Re: [PATCH] ALSA: pcmtest: fix reference leak on failed device registration

[Takashi Iwai]

On Thu, 23 Apr 2026 07:19:20 +0200,
Jiri Slaby wrote:
> 
> On 17. 04. 26, 12:52, Takashi Iwai wrote:
> > On Fri, 17 Apr 2026 09:57:20 +0200,
> > Takashi Iwai wrote:
> >> 
> >> On Wed, 15 Apr 2026 21:31:38 +0200,
> >> Guangshuo Li wrote:
> >>> 
> >>> When platform_device_register() fails in mod_init(), the embedded struct
> >>> device in pcmtst_pdev has already been initialized by
> >>> device_initialize(), but the failure path returns the error without
> >>> dropping the device reference for the current platform device:
> >>> 
> >>>    mod_init()
> >>>      -> platform_device_register(&pcmtst_pdev)
> >>>         -> device_initialize(&pcmtst_pdev.dev)
> >>>         -> setup_pdev_dma_masks(&pcmtst_pdev)
> >>>         -> platform_device_add(&pcmtst_pdev)
> >>> 
> >>> This leads to a reference leak when platform_device_register() fails.
> >>> Fix this by calling platform_device_put() before returning the error.
> >>> 
> >>> The issue was identified by a static analysis tool I developed and
> >>> confirmed by manual review.
> >>> 
> >>> Fixes: 315a3d57c64c5 ("ALSA: Implement the new Virtual PCM Test Driver")
> >>> Cc: stable@xxxxxxxxxxxxxxx
> >>> Signed-off-by: Guangshuo Li <lgs201920130244@xxxxxxxxx>
> >> 
> >> Thanks, applied now.
> > 
> > ... and now I looked through the whole tree, and noticed that the
> > majority of callers of platform_device_register() don't care the error
> > cases without calling platform_device_put().  There are over a hundred
> > callers of platform_device_register() while only 5 or so are doing the
> > proper cleanup at the error.
> 
> Moreover, unless the static pcmtst_pdev has ->type->release or
> ->release set, the patch triggers a warning upon put().

Ah a good point, the call should be conditional.

> > Judging from the numbers above, it might be better to change the
> > behavior of platform_device_register() itself to call *_put() at the
> > error case internally.
> 
> Yes, that's what should be done instead.

OK, I'm going to revert this patch for pcmtest.c for now (so back to
the upstream state again), so that we can work on the
platform_device_register() side.


thanks,

Takashi