next/master bisection: baseline.dmesg.alert on meson-sm1-sei610

Jerome Brunet jbrunet at baylibre.com
Fri Apr 17 17:48:08 CEST 2020


On Fri 17 Apr 2020 at 14:27, Mark Brown <broonie at kernel.org> wrote:

> On Fri, Apr 17, 2020 at 02:53:36AM -0700, kernelci.org bot wrote:
>
>> next/master bisection: baseline.dmesg.alert on meson-sm1-sei610
>
>> Summary:
>>   Start:      a3ca59b9af21e Add linux-next specific files for 20200416
>>   Plain log:  https://storage.kernelci.org//next/master/next-20200416/arm64/defconfig+CONFIG_RANDOMIZE_BASE=y/gcc-8/lab-baylibre/baseline-meson-sm1-sei610.txt
>>   HTML log:   https://storage.kernelci.org//next/master/next-20200416/arm64/defconfig+CONFIG_RANDOMIZE_BASE=y/gcc-8/lab-baylibre/baseline-meson-sm1-sei610.html
>>   Result:     9b5db059366ae ASoC: soc-pcm: dpcm: Only allow playback/capture if supported
>
> This change to check the playback and capture constraints is also
> causing issues on at least Meson - I'm a bit worried that this is also
> causing oopses here, not just audio problems so copying in Morimoto-san
> too.  We should fix the things that are broken by this change since it's
> a thing we should be able to rely on and do enforce for non-DPCm links
> but I'm wondering if we should revert for 5.7 and reapply for 5.8.
>
> Including complete report so people have it.

The Oops on this device is due to the CPU side of a codec-to-codec
link. Since these are no supposed to be exposed to the userspace,
I have marked them with ".no_pcm" so the dai_link is not going through
the appropriate code path. It was working so far, so it went unnoticed.

I'll send a fix for this.

On a more general note, the dpcm_playback and dcpm_capture is only there
to force a particular direction off, since it a bit difficult for ASoC
to figure out when the link are mostly made of dummy codecs.

This seems like a duplicate of the properties "playback_only" and
"capture_only"

Maybe we could:
1) Merge the 2 kind of properties. The "_only" form might be a bit
   annoying to work with in dpcm.
   What about a boolean table "force_disable" with the stream direction
   as indices ?
2) Get rid of the if clause below and handle DPCM and non-DPCM with a
   single code path.

Thoughts ?

>
>> Checks:
>>   revert:     PASS
>>   verify:     PASS
>
>> Parameters:
>>   Tree:       next
>>   URL:        https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
>>   Branch:     master
>>   Target:     meson-sm1-sei610
>>   CPU arch:   arm64
>>   Lab:        lab-baylibre
>>   Compiler:   gcc-8
>>   Config:     defconfig+CONFIG_RANDOMIZE_BASE=y
>>   Test case:  baseline.dmesg.alert
>> 
>> Breaking commit found:
>> 
>> -------------------------------------------------------------------------------
>> commit 9b5db059366ae2087e07892b5fc108f81f4ec189
>> Author: Stephan Gerhold <stephan at gerhold.net>
>> Date:   Wed Apr 15 12:49:28 2020 +0200
>> 
>>     ASoC: soc-pcm: dpcm: Only allow playback/capture if supported
>>     
>>     At the moment, PCM devices for DPCM are only created based on the
>>     dpcm_playback/capture parameters of the DAI link, without considering
>>     if the CPU/FE DAI is actually capable of playback/capture.
>>     
>>     Normally the dpcm_playback/capture parameter should match the
>>     capabilities of the CPU DAI. However, there is no way to set that
>>     parameter from the device tree (e.g. with simple-audio-card or
>>     qcom sound cards). dpcm_playback/capture are always both set to 1.
>>     
>>     This causes problems when the CPU DAI does only support playback
>>     or capture. Attemting to open that PCM device with an unsupported
>>     stream type then results in a null pointer dereference:
>>     
>>         Unable to handle kernel NULL pointer dereference at virtual address 0000000000000128
>>         Internal error: Oops: 96000044 [#1] PREEMPT SMP
>>         CPU: 3 PID: 1582 Comm: arecord Not tainted 5.7.0-rc1
>>         pc : invalidate_paths_ep+0x30/0xe0
>>         lr : snd_soc_dapm_dai_get_connected_widgets+0x170/0x1a8
>>         Call trace:
>>          invalidate_paths_ep+0x30/0xe0
>>          snd_soc_dapm_dai_get_connected_widgets+0x170/0x1a8
>>          dpcm_path_get+0x38/0xd0
>>          dpcm_fe_dai_open+0x70/0x920
>>          snd_pcm_open_substream+0x564/0x840
>>          snd_pcm_open+0xfc/0x228
>>          snd_pcm_capture_open+0x4c/0x78
>>          snd_open+0xac/0x1a8
>>          ...
>>     
>>     ... because the DAI playback/capture_widget is not set in that case.
>>     
>>     We could add checks there to fix the problem (maybe we should
>>     anyway), but much easier is to not expose the device as
>>     playback/capture in the first place. Attemting to use that
>>     device would always fail later anyway.
>>     
>>     Add checks for snd_soc_dai_stream_valid() to the DPCM case
>>     to avoid exposing playback/capture if it is not supported.
>>     
>>     Signed-off-by: Stephan Gerhold <stephan at gerhold.net>
>>     Link: https://lore.kernel.org/r/20200415104928.86091-1-stephan@gerhold.net
>>     Signed-off-by: Mark Brown <broonie at kernel.org>
>> 
>> diff --git a/sound/soc/soc-pcm.c b/sound/soc/soc-pcm.c
>> index 289aebc155293..1f302de440525 100644
>> --- a/sound/soc/soc-pcm.c
>> +++ b/sound/soc/soc-pcm.c
>> @@ -2911,8 +2911,17 @@ int soc_new_pcm(struct snd_soc_pcm_runtime *rtd, int num)
>>  	int i;
>>  
>>  	if (rtd->dai_link->dynamic || rtd->dai_link->no_pcm) {
>> -		playback = rtd->dai_link->dpcm_playback;
>> -		capture = rtd->dai_link->dpcm_capture;
>> +		cpu_dai = asoc_rtd_to_cpu(rtd, 0);
>> +		if (rtd->num_cpus > 1) {
>> +			dev_err(rtd->dev,
>> +				"DPCM doesn't support Multi CPU yet\n");
>> +			return -EINVAL;
>> +		}
>> +
>> +		playback = rtd->dai_link->dpcm_playback &&
>> +			   snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_PLAYBACK);
>> +		capture = rtd->dai_link->dpcm_capture &&
>> +			  snd_soc_dai_stream_valid(cpu_dai, SNDRV_PCM_STREAM_CAPTURE);
>>  	} else {
>>  		/* Adapt stream for codec2codec links */
>>  		int cpu_capture = rtd->dai_link->params ?
>> -------------------------------------------------------------------------------
>> 
>> 
>> Git bisection log:
>> 
>> -------------------------------------------------------------------------------
>> git bisect start
>> # good: [87b0f983f66f23762921129fd35966eddc3f2dae] net: mscc: ocelot: fix untagged packet drops when enslaving to vlan aware bridge
>> git bisect good 87b0f983f66f23762921129fd35966eddc3f2dae
>> # bad: [a3ca59b9af21e68069555ffff1ad89bd2a7c40fc] Add linux-next specific files for 20200416
>> git bisect bad a3ca59b9af21e68069555ffff1ad89bd2a7c40fc
>> # bad: [feb09551bfe34ccf0ba462188a1aee651be0f2c3] Merge remote-tracking branch 'i2c/i2c/for-next'
>> git bisect bad feb09551bfe34ccf0ba462188a1aee651be0f2c3
>> # good: [a4721ced760684d1776bf31f7925aa41bb3f4846] Merge v5.7-rc1 into drm-misc-fixes
>> git bisect good a4721ced760684d1776bf31f7925aa41bb3f4846
>> # bad: [e75043c693af6a10c9e2087adeef243cf05ce3bd] Merge remote-tracking branch 'rockchip/for-next'
>> git bisect bad e75043c693af6a10c9e2087adeef243cf05ce3bd
>> # bad: [08a315860a1b462fef5969d2e6ab4cdab26df8e3] Merge remote-tracking branch 'arm-soc/for-next'
>> git bisect bad 08a315860a1b462fef5969d2e6ab4cdab26df8e3
>> # bad: [80dbfa57508c36c2b3798b387ef3eb92954b9c09] Merge remote-tracking branch 'slave-dma-fixes/fixes'
>> git bisect bad 80dbfa57508c36c2b3798b387ef3eb92954b9c09
>> # good: [a723380c6f0a3723db72da6c6097b32b23f5564c] Merge remote-tracking branch 'sound-current/for-linus'
>> git bisect good a723380c6f0a3723db72da6c6097b32b23f5564c
>> # bad: [5bd70440cb0a6f5c6a84019bb2aa93ab8310a5cd] ASoC: soc-dai: revert all changes to DAI startup/shutdown sequence
>> git bisect bad 5bd70440cb0a6f5c6a84019bb2aa93ab8310a5cd
>> # good: [4d1a015a203c0249e3332ea217a38ec978118daa] ASoC: convert rockchip spdif bindings to yaml
>> git bisect good 4d1a015a203c0249e3332ea217a38ec978118daa
>> # good: [0f2a3b02274c02eb97697c4d89c019d1d21ac225] ASoC: wsa881x: mark read_only_wordlength flag
>> git bisect good 0f2a3b02274c02eb97697c4d89c019d1d21ac225
>> # good: [aa7812737f2877e192d57626cbe8825cc7cf6de9] ASoC: sgtl5000: Fix VAG power-on handling
>> git bisect good aa7812737f2877e192d57626cbe8825cc7cf6de9
>> # bad: [0c824ec094b5cda766c80d88c2036e28c24a4cb1] ASoC: q6dsp6: q6afe-dai: add missing channels to MI2S DAIs
>> git bisect bad 0c824ec094b5cda766c80d88c2036e28c24a4cb1
>> # bad: [9b5db059366ae2087e07892b5fc108f81f4ec189] ASoC: soc-pcm: dpcm: Only allow playback/capture if supported
>> git bisect bad 9b5db059366ae2087e07892b5fc108f81f4ec189
>> # first bad commit: [9b5db059366ae2087e07892b5fc108f81f4ec189] ASoC: soc-pcm: dpcm: Only allow playback/capture if supported
>> -------------------------------------------------------------------------------



More information about the Alsa-devel mailing list