openwrt/staging/blogic.git
10 years agodrm/radeon: use the new drm helpers for dp aux
Alex Deucher [Fri, 21 Mar 2014 14:34:07 +0000 (10:34 -0400)]
drm/radeon: use the new drm helpers for dp aux

Switch to the new dp helpers.  The main difference is
that the DP helpers don't allow an adjustable delay in
the aux transaction, but I don't know that this is
necessary.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/dp: make aux retries less chatty
Alex Deucher [Fri, 21 Mar 2014 14:34:06 +0000 (10:34 -0400)]
drm/dp: make aux retries less chatty

Switch to debug only to avoid flooding the logs.
This mirrors the behavior in some other drivers.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/radeon: clarify special handling in i2c over aux
Alex Deucher [Tue, 18 Mar 2014 03:48:17 +0000 (23:48 -0400)]
drm/radeon: clarify special handling in i2c over aux

We need a special packet for the start and end of the
transaction.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/radeon/atom: rework encoder enable/disable sequence
Alex Deucher [Tue, 18 Mar 2014 03:48:16 +0000 (23:48 -0400)]
drm/radeon/atom: rework encoder enable/disable sequence

This more closely matches what the vbios does and also
adds a quirk for travis lvds displays and powers down
the sink on DP displays which saves some power and may
fix display issues in some cases.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/radeon/dp: move sink power control to a separate function
Alex Deucher [Tue, 18 Mar 2014 03:48:15 +0000 (23:48 -0400)]
drm/radeon/dp: move sink power control to a separate function

This will be used elsewhere.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/radeon/dp: use i2c_get_adapdata rather than casting
Alex Deucher [Tue, 18 Mar 2014 03:48:14 +0000 (23:48 -0400)]
drm/radeon/dp: use i2c_get_adapdata rather than casting

Minor code cleanup.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
10 years agodrm/bridge/ptn3460: fix modular build
Dave Airlie [Sun, 23 Mar 2014 23:21:15 +0000 (09:21 +1000)]
drm/bridge/ptn3460: fix modular build

This failed to build =m, quick fix.

Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daein...
Dave Airlie [Sun, 23 Mar 2014 22:58:20 +0000 (08:58 +1000)]
Merge branch 'exynos-drm-next' of git://git./linux/kernel/git/daeinki/drm-exynos into drm-next

Highlights
----------

Re-factoring works over the exynos drm framework.
  - drm_crtc, drm_encoder/drm_connector are implemented by sub drivers
    directly.
  - Removing pm interfaces from each sub driver, and implementing them
    at top level of exynos drm.
Add DisplayPort Transmitter driver.
  - Just moving existing driver from drivers/vides/exynos into
    drivers/gpu/drm/exynos.
Add new LVDS bridge driver, PTN3460.
  - Placed in drivers/gpu/drm/bridge, and this device is used to transfer
    image signal from DP(DisplayPort) to LVDS Panel.
    So this driver will be used with DP driver moved into exynos drm.
Add parallel panel support
  - With the re-factoring patch series, existing parallel panel support was
    broken by moving exynos_drm_display ops into each real connector driver,
    DP. So this patch series adds a new parallel panel module,
    exynos_drm_dpi, for supporting parallel panel, and also adds relevant
    bindings.
Some fixups and cleanups.

* 'exynos-drm-next' of git://git.kernel.org/pub/scm/linux/kernel/git/daeinki/drm-exynos: (45 commits)
  drm/exynos: fimd: remove unused variable
  ARM: dts: exynos4210-universal: add exynos/fimd node
  drm/exynos: restore parallel output interface support
  exynos/fimd: add parallel output related bindings
  drm/exynos: correct timing porch conversion
  drm/exynos: init kms poll after creation of connectors
  drm/exynos: delay fbdev initialization until an output is connected
  drm/exynos: fix unnecessary resource cleanup
  drm/exynos: hdmi: use i2c_adapter instead of i2c_client
  drm/exynos: hdmi: consider APB PHY
  drm/exynos: Remove the exynos_drm_connector shim
  drm/exynos: Implement lvds bridge discovery to DP driver
  drm/bridge: Add PTN3460 bridge driver
  drm/exynos: Implement drm_connector directly in vidi driver
  drm/exynos: Implement drm_connector directly in dp driver
  drm/exynos: Implement drm_connector in hdmi directly
  drm/exynos: Add create_connector callback
  drm/exynos: Consolidate suspend/resume in drm_drv
  drm/exynos: Clean up FIMD power on/off routines
  drm/exynos: Implement dpms display callback in DP
  ...

10 years agodrm/gma500: add locking to fixed panel edid probing
Daniel Vetter [Fri, 21 Mar 2014 22:22:36 +0000 (23:22 +0100)]
drm/gma500: add locking to fixed panel edid probing

With the recent addition of locking checks in

commit 62ff94a5492175759546f8bc61383189d6b49122
Author:     Daniel Vetter <daniel.vetter@ffwll.ch>
AuthorDate: Thu Jan 23 22:18:47 2014 +0100

    drm/crtc-helper: remove LOCKING from kerneldoc

drm_add_edid_modes started to WARN about the mode_config.mutex not
being held in the lvds and dp initialization code.

Now since this is init code locking is fairly redudant if it wouldn't
be for the drm core registering sysfs files a bit early. And the
locking WARNINGs nicely enforce that indeed all access to the mode
lists are properly protected. And a full audit shows that only i915
and gma500 touch the modes lists at init time.

Hence I've opted to wrap up this entire mode detection sequence for
fixed panels with the mode_config mutex for both lvds and edp outputs.

Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.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/i915: add locking to fixed panel edid probing
Daniel Vetter [Fri, 21 Mar 2014 22:22:35 +0000 (23:22 +0100)]
drm/i915: add locking to fixed panel edid probing

With the recent addition of locking checks in

commit 62ff94a5492175759546f8bc61383189d6b49122
Author:     Daniel Vetter <daniel.vetter@ffwll.ch>
AuthorDate: Thu Jan 23 22:18:47 2014 +0100

    drm/crtc-helper: remove LOCKING from kerneldoc

drm_add_edid_modes started to WARN about the mode_config.mutex not
being held in the lvds and dp initialization code.

Now since this is init code locking is fairly redudant if it wouldn't
be for the drm core registering sysfs files a bit early. And the
locking WARNINGs nicely enforce that indeed all access to the mode
lists are properly protected. And a full audit shows that only i915
and gma500 touch the modes lists at init time.

Hence I've opted to wrap up this entire mode detection sequence for
fixed panels with the mode_config mutex for both lvds and edp outputs.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/exynos: fimd: remove unused variable
Andrzej Hajda [Thu, 20 Mar 2014 08:09:00 +0000 (17:09 +0900)]
drm/exynos: fimd: remove unused variable

The patch removes unused vidcon0 field from fimd_context structure.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agoARM: dts: exynos4210-universal: add exynos/fimd node
Andrzej Hajda [Mon, 17 Mar 2014 10:27:22 +0000 (11:27 +0100)]
ARM: dts: exynos4210-universal: add exynos/fimd node

The patch adds fimd node with display timings for exynos4210-universal device.
It also makes LCD regulators always on. This allow to re-use panel initialized
by boot loader.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: restore parallel output interface support
Andrzej Hajda [Mon, 17 Mar 2014 12:03:56 +0000 (13:03 +0100)]
drm/exynos: restore parallel output interface support

The patch adds parallel output interface to FIMD device driver.
It also restores support for panels initialized by boot loader,
but without proper kernel driver.
Driver uses video interface bindings to find connected panel.
It uses drm_panel interface to interact with the panel.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agoexynos/fimd: add parallel output related bindings
Andrzej Hajda [Mon, 17 Mar 2014 10:27:20 +0000 (11:27 +0100)]
exynos/fimd: add parallel output related bindings

The patch adds bindings required to add support
for parallel output.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: correct timing porch conversion
Andrzej Hajda [Mon, 17 Mar 2014 10:27:19 +0000 (11:27 +0100)]
drm/exynos: correct timing porch conversion

The patch corrects porch calculation. It should
be calculated as a difference between adjacent
respective fields of drm_display_mode.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Daniel Kurtz <djkurtz@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: init kms poll after creation of connectors
Andrzej Hajda [Mon, 17 Mar 2014 10:27:18 +0000 (11:27 +0100)]
drm/exynos: init kms poll after creation of connectors

KMS poll init helper should be run when connectors are created,
otherwise it will not schedule connection detector.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: delay fbdev initialization until an output is connected
Andrzej Hajda [Mon, 17 Mar 2014 10:27:17 +0000 (11:27 +0100)]
drm/exynos: delay fbdev initialization until an output is connected

In case fbdev is initialized before any output is connected,
fb resolution defaults to 1024x768. After that any output with
bigger resolution is ignored and fbdev is not displayed.
The patch postpones fbdev initialization to avoid such situation.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: fix unnecessary resource cleanup
Inki Dae [Fri, 28 Feb 2014 09:37:02 +0000 (18:37 +0900)]
drm/exynos: fix unnecessary resource cleanup

This patch removes unnecessary drm_mode_config_cleanup call.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
10 years agodrm/exynos: hdmi: use i2c_adapter instead of i2c_client
Inki Dae [Thu, 13 Mar 2014 07:38:31 +0000 (16:38 +0900)]
drm/exynos: hdmi: use i2c_adapter instead of i2c_client

This patch changes i2c_client for ddc to i2c_adapter
because ddc needs only i2c_adapter.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
10 years agodrm/exynos: hdmi: consider APB PHY
Inki Dae [Thu, 6 Mar 2014 05:18:17 +0000 (14:18 +0900)]
drm/exynos: hdmi: consider APB PHY

This patch returns error in case of using APB PHY.

Exynos5420 SoC and maybe later would use APB PHY instead of
I2C PHY so such case should be considered.

Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
10 years agodrm/exynos: Remove the exynos_drm_connector shim
Sean Paul [Mon, 24 Feb 2014 10:25:42 +0000 (19:25 +0900)]
drm/exynos: Remove the exynos_drm_connector shim

This path removes the exynos_drm_connector code since it was just
passing hooks through display_ops. The individual device drivers are now
responsible for implementing drm_connector directly.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement lvds bridge discovery to DP driver
Sean Paul [Mon, 24 Feb 2014 10:20:15 +0000 (19:20 +0900)]
drm/exynos: Implement lvds bridge discovery to DP driver

This patch implements the lvds bridge discovery and connector pre-emption
code to the dp driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/bridge: Add PTN3460 bridge driver
Sean Paul [Mon, 24 Feb 2014 10:31:24 +0000 (19:31 +0900)]
drm/bridge: Add PTN3460 bridge driver

This patch adds a drm_bridge driver for the PTN3460 DisplayPort to LVDS
bridge chip.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement drm_connector directly in vidi driver
Sean Paul [Thu, 30 Jan 2014 21:38:07 +0000 (16:38 -0500)]
drm/exynos: Implement drm_connector directly in vidi driver

This patch implements drm_connector directly in the vidi
driver, this will allow us to move away from the exynos_drm_connector
layer.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement drm_connector directly in dp driver
Sean Paul [Thu, 30 Jan 2014 21:19:30 +0000 (16:19 -0500)]
drm/exynos: Implement drm_connector directly in dp driver

This patch implements drm_connector directly in the dp driver, this will
allow us to move away from the exynos_drm_connector layer.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement drm_connector in hdmi directly
Sean Paul [Thu, 30 Jan 2014 21:19:29 +0000 (16:19 -0500)]
drm/exynos: Implement drm_connector in hdmi directly

This patch implements drm_connector in the hdmi driver directly, instead
of using exynos_drm_connector.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Add create_connector callback
Sean Paul [Mon, 24 Feb 2014 10:15:38 +0000 (19:15 +0900)]
drm/exynos: Add create_connector callback

This creates a new display hook called create_connector. The purpose is
to allow the display driver to create its own drm_connector instead of
using the exynos_drm_connector. This moves things closer to completely
removing the exynos_drm_connector abstraction.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Consolidate suspend/resume in drm_drv
Sean Paul [Thu, 30 Jan 2014 21:19:27 +0000 (16:19 -0500)]
drm/exynos: Consolidate suspend/resume in drm_drv

This patch removes all of the suspend/resume logic from the individual
drivers and consolidates it in drm_drv. This consolidation reduces the
number of functions which enable/disable the hardware to just one -- the
dpms callback. This ensures that we always power up/down in a consistent
manner.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Clean up FIMD power on/off routines
Sean Paul [Thu, 30 Jan 2014 21:19:26 +0000 (16:19 -0500)]
drm/exynos: Clean up FIMD power on/off routines

This patch separates the fimd_activate function into poweron/poweroff
functions to be more consistent with the other drivers in exynos drm. It
also properly cleans up after failures in poweron. The functions have
also been shuffled around such that they are all in the same
spot in the file and poweron/poweroff can be called from the dpms function.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement dpms display callback in DP
Sean Paul [Thu, 30 Jan 2014 21:19:25 +0000 (16:19 -0500)]
drm/exynos: Implement dpms display callback in DP

This patch implements the dpms display callback for the DP driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Move display implementation into dp
Sean Paul [Thu, 30 Jan 2014 21:19:23 +0000 (16:19 -0500)]
drm/exynos: Move display implementation into dp

This patch moves the exynos_drm_display implementation from fimd into
the dp driver. This will allow for tighter integration of the dp driver
into the exynos drm driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Move dp driver from video/ to drm/
Sean Paul [Thu, 30 Jan 2014 21:19:22 +0000 (16:19 -0500)]
drm/exynos: Move dp driver from video/ to drm/

This patch moves the code from video/ to drm/. This is required the DP
driver needs to power on/off in the correct order in relation to fimd.
This will also allow the DP driver to participate in drm modeset as well
as provide accurate connection detection and edid.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Remove unused/useless fimd_context members
Sean Paul [Thu, 30 Jan 2014 21:19:21 +0000 (16:19 -0500)]
drm/exynos: Remove unused/useless fimd_context members

This patch removes a few fimd_context members which are either entirely
unused or unneeded.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Use mode_set to configure fimd
Sean Paul [Thu, 30 Jan 2014 21:19:20 +0000 (16:19 -0500)]
drm/exynos: Use mode_set to configure fimd

This patch uses the mode passed into mode_set to configure fimd instead
of directly using the panel from context. This will allow us to move
the exynos_drm_display implementation out of fimd, where it doesn't
belong.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Implement mode_fixup manager operation
Sean Paul [Thu, 30 Jan 2014 21:19:19 +0000 (16:19 -0500)]
drm/exynos: Implement mode_fixup manager operation

This patch adds a new manager callback for mode_fixup and pipes it
through exynos_drm_crtc. This will allow the manager drivers to
alter the mode during modeset.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Add mode_set manager operation
Sean Paul [Thu, 30 Jan 2014 21:19:18 +0000 (16:19 -0500)]
drm/exynos: Add mode_set manager operation

This patch adds a mode_set callback to the manager operations which
sets the crtc's current mode to the manager driver. This will allow the
fimd  driver to set its mode using values from drm, instead of the dt.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Disable unused crtc planes from crtc
Sean Paul [Thu, 30 Jan 2014 21:19:17 +0000 (16:19 -0500)]
drm/exynos: Disable unused crtc planes from crtc

This patch moves the code which disables unused crtc planes from the
encoder to the crtc. Since there is a 1:1 encoder/crtc mapping in
exynos, the only valid crtc change the pre-existing code could catch is
disconnecting an active crtc from the encoder. Thus it is functionally
equivalent to just disable all planes attached to a crtc when the crtc
is disabled.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Use drm_mode_copy to copy modes
Sean Paul [Thu, 30 Jan 2014 21:19:16 +0000 (16:19 -0500)]
drm/exynos: Use drm_mode_copy to copy modes

This patch changes the manual copying of mode to adjusted_mode in
mode_fixup to use drm_mode_copy instead of handling things manually.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Remove exynos_drm_hdmi shim
Sean Paul [Thu, 30 Jan 2014 21:19:15 +0000 (16:19 -0500)]
drm/exynos: Remove exynos_drm_hdmi shim

This patch trims exynos_drm_hdmi out of the driver. The reason it
existed in the first place was to make up for the mixture of
display/overlay/manager ops being spread across hdmi and mixer. With
that code now rationalized, mixer and hdmi map directly to
exynos_drm_crtc and exynos_drm_encoder, respectively. Since there is a
1:1 mapping, we no longer need this layer.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: hdmi: remove the i2c drivers and use
Daniel Kurtz [Mon, 24 Feb 2014 09:52:51 +0000 (18:52 +0900)]
drm/exynos: hdmi: remove the i2c drivers and use

The i2c client was previously being passed into the hdmi driver via a
dedicated i2c driver, and then a global variable. This patch removes all
of that and just uses the device tree to get the i2c_client. This patch
also properly references the client so we don't lose it before we're
done with it.

Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
[seanpaul changed to phandle lookup instead of using of node name]
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Split manager/display/subdrv
Sean Paul [Wed, 19 Feb 2014 12:02:55 +0000 (21:02 +0900)]
drm/exynos: Split manager/display/subdrv

This patch splits display and manager from subdrv. The result is that
crtc functions can directly call into manager callbacks and encoder
functions can directly call into display callbacks. This will allow
us to remove the exynos_drm_hdmi shim and support mixer/hdmi & fimd/dp
with common code.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Use unsigned long for possible_crtcs
Sean Paul [Thu, 30 Jan 2014 21:19:11 +0000 (16:19 -0500)]
drm/exynos: Use unsigned long for possible_crtcs

Change all instances of possible_crtcs in the exynos drm driver to be
unsigned long. This matches the type used in the drm layer.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Don't keep dpms state in encoder
Sean Paul [Thu, 30 Jan 2014 21:19:10 +0000 (16:19 -0500)]
drm/exynos: Don't keep dpms state in encoder

This patch removes the dpms state tracking in encoder. This
state is at best confusing and at worst incorrect since the display
drivers can turn on and off without propagating the value.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Rename display_op power_on to dpms
Sean Paul [Thu, 30 Jan 2014 21:19:09 +0000 (16:19 -0500)]
drm/exynos: Rename display_op power_on to dpms

This patch renames the display_op power_on to dpms to accurately reflect
what the function does.

The side-effect of this patch is that the new hdmi dpms callback is now
invoked twice in the dpms path. This is safe and will be dealt with when
the exynos_drm shim goes away.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Remove dpms link between encoder/connector
Sean Paul [Thu, 30 Jan 2014 21:19:08 +0000 (16:19 -0500)]
drm/exynos: Remove dpms link between encoder/connector

This patch removes the call from encoder dpms into connector dpms (which
will then call back into encoder dpms through the helper function). The
callback is likely to keep connector->dpms in the right state when
initiating dpms from crtc or encoder, but this isn't the right way to do
it. This patch is the first step towards rationalizing power management
in the exynos drm driver.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Remove apply manager callback
Sean Paul [Thu, 30 Jan 2014 21:19:07 +0000 (16:19 -0500)]
drm/exynos: Remove apply manager callback

This patch removes the apply() manager callback in favor of putting the
relevant commits in the individual drivers. This will mitigate some of
the difference between the suspend/resume path and the dpms path

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Pass exynos_drm_manager in manager ops instead of dev
Sean Paul [Thu, 30 Jan 2014 21:19:06 +0000 (16:19 -0500)]
drm/exynos: Pass exynos_drm_manager in manager ops instead of dev

This patch changes the manager ops callbacks from accepting the subdrv
device pointer to taking a pointer to the manager. This will allow us
to move closer to decoupling manager/display from subdrv, and subsequently
decoupling the crtc/plane from the encoder.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: hdmi: Implement initialize op for hdmi
Sean Paul [Thu, 30 Jan 2014 21:19:05 +0000 (16:19 -0500)]
drm/exynos: hdmi: Implement initialize op for hdmi

This patch implements the initialize callback in the hdmi and mixer
manager. This allows us to get rid of drm_dev in the drm_hdmi level and
track it in the mixer and hdmi drivers. This is one of the things
holding back the complete removal of the drm_hdmi layer.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Use manager_op initialize in fimd
Sean Paul [Thu, 30 Jan 2014 21:19:04 +0000 (16:19 -0500)]
drm/exynos: Use manager_op initialize in fimd

This patch implements the intitialize manager op in fimd. This will
allow us to keep track of drm_dev in context instead of using subdev,
which in turn makes it easier to remove subdev from fimd.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Add an initialize function to manager and display
Sean Paul [Thu, 30 Jan 2014 21:19:03 +0000 (16:19 -0500)]
drm/exynos: Add an initialize function to manager and display

This patch adds an initialize function to the manager and display
operations. This allows them to keep track of drm_device in their
local context, as well as adds an initialization hook right after
the encoder is created.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Merge overlay_ops into manager_ops
Sean Paul [Thu, 30 Jan 2014 21:19:02 +0000 (16:19 -0500)]
drm/exynos: Merge overlay_ops into manager_ops

This patch merges overlay_ops into manager_ops. In all cases,
overlay_ops is implemented in the same place as manager ops, it doesn't
serve a functional purpose, and doesn't make things more clear.

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Remove useless slab.h include
Stephane Marchesin [Thu, 30 Jan 2014 21:19:01 +0000 (06:19 +0900)]
drm/exynos: Remove useless slab.h include

Signed-off-by: Stephane Marchesin <marcheu@chromium.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: set the active aspect ratio as per mode
Shirish S [Thu, 13 Mar 2014 05:28:28 +0000 (10:58 +0530)]
drm/exynos: set the active aspect ratio as per mode

Now that the drm_display_mode also provides aspect
ratio for all resolutions, this patch adds its usage
to set the active aspect ratio of AVI info frame
packets as per CEA-861-D standard's Table 9.

This is also needed to abide by the 7-27
compliance test of HDMI.

Signed-off-by: Shirish S <s.shirish@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: add phy settings for RB resolutions
Shirish S [Thu, 13 Mar 2014 05:28:27 +0000 (14:28 +0900)]
drm/exynos: add phy settings for RB resolutions

This patch adds support for the below mentioned
pixel clocks in Exynos5250.
Without them, following display modes won¡¯t
be supported:

71 MHz  - 1280x800@60Hz RB
73.25 MHz  - 800x600@120Hz RB
88.75 MHz  - 1440x900@60Hz RB
115.5 MHz  - 1024x768@120Hz RB
119 MHz  - 1680x1050@60Hz RB

Signed-off-by: Shirish S <s.shirish@samsung.com>
Reviewed-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/exynos: Fix (more) freeing issues in exynos_drm_drv.c
Daniel Kurtz [Mon, 17 Mar 2014 03:28:06 +0000 (11:28 +0800)]
drm/exynos: Fix (more) freeing issues in exynos_drm_drv.c

The following commit [0] fixed a use-after-free, but left the subdrv open
in the error path.

[0] commit 6ca605f7c70895a35737435f17ae9cc5e36f1466
drm/exynos: Fix freeing issues in exynos_drm_drv.c

Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
10 years agodrm/helper: lock all around force mode restore
Dave Airlie [Fri, 21 Mar 2014 00:45:40 +0000 (10:45 +1000)]
drm/helper: lock all around force mode restore

Since Daniel documented things with a sledge hammer, we got lots of
nice backtraces in suspend/resume operations, I've check the callers
of this and they all seems safe to me,

This fixes one set of warns I reported.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/fb-helper: improve drm_fb_helper_initial_config locking
Daniel Vetter [Thu, 20 Mar 2014 13:26:35 +0000 (14:26 +0100)]
drm/fb-helper: improve drm_fb_helper_initial_config locking

The locking in drm_fb_helper_initial_config is a bit troublesome for a
few reasons:

- We can't just wrap the entire function up into modeset locks since
  the fbdev registration might call down into fbcon code, which then
  through our ->set_par implementation needs to be able to grab all
  modeset locks. So we'd have a neat deadlock.

- This implies though that all current callers don't hold any modeset
  locks by necessity, so we have free reign to grab any modeset locks
  we need to grab.

- The private state of the fbdev helper doesn't need any protection
  through locks, since once we have the fbdev registered it is mostly
  invariant or protected through the modeset locking in ->set_par and
  other callbacks. We can fully rely on driver having non-racy setup
  sequences here. For the initial config computation we actually may
  not grab locks since drivers which provide their own magic sauce
  (like i915) might need to grab locks themselves.

- We should grab locks though when we probe outputs. Currently there's
  not much risk, but already now userspace could start poking at sysfs
  files and so probe concurrently. I expect that in the future driver
  init will be much more async, and since probing is really
  time-consuming this is a prime candidate.

- We must not hold any crtc->mutex locks while calling probe functions
  since those might need to lock a crtc for e.g. load detection. i915
  is such a driver.

Also it's the probing calls which hit upon piles of new locking
asserts I've recently added in

commit 62ff94a5492175759546f8bc61383189d6b49122
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jan 23 22:18:47 2014 +0100

    drm/crtc-helper: remove LOCKING from kerneldoc

and

commit 63951385052f7974155fa38f962f0f4e9847f90a
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jan 23 15:14:15 2014 +0100

    drm/doc: Repleace LOCKING kerneldoc sections in drm_modes.c

Hence the right fix is to grab the mode_config mutex, but only that
and only right around the probe calls.

It seems to be sufficient to shut up all the locking WARNINGs I see on
i915 and nouveau in drm_fb_helper_initial_config.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/crtc-helper: fix locking for drm_helper_disable_unused_functions
Daniel Vetter [Thu, 20 Mar 2014 13:26:34 +0000 (14:26 +0100)]
drm/crtc-helper: fix locking for drm_helper_disable_unused_functions

We have two calling contexts for thise function:

- In the crtc helper code itself as part of the ->set_config
  implementation. In this calling context all modeset locks are
  already held, as they should.

- In drivers not implementing fastboot before the fbdev/fbcon setup
  and initialization. This has been added for all drivers in

  commit 76a39dbfb2d1bc45219839e5a95d4ceaf6ca114f
  Author: Daniel Vetter <daniel.vetter@ffwll.ch>
  Date:   Sun Jan 20 23:12:54 2013 +0100

      drm/fb-helper: don't disable everything in initial_config

  In this calling context we do not hold any modeset locks since the
  immediately following call to initialize the fbev emulation grabs
  all these locks themselves.

- There are two exceptions to the above rule: shmob doesn't have fbdev
  emulation support. I've manually checked the callchain up to the
  driver load function and no kms locks are held.

The right fix therefore is to split this helper into an internal and
external version and add the required locking to the function exported
to drivers.

This remedies locking inconsistencies exposed by me adding locking
WARNs as part of the recent kerneldoc abi polishing done in

commit 62ff94a5492175759546f8bc61383189d6b49122
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jan 23 22:18:47 2014 +0100

    drm/crtc-helper: remove LOCKING from kerneldoc

and

commit 63951385052f7974155fa38f962f0f4e9847f90a
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jan 23 15:14:15 2014 +0100

    drm/doc: Repleace LOCKING kerneldoc sections in drm_modes.c

v2: It helps when I actually git add the entire thing.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'gma500-next' of git://github.com/patjak/drm-gma500 into drm-next
Dave Airlie [Thu, 20 Mar 2014 23:27:07 +0000 (09:27 +1000)]
Merge branch 'gma500-next' of git://github.com/patjak/drm-gma500 into drm-next

Summary of what's included:

- SGX MMU support
- SGX IRQ handling (Page faults and blitter fences)
- Minor Cedarview and Poulsbo unification
- Work queue for ASLE interrupt work
- Various cleanups, style fixes and removal of dead code

* 'gma500-next' of git://github.com/patjak/drm-gma500:
  drm/gma500: remove stub .open/postclose
  drm/gma500: Code cleanup - inline documentation
  drm/gma500: Code cleanup - style fixes
  drm/gma500: Code cleanup - removal of centralized exiting of function
  drm/gma500/cdv: Cedarview display cleanups
  drm/gma500: Unify encoder mode fixup
  drm/gma500: Unify _get_core_freq for cdv and psb
  drm/gma500: Move asle interrupt work into a work task
  drm/gma500: Remove dead code
  drm/gma500: Add backing type and base align to psb_gem_create()
  drm/gma500: Remove unused ioctls
  drm/gma500: Always trap MMU page faults
  drm/gma500: Hook up the MMU
  drm/gma500: Add first piece of blitter code
  drm/gma500: Give MMU code it's own header file
  drm/gma500: Add support for SGX interrupts
  drm/gma500: Make SGX MMU driver actually do something

10 years agodrm/radeon/kms: merge conflicted badly
Dave Airlie [Thu, 20 Mar 2014 01:08:11 +0000 (11:08 +1000)]
drm/radeon/kms: merge conflicted badly

Not sure why git didn't flag this, but the result of automerge
from 3.14-rc7 screwed up the radeon init procedure.

Reported-by: Fireburn on #radeon
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'topic/core-stuff' of git://git.freedesktop.org/git/drm-intel into drm...
Dave Airlie [Tue, 18 Mar 2014 09:23:22 +0000 (19:23 +1000)]
Merge branch 'topic/core-stuff' of git://git.freedesktop.org/git/drm-intel into drm-next

Merge straggling core drm patches.

* 'topic/core-stuff' of git://git.freedesktop.org/git/drm-intel:
  drm: Fix use-after-free in the shadow-attache exit code
  drm/fb-helper: Do the 'max_conn_count' zero check
  drm: Check if the allocation has succeeded before dereferencing newmode
  drm/fb-helper: Use drm_fb_helper_restore_fbdev_mode() in drm_fb_helper_set_par()
  drm/edid: request HDMI underscan by default

10 years agodrm/cma: remove to make sg_table when gem cma is created
Joonyoung Shim [Sat, 15 Mar 2014 07:30:28 +0000 (16:30 +0900)]
drm/cma: remove to make sg_table when gem cma is created

The sg_table made when gem cma is created isn't used anywhere. The sgt
of struct drm_gem_cma_object will have only sg_tabel imported.

Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'drm-next' of git://people.freedesktop.org/~dvdhrm/linux into drm-next
Dave Airlie [Tue, 18 Mar 2014 09:17:02 +0000 (19:17 +1000)]
Merge branch 'drm-next' of git://people.freedesktop.org/~dvdhrm/linux into drm-next

This is the 3rd respin of the drm-anon patches. They allow module unloading, use
the pin_fs_* helpers recommended by Al and are rebased on top of drm-next. Note
that there are minor conflicts with the "drm-minor" branch.

* 'drm-next' of git://people.freedesktop.org/~dvdhrm/linux:
  drm: init TTM dev_mapping in ttm_bo_device_init()
  drm: use anon-inode instead of relying on cdevs
  drm: add pseudo filesystem for shared inodes

10 years agoMerge tag 'v3.14-rc7' into drm-next
Dave Airlie [Tue, 18 Mar 2014 09:12:31 +0000 (19:12 +1000)]
Merge tag 'v3.14-rc7' into drm-next

Linux 3.14-rc7

Backmerge to help out Intel guys.

10 years agoMerge branch 'drm-docs' of ssh://people.freedesktop.org/~danvet/drm into drm-next
Dave Airlie [Tue, 18 Mar 2014 09:09:10 +0000 (19:09 +1000)]
Merge branch 'drm-docs' of ssh://people.freedesktop.org/~danvet/drm into drm-next

Here's my drm documentation update and driver api polish pull request.
Alex reviewed the entire pile, I've applied a little bit of spelling
polish in a few places since then and otherwise the Usual Suspects (David,
Rob, ...) don't seem up to have another look at it (I've poked them on
irc). So I think it's as good as it gets ;-)

Note that I've dropped the final imx breaker patch since that's blocked on
imx getting sane. Once that's landed I'll ping you to pick up that
straggler.

* 'drm-docs' of ssh://people.freedesktop.org/~danvet/drm: (34 commits)
  drm/imx: remove drm_mode_connector_detach_encoder harder
  drm: kerneldoc polish for drm_crtc.c
  drm: kerneldoc polish for drm_crtc_helper.c
  drm: drop error code for drm_helper_resume_force_mode
  drm/crtc-helper: remove LOCKING from kerneldoc
  drm: remove return value from drm_helper_mode_fill_fb_struct
  drm/doc: Fix misplaced </para>
  drm: remove drm_display_mode->private_size
  drm: polish function kerneldoc for drm_modes.[hc]
  drm/modes: drop maxPitch from drm_mode_validate_size
  drm/modes: drop return value from drm_display_mode_from_videomode
  drm/modes: remove drm_mode_height/width
  drm: extract drm_modes.h for drm_crtc.h functions
  drm: move drm_mode related functions into drm_modes.c
  drm/doc: Repleace LOCKING kerneldoc sections in drm_modes.c
  drm/doc: Integrate drm_modes.c kerneldoc
  drm/kms: rip out drm_mode_connector_detach_encoder
  drm/doc: Add function reference documentation for drm_mm.c
  drm/doc: Overview documentation for drm_mm.c
  drm/mm: Remove MM_UNUSED_TARGET
  ...

10 years agoMerge branch 'drm-intel-next' of git://git.freedesktop.org/git/drm-intel into drm...
Dave Airlie [Tue, 18 Mar 2014 09:06:53 +0000 (19:06 +1000)]
Merge branch 'drm-intel-next' of git://git.freedesktop.org/git/drm-intel into drm-next

- fine-grained display power domains for byt (Imre)
- runtime pm prep patches for !hsw from Paulo
- WiZ hashing flag updates from Ville
- ppgtt setup cleanup and enabling of full 4G range on bdw (Ben)
- fixes from Jesse for the inherited intial config code
- gpu reset code improvements from Mika
- per-pipe num_planes refactoring from Damien
- stability fixes around bdw forcewake handling and other bdw w/a from Mika
  Ken
- and as usual a pile of smaller fixes all over

* 'drm-intel-next' of git://git.freedesktop.org/git/drm-intel: (107 commits)
  drm/i915: Go OCD on the Makefile
  drm/i915: Implement command buffer parsing logic
  drm/i915: Refactor shmem pread setup
  drm/i915: Avoid div by zero when pixel clock is large
  drm/i915: power domains: add vlv power wells
  drm/i915: factor out intel_set_cpu_fifo_underrun_reporting_nolock
  drm/i915: vlv: factor out valleyview_display_irq_install
  drm/i915: sanity check power well sw state against hw state
  drm/i915: factor out reset_vblank_counter
  drm/i915: sanitize PUNIT register macro definitions
  drm/i915: vlv: keep first level vblank IRQs masked
  drm/i915: check pipe power domain when reading its hw state
  drm/i915: check port power domain when reading the encoder hw state
  drm/i915: get port power domain in connector detect handlers
  drm/i915: add port power domains
  drm/i915: add noop power well handlers instead of NULL checking them
  drm/i915: split power well 'set' handler to separate enable/disable/sync_hw
  drm/i915: add init power domain to always-on power wells
  drm/i915: move power domain macros to intel_pm.c
  drm/i915: Disable full ppgtt by default
  ...

10 years agoMerge remote-tracking branch 'airlied/drm-next' into drm-intel-next
Daniel Vetter [Tue, 18 Mar 2014 08:43:56 +0000 (09:43 +0100)]
Merge remote-tracking branch 'airlied/drm-next' into drm-intel-next

Conflicts:
drivers/gpu/drm/i915/Makefile

Makefile cleanup in drm-intel-next conflicts with a build-fix to move
intel_opregion under CONFIG_ACPI.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/gma500: remove stub .open/postclose
David Herrmann [Wed, 29 Jan 2014 13:39:34 +0000 (14:39 +0100)]
drm/gma500: remove stub .open/postclose

These are unused and can safely be dropped. DRM core verifies they're
non-NULL before it calls them.

Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Code cleanup - inline documentation
Arthur Borsboom [Sat, 15 Mar 2014 21:12:18 +0000 (22:12 +0100)]
drm/gma500: Code cleanup - inline documentation

Improve readability by adding/changing inline documentation

Signed-off-by: Arthur Borsboom <arthurborsboom@gmail.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Code cleanup - style fixes
Arthur Borsboom [Sat, 15 Mar 2014 21:12:17 +0000 (22:12 +0100)]
drm/gma500: Code cleanup - style fixes

Code cleanup by following i915 constant/variable names and ordering
Code cleanup by following directions from kernel doc: Codingstyle
Code cleanup by following directions from kernel doc: DRM

Signed-off-by: Arthur Borsboom <arthurborsboom@gmail.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Code cleanup - removal of centralized exiting of function
Arthur Borsboom [Sat, 15 Mar 2014 21:12:16 +0000 (22:12 +0100)]
drm/gma500: Code cleanup - removal of centralized exiting of function

Removed centralized exiting of function (goto statement), since it was
the only used in one single location with only a return statement.

Signed-off-by: Arthur Borsboom <arthurborsboom@gmail.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500/cdv: Cedarview display cleanups
Patrik Jakobsson [Tue, 11 Mar 2014 22:57:04 +0000 (23:57 +0100)]
drm/gma500/cdv: Cedarview display cleanups

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Unify encoder mode fixup
Patrik Jakobsson [Tue, 11 Mar 2014 22:14:06 +0000 (23:14 +0100)]
drm/gma500: Unify encoder mode fixup

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Unify _get_core_freq for cdv and psb
Patrik Jakobsson [Tue, 11 Mar 2014 21:53:43 +0000 (22:53 +0100)]
drm/gma500: Unify _get_core_freq for cdv and psb

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Move asle interrupt work into a work task
Patrik Jakobsson [Tue, 11 Mar 2014 17:51:20 +0000 (18:51 +0100)]
drm/gma500: Move asle interrupt work into a work task

Previously the backlight code was called from IRQ context which isn't
allowed. This patch moves all the asle work into a work task which takes
care of the locking bug reported by users.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=64221
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Remove dead code
Thierry Reding [Fri, 21 Feb 2014 07:55:25 +0000 (08:55 +0100)]
drm/gma500: Remove dead code

The gma500 driver sets DRIVER_GEM unconditionally, so testing for the
absence of the feature will always fail.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Add backing type and base align to psb_gem_create()
Patrik Jakobsson [Mon, 6 Jan 2014 01:39:10 +0000 (02:39 +0100)]
drm/gma500: Add backing type and base align to psb_gem_create()

We'll need this for our gem create ioctl in a later patch.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Remove unused ioctls
Patrik Jakobsson [Sun, 5 Jan 2014 00:01:14 +0000 (01:01 +0100)]
drm/gma500: Remove unused ioctls

All of these ioctls are unused and most of them just duplicate what drm
already provides.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Always trap MMU page faults
Patrik Jakobsson [Sat, 4 Jan 2014 23:27:51 +0000 (00:27 +0100)]
drm/gma500: Always trap MMU page faults

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Hook up the MMU
Patrik Jakobsson [Sat, 4 Jan 2014 21:11:17 +0000 (22:11 +0100)]
drm/gma500: Hook up the MMU

Properly init the MMU and add MMU entries when adding GTT entries

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Add first piece of blitter code
Patrik Jakobsson [Sat, 4 Jan 2014 19:58:35 +0000 (20:58 +0100)]
drm/gma500: Add first piece of blitter code

Right now, all we need to know about the blitter is that it's not doing
anything that can be messed up when fiddling with MMU mappings.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Give MMU code it's own header file
Patrik Jakobsson [Sat, 4 Jan 2014 18:35:20 +0000 (19:35 +0100)]
drm/gma500: Give MMU code it's own header file

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Add support for SGX interrupts
Patrik Jakobsson [Fri, 3 Jan 2014 00:52:46 +0000 (01:52 +0100)]
drm/gma500: Add support for SGX interrupts

Add 2D blit status and MMU fault interrupts to the IRQ handler.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm/gma500: Make SGX MMU driver actually do something
Patrik Jakobsson [Thu, 2 Jan 2014 00:27:00 +0000 (01:27 +0100)]
drm/gma500: Make SGX MMU driver actually do something

Old MMU code never wrote PDs or PTEs to any registers. Now we do, and
that's a good start.

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
10 years agodrm: Fix use-after-free in the shadow-attache exit code
Daniel Vetter [Thu, 30 Jan 2014 16:58:38 +0000 (17:58 +0100)]
drm: Fix use-after-free in the shadow-attache exit code

This regression has been introduced in

commit b3f2333de8e81b089262b26d52272911523e605f
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Dec 11 11:34:31 2013 +0100

    drm: restrict the device list for shadow attached drivers

Reported-by: Dave Jones <davej@redhat.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agodrm/fb-helper: Do the 'max_conn_count' zero check
Xiubo Li [Mon, 10 Mar 2014 01:33:58 +0000 (09:33 +0800)]
drm/fb-helper: Do the 'max_conn_count' zero check

Since we cannot make sure the 'max_conn_count' will always be none
zero from the users, and then if max_conn_count equals to zero, the
kcalloc() will return ZERO_SIZE_PTR, which equals to ((void *)16).

So this patch fix this with just doing the 'max_conn_count' zero check
in the front of drm_fb_helper_init().

Signed-off-by: Xiubo Li <Li.Xiubo@freescale.com>
CC: Jani Nikula <jani.nikula@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: Check if the allocation has succeeded before dereferencing newmode
Damien Lespiau [Mon, 3 Mar 2014 23:59:07 +0000 (23:59 +0000)]
drm: Check if the allocation has succeeded before dereferencing newmode

We allocate memory in drm_display_mode_from_vic_index() and use it
without checking the pointer is valid. Fix that.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/fb-helper: Use drm_fb_helper_restore_fbdev_mode() in drm_fb_helper_set_par()
Ville Syrjälä [Wed, 26 Feb 2014 20:16:03 +0000 (22:16 +0200)]
drm/fb-helper: Use drm_fb_helper_restore_fbdev_mode() in drm_fb_helper_set_par()

Use drm_fb_helper_restore_fbdev_mode() in drm_fb_helper_set_par() to
make sure extra planes get disabled whenever fbcon takes over.

Otherwise the code in drm_fb_helper_set_par() was already doing the
exact same thing as drm_fb_helper_restore_fbdev_mode(), so this doesn't
change the behaviour in any other way.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoMerge branch 'drm-minor' of git://people.freedesktop.org/~dvdhrm/linux into drm-next
Dave Airlie [Mon, 17 Mar 2014 02:29:29 +0000 (12:29 +1000)]
Merge branch 'drm-minor' of git://people.freedesktop.org/~dvdhrm/linux into drm-next

This series contains several cleanups for the DRM-minor handling. All but the
last one reviewed by Daniel and tested by Thierry. Initially, the series
included patches to convert minor-handling to a common base-ID, but have
been NACKed by Daniel so I dropped them and only included the main part in the
last patch. With this in place, drm_global_mutex is no longer needed for
minor-handling (but still for device unregistration..).
There are some pending patches that try to remove the global mutex entirely, but
they need some more reviews and thus are not included.
* 'drm-minor' of git://people.freedesktop.org/~dvdhrm/linux:
  drm: make minors independent of global lock
  drm: inline drm_minor_get_id()
  drm: coding-style fixes in minor handling
  drm: remove redundant minor->device field
  drm: remove unneeded #ifdef CONFIG_DEBUGFS
  drm: rename drm_unplug/get_minor() to drm_minor_register/unregister()
  drm: move drm_put_minor() to drm_minor_free()
  drm: allocate minors early
  drm: add minor-lookup/release helpers
  drm: provide device-refcount
  drm: turn DRM_MINOR_* into enum
  drm: remove unused DRM_MINOR_UNASSIGNED
  drm: skip redundant minor-lookup in open path
  drm: group dev-lifetime related members

10 years agoLinux 3.14-rc7
Linus Torvalds [Mon, 17 Mar 2014 01:51:24 +0000 (18:51 -0700)]
Linux 3.14-rc7

10 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~dvdhrm/linux into drm-next
Dave Airlie [Mon, 17 Mar 2014 00:42:58 +0000 (10:42 +1000)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~dvdhrm/linux into drm-next

This branch includes 6 minor fixes mainly for udl. Everything non-trivial was
reviewed by Daniel and the patches have been on the list for quite some time.

* 'drm-fixes' of git://people.freedesktop.org/~dvdhrm/linux:
  drm/gem: dont init "ret" in drm_gem_mmap()
  drm/crtc: add sanity checks to create_dumb()
  drm/gem: free vma-node during object-cleanup
  drm/gem: fix indentation
  drm/udl: fix Bpp calculation in dumb_create()
  drm/udl: fix error-path when damage-req fails

10 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 16 Mar 2014 17:42:07 +0000 (10:42 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Three small fixes"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/clock: Prevent tracing recursion in sched_clock_cpu()
  stop_machine: Fix^2 race between stop_two_cpus() and stop_cpus()
  sched/deadline: Deny unprivileged users to set/change SCHED_DEADLINE policy

10 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 16 Mar 2014 17:41:21 +0000 (10:41 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Misc smaller fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86: Fix leak in uncore_type_init failure paths
  perf machine: Use map as success in ip__resolve_ams
  perf symbols: Fix crash in elf_section_by_name
  perf trace: Decode architecture-specific signal numbers

10 years agoipc: Fix 2 bugs in msgrcv() MSG_COPY implementation
Michael Kerrisk [Mon, 10 Mar 2014 13:46:07 +0000 (14:46 +0100)]
ipc: Fix 2 bugs in msgrcv() MSG_COPY implementation

While testing and documenting the msgrcv() MSG_COPY flag that Stanislav
Kinsbursky added in commit 4a674f34ba04 ("ipc: introduce message queue
copy feature" => kernel 3.8), I discovered a couple of bugs in the
implementation.  The two bugs concern MSG_COPY interactions with other
msgrcv() flags, namely:

 (A) MSG_COPY + MSG_EXCEPT
 (B) MSG_COPY + !IPC_NOWAIT

The bugs are distinct (and the fix for the first one is obvious),
however my fix for both is a single-line patch, which is why I'm
combining them in a single mail, rather than writing two mails+patches.

 ===== (A) MSG_COPY + MSG_EXCEPT =====

With the addition of the MSG_COPY flag, there are now two msgrcv()
flags--MSG_COPY and MSG_EXCEPT--that modify the meaning of the 'msgtyp'
argument in unrelated ways.  Specifying both in the same call is a
logical error that is currently permitted, with the effect that MSG_COPY
has priority and MSG_EXCEPT is ignored.  The call should give an error
if both flags are specified.  The patch below implements that behavior.

 ===== (B) (B) MSG_COPY + !IPC_NOWAIT =====

The test code that was submitted in commit 3a665531a3b7 ("selftests: IPC
message queue copy feature test") shows MSG_COPY being used in
conjunction with IPC_NOWAIT.  In other words, if there is no message at
the position 'msgtyp'.  return immediately with the error in ENOMSG.

What was not (fully) tested is the behavior if MSG_COPY is specified
*without* IPC_NOWAIT, and there is an odd behavior.  If the queue
contains less than 'msgtyp' messages, then the call blocks until the
next message is written to the queue.  At that point, the msgrcv() call
returns a copy of the newly added message, regardless of whether that
message is at the ordinal position 'msgtyp'.  This is clearly bogus, and
problematic for applications that might want to make use of the MSG_COPY
flag.

I considered the following possible solutions to this problem:

 (1) Force the call to block until a message *does* appear at the
     position 'msgtyp'.

 (2) If the MSG_COPY flag is specified, the kernel should implicitly add
     IPC_NOWAIT, so that the call fails with ENOMSG for this case.

 (3) If the MSG_COPY flag is specified, but IPC_NOWAIT is not, generate
     an error (probably, EINVAL is the right one).

I do not know if any application would really want to have the
functionality of solution (1), especially since an application can
determine in advance the number of messages in the queue using msgctl()
IPC_STAT.  Obviously, this solution would be the most work to implement.

Solution (2) would have the effect of silently fixing any applications
that tried to employ broken behavior.  However, it would mean that if we
later decided to implement solution (1), then user-space could not
easily detect what the kernel supports (but, since I'm somewhat doubtful
that solution (1) is needed, I'm not sure that this is much of a
problem).

Solution (3) would have the effect of informing broken applications that
they are doing something broken.  The downside is that this would cause
a ABI breakage for any applications that are currently employing the
broken behavior.  However:

a) Those applications are almost certainly not getting the results they
   expect.
b) Possibly, those applications don't even exist, because MSG_COPY is
   currently hidden behind CONFIG_CHECKPOINT_RESTORE.

The upside of solution (3) is that if we later decided to implement
solution (1), user-space could determine what the kernel supports, via
the error return.

In my view, solution (3) is mildly preferable to solution (2), and
solution (1) could still be done later if anyone really cares.  The
patch below implements solution (3).

PS.  For anyone out there still listening, it's the usual story:
documenting an API (and the thinking about, and the testing of the API,
that documentation entails) is the one of the single best ways of
finding bugs in the API, as I've learned from a lot of experience.  Best
to do that documentation before releasing the API.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Acked-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Cc: stable@vger.kernel.org
Cc: Serge Hallyn <serge.hallyn@canonical.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge branch 'drm-minor' into drm-next
David Herrmann [Sun, 16 Mar 2014 12:13:51 +0000 (13:13 +0100)]
Merge branch 'drm-minor' into drm-next

Fix minor conflicts with drm-anon:
 - allocation/free order
 - drm_device header cleanups

10 years agoMerge branch 'drm-anon' into drm-next
David Herrmann [Sun, 16 Mar 2014 12:04:11 +0000 (13:04 +0100)]
Merge branch 'drm-anon' into drm-next

10 years agodrm: make minors independent of global lock
David Herrmann [Mon, 24 Feb 2014 14:53:25 +0000 (15:53 +0100)]
drm: make minors independent of global lock

We used to protect minor-lookup and setup by the global drm lock. To
continue our attempts of dropping drm_global_mutex, this patch makes the
minor management independent of it. Furthermore, we make it all atomic and
switch to spin-locks instead of a mutex.

Now that minor-lookup is independent, we also move the
"drm_is_unplugged()" test into the minor-lookup path. There is no reason
to ever return a minor for unplugged objects, so keep that logic internal.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
10 years agodrm: inline drm_minor_get_id()
David Herrmann [Mon, 24 Feb 2014 14:35:09 +0000 (15:35 +0100)]
drm: inline drm_minor_get_id()

We can significantly simplify this helper by using plain multiplication.
Note that we converted the minor-type to an enum earlier so this didn't
work before.

We also fix a minor range-bug here: the limit argument of idr_alloc() is
*exclusive*, not inclusive, so we should use 64 instead of 63 as offset.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
10 years agodrm: coding-style fixes in minor handling
David Herrmann [Mon, 24 Feb 2014 14:32:00 +0000 (15:32 +0100)]
drm: coding-style fixes in minor handling

Properly name goto-labels, remove empty lines and use DRM_ERROR if
possible.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
10 years agodrm: remove redundant minor->device field
David Herrmann [Wed, 29 Jan 2014 12:12:31 +0000 (13:12 +0100)]
drm: remove redundant minor->device field

Whenever we access minor->device, we are in a minor->kdev->...->fops
callback so the minor->kdev pointer *must* be valid. Thus, simply use
minor->kdev->devt instead of minor->device and remove the redundant field.

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>