On Mon, Jun 08, 2020 at 01:18:12PM +0300, Slava Aseev wrote:
Hello!
I recently tested the OSPF protocol implementation with AFL fuzzer and found some problem, that causes invalid memory read.
https://gitlab.labs.nic.cz/labs/bird/-/blob/master/proto/ospf/packet.c#L463 The problem is that 'plen' may be less than 'hlen'. And if this happens, then 'blen' will wraps around:
Hello Thanks for the bugreport. Seems like when union ospf_auth2 was removed from struct ospf_packet, the invalid length check for plen was not properly updated. Will fix that. At first glance, it seems that it would not cause a crash, as in ipsum_verify() it would overflow back to a pointer before 'body' and access no memory instead of accessing all memory behind 'body'.
if (ospf_is_v2(p) && (pkt->autype != OSPF_AUTH_CRYPT)) { uint hlen = sizeof(struct ospf_packet) + sizeof(union ospf_auth2); uint blen = plen - hlen; void *body = ((void *) pkt) + hlen;
if (!ipsum_verify(pkt, sizeof(struct ospf_packet), body, blen, NULL)) DROP("invalid checksum", ntohs(pkt->checksum)); }
'plen' declared here: https://gitlab.labs.nic.cz/labs/bird/-/blob/master/proto/ospf/packet.c#L443
uint plen = ntohs(pkt->length); if ((plen < sizeof(struct ospf_packet)) || ((plen % 4) != 0)) DROP("invalid length", plen);
-- Best regards, Slava Aseev
-- Elen sila lumenn' omentielvo Ondrej 'Santiago' Zajicek (email: santiago@crfreenet.org) OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net) "To err is human -- to blame it on a computer is even more so."