firefly-linux-kernel-4.4.55.git
13 years agoMerge branch 'edp-training-fixes' into drm-intel-next
Keith Packard [Thu, 20 Oct 2011 20:40:33 +0000 (13:40 -0700)]
Merge branch 'edp-training-fixes' into drm-intel-next

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

Just whitespace change conflicts

13 years agodrm/i915: No need to wait for eDP power off delay if panel is on
Keith Packard [Thu, 29 Sep 2011 23:51:26 +0000 (16:51 -0700)]
drm/i915: No need to wait for eDP power off delay if panel is on

If the panel is powered up, there's no need to delay for the 'off'
interval when turning the panel on.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Restrict ILK-specific eDP power hack to ILK
Keith Packard [Thu, 29 Sep 2011 23:33:01 +0000 (16:33 -0700)]
drm/i915: Restrict ILK-specific eDP power hack to ILK

This eliminates a fairly long delay when power sequencing newer
hardware

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Disable eDP VDD in a delayed work proc instead of synchronously
Keith Packard [Mon, 19 Sep 2011 06:09:52 +0000 (23:09 -0700)]
drm/i915: Disable eDP VDD in a delayed work proc instead of synchronously

There's no good reason to turn off the eDP force VDD bit synchronously
while probing devices; that just sticks a huge delay into all mode
setting paths. Instead, queue a delayed work proc to disable the VDD
force bit and then remember when that fires to ensure that the
appropriate delay is respected before trying to turn it back on.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Create helper functions to determine eDP power state
Keith Packard [Thu, 29 Sep 2011 22:53:27 +0000 (15:53 -0700)]
drm/i915: Create helper functions to determine eDP power state

We need to check eDP VDD force and panel on in several places, so
create some simple helper functions to avoid duplicating code.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: edp_panel_on does not need to return a bool
Keith Packard [Thu, 29 Sep 2011 23:05:34 +0000 (16:05 -0700)]
drm/i915: edp_panel_on does not need to return a bool

The return value was unused, so just stop doing that.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Move eDP panel fixed mode from dev_priv to intel_dp
Keith Packard [Mon, 19 Sep 2011 00:35:47 +0000 (17:35 -0700)]
drm/i915: Move eDP panel fixed mode from dev_priv to intel_dp

This value doesn't come directly from the VBT, and so is rather
specific to the particular DP output.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Correct eDP panel power sequencing delay computations
Keith Packard [Wed, 28 Sep 2011 23:48:10 +0000 (16:48 -0700)]
drm/i915: Correct eDP panel power sequencing delay computations

Store the panel power sequencing delays in the dp private structure,
rather than the global device structure. Who knows, maybe we'll get
more than one eDP device in the future.

From the eDP spec, we need the following numbers:

 T1 + T3 Power on to Aux Channel operation (panel_power_up_delay)

This marks how long it takes the panel to boot up and
get ready to receive aux channel communications.

 T8 Video signal to backlight on (backlight_on_delay)

Once a valid video signal is being sent to the device,
it can take a while before the panel is actuall
showing useful data. This delay allows the panel
to get something reasonable up before the backlight
is turned on.

 T9 Backlight off to video off (backlight_off_delay)

Turning the backlight off can take a moment, so
this delay makes sure there is still valid video
data on the screen.

 T10 Video off to power off (panel_power_down_delay)

Presumably this delay allows the panel to perform
an orderly shutdown of the display.

 T11 + T12 Power off to power on (panel_power_cycle_delay)

So, once you turn the panel off, you have to wait a
while before you can turn it back on. This delay is
usually the longest in the entire sequence.

Neither the VBIOS source code nor the hardware documentation has a
clear mapping between the delay values they provide and those required
by the eDP spec. The VBIOS code actually uses two different labels for
the delay values in the five words of the relevant VBT table.

**** MORE LATER ***

Look at both the current hardware register settings and the VBT
specified panel power sequencing timings. Use the maximum of the two
delays, to make sure things work reliably. If there is no VBT data,
then those values will be initialized to zero, so we'll just use the
values as programmed in the hardware. Note that the BIOS just fetches
delays from the VBT table to place in the hardware registers, so we
should get the same values from both places, except for rounding.

VBT doesn't provide any values for T1 or T2, so we'll always just use
the hardware value for that.

The panel power up delay is thus T1 + T2 + T3, which should be
sufficient in all cases.

The panel power down delay is T1 + T2 + T12, using T1+T2 as a proxy
for T11, which isn't available anywhere.

For the backlight delays, the eDP spec says T6 + T8 is the delay from the
end of link training to backlight on and T9 is the delay from
backlight off until video off. The hardware provides a 'backlight on'
delay, which I'm taking to be T6 + T8 while the VBT provides something
called 'T7', which I'm assuming is s

On the macbook air I'm testing with, this yields a power-up delay of
over 200ms and a power-down delay of over 600ms. It all works now, but
we're frobbing these power controls several times during mode setting,
making the whole process take an awfully long time.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Ensure eDP powered up during DP_SET_POWER operation in dp_prepare
Keith Packard [Wed, 28 Sep 2011 23:44:14 +0000 (16:44 -0700)]
drm/i915: Ensure eDP powered up during DP_SET_POWER operation in dp_prepare

Any call to intel_dp_sink_dpms must ensure that the panel has power so
that the DP_SET_POWER operation will be correctly received. The only
one missing this was in intel_dp_prepare.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Enable eDP panel power during I2C initialization sequence
Keith Packard [Wed, 28 Sep 2011 23:41:05 +0000 (16:41 -0700)]
drm/i915: Enable eDP panel power during I2C initialization sequence

The DP i2c initialization code does a couple of i2c transactions,
which means that an eDP panel must be powered up.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Wrap DP EDID fetch functions to enable eDP panel power
Keith Packard [Wed, 28 Sep 2011 23:38:44 +0000 (16:38 -0700)]
drm/i915: Wrap DP EDID fetch functions to enable eDP panel power

Talking to the eDP DDC channel requires that the panel be powered
up. Wrap both the EDID and modes fetch code with calls to turn the vdd
power on and back off.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Delay DP i2c initialization until panel power timings are computed
Keith Packard [Wed, 28 Sep 2011 23:31:53 +0000 (16:31 -0700)]
drm/i915: Delay DP i2c initialization until panel power timings are computed

On eDP, DDC requires panel power, but turning that on uses the panel
power sequencing timing values fetch from the DPCD data.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Ensure panel is on during DPMS off
Keith Packard [Thu, 6 Oct 2011 02:53:09 +0000 (19:53 -0700)]
drm/i915: Ensure panel is on during DPMS off

If the panel is already off, we'll need to turn VDD on to execute the
(useless) DPMS off code. Yes, it would be better to just not do any of
this, but correctness, and *then* performance.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Turn force VDD back off when panel running in intel_dp_dpms
Keith Packard [Wed, 28 Sep 2011 23:28:00 +0000 (16:28 -0700)]
drm/i915: Turn force VDD back off when panel running in intel_dp_dpms

The VDD force bit is turned on before touching the panel, but if it
was enabled, there was no call to turn it back off. Add a call.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Check for eDP inside edp panel on/off funcs
Keith Packard [Wed, 28 Sep 2011 23:23:51 +0000 (16:23 -0700)]
drm/i915: Check for eDP inside edp panel on/off funcs

Cleans up code dealing with eDP a bit. Remove redundant checks in
callers

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Unlock PCH_PP_CONTROL always
Keith Packard [Mon, 19 Sep 2011 20:59:29 +0000 (13:59 -0700)]
drm/i915: Unlock PCH_PP_CONTROL always

Avoid any question about locked registers by just writing the unlock
pattern with every write to the register.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Check eDP power when doing aux channel communications
Keith Packard [Mon, 19 Sep 2011 20:54:47 +0000 (13:54 -0700)]
drm/i915: Check eDP power when doing aux channel communications

Verify that the eDP VDD is on, either with the panel being on or with
the VDD force-on bit being set.

This demonstrates that in many instances, VDD is not on when needed,
which leads to failed EDID communications.

Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Only use VBT panel mode on eDP if no EDID is found
Keith Packard [Mon, 19 Sep 2011 21:33:26 +0000 (14:33 -0700)]
drm/i915: Only use VBT panel mode on eDP if no EDID is found

We're going to assume that EDID is more reliable than the VBT tables
for eDP panels, which is notably true on MacBook machines where the
VBT contains completely bogus data.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Shut down PCH interrupts during irq_uninstall
Keith Packard [Tue, 20 Sep 2011 17:12:44 +0000 (10:12 -0700)]
drm/i915: Shut down PCH interrupts during irq_uninstall

This masks out all interrupts and ack's any pending ones at IRQ
uninstall time to make sure we don't receive any unexpected interrupts
later on.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agodrm/i915: Enable digital port hotplug on PCH systems
Keith Packard [Mon, 19 Sep 2011 20:31:02 +0000 (13:31 -0700)]
drm/i915: Enable digital port hotplug on PCH systems

We were relying on the BIOS to set these bits, which doesn't always
happen.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agoMerge branch 'drm-intel-fixes' into drm-intel-next
Keith Packard [Wed, 28 Sep 2011 21:44:38 +0000 (14:44 -0700)]
Merge branch 'drm-intel-fixes' into drm-intel-next

13 years agodrm/i915: FBC off for ironlake and older, otherwise on by default
Keith Packard [Tue, 20 Sep 2011 04:34:19 +0000 (21:34 -0700)]
drm/i915: FBC off for ironlake and older, otherwise on by default

Make the default FBC behaviour chipset specific, allowing us to turn
it on by default for Ironlake and older where it has been seen to
cause trouble with screen updates.

Signed-off-by: Keith Packard <keithp@keithp.com>
Tested-by: Francis Moreau <francis.moro@gmail.com>
13 years agodrm/i915: Enable SDVO hotplug interrupts for HDMI and DVI
Simon Farnsworth [Wed, 21 Sep 2011 16:13:30 +0000 (17:13 +0100)]
drm/i915: Enable SDVO hotplug interrupts for HDMI and DVI

I was seeing a nasty 5 frame glitch every 10 seconds, caused by the
poll for connection on DVI attached by SDVO.

As my SDVO DVI supports hotplug detect interrupts, the fix is to
enable them, and hook them in to the various bits of driver
infrastructure so that they work reliably.

Note that this is only tested on single-function DVI-D SDVOs, on two
platforms (965GME and 945GSE), and has not been checked against a
specification document.

With lots of help from Adam Jackson <ajax@redhat.com> on IRC.

Signed-off-by: Simon Farnsworth <simon.farnsworth@onelan.co.uk>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agoMerge branch 'drm-intel-fixes' into drm-intel-next
Keith Packard [Wed, 21 Sep 2011 21:52:46 +0000 (14:52 -0700)]
Merge branch 'drm-intel-fixes' into drm-intel-next

13 years agodrm/i915: Dumb down the semaphore logic
Ben Widawsky [Thu, 15 Sep 2011 03:32:47 +0000 (20:32 -0700)]
drm/i915: Dumb down the semaphore logic

While I think the previous code is correct, it was hard to follow and
hard to debug. Since we already have a ring abstraction, might as well
use it to handle the semaphore updates and compares.

I don't expect this code to make semaphores better or worse, but you
never know...

v2:
Remove magic per Keith's suggestions.
Ran Daniel's gem_ring_sync_loop test on this.

v3:
Ignored one of Keith's suggestions.

v4:
Removed some bloat per Daniel's recommendation.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: pass ELD to HDMI/DP audio driver
Wu Fengguang [Mon, 5 Sep 2011 06:25:34 +0000 (14:25 +0800)]
drm/i915: pass ELD to HDMI/DP audio driver

Add ELD support for Intel Eaglelake, IbexPeak/Ironlake,
SandyBridge/CougarPoint and IvyBridge/PantherPoint chips.

ELD (EDID-Like Data) describes to the HDMI/DP audio driver the audio
capabilities of the plugged monitor. It's built and passed to audio
driver in 2 steps:

(1) at get_modes time, parse EDID and save ELD to drm_connector.eld[]

(2) at mode_set time, write drm_connector.eld[] to the Transcoder's hw
    ELD buffer and set the ELD_valid bit to inform HDMI/DP audio driver

This patch is tested OK on G45/HDMI, IbexPeak/HDMI and IvyBridge/HDMI+DP.
Test scheme: plug in the HDMI/DP monitor, and run

        cat /proc/asound/card0/eld*

to check if the monitor name, HDMI/DP type, etc. show up correctly.

Minor imperfection: the GEN5_AUD_CNTL_ST/DIP_Port_Select field always
reads 0 (reserved). Without knowing the port number, I worked it around
by setting the ELD_valid bit for ALL the three ports. It's tested to not
be a problem, because the audio driver will find invalid ELD data and
hence rightfully abort, even when it sees the ELD_valid indicator.

Thanks to Zhenyu and Pierre-Louis for a lot of valuable help and testing.

CC: Zhao Yakui <yakui.zhao@intel.com>
CC: Wang Zhenyu <zhenyu.z.wang@intel.com>
CC: Jeremy Bush <contractfrombelow@gmail.com>
CC: Christopher White <c.white@pulseforce.com>
CC: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
CC: Paul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm: support routines for HDMI/DP ELD
Wu Fengguang [Mon, 5 Sep 2011 06:23:20 +0000 (14:23 +0800)]
drm: support routines for HDMI/DP ELD

ELD (EDID-Like Data) describes to the HDMI/DP audio driver the audio
capabilities of the plugged monitor.

This adds drm_edid_to_eld() for converting EDID to ELD. The converted
ELD will be saved in a new drm_connector.eld[128] data field. This is
necessary because the graphics driver will need to fixup some of the
data fields (eg. HDMI/DP connection type, AV sync delay) before writing
to the hardware ELD buffer. drm_av_sync_delay() will help the graphics
drivers dynamically compute the AV sync delay for fixing-up the ELD.

ELD selection policy: it's possible for one encoder to be associated
with multiple connectors (ie. monitors), in which case the first found
ELD will be returned by drm_select_eld(). This policy may not be
suitable for all users, but let's start it simple first.

The impact of ELD selection policy: assume there are two monitors, one
supports stereo playback and the other has 8-channel output; cloned
display mode is used, so that the two monitors are associated with the
same internal encoder. If only the stereo playback capability is reported,
the user won't be able to start 8-channel playback; if the 8-channel ELD
is reported, then user space applications may send 8-channel samples
down, however the user may actually be listening to the 2-channel
monitor and not connecting speakers to the 8-channel monitor.

According to James, many TVs will either refuse the display anything or
pop-up an OSD warning whenever they receive hdmi audio which they cannot
handle. Eventually we will require configurability and/or per-monitor
audio control even when the video is cloned.

CC: Zhao Yakui <yakui.zhao@intel.com>
CC: Wang Zhenyu <zhenyu.z.wang@intel.com>
CC: Jeremy Bush <contractfrombelow@gmail.com>
CC: Christopher White <c.white@pulseforce.com>
CC: Pierre-Louis Bossart <pierre-louis.bossart@intel.com>
CC: Paul Menzel <paulepanter@users.sourceforge.net>
CC: James Cloos <cloos@jhcloos.com>
CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
13 years agodrm/i915: Enable dither whenever display bpc < frame buffer bpc
Keith Packard [Mon, 5 Sep 2011 18:53:21 +0000 (11:53 -0700)]
drm/i915: Enable dither whenever display bpc < frame buffer bpc

We want to enable dithering on any pipe where the frame buffer has
more color resolution than the output device.

The previous code was incorrectly clamping the frame buffer bpc to the
display bpc, effectively disabling dithering all of the time as the
computed frame buffer bpc would never be larger than the display bpc.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reported-by: Oliver Hartkopp <socketcan@hartkopp.net>
Tested-by: Oliver Hartkopp <socketcan@hartkopp.net>
13 years agoMerge branch 'drm-intel-next' of git://people.freedesktop.org/~keithp/linux into...
Dave Airlie [Tue, 20 Sep 2011 08:36:22 +0000 (09:36 +0100)]
Merge branch 'drm-intel-next' of git://people.freedesktop.org/~keithp/linux into drm-next

* 'drm-intel-next' of git://people.freedesktop.org/~keithp/linux:
  Drivers: i915: Fix all space related issues.

13 years agoMerge branch 'drm-nouveau-next' of git://git.freedesktop.org/git/nouveau/linux-2...
Dave Airlie [Tue, 20 Sep 2011 08:35:22 +0000 (09:35 +0100)]
Merge branch 'drm-nouveau-next' of git://git.freedesktop.org/git/nouveau/linux-2.6 into drm-next

* 'drm-nouveau-next' of git://git.freedesktop.org/git/nouveau/linux-2.6: (353 commits)
  drm/nouveau: remove allocations from gart populate() hook
  drm/nvc0/fb: slightly improve PMFB intr handling, move out of nvc0_graph.c
  drm/nvc0/fifo: avoid touching missing subfifos
  drm/nvd9/disp: bail out of mode_set_base if no fb bound to crtc
  drm/nvd9/disp: stub some more api hooks so we don't oops on resume
  drm/nouveau: fix printk typo in ioremap failure path
  drm/nvc0/pm: minor clock readback fixes
  drm/nv40/pm: execute memory reset script from vbios
  drm/nv50/gr: refactor initialisation
  drm/nouveau: if requested, try harder at disabling sysmem pushbufs
  drm/nv50/gr: enable ctxprog xfer only when we need it to save power
  drm/nouveau/dp: add support for displayport table 0x30
  drm/nouveau/dp: return master dp table pointer too when looking up encoder
  drm/nouveau/bios: simplify U/d table hash matching func to just match
  drm/nouveau/dp: preserve non-pattern bits in DP_TRAINING_PATTERN_SET
  drm/nvc0/gr: remove MODULE_FIRMWARE() lines
  drm/nouveau/dp: use alternate lane mask for nvaf
  drm/nouveau/dp: link rate scripts are selected with a comparison table
  drm/nv40/pm: write nv40-specific reclocking routines
  drm/nv40/pm: parse geometric delta clock from vbios
  ...

13 years agodrm/nouveau: remove allocations from gart populate() hook
Ben Skeggs [Tue, 30 Aug 2011 04:30:11 +0000 (14:30 +1000)]
drm/nouveau: remove allocations from gart populate() hook

Since some somewhat questionable changes a while back, TTM provides a
completely empty array of struct dma_address that stays around for the
entire lifetime of the TTM object.

Lets use this array, *always*, rather than wasting yet more memory on
another array who's purpose is identical, as well as yet another bool array
of the same size saying *which* of the previous two arrays to use...

This change will also solve the high order allocation failures seen by
some people while using nouveau.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvc0/fb: slightly improve PMFB intr handling, move out of nvc0_graph.c
Ben Skeggs [Fri, 26 Aug 2011 03:05:23 +0000 (13:05 +1000)]
drm/nvc0/fb: slightly improve PMFB intr handling, move out of nvc0_graph.c

I'm still not certain how to determine the number of SUBPs are present on
a given board.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvc0/fifo: avoid touching missing subfifos
Ben Skeggs [Thu, 25 Aug 2011 05:53:57 +0000 (15:53 +1000)]
drm/nvc0/fifo: avoid touching missing subfifos

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd9/disp: bail out of mode_set_base if no fb bound to crtc
Ben Skeggs [Thu, 25 Aug 2011 23:40:39 +0000 (09:40 +1000)]
drm/nvd9/disp: bail out of mode_set_base if no fb bound to crtc

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd9/disp: stub some more api hooks so we don't oops on resume
Ben Skeggs [Thu, 25 Aug 2011 04:09:43 +0000 (14:09 +1000)]
drm/nvd9/disp: stub some more api hooks so we don't oops on resume

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: fix printk typo in ioremap failure path
Marcin Slusarz [Mon, 22 Aug 2011 21:28:56 +0000 (23:28 +0200)]
drm/nouveau: fix printk typo in ioremap failure path

Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvc0/pm: minor clock readback fixes
Ben Skeggs [Thu, 23 Jun 2011 06:37:00 +0000 (16:37 +1000)]
drm/nvc0/pm: minor clock readback fixes

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv40/pm: execute memory reset script from vbios
Ben Skeggs [Fri, 12 Aug 2011 00:05:43 +0000 (10:05 +1000)]
drm/nv40/pm: execute memory reset script from vbios

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv50/gr: refactor initialisation
Ben Skeggs [Thu, 11 Aug 2011 23:30:58 +0000 (09:30 +1000)]
drm/nv50/gr: refactor initialisation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: if requested, try harder at disabling sysmem pushbufs
Ben Skeggs [Thu, 11 Aug 2011 22:56:06 +0000 (08:56 +1000)]
drm/nouveau: if requested, try harder at disabling sysmem pushbufs

On >=nv50, userspace would still end up allocating pushbufs in GART.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv50/gr: enable ctxprog xfer only when we need it to save power
Martin Peres [Sat, 30 Jul 2011 21:08:45 +0000 (23:08 +0200)]
drm/nv50/gr: enable ctxprog xfer only when we need it to save power

This patch adds instructions to ctxprog and by doing, impacts context
switching performance.  My testcase showed a 1% performance cost using
glxgears that is a context-switch bound application.

Please test and report bugs/performance/power/other.

Many thanks to Maxim Levitsky for his dedicated work on lowering power
consumption with nouveau.

More patches are coming thanks to his work:

https://bugs.freedesktop.org/show_bug.cgi?id=37922

Signed-off-by: Martin Peres <martin.peres@ensi-bourges.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: add support for displayport table 0x30
Ben Skeggs [Fri, 5 Aug 2011 04:47:28 +0000 (14:47 +1000)]
drm/nouveau/dp: add support for displayport table 0x30

Written from observations of my NVD9's vbios, completely untested due to
my NVD9 lacking actual DisplayPort connectors..

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: return master dp table pointer too when looking up encoder
Ben Skeggs [Fri, 5 Aug 2011 04:07:04 +0000 (14:07 +1000)]
drm/nouveau/dp: return master dp table pointer too when looking up encoder

Will need to be able to distinguish 2.0/2.1 from 3.0 soon.  Also, move
the vbios parsing to nouveau_dp where it belongs.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/bios: simplify U/d table hash matching func to just match
Ben Skeggs [Fri, 5 Aug 2011 03:42:49 +0000 (13:42 +1000)]
drm/nouveau/bios: simplify U/d table hash matching func to just match

The caller is now responsible for parsing its own lists (or whatever) of
possible encoders.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: preserve non-pattern bits in DP_TRAINING_PATTERN_SET
Ben Skeggs [Fri, 5 Aug 2011 05:56:53 +0000 (15:56 +1000)]
drm/nouveau/dp: preserve non-pattern bits in DP_TRAINING_PATTERN_SET

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvc0/gr: remove MODULE_FIRMWARE() lines
Ben Skeggs [Sun, 7 Aug 2011 22:57:55 +0000 (08:57 +1000)]
drm/nvc0/gr: remove MODULE_FIRMWARE() lines

We don't use these by default anymore, and there's been complaints from a
number of places thinking that the firmware blobs are required still.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: use alternate lane mask for nvaf
Ben Skeggs [Fri, 5 Aug 2011 01:09:21 +0000 (11:09 +1000)]
drm/nouveau/dp: use alternate lane mask for nvaf

Naturally...  Because Macs can't just be the same as everything else
now can they?

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: link rate scripts are selected with a comparison table
Ben Skeggs [Fri, 5 Aug 2011 00:28:52 +0000 (10:28 +1000)]
drm/nouveau/dp: link rate scripts are selected with a comparison table

Not hardcoded as originally thought.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv40/pm: write nv40-specific reclocking routines
Ben Skeggs [Mon, 18 Jul 2011 05:15:34 +0000 (15:15 +1000)]
drm/nv40/pm: write nv40-specific reclocking routines

Not 100% perfect yet, but a good start towards what it'll look like in the
end.

Actually seems stable on a NV44 I have here, as much as running around OA
for a fair amount of time constantly switching between performance levels
can prove..

My NV49 isn't quite so happy, and semaphores mess up somehow (sometimes) as
a result of the memory reclocking.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv40/pm: parse geometric delta clock from vbios
Ben Skeggs [Mon, 18 Jul 2011 06:02:37 +0000 (16:02 +1000)]
drm/nv40/pm: parse geometric delta clock from vbios

This changes the meaning of what we reported as "core" clock previously.

The shader/rop units are allegedly supposed to be run at the base clock
listed in the perf table, while the geometric clock can be bumped from
this value on some boards.

So that we can report both, we'll report the base clock as "shader" (since
the shaders *do* run at it), and the geometric clock as "core".

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: enable down-spread if vbios and sink support it
Ben Skeggs [Thu, 4 Aug 2011 04:31:28 +0000 (14:31 +1000)]
drm/nouveau/dp: enable down-spread if vbios and sink support it

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: execute some more vbios tables relating to link rate
Ben Skeggs [Thu, 4 Aug 2011 04:16:45 +0000 (14:16 +1000)]
drm/nouveau/dp: execute some more vbios tables relating to link rate

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: store unencoded link_bw everywhere
Ben Skeggs [Wed, 3 Aug 2011 23:55:44 +0000 (09:55 +1000)]
drm/nouveau/dp: store unencoded link_bw everywhere

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: restructure link training code
Ben Skeggs [Wed, 3 Aug 2011 23:26:44 +0000 (09:26 +1000)]
drm/nouveau/dp: restructure link training code

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: pass in required datarate to link training
Ben Skeggs [Thu, 4 Aug 2011 01:04:47 +0000 (11:04 +1000)]
drm/nouveau/dp: pass in required datarate to link training

Not used currently, but it will be used in preference to pre-determined
lane/bandwidth numbers at a later point.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/bios: check for null script pointers in parser
Ben Skeggs [Thu, 4 Aug 2011 03:57:33 +0000 (13:57 +1000)]
drm/nouveau/bios: check for null script pointers in parser

Allows us to be lazy elsewhere...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nva3/backlight: add suppport for newer style backlight regs
Ben Skeggs [Tue, 2 Aug 2011 22:52:39 +0000 (08:52 +1000)]
drm/nva3/backlight: add suppport for newer style backlight regs

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv50/backlight: express brightness level in percent
Ben Skeggs [Tue, 2 Aug 2011 10:45:35 +0000 (20:45 +1000)]
drm/nv50/backlight: express brightness level in percent

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nv50/backlight: take the sor into account when bashing regs
Ben Skeggs [Tue, 2 Aug 2011 09:29:37 +0000 (19:29 +1000)]
drm/nv50/backlight: take the sor into account when bashing regs

I'm sure that out there somewhere, someone will need this.  We currently
haven't seen an example of LVDS being on a non-0 SOR so far though.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/backlight: make more consistent with rest of driver style
Ben Skeggs [Tue, 2 Aug 2011 08:54:43 +0000 (18:54 +1000)]
drm/nouveau/backlight: make more consistent with rest of driver style

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: tidy connector hotplug handler, punt messages to debug
Ben Skeggs [Tue, 2 Aug 2011 03:57:10 +0000 (13:57 +1000)]
drm/nouveau: tidy connector hotplug handler, punt messages to debug

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: remove reliance on vbios for native displayport
Ben Skeggs [Fri, 1 Jul 2011 05:51:49 +0000 (15:51 +1000)]
drm/nouveau/dp: remove reliance on vbios for native displayport

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/dp: rewrite auxch transaction routines
Ben Skeggs [Wed, 20 Jul 2011 05:50:14 +0000 (15:50 +1000)]
drm/nouveau/dp: rewrite auxch transaction routines

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: workaround semaphore hw bug causing unnecessary interrupts
Ben Skeggs [Mon, 25 Jul 2011 10:26:19 +0000 (20:26 +1000)]
drm/nouveau: workaround semaphore hw bug causing unnecessary interrupts

The HW will only accept the DMA_FROM_MEMORY class for DMA_SEMAPHORE without
asking the driver to intervene.

It appears that semaphores will work correctly even without DMA_IN_MEMORY,
so lets avoid the large amount of interrupts generated by x-chan sync.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/tmr: fix miscalculation of ratio on pre-nv4x chipsets
Ben Skeggs [Thu, 21 Jul 2011 06:12:58 +0000 (16:12 +1000)]
drm/nouveau/tmr: fix miscalculation of ratio on pre-nv4x chipsets

The clock_get() hook returns KHz, not Hz.

Also fixed to use crystal freq from dev_priv.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nva3/pm: fixup for NVAF special
Ben Skeggs [Thu, 21 Jul 2011 05:54:48 +0000 (15:54 +1000)]
drm/nva3/pm: fixup for NVAF special

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nva3/pm: use crystal freq where appropriate
Ben Skeggs [Thu, 21 Jul 2011 05:52:52 +0000 (15:52 +1000)]
drm/nva3/pm: use crystal freq where appropriate

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: determine timing crystal freq from straps
Ben Skeggs [Thu, 21 Jul 2011 05:39:06 +0000 (15:39 +1000)]
drm/nouveau: determine timing crystal freq from straps

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nva3/pm: pll disabled if bit 0 of ctrl not set
Ben Skeggs [Tue, 19 Jul 2011 23:59:05 +0000 (09:59 +1000)]
drm/nva3/pm: pll disabled if bit 0 of ctrl not set

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: magic to make auxch on new macbooks booted in EFI mode work
Ben Skeggs [Tue, 19 Jul 2011 05:59:39 +0000 (15:59 +1000)]
drm/nouveau: magic to make auxch on new macbooks booted in EFI mode work

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau/pm: Document and expose CL and WR for 0x1002Cx
Roy Spliet [Thu, 14 Jul 2011 18:40:10 +0000 (20:40 +0200)]
drm/nouveau/pm: Document and expose CL and WR for 0x1002Cx

Signed-off-by: Roy Spliet <r.spliet@student.tudelft.nl>
13 years agodrm/nouveau/pm: add initial NV3x/NVCx memtiming support, improve other cards
Roy Spliet [Sat, 9 Jul 2011 19:18:11 +0000 (21:18 +0200)]
drm/nouveau/pm: add initial NV3x/NVCx memtiming support, improve other cards

NV30: Create framework for memtm
NV50: Improve reg creation,
NV50: Use P.version instead of card codename/stepping,
NVC0: Initial memtiming code for Fermi,
Renamed regs for consistency,
Overall redesign to improve readability,
Avoid kfree on null-pointer

Signed-off-by: Roy Spliet <r.spliet@student.tudelft.nl>
13 years agodrm/nouveau: remove special-casing of hotplug detection type
Ben Skeggs [Wed, 13 Jul 2011 06:15:57 +0000 (16:15 +1000)]
drm/nouveau: remove special-casing of hotplug detection type

If we support PGPIO interrupts, and know a hotplug GPIO tag for a
connector we use HPD, otherwise POLL_CONNECT.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0: no page flipping at the moment
Ben Skeggs [Tue, 12 Jul 2011 02:06:36 +0000 (12:06 +1000)]
drm/nvd0: no page flipping at the moment

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvc0/gr: remove max tpc count info
Ben Skeggs [Mon, 11 Jul 2011 05:54:20 +0000 (15:54 +1000)]
drm/nvc0/gr: remove max tpc count info

Just assume a max of 16 everywhere, and hope it's okay.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0: lets not attempt to dereference a nv50_display pointer
Ben Skeggs [Mon, 11 Jul 2011 05:46:01 +0000 (15:46 +1000)]
drm/nvd0: lets not attempt to dereference a nv50_display pointer

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nouveau: enable hwmon support when both nouveau/hwmon are built as modules.
Ken Milmore [Sun, 3 Jul 2011 18:54:28 +0000 (19:54 +0100)]
drm/nouveau: enable hwmon support when both nouveau/hwmon are built as modules.

The nouveau hwmon temperature support currently only functions when hwmon is
compiled into the kernel. There's no reason why this shouldn't also work when
both hwmon and nouveau are modularised (as is the case with Slackware's stock
kernels).

Signed-off-by: Ken Milmore <ken.milmore@googlemail.com>
Reviewed-by: Martin Peres <martin.peres@ensi-bourges.fr>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: tidy up what we have so far
Ben Skeggs [Fri, 8 Jul 2011 04:43:19 +0000 (14:43 +1000)]
drm/nvd0/disp: tidy up what we have so far

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: rewrite irq handler, should be somewhat sturdier now
Ben Skeggs [Fri, 8 Jul 2011 04:34:45 +0000 (14:34 +1000)]
drm/nvd0/disp: rewrite irq handler, should be somewhat sturdier now

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: do modeset irq handling from tasklet
Ben Skeggs [Fri, 8 Jul 2011 03:17:01 +0000 (13:17 +1000)]
drm/nvd0/disp: do modeset irq handling from tasklet

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: untested LVDS support
Ben Skeggs [Fri, 8 Jul 2011 02:52:14 +0000 (12:52 +1000)]
drm/nvd0/disp: untested LVDS support

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: track down fb positioning method
Ben Skeggs [Fri, 8 Jul 2011 02:11:58 +0000 (12:11 +1000)]
drm/nvd0/disp: track down fb positioning method

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: determine U table config in or_mode_set()
Ben Skeggs [Fri, 8 Jul 2011 01:53:37 +0000 (11:53 +1000)]
drm/nvd0/disp: determine U table config in or_mode_set()

Takes a gamble and presumes that we can safely store something random in
OR_MODE_CTRL+4, the hw doesn't seem to mind...

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: dac load detect
Ben Skeggs [Fri, 8 Jul 2011 01:14:50 +0000 (11:14 +1000)]
drm/nvd0/disp: dac load detect

VBIOS does more than this, as does nv50/nvc0 driver in nouveau.  Traces
of the NVIDIA binary driver however, show pretty much just this being
done...  Seems to work for me, it'll be fine for the moment.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: handle sync polarity, kill off some unknown
Ben Skeggs [Thu, 7 Jul 2011 23:43:20 +0000 (09:43 +1000)]
drm/nvd0/disp: handle sync polarity, kill off some unknown

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: scaling
Ben Skeggs [Thu, 7 Jul 2011 06:01:57 +0000 (16:01 +1000)]
drm/nvd0/disp: scaling

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: push the update button in mode_set_base()
Ben Skeggs [Thu, 7 Jul 2011 05:23:48 +0000 (15:23 +1000)]
drm/nvd0/disp: push the update button in mode_set_base()

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: fixup clut so it actually works
Ben Skeggs [Thu, 7 Jul 2011 04:49:24 +0000 (14:49 +1000)]
drm/nvd0/disp: fixup clut so it actually works

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: initial attempt at modeset irq handling
Ben Skeggs [Thu, 7 Jul 2011 00:47:10 +0000 (10:47 +1000)]
drm/nvd0/disp: initial attempt at modeset irq handling

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: stub dac load detect, prevents oops
Ben Skeggs [Wed, 6 Jul 2011 23:51:29 +0000 (09:51 +1000)]
drm/nvd0/disp: stub dac load detect, prevents oops

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: dac encoder module
Ben Skeggs [Wed, 6 Jul 2011 05:25:47 +0000 (15:25 +1000)]
drm/nvd0/disp: dac encoder module

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: extend the init voodoo to cover crtcs
Ben Skeggs [Wed, 6 Jul 2011 04:39:23 +0000 (14:39 +1000)]
drm/nvd0/disp: extend the init voodoo to cover crtcs

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: remove lp reordering from vram dmaobj, create fb dmaobjs
Ben Skeggs [Wed, 6 Jul 2011 01:40:45 +0000 (11:40 +1000)]
drm/nvd0/disp: remove lp reordering from vram dmaobj, create fb dmaobjs

Fixes CLUT being messed up.  Mostly.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: some magic to make evo happeir
Ben Skeggs [Wed, 6 Jul 2011 00:29:49 +0000 (10:29 +1000)]
drm/nvd0/disp: some magic to make evo happeir

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: they moved the linear flag..
Ben Skeggs [Tue, 5 Jul 2011 23:59:40 +0000 (09:59 +1000)]
drm/nvd0/disp: they moved the linear flag..

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: initial crtc object implementation
Ben Skeggs [Tue, 5 Jul 2011 06:48:06 +0000 (16:48 +1000)]
drm/nvd0/disp: initial crtc object implementation

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: skeletal handling of modeset interrupts
Ben Skeggs [Tue, 5 Jul 2011 04:16:05 +0000 (14:16 +1000)]
drm/nvd0/disp: skeletal handling of modeset interrupts

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: start on SOR encoder functions
Ben Skeggs [Tue, 5 Jul 2011 03:08:40 +0000 (13:08 +1000)]
drm/nvd0/disp: start on SOR encoder functions

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: setup a couple of dma objects we'll need
Ben Skeggs [Tue, 5 Jul 2011 01:58:58 +0000 (11:58 +1000)]
drm/nvd0/disp: setup a couple of dma objects we'll need

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agodrm/nvd0/disp: start on interrupt handling
Ben Skeggs [Tue, 5 Jul 2011 01:01:13 +0000 (11:01 +1000)]
drm/nvd0/disp: start on interrupt handling

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>