[PATCH] ASoC: soc-core: remove error due to probe deferral

[PATCH] ASoC: soc-core: remove error due to probe deferral

[Stefan Agner]

From: Stefan Agner <stefan.agner@toradex.com>

Deferred probes shouldn't cause error messages in the boot log. Avoid
printing with dev_err() in case EPROBE_DEFER is the return value.

Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
---
 sound/soc/soc-core.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c
index fd6eaae6c0ed..98e1e80b5493 100644
--- a/sound/soc/soc-core.c
+++ b/sound/soc/soc-core.c
@@ -1985,9 +1985,11 @@ static int snd_soc_instantiate_card(struct snd_soc_card *card)
 	mutex_lock(&client_mutex);
 	for_each_card_prelinks(card, i, dai_link) {
 		ret = soc_init_dai_link(card, dai_link);
-		if (ret) {
+		if (ret && ret != -EPROBE_DEFER) {
 			dev_err(card->dev, "ASoC: failed to init link %s: %d\n",
 				dai_link->name, ret);
+		}
+		if (ret) {
 			mutex_unlock(&client_mutex);
 			return ret;
 		}
-- 
2.22.0

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 441 bytes --]

On Thu, Aug 08, 2019 at 02:36:55PM +0200, Stefan Agner wrote:
> From: Stefan Agner <stefan.agner@toradex.com>
> 
> Deferred probes shouldn't cause error messages in the boot log. Avoid
> printing with dev_err() in case EPROBE_DEFER is the return value.

No, they absolutely should tell the user why they are deferring so the
user has some information to go on when they're trying to figure out why
their device isn't instantiating.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Stefan Agner]

On 2019-08-08 14:44, Mark Brown wrote:
> On Thu, Aug 08, 2019 at 02:36:55PM +0200, Stefan Agner wrote:
>> From: Stefan Agner <stefan.agner@toradex.com>
>>
>> Deferred probes shouldn't cause error messages in the boot log. Avoid
>> printing with dev_err() in case EPROBE_DEFER is the return value.
> 
> No, they absolutely should tell the user why they are deferring so the
> user has some information to go on when they're trying to figure out why
> their device isn't instantiating.

Hm, I see, if the driver defers and does not manage in the end, then the
messages are indeed helpful.

But can we lower severity, e.g. to dev_info? In my case it succeeds in
the end, just defers about 6 times. I have 3 links which then leads to
18 error messages which confuse users... From what I can see
soc_init_dai_link() would print dev_err in case there is an actual
error.

--
Stefan

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Takashi Iwai]

On Thu, 08 Aug 2019 14:44:37 +0200,
Mark Brown wrote:
> 
> On Thu, Aug 08, 2019 at 02:36:55PM +0200, Stefan Agner wrote:
> > From: Stefan Agner <stefan.agner@toradex.com>
> > 
> > Deferred probes shouldn't cause error messages in the boot log. Avoid
> > printing with dev_err() in case EPROBE_DEFER is the return value.
> 
> No, they absolutely should tell the user why they are deferring so the
> user has some information to go on when they're trying to figure out why
> their device isn't instantiating.

But it's no real error that *must* be printed on the console, either.
Maybe downgrading the printk level?


thanks,

Takashi

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 443 bytes --]

On Thu, Aug 08, 2019 at 03:00:06PM +0200, Takashi Iwai wrote:
> Mark Brown wrote:

> > No, they absolutely should tell the user why they are deferring so the
> > user has some information to go on when they're trying to figure out why
> > their device isn't instantiating.

> But it's no real error that *must* be printed on the console, either.
> Maybe downgrading the printk level?

Yes, downgrading can be OK though it does bloat the code.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Takashi Iwai]

On Thu, 08 Aug 2019 15:02:17 +0200,
Mark Brown wrote:
> 
> On Thu, Aug 08, 2019 at 03:00:06PM +0200, Takashi Iwai wrote:
> > Mark Brown wrote:
> 
> > > No, they absolutely should tell the user why they are deferring so the
> > > user has some information to go on when they're trying to figure out why
> > > their device isn't instantiating.
> 
> > But it's no real error that *must* be printed on the console, either.
> > Maybe downgrading the printk level?
> 
> Yes, downgrading can be OK though it does bloat the code.

I guess we can use dev_printk() with the conditional level choice.


Takashi

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Stefan Agner]

On 2019-08-08 15:14, Takashi Iwai wrote:
> On Thu, 08 Aug 2019 15:02:17 +0200,
> Mark Brown wrote:
>>
>> On Thu, Aug 08, 2019 at 03:00:06PM +0200, Takashi Iwai wrote:
>> > Mark Brown wrote:
>>
>> > > No, they absolutely should tell the user why they are deferring so the
>> > > user has some information to go on when they're trying to figure out why
>> > > their device isn't instantiating.
>>
>> > But it's no real error that *must* be printed on the console, either.
>> > Maybe downgrading the printk level?
>>
>> Yes, downgrading can be OK though it does bloat the code.
> 
> I guess we can use dev_printk() with the conditional level choice.
> 

How about use dev_info always? We get a dev_err message from
soc_init_dai_link in error cases...

		ret = soc_init_dai_link(card, dai_link);
		if (ret && ret != -EPROBE_DEFER) {
			dev_info(card->dev, "ASoC: failed to init link %s: %d\n",
				 dai_link->name, ret);
		}
		if (ret) {
			soc_cleanup_platform(card);
			mutex_unlock(&client_mutex);
			return ret;
		}

--
Stefan

Re: [PATCH] ASoC: soc-core: remove error due to probe deferral

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 1013 bytes --]

On Thu, Aug 08, 2019 at 03:16:53PM +0200, Stefan Agner wrote:
> On 2019-08-08 15:14, Takashi Iwai wrote:
> > Mark Brown wrote:

> > I guess we can use dev_printk() with the conditional level choice.

> How about use dev_info always? We get a dev_err message from
> soc_init_dai_link in error cases...

> 		ret = soc_init_dai_link(card, dai_link);
> 		if (ret && ret != -EPROBE_DEFER) {
> 			dev_info(card->dev, "ASoC: failed to init link %s: %d\n",
> 				 dai_link->name, ret);
> 		}

Well, if there's adequate error reporting in init_dai_link() it's a bit
different - we can just remove the print entirely regardless of what the
return code is.  The point is to ensure that we don't just silently
fail.  Unfortunately there's no prints in the probe deferral case there
so they need adding, that'll actually improve things though since we can
make it print the name of the thing it's mising which will be useful to
people trying to figure out what's going on (we used to do that but it
got lost in reshufflings).

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH] ASoC: soc-core: remove error due to probe deferral

[Stefan Agner]

Deferred probes shouldn't cause error messages in the boot log, so
change the dev_err() to the more harmless dev_info().

Signed-off-by: Stefan Agner <stefan@agner.ch>
---
 sound/soc/soc-core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sound/soc/soc-core.c b/sound/soc/soc-core.c
index b29d0f65611e..ba9b09517c5f 100644
--- a/sound/soc/soc-core.c
+++ b/sound/soc/soc-core.c
@@ -903,8 +903,8 @@ static int soc_bind_dai_link(struct snd_soc_card *card,
 	for (i = 0; i < rtd->num_codecs; i++) {
 		codec_dais[i] = snd_soc_find_dai(&codecs[i]);
 		if (!codec_dais[i]) {
-			dev_err(card->dev, "ASoC: CODEC DAI %s not registered\n",
-				codecs[i].dai_name);
+			dev_info(card->dev, "ASoC: CODEC DAI %s not registered\n",
+				 codecs[i].dai_name);
 			goto _err_defer;
 		}
 		snd_soc_rtdcom_add(rtd, codec_dais[i]->component);
-- 
2.20.1