Re: [PATCH 4.19 062/125] ALSA: usb-audio: Fix possible NULL dereference at create_yamaha_midi_quirk()

From: Takashi Iwai
Date: Wed Nov 13 2019 - 05:36:49 EST


On Wed, 13 Nov 2019 11:35:16 +0100,
Takashi Iwai wrote:
>
> On Wed, 13 Nov 2019 11:19:54 +0100,
> Pavel Machek wrote:
> >
> > Hi!
> >
> > > From: Takashi Iwai <tiwai@xxxxxxx>
> > >
> > > commit 60849562a5db4a1eee2160167e4dce4590d3eafe upstream.
> > >
> > > The previous addition of descriptor validation may lead to a NULL
> > > dereference at create_yamaha_midi_quirk() when either injd or outjd is
> > > NULL. Add proper non-NULL checks.
> >
> > This is wrong.
> >
> > > @@ -259,8 +259,8 @@ static int create_yamaha_midi_quirk(stru
> > > NULL, USB_MS_MIDI_OUT_JACK);
> > > if (!injd && !outjd)
> > > return -ENODEV;
> >
> > Clearly code wants to allow at most one of them to be NULL.
> >
> > > - if (!snd_usb_validate_midi_desc(injd) ||
> > > - !snd_usb_validate_midi_desc(outjd))
> > > + if (!(injd && snd_usb_validate_midi_desc(injd)) ||
> > > + !(outjd && snd_usb_validate_midi_desc(outjd)))
> > > return -ENODEV;
> >
> > Yet it will return here if it is. Correct check would be
> >
> > + if (!(!injd || snd_usb_validate_midi_desc(injd)) ||
> > + !(!outjd || snd_usb_validate_midi_desc(outjd)))
> >
> > AFAICT.
>
> Erm, right, but a better representation is:
>
> if ((injd && !snd_usb_validate_midi_desc(injd)) ||
> (outjd && !snd_usb_validate_midi_desc(injd)))

Of course, another typo:

if ((injd && !snd_usb_validate_midi_desc(injd)) ||
(outjd && !snd_usb_validate_midi_desc(outjd)))


Takashi