On Tue, Dec 11, 2018 at 09:22:22AM -0800, Linus Torvalds wrote:
> On Tue, Dec 11, 2018 at 7:21 AM Thomas Preston
> <thomas.pres...@codethink.co.uk> wrote:
> >
> > Stop using the obsolete functions simple_strtoul() and
> > simple_strtoull(). Instead, we should use the improved kstrtol() and
> > kstrtoll() functions. To do this, we must copy the current field into a
> > null-terminated tmpstr and advance the variable `next` manually.
> 
> I see what you're trying to do, but this fix is much much worse than
> the bug was.
> 
> > +               if (field_width > 0) {
> > +                       char tmpstr[INT_BUF_LEN];
> > +                       int ret;
> > +
> > +                       strscpy(tmpstr, str, field_width+1);
> 
> If field_width is larger than INT_BUF_LEN, you are now corrupting kernel 
> stack.
> 
> And no, you can't fix it by limiting field_width, since a large
> field_width is quite possible and might even be valid - and still fit
> in an int. Maybe the number is
> 
>     000000000000000000000001
> 
> or something?
> 
> A fix might be to skip leading zeroes.
> 
> Honestly, just do it by hand. Don't use kstrol and friends at all.
> Just do something like
> 
>     unsigned long long val = 0;
>     p = str;
>     for (;;) {
>         int c;
>         if (field_width > 0 && p - str >= field_width)
>             break;
>         c = hexval(*p++);
>         if (c < 0 || c > base)
>             break;
>         val = val * base + c;
>         // check for overflow

I think it's slightly more complicated, I run the following test case on glibc:

        uint32_t hi, lo, t;

        sscanf("00fafafafa0d0b0b0b0c000000", "%8x%8x%x", &hi, &lo, &t);

64-bit:
        HI: 00fafafa LO: fa0d0b0b (c000000)
32-bit:
        HI: 00fafafa LO: fa0d0b0b (ffffffff)

>     }
>     /* Now do "sign" and range checking on val */
>     /* Ta-daa, all done */
> 
> or similar.  Treat the above as pseudo-code, I didn't fill in all the details.
> 
>             Linus

-- 
With Best Regards,
Andy Shevchenko


Reply via email to