Re: [PATCH] staging: gdm724x: gdm_tty: Fix for possible null pointer dereference

[Dan Carpenter <dan.carpenter@oracle.com>]

On Wed, Feb 04, 2015 at 06:46:34PM +0100, Rickard Strandqvist wrote:
> 2015-02-02 17:36 GMT+01:00 Sudip Mukherjee <sudipm.mukherjee@gmail.com>:
> > On Thu, Jan 29, 2015 at 07:46:10PM +0100, Rickard Strandqvist wrote:
> >> diff --git a/drivers/staging/gdm724x/gdm_tty.c b/drivers/staging/gdm724x/gdm_tty.c
> >> index 001348c..66b356e 100644
> >> --- a/drivers/staging/gdm724x/gdm_tty.c
> >> +++ b/drivers/staging/gdm724x/gdm_tty.c
> >> @@ -145,7 +145,7 @@ static int gdm_tty_recv_complete(void *data,
> >>       struct gdm *gdm = tty_dev->gdm[index];
> >>
> >>       if (!GDM_TTY_READY(gdm)) {
> >> -             if (complete == RECV_PACKET_PROCESS_COMPLETE)
> >> +             if (gdm && complete == RECV_PACKET_PROCESS_COMPLETE)
> > GDM_TTY_READY() is already checking for gdm, there is no chance that gdm can be null at this point. so this additional check is not required.
> >
> > regards
> > sudip
> >>                       gdm_tty_recv(gdm, gdm_tty_recv_complete);
> >>               return TO_HOST_PORT_CLOSE;
> >>       }
> 
> Hi Sudip
> 
> Yes, GDM_TTY_READY checks gdm, but this is a if(! )
> 

You're right.  But, by that same logic, we should also test
gdm->tty_dev.  So it looks like this:

	if (!GDM_TTY_READY(gdm))  {
		if (gdm && gdm->tty_dev && complete == RECV_PACKET_PROCESS_COMPLETE)
			gdm_tty_recv(gdm, gdm_tty_recv_complete);
		return TO_HOST_PORT_CLOSE;
	}

That is really sucky...  Garbage code like this is why kernel style
doesn't favour macros.  We should just open code GDM_TTY_READY() and
gdm_tty_recv() so that people can read the code.

I wonder if "gdm->tty_dev" is the same as the "tty_dev" parameter?

regards,
dan carpenter