[Intel-gfx] [PATCH 0/4 V7] Power-well API implementation for Haswell

[Paulo Zanoni]

2013/7/4 Wang xingchao <xingchao.wang at linux.intel.com>:
> On Thu, Jul 04, 2013 at 10:24:15AM -0300, Paulo Zanoni wrote:
>> 2013/7/4 Wang xingchao <xingchao.wang at linux.intel.com>:
>> > On Wed, Jul 03, 2013 at 05:00:51PM -0300, Paulo Zanoni wrote:
>> >> 2013/6/6 Daniel Vetter <daniel at ffwll.ch>:
>> >> > On Thu, May 30, 2013 at 10:07:07PM +0800, Wang Xingchao wrote:
>> >> >> Hi all,
>> >> >>
>> >> >>    This is V7 and here're some changes notes:
>> >> >>    change from V6-->V7:
>> >> >>    - rename variable
>> >> >>    - use HAS_POWER_WELL instead of IS_HASWELL
>> >> >>    - put structure inside drm_i915_private
>> >> >>    - use WARN_ON for global pointer check
>> >> >>
>> >> >>    change from V5-->V6:
>> >> >>    - Remove duplication code in new introduced probe work
>> >> >>    - move duplication code in azx_probe_continue
>> >> >>    - remove unused #ifdef
>> >> >>    - replace request_module with symbol_request
>> >> >>    - replace spin_lock_irq with spin_lock_irqsave in gfx side
>> >> >>    - other typo fixes
>> >> >>    (review by Takashi)
>> >> >>
>> >> >>    change from V4-->V5:
>> >> >>    - fix reference count bug
>> >> >>    - new patch on general runtime pm support for audio pci device
>> >> >>    - new patch to avoid request_module() deadlock
>> >> >>
>> >> >>    change between V3-->V4:
>> >> >>    - add new structure i915_power_well
>> >> >>    - initialize drm_device pointer at module init time
>> >> >>    - change function name
>> >> >>
>> >> >>    change between V2-->V3:
>> >> >>    - make SND_HDA_I915 selectable
>> >> >>    - use snd_printdd to output message
>> >> >>    - add return error code check
>> >> >>    - use symbol_request to replace symbol_get
>> >> >>    - release power_well at azx_free
>> >> >>    - some typo fixes
>> >> >>
>> >> >>    changes between V1-->V2:
>> >> >>    - use reference count to track power-well usage
>> >> >>    - remove external module, compiled into snd-hda-intel instead
>> >> >>    - manage symbols and module loading properly
>> >> >>    - remove IS_HSW macro, use flag instead
>> >> >>    - remove audio callback for gfx driver to avoid dependency
>> >> >>    - split whole patch into two pieces for easy review
>> >> >>    - more typo fixes
>> >> >>
>> >> >>
>> >> >> Takashi Iwai (1):
>> >> >>   ALSA: hda - Move azx_first_init() into azx_probe_continue()
>> >> >>
>> >> >> Wang Xingchao (3):
>> >> >>   ALSA: hda - Fix runtime PM check
>> >> >>   ALSA: hda - Add power-welll support for haswell HDA
>> >> >>   i915/drm: Add private api for power well usage
>> >> >
>> >> > After discussion with Dave and Takashi I've merged the entire series to
>> >> > drm-intel-next. I'll show up in the next linux-next and I'll forward it to
>> >> > Dave for mergin into drm-next in roughly 2 weeks.
>> >>
>> >> So today I unblacklisted the audio modules on one of my Haswell
>> >> machines and booted it with i915.disable_power_well=1. I only have an
>> >> eDP output (it doesn't have audio) and I see the power well is
>> >> enabled. This is wrong, the power well should be disabled since we
>> >
>> > right, if no application using audio it should be in runtime suspend mode.
>> > And maybe your system didnot enable runtime suspend by default, would you
>> > tell me the output below?
>> > cat /sys/devices/pci0000:00/0000:00:03.0/power/control
>>
>> It says "on".
>
> would you change it to "auto" and test again.
> Runtime power save should be enabled with "auto".

Doesn't solve the problem. Should I open a bug report somewhere?
Having the power well enabled prevents some power saving features from
the graphics driver.


>
> --xingchao
>>
>>
>> >
>> > thanks
>> > --xingchao
>> >> only have an eDP panel, and we don't support audio on eDP. I checked
>> >> on dmesg and the audio driver requests the power well but never
>> >> releases it.
>> >>
>> >> So I decided to do the same test on another Haswell machine, and on
>> >> that specific machine the audio driver gets the power well and then
>> >> releases it at azx_runtime_suspend. This machine is also eDP-only
>> >>
>> >> I was expecting that on both cases the audio driver would release the
>> >> power well as soon as it sees there's no connected output capable of
>> >> HD audio.
>> >>
>> >> How can I help debugging this?
>> >>
>> >> Thanks,
>> >> Paulo
>> >>
>> >> >
>> >> > Thanks, Daniel
>> >> >
>> >> >>
>> >> >>  drivers/gpu/drm/i915/i915_dma.c  |    6 +++
>> >> >>  drivers/gpu/drm/i915/i915_drv.h  |   12 ++++++
>> >> >>  drivers/gpu/drm/i915/intel_drv.h |    4 ++
>> >> >>  drivers/gpu/drm/i915/intel_pm.c  |   81 ++++++++++++++++++++++++++++++++---
>> >> >>  include/drm/i915_powerwell.h     |   36 ++++++++++++++++
>> >> >>  sound/pci/hda/Kconfig            |   10 +++++
>> >> >>  sound/pci/hda/Makefile           |    2 +
>> >> >>  sound/pci/hda/hda_i915.c         |   75 ++++++++++++++++++++++++++++++++
>> >> >>  sound/pci/hda/hda_i915.h         |   35 +++++++++++++++
>> >> >>  sound/pci/hda/hda_intel.c        |   87 ++++++++++++++++++++++++++++++--------
>> >> >>  10 files changed, 324 insertions(+), 24 deletions(-)
>> >> >>  create mode 100644 include/drm/i915_powerwell.h
>> >> >>  create mode 100644 sound/pci/hda/hda_i915.c
>> >> >>  create mode 100644 sound/pci/hda/hda_i915.h
>> >> >>
>> >> >> --
>> >> >> 1.7.9.5
>> >> >>
>> >> >
>> >> > --
>> >> > Daniel Vetter
>> >> > Software Engineer, Intel Corporation
>> >> > +41 (0) 79 365 57 48 - http://blog.ffwll.ch
>> >> > _______________________________________________
>> >> > Intel-gfx mailing list
>> >> > Intel-gfx at lists.freedesktop.org
>> >> > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
>> >>
>> >>
>> >>
>> >> --
>> >> Paulo Zanoni
>>
>>
>>
>> --
>> Paulo Zanoni



-- 
Paulo Zanoni