Bug 216363

Summary:	sometimes a short period of noise or silence at beginning of sound
Product:	Drivers	Reporter:	qbl
Component:	Sound(ALSA)	Assignee:	Jaroslav Kysela (perex)
Status:	RESOLVED CODE_FIX
Severity:	normal	CC:	pierre.juhen, tiwai
Priority:	P1
Hardware:	AMD
OS:	Linux
Kernel Version:	5.18.8 5.18.9 ... 5.19.1	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	Test fix patch Test fix patch

Description qbl 2022-08-14 20:21:30 UTC

sometimes a short period of noise or silence at beginning of sound or after pausing

Hardware:
00:01.1 Audio device: Advanced Micro Devices, Inc. [AMD/ATI] Kaveri HDMI/DP Audio Controller
00:01.1 0403: 1002:1308
Output: HDMI
Same on Radeon RX560/DP
Sometimes there is a short period of silence or random noise when sound statrts to play ( about a second). This somtimes happens too, after playback paused.
Affected Kernels 5.18.8, 5.18.9, ..., 5.19.1

bisecting showed up d35135ebebe4090f3e0a5fc7f50bd6d8848fda3f as the likely cause.
Since this kind of Bug is not easy to track down, this might be a false positive.
But i've applied a reverse patch to some of the newer Kernels - including 5.19.1 - and it looks like this is the bad boy. Maybe its a solution to implement the trouble causing patch in a per machine fashion, for example via cpuid, so the weakness of Intels Haswell could be worked around.

Comment 1 Takashi Iwai 2022-08-15 07:16:22 UTC

Hmm, it's not really CPU-dependent but rather IOMMU-dependent.

How about the patch below instead?

Comment 2 Takashi Iwai 2022-08-15 07:16:46 UTC

Created attachment 301568 [details]
Test fix patch

Comment 3 Takashi Iwai 2022-08-15 07:45:48 UTC

Scratch the above, it will break things obviously...

Comment 4 Takashi Iwai 2022-08-15 13:12:27 UTC

I guess we have to put back x86-specific workaround again, but in a different way.  Could you try the patch below instead of the previous one?

Comment 5 Takashi Iwai 2022-08-15 13:13:12 UTC

Created attachment 301569 [details]
Test fix patch

Comment 6 qbl 2022-08-15 17:15:59 UTC

I've applied this patch to a vanilla 19.1, which is your intention, i think.
Now we have to wait.

Comment 7 qbl 2022-08-20 14:23:33 UTC

It looks like this patch works.

Comment 8 Pierre Juhen 2022-08-22 07:20:54 UTC

Hi,

also an issue with iommu with chipset AMD 970;

My computer freezes with access to the disk to a disk each second.

Have to stop the PC for some minutes before being able to restart.

Sometimes, when I reboot, BIOS says the main BIOS is corrupted...

iommu=soft in the kernel parameters seems to solve the issue..

Regards,

Comment 9 Takashi Iwai 2022-08-23 08:13:53 UTC

Now the fix is merged to sound git tree (commit a8d302a0b77057568350fe0123e639d02dba0745), possibly included in the upstream 6.0-rc3 or rc4.

Comment 10 Pierre Juhen 2022-08-24 07:16:30 UTC

Shame on me ! 

The troubles I had where not linked to the kernel, but to a dying CMOS battery on the motherboard.

Sorry,

Regards,


(In reply to Pierre Juhen from comment #8)
> Hi,
> 
> also an issue with iommu with chipset AMD 970;
> 
> My computer freezes with access to the disk to a disk each second.
> 
> Have to stop the PC for some minutes before being able to restart.
> 
> Sometimes, when I reboot, BIOS says the main BIOS is corrupted...
> 
> iommu=soft in the kernel parameters seems to solve the issue..
> 
> Regards,

Comment 11 Takashi Iwai 2022-09-04 09:07:01 UTC

It turned out that the fix caused regressions on another machines with AMD GPU.

I posted two possible fix patches below.
Could you check each of them and verify whether it still works with your device?
  https://lore.kernel.org/r/87ilm3vbzq.wl-tiwai@suse.de
  https://lore.kernel.org/r/875yi3froa.wl-tiwai@suse.de

Note that the first one might bring back to the origianl problem.

Comment 12 Takashi Iwai 2022-09-05 05:38:23 UTC

There is a third one, and that should be tested at first:
  https://lore.kernel.org/r/874jxml7a4.wl-tiwai@suse.de

It'd be appreciated if you can confirm that it doesn't bring further regression.

Comment 13 qbl 2022-09-05 11:54:51 UTC

I am going to test the third patch now.

Comment 14 qbl 2022-10-01 17:59:04 UTC

There hasn't been any trouble. So it's time to classify this bug as fixed.
Thanks for your effort, greetings to all maintainers and supporters and keep up the good work.