On Wed, 27 Jun 2018 17:54:31 +0200, Chris Wilson wrote:
Quoting Takashi Iwai (2018-06-27 10:10:32)
Although snd_hda_power_up() and snd_hda_power_up_pm() may fail, we haven't dealt with the error properly in many places. It's an unusual situation but still possible.
This patch spots these places and adds the proper error paths.
Signed-off-by: Takashi Iwai tiwai@suse.de
Verdict from CI,
https://intel-gfx-ci.01.org/tree/drm-tip/Trybot_2430/issues.html
is that this one causes a bunch of pm fallout.
Do you mind doing a quick revert? Or working with our CI to find the bad chunk?
Hrm, it doesn't look good -- I revert the branch merge now. Thanks for the quick heads up.
Takashi
On a second look,
@@ -7310,7 +7339,8 @@ static void ca0132_free(struct hda_codec *codec) struct ca0132_spec *spec = codec->spec;
cancel_delayed_work_sync(&spec->unsol_hp_work);
snd_hda_power_up(codec);
if (snd_hda_power_up(codec) < 0)goto skip_shutdown; switch (spec->quirk) { case QUIRK_SBZ: sbz_exit_chip(codec);@@ -7326,6 +7356,7 @@ static void ca0132_free(struct hda_codec *codec) break; } snd_hda_power_down(codec);
- skip_shutdown: if (spec->mem_base) iounmap(spec->mem_base); kfree(spec->spec_init_verbs);
would seem to be the only chunk more complicated than the rest. -Chris