Re: kernel BUG at lib/string.c:LINE! (6)

[Linus Torvalds <torvalds@linux-foundation.org>]

On Tue, Dec 22, 2020 at 6:44 AM syzbot
<syzbot+e86f7c428c8c50db65b4@syzkaller.appspotmail.com> wrote:
>
> The issue was bisected to:
>
> commit 2f78788b55ba ("ilog2: improve ilog2 for constant arguments")

That looks unlikely, although possibly some constant folding
improvement might make the fortify code notice something with it.

> detected buffer overflow in strlen
> ------------[ cut here ]------------
> kernel BUG at lib/string.c:1149!
> Call Trace:
>  strlen include/linux/string.h:325 [inline]
>  strlcpy include/linux/string.h:348 [inline]
>  xt_rateest_tg_checkentry+0x2a5/0x6b0 net/netfilter/xt_RATEEST.c:143

Honestly, this just looks like the traditional bug in "strlcpy()".

That BSD function is complete garbage, exactly because it doesn't
limit the source length. People tend to _think_ it does ("what's that
size_t argument for?") but strlcpy() only limits the *destination*
size, and the source is always read fully.

So it's a completely useless function if you can't implicitly trust
the source string - but that is almost always why people think they
should use it!

Nobody should use it. I really would like to remove it, and let
everybody know how incredibly broken sh*t that function is.

Can we please have everybody stop using strlcpy(). But in this
particular case, it's that xt_rateest_tg_checkentry() in
net/netfilter/xt_RATEEST.c

That said, this may be a real FORTIFY report if that info->name is
*supposed* to be trustworthy? The xt_RATETEST code does use
"info->name" a few lines earlier when it does

        est = __xt_rateest_lookup(xn, info->name);

or maybe the bisection is right, and this points to some problem with
__builtin_clzll?

                 Linus