On Thu, Feb 11, 2021 at 10:09 AM Geert Uytterhoeven
<ge...@linux-m68k.org> wrote:
> On Wed, Feb 10, 2021 at 11:33 PM Drew Fustini <d...@beagleboard.org> wrote:
> > +#define PINMUX_MAX_NAME 64
> > + if (len > (PINMUX_MAX_NAME * 2)) {
> > + dev_err(pctldev->dev, "write too big for buffer");
> > + return -EINVAL;
> > + }
> > +
> > + buf = kzalloc(PINMUX_MAX_NAME * 2, GFP_KERNEL);
> > + ret = strncpy_from_user(buf, user_buf, PINMUX_MAX_NAME * 2);
>
> While this guarantees buf is not overflowed...
>
> > + if (ret < 0) {
> > + dev_err(pctldev->dev, "failed to copy buffer from
> > userspace");
> > + goto free_gname;
> > + }
> > + buf[len-1] = '\0';
> > +
> > + ret = sscanf(buf, "%s %s", fname, gname);
>
> ... one of the two strings can still be longer than PINMUX_MAX_NAME,
> thus overflowing fname or gname.
>
> As buf is already a copy, it may be easier to just find the strings in
> buf, write the NUL terminators into buf, and set up fname and gname
> to point to the strings inside buf.
You beat me up to it. I was about to comment the same.
So, indeed, instead of sscanf it's simply better and faster to do just
something like
fname = strstrip(buf) ;
if (*fname == '\0') {
...
return -EINVAL;
}
gname = strchr(fname, ' ');
if (!gname) {
...
return -EINVAL;
}
*gname++ = '\0';
on top of the buf pointer.
> > + if (ret != 2) {
> > + dev_err(pctldev->dev, "expected format: <function-name>
> > <group-name>");
> > + goto free_gname;
> > + }
--
With Best Regards,
Andy Shevchenko
