Re: [PATCH 3v] char: hw_random: core.c: Changed from using strncat to strlcat

[Joe Perches <joe@perches.com>]

On Thu, 2014-10-16 at 19:41 +0200, Rickard Strandqvist wrote:
> 2014-10-16 19:25 GMT+02:00 Joe Perches <joe@perches.com>:
> > On Thu, 2014-10-16 at 19:15 +0200, Rickard Strandqvist wrote:
> >> The buf is used to hold the list of hwrng devices registered.
> >> The old code ensures we don't walk off the end of buf as we
> >> fill it, but it's unnecessarily complicated and thus difficult
> >> to maintain. Simplify it by using strlcat.
> >> We also ensure the string within buf is NULL terminated
> >> so the final strlen is ok.
> > []
> >> diff --git a/drivers/char/hw_random/core.c b/drivers/char/hw_random/core.c
> > []
> >> @@ -289,16 +288,13 @@ static ssize_t hwrng_attr_available_show(struct device *dev,
> >>               return -ERESTARTSYS;
> >>       buf[0] = '\0';
> >>       list_for_each_entry(rng, &rng_list, list) {
> >> -             strncat(buf, rng->name, PAGE_SIZE - ret - 1);
> >> -             ret += strlen(rng->name);
> >> -             strncat(buf, " ", PAGE_SIZE - ret - 1);
> >> -             ret++;
> >> +             strlcat(buf, rng->name, PAGE_SIZE);
> >> +             strlcat(buf, " ", PAGE_SIZE);
> >>       }
> >> -     strncat(buf, "\n", PAGE_SIZE - ret - 1);
> >> -     ret++;
> >> +     strlcat(buf, "\n", PAGE_SIZE);
> >>       mutex_unlock(&rng_mutex);
> >>
> >> -     return ret;
> >> +     return strlen(buf);
> >>  }
> >>
> >>  static DEVICE_ATTR(rng_current, S_IRUGO | S_IWUSR,
> >
> > Rickard, can you please use some optimizations here
> > (and elsewhere) so that strlcat doesn't always have
> > to strlen the first string and the return doesn't
> > have to do the strlen too?
> >
> > You could use a temporary for the returned length
> > of the strlcat so that if it's shorter than
> > the buffer, the next strlcat can start at the
> > appropriate known position instead of having
> > to do the initial strlen again and again.
[]
> But the others, I am very hesitant about. Because strlcat like
> snprintf and strlcpy returns the length that would be copied rather
> than what is actually copied. Hence such a code to be even more
> complex than before.

None of the conversions you've done seem to use the
return value so maybe there should be yet another
function that doesn't return an overrun value.