Re: [PATCH v6] ALSA: control: Verify put() result when in debug mode

[Takashi Iwai <tiwai@suse.de>]

On Tue, 24 Feb 2026 16:46:55 +0100,
Cezary Rojewski wrote:
> 
> On 2026-02-24 1:09 PM, Takashi Iwai wrote:
> > On Mon, 23 Feb 2026 15:49:37 +0100,
> > Cezary Rojewski wrote:
> 
> ...
> 
> >> --- a/sound/core/init.c
> >> +++ b/sound/core/init.c
> > (snip)
> >> @@ -537,6 +542,9 @@ void snd_card_disconnect(struct snd_card *card)
> >>   		synchronize_irq(card->sync_irq);
> >>     	snd_info_card_disconnect(card);
> >> +#ifdef CONFIG_SND_CTL_DEBUG
> >> +	kfree(card->value_buf);
> >> +#endif
> >>   #ifdef CONFIG_SND_DEBUG
> >>   	debugfs_remove(card->debugfs_root);
> >>   	card->debugfs_root = NULL;
> > 
> > IMO, it's better to be put in snd_card_do_free() which is the place to
> > release the actual resources.
> 
> No problem, a question though - I've written the above to be cohesive
> as CONFIG_SND_DEBUG-part isn't part of snd_card_do_free() and what I'm
> adding is definitely also of debug-related capability. IIRC
> debugfs_remove() may invoke kfree() operations. Why isn't it part of
> snd_card_do_free() too?

The *_disconnect() is a stage to stop the operation and disconnect the
interfaces from user-space, so debugfs_remove() is invoked there.
You can do kfree() if all things are detached, but in general we do
the release in a later phase, as some objects may be still alive.

snd_card_do_free() is the actual destructor for the resources, where
all things can be safely freed.


Takashi