cascardo/linux.git
9 years agoMerge tag 'drm-intel-next-2015-02-27' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Mon, 9 Mar 2015 09:41:15 +0000 (19:41 +1000)]
Merge tag 'drm-intel-next-2015-02-27' of git://anongit.freedesktop.org/drm-intel into drm-next

- Y tiling support for scanout from Tvrtko&Damien
- Remove more UMS support
- some small prep patches for OLR removal from John Harrison
- first few patches for dynamic pagetable allocation from Ben Widawsky, rebased
  by tons of other people
- DRRS support patches (Sonika&Vandana)
- fbc patches from Paulo
- make sure our vblank callbacks aren't called when the pipes are off
- various patches all over

* tag 'drm-intel-next-2015-02-27' of git://anongit.freedesktop.org/drm-intel: (61 commits)
  drm/i915: Update DRIVER_DATE to 20150227
  drm/i915: Clarify obj->map_and_fenceable
  drm/i915/skl: Allow Y (and Yf) frame buffer creation
  drm/i915/skl: Update watermarks for Y tiling
  drm/i915/skl: Updated watermark programming
  drm/i915/skl: Adjust get_plane_config() to support Yb/Yf tiling
  drm/i915/skl: Teach pin_and_fence_fb_obj() about Y tiling constraints
  drm/i915/skl: Adjust intel_fb_align_height() for Yb/Yf tiling
  drm/i915/skl: Allow scanning out Y and Yf fbs
  drm/i915/skl: Add new displayable tiling formats
  drm/i915: Remove DRIVER_MODESET checks from modeset code
  drm/i915: Remove regfile code&data for UMS suspend/resume
  drm/i915: Remove DRIVER_MODESET checks from gem code
  drm/i915: Remove DRIVER_MODESET checks in the gpu reset code
  drm/i915: Remove DRIVER_MODESET checks from suspend/resume code
  drm/i915: Remove DRIVER_MODESET checks in load/unload/close code
  drm/i915: fix a printk format
  drm/i915: Add media rc6 residency file to sysfs
  drm/i915: Add missing description to parameter in alloc_pt_range
  drm/i915: Removed the read of RP_STATE_CAP from sysfs/debugfs functions
  ...

9 years agodrm: Pass in new and old plane state to prepare_fb and cleanup_fb
Tvrtko Ursulin [Tue, 3 Mar 2015 14:22:31 +0000 (14:22 +0000)]
drm: Pass in new and old plane state to prepare_fb and cleanup_fb

Use cases like rotation require these hooks to have some context so they
know how to prepare and cleanup the frame buffer correctly.

For i915 specifically, object backing pages need to be mapped differently
for different rotation modes and the driver needs to know which mapping to
instantiate and which to tear down when transitioning between them.

v2: Made passed in states const. (Daniel Vetter)

[airlied: add mdp5 and atmel fixups]
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoMerge tag 'drm-intel-next-2015-02-14' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Wed, 4 Mar 2015 23:41:09 +0000 (09:41 +1000)]
Merge tag 'drm-intel-next-2015-02-14' of git://anongit.freedesktop.org/drm-intel into drm-next

- use the atomic helpers for plane_upate/disable hooks (Matt Roper)
- refactor the initial plane config code (Damien)
- ppgtt prep patches for dynamic pagetable alloc (Ben Widawsky, reworked and
  rebased by a lot of other people)
- framebuffer modifier support from Tvrtko Ursulin, drm core code from Rob Clark
- piles of workaround patches for skl from Damien and Nick Hoath
- vGPU support for xengt on the client side (Yu Zhang)
- and the usual smaller things all over

* tag 'drm-intel-next-2015-02-14' of git://anongit.freedesktop.org/drm-intel: (88 commits)
  drm/i915: Update DRIVER_DATE to 20150214
  drm/i915: Remove references to previously removed UMS config option
  drm/i915/skl: Use a LRI for WaDisableDgMirrorFixInHalfSliceChicken5
  drm/i915/skl: Fix always true comparison in a revision id check
  drm/i915/skl: Implement WaEnableLbsSlaRetryTimerDecrement
  drm/i915/skl: Implement WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken
  drm/i915: Add process identifier to requests
  drm/i915/skl: Implement WaBarrierPerformanceFixDisable
  drm/i915/skl: Implement WaCcsTlbPrefetchDisable:skl
  drm/i915/skl: Implement WaDisableChickenBitTSGBarrierAckForFFSliceCS
  drm/i915/skl: Implement WaDisableHDCInvalidation
  drm/i915/skl: Implement WaDisableLSQCROPERFforOCL
  drm/i915/skl: Implement WaDisablePartialResolveInVc
  drm/i915/skl: Introduce a SKL specific init_workarounds()
  drm/i915/skl: Document that we implement WaRsClearFWBitsAtReset
  drm/i915/skl: Implement WaSetGAPSunitClckGateDisable
  drm/i915/skl: Make the init clock gating function skylake specific
  drm/i915/skl: Provide a gen9 specific init_render_ring()
  drm/i915/skl: Document the WM read latency W/A with its name
  drm/i915/skl: Also detect eDRAM on SKL
  ...

9 years agoMerge branch 'drm-atmel-hlcdc-devel' of git://github.com/bbrezillon/linux-at91 into...
Dave Airlie [Wed, 4 Mar 2015 23:38:56 +0000 (09:38 +1000)]
Merge branch 'drm-atmel-hlcdc-devel' of git://github.com/bbrezillon/linux-at91 into drm-next

This pull request includes:
- support for Atomic mode setting
- the discard area optimization
- implementation of PM primitives

* 'drm-atmel-hlcdc-devel' of git://github.com/bbrezillon/linux-at91:
  drm: atmel-hlcdc: Add pinctrl PM select sleep,default state in CRTC suspend/resume
  drm: atmel-hlcdc: Add PM suspend/resume support
  drm: atmel-hlcdc: add discard area support
  drm: atmel-hlcdc: Atomic mode-setting conversion

9 years agoMerge branch 'drm/next/atomic' of git://linuxtv.org/pinchartl/fbdev into drm-next
Dave Airlie [Wed, 4 Mar 2015 23:37:58 +0000 (09:37 +1000)]
Merge branch 'drm/next/atomic' of git://linuxtv.org/pinchartl/fbdev into drm-next

rcar-du atomic modesetting support
* 'drm/next/atomic' of git://linuxtv.org/pinchartl/fbdev: (32 commits)
  drm: rcar-du: Fix race condition in hardware plane allocator
  drm: rcar-du: Move group locking inside rcar_du_crtc_update_planes()
  drm: rcar-du: Move plane commit code from CRTC start to CRTC resume
  drm: rcar-du: Move plane format to plane state
  drm: rcar-du: Remove unneeded rcar_du_crtc plane field
  drm: rcar-du: Replace plane crtc and enabled fields by plane state
  drm: rcar-du: Rework plane setup code
  drm: rcar-du: Switch plane set_property to atomic helpers
  drm: rcar-du: Switch page flip to atomic helpers
  drm: rcar-du: Implement asynchronous commit support
  drm: rcar-du: Replace encoder mode_fixup with atomic_check
  drm: rcar-du: Switch connector DPMS to atomic helpers
  drm: rcar-du: Switch mode config to atomic helpers
  drm: rcar-du: Switch plane update to atomic helpers
  drm: rcar-du: Rework CRTC enable/disable for atomic updates
  drm: rcar-du: Rework HDMI encoder enable/disable for atomic updates
  drm: rcar-du: Rework encoder enable/disable for atomic updates
  drm: rcar-du: Replace LVDS encoder DPMS by enable/disable
  drm: rcar-du: Remove private copy of plane size and position
  drm: rcar-du: Wire up atomic state object scaffolding
  ...

9 years agoMerge branch 'drm-tda998x-devel' of git://ftp.arm.linux.org.uk/~rmk/linux-arm into...
Dave Airlie [Wed, 4 Mar 2015 23:37:19 +0000 (09:37 +1000)]
Merge branch 'drm-tda998x-devel' of git://ftp.arm.linux.org.uk/~rmk/linux-arm into drm-next

A number of TDA998x updates for the next merge window.  Patches
included in this set are:
* adding support for finding the attached CRTCs from DT
* a fix function name mis-spelling in a dev_err()
* simplify the EDID reading by using the drm_do_get_edid() function
  instead of coding this ourselves.

* 'drm-tda998x-devel' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  drm/i2c: tda998x: use drm_do_get_edid()
  drm/i2c: tda998x: fix misspelling of current function in string
  drm/i2c: tda998x: add OF support for finding attached CRTCs

9 years agodrm: rcar-du: Fix race condition in hardware plane allocator
Laurent Pinchart [Wed, 25 Feb 2015 16:27:19 +0000 (18:27 +0200)]
drm: rcar-du: Fix race condition in hardware plane allocator

The plane allocator has been inherently racy since the beginning of the
transition to atomic updates, as the allocator lock is released between
free plane check (at .atomic_check() time) and the reservation (at
.atomic_update() time).

To fix it, create a new allocator solely based on the atomic plane
states without keeping any external state and perform allocation in the
.atomic_check() handler. The core idea is to replace the free planes
bitmask with a collective knowledge based on the allocated hardware
plane(s) for each KMS plane. The allocator then loops over all plane
states to compute the free planes bitmask, allocates hardware planes
based on that bitmask, and stores the result back in the plane states.

For this to work we need to access the current state of planes not
touched by the atomic update. To ensure that it won't be modified, we
need to lock all planes using drm_atomic_get_plane_state(). This
effectively serializes atomic updates from .atomic_check() up to
completion, either when swapping the states if the check step has
succeeded, or when freeing the states if the check step has failed.

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Move group locking inside rcar_du_crtc_update_planes()
Laurent Pinchart [Mon, 23 Feb 2015 14:55:56 +0000 (16:55 +0200)]
drm: rcar-du: Move group locking inside rcar_du_crtc_update_planes()

Only the planes to CRTCs association control register DPTSR needs to be
protected by custom locking, don't hold the mutex around the whole code.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Move plane commit code from CRTC start to CRTC resume
Laurent Pinchart [Sun, 22 Feb 2015 23:39:13 +0000 (01:39 +0200)]
drm: rcar-du: Move plane commit code from CRTC start to CRTC resume

As the DRM core will commit plane states when performing atomic updates,
those don't need to be committed manually when the CRTC is started except
in the system resume code path.

However, the atomic plane commit step is currently performed between
mode set disable and mode set enable to mimick the legacy mode setting
operations order. This causes the device clocks to be disabled after
applying plane settings and reenabled when enabling the CRTC,
potentially losing hardware in between.

Reorder the operations to enable the CRTC first and only then apply
plane settings, removing the need to manage clocks in the atomic begin
and flush handlers. We can then move the plane state commit code out of
the CRTC start handler to the system resume handler.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Move plane format to plane state
Laurent Pinchart [Mon, 23 Feb 2015 00:59:35 +0000 (02:59 +0200)]
drm: rcar-du: Move plane format to plane state

The format stored in the rcar_du_plane structure is part of the plane
state. Move it to the rcar_du_plane_state structure and precompute it in
the .atomic_check() handler.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Remove unneeded rcar_du_crtc plane field
Laurent Pinchart [Mon, 23 Feb 2015 01:20:39 +0000 (03:20 +0200)]
drm: rcar-du: Remove unneeded rcar_du_crtc plane field

The rcar_du_crtc plane field is only used to check for an error that
can't occur. Remove it.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Replace plane crtc and enabled fields by plane state
Laurent Pinchart [Sun, 22 Feb 2015 17:24:59 +0000 (19:24 +0200)]
drm: rcar-du: Replace plane crtc and enabled fields by plane state

The crtc and enabled fields duplicates information stored in the plane
state. Use the plane state instead and remove the fields.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Rework plane setup code
Laurent Pinchart [Sun, 22 Feb 2015 23:25:19 +0000 (01:25 +0200)]
drm: rcar-du: Rework plane setup code

Now that the plane setup code isn't called outside of the plane
implementation, it can be simplified by merging the
rcar_du_plane_compute_base() and rcar_du_plane_update_base() functions.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Switch plane set_property to atomic helpers
Laurent Pinchart [Mon, 23 Feb 2015 00:36:31 +0000 (02:36 +0200)]
drm: rcar-du: Switch plane set_property to atomic helpers

Allow setting up plane properties atomically using the plane
set_property atomic helper. The properties are now stored in the plane
state (requiring subclassing it) and applied when updating the planes.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Switch page flip to atomic helpers
Laurent Pinchart [Sun, 22 Feb 2015 23:04:21 +0000 (01:04 +0200)]
drm: rcar-du: Switch page flip to atomic helpers

The atomic page flip helper implements the page flip operation using
asynchronous commits.

As the legacy page flip was the last CRTC operation that needed direct
access to plane setup, the plane setup functions can now become private
to the plane implementation.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Implement asynchronous commit support
Laurent Pinchart [Sun, 22 Feb 2015 23:02:15 +0000 (01:02 +0200)]
drm: rcar-du: Implement asynchronous commit support

Implement a custom .atomic_commit() handler that supports asynchronous
commits using a work queue. This can be used for userspace-driven
asynchronous commits, as well as for an atomic page flip implementation.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Replace encoder mode_fixup with atomic_check
Laurent Pinchart [Sun, 22 Feb 2015 19:02:39 +0000 (21:02 +0200)]
drm: rcar-du: Replace encoder mode_fixup with atomic_check

The encoder .mode_fixup() operation is legacy, atomic updates uses the
new .atomic_check() operation. Convert the encoders drivers.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Switch connector DPMS to atomic helpers
Laurent Pinchart [Sat, 21 Feb 2015 23:49:11 +0000 (01:49 +0200)]
drm: rcar-du: Switch connector DPMS to atomic helpers

The atomic connector DPMS helper implements the connector DPMS operation
using atomic commit, removing the need for DPMS helper operations on
CRTCs and encoders.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Switch mode config to atomic helpers
Laurent Pinchart [Fri, 20 Feb 2015 13:16:55 +0000 (15:16 +0200)]
drm: rcar-du: Switch mode config to atomic helpers

This removes the legacy mode config code. The CRTC and encoder prepare
and commit operations are not used anymore, remove them.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Switch plane update to atomic helpers
Laurent Pinchart [Fri, 20 Feb 2015 11:18:56 +0000 (13:18 +0200)]
drm: rcar-du: Switch plane update to atomic helpers

This removes the legacy plane update code. Wire up the default atomic
check and atomic commit mode config helpers as needed by the plane
update atomic helpers.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Rework CRTC enable/disable for atomic updates
Laurent Pinchart [Fri, 20 Feb 2015 12:05:21 +0000 (14:05 +0200)]
drm: rcar-du: Rework CRTC enable/disable for atomic updates

When using atomic updates the CRTC .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and rework .disable() to not depend on
DPMS, easing DPMS removal later on.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Rework HDMI encoder enable/disable for atomic updates
Laurent Pinchart [Fri, 20 Feb 2015 12:05:21 +0000 (14:05 +0200)]
drm: rcar-du: Rework HDMI encoder enable/disable for atomic updates

When using atomic updates the encoder .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and .disable() and rework .prepare(),
.commit() and .dpms() as wrappers around .enable() and .disable(),
easing their future removal.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Rework encoder enable/disable for atomic updates
Laurent Pinchart [Fri, 20 Feb 2015 12:05:21 +0000 (14:05 +0200)]
drm: rcar-du: Rework encoder enable/disable for atomic updates

When using atomic updates the encoder .enable() and .disable() helper
operations are preferred over the (then legacy) .prepare() and .commit()
operations. Implement .enable() and .disable() and rework .prepare(),
.commit() and .dpms() as wrappers around .enable() and .disable(),
easing their future removal.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Replace LVDS encoder DPMS by enable/disable
Laurent Pinchart [Fri, 20 Feb 2015 12:59:58 +0000 (14:59 +0200)]
drm: rcar-du: Replace LVDS encoder DPMS by enable/disable

The LVDS encoder doesn't support DPMS states, replace the DPMS operation
by enable/disable to avoid propagating DPMS states down to the encoder
code.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Remove private copy of plane size and position
Laurent Pinchart [Fri, 20 Feb 2015 13:58:38 +0000 (15:58 +0200)]
drm: rcar-du: Remove private copy of plane size and position

The plane source and destination size and positions are stored in the
plane state, and a private copy is kept in the rcar_du_plane objects.
Remove the private copy as it just duplicates the state.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Wire up atomic state object scaffolding
Laurent Pinchart [Fri, 20 Feb 2015 09:30:59 +0000 (11:30 +0200)]
drm: rcar-du: Wire up atomic state object scaffolding

Hook up the default .reset(), .atomic_duplicate_state() and
.atomic_free_state() helpers to ensure that state objects are properly
created and destroyed, and call drm_mode_config_reset() at init time to
create the initial state objects.

Framebuffer reference count also gets maintained automatically by the
transitional helpers except for the legacy page flip operation. Maintain
it explicitly there.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Handle primary plane config through atomic plane ops
Laurent Pinchart [Wed, 18 Feb 2015 13:47:27 +0000 (15:47 +0200)]
drm: rcar-du: Handle primary plane config through atomic plane ops

Use the new CRTC atomic transitional helpers drm_helper_crtc_mode_set()
and drm_helper_crtc_mode_set_base() to implement the CRTC .mode_set and
.mode_set_base operations. This delegates primary plane configuration to
the plane .atomic_update and .atomic_disable operations, removing
duplicate code from the CRTC implementation.

There is now no code path available to the driver in which to drop the
reference to the CRTC acquired in the .prepare() operation if an error
then occurs. The driver thus now leaks a reference if an error occurs
during mode set. So be it, this will be fixed in a further step of the
atomic update transition.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Implement planes atomic operations
Laurent Pinchart [Wed, 18 Feb 2015 10:18:05 +0000 (12:18 +0200)]
drm: rcar-du: Implement planes atomic operations

Implement the CRTC .atomic_begin() and .atomic_flush() operations, the
plane .atomic_check(), .atomic_update() and operations, and use the
transitional atomic helpers to implement the plane update and disable
operations on top of the new atomic operations.

The plane setup code can't be moved out of the CRTC start function
completely yet, as the atomic code paths are not taken every time the
CRTC needs to be started. This results in some code duplication that
will be fixed after switching to atomic updates completely.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Fix hardware plane allocation
Laurent Pinchart [Wed, 25 Feb 2015 16:38:25 +0000 (18:38 +0200)]
drm: rcar-du: Fix hardware plane allocation

The hardware plane allocator loops over all planes to find free
candidates. However, instead of looping over the number of hardware
planes, it loops over the number of software planes, which happens to be
larger by one unit. This has no effect in practise as the extra plane is
always cleared in the mask of free planes, but it should still be fixed
for correctness.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Implement universal plane support
Laurent Pinchart [Tue, 17 Feb 2015 16:34:17 +0000 (18:34 +0200)]
drm: rcar-du: Implement universal plane support

Explicitly create the CRTC primary plane instead of relying on the core
helpers to do so. This simplifies the plane logic by merging the KMS and
software planes.

Reject plane API operations on the primary planes for now, as that code
will anyway be refactored when implementing support for atomic updates.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Define macros for the max number of groups, CRTCs and LVDS
Laurent Pinchart [Wed, 25 Feb 2015 16:21:12 +0000 (18:21 +0200)]
drm: rcar-du: Define macros for the max number of groups, CRTCs and LVDS

Let's avoid magic constants. Beside increasing code readability, it will
also ensure that no location will be forgotten when raising the maximum
number of groups, CRTCs or LVDS encoders

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Disable fbdev emulation when no connector is present
Laurent Pinchart [Tue, 24 Feb 2015 01:51:26 +0000 (03:51 +0200)]
drm: rcar-du: Disable fbdev emulation when no connector is present

fbdev emulation requires at least one connector, and will fail to
initialize if no connector has been successfully instantiated. Disable
it in that case and print an informational message instead of failing
probe with a confusing fbdev emulation error message.

It could be argued that probe should fail when no connector is present,
but the DU could still be useful in that case with the to-be-implemented
memory write-back support.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Turn vblank on/off when enabling/disabling CRTC
Laurent Pinchart [Wed, 18 Feb 2015 11:14:46 +0000 (13:14 +0200)]
drm: rcar-du: Turn vblank on/off when enabling/disabling CRTC

The DRM core vblank handling mechanism requires drivers to forcefully
turn vblank reporting off when disabling the CRTC, and to restore the
vblank reporting status when enabling the CRTC.

Implement this using the drm_crtc_vblank_on/off helpers. When disabling
vblank we must first wait for page flips to complete, so implement page
flip completion wait as well.

Finally, drm_crtc_vblank_off() must be called at startup to synchronize
the state of the vblank core code with the hardware, which is initially
disabled. This is performed at CRTC creation time, requiring vertical
blanking to be initialized before creating CRTCs.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Wait for page flip completion when turning the CRTC off
Laurent Pinchart [Wed, 18 Feb 2015 11:21:56 +0000 (13:21 +0200)]
drm: rcar-du: Wait for page flip completion when turning the CRTC off

Turning a CRTC off will prevent a queued page flip from ever completing,
potentially confusing userspace. Wait for queued page flips to complete
before turning the CRTC off to avoid this.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Reorder CRTC functions
Laurent Pinchart [Wed, 18 Feb 2015 11:42:40 +0000 (13:42 +0200)]
drm: rcar-du: Reorder CRTC functions

The next commit will need functions to be reordered to avoid forward
declarations. Do it separately to help review.

This only moves functions without any change to the code.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Don't set connector->encoder at init time
Laurent Pinchart [Mon, 23 Feb 2015 11:22:44 +0000 (13:22 +0200)]
drm: rcar-du: Don't set connector->encoder at init time

The drm_connector encoder field points to the encoder driving the
connector. No such association exists at init time, as all pipelines are
disabled. Don't set the field.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Remove drm_fbdev_cma_restore_mode() call at init time
Laurent Pinchart [Fri, 20 Feb 2015 20:35:52 +0000 (22:35 +0200)]
drm: rcar-du: Remove drm_fbdev_cma_restore_mode() call at init time

The function is meant to restore the fbdev mode in the lastclose
handler, not to be called at init time. Remove the call.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm: rcar-du: Don't disable unused functions at init time
Laurent Pinchart [Fri, 20 Feb 2015 15:46:14 +0000 (17:46 +0200)]
drm: rcar-du: Don't disable unused functions at init time

All encoders and CRTCs start disabled, re-disabling them is a no-op.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
9 years agodrm/i915: Update DRIVER_DATE to 20150227
Daniel Vetter [Fri, 27 Feb 2015 18:12:46 +0000 (19:12 +0100)]
drm/i915: Update DRIVER_DATE to 20150227

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Clarify obj->map_and_fenceable
Chris Wilson [Fri, 27 Feb 2015 13:58:43 +0000 (13:58 +0000)]
drm/i915: Clarify obj->map_and_fenceable

For an object right on the boundary of mappable space, as the fenceable
size is stricly greater than the actual size, its fence region may extend
out of mappable space.

Note that only pnv/g33 has fence_size > obj.size and an unmappable
range in the gtt, and there alignment constraints prevent bad things
from happening.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Clarify why this shouldn't change anything as per the
discussion on intel-gfx.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Allow Y (and Yf) frame buffer creation
Tvrtko Ursulin [Fri, 27 Feb 2015 11:15:24 +0000 (11:15 +0000)]
drm/i915/skl: Allow Y (and Yf) frame buffer creation

By this patch all underlying bits have been implemented and this
patch actually enables the feature.

v2: Validate passed in fb modifiers to reject garbage. (Daniel Vetter)
v3: Rearrange validation checks per code review comments. (Daniel Vetter)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Update watermarks for Y tiling
Tvrtko Ursulin [Fri, 27 Feb 2015 15:12:35 +0000 (15:12 +0000)]
drm/i915/skl: Update watermarks for Y tiling

Display watermarks need different programming for different tiling
modes.

Set the relevant flag so this happens during the plane commit and
add relevant data into a structure made available to the watermark
computation code.

v2: Pass in tiling info to sprite plane updates as well.
v3: Rebased for plane handling changes.
v4: Handle fb == NULL when plane is disabled.
v5: Refactored for addfb2 interface.
v6: Refactored for fb modifier changes.
v7: Updated for atomic commit by only updating watermarks when tiling changes.
v8: BSpec watermark calculation updates.
v9: Restrict scope of y_tile_minimum variable. (Damien Lespiau)
v10: Get fb from plane state otherwise we are working on old state.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Acked-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v9)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Updated watermark programming
Tvrtko Ursulin [Fri, 27 Feb 2015 11:15:22 +0000 (11:15 +0000)]
drm/i915/skl: Updated watermark programming

Recent BSpect updates have changed the watermark calculation to avoid
display flickering in some cases.

v2: Fix check against DDB allocation and tidy the code a bit. (Damien Lespiau)

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Adjust get_plane_config() to support Yb/Yf tiling
Damien Lespiau [Fri, 27 Feb 2015 11:15:21 +0000 (11:15 +0000)]
drm/i915/skl: Adjust get_plane_config() to support Yb/Yf tiling

v2: Rebased for addfb2 interface and consolidated a bit. (Tvrtko Ursulin)
v3: Rebased for fb modifier changes. (Tvrtko Ursulin)
v4: Use intel_fb_stride_alignment instead of open coding. (Damien Lespiau)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Teach pin_and_fence_fb_obj() about Y tiling constraints
Damien Lespiau [Fri, 27 Feb 2015 11:15:20 +0000 (11:15 +0000)]
drm/i915/skl: Teach pin_and_fence_fb_obj() about Y tiling constraints

1Mb!

v2: Rebased for addfb2 interface. (Tvrtko Ursulin)
v3: Rebased for fb modifier changes. (Tvrtko Ursulin)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Adjust intel_fb_align_height() for Yb/Yf tiling
Damien Lespiau [Fri, 27 Feb 2015 11:15:19 +0000 (11:15 +0000)]
drm/i915/skl: Adjust intel_fb_align_height() for Yb/Yf tiling

We now need the bpp of the fb as Yf tiling has different tile widths
depending on it.

v2: Rebased for the new addfb2 interface. (Tvrtko Ursulin)
v3: Rebased for fb modifier changes. (Tvrtko Ursulin)
v4: Added missing case and 128-bit pixel warning. (Damien Lespiau)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v3)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Allow scanning out Y and Yf fbs
Damien Lespiau [Fri, 27 Feb 2015 11:15:18 +0000 (11:15 +0000)]
drm/i915/skl: Allow scanning out Y and Yf fbs

Skylake is able to scannout those tiling formats. We need to allow them
in the ADDFB ioctl and tell the harware about it.

v2: Rebased for addfb2 interface. (Tvrtko Ursulin)
v3: Rebased for fb modifier changes. (Tvrtko Ursulin)
v4: Don't allow Y tiled fbs just yet. (Tvrtko Ursulin)
v5: Check for stride alignment and max pitch. (Tvrtko Ursulin)
v6: Simplify maximum pitch check. (Ville Syrjälä)
v7: Drop the gen9 check since requirements are no different. (Ville Syrjälä)
v8: Gen2 has different X tiling stride. (Ville Syrjälä)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v7)
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Add new displayable tiling formats
Tvrtko Ursulin [Fri, 27 Feb 2015 11:15:17 +0000 (11:15 +0000)]
drm/i915/skl: Add new displayable tiling formats

Starting with SKL display engine can scan out Y, and newly introduced Yf
tiling formats so add the latter to the frame buffer modifier space.

v2: Definitions moved to drm_fourcc.h.
v3: Try to document the format better.

Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove DRIVER_MODESET checks from modeset code
Daniel Vetter [Mon, 23 Feb 2015 11:03:31 +0000 (12:03 +0100)]
drm/i915: Remove DRIVER_MODESET checks from modeset code

Mostly just checks in i915-private modeset ioctls.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove regfile code&data for UMS suspend/resume
Daniel Vetter [Mon, 23 Feb 2015 11:03:30 +0000 (12:03 +0100)]
drm/i915: Remove regfile code&data for UMS suspend/resume

Lots of lines to remove!

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
[danvet: Fixup makefile.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove DRIVER_MODESET checks from gem code
Daniel Vetter [Mon, 23 Feb 2015 11:03:29 +0000 (12:03 +0100)]
drm/i915: Remove DRIVER_MODESET checks from gem code

Hooray!

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove DRIVER_MODESET checks in the gpu reset code
Daniel Vetter [Mon, 23 Feb 2015 11:03:27 +0000 (12:03 +0100)]
drm/i915: Remove DRIVER_MODESET checks in the gpu reset code

Again, good riddance to UMS!

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove DRIVER_MODESET checks from suspend/resume code
Daniel Vetter [Mon, 23 Feb 2015 11:03:26 +0000 (12:03 +0100)]
drm/i915: Remove DRIVER_MODESET checks from suspend/resume code

UMS is dead, yay!

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Remove DRIVER_MODESET checks in load/unload/close code
Daniel Vetter [Mon, 23 Feb 2015 11:03:25 +0000 (12:03 +0100)]
drm/i915: Remove DRIVER_MODESET checks in load/unload/close code

UMS is gone, this is dead code.

Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: fix a printk format
Dan Carpenter [Thu, 26 Feb 2015 16:53:54 +0000 (19:53 +0300)]
drm/i915: fix a printk format

This printk leads to the following Smatch warning:

drivers/gpu/drm/i915/i915_gem_gtt.c:336 alloc_pt_range()
error: '%pa' expects argument of type 'phys_addr_t*',
argument 5 has type 'struct i915_page_table_entry*'

It looks like a simple typo to me where "%p" was intended instead of
"%pa".

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Add media rc6 residency file to sysfs
Ville Syrjälä [Thu, 26 Feb 2015 15:40:27 +0000 (21:10 +0530)]
drm/i915: Add media rc6 residency file to sysfs

On VLV/CHV the media well rc6 residency gets reported separately
from the render well, so add another file to sysfs so that we can
report the residency to the user.

Testcase: igt/pm_rc6_residency --run-subtest media-rc6-accuracy
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Add missing description to parameter in alloc_pt_range
Michel Thierry [Thu, 26 Feb 2015 11:28:13 +0000 (11:28 +0000)]
drm/i915: Add missing description to parameter in alloc_pt_range

The patch "drm/i915: Plumb drm_device through page tables operations"
added an extra parameter, but it didn't update the function description.
Also remove unnecessary blank line added by the same patch.

Found by kbuild test robot.

Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Removed the read of RP_STATE_CAP from sysfs/debugfs functions
Akash Goel [Thu, 26 Feb 2015 10:39:47 +0000 (16:09 +0530)]
drm/i915: Removed the read of RP_STATE_CAP from sysfs/debugfs functions

The frequency values(Rp0, Rp1, Rpn) reported by RP_STATE_CAP register
are stored, initially by the Driver, inside the dev_priv->rps structure.
Since these values are expected to remain same throughout, there is no real
need to read this register, on dynamic basis, from certain debugfs/sysfs
functions and the values can be instead retrieved from the dev_priv->rps
structure when needed.
For the i915_frequency_info debugfs interface, the frequency values from the
RP_STATE_CAP register only should be used, to indicate the actual Hw state,
since it is principally used for the debugging purpose.

v2: Reverted the changes in i915_frequency_info function, to continue report
    back the frequency values, as per the actual Hw state (Chris)

Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Look at staged config when fixing pipe_src_w for LVDS
Ander Conselvan de Oliveira [Thu, 26 Feb 2015 07:44:45 +0000 (09:44 +0200)]
drm/i915: Look at staged config when fixing pipe_src_w for LVDS

The code in function intel_crtc_compute_config() that evens pipe_src_w
if necessary would look at the current config instead of the staged one
when deciding if there is an LVDS encoder in use. This could potentially
lead to the value not being updated, if during the modeset a crtc wasn't
driving an LVDS encoder.

Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agoMerge tag 'topic/drm-misc-2015-02-25' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Thu, 26 Feb 2015 00:32:55 +0000 (10:32 +1000)]
Merge tag 'topic/drm-misc-2015-02-25' of git://anongit.freedesktop.org/drm-intel into drm-next

misc atomic and dp macros

* tag 'topic/drm-misc-2015-02-25' of git://anongit.freedesktop.org/drm-intel:
  drm: Adding edp1.4 specific dpcd macros
  drm/atomic-helpers: make mode_set hooks optional
  drm/atomic-helper: Rename commmit_post/pre_planes
  drm/atomic: Rename drm_atomic_helper_commit_pre_planes() state argument
  drm: If available use atomic state in getcrtc ioctl
  drm: Add DRM_DEBUG_ATOMIC
  drm/atomic-helpers: Fix documentation typos and wrong copy&paste
  drm: Fix the CRTC_STEREO_DOUBLE_ONLY define to include stero modes
  drm: Fix drm_crtc_vblank_get() documentation

9 years agodrm/i915: Ensure crtc_state backpointer is always initialized
Matt Roper [Wed, 25 Feb 2015 19:43:26 +0000 (11:43 -0800)]
drm/i915: Ensure crtc_state backpointer is always initialized

As we transition to full atomic modesetting, we want to be able to pass
intel_crtc_state around in various places that we pass intel_crtc
directly today.  Ensure that the ->crtc backpointer is properly
initialized in case we need to get back to the associated CRTC.

Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Use enabled value from crtc_state rather than crtc (v2)
Matt Roper [Wed, 25 Feb 2015 21:12:16 +0000 (13:12 -0800)]
drm/i915: Use enabled value from crtc_state rather than crtc (v2)

As vendors transition their drivers from legacy to atomic there's some
duplication of data between drm_crtc and drm_crtc_state (since
unconverted drivers likely won't have a state structure).

i915 is partially converted and does have a crtc->state structure, but
still uses direct crtc fields internally in many places, which causes
the two sets of data to get out of sync.  As of commit

        commit 31c946e85ce6b48ce0f25e3cdca8362e4fe8b300
        Author: Daniel Vetter <daniel.vetter@ffwll.ch>
        Date:   Sun Feb 22 12:24:17 2015 +0100

            drm: If available use atomic state in getcrtc ioctl

            This way drivers fully converted to atomic don't need to update these
            legacy state variables in their modeset code any more.

Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
the DRM core starts assuming that the presence of a ->state structure
implies that it should make use of the values stored there which, on
i915, leads to the core code using stale values for CRTC 'enabled'
status.

Let's switch over to using the state value of 'enable' internally rather
than using the drm_crtc field.  This ensures that our driver internals
are working from the same data that the DRM core is, avoiding
mismatches.

This patch was generated with Coccinelle using the following semantic
patch:

        <smpl>
        @@
        struct drm_crtc C;
        struct drm_crtc *CP;
        @@
        (
        - C.enabled
        + C.state->enable
        |
        - CP->enabled
        + CP->state->enable
        )

        // For assignments, we still update the legacy value as well as the state value
        // so add an extra assignment statement for that.
        @@
        struct drm_crtc C;
        struct drm_crtc *CP;
        expression E;
        @@
        (
          C.state->enable = E;
        + C.enabled = E;
        |
          CP->state->enable = E;
        + CP->enabled = E;
        )
        </smpl>

The crtc->mode and crtc->hwmode fields should probably be transitioned
over as well eventually, but we seem to do an okay job of keeping those
up-to-date already so I want to minimize the changes that will clash
with Ander's in-progress atomic work.

v2: Don't remove the assignments to the legacy value when we assign to
    the state value.  A second cocci stanza takes care of adding the
    legacy assignment back where appropriate.  (Daniel)

Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Cache ringbuf pointer in request structure
John Harrison [Fri, 13 Feb 2015 11:48:12 +0000 (11:48 +0000)]
drm/i915: Cache ringbuf pointer in request structure

In execlist mode, the ringbuf is a function of the ring and context whereas in
legacy mode, it is derived from the ring alone. Thus the calculation required to
determine the ringbuf pointer from the ring (and context) also needs to test
execlist mode or not. This is messy.

Further, the request structure holds a pointer to both the ring and the context
for which it was created. Thus, given a request, it is possible to derive the
ringbuf in either legacy or execlist mode. Hence it is necessary to pass just
the request in to all the low level functions rather than some combination of
request, ring, context and ringbuf. However, rather than recalculating it each
time, it is much simpler to just cache the ringbuf pointer in the request
structure itself.

Caching the pointer means the calculation is done once at request creation time
and all further code and simply read it directly from the request structure.

OTC-Jira: VIZ-5115
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
[danvet: Drop contentless comment in lrc alloc request entirely. And
spelling fix in the commit message.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Add missing trace point to LRC execbuff code path
John Harrison [Fri, 13 Feb 2015 11:48:11 +0000 (11:48 +0000)]
drm/i915: Add missing trace point to LRC execbuff code path

There is a trace point in the legacy execbuffer execution path that is missing
from the execlist path. Trace points are extremely useful for debugging and are
used by various automated validation tests. Hence, this patch adds the missing
trace point back in.

OTC-Jira: VIZ-5115
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Rename 'flags' to 'dispatch_flags' for better code reading
John Harrison [Fri, 13 Feb 2015 11:48:10 +0000 (11:48 +0000)]
drm/i915: Rename 'flags' to 'dispatch_flags' for better code reading

There is a flags word that is passed through the execbuffer code path all the
way from initial decoding of the user parameters down to the very final dispatch
buffer call. It is simply called 'flags'. Unfortuantely, there are many other
flags words floating around in the same blocks of code. Even more once the GPU
scheduler arrives.

This patch makes it more obvious exactly which flags word is which by renaming
'flags' to 'dispatch_flags'. Note that the bit definitions for this flags word
already have an 'I915_DISPATCH_' prefix on them and so are not quite so
ambiguous.

OTC-Jira: VIZ-1587
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
[danvet: Resolve conflict with Chris' rework of the bb parsing.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Plumb drm_device through page tables operations
Michel Thierry [Tue, 24 Feb 2015 16:22:37 +0000 (16:22 +0000)]
drm/i915: Plumb drm_device through page tables operations

The next patch in the series will require it for alloc_pt_single.

v2: Rebased after s/page_tables/page_table/.

Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Create page table allocators
Ben Widawsky [Tue, 24 Feb 2015 16:22:36 +0000 (16:22 +0000)]
drm/i915: Create page table allocators

As we move toward dynamic page table allocation, it becomes much easier
to manage our data structures if break do things less coarsely by
breaking up all of our actions into individual tasks.  This makes the
code easier to write, read, and verify.

Aside from the dissection of the allocation functions, the patch
statically allocates the page table structures without a page directory.
This remains the same for all platforms,

The patch itself should not have much functional difference. The primary
noticeable difference is the fact that page tables are no longer
allocated, but rather statically declared as part of the page directory.
This has non-zero overhead, but things gain additional complexity as a
result.

This patch exists for a few reasons:
1. Splitting out the functions allows easily combining GEN6 and GEN8
code. Page tables have no difference based on GEN8. As we'll see in a
future patch when we add the DMA mappings to the allocations, it
requires only one small change to make work, and error handling should
just fall into place.

2. Unless we always want to allocate all page tables under a given PDE,
we'll have to eventually break this up into an array of pointers (or
pointer to pointer).

3. Having the discrete functions is easier to review, and understand.
All allocations and frees now take place in just a couple of locations.
Reviewing, and catching leaks should be easy.

4. Less important: the GFP flags are confined to one location, which
makes playing around with such things trivial.

v2: Updated commit message to explain why this patch exists

v3: For lrc, s/pdp.page_directory[i].daddr/pdp.page_directory[i]->daddr/

v4: Renamed free_pt/pd_single functions to unmap_and_free_pt/pd (Daniel)

v5: Added additional safety checks in gen8 clear/free/unmap.

v6: Use WARN_ON and return -EINVAL in alloc_pt_range (Mika).

v7: Make err_out loop symmetrical to the way we allocate in
alloc_pt_range. Also s/page_tables/page_table and correct commit
message (Mika)

Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v3+)
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Complete page table structures
Ben Widawsky [Tue, 24 Feb 2015 16:22:35 +0000 (16:22 +0000)]
drm/i915: Complete page table structures

Move the remaining members over to the new page table structures.

This can be squashed with the previous commit if desire. The reasoning
is the same as that patch. I simply felt it is easier to review if split.

v2: In lrc: s/ppgtt->pd_dma_addr[i]/ppgtt->pdp.page_directory[i].daddr/
v3: Rebase.
v4: Rebased after s/page_tables/page_table/.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2+)
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: page table abstractions
Ben Widawsky [Tue, 24 Feb 2015 16:22:34 +0000 (16:22 +0000)]
drm/i915: page table abstractions

When we move to dynamic page allocation, keeping page_directory and pagetabs as
separate structures will help to break actions into simpler tasks.

To help transition the code nicely there is some wasted space in gen6/7.
This will be ameliorated shortly.

Following the x86 pagetable terminology:
PDPE = struct i915_page_directory_pointer_entry.
PDE = struct i915_page_directory_entry [page_directory].
PTE = struct i915_page_table_entry [page_tables].

v2: fixed mismatches after clean-up/rebase.

v3: Clarify the names of the multiple levels of page tables (Daniel)

v4: Addressing Mika's review comments.
s/gen8_free_page_directories/gen8_free_page_directory and free the
page tables for the directory there.
In gen8_ppgtt_allocate_page_directories, do not leak previously allocated
pt in case the page_directory alloc fails.
Update error return handling in gen8_ppgtt_alloc.

v5: Do not leak pt on error in gen6_ppgtt_allocate_page_tables. (Mika)

v6: s/page_tables/page_table/. (Mika)

Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Michel Thierry <michel.thierry@intel.com> (v2+)
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Add support for edp1.4 low vswing
Sonika Jindal [Wed, 25 Feb 2015 04:59:12 +0000 (10:29 +0530)]
drm/i915/skl: Add support for edp1.4 low vswing

Based upon vbt's vswing preemph settings value select the appropriate
translations for edp.

v2: Incorporating bspec changes for vswing and preemph levels, adding edp
translation table. Removed HSW from selection 9 which is specific to skl and
correcting the returning of level2 from max pre emph (Damien)

v3: Rebasing on top of renaming patches. Adding level(3,0) since level(2,2) as
mentioned in bspec is invalid as per edp spec. Also changed the determining of
size of the table selected (Satheesh).

v4: Adding level 3 in max voltage selection if low vswing is selected (Satheesh)

v5: Add a comment stating that skl_ddi_translations_edp is for eDP 1.4
    low vswing panels.

v6: Updating recommended DDI translation table for edp 1.4

Reviewed-by: Satheeshakrishna M <satheeshakrishna.m@intel.com> (v4)
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v6)
Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Support for edp low_vswing param in vbt
Sonika Jindal [Wed, 25 Feb 2015 04:59:11 +0000 (10:29 +0530)]
drm/i915/skl: Support for edp low_vswing param in vbt

v2: Adding VBT version check for low_vswing field, and correcting parsing

v3: (Damien)
 - Restrain the scope of the 'vswing' variable
 - Use the more idiomatic "ev_priv->vbt.edp_low_vswing = vswing == 0;"
  instead of if (foo) var = true; else var = false;
 - Shorten edp_vswing_premph_setting to edp_vswing_premph to fit in 80 chars
 - Add the version from which the edp_vswing_premph field is valid in the
  struct definition

Reviewed-by: Satheeshakrishna M <satheeshakrishna.m@intel.com> (v2)
Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: cleanup some indenting
Dan Carpenter [Wed, 25 Feb 2015 13:17:48 +0000 (16:17 +0300)]
drm/i915: cleanup some indenting

Static checkers complain that we should probably add curly braces
because, from the indenting, it looks like seq_printf() should be inside
the list_for_each_entry() loop.  But the code is actually correct, it's
just the indenting which is off.

Besides fixing the indenting on seq_printf(), I did add curly braces,
because generally mult-line indents should have curly braces to make
them more readable.

The unintended indent was left behind and not unindented in

commit d7f46fc4e7323887494db13f063a8e59861fefb0
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Fri Dec 6 14:10:55 2013 -0800

    drm/i915: Make pin count per VMA
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Make sure to allocate mininum sizes in the DDB
Damien Lespiau [Mon, 9 Feb 2015 13:35:10 +0000 (13:35 +0000)]
drm/i915/skl: Make sure to allocate mininum sizes in the DDB

I overlooked the fact that we need to allocate a minimum 8 blocks and
that just allocating the planes depending on how much they need to fetch
from the DDB in proportion of how much memory bw is necessary for the
whole display can lead to cases where we don't respect those minima (and
thus overrun).

So, instead, start by allocating 8 blocks to each active display plane
and then allocate the remaining blocks like before.

v2: Rebase on top of -nightly

Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Enhancing eDP DRRS debug message
Ramalingam C [Mon, 23 Feb 2015 12:08:33 +0000 (17:38 +0530)]
drm/i915: Enhancing eDP DRRS debug message

When Downclock mode is not found, the same info is added to the
corresponding debug log.

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agoDocumentation/drm: DocBook integration for DRRS
Vandana Kannan [Fri, 13 Feb 2015 10:03:03 +0000 (15:33 +0530)]
Documentation/drm: DocBook integration for DRRS

Adding an overview of DRRS in general and the implementation for eDP DRRS.
Also, describing the functions related to eDP DRRS.

Signed-off-by: Vandana Kannan <vandana.kannan@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Enable eDP DRRS for CHV
Durgadoss R [Fri, 13 Feb 2015 10:03:02 +0000 (15:33 +0530)]
drm/i915: Enable eDP DRRS for CHV

This patch enables eDP DRRS for CHV by adding the
required IS_CHERRYVIEW() checks.
CHV uses the same register bit as VLV.

[Vandana]: Since CHV has 2 sets of M_N registers, it will follow the same code
path as gen < 8. Added CHV check in dp_set_m_n()

[Ram]: Rebased on top of previous patch modifications

Signed-off-by: Durgadoss R <durgadoss.r@intel.com>
Signed-off-by: Vandana Kannan <vandana.kannan@intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Support for RR switching on VLV
Vandana Kannan [Fri, 13 Feb 2015 10:03:01 +0000 (15:33 +0530)]
drm/i915: Support for RR switching on VLV

Definition of VLV RR switch bit and corresponding toggling in
set_drrs function.

Signed-off-by: Vandana Kannan <vandana.kannan@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/bdw: Add support for DRRS to switch RR
Vandana Kannan [Fri, 13 Feb 2015 10:03:00 +0000 (15:33 +0530)]
drm/i915/bdw: Add support for DRRS to switch RR

For Broadwell, there is one instance of Transcoder MN values per transcoder.
For dynamic switching between multiple refreshr rates, M/N values may be
reprogrammed on the fly. Link N programming triggers update of all data and
link M & N registers and the new M/N values will be used in the next frame
that is output.

V2: [By Ram]: intel_dp_set_m_n() is rewritten to accommodate
gen >= 8 [Rodrigo]
V3: Coding style correction [Ram]
V4: [By Ram] intel_dp_set_m_n modifications are moved into a
separate patch, retaining only DRRS related changes here [Rodrigo]

Signed-off-by: Vandana Kannan <vandana.kannan@intel.com>
Signed-off-by: Pradeep Bhat <pradeep.bhat@intel.com>
Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Add support for DRRS in intel_dp_set_m_n
Ramalingam C [Fri, 13 Feb 2015 10:02:59 +0000 (15:32 +0530)]
drm/i915: Add support for DRRS in intel_dp_set_m_n

Till Gen 7 we have two sets of M_N registers, but Gen 8 onwards
we have only one M_N register set. To support DRRS on both scenarios
a input parameter to intel_dp_set_m_n is added.

In case of DRRS, When platform provides two set of M_N registers for dp,
we can program them with two different dividers and switch between them.
But when only one such register set is provided, we have to program
the required divider M_N value on that registers itself.

Two enum members M1_N1 and M2_N2 are defined to represent the above
scenarios.

M1_N1        : Program dp_m_n on M1_N1 registers
dp_m2_n2 on M2_N2 registers (If supported)

M2_N2        : Program dp_m2_n2 on M1_N1 registers
M2_N2 registers are not supported

Signed-off-by: Ramalingam C <ramalingam.c@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Shift driver's HWSP usage out of reserved range
Thomas Daniel [Wed, 18 Feb 2015 11:48:21 +0000 (11:48 +0000)]
drm/i915: Shift driver's HWSP usage out of reserved range

As of Gen6, the general purpose area of the hardware status page has shrunk and
now begins at dword 0x30.  i915 driver uses dword 0x20 to store the seqno which
is now reserved.  So shift our HWSP dwords up into the general purpose range
before this bites us.

Note that all available documentation just says this is reserved
without going into details about what it's used for.

Signed-off-by: Thomas Daniel <thomas.daniel@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
[danvet: Add clarification from Thomas that unfortunately Bspec is
silent on what "reserverd" precisely means.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm: atmel-hlcdc: Add pinctrl PM select sleep,default state in CRTC suspend/resume
Sylvain Rochet [Sun, 22 Feb 2015 17:51:04 +0000 (18:51 +0100)]
drm: atmel-hlcdc: Add pinctrl PM select sleep,default state in CRTC suspend/resume

Some LCD panels have back-powering issue when un-powered, allows users
to use an alternate pinctrl "sleep" in order to clamp outputs to a
wanted state at suspend.

Signed-off-by: Sylvain Rochet <sylvain.rochet@finsecur.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
9 years agodrm: atmel-hlcdc: Add PM suspend/resume support
Sylvain Rochet [Sun, 22 Feb 2015 17:51:03 +0000 (18:51 +0100)]
drm: atmel-hlcdc: Add PM suspend/resume support

On suspend: switch off CRTC if not already suspended with runtime PM

On resume: switch on CRTC if we were not already suspended from runtime
PM while suspending.

Signed-off-by: Sylvain Rochet <sylvain.rochet@finsecur.com>
Reviewed-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
9 years agodrm: Adding edp1.4 specific dpcd macros
Sonika Jindal [Thu, 19 Feb 2015 07:46:44 +0000 (13:16 +0530)]
drm: Adding edp1.4 specific dpcd macros

Adding dpcd macros related to edp1.4 and link rates

v2: Added DP_SUPPORTED_LINK_RATES macros

Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Todd Previte <tprevite@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/atomic-helpers: make mode_set hooks optional
Daniel Vetter [Sun, 22 Feb 2015 11:24:20 +0000 (12:24 +0100)]
drm/atomic-helpers: make mode_set hooks optional

With runtime PM the hw might still be off while doing the ->mode_set
callbacks - runtime PM get/put should only happen in the
enable/disable hooks to properly support DPMS. Which essentially makes
these callbacks useless for drivers support runtime PM, so make them
optional. Again motivated by discussions with Laurent.

Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agodrm/atomic-helper: Rename commmit_post/pre_planes
Daniel Vetter [Sun, 22 Feb 2015 11:24:19 +0000 (12:24 +0100)]
drm/atomic-helper: Rename commmit_post/pre_planes

These names only make sense because of backwards compatability with
the order used by the crtc helper library. There's not really any real
requirement in the ordering here.

So rename them to something more descriptive and update the kerneldoc
a bit. Motivated in a discussion with Laurent about how to restore
plane state for dpms for drivers with runtime pm.

v2: Squash in fixup from Stephen Rothwell to fix a conflict with
tegra.

Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Acked-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
9 years agodrm/i915: Reset logical ring contexts' head and tail during GPU reset
Thomas Daniel [Mon, 16 Feb 2015 16:12:53 +0000 (16:12 +0000)]
drm/i915: Reset logical ring contexts' head and tail during GPU reset

Work was getting left behind in LRC contexts during reset.  This causes a hang
if the GPU is reset when HEAD==TAIL because the context's ringbuffer head and
tail don't get reset and retiring a request doesn't alter them, so the ring
still appears full.

Added a function intel_lr_context_reset() to reset head and tail on a LRC and
its ringbuffer.

Call intel_lr_context_reset() for each context in i915_gem_context_reset() when
in execlists mode.

Testcase: igt/pm_rps --run-subtest reset #bdw
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88096
Signed-off-by: Thomas Daniel <thomas.daniel@intel.com>
Reviewed-by: Dave Gordon <david.s.gordon@intel.com>
[danvet: Flatten control flow in the lrc reset code a notch.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Tune IZ hashing when subslices are unbalanced
Damien Lespiau [Sat, 14 Feb 2015 18:30:29 +0000 (18:30 +0000)]
drm/i915/skl: Tune IZ hashing when subslices are unbalanced

When one EU is disabled in a particular subslice, we can tune how the
work is spread between subslices to improve EU utilization.

v2: - Use a bitfield to record which subslice(s) has(have) 7 EUs. That
      will also make the machinery work if several sublices have 7 EUs.
      (Jeff Mcgee)
    - Only apply the different hashing algorithm if the slice is
      effectively unbalanced by checking there's a single subslice with
      7 EUs. (Jeff Mcgee)

v3: Fix typo in comment (Jeff Mcgee)

Issue: VIZ-3845
Cc: Jeff Mcgee <jeff.mcgee@intel.com>
Reviewed-by: Jeff Mcgee <jeff.mcgee@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: don't reallocate the compressed FB at every frame
Paulo Zanoni [Fri, 13 Feb 2015 19:23:47 +0000 (17:23 -0200)]
drm/i915: don't reallocate the compressed FB at every frame

With the current code we just reallocate the compressed FB at every
FBC update: we have X in one frame, then in the other frame we need X
again, but we check "needed < have" instead of "needed <= have".

v2: Rebase after Jani addressed the other problems described in v1.

Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: gen5+ can have FBC with multiple pipes
Paulo Zanoni [Fri, 13 Feb 2015 19:23:43 +0000 (17:23 -0200)]
drm/i915: gen5+ can have FBC with multiple pipes

So allow it.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: HSW+ FBC is tied to pipe A
Paulo Zanoni [Fri, 13 Feb 2015 19:23:42 +0000 (17:23 -0200)]
drm/i915: HSW+ FBC is tied to pipe A

So add code to consider this case.

v2: Reorder the series, so drop the possible_framebuffer_bits chunk.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com> (v1)
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: extract intel_fbc_find_crtc()
Paulo Zanoni [Fri, 13 Feb 2015 19:23:41 +0000 (17:23 -0200)]
drm/i915: extract intel_fbc_find_crtc()

I want to make this code a little more complicated, so let's extract
the function first.

Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Request full SSEU enablement on Gen9
Jeff McGee [Fri, 13 Feb 2015 16:27:56 +0000 (10:27 -0600)]
drm/i915: Request full SSEU enablement on Gen9

On Gen9 the render power gating can leave slice/subslice/EU in
a partially enabled state. We must make an explicit request for
full SSEU enablement through the Render Power Clock State
register when resuming render work. This register is save/
restored in the logical ring context image for execlist
submission mode. Initialize its value in each LRC image to
request full enablement according to the device SSEU config.

Thanks to Sharma Ankitprasad and Akash Goel for highlighting the
issue and proposing the initial fix on which this patch is based.

v2: Adjusted the names of the power gating support flags to fit
    update of an earlier patch.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Reviewed-by: "Akash Goel <akash.goel@intel.com>"
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Add SKL HW status to SSEU status
Jeff McGee [Fri, 13 Feb 2015 16:27:55 +0000 (10:27 -0600)]
drm/i915/skl: Add SKL HW status to SSEU status

Add a new section to the 'i915_sseu_status' debugfs entry to
report the currently enabled counts of slice, subslice, and
execution units on the device. The count of enabled subslice
per slice represents the most enabled subslice on any one
slice for devices where imbalances may exist. Similarly, the
count of enabled EU per subslice represents the most enabled
EU on any one subslice.

Collect this device status for Skylake by reading the Gen9
power gate control ack message registers. Power gate control
operates on EU in pairs, therefore our reported counts of
enabled EU can be overestimated by one for each pair in which
one EU is fused-off.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Determine SKL slice/subslice/EU info
Jeff McGee [Fri, 13 Feb 2015 16:27:54 +0000 (10:27 -0600)]
drm/i915/skl: Determine SKL slice/subslice/EU info

Read fuse registers to determine the available slice total,
subslice total, subslice per slice, EU total, and EU per subslice
counts of the SKL device. The EU per subslice attribute is more
precisely defined as the maximum EU available on any one subslice,
since available EU counts may vary across subslices due to fusing.
Set flags indicating the SKL device's slice/subslice/EU (SSEU)
power gating capability. Make all values available via debugfs
entry 'i915_sseu_status'.

v2: Several small clean-ups suggested by Damien. Most notably,
    used smaller types for the new device info fields to reduce
    memory usage and improved the clarity/readability of the
    method used to extract attribute values from the fuse
    registers.

Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915/skl: Implement WaDisablePowerCompilerClockGating
Damien Lespiau [Mon, 9 Feb 2015 19:33:16 +0000 (19:33 +0000)]
drm/i915/skl: Implement WaDisablePowerCompilerClockGating

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Limit max VCO supported in CHV to 6.48GHz
Vijay Purushothaman [Mon, 16 Feb 2015 09:37:59 +0000 (15:07 +0530)]
drm/i915: Limit max VCO supported in CHV to 6.48GHz

As per the recommendation from PHY team, limit the max vco supported in CHV to 6.48 GHz

Signed-off-by: Vijay Purushothaman <vijay.a.purushothaman@linux.intel.com>
Acked-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Add new PHY reg definitions for lock threshold
Vijay Purushothaman [Mon, 16 Feb 2015 09:37:58 +0000 (15:07 +0530)]
drm/i915: Add new PHY reg definitions for lock threshold

Added new PHY register definitions to control TDC buffer calibration and
digital lock threshold.

Signed-off-by: Vijay Purushothaman <vijay.a.purushothaman@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: Trim the command parser allocations
Chris Wilson [Wed, 14 Jan 2015 11:20:57 +0000 (11:20 +0000)]
drm/i915: Trim the command parser allocations

Currently, the command parser tries to create a secondary batch exactly
as large as the original, and vmap both. This is open to abuse by
userspace using extremely large batch objects, but only executing very
short batches. For example, this would be if userspace were to implement
a command submission ringbuffer. However, we only need to allocate pages
for just the contents of the command sequence in the batch - all
relocations copied to the secondary batch will reference the original
batch and so there can be no access to the secondary batch outside of
the explicit execution region.

Testcase: igt/gem_exec_big #ivb,byt,hsw
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88308
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/i915: FIFO space query code refactor
Dave Gordon [Wed, 10 Dec 2014 18:12:12 +0000 (18:12 +0000)]
drm/i915: FIFO space query code refactor

When querying the GTFIFOCTL register to check the FIFO space, the read value
must be masked. The operation is repeated explicitly in several places. This
change refactors the read-and-mask code into a function call.

v2: rebased on top of Mika's forcewake patch set, specifically:
[PATCH 8/8] drm/i915: Enum forcewake domains and domain identifiers

Change-Id: Id1a9f3785cb20b82d4caa330c37b31e4e384a3ef
Signed-off-by: Dave Gordon <david.s.gordon@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
9 years agodrm/atomic: Rename drm_atomic_helper_commit_pre_planes() state argument
Laurent Pinchart [Sun, 22 Feb 2015 11:24:18 +0000 (12:24 +0100)]
drm/atomic: Rename drm_atomic_helper_commit_pre_planes() state argument

The argument contains a pointer to the old state, rename it to old_state
like in all other commit helper functions.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>