Re: [PATCH 2/2] lib: call native hex_to_bin() inside _kstrtoull()

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Mon, 2011-07-18 at 13:31 +0300, Alexey Dobriyan wrote: 
> On Mon, Jul 18, 2011 at 11:23 AM, Andy Shevchenko
> <andriy.shevchenko@linux.intel.com> wrote:
> > --- a/lib/kstrtox.c
> > +++ b/lib/kstrtox.c
> > @@ -39,25 +39,18 @@ static int _kstrtoull(const char *s, unsigned int base, unsigned long long *res)
> >        acc = 0;
> >        ok = 0;
> >        while (*s) {
> > -               unsigned int val;
> > +               int val;
> >
> > -               if ('0' <= *s && *s <= '9')
> > -                       val = *s - '0';
> > -               else if ('a' <= TOLOWER(*s) && TOLOWER(*s) <= 'f')
> > -                       val = TOLOWER(*s) - 'a' + 10;
> > -               else if (*s == '\n' && *(s + 1) == '\0')
> > +               if (unlikely(*s == '\n' && *(s + 1) == '\0'))
> >                        break;
> > -               else
> > -                       return -EINVAL;
> >
> > -               if (val >= base)
> > +               val = hex_to_bin(*s++);
> > +               if (val >= base || val < 0)
> >                        return -EINVAL;
> >                if (acc > div_u64(ULLONG_MAX - val, base))
> >                        return -ERANGE;
> >                acc = acc * base + val;
> >                ok = 1;
> > -
> > -               s++;
> >        }
> >        if (!ok)
> >                return -EINVAL;
> 
> 1. unlikely() and s++ move don't have anything to do with changes
That's true. I remove those changes.

> 2. I don't understand desire to use some half-thought out API,
>    in fact, restricting to radix 16 is arbitrary.
>    Without such restriction hex_to_bin doesn't make sense.
I doubt I get it. hex_to_bin validates its input indirectly. The
behaviour of your code (without changes mentioned in 1.) is the same.

-- 
Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Intel Finland Oy