Re: [PATCH] ASoC: soc-dai: don't use dev_{set/get}_drvdata() on DAI

[Charles Keepax <ckeepax@opensource.cirrus.com>]

On Thu, Mar 05, 2026 at 01:56:44AM +0000, Kuninori Morimoto wrote:
> We have 2 type of driver data set functions.
> 
> (A)	snd_soc_component_set_drvdata()
> (B)	snd_soc_dai_set_drvdata()
> 
> Both are using dev_set_drvdata() with own dev
> 
> (A)	static inline void snd_soc_component_set_drvdata(...)
> 	{
> 		dev_set_drvdata(component->dev, data);
> 	}			^^^^^^^^^^^^^^
> 
> (B)	static inline void snd_soc_dai_set_drvdata(...)
> 	{
> 		dev_set_drvdata(dai->dev, data);
> 	}			^^^^^^^^
> 
> But, these are same dev.
> 
> 	struct snd_soc_dai *snd_soc_register_dai(...)
> 	{
> (A)		struct device *dev = component->dev;
> 		...
> (B)		dai->dev = dev;
> 		...
> 	}
> 
> This means, if (A) and (B) were called in the same time, private data
> will be overwritten.
> 
> But, in the same time, commit 5f86d41d0410 ("ASoC: soc-dai: Add private
> data to snd_soc_dai") added dai->priv for private data, and sdca_asoc.c
> is only user of it.
> 
> We can re-use this dai->priv for snd_soc_dai_set_drvdata() (B) instead,
> and avoid the issue.
> 
> Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
> ---

Hm... the patch looks pretty reasonable to me but does seem to
cause issues on my system. I will try to find some time to
investigate further:

Mar 05 13:06:17.645064 ck-upx-mtl kernel: i915 0000:00:02.0: [drm] *ERROR* DC state mismatch (0x0 -> 0x2)
Mar 05 13:06:17.647247 ck-upx-mtl kernel: BUG: kernel NULL pointer dereference, address: 0000000000000000
Mar 05 13:06:17.647257 ck-upx-mtl kernel: #PF: supervisor read access in kernel mode
Mar 05 13:06:17.647265 ck-upx-mtl kernel: #PF: error_code(0x0000) - not-present page
Mar 05 13:06:17.647270 ck-upx-mtl kernel: PGD 0 P4D 0 
Mar 05 13:06:17.647276 ck-upx-mtl kernel: Oops: Oops: 0000 [#1] SMP NOPTI
Mar 05 13:06:17.647281 ck-upx-mtl kernel: CPU: 0 UID: 0 PID: 98 Comm: (udev-worker) Tainted: G            E 7.0.0-rc2-mykernel #1 PREEMPT(lazy) Mar 05 13:06:17.647286 ck-upx-mtl kernel: Tainted: [E]=UNSIGNED_MODULE
Mar 05 13:06:17.647291 ck-upx-mtl kernel: Hardware name: AAEON UPX-MTL01/UPX-MTL01, BIOS UXMTAM13 02/13/2025
Mar 05 13:06:17.647295 ck-upx-mtl kernel: RIP: 0010:intel_dmc_update_dc6_allowed_count+0x17/0xb0 [i915]

I wonder if maybe something is mixing these functions and
expecting them to set the same thing?

Thanks,
Charles