On Fri, Oct 18, 2013 at 04:04:42PM +0200, Willy Tarreau wrote: > Greg, David, > > one of our customers faced a panic in latest 2.6.32 when both somaxconn > and the listen backlog are large on an IPv6 socket. It was also reported > by one haproxy user on the latest RHEL6 kernel a few months ago. We found > that the same bug affects 3.0 up to and including 3.0.100. > > Eric had already spotted that bug and fixed it in 3.2 with the following > patch : > > commit c16a98ed91597b40b22b540c6517103497ef8e74 > Author: Eric Dumazet <eric.dumazet@xxxxxxxxx> > Date: Wed Nov 23 15:49:31 2011 -0500 > > ipv6: tcp: fix panic in SYN processing > > commit 72a3effaf633bc ([NET]: Size listen hash tables using backlog > hint) added a bug allowing inet6_synq_hash() to return an out of bound > array index, because of u16 overflow. > > Bug can happen if system admins set net.core.somaxconn & > net.ipv4.tcp_max_syn_backlog sysctls to values greater than 65536 > > Signed-off-by: Eric Dumazet <eric.dumazet@xxxxxxxxx> > Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx> > > In practice, the bug extends to lower values as well (32768 and above), > because reqsk_queue_alloc() can round the number of entries to double of > the backlog by doing roundup_pow_of_two(backlog+1), resulting in > inet6_csk_search_req() calling inet6_synq_hash() with too large an integer. > > Could we please apply it to 3.0 before it finishes its life ? Unless David objects, I can queue this up just in time for the last 3.0.stable. David? -- To unsubscribe from this list: send the line "unsubscribe stable" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html