Re: [PATCH 37/69] ALSA: usx2y: check for failure of usb_alloc_urb()

[Greg Kroah-Hartman]

On Tue, May 04, 2021 at 10:27:34AM +0200, Takashi Iwai wrote:
> On Mon, 03 May 2021 22:33:52 +0200,
> Jaroslav Kysela wrote:
> > 
> > Dne 03. 05. 21 v 13:57 Greg Kroah-Hartman napsal(a):
> > > While it is almost impossible to hit an error calling usb_alloc_urb(),
> > > to make systems like syzbot which does error injection, and some static
> > > analysis tools happy, properly handle errors on this path by unwinding
> > > the previously allocated urbs and freeing them.
> > 
> > Perhaps, I miss something, but this revert and add the cleanup to the wrong
> > place makes things worse:
> > 
> > The usb_stream_free() is called when init_urbs() fails (returns an error), so
> > all urbs are freed there and pointers are reset to NULL. Your code frees urbs
> > twice on an allocation error.
> > 
> > The reverted code does the job better.
> 
> Right, the suggested patch will cause the double-free, and the
> reverted code is fine in this regard.

That wasn't obvious at all, sorry about that.

> In anyway, the cleanup of this driver code is on my post-15.3 TODO
> list.  So, Greg, could you drop the revert and the additional fix at
> this round for usx2y driver?

Ok, will drop the revert and this patch, thanks for the review and sorry
for the noise.

greg k-h