[SQUASH ME] ASoC: SOF: Intel: hda: release display power for D0ix - #1767
[SQUASH ME] ASoC: SOF: Intel: hda: release display power for D0ix#1767kv2019i wants to merge 3 commits into
Conversation
Display driver will emit a warning if system enters S3 with wakerefs to some power domain active. Make sure SOF releases display power when it enters D0ix, and wakes it back up upon resume. Now that we have more invocations of the display power helper functions, put the the ifdefs and the check for codec_mask, into hda_codec_i915_display_power() directly to reduce clutter. Should be squashed with "ASoC: SOF: Intel: hda: fix ordering bug in resume flow" Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
|
This is in immediate follow-up to #1764 which has a regression to D0ix flow. @plbossart -- if this looks ok, I can squash with #1764 and send to alsa-devel. |
snd_hdac_ext_bus_link_get() does not work correctly in case there are multiple codecs on the bus. It unconditionally resets the bus->codec_mask value. This overwrites the initial value discovered at probe time when all codecs on the link were powered up. Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
|
@kv2019i I just realized that these won't apply on Marks tree without the D0i3 patches. Maybe we should combine them? |
@ranj063 True. I can either rebase on top of Mark's tree and send (this needs to be squashed to the first patch of #1764)), or we send these via @plbossart 's next pull and avoid the conflict with your D0i3 changes. I'm ok with both. |
| dev_dbg(bus->dev, "codec_mask = 0x%lx\n", codec_mask); | ||
| snd_hdac_chip_writew(bus, STATESTS, codec_mask); | ||
| if (!bus->codec_mask) | ||
| bus->codec_mask = codec_mask; |
There was a problem hiding this comment.
can you clarify in the commit message what the impact of overriding the bus->codec_mask is? the patch looks ok, but it's not clear why it's needed.
There was a problem hiding this comment.
@plbossart Ack. So basicly my patch to power up/down display, at runtime suspend/resume, depends on codec_mask being correct. If hdac library changes it on its own, our PM runtime flow stops working correctly. We could copy the data to a SOF local variable outside HDAC, but it seems that's not really reasonable -- there doesn't seem to be any reason for hdac_ext_controller.c to modify codec_mask like this.
When sending upstream, this should patch should be first in the series (as later patches depend on it). I can still do this :)
There was a problem hiding this comment.
ok, I'll approve then and let you send all the i915-related upstream on top of Mark's tree. Please add my reviewed-by tag.
when it's merged, we'll deal with conflicts with Ranjani's changes and send a v2.
There was a problem hiding this comment.
if there's really no reason for hdac_ext_controller to modify .codec_mask maybe that should be fixed instead?
It's probably better to have first those i915 fixes first, then re-send the v2 version of Ranjani's patches. It makes our life more complicated but others doing backports will thank us. And this is easier to apply to -stable as well. |
Ack @plbossart and @ranj063 . I'll send to the list with your signed-offs. Takashi is also making lot of HDMI changes, so he might have some further comments, let's see. |
|
Sent upstream (squashed and a few more very minor modifications): https://mailman.alsa-project.org/pipermail/alsa-devel/2020-February/162509.html |
Incremental changes to following patches sent usptream: ALSA: hda: do not override bus codec_mask in link_get() ASoC: SOF: Intel: hda: fix ordering bug in resume flow ASoC: SOF: Intel: hda: move i915 init earlier Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
|
Added one incremental patch to align with the version sent upstream. |
|
@kv2019i this can be closed, yes? |
|
Merged via upstream rebase, closing. |
4 participants
Display driver will emit a warning if system enters S3 with
wakerefs to some power domain active. Make sure SOF releases
display power when it enters D0ix, and wakes it back up upon
resume.
Now that we have more invocations of the display power helper
functions, put the the ifdefs and the check for codec_mask,
into hda_codec_i915_display_power() directly to reduce
clutter.
Should be squashed with
"ASoC: SOF: Intel: hda: fix ordering bug in resume flow"
Signed-off-by: Kai Vehmanen kai.vehmanen@linux.intel.com