Re: possible integer underflow in __sctp_auth_cid()

On 06/30/2012 08:17 PM, Dan Carpenter wrote:

In 555d3d5d "SCTP: Fix chunk acceptance when no authenticated chunks
were listed.", we added a check for if (param->param_hdr.length == 0).
Shouldn't that check be a check for if
(param->param_hdr.length < sizeos(sizeof(sctp_paramhdr_t)))?  Otherwise,
when we do the substraction on the next line we would unintentionally
end up with a high positive number.

I had a similar question about sctp_auth_ep_add_chunkid():

net/sctp/auth.c
   770          /* Check if we can add this chunk to the array */
   771          param_len = ntohs(p->param_hdr.length);
   772          nchunks = param_len - sizeof(sctp_paramhdr_t);
   773          if (nchunks == SCTP_NUM_CHUNK_TYPES)
   774                  return -EINVAL;
   775
   776          p->chunks[nchunks] = chunk_id;

If param_len is less than sizeof(sctp_paramhdr_t) we could write past
the end of the array.  There are a couple other places with this same
subtraction as well.

This will not happen because that p which means ep->auth_chunk_list
is maintained by the endpoint itself, and the default value is
sizeof(sctp_paramhdr_t) or sizeof(sctp_paramhdr_t) + 2.

Instead, I found that if user enable AUTH after the endpoint is
created, set the AUTH chunk may cause panic, because the
ep->auth_chunk_list is NULL.

I think we should introduce a help function to check whether the
AUTH is enabled instead only check sctp_auth_enable, may like:

bool sctp_auth_ep_enabled(struct sctp_endpoint *ep)
{
	if (!sctp_auth_enable)
		return false;

	if (!ep->auth_chunk_list || !ep->auth_hmacs_list)
		return false;

	return true;
}


Regards
Yongjun Wei

regards,
dan carpenter
--
To unsubscribe from this list: send the line "unsubscribe linux-sctp" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html