netdev - Re: BUG: net/netfilter: KASAN: stack-out-of-bounds in tcp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Mon, 28 Mar 2016 18:17:02 -0700
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	jengelh@...i.de, kadlec@...ckhole.kfki.hu, sploving1@...il.com,
	pablo@...filter.org, kaber@...sh.net,
	netfilter-devel@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: BUG: net/netfilter: KASAN: stack-out-of-bounds in tcp_packet

On Mon, 2016-03-28 at 19:54 -0400, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@...il.com>
> Date: Mon, 28 Mar 2016 13:51:46 -0700
> 
> > On Mon, 2016-03-28 at 13:46 -0700, Eric Dumazet wrote:
> > 
> >> We have at least 384 bytes of padding in skb->head (this is struct
> >> skb_shared_info).
> >> 
> >> Whatever garbage we might read, current code is fine.
> >> 
> >> We have to deal with a false positive here.
> > 
> > Very similar to the one fixed in 
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=10ec9472f05b45c94db3c854d22581a20b97db41
> 
> I don't see them as similar.
> 
> The current options code we are talking about here never references
> past legitimate parts of the packet data.  We always check 'length',
> and we never access past the boundary it describes.
> 
> This was the entire point of my posting.
> 
> Talking about padding, rather than the logical correctness of the
> code, is therefore a distraction I think :-)

Not really, we do read one out of bound byte David.

length = 1;
...
while (length > 0) {
	int opcode = *ptr++; // Note that length is still 1
        switch (opcode) {
        ...
        default:
            opsize = *ptr++; // Note that length is still 1


        ...
        length -= opsize;
}

So we do read 2 bytes, while length was 1.

opsize definitely can read garbage.
Call it padding or redzone or whatever ;)