GitHub/MotorolaMobilityLLC/kernel-slsi.git
10 years agodrm/i915: Include HW status page in error capture
Chris Wilson [Thu, 23 Jan 2014 22:40:36 +0000 (22:40 +0000)]
drm/i915: Include HW status page in error capture

Many times in the past we have concluded that the cause of the GPU hang
has been that the hw status page was stale, usually because the GPU and
CPU disagreed over the address of the page. Having stumbled across yet
another issue that seems to be related to the HWSP, it is time to
include that information in the GPU error dump.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Always pin the default context
Chris Wilson [Thu, 23 Jan 2014 19:40:02 +0000 (19:40 +0000)]
drm/i915: Always pin the default context

Through a twisty and circuituous path it is possible to currently trick
the code into creating a default context and forgetting to pin it
immediately into the GGTT. (This requires a system using contexts without
an aliasing ppgtt, which is currently restricted to Baytrails machines
manually specifying a module parameter to force enable contexts, or
on Sandybridge and later that manually disable the aliasing ppgtt.) The
consequence is that during module unload we attempt to unpin the default
context twice and encounter a BUG remonstrating that we attempt to unpin
an unbound object.

[  161.002869] Kernel BUG at f84861f8 [verbose debug info unavailable]
[  161.002875] invalid opcode: 0000 [#1] SMP
[  161.002882] Modules linked in: coretemp kvm_intel kvm crc32_pclmul aesni_intel aes_i586 xts lrw gf128mul ablk_helper cryptd hid_sensor_accel_3d hid_sensor_gyro_3d hid_sensor_magn_3d hid_sensor_trigger industrialio_triggered_buffer kfifo_buf industrialio hid_sensor_iio_common snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm snd_page_alloc snd_seq_midi snd_seq_midi_event dm_multipath scsi_dh asix ppdev usbnet snd_rawmidi mii hid_sensor_hub microcode snd_seq rfcomm bnep snd_seq_device bluetooth snd_timer snd parport_pc binfmt_misc soundcore dw_dmac_pci dw_dmac_core mac_hid lp parport dm_mirror dm_region_hash dm_log hid_generic usbhid hid i915(O-) drm_kms_helper(O) igb dca ptp pps_core i2c_algo_bit drm(O) ahci libahci video
[  161.002991] CPU: 0 PID: 2114 Comm: rmmod Tainted: G        W  O 3.13.0-rc8+ #2
[  161.002997] Hardware name: NEXCOM VTC1010/Aptio CRB, BIOS 5.6.5 09/24/2013
[  161.003004] task: dbdd6800 ti: dbe0e000 task.ti: dbe0e000
[  161.003010] EIP: 0060:[<f84861f8>] EFLAGS: 00010246 CPU: 0
[  161.003044] EIP is at i915_gem_object_ggtt_unpin+0x88/0x90 [i915]
[  161.003050] EAX: dfce3840 EBX: 00000000 ECX: dfafd690 EDX: dfce3874
[  161.003056] ESI: c0086b40 EDI: df962e00 EBP: dbe0fe1c ESP: dbe0fe0c
[  161.003062]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[  161.003068] CR0: 8005003b CR2: b7718000 CR3: 1bec0000 CR4: 001007f0
[  161.003076] Stack:
[  161.003081]  00afc014 00000004 c0086b40 dfafc000 dbe0fe38 f8487e5a dfaa5400 c0086b40
[  161.003099]  dfafc000 dfaa5400 dfaa5414 dbe0fe58 f84741aa 00000000 f89c34b9 dfaa5414
[  161.003117]  dfaa5400 dfaa5400 f644b000 dbe0fe6c f89a5443 dfaa5400 f8505000 f644b000
[  161.003134] Call Trace:
[  161.003169]  [<f8487e5a>] i915_gem_context_fini+0xba/0x1c0 [i915]
[  161.003202]  [<f84741aa>] i915_driver_unload+0x1fa/0x2f0 [i915]
[  161.003232]  [<f89a5443>] drm_dev_unregister+0x23/0x90 [drm]
[  161.003259]  [<f89a54ed>] drm_put_dev+0x3d/0x70 [drm]
[  161.003294]  [<f8470615>] i915_pci_remove+0x15/0x20 [i915]
[  161.003306]  [<c1338a6f>] pci_device_remove+0x2f/0xa0
[  161.003317]  [<c140c871>] __device_release_driver+0x61/0xc0
[  161.003328]  [<c140d12f>] driver_detach+0x8f/0xa0
[  161.003341]  [<c140c54f>] bus_remove_driver+0x4f/0xc0
[  161.003353]  [<c140d708>] driver_unregister+0x28/0x60
[  161.003362]  [<c10cee42>] ? stop_cpus+0x32/0x40
[  161.003372]  [<c10bd510>] ? module_refcount+0x90/0x90
[  161.003383]  [<c13378c5>] pci_unregister_driver+0x15/0x60
[  161.003413]  [<f89a739f>] drm_pci_exit+0x9f/0xb0 [drm]
[  161.003458]  [<f84e624a>] i915_exit+0x1b/0x1d [i915]
[  161.003468]  [<c10bf8a8>] SyS_delete_module+0x158/0x1f0
[  161.003480]  [<c1173d5d>] ? ____fput+0xd/0x10
[  161.003488]  [<c106f0fe>] ? task_work_run+0x7e/0xb0
[  161.003499]  [<c165a68d>] sysenter_do_call+0x12/0x28
[  161.003505] Code: 0f b6 4d f3 8d 51 0f 83 e1 f0 83 e2 0f 09 d1 84 d2 88 48 54 75 07 80 a7 91 00 00 00 7f 83 c4 04 5b 5e 5f 5d c3 8d b6 00 00 00 00 <0f> 0b 8d b6 00 00 00 00 55 89 e5 57 56 53 83 ec 64 3e 8d 74 26
[  161.003586] EIP: [<f84861f8>] i915_gem_object_ggtt_unpin+0x88/0x90 [i915] SS:ESP 0068:dbe0fe0c

v2: Rename the local variable (is_default_ctx) to avoid confusion with
the function is_default_ctx(). And correct Jesse's email address.

Reported-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73985
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Ben Widawsky <benjamin.widawsky@intel.com>
Tested-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ben Widawsky <benjamin.widawsky@intel.com>
[danvet: Fix up the rebase fail from my first attempt, thankfully
pointed out by Ville.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: debugfs: Add support for probing DP sink CRC.
Rodrigo Vivi [Fri, 24 Jan 2014 15:36:17 +0000 (13:36 -0200)]
drm/i915: debugfs: Add support for probing DP sink CRC.

This debugfs interface will allow intel-gpu-tools test case
to verify if screen has been updated properly on cases like PSR.

v2: Accepted all Daniel's suggestions:
    * grab modeset lock
    * loop over connector and check DPMS on
    * return errors
    * use _eDP1 suffix for easy future extension
    * don't cache crc_supported neither latest crc
    * return crc as a full array and read it at once with aux.
    * use 0 to turn TEST_SINK off.
    * split the drm_helpers definitions in another patch.

v3: Accepted 2 Damien's suggestion: remove h from printf hexa
    and return ENODEV when eDP not present instead of EAGAIN.

v4: Accepted 2 Jani' s suggestion: 1 path for unlock and remove
    _retry from aux read.

v5: removing last missing useless _retry (by Damien)

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: dp helper: Add DP test sink CRC definition.
Rodrigo Vivi [Tue, 14 Jan 2014 18:21:49 +0000 (16:21 -0200)]
drm: dp helper: Add DP test sink CRC definition.

This address will be used to verify panel CRC for test and
validation purposes.

Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
[danvet: Fix whitespace fail.]
Acked-by: Dave Airlie <airlied@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Fix FBC_FENCE_OFF
Ville Syrjälä [Thu, 23 Jan 2014 14:49:17 +0000 (16:49 +0200)]
drm/i915: Fix FBC_FENCE_OFF

Having a 4 byte register at 0x321b seems unlikely as that's not
4 byte aligned. Since later platforms have more or less the same FBC
registers with new names, assume that FBC_FENCE_OFF is at 0x3218 just
like DPFC_FENCE_YOFF.

This feels like a simple typo in BSpec. 321Bh looks a lot like 3218h
after all.

Should still be tested on real hardware of course. But I don't have
any mobile gen4 systems.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Fix FBC1 enable message
Ville Syrjälä [Thu, 23 Jan 2014 14:49:16 +0000 (16:49 +0200)]
drm/i915: Fix FBC1 enable message

The debug message telling FBC1 has been enabled is missing a newline.
Add it.

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: Don't preserve DPFC_CONTROL bits ILK/SNB
Ville Syrjälä [Thu, 23 Jan 2014 14:49:14 +0000 (16:49 +0200)]
drm/i915: Don't preserve DPFC_CONTROL bits ILK/SNB

On CTG and IVB+ we don't try to preserve any bits from the
DPFC_CONTROL register. Follow suit on ILK/SNB.

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: Kill most of the FBC register save/restore
Ville Syrjälä [Thu, 23 Jan 2014 14:49:15 +0000 (16:49 +0200)]
drm/i915: Kill most of the FBC register save/restore

We will anyway re-enable FBC normally after resume, so trying to save
and restore the register makes little sense.

We do need to preserve the FBC1 interval bits in FBC_CONTROL since
we only initialize them during driver load, and try to preserve them
after that.

v2: s/I915_HAS_FBC/HAS_FBC/ and fix the check for gen4

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: Actually write the correct bits to DPFC_CONTROL on CTG
Ville Syrjälä [Thu, 23 Jan 2014 14:49:13 +0000 (16:49 +0200)]
drm/i915: Actually write the correct bits to DPFC_CONTROL on CTG

We set up all the bits for DPFC_CONTROL but forgot to actually
write them to the register. Oops.

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: Use 1/2 compression ratio limit for 16bpp on FBC2
Ville Syrjälä [Thu, 23 Jan 2014 14:49:12 +0000 (16:49 +0200)]
drm/i915: Use 1/2 compression ratio limit for 16bpp on FBC2

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: Improve FBC plane defines a bit
Ville Syrjälä [Thu, 23 Jan 2014 14:49:11 +0000 (16:49 +0200)]
drm/i915: Improve FBC plane defines a bit

Make the FBC plane macros take the plane as a parameter.

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: Don't set DPFC_HT_MODIFY bit on CTG/ILK/SNB
Ville Syrjälä [Thu, 23 Jan 2014 14:49:10 +0000 (16:49 +0200)]
drm/i915: Don't set DPFC_HT_MODIFY bit on CTG/ILK/SNB

The ILK/SNB docs don't really mention the the DPFC_HT_MODIFY bit.
CTG docs clearly state that it should be set only when tracking
back buffer modification in persistent mode. The bit is supposed
to be set by software after the first CPU modification to the
back buffer, and it would get automagically cleared by the hardware
on the next page flip.

Since we only track front buffer modification we don't need to set
this bit. GTT modification tracking still appears to work on ILK
and SNB with the bit unset. I don't have a CTG to verify how that
behaves.

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: Don't set persistent FBC mode on ILK/SNB
Ville Syrjälä [Thu, 23 Jan 2014 14:49:09 +0000 (16:49 +0200)]
drm/i915: Don't set persistent FBC mode on ILK/SNB

The ILK/SNB docs are a bit unclear what the persistent mode does, but
the CTG docs clearly state that it was meant to be used when we're
tracking back buffer modifications. We never do that, so leave it in
non-persistent mode.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Don't write IVB_FBC_RT_BASE
Ville Syrjälä [Thu, 23 Jan 2014 14:49:08 +0000 (16:49 +0200)]
drm/i915: Don't write IVB_FBC_RT_BASE

We use nuking instead of render tracking on IVB+, so there's
no point in writing IVB_FBC_RT_BASE.

v2: Drop the IVB_FBC_RT_BASE write too
v3: Move the SNB stuff elsewhere, leaving only IVB+ here

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 agoMerge branch 'topic/ppgtt' into drm-intel-next-queued
Daniel Vetter [Sat, 25 Jan 2014 20:14:57 +0000 (21:14 +0100)]
Merge branch 'topic/ppgtt' into drm-intel-next-queued

Because whatever.*

* This should contain a fairly long list of issues and still
unresolved resgressions, but I didn't really get a vote.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Make underruns DRM_ERROR
Ville Syrjälä [Fri, 17 Jan 2014 09:44:32 +0000 (11:44 +0200)]
drm/i915: Make underruns DRM_ERROR

I want to see these without having full debugs enabled.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: fix the gen8 irq handler as spotted by Paulo in his review.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Limit FIFO underrun reports on GMCH platforms
Ville Syrjälä [Fri, 17 Jan 2014 09:44:31 +0000 (11:44 +0200)]
drm/i915: Limit FIFO underrun reports on GMCH platforms

Currently we print all pipe underruns on GMCH platforms. Hook up the
same logic we use on PCH platforms where we disable the underrun
reporting after the first underrun.

Underruns don't actually generate interrupts themselves on GMCH
platforms, we just can detect them whenever we service other
interrupts. So we don't have any enable bits to worry about. We just
need to remember to clear the underrun status when enabling underrun
reporting.

Note that the underrun handling needs to be moved to the non-locked
pipe_stats[] loop in the interrupt handlers to avoid having to rework
the locking in intel_set_cpu_fifo_underrun_reporting().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Place the Global GTT VM first in the list of VM
Chris Wilson [Thu, 9 Jan 2014 22:57:22 +0000 (22:57 +0000)]
drm/i915: Place the Global GTT VM first in the list of VM

This is useful for debugging as we then know that the first entry is
always the global GTT, and all later entries the per-process GTT VM.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: fix WRPLL clock calculation
Jesse Barnes [Wed, 22 Jan 2014 20:58:04 +0000 (12:58 -0800)]
drm/i915: fix WRPLL clock calculation

Forgot to convert to using the refclk variable when I added refclk
readout support, and Paulo noticed the resulting calculation was off due
to the way p & r are stored.

Reported-by: Paulo Zanoni <przanoni@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Remove incorrect comment about struct mutex
Ben Widawsky [Wed, 22 Jan 2014 00:55:01 +0000 (16:55 -0800)]
drm/i915: Remove incorrect comment about struct mutex

This statenment became false here:

commit 4fc688ce79772496503d22263d61b071a8fb596e
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Fri Nov 2 11:14:01 2012 -0700

    drm/i915: protect RPS/RC6 related accesses (including PCU) with a new mutex

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: always check clocks when comparing pipe configs
Jesse Barnes [Mon, 20 Jan 2014 22:18:04 +0000 (14:18 -0800)]
drm/i915: always check clocks when comparing pipe configs

Now that we have DDI support, we can check these all the time.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: clock readout support for DDI v3
Jesse Barnes [Tue, 21 Jan 2014 20:42:10 +0000 (12:42 -0800)]
drm/i915: clock readout support for DDI v3

Read out and calculate the port and pixel clocks on DDI configs as well.
This means we have to grab the DP divider values and look at the port
mapping to figure out which clock select reg to read out.

v2: do the work from ddi_get_config (Ville)
v3: check WRPLL reference clock (Ville)
    add additional SPLL freqs (Ville)
    clean up port/crtc clock calc (Ville)
    fix up crtc_clock conditionals (Ville)
    drop superfluous dp_get_m_n from get_config (Ville)

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: Introduce a get_aux_send_ctl() vfunc
Damien Lespiau [Tue, 21 Jan 2014 13:37:15 +0000 (13:37 +0000)]
drm/i915: Introduce a get_aux_send_ctl() vfunc

We need a bit more flexibility here in the future, bits get shuffled
around.

v2: more descriptive commit message (Jani Nikula)

Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Reorder the AUX_CTL bits in descending order
Damien Lespiau [Mon, 20 Jan 2014 15:52:31 +0000 (15:52 +0000)]
drm/i915: Reorder the AUX_CTL bits in descending order

So it's easier to compare what we program with the documentation, not
having to jump at all.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Factor out a function returning the AUX_CTL value to start a send
Damien Lespiau [Mon, 20 Jan 2014 15:52:30 +0000 (15:52 +0000)]
drm/i915: Factor out a function returning the AUX_CTL value to start a send

Also, move that computation outside of the for loop that tries 5 times,
this value doesn't change between tries.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Turn get_aux_clock_divider() into per-platform vfuncs
Damien Lespiau [Tue, 21 Jan 2014 13:35:39 +0000 (13:35 +0000)]
drm/i915: Turn get_aux_clock_divider() into per-platform vfuncs

A tiny clean-up to allow better code separation between platforms.

v2: Fix comment placement (put in in i9xx_get_aux_clock_divider()) and
    nuke the outdated PCH eDP comment (Jani Nikula)

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: quirk invert brightness for Acer Aspire 5336
Jani Nikula [Mon, 13 Jan 2014 15:30:34 +0000 (17:30 +0200)]
drm/i915: quirk invert brightness for Acer Aspire 5336

Since
commit ee1452d7458451a7508e0663553ce88d63958157
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Fri Sep 20 15:05:30 2013 +0300

    drm/i915: assume all GM45 Acer laptops use inverted backlight PWM

failed and was later reverted in
commit be505f643925e257087247b996cd8ece787c12af
Author: Alexander van Heukelum <heukelum@fastmail.fm>
Date:   Sat Dec 28 21:00:39 2013 +0100

    Revert "drm/i915: assume all GM45 Acer laptops use inverted backlight PWM"

fix the individual broken machine instead.

Note to backporters:

http://patchwork.freedesktop.org/patch/17837/

is the patch you want for 3.13 and older.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=54171
Reference: http://mid.gmane.org/DUB115-W7628C7C710EA51AA110CD4A5000@phx.gbl
CC: stable@vger.kernel.org
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Patch mangling for 3.14 plus adding the link to the original
for 3.13.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: drop the i915.fbpercrtc module parameter
Jani Nikula [Tue, 21 Jan 2014 09:24:24 +0000 (11:24 +0200)]
drm/i915: drop the i915.fbpercrtc module parameter

It's unused, and nowadays specifying unknown parameters no longer
prevents modules from being loaded.

Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Enable 5.4Ghz (HBR2) link rate for Displayport 1.2-capable devices
Todd Previte [Mon, 20 Jan 2014 17:19:39 +0000 (10:19 -0700)]
drm/i915: Enable 5.4Ghz (HBR2) link rate for Displayport 1.2-capable devices

For HSW+ platforms, enable the 5.4Ghz (HBR2) link rate for devices that support it. The
sink device must report that is supports Displayport 1.2 and the HBR2 bit rate in the
DPCD in order to use HBR2.

Signed-off-by: Todd Previte <tprevite@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Shuffle sprite register writes into a tighter group
Ville Syrjälä [Fri, 17 Jan 2014 18:09:03 +0000 (20:09 +0200)]
drm/i915: Shuffle sprite register writes into a tighter group

Group the sprite register writes a bit tighter. We want to write
the registers atomically, and so doing the base address/offset
artihmetic within the critical section is pointless when it can
all be done beforehand.

Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Shuffle modeset reset handling around
Daniel Vetter [Thu, 16 Jan 2014 21:28:44 +0000 (22:28 +0100)]
drm/i915: Shuffle modeset reset handling around

Currently we're doing the reset handling a bit late, and we're doing
it both in the driver load code and on resume. This makes it unusable
for e.g. resetting the panel power sequence state like Paulo wants to.

Instead of adding yet another single-use callback shuffle things
around:
- Output handling code is responsible to reset/init all state on its
  own at driver load time.
- We call the reset functions much earlier, before we start using any
  of the modeset code.

Compared to Paulo's new ->resume callback the only difference in
placement is that ->reset is still called without dev->struct_mutex
held. Which is imo a feature.

v2: Rebase on top of the now merge dinq.

Cc: Paulo Zanoni <przanoni@gmail.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: set the backlight panel delays registers to 1
Paulo Zanoni [Thu, 19 Dec 2013 16:29:44 +0000 (14:29 -0200)]
drm/i915: set the backlight panel delays registers to 1

Because we already do the wait in software: see
ironlake_wait_backlight_on and ironlake_edp_wait_backlight_off.

For the "backlight on" delay, even BSpec says we need to program 0x1
to PP_ON_DELAYS 12:0.

For the "backlight off" delay, if we don't do the same thing, when we
call ironlake_wait_panel_off we'll end up waiting for the it again.

On my machine the off delay is 200ms, so we save this amount of time
whenever we disable the panel (e.g, suspend).

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Fix new_config and new_enabled for load detect
Ville Syrjälä [Fri, 17 Jan 2014 13:59:39 +0000 (15:59 +0200)]
drm/i915: Fix new_config and new_enabled for load detect

I forgot to set new_config and new_enabled appropriately in the load
detect code. Fix it up.

v2: Handle the other error path in intel_get_load_detect_pipe() too (Imre)

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Kill dev_priv->irq_received
Ville Syrjälä [Fri, 17 Jan 2014 09:35:16 +0000 (11:35 +0200)]
drm/i915: Kill dev_priv->irq_received

Not sure anyone cares about this information. I suppose most people
would just look at /proc/interrupts instead.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Make irq_received bool
Ville Syrjälä [Fri, 17 Jan 2014 09:35:15 +0000 (11:35 +0200)]
drm/i915: Make irq_received bool

irq_received is used as a boolean in i965_irq_handler(), so make it
bool. This also makes i965_irq_handler() closer to i915_irq_handler().

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewd-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Add intel_hpd_irq_uninstall()
Ville Syrjälä [Fri, 17 Jan 2014 11:43:51 +0000 (13:43 +0200)]
drm/i915: Add intel_hpd_irq_uninstall()

Add intel_hpd_irq_uninstall() which will cancel the hotplug re-enable
timer.

Also s/i915_reenable_hotplug_timer_func/intel_hpd_irq_reenable/

Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
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: don't wait for power cycle when waiting for power off
Paulo Zanoni [Thu, 19 Dec 2013 16:29:43 +0000 (14:29 -0200)]
drm/i915: don't wait for power cycle when waiting for power off

Function ironlake_wait_panel_off should just wait for the power off
delay, while function ironlake_wait_panel_power_cycle should wait for
the panel cycle (that's required after we turn the panel off, before
we enable it again).

The problem is that, currently, ironlake_wait_panel_off is waiting not
just for the panel to be off, but also for the power cycle delay and
the backlight off delay. This function relies on the PP_STATUS bits
3:0, which are not documented and not supposed to be used. A quick
analysis of the values we get while waiting quickly shows that power
off is reached while bits 3:0 are still 0x1, and the time it takes to
become 0x0 is the power cycle delay.

On my system with backlight off delay of 200ms, power down delay of
50ms and power cycle delay of 500ms, this is what I get:
 - Start waiting with value 0x80000008, timestamp 6.429364.
 - Jumps to 0xa0000003, timestamp 6.431360 (time waited: 0.001996)
 - Jumps to 0xa0000002, timestamp 6.631277 (time waited: 0.201913)
 - Jumps to 0x08000001, timestamp 6.681258 (time waited: 0.251894)
 - Jumps to 0x00000000, timestamp 7.192012 (time waited: 0.762648)

As you can see, ironlake_wait_panel_off is sleeping 760ms instead of
the expected 50ms: the first 200ms matches the backlight off delay
(which we should already have waited for!), then the 50ms for the real
panel off delay, then the 500ms for the panel power cycle.

This patch makes is look just at bits 31 and 29:28, which will ignore
the panel power cycle.

And just to be clear: this saves 500ms on my system every time we
disable the panel. But we can still save 200ms more (the backlight off
delay) on the next patches.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuougseek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: remove a column of zeros from the eDP wait definitions
Paulo Zanoni [Thu, 19 Dec 2013 16:29:42 +0000 (14:29 -0200)]
drm/i915: remove a column of zeros from the eDP wait definitions

I like how the macros are nicely column-aligned, so we can properly
compare what each macro waits for, but a column full of zeroes doesn't
really help anything: it just makes the lines bigger, and they're
already way past 80 columns. I imagine this column was used in the
past, but IMHO now we can get rid of it.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: drop ironlake_ prefix from edp panel/backlight functions
Daniel Vetter [Fri, 17 Jan 2014 13:39:48 +0000 (14:39 +0100)]
drm/i915: drop ironlake_ prefix from edp panel/backlight functions

They now also work on vlv, which has the regs somewhere else. And
daring a glance into the looking glass it seems like this
functionality will continue to work the same for the next few hardware
platforms.

So it's better to just remove that misleading prefix and have a bit
shorter code for better readability.

The only exceptions are the panel/backlight functions shared with
intel_ddi.c, those get an intel_ prefix.

While at it make the vdd_on/off functions static.

And one straggler was missing the edp_ in the name, so make everything
neatly OCD.

Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: save some time when waiting the eDP timings
Paulo Zanoni [Thu, 19 Dec 2013 16:29:40 +0000 (14:29 -0200)]
drm/i915: save some time when waiting the eDP timings

The eDP spec defines some points where after you do action A, you have
to wait some time before action B. The thing is that in our driver
action B does not happen exactly after action A, but we still use
msleep() calls directly. What this patch does is that we record the
timestamp of when action A happened, then, just before action B, we
look at how much time has passed and only sleep the remaining amount
needed.

With this change, I am able to save about 5-20ms (out of the total
200ms) of the backlight_off delay and completely skip the 1ms
backlight_on delay. The 600ms vdd_off delay doesn't happen during
normal usage anymore due to a previous patch.

v2: - Rename ironlake_wait_jiffies_delay to intel_wait_until_after and
      move it to intel_display.c
    - Fix the msleep call: diff is in jiffies
v3: - Use "tmp_jiffies" so we don't need to worry about the value of
      "jiffies" advancing while we're doing the math.
v4: - Rename function again.
    - Move function to i915_drv.h.
    - Store last_power_cycle at edp_panel_off too.
    - Use msecs_to_jiffies_timeout, then replace the msleep with an
      open-coded version that avoids the extra +1 jiffy.
    - Try to add units to every variable name so we don't confuse
      jiffies with milliseconds.

Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: init the DP panel power seq variables earlier
Paulo Zanoni [Thu, 19 Dec 2013 16:29:39 +0000 (14:29 -0200)]
drm/i915: init the DP panel power seq variables earlier

Our driver has two different ways of waiting for panel power
sequencing delays. One of these ways is through
ironlake_wait_panel_status, which implicitly uses the values written
to our registers. The other way is through the functions that call
intel_wait_until_after, and on this case we do direct msleep() calls
on the intel_dp->xxx_delay variables.

Function intel_dp_init_panel_power_sequencer is responsible for
initializing the _delay variables and deciding which values we need to
write to the registers, but it does not write these values to the
registers. Only at intel_dp_init_panel_power_sequencer_registers we
actually do this write.

Then problem is that when we call intel_dp_i2c_init, we will get some
I2C calls, which will trigger a VDD enable, which will make use of the
panel power sequencing registers and the _delay variables, so we need
to have both ready by this time. Today, when this happens, the _delay
variables are zero (because they were not computed) and the panel
power sequence registers contain whatever values were written by the
BIOS (which are usually correct).

What this patch does is to make sure that function
intel_dp_init_panel_power_sequencer is called earlier, so by the time
we call intel_dp_i2c_init, the _delay variables will already be
initialized. The actual registers won't contain their final values,
but at least they will contain the values set by the BIOS.

The good side is that we were reading the values, but were not using
them for anything (because we were just skipping the msleep(0) calls),
so this "fix" shouldn't fix any real existing bugs. I was only able to
identify the problem because I added some debug code to check how much
time time we were saving with my previous patch.

Regression introduced by:
    commit ed92f0b239ac971edc509169ae3d6955fbe0a188
    Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
    Date:   Wed Jun 12 17:27:24 2013 -0300
        drm/i915: extract intel_edp_init_connector

v2: - Rewrite commit message.

Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Only restore backlight combination mode reg for ums
Daniel Vetter [Thu, 16 Jan 2014 15:42:54 +0000 (16:42 +0100)]
drm/i915: Only restore backlight combination mode reg for ums

This was forgotten in

commit 565ee3897f0cb1e9b09905747b3784e6605767e8
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Wed Nov 13 12:56:29 2013 +0200

    drm/i915: do not save/restore backlight registers in KMS

Since the confusion was likely due to the duplicated definition for
this pci config register, let's unify that, too.

Cc: Jani Nikula <jani.nikula@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: clean up HPD IRQ debug printing
Imre Deak [Thu, 16 Jan 2014 17:56:54 +0000 (19:56 +0200)]
drm/i915: clean up HPD IRQ debug printing

Atm, we don't print these events for all platforms and for VLV/G4X we
also print them for DP AUX completion events which is unnecessary spam.
Fix both issues.

Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Don't use i915_preliminary_hw_support to mean pre-production
Damien Lespiau [Thu, 16 Jan 2014 16:51:35 +0000 (16:51 +0000)]
drm/i915: Don't use i915_preliminary_hw_support to mean pre-production

Those are two distinct concepts. Just use a comment to remind us to
remove that W/A at some point.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Set crtc->new_config to NULL for pipes that are about to be disabled
Ville Syrjälä [Tue, 14 Jan 2014 12:31:38 +0000 (14:31 +0200)]
drm/i915: Set crtc->new_config to NULL for pipes that are about to be disabled

crtc->new_config is only relevant for pipes that are going to be active
post-modeset. Set the pointer to NULL for all pipes that are going to
be disabled. This is done to help catch bugs where some piece of code
would go looking at crtc->new_config even if the data there is stale.

v2: Clear new_config in disable_crtc_nofb() too (Imre)

Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Don't oops if the initial modeset fails
Ville Syrjälä [Fri, 10 Jan 2014 09:28:09 +0000 (11:28 +0200)]
drm/i915: Don't oops if the initial modeset fails

If the first modeset operation fails, we will attempt to restore the
previous configuration that we read out from the hardware. But as we
don't yet reconstruct the framebuffer information, we end up calling
the modeset code with an enabled crtc but with fb==NULL. This will
lead to an oops within the modeset code.

Check for NULL fb when restoring the configuration, and instead of
oopsing simply disable the pipe.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Use new_config and new_enabled to simplify the VLV cdclk code
Ville Syrjälä [Fri, 10 Jan 2014 09:28:08 +0000 (11:28 +0200)]
drm/i915: Use new_config and new_enabled to simplify the VLV cdclk code

On VLV we need to compute the new cdclk before we've updated the current
state. The code achieved that in a somewhat complex way. Now that we
have new_enabled and new_config, we can simplify the code quite a bit.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Prepare to track new pipe config per pipe
Ville Syrjälä [Fri, 10 Jan 2014 09:28:07 +0000 (11:28 +0200)]
drm/i915: Prepare to track new pipe config per pipe

Add a new_config pointer to intel_crtc which will point to the new pipe
config for said crtc while intel_crtc.config will still contain the old
config during first parts of the modeset operation. This is a step
towards having the entire new state available during the compute phase,
so that we can make accurate decisions about global resource usage.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Pre-compute pipe enabled state
Ville Syrjälä [Fri, 10 Jan 2014 09:28:06 +0000 (11:28 +0200)]
drm/i915: Pre-compute pipe enabled state

Add 'new_enabled' to intel_crtc and precompute it alongside new_encoder
and new_crtc. This will allow making decisions about shared resources
that are affected by the set of active pipes, before we've clobbered
anything for real.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoRevert "drm/i915: Mask reserved bits in display/sprite address registers"
Daniel Vetter [Fri, 24 Jan 2014 09:31:44 +0000 (10:31 +0100)]
Revert "drm/i915: Mask reserved bits in display/sprite address registers"

This reverts commit 446f254566ea8911c9e19c7bc8a162fc0e53cf31.

I've left the masking in the pageflip code since that seems to be some
useful piece of preemptive robustness.

Iirc I've merged this patch under the assumption that the BIOS leaves
some random gunk in the lower bits and gets unhappy if we trample on
them. We have quite a few case like this, so this made sense.

Now I've just learned that there's actual hardware features bits in
the low 12 bits, and the kernel needs to preserve them to allow a
userspace blob to do its job. Given Dave Airlie's clear stance on
userspace blob drivers I've quickly chatted with him and he doesn't
seem too happy. So let's revert this.

If there are indeed bits that we must preserve in this range then we
can ressurrect this patch, but with proper documentation for those
bits supplied. And we probably also need to think a bit about
interactions with our driver.

Cc: Armin Reese <armin.c.reese@intel.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Dave Airlie <airlied@linux.ie>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: VLV2 - Fix hotplug detect bits
Todd Previte [Thu, 23 Jan 2014 07:13:41 +0000 (00:13 -0700)]
drm/i915: VLV2 - Fix hotplug detect bits

Add new definitions for hotplug live status bits for VLV2 since they're
in reverse order from the gen4x ones.

Changelog:
- Restored gen4 bit definitions
- Added new definitions for VLV2
- Added platform check for IS_VALLEYVIEW() in dp_detect to use the correct
  bit defintions
- Replaced a lost trailing brace for the added switch()

Signed-off-by: Todd Previte <tprevite@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73951
[danvet: Switch to _VLV postfix instead of prefix and regroupg
comments again so that the g4x warning is right next to those defines.
Also add a _G4X suffix for those special ones. Also cc stable.]
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Allow reading the TIMESTAMP register on Gen8.
Kenneth Graunke [Tue, 21 Jan 2014 22:42:38 +0000 (14:42 -0800)]
drm/i915: Allow reading the TIMESTAMP register on Gen8.

Nothing's changed here; we just need to bump the generation check.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Repeat evictions whilst pageflip completions are outstanding
Chris Wilson [Mon, 20 Jan 2014 10:17:37 +0000 (10:17 +0000)]
drm/i915: Repeat evictions whilst pageflip completions are outstanding

Since an old pageflip will keep its scanout buffer object pinned until
it has executed its unpin task on the common workqueue, we can clog up
our GGTT with stale pinned objects. As we cannot flush those workqueues
without dropping our locks, we have to resort to falling back to
userspace and telling them to repeat the operation in order to have a
chance to run our workqueues and free up the required memory. If we
fail, then we are forced to report ENOSPC back to userspace causing the
operation to fail and best-case scenario is that it introduces temporary
corruption.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Wait for completion of pending flips when starved of fences
Chris Wilson [Mon, 20 Jan 2014 10:17:36 +0000 (10:17 +0000)]
drm/i915: Wait for completion of pending flips when starved of fences

On older generations (gen2, gen3) the GPU requires fences for many
operations, such as blits. The display hardware also requires fences for
scanouts and this leads to a situation where an arbitrary number of
fences may be pinned by old scanouts following a pageflip but before we
have executed the unpin workqueue. This is unpredictable by userspace
and leads to random EDEADLK when submitting an otherwise benign
execbuffer. However, we can detect when we have an outstanding flip and
so cause userspace to wait upon their completion before finally
declaring that the system is starved of fences. This is really no worse
than forcing the GPU to stall waiting for older execbuffer to retire and
release their fences before we can reallocate them for the next
execbuffer.

v2: move the test for a pending fb unpin to a common routine for
later reuse during eviction

Reported-and-tested-by: dimon@gmx.net
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73696
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: don't disable DP port after a failed link training
Imre Deak [Fri, 17 Jan 2014 13:46:43 +0000 (15:46 +0200)]
drm/i915: don't disable DP port after a failed link training

Atm after a failed link training we disable the DP port. This can happen
during a modeset-enable or a DP link re-establishment. The latter can be
a problem and we shouldn't disable the DP port, see the previous patch for
the reasoning. In the former case the right thing would be to disable
the DP port, but also the rest of the pipe.

As a stop-gap solution leave the DP port enabled in both cases. It is an
improvement on its own (avoiding HW lock ups) and the proper solution
for the first case requires a bigger change, so let's keep that on the
TODO list.

v2:
- fix explanation of change impact (Chris)

Suggested-by: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: don't disable the DP port if the link is lost
Imre Deak [Thu, 16 Jan 2014 16:35:57 +0000 (18:35 +0200)]
drm/i915: don't disable the DP port if the link is lost

Currently if the DP link is lost (either because of a hot unplug, or
failed link status check) we disable the DP port, but leave the rest
of the pipe running. This is incompatible with the modeset disabling
sequence of some platforms/configurations. At least this is the case for
DP ports on the CPU as opposed to PCH.

Atm we'll also get a warning when we do a modeset disable after the
above link lost event, since we expect the DP port to be enabled at this
point (see the bugzilla ticket for the related dmesg).

Note that with this patch we'll still end up disabling the port, thanks
to the HPD uevent and subsequent modeset disable.

See also the next patch fixing the other half of this issue.

Solution suggested by Ville.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70570
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Eliminate lots of WARNs when there's no backlight present
Ville Syrjälä [Thu, 16 Jan 2014 16:27:15 +0000 (18:27 +0200)]
drm/i915: Eliminate lots of WARNs when there's no backlight present

My 855gm doesn't register the intel backlight but it still ends up
calling the backlight code to enable/disable the backlight via the
LVDS code. This leads to some WARNs due to backlight.max being 0.

Let's have intel_panel_enable_backlight() and intel_panel_disable_backlight()
check whether there's a backlight present or not.

Also move the backlight.present check from asle_set_backlight() into
intel_panel_set_backlight() for some extra symmetry.

Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: g4x/vlv: fix dp aux interrupt mask
Imre Deak [Thu, 16 Jan 2014 17:56:53 +0000 (19:56 +0200)]
drm/i915: g4x/vlv: fix dp aux interrupt mask

Fix typo possibly leading to timed out DP aux transactions on ports C,D.

Introduced in:

Commmit 4aeebd7443e36b0a40032e518a9338f48bd27efc
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Oct 31 09:53:36 2013 +0100

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72210
Signed off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/ppgtt: Defer request freeing on reset
Ben Widawsky [Wed, 1 Jan 2014 18:15:13 +0000 (10:15 -0800)]
drm/i915/ppgtt: Defer request freeing on reset

We need to defer the free request until the object/vma is capable of
being freed - or else we have a problem  when we try to destroy the
context.

The exact same issue is described and fixed here:
commit e20780439b26ba95aeb29d3e27cd8cc32bc82a4c
Author: Ben Widawsky <ben@bwidawsk.net>
Date:   Fri Dec 6 14:11:22 2013 -0800

    drm/i915: Defer request freeing

I had this fix previously, but decided not to keep it for some reason I
can no longer remember.

gem_reset_stats is a really good test at hitting the problem.

For the inquisitive:
[  170.516392] ------------[ cut here ]------------
[  170.517227] WARNING: CPU: 1 PID: 105 at drivers/gpu/drm/drm_mm.c:578 drm_mm_takedown+0x2e/0x30 [drm]()
[  170.518064] Memory manager not clean during takedown.
[  170.518941] CPU: 1 PID: 105 Comm: kworker/1:1 Not tainted 3.13.0-rc4-BEN+ #28
[  170.519787] Hardware name: Hewlett-Packard HP EliteBook 8470p/179B, BIOS 68ICF Ver. F.02 04/27/2012
[  170.520662] Call Trace:
[  170.521517]  [<ffffffff814f0589>] dump_stack+0x4e/0x7a
[  170.522373]  [<ffffffff81049e6d>] warn_slowpath_common+0x7d/0xa0
[  170.523227]  [<ffffffff81049edc>] warn_slowpath_fmt+0x4c/0x50
[  170.524079]  [<ffffffffa06c414e>] drm_mm_takedown+0x2e/0x30 [drm]
[  170.524934]  [<ffffffffa07213f3>] gen6_ppgtt_cleanup+0x23/0x110
[i915]
[  170.525777]  [<ffffffffa07837ed>] ppgtt_release.part.5+0x24/0x29
[i915]
[  170.526603]  [<ffffffffa071aaa5>] i915_gem_context_free+0x195/0x1a0
[i915]
[  170.527423]  [<ffffffffa071189d>] i915_gem_free_request+0x9d/0xb0
[i915]
[  170.528247]  [<ffffffffa0718af9>] i915_gem_reset+0x1f9/0x3f0 [i915]
[  170.529065]  [<ffffffffa0700cce>] i915_reset+0x4e/0x180 [i915]
[  170.529870]  [<ffffffffa070829d>] i915_error_work_func+0xcd/0x120
[i915]
[  170.530666]  [<ffffffff8106c13a>] process_one_work+0x1fa/0x6d0
[  170.531453]  [<ffffffff8106c0d8>] ? process_one_work+0x198/0x6d0
[  170.532230]  [<ffffffff8106c72b>] worker_thread+0x11b/0x3a0
[  170.532996]  [<ffffffff8106c610>] ? process_one_work+0x6d0/0x6d0
[  170.533771]  [<ffffffff810743ef>] kthread+0xff/0x120
[  170.534548]  [<ffffffff810742f0>] ? insert_kthread_work+0x80/0x80
[  170.535322]  [<ffffffff814f97ac>] ret_from_fork+0x7c/0xb0
[  170.536089]  [<ffffffff810742f0>] ? insert_kthread_work+0x80/0x80
[  170.536847] ---[ end trace 3d4c12892e42d58f ]---

v2: Whitespace fix. (Chris)

Note: This is a bug that only hits the ppgtt topic branch but I've
figured that doing the request cleanup in this order is generally the
right thing to do.

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Add a code comment to clarify what's actually going on since
the lifetime rules aroung ppgtt cleanup are ... fuzzy a best atm. Also
add a note about why we need this.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoi915: send D1 opregion notification
Kristen Carlson Accardi [Tue, 14 Jan 2014 23:36:15 +0000 (15:36 -0800)]
i915: send D1 opregion notification

The opregion notification for runtime suspend is currently D1, not D3.

Signed-off-by: Kristen Carlson Accardi <kristen@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915/bdw: remove preliminary_hw_support flag from BDW
Jesse Barnes [Tue, 14 Jan 2014 19:04:16 +0000 (11:04 -0800)]
drm/i915/bdw: remove preliminary_hw_support flag from BDW

It ought to work ok in 3.14.  We have some fun stuff coming after that,
but all the basics are in place now and seem relatively stable.

Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Acked by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Tune down reset_stat output from ERROR to debug
Daniel Vetter [Tue, 14 Jan 2014 10:40:54 +0000 (11:40 +0100)]
drm/i915: Tune down reset_stat output from ERROR to debug

This is user-triggerable and hence we should not allow it to spam
dmesg. Also, it upsets the nice dmesg tracking piglit does.

Note that this is just extra debugging information, mostly
unwanted, in case of a hang and that there is a separate message to the
user giving instructions on how to report a bug for a GPU hang.

v2: Add note as suggests in Chris' reply.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=72740
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Make semaphore modparam RO
Ben Widawsky [Tue, 17 Dec 2013 23:06:42 +0000 (15:06 -0800)]
drm/i915: Make semaphore modparam RO

A couple patches in the upcoming rework of semaphores will break if
semaphores are toggled by the user at various times. Since the code
cleanups there seem to be an overall win, and toggling semaphores at
runtime is not a terribly useful thing to do, simply make the module
parameter read-only.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Fix disabled semaphores
Ben Widawsky [Wed, 18 Dec 2013 04:06:00 +0000 (20:06 -0800)]
drm/i915: Fix disabled semaphores

The ring will emit too many if semaphores are disabled since we do not
add the correct number to num_dwords anymore.

This was introduced:
commit 52ed23253b68e1cf154b03d91bed619504cf955b
Author: Ben Widawsky <benjamin.widawsky@intel.com>
Date:   Mon Dec 16 20:50:38 2013 -0800

    drm/i915: Don't emit mbox updates without semaphores

FWIW, the bug was fixed later in the series.

/me hangs head in shame.

Daniel: Also note that we should have merged the read-only semaphore
modparam before this patch.

Reported-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Clarify relocation errnos
Ben Widawsky [Thu, 26 Dec 2013 21:39:50 +0000 (13:39 -0800)]
drm/i915: Clarify relocation errnos

While trying to find a random -EINVAL from a failing test, I noticed we
had a few hard to follow return values.

The first two hunks in this patch replace completely useless
initialization of ret. The last several hunks help to distinguish
between altering 'return ret' and 'return <ERROR>'

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Spelling s/auxilliary/auxiliary/
Geert Uytterhoeven [Sun, 12 Jan 2014 13:08:43 +0000 (14:08 +0100)]
drm/i915: Spelling s/auxilliary/auxiliary/

Signed-off-by: Geert Uytterhoeven <geert+renesas@linux-m68k.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoMerge branch 'drm-vbl-timestamp' of git://gitorious.org/vsyrjala/linux into drm-next
Dave Airlie [Tue, 21 Jan 2014 23:13:13 +0000 (09:13 +1000)]
Merge branch 'drm-vbl-timestamp' of git://gitorious.org/vsyrjala/linux into drm-next

Here's the vblank timestamp pull request you wanted.

I addressed the few bugs that Mario pointed out and added
the r-bs.

As it has been a while since I made the changes, I gave it a
quick spin on a few different i915 machines. Fortunately
everything still seems to be fine.

* 'drm-vbl-timestamp' of git://gitorious.org/vsyrjala/linux:
  drm/i915: Add a kludge for DSL incrementing too late and ISR not working
  drm/radeon: Move the early vblank IRQ fixup to radeon_get_crtc_scanoutpos()
  drm: Pass 'flags' from the caller to .get_scanout_position()
  drm: Fix vblank timestamping constants for interlaced modes
  drm/i915: Fix scanoutpos calculations for interlaced modes
  drm: Change {pixel,line,frame}dur_ns from s64 to int
  drm: Use crtc_clock in drm_calc_timestamping_constants()
  drm/radeon: Populate crtc_clock in radeon_atom_get_tv_timings()
  drm: Simplify the math in drm_calc_timestamping_constants()
  drm: Improve drm_calc_timestamping_constants() documentation
  drm/i915: Call drm_calc_timestamping_constants() earlier
  drm/i915: Kill hwmode save/restore
  drm: Pass the display mode to drm_calc_vbltimestamp_from_scanoutpos()
  drm: Pass the display mode to drm_calc_timestamping_constants()

10 years agoMerge branch 'topic/core-stuff' of git://people.freedesktop.org/~danvet/drm-intel...
Dave Airlie [Tue, 21 Jan 2014 23:11:39 +0000 (09:11 +1000)]
Merge branch 'topic/core-stuff' of git://people.freedesktop.org/~danvet/drm-intel into drm-next

Some straggling drm core patches

* 'topic/core-stuff' of git://people.freedesktop.org/~danvet/drm-intel:
  drm/gem: Always initialize the gem object in object_init
  drm/edid: Populate picture aspect ratio for CEA modes
  drm/edid: parse the list of additional 3D modes
  drm/edid: split VIC display mode lookup into a separate function
  drm: Make the connector mode_valid() vfunc return a drm_mode_status enum

10 years agoMerge branch 'vmwgfx-next' of git://people.freedesktop.org/~thomash/linux into drm...
Dave Airlie [Tue, 21 Jan 2014 23:11:09 +0000 (09:11 +1000)]
Merge branch 'vmwgfx-next' of git://people.freedesktop.org/~thomash/linux into drm-next

Just a single fix for sparse/smatch warnings introduced by the previous
vmwgfx-next pull.

* 'vmwgfx-next' of git://people.freedesktop.org/~thomash/linux:
  drm/vmwgfx: Fix recently introduced sparse / smatch warnings and errors

10 years agodrm/vmwgfx: Fix recently introduced sparse / smatch warnings and errors
Thomas Hellstrom [Mon, 20 Jan 2014 10:33:04 +0000 (11:33 +0100)]
drm/vmwgfx: Fix recently introduced sparse / smatch warnings and errors

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrant <jakob@vmware.com>
10 years agodrm/gem: Always initialize the gem object in object_init
Daniel Vetter [Mon, 20 Jan 2014 07:21:54 +0000 (08:21 +0100)]
drm/gem: Always initialize the gem object in object_init

At least drm/i915 expects that the obj->dev pointer is set even in
failure paths. Specifically when the shmem initialization fails we
call i915_gem_object_free which needs to deref obj->base.dev to get at
the slab pointer in the device private structure. And the shmem
allocation can easily fail when userspace is hitting open file limits.

Doing the structure init even when the shmem file allocation fails
prevents this Oops.

This is a regression from

commit 89c8233f82d9c8af5b20e72e4a185a38a7d3c50b
Author: David Herrmann <dh.herrmann@gmail.com>
Date:   Thu Jul 11 11:56:32 2013 +0200

    drm/gem: simplify object initialization

v2: Add regression note which Chris supplied.

Testcase: igt/gem_fd_exhaustion
Reported-and-Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
References: http://lists.freedesktop.org/archives/intel-gfx/2014-January/038433.html
Cc: stable@vger.kernel.org
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agoMerge branch 'drm-next-3.14' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Tue, 21 Jan 2014 00:26:50 +0000 (10:26 +1000)]
Merge branch 'drm-next-3.14' of git://people.freedesktop.org/~agd5f/linux into drm-next

New tree with the INFO ioctl merge fixed up.  This also adds a couple
of additional minor fixes.

A few more changes for 3.14, mostly just bug fixes.  Note that:
drm/radeon: add query to fetch the max engine clock.
will conflict with 3.13 final, but the fix is pretty obvious.

* 'drm-next-3.14' of git://people.freedesktop.org/~agd5f/linux: (22 commits)
  drm/radeon: add UVD support for OLAND
  drm/radeon: fix minor typos in si_dpm.c
  drm/radeon: set the full cache bit for fences on r7xx+
  drm/radeon: fix surface sync in fence on cayman (v2)
  drm/radeon/dpm: disable mclk switching on desktop RV770
  drm/radeon: fix endian handling in radeon_atom_init_mc_reg_table
  drm/radeon: write gfx pg bases even when gfx pg is disabled
  drm/radeon: bail early from enable ss in certain cases
  drm/radeon: handle ss percentage divider properly
  drm/radeon: add query to fetch the max engine clock (v2)
  drm/radeon/dp: sleep after powering up the display
  drm/radeon/dp: use usleep_range rather than udelay
  drm/radeon/dp: bump i2c-over-aux retries to 7
  drm/radeon: disable ss on DP for DCE3.x
  drm/radeon/cik: use hw defaults for TC_CFG registers
  drm/radeon: disable dpm on BTC
  drm/radeon/cik: use WAIT_REG_MEM special op for CP HDP flush
  drm/radeon/cik: use POLL_REG_MEM special op for sDMA HDP flush
  drm/radeon: consolidate sdma hdp flushing code for CIK
  drm/radeon: consolidate cp hdp flushing code for CIK
  ...

10 years agodrm/radeon: add UVD support for OLAND
Alex Deucher [Mon, 20 Jan 2014 16:25:35 +0000 (11:25 -0500)]
drm/radeon: add UVD support for OLAND

It seems this got dropped when we merged UVD support
last year.  Add this back now.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
10 years agodrm/radeon: fix minor typos in si_dpm.c
Alex Deucher [Fri, 17 Jan 2014 17:34:55 +0000 (12:34 -0500)]
drm/radeon: fix minor typos in si_dpm.c

Copy/paste typos from the ni code. Should not
have any functional change.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: set the full cache bit for fences on r7xx+
Alex Deucher [Thu, 16 Jan 2014 23:11:47 +0000 (18:11 -0500)]
drm/radeon: set the full cache bit for fences on r7xx+

Needed to properly flush the read caches for fences.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
10 years agodrm/radeon: fix surface sync in fence on cayman (v2)
Alex Deucher [Thu, 16 Jan 2014 23:02:59 +0000 (18:02 -0500)]
drm/radeon: fix surface sync in fence on cayman (v2)

We need to set the engine bit to select the ME and
also set the full cache bit.  Should help stability
on TN and cayman.

V2: fix up surface sync in ib execute as well

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
10 years agodrm/radeon/dpm: disable mclk switching on desktop RV770
Alex Deucher [Tue, 7 Jan 2014 18:51:51 +0000 (13:51 -0500)]
drm/radeon/dpm: disable mclk switching on desktop RV770

Mclk switching doesn't seem to work reliably on these
cards.  Most RV770 boards specify the same mclk for all
performance levels anyway so in most cases, this has
no affect.

Bug:
https://bugs.freedesktop.org/show_bug.cgi?id=73067

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
10 years agodrm/radeon: fix endian handling in radeon_atom_init_mc_reg_table
Alex Deucher [Thu, 16 Jan 2014 15:53:50 +0000 (10:53 -0500)]
drm/radeon: fix endian handling in radeon_atom_init_mc_reg_table

Need to swap the data for big endian.
Notcied by sylware in IRC.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: write gfx pg bases even when gfx pg is disabled
Alex Deucher [Thu, 16 Jan 2014 15:39:17 +0000 (10:39 -0500)]
drm/radeon: write gfx pg bases even when gfx pg is disabled

For consistency.  These buffers aren't used when pg is
disabled.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: bail early from enable ss in certain cases
Alex Deucher [Wed, 15 Jan 2014 18:59:47 +0000 (13:59 -0500)]
drm/radeon: bail early from enable ss in certain cases

If the ss percentage is 0 or we are using external ss,
just bail when enabling ss.  We disable it explicitly
earlier in the modeset already.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: handle ss percentage divider properly
Alex Deucher [Wed, 15 Jan 2014 18:41:31 +0000 (13:41 -0500)]
drm/radeon: handle ss percentage divider properly

It's either 100 or 1000 depending on the flags in the
table.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: add query to fetch the max engine clock (v2)
Alex Deucher [Mon, 20 Jan 2014 23:20:29 +0000 (18:20 -0500)]
drm/radeon: add query to fetch the max engine clock (v2)

This is needed for reporting the max GPU engine clock
in OpenCL.  This just reports the max possible engine
clock, it does not take into account current conditions
that may limit that clock.

v2: fix query number for merge with 3.13

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/edid: Populate picture aspect ratio for CEA modes
Vandana Kannan [Thu, 19 Dec 2013 10:04:07 +0000 (15:34 +0530)]
drm/edid: Populate picture aspect ratio for CEA modes

Adding picture aspect ratio for CEA modes based on CEA-861D Table 3 or
CEA-861E Table 4. This is useful for filling up the detail in AVI
infoframe.

v2: Ville's review comments incorporated
Added picture aspect ratio as part of edid_cea_modes instead of DRM_MODE

Signed-off-by: Vandana Kannan <vandana.kannan@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/radeon/dp: sleep after powering up the display
Alex Deucher [Tue, 14 Jan 2014 15:45:51 +0000 (10:45 -0500)]
drm/radeon/dp: sleep after powering up the display

According to the DP 1.1 spec, the sink must power
up within 1ms.  Noticed while reviewing Thierry's
drm/dp patches.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon/dp: use usleep_range rather than udelay
Alex Deucher [Tue, 14 Jan 2014 15:37:33 +0000 (10:37 -0500)]
drm/radeon/dp: use usleep_range rather than udelay

Based on common dp code proposed by Thierry Reding.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon/dp: bump i2c-over-aux retries to 7
Alex Deucher [Tue, 14 Jan 2014 15:29:59 +0000 (10:29 -0500)]
drm/radeon/dp: bump i2c-over-aux retries to 7

As per the DP1.2 spec.  Noticed while reviewing
Thierry's drm/dp patches. Also bump native aux
retries to 7 for consistency.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: disable ss on DP for DCE3.x
Alex Deucher [Mon, 13 Jan 2014 21:47:05 +0000 (16:47 -0500)]
drm/radeon: disable ss on DP for DCE3.x

Seems to cause problems with certain DP monitors.

Bug:
https://bugs.freedesktop.org/show_bug.cgi?id=40699

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
10 years agodrm/radeon/cik: use hw defaults for TC_CFG registers
Alex Deucher [Mon, 13 Jan 2014 15:18:03 +0000 (10:18 -0500)]
drm/radeon/cik: use hw defaults for TC_CFG registers

Use the hw power up values rather than 0.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: disable dpm on BTC
Alex Deucher [Sat, 11 Jan 2014 15:55:55 +0000 (10:55 -0500)]
drm/radeon: disable dpm on BTC

Still unstable on some boards.

Bugs:
https://bugs.freedesktop.org/show_bug.cgi?id=73053
https://bugzilla.kernel.org/show_bug.cgi?id=68571

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: 3.13 <stable@vger.kernel.org> # 3.13
10 years agodrm/radeon/cik: use WAIT_REG_MEM special op for CP HDP flush
Alex Deucher [Thu, 9 Jan 2014 21:51:56 +0000 (16:51 -0500)]
drm/radeon/cik: use WAIT_REG_MEM special op for CP HDP flush

This is the preferred flushing method on CIK.

Note, this only works on the PFP so the engine bit must be
set.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon/cik: use POLL_REG_MEM special op for sDMA HDP flush
Alex Deucher [Thu, 9 Jan 2014 21:35:39 +0000 (16:35 -0500)]
drm/radeon/cik: use POLL_REG_MEM special op for sDMA HDP flush

This is the preferred flushing method on CIK.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: consolidate sdma hdp flushing code for CIK
Alex Deucher [Thu, 9 Jan 2014 21:23:37 +0000 (16:23 -0500)]
drm/radeon: consolidate sdma hdp flushing code for CIK

It's used in several places so move to a common shared
function.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: consolidate cp hdp flushing code for CIK
Alex Deucher [Thu, 9 Jan 2014 21:18:11 +0000 (16:18 -0500)]
drm/radeon: consolidate cp hdp flushing code for CIK

It's used in several places so move to a common shared
function.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agodrm/radeon: don't power gate paused UVD streams
Christian König [Fri, 10 Jan 2014 15:05:05 +0000 (16:05 +0100)]
drm/radeon: don't power gate paused UVD streams

Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agoRevert "drm/radeon: disable CIK CP semaphores for now"
Alex Deucher [Thu, 9 Jan 2014 15:59:56 +0000 (10:59 -0500)]
Revert "drm/radeon: disable CIK CP semaphores for now"

This reverts commit 99b4f25122f43210278cde17a9d100906235a074.

Semaphores work fine after further review and testing.
Cc: 3.13 <stable@vger.kernel.org> # 3.13
10 years agodrm/edid: parse the list of additional 3D modes
Thomas Wood [Fri, 29 Nov 2013 18:18:58 +0000 (18:18 +0000)]
drm/edid: parse the list of additional 3D modes

Parse 2D_VIC_order_X and 3D_Structure_X from the list at the end of the
HDMI Vendor Specific Data Block.

v2: Use an offset value depending on 3D_Multi_present and add
    detail_present. (Ville Syrjälä)
v3: Make sure the list is parsed even if 3D_Structure_ALL/MASK is not
    present. (Ville Syrjälä)
    Fix one length check and remove another. (Ville Syrjälä)

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/edid: split VIC display mode lookup into a separate function
Thomas Wood [Fri, 29 Nov 2013 15:33:27 +0000 (15:33 +0000)]
drm/edid: split VIC display mode lookup into a separate function

Signed-off-by: Thomas Wood <thomas.wood@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm: Make the connector mode_valid() vfunc return a drm_mode_status enum
Damien Lespiau [Thu, 28 Nov 2013 15:29:17 +0000 (15:29 +0000)]
drm: Make the connector mode_valid() vfunc return a drm_mode_status enum

To make it clear what exactly mode_valid() should return.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
10 years agodrm/i915: Add a kludge for DSL incrementing too late and ISR not working
Ville Syrjälä [Mon, 28 Oct 2013 22:04:43 +0000 (00:04 +0200)]
drm/i915: Add a kludge for DSL incrementing too late and ISR not working

On pre-PCH platforms ISR doesn't seem to be an actual ISR, at least as
far as display interrupts are concerned. Instead it sort of looks like
some ISR bits just directly reflect the corresponding bit from PIPESTAT.
The bit appears in the ISR only if the PIPESTAT interrupt is enabled. So
in that sense it sort of looks a bit like the south interrupt scheme on
PCH platforms. So it goes something a bit like this:
PIPESTAT.status & PIPESTAT.enable -> ISR -> IMR -> IIR -> IER -> actual
interrupt

In any case that means the intel_pipe_in_vblank_locked() doesn't actually
work for pre-PCH platforms. As a last resort, add a similar kludge as radeon
has that fixes things up if we got called from the vblank interrupt,
but the scanline counter value indicates that we're not quite there yet.
We know that the scanline counter increments at hsync but is otherwise
accurate, so we can limit the kludge to the line just prior to vblank
start, instead of the relative distance that radeon uses.

Reviewed-by: mario.kleiner.de@gmail.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
10 years agodrm/radeon: Move the early vblank IRQ fixup to radeon_get_crtc_scanoutpos()
Ville Syrjälä [Mon, 28 Oct 2013 19:22:52 +0000 (21:22 +0200)]
drm/radeon: Move the early vblank IRQ fixup to radeon_get_crtc_scanoutpos()

i915 doesn't need this kludge for most platforms. Although we do
appear to need something similar on certain platforms, but we can
be more accurate when we apply the adjustment since we know exactly
why the scanline counter doesn't always quite match the vblank
status.

Also the current code doesn't handle interlaced modes correctly,
and we already deal with interlaced modes in i915 code.

So let's just move the current code to radeon_get_crtc_scanoutpos()
since that's why it was added. For i915 we'll add a more finely
targeted variant.

v2: Fix vpos vs. *vpos bug (Mario)

Reviewed-by: mario.kleiner.de@gmail.com
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>