Bug 105069

*** Bug 105070 has been marked as a duplicate of this bug. ***

This is the patch that cause the regression:

commit 3b5b899ca67db07a4c4825911072221f99e157e2
Author: Abhijeet Kumar <abhijeet.kumar@intel.com>
Date:   Tue Jan 23 23:00:52 2018 +0530

    ALSA: hda: Make use of core codec functions to sync power state
    
    Since sync_power_state is moved to core it's better to use the helper
    function to ensure the actual power state reaches target instead of
    using the local helper functions already exsisting in hda code.
    
    Signed-off-by: Abhijeet Kumar <abhijeet.kumar@intel.com>
    Signed-off-by: Takashi Iwai <tiwai@suse.de>

It is reverted on this patchwork:
https://patchwork.freedesktop.org/series/38097/

and the results are back to normal.

Reference: https://patchwork.freedesktop.org/series/38140/

Hello, this is a "me too" from QA. Same results on both platforms:

igt@pm_rpm@basic-pci-d3-state:
Stdout	
IGT-Version: 1.21-g37bd27f (x86_64) (Linux: 4.16.0-rc1-drm-intel-qa-ww7-commit-ec41124+ x86_64)
Runtime PM support: 1
PC8 residency support: 1
Stack trace:
  #0 [__igt_fail_assert+0x101]
  #1 [main+0xfa8]
  #2 [__libc_start_main+0xf1]
  #3 [_start+0x2a]
  #4 [<unknown>+0x2a]
Subtest basic-pci-d3-state: FAIL (10.206s)
Stderr	
(pm_rpm:18811) CRITICAL: Test assertion failure function pci_d3_state_subtest, file pm_rpm.c:1336:
(pm_rpm:18811) CRITICAL: Failed assertion: wait_for_suspended()
Subtest basic-pci-d3-state failed.
**** DEBUG ****
(pm_rpm:18811) DEBUG: Test requirement passed: has_runtime_pm
(pm_rpm:18811) CRITICAL: Test assertion failure function pci_d3_state_subtest, file pm_rpm.c:1336:
(pm_rpm:18811) CRITICAL: Failed assertion: wait_for_suspended()
(pm_rpm:18811) igt-core-INFO: Stack trace:
(pm_rpm:18811) igt-core-INFO:   #0 [__igt_fail_assert+0x101]
(pm_rpm:18811) igt-core-INFO:   #1 [main+0xfa8]
(pm_rpm:18811) igt-core-INFO:   #2 [__libc_start_main+0xf1]
(pm_rpm:18811) igt-core-INFO:   #3 [_start+0x2a]
(pm_rpm:18811) igt-core-INFO:   #4 [<unknown>+0x2a]
****  END  ****

igt@pm_rpm@basic-rte:
Stdout	
IGT-Version: 1.21-g37bd27f (x86_64) (Linux: 4.16.0-rc1-drm-intel-qa-ww7-commit-ec41124+ x86_64)
Runtime PM support: 1
PC8 residency support: 1
Stack trace:
  #0 [__igt_fail_assert+0x101]
  #1 [main+0xd35]
  #2 [__libc_start_main+0xf1]
  #3 [_start+0x2a]
  #4 [<unknown>+0x2a]
Subtest basic-rte: FAIL (10.048s)
Stderr	
(pm_rpm:19386) CRITICAL: Test assertion failure function basic_subtest, file pm_rpm.c:720:
(pm_rpm:19386) CRITICAL: Failed assertion: wait_for_suspended()
Subtest basic-rte failed.
**** DEBUG ****
(pm_rpm:19386) CRITICAL: Test assertion failure function basic_subtest, file pm_rpm.c:720:
(pm_rpm:19386) CRITICAL: Failed assertion: wait_for_suspended()
(pm_rpm:19386) igt-core-INFO: Stack trace:
(pm_rpm:19386) igt-core-INFO:   #0 [__igt_fail_assert+0x101]
(pm_rpm:19386) igt-core-INFO:   #1 [main+0xd35]
(pm_rpm:19386) igt-core-INFO:   #2 [__libc_start_main+0xf1]
(pm_rpm:19386) igt-core-INFO:   #3 [_start+0x2a]
(pm_rpm:19386) igt-core-INFO:   #4 [<unknown>+0x2a]
****  END  ****

Created attachment 137340 [details]
dmesg_log_HSW_pm_rpm

Created attachment 137341 [details]
dmesg_log_BDW_pm_rpm

commit 7a3da83bff6a8ab9635d3b8c958756265d80df7c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Feb 14 11:34:21 2018 +0000

    Revert "ALSA: hda: Make use of core codec functions to sync power state"
    
    This reverts commit 3b5b899ca67db07a4c4825911072221f99e157e2.
    
    References: https://bugs.freedesktop.org/show_bug.cgi?id=105069

OK we can archive from cibuglog, but if audio is not going to fix this we will end up in same mess sooner or later.

I will assign the bug to Jani Saarinen so that he can follow up with the audio people. It is then up to him to decide if audio wants to use this FDO bug to fix the issue or whatever.

Hi @Elizabeth,

In both attached logs I can't find the IGT exiting statement for subtest basic-rte. Is it something expected ?

(In reply to Abhijeet Kumar from comment #9)
> Hi @Elizabeth,
> 
> In both attached logs I can't find the IGT exiting statement for subtest
> basic-rte. Is it something expected ?
Not sure if it was because I ran them manually like this to get the log:

$ sudo -E ./pm_rpm --r basic*

Guess I can try one at time and check if the exiting appears twice as in CI logs.

(In reply to Elizabeth from comment #10)
> (In reply to Abhijeet Kumar from comment #9)
> > Hi @Elizabeth,
> > 
> > In both attached logs I can't find the IGT exiting statement for subtest
> > basic-rte. Is it something expected ?
> Not sure if it was because I ran them manually like this to get the log:
> 
> $ sudo -E ./pm_rpm --r basic*
> 
> Guess I can try one at time and check if the exiting appears twice as in CI
> logs.
Yep, that was the case. Running one at time, I get it twice:
One by one:
[11310.082596] [IGT] pm_rpm: executing
[11311.181291] [IGT] pm_rpm: starting subtest basic-pci-d3-state
[11311.496757] [IGT] pm_rpm: exiting, ret=0
[11327.443891] [IGT] pm_rpm: executing
[11328.535669] [IGT] pm_rpm: starting subtest basic-rte
[11328.845933] [IGT] pm_rpm: exiting, ret=0
And with basic*
[11353.213800] [IGT] pm_rpm: executing
[11354.309595] [IGT] pm_rpm: starting subtest basic-rte
[11354.551537] [IGT] pm_rpm: starting subtest basic-pci-d3-state
[11354.909636] [IGT] pm_rpm: exiting, ret=0

[11328.845933] [IGT] pm_rpm: exiting, ret=0

I see now it is exiting without any failure. Is it with 3b5b899ca67db07a4c4825911072221f99e157e2 revert ? or without that fix ?

(In reply to Abhijeet Kumar from comment #12)
> [11328.845933] [IGT] pm_rpm: exiting, ret=0
> 
> I see now it is exiting without any failure. Is it with
> 3b5b899ca67db07a4c4825911072221f99e157e2 revert ? or without that fix ?

It looks like it's on updated drm-tip. Which has the reversion.

Hey you mind sharing the passing logs for anyone device "where you ran subtest one by one" ? It will helpful for us to understand the difference in code path for the two. And please enable dynamic debug logs for snd and had codec as well in both passing and failing.

(In reply to Abhijeet Kumar from comment #13)
> (In reply to Abhijeet Kumar from comment #12)
> > [11328.845933] [IGT] pm_rpm: exiting, ret=0
> > 
> > I see now it is exiting without any failure. Is it with
> > 3b5b899ca67db07a4c4825911072221f99e157e2 revert ? or without that fix ?
> 
> It looks like it's on updated drm-tip. Which has the reversion.
> 
> Hey you mind sharing the passing logs for anyone device "where you ran
> subtest one by one" ? It will helpful for us to understand the difference in
> code path for the two. And please enable dynamic debug logs for snd and had
> codec as well in both passing and failing.

Yes, Chris reverted the patch.

Abhijeet, if you want to experiment you can always use the i915 try-bot. I.e. revert the revert and send to try-bot list together with whatever you want to test. 
If you haven't already subscribed you can do that here: https://lists.freedesktop.org/mailman/listinfo/intel-gfx-trybot

Closing, please re-open if still occurs.

commit 009f8c90f571d87855914dbc20e6c0ea2a3b19ae
Author: Lukas Wunner <lukas@wunner.de>
Date:   Thu May 24 19:01:07 2018 +0200

    ALSA: hda - Fix runtime PM