Re: [PATCH 0/2][RFC] Refactor snd primitives refcounters

[Takashi Iwai <tiwai@suse.de>]

On Thu, 24 Aug 2023 23:02:51 +0200,
cujomalainey@chromium.org wrote:
> 
> From: Curtis Malainey <cujomalainey@chromium.org>
> 
> As previously identified in [1] there are some issues with how kobjs are
> handled in sound/core. The solution provided in [2] is a workaround for
> the issues to fix the failures.
> 
> This series is a first attempt at the larger refactor needed to properly
> handle the objects.
> 
> [1] https://mailman.alsa-project.org/hyperkitty/list/alsa-devel@alsa-project.org/message/3B2VMPFATCJCRD76DM36LZAICPDP2R6A/
> [2] https://mailman.alsa-project.org/hyperkitty/list/alsa-devel@alsa-project.org/message/JZIPB4OIC3OK6YAGTPMAEKXMLUWM5PTA/
> 
> Curtis Malainey (2):
>   ALSA: core: add snd_device_init
>   ALSA: core: split control primitives out of snd_card

Thanks for the patches.  But, as the 6.6 merge window open is pretty
close, I'd postpone those unless it's urgently needed.

Also, before moving the resource tied with the device object, we'll
need a refcount to the ctl dev from pcm dev, as PCM does release chmap
at its free path (calling free_chmap()).  Otherwise it'll lead to
another UAF, if both objects releases are done asynchronously without
dependency.


BTW, the cover letter and the subject prefix of the patches don't
match, and also the cover letter didn't include Cc.  Please try to
make both cover letter and patches sent properly.


thanks,


Takashi