firefly-linux-kernel-4.4.55.git
10 years agodrm/nouveau/graph: enable when using external fw
Alexandre Courbot [Fri, 2 May 2014 09:32:39 +0000 (18:32 +0900)]
drm/nouveau/graph: enable when using external fw

nvc0_graph_ctor() would only let the graphics engine be enabled if its
oclass has a proper microcode linked to it. This prevents GR from being
enabled at all on chips that rely exclusively on external firmware, even
though such a use-case is valid.

Relax the conditions enabling the GR engine to also include the case
where an external firmware has also been loaded.

Also switch to external firmware if the graph class has no microcode
linked to it.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/nouveau/fifo: add GK20A support
Alexandre Courbot [Fri, 2 May 2014 09:32:38 +0000 (18:32 +0900)]
drm/nouveau/fifo: add GK20A support

GK20A's FIFO is compatible with NVE0, but only features 128 channels and
1 runlist.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/nouveau/fb: add GK20A support
Alexandre Courbot [Fri, 2 May 2014 09:32:37 +0000 (18:32 +0900)]
drm/nouveau/fb: add GK20A support

Add a simple FB device for GK20A, as well as a RAM implementation
suitable for chips that use system memory as video RAM.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/nouveau/ibus: add GK20A support
Alexandre Courbot [Fri, 2 May 2014 09:32:36 +0000 (18:32 +0900)]
drm/nouveau/ibus: add GK20A support

Add support for initializing the priv ring of GK20A. This is done by the
BIOS on desktop GPUs, but needs to be done by hand on Tegra.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/nvc0/bar: support chips without BAR3
Alexandre Courbot [Fri, 2 May 2014 09:32:35 +0000 (18:32 +0900)]
drm/nvc0/bar: support chips without BAR3

Adapt the NVC0 BAR driver to make it able to support chips that do not
expose a BAR3. When this happens, BAR1 is then used for USERD mapping
and the BAR alloc() functions is disabled, making GPU objects unable
to rely on BAR for data access and falling back to PRAMIN.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/nouveau/bar: only ioremap BAR3 if it exists
Alexandre Courbot [Fri, 2 May 2014 09:32:34 +0000 (18:32 +0900)]
drm/nouveau/bar: only ioremap BAR3 if it exists

Some chips that use system memory exclusively (e.g. GK20A) do not
expose 2 BAR regions. For them only BAR1 exists, and it should be used
for USERD mapping. Do not map BAR3 if its resource does not exist.

Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
10 years agodrm/doc: Fix nouveau typo
Damien Lespiau [Mon, 9 Jun 2014 13:40:35 +0000 (14:40 +0100)]
drm/doc: Fix nouveau typo

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge tag 'drm/panel/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux into...
Dave Airlie [Mon, 9 Jun 2014 22:53:00 +0000 (08:53 +1000)]
Merge tag 'drm/panel/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/panel: Changes for v3.16-rc1

This set of commits contains a couple of fixes to existing panel drivers
and support for some new panels.

One commit touches the DRM core in that in modifies the MIPI DSI support
to hook up the shutdown function so that drivers can provide code that's
run on shutdown. This is used by a subsequent commit to make the simple
panel driver power off the backlight on shutdown.

* tag 'drm/panel/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux:
  drm/panel: simple - Add AUO B133XTN01 panel support
  drm/panel: simple - Disable panel on shutdown
  drm/panel: add support for EDT ET057090DHU panel
  drm/panel: Add support for EDT ETM0700G0DH6 and ET070080DH6 panels
  drm/panel: ld9040: add power control sequence
  drm/panel: s6e8aa0: silence array overflow warning
  drm/dsi: Support device shutdown

10 years agoMerge tag 'drm/tegra/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux into...
Dave Airlie [Mon, 9 Jun 2014 22:51:19 +0000 (08:51 +1000)]
Merge tag 'drm/tegra/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux into drm-next

drm/tegra: Changes for v3.16-rc1

The majority of these changes are a slew of cleanups across the board.
A more noteworthy change is the addition of drm_dev_set_unique() and the
conversion of the Tegra DRM driver to use it. This allows us to get rid
of the host1x drm_bus implementation. Other USB and platform drivers can
be changed in a similar way. Unfortunately for most PCI devices there is
some userspace that relies on the old functionality and cannot be as
easily converted.

HDMI and hardware cursor support is added for Tegra124. The SOR output
gains support for exposing CRCs via debugfs, which can be used for
automated testing. Many values that were hardcoded in the SOR/eDP code
are now computed at runtime to increase compatibility with more devices.

* tag 'drm/tegra/for-3.16-rc1' of git://anongit.freedesktop.org/tegra/linux: (47 commits)
  drm/tegra: sor - Remove obsolete comment
  drm/tegra: sor - Enable only the necessary number of lanes
  drm/tegra: sor - Power on only the necessary lanes
  drm/tegra: sor - Do not program interlaced mode registers
  drm/tegra: sor - Do not hardcode link speed
  drm/tegra: sor - Do not hardcode number of blank symbols
  drm/tegra: sor - Don't hardcode link parameters
  drm/tegra: sor - Change power down ordering
  drm/tegra: sor - Fix copy/paste error
  drm/tegra: sor - Remove pixel clock rounding
  drm/tegra: sor - Make debugfs setup consistent
  drm/tegra: sor - Recursively remove debugfs tree
  drm/tegra: dp - Mark the connector as hotplug capable
  drm/tegra: dp - Implement hotplug detection in work queue
  drm/tegra: Add hardware cursor support
  drm/tegra: Remove host1x drm_bus implementation
  drm: Document how to register devices without struct drm_bus
  drm: Add device registration documentation
  drm: Introduce drm_dev_set_unique()
  gpu: host1x: Rename internal functions for clarity
  ...

10 years agodrm/panel: simple - Add AUO B133XTN01 panel support
Stéphane Marchesin [Sat, 24 May 2014 02:27:59 +0000 (19:27 -0700)]
drm/panel: simple - Add AUO B133XTN01 panel support

This panel is used by nyan-big and can be supported by the simple-panel
driver.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
[treding@nvidia.com: add device tree binding document]
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Remove obsolete comment
Thierry Reding [Thu, 5 Jun 2014 14:20:27 +0000 (16:20 +0200)]
drm/tegra: sor - Remove obsolete comment

According to the DP specification the disparity of the first symbol
should always be negative. It is therefore safe to assume that panels
will conform to that and therefore parameterizing this field should
never be necessary.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Enable only the necessary number of lanes
Thierry Reding [Thu, 5 Jun 2014 14:29:46 +0000 (16:29 +0200)]
drm/tegra: sor - Enable only the necessary number of lanes

Instead of always enabling all four lanes, enable only the number probed
from the link.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Power on only the necessary lanes
Thierry Reding [Thu, 5 Jun 2014 14:19:48 +0000 (16:19 +0200)]
drm/tegra: sor - Power on only the necessary lanes

Power on only those lanes required for the specified link.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Do not program interlaced mode registers
Thierry Reding [Thu, 5 Jun 2014 14:17:25 +0000 (16:17 +0200)]
drm/tegra: sor - Do not program interlaced mode registers

Interlaced mode is currently not supported on the SOR, so don't program
any associated registers.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Do not hardcode link speed
Thierry Reding [Thu, 5 Jun 2014 14:16:23 +0000 (16:16 +0200)]
drm/tegra: sor - Do not hardcode link speed

Use the speed probed from the link at runtime rather than relying on a
hardcoded default.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Do not hardcode number of blank symbols
Thierry Reding [Thu, 5 Jun 2014 14:12:46 +0000 (16:12 +0200)]
drm/tegra: sor - Do not hardcode number of blank symbols

The number of HBLANK and VBLANK symbols can be computed at runtime so
that they can be set appropriately depending on the video mode and DP
link.

These values are used by the packet generation logic to determine how
many audio samples can be transferred during the blanking intervals.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Don't hardcode link parameters
Thierry Reding [Thu, 5 Jun 2014 14:31:10 +0000 (16:31 +0200)]
drm/tegra: sor - Don't hardcode link parameters

The currently hardcoded link parameters don't work on all eDP panels, so
compute the parameters at runtime depending on the mode and panel type
to allow the driver to cope with a wider variety of panels.

Note that the number of bits per pixel of the panel is still hardcoded,
but this can be addressed in a separate patch.

This is largely based on a patch by Stéphane Marchesin but the algorithm
was largely rewritten to be more readable and concise.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Change power down ordering
Stéphane Marchesin [Fri, 23 May 2014 03:32:48 +0000 (20:32 -0700)]
drm/tegra: sor - Change power down ordering

Lanes are powered up in decreasing order. Power them down in increasing
order for consistency.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Fix copy/paste error
Stéphane Marchesin [Fri, 23 May 2014 03:32:47 +0000 (20:32 -0700)]
drm/tegra: sor - Fix copy/paste error

The comment above mentions link A/B but this isn't what the code does,
so let's fix that.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Remove pixel clock rounding
Stéphane Marchesin [Fri, 23 May 2014 03:32:46 +0000 (20:32 -0700)]
drm/tegra: sor - Remove pixel clock rounding

The code currently rounds up the clock to the next MHZ, which is
rounding up a 69.5MHz clock to 70MHz on my machine. This in turn
prevents the display from syncing. Removing this rounding fixes eDP
for me.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agoMerge tag 'drm-intel-fixes-2014-06-06' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Fri, 6 Jun 2014 09:07:09 +0000 (19:07 +1000)]
Merge tag 'drm-intel-fixes-2014-06-06' of git://anongit.freedesktop.org/drm-intel into drm-next

> Bunch of stuff for 3.16 still:
> - Mipi dsi panel support for byt. Finally! From Shobhit&others. I've
>   squeezed this in since it's a regression compared to vbios and we've
>   been ridiculed about it a bit too often ...
> - connection_mutex deadlock fix in get_connector (only affects i915).
> - Core patches from Matt's primary plane from Matt Roper, I've pushed the
>   i915 stuff to 3.17.
> - vlv power well sequencing fixes from Jesse.
> - Fix for cursor size changes from Chris.
> - agpbusy fixes from Ville.
> - A few smaller things.
>

* tag 'drm-intel-fixes-2014-06-06' of git://anongit.freedesktop.org/drm-intel: (32 commits)
  drm/i915: BDW: Adding missing cursor offsets.
  drm: Fix getconnector connection_mutex locking
  drm/i915/bdw: Only use 2g GGTT for 32b platforms
  drm/i915: Nuke pipe A quirk on i830M
  drm/i915: fix display power sw state reporting
  drm/i915: Always apply cursor width changes
  drm/i915: tell the user if both KMS and UMS are disabled
  drm/plane-helper: Add drm_plane_helper_check_update() (v3)
  drm: Check CRTC compatibility in setplane
  drm/i915: use VBT to determine whether to enumerate the VGA port
  drm/i915: Don't WARN about ring idle bit on gen2
  drm/i915: Silence the WARN if the user tries to GTT mmap an incoherent object
  drm/i915: Move the C3 LP write bit setup to gen3_init_clock_gating() for KMS
  drm/i915: Enable interrupt-based AGPBUSY# enable on 85x
  drm/i915: Flip the sense of AGPBUSY_DIS bit
  drm/i915: Set AGPBUSY# bit in init_clock_gating
  drm/i915/vlv: add pll assertion when disabling DPIO common well
  drm/i915/vlv: move DPIO common reset de-assert into __vlv_set_power_well
  drm/i915/vlv: re-order power wells so DPIO common comes after TX
  drm/i915/vlv: move CRI refclk enable into __vlv_set_power_well
  ...

10 years agodrm/tegra: sor - Make debugfs setup consistent
Thierry Reding [Wed, 28 May 2014 11:46:12 +0000 (13:46 +0200)]
drm/tegra: sor - Make debugfs setup consistent

Other output drivers set up debugfs slightly differently. Bring the SOR
driver in line with those for consistency.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Recursively remove debugfs tree
Thierry Reding [Fri, 25 Apr 2014 14:48:36 +0000 (16:48 +0200)]
drm/tegra: sor - Recursively remove debugfs tree

Removing only the root directory will fail when there are still files in
it. Instead of manually removing all files, remove the whole directory
recursively.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dp - Mark the connector as hotplug capable
Thierry Reding [Fri, 25 Apr 2014 14:44:48 +0000 (16:44 +0200)]
drm/tegra: dp - Mark the connector as hotplug capable

Doing so allows the hotplug events generated by the connector to be
properly handled by the DRM poll helpers.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dp - Implement hotplug detection in work queue
Thierry Reding [Fri, 25 Apr 2014 14:42:32 +0000 (16:42 +0200)]
drm/tegra: dp - Implement hotplug detection in work queue

Calling the drm_helper_hpd_irq_event() helper can sleep, so instead of
invoking it directly from the interrupt handler, schedule a work queue
and run it from there.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Add hardware cursor support
Thierry Reding [Fri, 20 Dec 2013 12:58:33 +0000 (13:58 +0100)]
drm/tegra: Add hardware cursor support

Enable hardware cursor support on Tegra124. Earlier generations support
the hardware cursor to some degree as well, but not in a way that can be
generically exposed.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Remove host1x drm_bus implementation
Thierry Reding [Thu, 22 May 2014 07:57:15 +0000 (09:57 +0200)]
drm/tegra: Remove host1x drm_bus implementation

The DRM core can now cope with drivers that don't have an associated
struct drm_bus, so the host1x implementation is no longer useful.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Document how to register devices without struct drm_bus
Thierry Reding [Wed, 23 Apr 2014 09:52:10 +0000 (11:52 +0200)]
drm: Document how to register devices without struct drm_bus

With the recent addition of the drm_set_unique() function, devices can
now be registered without requiring a drm_bus. Add a brief description
to the DRM docbook to show how that can be achieved.

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Add device registration documentation
Thierry Reding [Mon, 19 May 2014 11:39:07 +0000 (13:39 +0200)]
drm: Add device registration documentation

Describe how devices are registered using the drm_*_init() functions.
Adding this to docbook requires a largish set of changes to the comments
in drm_{pci,usb,platform}.c since they are doxygen-style rather than
proper kernel-doc and therefore mess with the docbook generation.

While at it, mark usage of drm_put_dev() as discouraged in favour of
calling drm_dev_unregister() and drm_dev_unref() directly.

Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Introduce drm_dev_set_unique()
Thierry Reding [Fri, 11 Apr 2014 13:23:00 +0000 (15:23 +0200)]
drm: Introduce drm_dev_set_unique()

Add a helper function that allows drivers to statically set the unique
name of the device. This will allow platform and USB drivers to get rid
of their DRM bus implementations and directly use drm_dev_alloc() and
drm_dev_register().

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agogpu: host1x: Rename internal functions for clarity
Thierry Reding [Thu, 22 May 2014 09:12:17 +0000 (11:12 +0200)]
gpu: host1x: Rename internal functions for clarity

The internal host1x_{,un}register_client() functions can potentially be
confused with public the host1x_client_{,un}register() functions.

Rename them to host1x_{add,del}_client() to remove some of the possible
confusion.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: gem - Make tegra_bo_import() static
Thierry Reding [Tue, 13 May 2014 14:46:11 +0000 (16:46 +0200)]
drm/tegra: gem - Make tegra_bo_import() static

The function is never used outside of the source file and therefore can
be locally scoped.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Add Tegra124 support
Thierry Reding [Fri, 15 Nov 2013 15:07:32 +0000 (16:07 +0100)]
drm/tegra: hdmi - Add Tegra124 support

Tegra124 is mostly backwards-compatible with Tegra114. However, Tegra124
supports a few more features (e.g. interlacing, ...). Introduce a new
compatible string and TMDS tables to cope with these differences.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Protect CRC debugfs against enable state
Thierry Reding [Wed, 26 Mar 2014 10:13:16 +0000 (11:13 +0100)]
drm/tegra: sor - Protect CRC debugfs against enable state

Accessing the CRC debugfs file will hang the system if the SOR is not
enabled, so make sure that it is stays enabled until the CRC has been
read.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Do not needlessly recompute pclk
Thierry Reding [Fri, 14 Mar 2014 13:20:42 +0000 (14:20 +0100)]
drm/tegra: dsi - Do not needlessly recompute pclk

In some cases the pixel clock used to not be correct, which is why it
had to be recomputed. It turns out that the reason why it wasn't correct
is that it was used wrongly. If used correctly there's not need for the
recomputation.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Compute shift clock divider in output drivers
Thierry Reding [Wed, 26 Mar 2014 12:32:21 +0000 (13:32 +0100)]
drm/tegra: dc - Compute shift clock divider in output drivers

The shift clock divider is highly dependent on the type of output, so
push computation of it down into the output drivers. The old code used
to work merely by accident.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Move around shift clock programming
Thierry Reding [Wed, 26 Mar 2014 11:32:14 +0000 (12:32 +0100)]
drm/tegra: dc - Move around shift clock programming

Program the shift clock divider in tegra_crtc_setup_clk() since that's
where the divider is computed, so passing it around can be avoided.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Reset controller on driver unload
Thierry Reding [Fri, 14 Mar 2014 13:25:43 +0000 (14:25 +0100)]
drm/tegra: dsi - Reset controller on driver unload

Assert the DSI controller's reset when the driver is unloaded to reduce
power consumption and to put the controller into a known state for
subsequent driver reloads.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Fix typo when disabling controller
Thierry Reding [Fri, 14 Mar 2014 13:19:17 +0000 (14:19 +0100)]
drm/tegra: dsi - Fix typo when disabling controller

When disabling the DSI controller, the code wasn't really doing what it
was supposed to.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Add enable guard
Thierry Reding [Fri, 14 Mar 2014 13:15:10 +0000 (14:15 +0100)]
drm/tegra: dsi - Add enable guard

To prevent the enable or disable operations to potentially be run
multiple times, add guards to return early when the output is already
in the targetted state.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Initialize proper packet sequences
Thierry Reding [Fri, 14 Mar 2014 13:13:15 +0000 (14:13 +0100)]
drm/tegra: dsi - Initialize proper packet sequences

The packet sequencer needs to be programmed depending on the video mode
of the attached peripheral. Add support for non-burst video modes with
sync events (as opposed to sync pulses) and select either sequence
depending on the video mode.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Implement VDD supply support
Thierry Reding [Fri, 14 Mar 2014 13:07:50 +0000 (14:07 +0100)]
drm/tegra: dsi - Implement VDD supply support

The DSI controllers are powered by a (typically 1.2V) regulator. Usually
this is always on, so there was no need to support enabling or disabling
it thus far. But in order not to consume any power when DSI is inactive,
give the driver a chance to enable or disable the supply as needed.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Remove unneeded code
Thierry Reding [Fri, 14 Mar 2014 13:02:28 +0000 (14:02 +0100)]
drm/tegra: dsi - Remove unneeded code

A bunch of registers are initialized to 0 upon during driver probe. It
turns out that none of these are actually needed, so they can simply be
dropped.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dsi - Use internal pixel format
Thierry Reding [Thu, 13 Mar 2014 07:50:39 +0000 (08:50 +0100)]
drm/tegra: dsi - Use internal pixel format

The pixel format enumeration values used by the Tegra DSI controller
don't match those defined by the DSI framework. Make sure to convert
them to the internal format before writing it to the register.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Fix disable sequence
Thierry Reding [Wed, 16 Apr 2014 08:55:25 +0000 (10:55 +0200)]
drm/tegra: hdmi - Fix disable sequence

For some reason when the PW*_ENABLE and PM*_ENABLE fields are cleared
during disable, the HDMI output stops working properly. Resetting and
initializing doesn't help.

Comment out those accesses for now until it has been determined what to
do about them.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Disable LVDS mode
Thierry Reding [Wed, 16 Apr 2014 08:47:36 +0000 (10:47 +0200)]
drm/tegra: hdmi - Disable LVDS mode

Disable LVDS mode according to register documentation. It seems like
this has no effect on the operation of HDMI, but it's probably a good
idea to do this anyway.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Use proper power-up sequence
Thierry Reding [Wed, 16 Apr 2014 08:46:24 +0000 (10:46 +0200)]
drm/tegra: hdmi - Use proper power-up sequence

This reflects the power-up sequence as described in the documentation,
but it doesn't seem to be strictly necessary to get HDMI to work.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Clean up clock usage
Thierry Reding [Wed, 16 Apr 2014 08:43:41 +0000 (10:43 +0200)]
drm/tegra: hdmi - Clean up clock usage

Clocks are never enabled or disabled in atomic context, so we can use
the clk_prepare_enable() and clk_disable_unprepare() helpers instead.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Reverse regulator enable ordering
Thierry Reding [Wed, 16 Apr 2014 08:24:12 +0000 (10:24 +0200)]
drm/tegra: hdmi - Reverse regulator enable ordering

Schematics indicate that the AVDD_HDMI_PLL supply should be enabled
prior to the AVDD_HDMI supply.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Remove duplicate code
Thierry Reding [Wed, 26 Mar 2014 11:21:25 +0000 (12:21 +0100)]
drm/tegra: hdmi - Remove duplicate code

The generic Tegra output code already sets up the clocks properly, so
there's no need to do it again when the HDMI output is enabled.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: hdmi - Add connector supply support
Thierry Reding [Fri, 28 Feb 2014 15:57:34 +0000 (16:57 +0100)]
drm/tegra: hdmi - Add connector supply support

Revert commit 18ebc0f404d5 "drm/tegra: hdmi: Enable VDD earlier for
hotplug/DDC" and instead add a new supply for the +5V pin on the HDMI
connector.

The vdd-supply property refers to the regulator that supplies the
AVDD_HDMI input on Tegra, rather than the +5V HDMI connector pin. This
was never a problem before, because all boards had that pin hooked up to
a regulator that was always on. Starting with Dalmore and continuing
with Venice2, the +5V pin is controllable via a GPIO. For reasons
unknown, the GPIO ended up as the controlling GPIO of the AVDD_HDMI
supply in the Dalmore and Venice2 DTS files. But that's not correct.
Instead, a separate supply must be introduced so that the +5V pin can be
controlled separately from the supplies that feed the HDMI block within
Tegra.

A new hdmi-supply property is introduced that takes the place of the
vdd-supply and vdd-supply is only enabled when HDMI is enabled rather
than all the time.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Use proper H/V ref-to-sync values
Thierry Reding [Wed, 16 Apr 2014 07:22:38 +0000 (09:22 +0200)]
drm/tegra: dc - Use proper H/V ref-to-sync values

For HDMI compliance both of these values need to be set to 1.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Do not touch power control register
Thierry Reding [Wed, 16 Apr 2014 07:52:31 +0000 (09:52 +0200)]
drm/tegra: dc - Do not touch power control register

Setting the bits in this register is dependent on the output type driven
by the display controller. All output drivers already set these properly
so there is no need to do it here again.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Reshuffle code to get rid of prototypes
Thierry Reding [Fri, 14 Mar 2014 08:54:58 +0000 (09:54 +0100)]
drm/tegra: dc - Reshuffle code to get rid of prototypes

The tegra_dc_format() and tegra_dc_setup_window() functions are only
used internally by the display controller driver. Move them upwards in
order to make them static and get rid of the function prototypes.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Rename INVERT_V to V_DIRECTION
Thierry Reding [Tue, 25 Feb 2014 11:04:06 +0000 (12:04 +0100)]
drm/tegra: dc - Rename INVERT_V to V_DIRECTION

V_DIRECTION is the name of the field in the documentation, so use that
for consistency. Also add the H_DIRECTION field for completeness.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: sor - Add CRC debugfs support
Thierry Reding [Fri, 31 Jan 2014 09:02:15 +0000 (10:02 +0100)]
drm/tegra: sor - Add CRC debugfs support

The SOR allows the computation of a 32 bit CRC of the content that it
transmits. This functionality is exposed via debugfs and is useful to
verify proper operation of the SOR.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: dc - Add YUYV support
Thierry Reding [Wed, 29 Jan 2014 19:31:17 +0000 (20:31 +0100)]
drm/tegra: dc - Add YUYV support

YUYV is UYVY with swapped bytes. Luckily the Tegra DC hardware can swap
bytes during scan-out, so supporting YUYV is simply a matter of writing
the correct value to the byteswap register.

This patch modifies tegra_dc_format() to return the byte swap parameter
via an output parameter in addition to returning the pixel format. Many
other formats can potentially be supported in a similar way.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/tegra: Cleanup header file
Thierry Reding [Wed, 16 Apr 2014 07:54:21 +0000 (09:54 +0200)]
drm/tegra: Cleanup header file

Remove extern keyword from function prototypes since it isn't needed and
drop an unnecessary forward declaration.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm: Fix getconnector connection_mutex locking
Daniel Vetter [Thu, 5 Jun 2014 09:12:03 +0000 (11:12 +0200)]
drm: Fix getconnector connection_mutex locking

I've fumbled my own idea and enthusiastically wrapped all the
getconnector code with the connection_mutex. But we only need it to
chase the connector->encoder link. Even there it's not really needed
since races with userspace won't matter, but better paranoid and
consistent about this stuff.

If we grap it everywhere connector probe callbacks can't grab it
themselves, which means they'll deadlock. i915 does that for the load
detect pipe. Furthermore i915 needs to do a ww dance since we also
need to grab the mutex of the load detect crtc.

This is a regression from

commit 6e9f798d91c526982cca0026cd451e8fdbf18aaf
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu May 29 23:54:47 2014 +0200

    drm: Split connection_mutex out of mode_config.mutex (v3)

Cc: Rob Clark <robdclark@gmail.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm/panel: simple - Disable panel on shutdown
Thierry Reding [Tue, 29 Apr 2014 15:21:21 +0000 (17:21 +0200)]
drm/panel: simple - Disable panel on shutdown

When a device is shut down, disable the panel to make sure the display
backlight doesn't stay lit.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: add support for EDT ET057090DHU panel
Stefan Agner [Thu, 15 May 2014 09:38:45 +0000 (11:38 +0200)]
drm/panel: add support for EDT ET057090DHU panel

This panel is sold by Toradex for Colibri T20/T30 and Apalis T30
evaluation kits.

Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: Add support for EDT ETM0700G0DH6 and ET070080DH6 panels
Philipp Zabel [Thu, 15 May 2014 10:25:47 +0000 (12:25 +0200)]
drm/panel: Add support for EDT ETM0700G0DH6 and ET070080DH6 panels

The EDT ETM0700G0DH6 and ET070080DH6 are 7" 800x480 panels,
which can be supported by the simple panel driver.

Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: ld9040: add power control sequence
Andrzej Hajda [Tue, 8 Apr 2014 10:58:05 +0000 (12:58 +0200)]
drm/panel: ld9040: add power control sequence

Some ld9040 panels do not start without providing power control sequence
during initialization. The patch fixes the driver by providing such
sequence for all panels.

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/panel: s6e8aa0: silence array overflow warning
Dan Carpenter [Wed, 9 Apr 2014 15:21:12 +0000 (18:21 +0300)]
drm/panel: s6e8aa0: silence array overflow warning

Smatch complains that we are reading beyond the end of the array here:

drivers/gpu/drm/panel/panel-s6e8aa0.c:852 s6e8aa0_read_mtp_id()
warn: buffer overflow 's6e8aa0_variants' 4 <= 4

We set the error code, so it's not harmful but it looks like a return
was intended here so lets add that and silence the warning.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/dsi: Support device shutdown
Thierry Reding [Tue, 29 Apr 2014 15:19:57 +0000 (17:19 +0200)]
drm/dsi: Support device shutdown

Hook up the MIPI DSI bus's .shutdown() function to allow drivers to
implement code that should be run when a device is shut down.

Signed-off-by: Thierry Reding <treding@nvidia.com>
10 years agodrm/i915: BDW: Adding missing cursor offsets.
Rodrigo Vivi [Thu, 5 Jun 2014 00:09:30 +0000 (17:09 -0700)]
drm/i915: BDW: Adding missing cursor offsets.

BDW uses IVB cursor offsets.

Whithout this patch it is not possible to use multiple outputs with cursor
on BDW.
The cursor gets completely crazy because update position uses the wrong
cursor register for the second pipe.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Cc: stable@vger.kernel.org
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=79621
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: Fix getconnector connection_mutex locking
Daniel Vetter [Thu, 5 Jun 2014 09:07:20 +0000 (11:07 +0200)]
drm: Fix getconnector connection_mutex locking

I've fumbled my own idea and enthusiastically wrapped all the
getconnector code with the connection_mutex. But we only need it to
chase the connector->encoder link. Even there it's not really needed
since races with userspace won't matter, but better paranoid and
consistent about this stuff.

If we grap it everywhere connector probe callbacks can't grab it
themselves, which means they'll deadlock. i915 does that for the load
detect pipe. Furthermore i915 needs to do a ww dance since we also
need to grab the mutex of the load detect crtc.

This is a regression from

commit 6e9f798d91c526982cca0026cd451e8fdbf18aaf
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu May 29 23:54:47 2014 +0200

    drm: Split connection_mutex out of mode_config.mutex (v3)

Cc: Rob Clark <robdclark@gmail.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoMerge commit '9e9a928eed8796a0a1aaed7e0b676db86ba84594' into drm-next
Dave Airlie [Thu, 5 Jun 2014 10:28:59 +0000 (20:28 +1000)]
Merge commit '9e9a928eed8796a0a1aaed7e0b676db86ba84594' into drm-next

Merge drm-fixes into drm-next.

Both i915 and radeon need this done for later patches.

Conflicts:
drivers/gpu/drm/drm_crtc_helper.c
drivers/gpu/drm/i915/i915_drv.h
drivers/gpu/drm/i915/i915_gem.c
drivers/gpu/drm/i915/i915_gem_execbuffer.c
drivers/gpu/drm/i915/i915_gem_gtt.c

10 years agodrm/i915/bdw: Only use 2g GGTT for 32b platforms
Ben Widawsky [Tue, 27 May 2014 23:53:08 +0000 (16:53 -0700)]
drm/i915/bdw: Only use 2g GGTT for 32b platforms

Daniel requested in the bug that I use a 3GB fallback size. Since this
is not in the spec as a valid size, I decided against it. We could
potentially add a patch to bump it to 3GB on top of this one.

This probably should be CC: stable - but I'll let the powers that be
decide that one.

Regression from a revert of the revert:
commit 7907f45bf9f67a1c5e5d4ae05bab428d7c2f43b2
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Wed Feb 19 22:05:46 2014 -0800

    Revert "drm/i915/bdw: Limit GTT to 2GB"

v2: Change ifdef to 32b, instead of ifndef
update comment

v3. Update comment to not wrap (Daniel).
Update commit message

v4: s/CONFIG_32/CONFIG_X86_32 (Jani).

v5: s/CONFIG_x86_32BIT/CONFIG_x86_32, as meant in v4
s/32B/32b (chris)

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=76619
Cc: stable@vger.kernel.org
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Tested-by: "Yang, Guang A" <guang.a.yang@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Nuke pipe A quirk on i830M
Daniel Vetter [Mon, 2 Jun 2014 15:57:18 +0000 (17:57 +0200)]
drm/i915: Nuke pipe A quirk on i830M

Apparently it does more harm than good. Thomas Richter reports that
it helps his machine (Thinkpad X31) and there's another report from a
Fujitsu S6010. Also, we've nuked it on i845G already to make Chris'
machine happy.

Cc: Thomas Richter <richter@rus.uni-stuttgart.de>
References: http://mid.mail-archive.com/538C54E0.8090507@rus.uni-stuttgart.de
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: fix display power sw state reporting
Imre Deak [Mon, 2 Jun 2014 11:21:10 +0000 (14:21 +0300)]
drm/i915: fix display power sw state reporting

Atm, we refcount both power domains and power wells and
intel_display_power_enabled_sw() returns the power domain refcount. What
the callers are really interested in though is the sw state of the
underlying power wells. Due to this we will report incorrectly that a
given power domain is off if its power wells were enabled via another
power domain, for example POWER_DOMAIN_INIT which enables all power
wells.

As a fix return instead the state based on the refcount of all power
wells included in the passed in power domain.

References: https://bugs.freedesktop.org/show_bug.cgi?id=79505
References: https://bugs.freedesktop.org/show_bug.cgi?id=79038
Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Always apply cursor width changes
Chris Wilson [Fri, 30 May 2014 13:35:26 +0000 (16:35 +0300)]
drm/i915: Always apply cursor width changes

It is possible for userspace to create a big object large enough for a
256x256, and then switch over to using it as a 64x64 cursor. This
requires the cursor update routines to check for a change in width on
every update, rather than just when the cursor is originally enabled.

This also fixes an issue with 845g/865g which cannot change the base
address of the cursor whilst it is active.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[Antti:rebased, adjusted macro names and moved some lines, no functional
changes]
Reviewed-by: Antti Koskipaa <antti.koskipaa@linux.intel.com>
Tested-by: Antti Koskipaa <antti.koskipaa@linux.intel.com>
Cc: stable@vger.kernel.org
Testcase: igt/kms_cursor_crc/cursor-size-change
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: tell the user if both KMS and UMS are disabled
Jani Nikula [Mon, 2 Jun 2014 13:58:30 +0000 (16:58 +0300)]
drm/i915: tell the user if both KMS and UMS are disabled

If both KMS is disabled (by i915.modeset=0 or nomodeset parameters) and
UMS is disabled (by CONFIG_DRM_I915_UMS=n, the default), the user might
not be aware his setup is not supported. Inform the users (and, by
extension, the poor i915 developers having to read their dmesgs in bug
reports) why their graphics experience might be lacking.

A similar message was added on the UMS path in
commit e147accbd19f55489dabdcc4dc3551cc3e3f2553
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Thu Oct 10 15:25:37 2013 +0300

    drm/i915: tell the user KMS is required for gen6+

but it won't be reached if CONFIG_DRM_I915_UMS=n since
commit b30324adaf8d2e5950a602bde63030d15a61826f
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Wed Nov 13 22:11:25 2013 +0100

    drm/i915: Deprecated UMS support

v2: Use DRM_DEBUG_DRIVER.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/plane-helper: Add drm_plane_helper_check_update() (v3)
Matt Roper [Thu, 29 May 2014 15:06:52 +0000 (08:06 -0700)]
drm/plane-helper: Add drm_plane_helper_check_update() (v3)

Pull the parameter checking from drm_primary_helper_update() out into
its own function; drivers that provide their own setplane()
implementations rather than using the helper may still want to share
this parameter checking logic.

A few of the checks here were also updated based on suggestions by
Ville Syrjälä.

v3:
 - s/primary_helper/plane_helper/ --- this checking logic may be useful
   for other types of planes as well.
 - Fix visibility check (need to dereference visibility pointer)
v2:
 - Pass src/dest/clip rects and min/max scaling down to helper to avoid
   duplication of effort between helper and drivers (suggested by
   Ville).
 - Allow caller to specify whether the primary plane should be
   updatable while the crtc is disabled.

Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Chon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
[danvet: Include header properly and fixup declaration mismatch to
make this compile.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: Check CRTC compatibility in setplane
Matt Roper [Thu, 29 May 2014 15:06:51 +0000 (08:06 -0700)]
drm: Check CRTC compatibility in setplane

The DRM core setplane code should check that the plane is usable on the
specified CRTC before calling into the driver.

Prior to this patch, a plane's possible_crtcs field was purely
informational for userspace and was never actually verified at the
kernel level (aside from the primary plane helper).

Cc: dri-devel@lists.freedesktop.org
Reviewed-by: Chon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: use VBT to determine whether to enumerate the VGA port
Jesse Barnes [Fri, 4 Apr 2014 23:12:07 +0000 (16:12 -0700)]
drm/i915: use VBT to determine whether to enumerate the VGA port

Some platforms may not have it, and enumerating it is both confusing and
time consuming due to the hotplug and DDC probing.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Don't WARN about ring idle bit on gen2
Ville Syrjälä [Wed, 28 May 2014 16:12:13 +0000 (19:12 +0300)]
drm/i915: Don't WARN about ring idle bit on gen2

Gen2 doesn't have the ring idle/stop bits in the SCPD/MI_MODE register,
so don't go spewing warnings about the state of those bits.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Silence the WARN if the user tries to GTT mmap an incoherent object
Chris Wilson [Wed, 28 May 2014 15:16:41 +0000 (16:16 +0100)]
drm/i915: Silence the WARN if the user tries to GTT mmap an incoherent object

If the user tries to mmap through the GTT an object that is marked as
snooped, we report an error rather than allow the GPU to hang the
machine. The choice of EINVAL, however, was unfortunate as we turn that
into a WARN rather than a quiet SIGBUS.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Move the C3 LP write bit setup to gen3_init_clock_gating() for KMS
Ville Syrjälä [Tue, 25 Feb 2014 13:13:41 +0000 (15:13 +0200)]
drm/i915: Move the C3 LP write bit setup to gen3_init_clock_gating() for KMS

Move the MI_ARB_STATE MI_ARB_C3_LP_WRITE_ENABLE setup to
gen3_init_clock_gating() from i915_gem_load() when KMS is enabled. Leave
it in i915_gem_load() for the UMS case, but add an explcit check, just
to make it easier to spot it when we eventually rip out UMS support.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Enable interrupt-based AGPBUSY# enable on 85x
Ville Syrjälä [Tue, 25 Feb 2014 13:13:40 +0000 (15:13 +0200)]
drm/i915: Enable interrupt-based AGPBUSY# enable on 85x

85x also has a similar AGPBUSY# bit as gen3. Enable it to make
sure vblank interrupts don't get dealyed during C3 state.

There's also another bit which controls whether AGPBUSY# is asserted
based on pending cacheable cycles and interrupts, or just based on
pending commands in the ring and interrupts. Select the cacheable
cycles mode since that seems to be the new way of doing things in
85x, and it does give slightly better C3 residency numbers with
glxgears running.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Flip the sense of AGPBUSY_DIS bit
Ville Syrjälä [Tue, 25 Feb 2014 13:13:39 +0000 (15:13 +0200)]
drm/i915: Flip the sense of AGPBUSY_DIS bit

My Gen3 Bspec lists the AGPBUSY# bit in INSTPM as an enable bit rather
than a disable bit. Our code has the opposite idea. Make the code match
the spec.

Might fix some gen3 C3 related interrupt delivery problems. Untested
due to lack of hardware.

v2: call it AGPBUSY_INT_EN to make it clearer it has to do with interrupts

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Set AGPBUSY# bit in init_clock_gating
Ville Syrjälä [Tue, 25 Feb 2014 13:13:38 +0000 (15:13 +0200)]
drm/i915: Set AGPBUSY# bit in init_clock_gating

I don't see why we wouldn't want interrupts to wake up the CPU from C3
always, so just set the AGPBUSY# bit in gen3_init_clock_gating().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: add pll assertion when disabling DPIO common well
Jesse Barnes [Fri, 23 May 2014 20:16:45 +0000 (13:16 -0700)]
drm/i915/vlv: add pll assertion when disabling DPIO common well

When doing this, all PLLs should be disabled.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: move DPIO common reset de-assert into __vlv_set_power_well
Jesse Barnes [Fri, 23 May 2014 20:16:44 +0000 (13:16 -0700)]
drm/i915/vlv: move DPIO common reset de-assert into __vlv_set_power_well

We need to do this anytime we power gate the DPIO common well.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: re-order power wells so DPIO common comes after TX
Jesse Barnes [Fri, 23 May 2014 20:16:43 +0000 (13:16 -0700)]
drm/i915/vlv: re-order power wells so DPIO common comes after TX

There may be a dependency here.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: move CRI refclk enable into __vlv_set_power_well
Jesse Barnes [Fri, 23 May 2014 20:16:42 +0000 (13:16 -0700)]
drm/i915/vlv: move CRI refclk enable into __vlv_set_power_well

This needs to be done before we power back on the CMN_BC well so the PHY
can calibrate properly.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: drop power well enable in uncore_sanitize
Jesse Barnes [Fri, 23 May 2014 20:16:41 +0000 (13:16 -0700)]
drm/i915/vlv: drop power well enable in uncore_sanitize

We do this at runtime and later on now.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: assert and de-assert sideband reset at boot and resume v3
Jesse Barnes [Fri, 23 May 2014 20:16:40 +0000 (13:16 -0700)]
drm/i915/vlv: assert and de-assert sideband reset at boot and resume v3

This is a bit like the CMN reset de-assert we do in DPIO_CTL, except
that it resets the whole common lane section of the PHY.  This is
required on machines where the BIOS doesn't do this for us on boot or
resume to properly re-calibrate and get the PHY ready to transmit data.

Without this patch, such machines won't resume correctly much of the time,
with the symptom being a 'port ready' timeout and/or a link training
failure.

Note that simply asserting reset at suspend and de-asserting at resume
is not sufficient, nor is simply de-asserting at boot.  Both of these
cases have been tested and have still been found to have failures on
some configurations.

v2: extract simpler set_power_well function for use in reset_dpio (Imre)
    move to reset_dpio (Daniel & Ville)
v3: don't reset if DPIO reset is already de-asserted (Imre)

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: dsi: fix pipe-off timeout due to port vs. pipe disable ordering
Imre Deak [Tue, 27 May 2014 16:00:09 +0000 (19:00 +0300)]
drm/i915: dsi: fix pipe-off timeout due to port vs. pipe disable ordering

If we disable first the port (by disabling DPI) and only then the
display pipe the pipe-off flag will never be set, possibly leading to a
hanged pipe state at the next modeset-enable.

Note that according to the VLV2 display cluster HAS, we should disable
the port before the pipe. This doesn't seem to match reality based on
the above and it's also asymmetric with the enabling sequence, where we
first enable the port and then the pipe.

v2:
- send the panel shutdown command before stopping the pipe, since this
  is the recommended sequence (Shobhit)

Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Shobhit Kumar <shobhit.kumar@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Detect if MIPI panel based on VBT and initialize only if present
Shobhit Kumar [Tue, 27 May 2014 14:03:59 +0000 (19:33 +0530)]
drm/i915: Detect if MIPI panel based on VBT and initialize only if present

It seems by default the VBT has MIPI configuration block as well. The
Generic driver will assume always MIPI if MIPI configuration block is found.
This is causing probelm when actually there is eDP. Fix this by looking
into general definition block which will have device configurations. From here
we can figure out what is the LFP type and initialize MIPI only if MIPI
is found.

v2: Addressed review comments by Damien
    - Moved PORT definitions to intel_bios.h and renamed as DVO_PORT_MIPIA
    - renamed is_mipi to has_mipi and moved definition as suggested
    - Check has_mipi inside parse_mipi and intel_dsi_init insted of outside

v3: Make has_mipi as a bitfield as suggested

Signed-off-by: Shobhit Kumar <shobhit.kumar@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: fold in conditions to pack everything neatly below 80 chars.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/vlv: Modifying WA 'WaDisableL3Bank2xClockGate for vlv
Akash Goel [Mon, 24 Mar 2014 17:30:07 +0000 (23:00 +0530)]
drm/i915/vlv: Modifying WA 'WaDisableL3Bank2xClockGate for vlv

For disabling L3 clock gating we need to set bit 25 of MMIO
register 940c. Earlier this was being done by just writing 1
into bit 25 and resetting all other bits.
This patch modifies the routine to read-modify-write of the
register, so that the values of other bits are not destroyed.

v2: Modifying the comments and the patch commit message (Chris)

Signed-off-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Sourab Gupta <sourab.gupta@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Apply checkpatch fixup.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Add support for Generic MIPI panel driver
Shobhit Kumar [Fri, 23 May 2014 16:05:27 +0000 (21:35 +0530)]
drm/i915: Add support for Generic MIPI panel driver

This driver makes use of the generic panel information from the VBT.
Panel information is classified into two - panel configuration and panel
power sequence which is unique to each panel. The generic driver uses the
panel configuration and sequence parsed from VBT block #52 and #53

v2: Address review comments by Jani
    - Move all of the things in driver c file from header
    - Make all functions static
    - Make use of video/mipi_display.c instead of redefining
    - Null checks during sequence execution

v3: Address review comments by Damien
    - Rename the panel driver file as intel_dsi_panel_vbt.c
    - Fix style changes as suggested
    - Correct comments for lp->hs and hs->lp count calculations
    - General updating comments to have more clarity
    - using max() instead of ternary operator
    - Fix names (ui_num, ui_den) while using UI in calculations
    - compute max of lp_to_hs switch and hs_to_lp switch while computing
      hs_lp_switch_count

Signed-off-by: Shobhit Kumar <shobhit.kumar@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Extract gen8_gt_irq_reset
Daniel Vetter [Thu, 22 May 2014 20:18:22 +0000 (22:18 +0200)]
drm/i915: Extract gen8_gt_irq_reset

Fallout from an intermediate patch revision that I deemed worth saving.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Improve irq handling after gpu resets
Daniel Vetter [Thu, 22 May 2014 20:18:21 +0000 (22:18 +0200)]
drm/i915: Improve irq handling after gpu resets

Currently we do a full re-init of all interrupts after a gpu hang.
Which is pretty bad since we don't restore the interrupts we've
enabled at runtime correctly. Even with that addressed it's rather
horribly race.

But on g4x and later we only reset the gt and not the entire gpu.
Which means we only need to reset the GT interrupt bits. Which has the
nice benefit that vblank waits, pipe CRC interrupts and everything
else display related just keeps on working.

The downside is that gt interrupt handling (i.e. ring->get/put_irq) is
still racy. But as long as the gpu hang reliably wakes all waters and
we have a short time where the refcount drops to 0 we'll recover. So
not that bad really.

v2: Ville noticed that GTIMR and PMIMR don't get cleared, only the
subordinate per-ring registers. So let's rip out all the interrupt dancing.
The FIXME comment is still required though since the ring irq handling
happens at the per-ring interrupt mask registers, too.

Testcase: igt/kms_flip/vblank-vs-hang
Testcase: igt/kms_pipe_crc_basic/hang-*
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Inline ilk/gen8_irq_reset
Daniel Vetter [Thu, 22 May 2014 15:56:34 +0000 (17:56 +0200)]
drm/i915: Inline ilk/gen8_irq_reset

No point in having this indirection.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Disable gpu reset on i965g/gm
Daniel Vetter [Thu, 22 May 2014 15:56:33 +0000 (17:56 +0200)]
drm/i915: Disable gpu reset on i965g/gm

Ville figured out that it needs a full display reset since apparently
a lot more goes down than just the GT. Until that's address it's
better to just diable gpu reset.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Fix up fifo underrun tracking, take N
Daniel Vetter [Thu, 22 May 2014 15:56:32 +0000 (17:56 +0200)]
drm/i915: Fix up fifo underrun tracking, take N

So apparently this is tricky.

We need to consider:
- We start out with all the hw enabling bits disabled, both the
  individual fifo underrun interrupts and the shared display error
  interrupts masked. Otherwise if the bios config is broken we'll blow
  up with a NULL deref in our interrupt handler since the crtc
  structures aren't set up yet at driver load time.
- On gmch we need to mask fifo underruns on the sw side, so always
  need to set that in sanitize_crtc for those platforms.
- On other platforms we try to set the sw tracking so that it reflects
  the real state. But since a few platforms have shared bits we must
  _not_ disable fifo underrun reporting. Otherwise we'll never enable
  the shared error interrupt.

This is the state before out patch, but unfortunately this is not good
enough. But after a suspend resume operation this is broken:
1. We don't enable the hw interrupts since the same code runs on
resume as on driver load.
2. The fifo underrun state adjustments we do in sanitize_crtc doesn't
fire on resume since (except for hilarious firmware) all pipes are off
at that point. But they also don't hurt since the subsequent crtc
enabling due to force_restore will enable fifo underruns.

Which means when we enable fifo underrun reporting we notice that the
per-crtc state is already correct and short-circuit everthing out. And
the interrupt doesn't get enabled.

A similar problem would happen if the bios doesn't light up anything
when the driver loads. Which is exactly what happens when we reload
the driver since our unload functions disables all outputs.

Now we can't just rip out the short-circuit logic and unconditionally
update the fifo underrun reporting interrupt masking: We have some
checks for shared error interrupts to catch issues that happened when
the shared error interrupt was disabled.

The right fix is to push down this logic so that we can always update
the hardware state, but only check for missed fifo underruns on a real
enabled->disabled transition and ignore them when we're already
disabled.

On platforms with shared error interrupt the pipe CRC interrupts are
grouped together with the fifo underrun reporting this fixes pipe CRC
support after suspend and driver reloads.

Testcase: igt/kms_pipe_crc_basic/suspend-*
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Add fifo underrun reporting state to debugfs
Daniel Vetter [Thu, 22 May 2014 15:56:31 +0000 (17:56 +0200)]
drm/i915: Add fifo underrun reporting state to debugfs

On platforms with shared interrupt enable bits (which are shared even
with the pipe CRC logic) there's some tricky corner cases. Add
information to make debugging those easier.

Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: add drm_fb_helper_restore_fbdev_mode_unlocked()
Rob Clark [Fri, 30 May 2014 16:29:48 +0000 (12:29 -0400)]
drm: add drm_fb_helper_restore_fbdev_mode_unlocked()

All drm_fb_helper_restore_fbdev_mode() call sites, save one, do the same
locking.  Simplify this into drm_fb_helper_restore_fbdev_mode_unlocked().

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agodrm: convert crtc and connection_mutex to ww_mutex (v5)
Rob Clark [Tue, 19 Nov 2013 17:10:12 +0000 (12:10 -0500)]
drm: convert crtc and connection_mutex to ww_mutex (v5)

For atomic, it will be quite necessary to not need to care so much
about locking order.  And 'state' gives us a convenient place to stash a
ww_ctx for any sort of update that needs to grab multiple crtc locks.

Because we will want to eventually make locking even more fine grained
(giving locks to planes, connectors, etc), split out drm_modeset_lock
and drm_modeset_acquire_ctx to track acquired locks.

Atomic will use this to keep track of which locks have been acquired
in a transaction.

v1: original
v2: remove a few things not needed until atomic, for now
v3: update for v3 of connection_mutex patch..
v4: squash in docbook
v5: doc tweaks/fixes

Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>