[PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

[PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

[Francesco Saverio Pavone]

[-- Attachment #1: Type: text/plain, Size: 607 bytes --]

Hi all,

Sending this via attachment because my Postfix relay setup wasn't usable.
Patch is straightforward: rate-limit two dev_info() calls in
snd_parse_eld() that get triggered on every PCM open against an
HDMI codec with no sink attached.

Equivalent fix to the one Mark Brown did for hdac_hdmi in June 2025
(lkml.kernel.org/lkml/2025/6/13/1380), just for the generic helper used
by ASoC hdmi-codec.

Tested on Rock 5B+ (RK3588): two HDMI outputs, only one cable plugged
in. Before: ~30 "HDMI: Unknown ELD version 0" lines per probe burst.
After: 0.

Let me know if anything needs adjusting.

Thanks,
Sav

[-- Attachment #2: 0001-ALSA-pcm_drm_eld-rate-limit-ELD-parsing-errors.patch --]
[-- Type: application/octet-stream, Size: 2081 bytes --]

From 5c740f52f5bc1629125588cc7e5a2731baf91e8b Mon Sep 17 00:00:00 2001
From: Sav <62543096+dongioia@users.noreply.github.com>
Date: Sun, 10 May 2026 10:54:00 +0000
Subject: [PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

snd_parse_eld() is the generic helper used by ASoC hdmi-codec, equivalent
to the hdac_hdmi parse path Mark Brown rate-limited in
lkml.kernel.org/lkml/2025/6/13/1380. The same spam shows up here for the
same reason.

When a HDMI sink is not connected (for example a board with two HDMI
outputs and a cable in only one of them), userspace audio servers like
PipeWire keep opening the empty card. snd_parse_eld() then logs

    HDMI: Unknown ELD version 0

on every retry. On rk3588 that easily produces 30+ messages per burst.
A malformed ELD (MNL out of range) hits the same code path. Neither is
worth a separate dev_info() per occurrence; switch both to
dev_info_ratelimited().

Signed-off-by: Sav <62543096+dongioia@users.noreply.github.com>
---
 sound/core/pcm_drm_eld.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sound/core/pcm_drm_eld.c b/sound/core/pcm_drm_eld.c
index cb2eebaac85f..1941ee520063 100644
--- a/sound/core/pcm_drm_eld.c
+++ b/sound/core/pcm_drm_eld.c
@@ -334,7 +334,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
 	e->eld_ver = GRAB_BITS(buf, 0, 3, 5);
 	if (e->eld_ver != ELD_VER_CEA_861D &&
 	    e->eld_ver != ELD_VER_PARTIAL) {
-		dev_info(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
+		dev_info_ratelimited(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
 		goto out_fail;
 	}
 
@@ -357,7 +357,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
 	e->product_id	  = get_unaligned_le16(buf + 18);
 
 	if (mnl > ELD_MAX_MNL) {
-		dev_info(dev, "HDMI: MNL is reserved value %d\n", mnl);
+		dev_info_ratelimited(dev, "HDMI: MNL is reserved value %d\n", mnl);
 		goto out_fail;
 	} else if (ELD_FIXED_BYTES + mnl > size) {
 		dev_info(dev, "HDMI: out of range MNL %d\n", mnl);
-- 
2.54.0

Re: [PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

[Takashi Iwai]

On Sun, 10 May 2026 16:17:47 +0200,
Francesco Saverio Pavone wrote:
> 
> Hi all,
> 
> Sending this via attachment because my Postfix relay setup wasn't usable.
> Patch is straightforward: rate-limit two dev_info() calls in
> snd_parse_eld() that get triggered on every PCM open against an
> HDMI codec with no sink attached.
> 
> Equivalent fix to the one Mark Brown did for hdac_hdmi in June 2025
> (lkml.kernel.org/lkml/2025/6/13/1380), just for the generic helper used
> by ASoC hdmi-codec.
> 
> Tested on Rock 5B+ (RK3588): two HDMI outputs, only one cable plugged
> in. Before: ~30 "HDMI: Unknown ELD version 0" lines per probe burst.
> After: 0.
> 
> Let me know if anything needs adjusting.
> 
> Thanks,
> Sav

Is the patch itself missing?  Please resubmit.


thanks,

Takashi

[PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

[Sav]

From: Sav <62543096+dongioia@users.noreply.github.com>

snd_parse_eld() is the generic helper used by ASoC hdmi-codec, equivalent
to the hdac_hdmi parse path Mark Brown rate-limited in
lkml.kernel.org/lkml/2025/6/13/1380. The same spam shows up here for the
same reason.

When a HDMI sink is not connected (for example a board with two HDMI
outputs and a cable in only one of them), userspace audio servers like
PipeWire keep opening the empty card. snd_parse_eld() then logs

    HDMI: Unknown ELD version 0

on every retry. On rk3588 that easily produces 30+ messages per burst.
A malformed ELD (MNL out of range) hits the same code path. Neither is
worth a separate dev_info() per occurrence; switch both to
dev_info_ratelimited().

Signed-off-by: Sav <62543096+dongioia@users.noreply.github.com>
---
 sound/core/pcm_drm_eld.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sound/core/pcm_drm_eld.c b/sound/core/pcm_drm_eld.c
index cb2eebaac85f..1941ee520063 100644
--- a/sound/core/pcm_drm_eld.c
+++ b/sound/core/pcm_drm_eld.c
@@ -334,7 +334,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
 	e->eld_ver = GRAB_BITS(buf, 0, 3, 5);
 	if (e->eld_ver != ELD_VER_CEA_861D &&
 	    e->eld_ver != ELD_VER_PARTIAL) {
-		dev_info(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
+		dev_info_ratelimited(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
 		goto out_fail;
 	}
 
@@ -357,7 +357,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
 	e->product_id	  = get_unaligned_le16(buf + 18);
 
 	if (mnl > ELD_MAX_MNL) {
-		dev_info(dev, "HDMI: MNL is reserved value %d\n", mnl);
+		dev_info_ratelimited(dev, "HDMI: MNL is reserved value %d\n", mnl);
 		goto out_fail;
 	} else if (ELD_FIXED_BYTES + mnl > size) {
 		dev_info(dev, "HDMI: out of range MNL %d\n", mnl);
-- 
2.54.0

Re: [PATCH] ALSA: pcm_drm_eld: rate-limit ELD parsing errors

[Takashi Iwai]

On Fri, 15 May 2026 17:58:36 +0200,
Sav wrote:
> 
> From: Sav <62543096+dongioia@users.noreply.github.com>
> 
> snd_parse_eld() is the generic helper used by ASoC hdmi-codec, equivalent
> to the hdac_hdmi parse path Mark Brown rate-limited in
> lkml.kernel.org/lkml/2025/6/13/1380. The same spam shows up here for the
> same reason.
> 
> When a HDMI sink is not connected (for example a board with two HDMI
> outputs and a cable in only one of them), userspace audio servers like
> PipeWire keep opening the empty card. snd_parse_eld() then logs
> 
>     HDMI: Unknown ELD version 0
> 
> on every retry. On rk3588 that easily produces 30+ messages per burst.
> A malformed ELD (MNL out of range) hits the same code path. Neither is
> worth a separate dev_info() per occurrence; switch both to
> dev_info_ratelimited().
> 
> Signed-off-by: Sav <62543096+dongioia@users.noreply.github.com>

Could you use the proper name (a real name or a known identity) and
the mail address for Signed-off-by and From, and resubmit?  It's a
legal requirement for the upstreaming.


thanks,

Takashi

> ---
>  sound/core/pcm_drm_eld.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/sound/core/pcm_drm_eld.c b/sound/core/pcm_drm_eld.c
> index cb2eebaac85f..1941ee520063 100644
> --- a/sound/core/pcm_drm_eld.c
> +++ b/sound/core/pcm_drm_eld.c
> @@ -334,7 +334,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
>  	e->eld_ver = GRAB_BITS(buf, 0, 3, 5);
>  	if (e->eld_ver != ELD_VER_CEA_861D &&
>  	    e->eld_ver != ELD_VER_PARTIAL) {
> -		dev_info(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
> +		dev_info_ratelimited(dev, "HDMI: Unknown ELD version %d\n", e->eld_ver);
>  		goto out_fail;
>  	}
>  
> @@ -357,7 +357,7 @@ int snd_parse_eld(struct device *dev, struct snd_parsed_hdmi_eld *e,
>  	e->product_id	  = get_unaligned_le16(buf + 18);
>  
>  	if (mnl > ELD_MAX_MNL) {
> -		dev_info(dev, "HDMI: MNL is reserved value %d\n", mnl);
> +		dev_info_ratelimited(dev, "HDMI: MNL is reserved value %d\n", mnl);
>  		goto out_fail;
>  	} else if (ELD_FIXED_BYTES + mnl > size) {
>  		dev_info(dev, "HDMI: out of range MNL %d\n", mnl);
> -- 
> 2.54.0
>