: Disable outputs for power-saving

When an output is disabled, its DPMS mode is usually set to off. Instead
of only disabling the panel (if one is attached), turn the output off
entirely to save more power.

HDMI doesn't have any panels attached, so it previously didn't save any
power at all. With this commit, however, the complete HDMI interface
will be turned off, therefore allowing an attached monitor to go into a
standby mode.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Track HDMI enable state
Thierry Reding [Tue, 29 Oct 2013 15:00:42 +0000 (16:00 +0100)]
drm/tegra: Track HDMI enable state

The DRM core doesn't track enable and disable state of encoders and/or
connectors, so calls to the output's .enable() and .disable() are not
guaranteed to be balanced. Track the enable state internally so that
calls to regulator and clock frameworks remain balanced.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Fix HDMI audio frequency typo
Thierry Reding [Mon, 16 Dec 2013 09:01:24 +0000 (10:01 +0100)]
drm/tegra: Fix HDMI audio frequency typo

The correct check is for 48 kHz, not 480 kHz. Found by Coverity.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Do not export tegra_bo_ops
Thierry Reding [Thu, 12 Dec 2013 09:10:46 +0000 (10:10 +0100)]
drm/tegra: Do not export tegra_bo_ops

These buffer object operations are never used outside of the GEM
implementation so there is no use in exporting them.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Remove spurious blank line
Thierry Reding [Thu, 12 Dec 2013 09:09:46 +0000 (10:09 +0100)]
drm/tegra: Remove spurious blank line

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Increase compile test coverage
Thierry Reding [Wed, 20 Nov 2013 19:39:35 +0000 (20:39 +0100)]
drm/tegra: Increase compile test coverage

The ARCH_MULTIPLATFORM dependency was introduced back when Tegra didn't
support multiplatform yet as a means to allow the driver to be easily
compile-tested along with other DRM drivers. In the meantime, the new
COMPILE_TEST Kconfig option has been introduced for exactly that
purpose, so use that instead to clarify the intention.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Allow the driver to be built as a module
Thierry Reding [Fri, 8 Nov 2013 10:43:15 +0000 (11:43 +0100)]
drm/tegra: Allow the driver to be built as a module

All APIs that the driver uses are exported, so the driver can now be
built as a module.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Add Tegra124 support
Thierry Reding [Fri, 15 Nov 2013 13:58:05 +0000 (14:58 +0100)]
gpu: host1x: Add Tegra124 support

Tegra124 has 192 syncpoints whereas its predecessors had 32 syncpoints.
This required changes to the hardware register layout.

Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: clk_round_rate() can return a zero upon error
Paul Walmsley [Tue, 10 Dec 2013 02:00:12 +0000 (18:00 -0800)]
gpu: host1x: clk_round_rate() can return a zero upon error

Treat both negative and zero return values from clk_round_rate() as
errors.  This is needed since subsequent patches will convert
clk_round_rate()'s return value to be an unsigned type, rather than a
signed type, since some clock sources can generate rates higher than
(2^31)-1 Hz.

Eventually, when calling clk_round_rate(), only a return value of zero
will be considered a error.  All other values will be considered valid
rates.  The comparison against values less than 0 is kept to preserve
the correct behavior in the meantime.

Signed-off-by: Paul Walmsley <pwalmsley@nvidia.com>
Cc: Mikko Perttunen <mperttunen@nvidia.com>
Cc: Arto Merilainen <amerilainen@nvidia.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Terje Bergström <tbergstrom@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Fix build warnings
Thierry Reding [Tue, 3 Dec 2013 10:44:48 +0000 (11:44 +0100)]
gpu: host1x: Fix build warnings

When debugfs support isn't enabled, gcc complains about some variables
being unused. To avoid further #ifdefery, move debugfs specific setup
code into static functions and use IS_ENABLED(CONFIG_DEBUG_FS) to have
the compiler, rather than the preprocessor, discard them when unused.
The advantage of doing it this way is that all the code will be
compile-tested whether or not debugfs support is enabled.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Increase compile test coverage
Thierry Reding [Wed, 20 Nov 2013 19:58:36 +0000 (20:58 +0100)]
gpu: host1x: Increase compile test coverage

The ARCH_MULTIPLATFORM dependency was introduced back when Tegra didn't
support multiplatform yet as a means to allow the driver to be easily
compile-tested along with other DRM drivers. In the meantime, the new
COMPILE_TEST Kconfig option has been introduced for exactly that
purpose, so use that instead to clarify the intention.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Fix more sparse warnings
Thierry Reding [Fri, 8 Nov 2013 12:28:34 +0000 (13:28 +0100)]
gpu: host1x: Fix more sparse warnings

Include the linux/host1x.h and dev.h headers so that function prototypes
are visible to keep sparse from suggesting that their implementations be
made static.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Use the correct HW headers for host1x02
Thierry Reding [Fri, 8 Nov 2013 12:24:28 +0000 (13:24 +0100)]
gpu: host1x: Use the correct HW headers for host1x02

An earlier patch added a subset of the required HW specific header files
but didn't actually include the right ones when compiling for host1x02.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Export public API
Thierry Reding [Fri, 8 Nov 2013 10:41:42 +0000 (11:41 +0100)]
gpu: host1x: Export public API

Make the public API symbols visible so that depending drivers can be
built as a module.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Add MIPI pad calibration support
Thierry Reding [Mon, 2 Sep 2013 07:48:53 +0000 (09:48 +0200)]
gpu: host1x: Add MIPI pad calibration support

This driver adds support to perform calibration of the MIPI pads for CSI
and DSI.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/dp: Clarify automated test constant and add constant for FAUX test pattern
Todd Previte [Fri, 4 Oct 2013 19:59:03 +0000 (12:59 -0700)]
drm/dp: Clarify automated test constant and add constant for FAUX test pattern

  - DP_TEST_LINK_PATTERN is ambiguous, rename to DP_TEST_LINK_VIDEO_PATTERN to clarify
  - Added DP_TEST_LINK_FAUX_PATTERN to support automated testing of Fast AUX

Signed-off-by: Todd Previte <tprevite@gmail.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: use memdup_user() as a cleanup
Dan Carpenter [Wed, 11 Dec 2013 10:35:11 +0000 (11:35 +0100)]
drm: use memdup_user() as a cleanup

drivers/gpu/drm/r128/r128_state.c:1014:10-17: WARNING opportunity for memdup_user
/c/kernel-tests/src/cocci/drivers/gpu/drm/r128/r128_state.c:1029:9-16: WARNING opportunity for memdup_user
/c/kernel-tests/src/cocci/drivers/gpu/drm/r128/r128_state.c:904:10-17: WARNING opportunity for memdup_user
/c/kernel-tests/src/cocci/drivers/gpu/drm/r128/r128_state.c:914:9-16: WARNING opportunity for memdup_user

 Use memdup_user rather than duplicating its implementation
 This is a little bit restricted to reduce false positives

Generated by: coccinelle/api/memdup_user.cocci

CC: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: remove dev->vma_count
Daniel Vetter [Wed, 11 Dec 2013 10:35:10 +0000 (11:35 +0100)]
drm: remove dev->vma_count

This is just used for a debugfs file, and we can easily reconstruct
this number by just walking the list twice. Which isn't really bad for
a debugfs file anyway.

So let's rip this out.

There's the other issue that the dev->vmalist itself is a bit useless,
since that can be reconstructed with all the memory mapping
information from proc. But remove that is a different topic entirely.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill file_priv->ioctl_count tracking
Daniel Vetter [Wed, 11 Dec 2013 10:35:09 +0000 (11:35 +0100)]
drm: Kill file_priv->ioctl_count tracking

It's racy, and it's only used in debugfs. There are simpler ways to
know whether something is going on (like looking at dmesg with full
debugging enabled). And they're all much more useful.

So let's just rip this out.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: rip out dev->ioctl_count tracking
Daniel Vetter [Wed, 11 Dec 2013 10:35:08 +0000 (11:35 +0100)]
drm: rip out dev->ioctl_count tracking

Now dev->ioctl_count tries to prevent the device from disappearing if
it's still in use. And if we'd actually need this code it would be
hopelessly racy and broken.

But luckily the vfs already takes care of this. So we can just rip it
out.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/gma500: Remove dead code
Daniel Vetter [Wed, 11 Dec 2013 10:34:47 +0000 (11:34 +0100)]
drm/gma500: Remove dead code

This has the nice advantage that we'll get rid of a DRM_WAIT_ON user
for free.

Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Cc: Alan Cox <alan@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_SUSER
Daniel Vetter [Wed, 11 Dec 2013 10:34:46 +0000 (11:34 +0100)]
drm: Kill DRM_SUSER

Checking directly for the right capability is simpler. Also this rids
us of a few places that use DRM_CURRENTPID.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_*MEMORYBARRIER
Daniel Vetter [Wed, 11 Dec 2013 10:34:45 +0000 (11:34 +0100)]
drm: Kill DRM_*MEMORYBARRIER

The real linux interfaces are soooo much easier on the eyes ...

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_COPY_(TO|FROM)_USER
Daniel Vetter [Wed, 11 Dec 2013 10:34:44 +0000 (11:34 +0100)]
drm: Kill DRM_COPY_(TO|FROM)_USER

Less yelling ftw!

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_WAKUP and DRM_INIT_WAITQUEUE
Daniel Vetter [Wed, 11 Dec 2013 10:34:43 +0000 (11:34 +0100)]
drm: Kill DRM_WAKUP and DRM_INIT_WAITQUEUE

Less yelling ftw!

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_IRQ_ARGS
Daniel Vetter [Wed, 11 Dec 2013 10:34:42 +0000 (11:34 +0100)]
drm: Kill DRM_IRQ_ARGS

I've killed them a long time ago in drm/i915, let's get rid of this
remnant of shared drm core days for good.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Kill DRM_HZ
Daniel Vetter [Wed, 11 Dec 2013 10:34:41 +0000 (11:34 +0100)]
drm: Kill DRM_HZ

We don't have any userspace interfaces that use HZ as a time unit, so
having our own DRM define is useless.

Remove this remnant from the shared drm core days.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: rip out DRM_AGP_MEM and DRM_AGP_KERN
Daniel Vetter [Wed, 11 Dec 2013 10:34:40 +0000 (11:34 +0100)]
drm: rip out DRM_AGP_MEM and DRM_AGP_KERN

The <linux/agp_backend.h> header provides dummy functions and
fallbacks, so no need for screaming macros.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: remove global_mutex locking around agp_init
Daniel Vetter [Wed, 11 Dec 2013 10:34:39 +0000 (11:34 +0100)]
drm: remove global_mutex locking around agp_init

David Herrmann dutifully moved this locking along when moving the
agp_init call out of the generic drm_dev_register into the pci
specific load helpers.

But afaict there's no need and the reason for that locking has been
purely a historical accident - we need the lock around the driver dev
node registration to paper over the midlayer init races, and the agp
init simply ended up in there. The real fix for all this is of course
to delay the dev (and sysfs/debugfs) interface registration until
everything is fully set up.

Until then stop the cargo-cult locking from spreading and remove the
locking.

Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/agpsupport: use kzalloc instead of kmalloc/memset
Dave Airlie [Wed, 18 Dec 2013 01:27:06 +0000 (11:27 +1000)]
drm/agpsupport: use kzalloc instead of kmalloc/memset

Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: kill the ->agp_destroy callback
Daniel Vetter [Wed, 11 Dec 2013 10:34:38 +0000 (11:34 +0100)]
drm: kill the ->agp_destroy callback

Call drm_pci_agp_destroy directly, there's no point in the
indirection. Long term we want to shuffle this into each driver's
unload logic, but that needs cleared-up drm lifetime rules first.

v2: Add a dummy function for !CONFIG_PCI, spotted my David Herrmann.

v3: Fixup for the coding style police.

Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: inline drm_agp_destroy
Daniel Vetter [Wed, 11 Dec 2013 10:34:37 +0000 (11:34 +0100)]
drm: inline drm_agp_destroy

Wrapping a kfree is pointless.

v2: Add a comment to the kerneldoc for drm_agp_init to explain where
the kfree happens as requested by David. Note that for modeset drivers
agp cleanup is fairly complicated anyway: The drm_agp_clear is a noop
and drivers must call drm_agp_release on their own. Which they all
seem to do properly.

Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: remove agp_init() bus callback
Daniel Vetter [Wed, 11 Dec 2013 10:34:36 +0000 (11:34 +0100)]
drm: remove agp_init() bus callback

The PCI bus helper is the only user of it. Call it directly before
device-registration to get rid of the callback.

Note that all drm_agp_*() calls are locked with the drm-global-mutex so we
need to explicitly lock it during initialization. It's not really clear
why it's needed, but lets be safe.

v2: Rebase on top of the agp_init interface change.

v3: Remove the rebase-fail where I've accidentally killed the ->irq_by_busid
callback a bit too early.

Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: rip out drm_core_has_AGP
Daniel Vetter [Wed, 11 Dec 2013 10:34:35 +0000 (11:34 +0100)]
drm: rip out drm_core_has_AGP

Most place actually want to just check for dev->agp (most do, but a
few don't so this fixes a few potential NULL derefs). The only
exception is the agp init code which should check for the AGP driver
feature flag.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: ->agp_init can't fail
Daniel Vetter [Wed, 11 Dec 2013 10:34:34 +0000 (11:34 +0100)]
drm: ->agp_init can't fail

Thanks to the removal of REQUIRE_AGP we can use a void return value
and shed a bit of complexity.

Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: kill DRIVER_REQUIRE_AGP
Daniel Vetter [Wed, 11 Dec 2013 10:34:33 +0000 (11:34 +0100)]
drm: kill DRIVER_REQUIRE_AGP

Only the two intel drivers need this and they can easily check for
working agp support in their driver ->load callbacks.

This is the only reason why agp initialization could fail, so allows
us to rip out a bit of error handling code in the next patch.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/rcar-du: Fix return value check in rcar_du_lvdsenc_get_resources()
Wei Yongjun [Tue, 10 Dec 2013 02:27:19 +0000 (03:27 +0100)]
drm/rcar-du: Fix return value check in rcar_du_lvdsenc_get_resources()

In case of error, the function devm_ioremap_resource() returns ERR_PTR()
and never returns NULL. The NULL test in the return value check should be
replaced with IS_ERR(). Also remove the dev_err call to avoid redundant
error message.

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
[Remove the unneeded mem == NULL check]
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/dp: Use AUX constants from specification
Thierry Reding [Mon, 16 Dec 2013 16:01:29 +0000 (17:01 +0100)]
drm/dp: Use AUX constants from specification

The current values seem to be defined in a format that's specific to the
i915, gma500 and radeon drivers. To make this more generally useful, use
the values as defined in the specification.

While at it, prefix the constants with DP_ for improved namespacing.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/bufs: remove handling of _DRM_GEM mappings
Daniel Vetter [Wed, 11 Dec 2013 10:34:32 +0000 (11:34 +0100)]
drm/bufs: remove handling of _DRM_GEM mappings

Gone with the new gem vma offset manager from David.

We can also ditch the uapi header definition from the enum since
userspace never used this. It ended up in there purely for historical
reasons (for reusing the old drm mmap code essentially), not because
userspace ever needed it.

Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: restrict the device list for shadow attached drivers
Daniel Vetter [Wed, 11 Dec 2013 10:34:31 +0000 (11:34 +0100)]
drm: restrict the device list for shadow attached drivers

There's really no need for the drm core to keep a list of all
devices of a given driver - the linux device model keeps perfect
track of this already for us.

The exception is old legacy ums drivers using pci shadow attaching.
So rename the lists to make the use case clearer and rip out everything
else.

v2: Rebase on top of David Herrmann's drm device register changes.
Also drop the bogus dev_set_drvdata for platform drivers that somehow
crept into the original version - drivers really should be in full
control of that field.

v3: Initialize driver->legacy_dev_list outside of the loop, spotted by
David Herrmann.

v4: Rebase on top of the newly created host1x drm_bus for tegra.

Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: rip out drm_platform_exit
Daniel Vetter [Wed, 11 Dec 2013 10:34:30 +0000 (11:34 +0100)]
drm: rip out drm_platform_exit

This very much looks like a remnant of the old legady ums shadow
attach days. Now with the last users gone we can rip it out since
we won't ever support an ums drm driver again.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/msm: call drm_put_dev directly in ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:29 +0000 (11:34 +0100)]
drm/msm: call drm_put_dev directly in ->remove

The drvdata pointer is already assigned to something useful.

Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/armada: directly call drm_put_dev in ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:28 +0000 (11:34 +0100)]
drm/armada: directly call drm_put_dev in ->remove

Again no apparent user of the driver data field.

Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/shmob: call drm_put_dev directly from ->remove hook
Daniel Vetter [Wed, 11 Dec 2013 10:34:27 +0000 (11:34 +0100)]
drm/shmob: call drm_put_dev directly from ->remove hook

We need to chase one pointer here.

Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/omap: call drm_put_dev directly in ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:26 +0000 (11:34 +0100)]
drm/omap: call drm_put_dev directly in ->remove

Again omap already sets the driver data pointer to the drm_device.

Also drop the driver unregister call, that should be (and already is)
done in the module unload hook.

Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/tilcdc: call drm_put_dev directly from ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:25 +0000 (11:34 +0100)]
drm/tilcdc: call drm_put_dev directly from ->remove

tilcdc already stores the drm_device in the driver data pointer. So
use that.

Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/imx: directly call drm_put_dev in ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:24 +0000 (11:34 +0100)]
drm/imx: directly call drm_put_dev in ->remove

Again no apparent user of the driver data field.

Cc: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/exynos: call drm_put_dev directly from ->remove
Daniel Vetter [Wed, 11 Dec 2013 10:34:23 +0000 (11:34 +0100)]
drm/exynos: call drm_put_dev directly from ->remove

I didn't find any user of the driver data yet, so store the
drm_device pointer in there.

Cc: Inki Dae <inki.dae@samsung.com>
Acked-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/rcar: call drm_put_dev directly in the ->remove hook
Daniel Vetter [Wed, 11 Dec 2013 10:34:22 +0000 (11:34 +0100)]
drm/rcar: call drm_put_dev directly in the ->remove hook

The magic dance drm_platform_exit does is actually a remnant of the
old legacy shadow attach support for platform devices. Modern modesetting
drm drivers shouldn't do this any more (and usb/pci devices actually don't
do this).

Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: shmob_drm: Check clk_prepare_enable() return value
Laurent Pinchart [Mon, 2 Dec 2013 00:52:20 +0000 (01:52 +0100)]
drm: shmob_drm: Check clk_prepare_enable() return value

The clk_prepare_enable() call can fail. Check it's return value. We
can't propagate it all the way to the user as the KMS operations in
which the clock is enabled return a void.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Implement dummies for debugfs helpers
Thierry Reding [Thu, 12 Dec 2013 14:44:04 +0000 (15:44 +0100)]
drm: Implement dummies for debugfs helpers

In case where debugfs support is disabled, define dummy functions to
avoid the need for #ifdefery in drivers.

Based on an earlier patch by Arnd Bergmann.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agovideo/fb: Propagate error code from failing to unregister conflicting fb
Chris Wilson [Mon, 16 Dec 2013 15:57:39 +0000 (15:57 +0000)]
video/fb: Propagate error code from failing to unregister conflicting fb

If we fail to remove a conflicting fb driver, we need to abort the
loading of the second driver to avoid likely kernel panics.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: linux-fbdev@vger.kernel.org
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Push dirtyfb ioctl kms locking down to drivers
Ville Syrjälä [Wed, 4 Dec 2013 12:13:58 +0000 (14:13 +0200)]
drm: Push dirtyfb ioctl kms locking down to drivers

Not all drivers will need take all the modeset locks for dirtyfb, so
push the locking down to the drivers.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Don't reference objects in the flink name idr
Kristian Hogsberg [Tue, 3 Dec 2013 01:36:17 +0000 (17:36 -0800)]
drm: Don't reference objects in the flink name idr

There's no reason to keep a reference to objects in the name idr.  Each
handle to an object has a reference to the object and just before we
destroy the last handle we take the object out of the name idr.  Thus,
if an object is in the name idr, there's at least one reference to the
object.

Or to put it another way, the name idr reference will never keep the
object alive.  It just looks like it, which is confusing.

Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: do not steal the display if we have a master
Paulo Zanoni [Wed, 27 Nov 2013 20:24:08 +0000 (18:24 -0200)]
drm: do not steal the display if we have a master

Sometimes we want to disable all the screens on a system, because that
will allow the graphics card to be put into low-power states. The
problem is that, for example, while all screens are disabled, if we
get a hotplug interrupt, fbcon will decide to set a mode instead of
keeping everything disabled, which will remove us from our low power
states.

Let's assume that if there's a DRM master, it will be able to do
whatever is appropriate when we get the hotplug.

This problem can be reproduced by the runtime PM test program from
intel-gpu-tools: we disable all the screens so the graphics device can
be put into D3, then something triggers a hotplug interrupt, fbcon
sets a mode and breaks our test suite. The problem can be reproduced
more easily by the "i2c" subtest.

Other approaches considered for the problem:
    - Return "false" if "bound == 0" and the caller of
      drm_fb_helper_is_bound is a hotplug handler. This would break
      the case where the machine boots with no outputs connected, then
      the user plugs a monitor.
    - Add a new IOCTL to force fbcon to not set modes. This would keep
      all the current applications behaving the same, but adding a new
      IOCTL is not always the greatest idea.
    - Return false only if "dev->primary->master && bound == 0". This
      was my first implementation, but Chris suggested we should do
      the check irrespective of the "bound" variable.

Thanks to Daniel Vetter for the investigation, ideas and the
implementation of the hotplug alternative.

v2: - Do the check first, irrespective of "bound".
    - Cc dri-devel

Cc: dri-devel@lists.freedesktop.org
Credits-to: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: Don't split up debug output
Daniel Vetter [Sun, 17 Nov 2013 21:25:02 +0000 (22:25 +0100)]
drm: Don't split up debug output

Otherwise we risk that the 2nd part of the line ends up on a line of
it's own, which means a kernel dmesg line without a log level. This
then upsets the dmesg checker in piglit.

Only really happens in some of the truly nasty igt testcases which
race cache dropping (through debugfs) with other gem operations.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: add DRM_ERROR_RATELIMITED
Rob Clark [Thu, 21 Nov 2013 19:29:51 +0000 (14:29 -0500)]
drm: add DRM_ERROR_RATELIMITED

For error traces in situations that can run away, it is nice to have a
rate-limited version of DRM_ERROR() to avoid massive log flooding.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/edid: Make edid_load() return a void *
Geert Uytterhoeven [Tue, 19 Nov 2013 11:15:05 +0000 (12:15 +0100)]
drm/edid: Make edid_load() return a void *

Always use "void *" for arbitrary memory buffers, as this allows to drop
casts in assignments.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge tag 'drm-intel-next-2013-11-29' of git://people.freedesktop.org/~danvet/drm...
Dave Airlie [Wed, 18 Dec 2013 00:39:56 +0000 (10:39 +1000)]
Merge tag 'drm-intel-next-2013-11-29' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

- some more ppgtt prep patches from Ben
- a few fbc fixes from Ville
- power well rework from Imre
- vlv forcewake improvements from Deepak S, Ville and Jesse
- a few smaller things all over

[airlied: fixup forwcewake conflict]
* tag 'drm-intel-next-2013-11-29' of git://people.freedesktop.org/~danvet/drm-intel: (97 commits)
  drm/i915: Fix port name in vlv_wait_port_ready() timeout warning
  drm/i915: Return a drm_mode_status enum in the mode_valid vfuncs
  drm/i915: add intel_display_power_enabled_sw() for use in atomic ctx
  drm/i915: drop DRM_ERROR in intel_fbdev init
  drm/i915/vlv: use parallel context restore when coming out of RC6
  drm/i915/vlv: use a lower RC6 timeout on VLV
  drm/i915/sdvo: Fix up debug output to not split lines
  drm/i915: make sparse happy for the new vlv mmio read function
  drm/i915: drop the right force-wake engine in the vlv mmio funcs
  drm/i915: Fix GT wake FIFO free entries for VLV
  drm/i915: Report all GTFIFODBG errors
  drm/i915: Enabling DebugFS for valleyview forcewake counts
  drm/i915/vlv: Valleyview support for forcewake Individual power wells.
  drm/i915: Add power well arguments to force wake routines.
  drm/i915: Do not attempt to re-enable an unconnected primary plane
  drm/i915: add a debugfs entry for power domain info
  drm/i915: add a default always-on power well
  drm/i915: don't do BDW/HSW specific powerdomains init on other platforms
  drm/i915: protect HSW power well check with IS_HASWELL in redisable_vga
  drm/i915: use IS_HASWELL/BROADWELL instead of HAS_POWER_WELL
  ...

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

10 years agogpu: host1x: Add MIPI pad calibration DT bindings
Thierry Reding [Mon, 2 Dec 2013 13:22:58 +0000 (14:22 +0100)]
gpu: host1x: Add MIPI pad calibration DT bindings

Introduce device tree bindings for the MIPI pad calibration controller
found on Tegra SoCs. The controller can be used to perform calibration
of pads used for DSI and CSI peripherals.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Update host1x device tree example
Thierry Reding [Fri, 13 Dec 2013 14:27:42 +0000 (15:27 +0100)]
gpu: host1x: Update host1x device tree example

The display controller primary clock was recently renamed to "dc", so
update the example to reflect that.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Implement panel support
Thierry Reding [Fri, 30 Aug 2013 13:22:36 +0000 (15:22 +0200)]
drm/tegra: Implement panel support

Use the DRM panel framework to attach a panel to an output. If the panel
attached to a connector supports supports the backlight brightness
accessors, a property will be available to allow the brightness to be
modified from userspace.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: Add support for Panasonic VVX10F004B0
Thierry Reding [Fri, 22 Nov 2013 18:27:11 +0000 (19:27 +0100)]
drm/panel: Add support for Panasonic VVX10F004B0

The Panasonic VVX10F004B0 is a 10.1" WUXGA TFT LCD panel connected using
four DSI lanes.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: Add simple panel support
Thierry Reding [Fri, 30 Aug 2013 13:10:14 +0000 (15:10 +0200)]
drm/panel: Add simple panel support

Add a driver for simple panels. Such panels can have a regulator that
provides the supply voltage and a separate GPIO to enable the panel.
Optionally the panels can have a backlight associated with them so it
can be enabled or disabled according to the panel's power management
mode.

Support is added for two panels: An AU Optronics 10.1" WSVGA and a
Chunghwa Picture Tubes 10.1" WXGA panel.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Add panel support
Thierry Reding [Fri, 30 Aug 2013 11:36:43 +0000 (13:36 +0200)]
drm: Add panel support

Add a very simple framework to register and lookup panels. Panel drivers
can initialize a DRM panel and register it with the framework, allowing
them to be retrieved and used by display drivers. Currently only support
for DPMS and obtaining panel modes is provided. However it should be
sufficient to enable a large number of panels. The framework should also
be easily extensible to support more sophisticated kinds of panels such
as DSI.

The framework hasn't been tied into the DRM core, even though it should
be easily possible to do so if that's what we want. In the current
implementation, display drivers can simple make use of it to retrieve a
panel, obtain its modes and control its DPMS mode.

Note that this is currently only tested on systems that boot from a
device tree. No glue code has been written yet for systems that use
platform data, but it should be easy to add.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Add MIPI DSI bus support
Andrzej Hajda [Wed, 4 Dec 2013 15:35:12 +0000 (16:35 +0100)]
drm: Add MIPI DSI bus support

MIPI DSI bus allows to model DSI hosts and DSI peripherals using the
Linux driver model. DSI hosts are registered by the DSI host drivers.
During registration DSI peripherals will be created from the children
of the DSI host's device tree node. Support for registration from
board-setup code will be added later when needed.

DSI hosts expose operations which can be used by DSI peripheral drivers
to access associated devices.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agoof: Add simple panel device tree binding
Thierry Reding [Fri, 22 Nov 2013 18:19:55 +0000 (19:19 +0100)]
of: Add simple panel device tree binding

This binding specifies a set of common properties for display panels. It
can be used as a basis by bindings for specific panels.

Bindings for three specific panels are provided to show how the simple
panel binding can be used.

Acked-by: Rob Herring <rob.herring@calxeda.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agoof: Add MIPI DSI bus device tree bindings
Thierry Reding [Mon, 2 Dec 2013 15:22:49 +0000 (16:22 +0100)]
of: Add MIPI DSI bus device tree bindings

Document the device tree bindings for the MIPI DSI bus. The MIPI Display
Serial Interface specifies a serial bus and a protocol for communication
between a host and up to four peripherals.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agoMerge tag 'tegra-for-3.14-powergate' into drm/for-next
Thierry Reding [Tue, 17 Dec 2013 17:09:23 +0000 (18:09 +0100)]
Merge tag 'tegra-for-3.14-powergate' into drm/for-next

ARM: tegra: powergate driver changes

This branch includes all the changes to Tegra's powergate driver for 3.14.
These are separate out, since the Tegra DRM changes for 3.14 rely on the
new APIs introduced here.

A few cleanups and fixes are included, plus additions of Tegra124 SoC
support, and a new API for manipulating Tegra's IO rail deep power down
states.

This branch is based on tag tegra-for-3.14-dmas-resets-rework, in order
to avoid conflicts with the addition of common reset controller support
to the powergate driver.

10 years agoMerge tag 'tegra-for-3.14-dmas-resets-rework' into drm/for-next
Thierry Reding [Tue, 17 Dec 2013 17:09:16 +0000 (18:09 +0100)]
Merge tag 'tegra-for-3.14-dmas-resets-rework' into drm/for-next

ARM: tegra: implement common DMA and resets DT bindings

This series converts the Tegra DTs and drivers to use the common/
standard DMA and reset bindings, rather than custom bindings. It also
adds complete documentation for the Tegra clock bindings without
actually changing any binding definitions.

This conversion relies on a few sets of patches in branches from outside
the Tegra tree:

1) A patch to add an DMA channel request API which allows deferred probe
   to be implemented.

2) A patch to implement a common part of the of_xlate function for DMA
   controllers.

3) Some ASoC patches (which in turn rely on (1) above), which support
   deferred probe during DMA channel allocation.

4) The Tegra clock driver changes for 3.14.

Consequently, this branch is based on a merge of all of those external
branches.

In turn, this branch is or will be pulled into a few places that either
rely on features introduced here, or would otherwise conflict with the
patches:

a) Tegra's own for-3.14/powergate and for-4.14/dt branches, to avoid
   conflicts.

b) The DRM tree, which introduces new code that relies on the reset
   controller framework introduced in this branch, and to avoid
   conflicts.

10 years agoARM: tegra: Add IO rail support
Thierry Reding [Mon, 16 Dec 2013 20:42:28 +0000 (21:42 +0100)]
ARM: tegra: Add IO rail support

Add tegra_io_rail_power_off() and tegra_io_rail_power_on() functions to
put IO rails into or out of deep powerdown mode, respectively.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Special-case the 3D clamps on Tegra124
Thierry Reding [Fri, 13 Dec 2013 16:31:04 +0000 (17:31 +0100)]
ARM: tegra: Special-case the 3D clamps on Tegra124

A separate register is used to remove the clamps for the GPU on
Tegra124. In order to be able to use the same API, special-case
this particular partition.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Add Tegra124 powergate support
Thierry Reding [Fri, 13 Dec 2013 16:31:03 +0000 (17:31 +0100)]
ARM: tegra: Add Tegra124 powergate support

Three new gates have been added for Tegra124: SOR, VIC and IRAM. In
addition, PCIe and SATA gates are again supported, like on Tegra20 and
Tegra30.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Export tegra_powergate_remove_clamping()
Thierry Reding [Fri, 6 Dec 2013 15:49:56 +0000 (16:49 +0100)]
ARM: tegra: Export tegra_powergate_remove_clamping()

Drivers can use the tegra_powergate_remove_clamping() API during
initialization. In order to allow such drivers to be built as modules,
export the symbol.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Export tegra_powergate_power_off()
Thierry Reding [Fri, 6 Dec 2013 15:49:55 +0000 (16:49 +0100)]
ARM: tegra: Export tegra_powergate_power_off()

This function can be used by drivers, which in turn may be built as
modules. Export the symbol so it is available to modules.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Rename cpu0 powergate to crail
Thierry Reding [Fri, 6 Dec 2013 15:27:13 +0000 (16:27 +0100)]
ARM: tegra: Rename cpu0 powergate to crail

This matches the name of the powergate as listed in the TRM.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoARM: tegra: Fix some whitespace oddities
Thierry Reding [Fri, 6 Dec 2013 15:27:12 +0000 (16:27 +0100)]
ARM: tegra: Fix some whitespace oddities

Some of the powergate code uses unusual spacing around == and has a tab
instead of a space before an opening parenthesis.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agospi: tegra: checking for ERR_PTR instead of NULL
Dan Carpenter [Mon, 16 Dec 2013 14:02:10 +0000 (17:02 +0300)]
spi: tegra: checking for ERR_PTR instead of NULL

dma_request_slave_channel() returns NULL on error and not ERR_PTRs.
I've fixed this by using dma_request_slave_channel_reason() which does
return ERR_PTRs.

Fixes: a915d150f68d ('spi: tegra: convert to standard DMA DT bindings')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Stephen Warren <swarren@nvidia.com>
10 years agoLinux 3.13-rc4
Linus Torvalds [Sun, 15 Dec 2013 20:31:33 +0000 (12:31 -0800)]
Linux 3.13-rc4

10 years agonull_blk: mem garbage on NUMA systems during init
Matias Bjorling [Tue, 10 Dec 2013 15:50:38 +0000 (16:50 +0100)]
null_blk: mem garbage on NUMA systems during init

For NUMA systems, initializing the blk-mq layer and using per node hctx.
We initialize submit queues to 1, while blk-mq nr_hw_queues is
initialized to the number of NUMA nodes.

This makes the null_init_hctx function overwrite memory outside of what
it allocated.  In my case it lead to writing garbage into struct
request_queue's mq_map.

Signed-off-by: Matias Bjorling <m@bjorling.me>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoradeon_pm: fix oops in hwmon_attributes_visible() and radeon_hwmon_show_temp_thresh()
Sergey Senozhatsky [Thu, 12 Dec 2013 23:25:57 +0000 (02:25 +0300)]
radeon_pm: fix oops in hwmon_attributes_visible() and radeon_hwmon_show_temp_thresh()

Since commit ec39f64bba34 ("drm/radeon/dpm: Convert to use
devm_hwmon_register_with_groups") radeon_hwmon_init() is using
hwmon_device_register_with_groups(), which sets `rdev' as a device
private driver_data, while hwmon_attributes_visible() and
radeon_hwmon_show_temp_thresh() are still waiting for `drm_device'.

Fix them by using dev_get_drvdata(), in order to avoid this oops:

  BUG: unable to handle kernel paging request at 0000000000001e28
  IP: [<ffffffffa02ae8b4>] hwmon_attributes_visible+0x18/0x3d [radeon]
  PGD 15057e067 PUD 151a8e067 PMD 0
  Oops: 0000 [#1] PREEMPT SMP
  Call Trace:
    internal_create_group+0x114/0x1d9
    sysfs_create_group+0xe/0x10
    sysfs_create_groups+0x22/0x5f
    device_add+0x34f/0x501
    device_register+0x15/0x18
    hwmon_device_register_with_groups+0xb5/0xed
    radeon_hwmon_init+0x56/0x7c [radeon]
    radeon_pm_init+0x134/0x7e5 [radeon]
    radeon_modeset_init+0x75f/0x8ed [radeon]
    radeon_driver_load_kms+0xc6/0x187 [radeon]
    drm_dev_register+0xf9/0x1b4 [drm]
    drm_get_pci_dev+0x98/0x129 [drm]
    radeon_pci_probe+0xa3/0xac [radeon]
    pci_device_probe+0x6e/0xcf
    driver_probe_device+0x98/0x1c4
    __driver_attach+0x5c/0x7e
    bus_for_each_dev+0x7b/0x85
    driver_attach+0x19/0x1b
    bus_add_driver+0x104/0x1ce
    driver_register+0x89/0xc5
    __pci_register_driver+0x58/0x5b
    drm_pci_init+0x86/0xea [drm]
    radeon_init+0x97/0x1000 [radeon]
    do_one_initcall+0x7f/0x117
    load_module+0x1583/0x1bb4
    SyS_init_module+0xa0/0xaf

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Alexander Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 15 Dec 2013 19:56:47 +0000 (11:56 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Revert CHECKSUM_COMPLETE optimization in pskb_trim_rcsum(), I can't
    figure out why it breaks things.

 2) Fix comparison in netfilter ipset's hash_netnet4_data_equal(), it
    was basically doing "x == x", from Dave Jones.

 3) Freescale FEC driver was DMA mapping the wrong number of bytes, from
    Sebastian Siewior.

 4) Blackhole and prohibit routes in ipv6 were not doing the right thing
    because their ->input and ->output methods were not being assigned
    correctly.  Now they behave properly like their ipv4 counterparts.
    From Kamala R.

 5) Several drivers advertise the NETIF_F_FRAGLIST capability, but
    really do not support this feature and will send garbage packets if
    fed fraglist SKBs.  From Eric Dumazet.

 6) Fix long standing user triggerable BUG_ON over loopback in RDS
    protocol stack, from Venkat Venkatsubra.

 7) Several not so common code paths can potentially try to invoke
    packet scheduler actions that might be NULL without checking.  Shore
    things up by either 1) defining a method as mandatory and erroring
    on registration if that method is NULL 2) defininig a method as
    optional and the registration function hooks up a default
    implementation when NULL is seen.  From Jamal Hadi Salim.

 8) Fix fragment detection in xen-natback driver, from Paul Durrant.

 9) Kill dangling enter_memory_pressure method in cg_proto ops, from
    Eric W Biederman.

10) SKBs that traverse namespaces should have their local_df cleared,
    from Hannes Frederic Sowa.

11) IOCB file position is not being updated by macvtap_aio_read() and
    tun_chr_aio_read().  From Zhi Yong Wu.

12) Don't free virtio_net netdev before releasing all of the NAPI
    instances.  From Andrey Vagin.

13) Procfs entry leak in xt_hashlimit, from Sergey Popovich.

14) IPv6 routes that are no cached routes should not count against the
    garbage collection limits.  We had this almost right, but were
    missing handling addrconf generated routes properly.  From Hannes
    Frederic Sowa.

15) fib{4,6}_rule_suppress() have to consider potentially seeing NULL
    route info when they are called, from Stefan Tomanek.

16) TUN and MACVTAP have had truncated packet signalling for some time,
    fix from Jason Wang.

17) Fix use after frrr in __udp4_lib_rcv(), from Eric Dumazet.

18) xen-netback does not interpret the NAPI budget properly for TX work,
    fix from Paul Durrant.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (132 commits)
  igb: Fix for issue where values could be too high for udelay function.
  i40e: fix null dereference
  xen-netback: fix gso_prefix check
  net: make neigh_priv_len in struct net_device 16bit instead of 8bit
  drivers: net: cpsw: fix for cpsw crash when build as modules
  xen-netback: napi: don't prematurely request a tx event
  xen-netback: napi: fix abuse of budget
  sch_tbf: use do_div() for 64-bit divide
  udp: ipv4: must add synchronization in udp_sk_rx_dst_set()
  net:fec: remove duplicate lines in comment about errata ERR006358
  Revert "8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature"
  8390 : Replace ei_debug with msg_enable/NETIF_MSG_* feature
  xen-netback: make sure skb linear area covers checksum field
  net: smc91x: Fix device tree based configuration so it's usable
  udp: ipv4: fix potential use after free in udp_v4_early_demux()
  macvtap: signal truncated packets
  tun: unbreak truncated packet signalling
  net: sched: htb: fix the calculation of quantum
  net: sched: tbf: fix the calculation of max_size
  micrel: add support for KSZ8041RNLI
  ...

10 years agoMerge branch 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Linus Torvalds [Sun, 15 Dec 2013 19:52:47 +0000 (11:52 -0800)]
Merge branch 'x86/urgent' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Peter Anvin:
 "This is a pretty small batch:

  The biggest single change is to stop using EFI time services on 32-bit
  platforms.  This matches our current behavior on 64-bit platforms as
  we already had ruled them out there as being too unreliable.  Turns
  out that affects 32-bit platforms, too.

  One NULL pointer fix for SGI UV.

  Two minor build fixes, one of which only affects icc and the other
  which affects icc and future versions or nonstandard default settings
  of gcc"

* 'x86/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86, efi: Don't use (U)EFI time services on 32 bit
  x86, build, icc: Remove uninitialized_var() from compiler-intel.h
  x86/UV: Fix NULL pointer dereference in uv_flush_tlb_others() if the 'nobau' boot option is used
  x86, build: Pass in additional -mno-mmx, -mno-sse options

10 years agoMerge tag 'pci-v3.13-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Sun, 15 Dec 2013 19:45:27 +0000 (11:45 -0800)]
Merge tag 'pci-v3.13-fixes-2' of git://git./linux/kernel/git/helgaas/pci

Pull PCI updates from Bjorn Helgaas:
 "PCI device hotplug
    - Move device_del() from pci_stop_dev() to pci_destroy_dev() (Rafael
      Wysocki)

  Host bridge drivers
    - Update maintainers for DesignWare, i.MX6, Armada, R-Car (Bjorn
      Helgaas)
    - mvebu: Return 'unsupported' for Interrupt Line and Interrupt Pin
      (Jason Gunthorpe)

  Miscellaneous
    - Avoid unnecessary CPU switch when calling .probe() (Alexander
      Duyck)
    - Revert "workqueue: allow work_on_cpu() to be called recursively"
      (Bjorn Helgaas)
    - Disable Bus Master only on kexec reboot (Khalid Aziz)
    - Omit PCI ID macro strings to shorten quirk names for LTO (Michal
      Marek)"

* tag 'pci-v3.13-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  MAINTAINERS: Add DesignWare, i.MX6, Armada, R-Car PCI host maintainers
  PCI: Disable Bus Master only on kexec reboot
  PCI: mvebu: Return 'unsupported' for Interrupt Line and Interrupt Pin
  PCI: Omit PCI ID macro strings to shorten quirk names
  PCI: Move device_del() from pci_stop_dev() to pci_destroy_dev()
  Revert "workqueue: allow work_on_cpu() to be called recursively"
  PCI: Avoid unnecessary CPU switch when calling driver .probe() method