af_unix: limit unix_tot_inflight
commit6034a281f077b621859fb7b2f362fc9d693ea218
authorEric Dumazet <eric.dumazet@gmail.com>
Wed, 24 Nov 2010 17:15:27 +0000 (24 09:15 -0800)
committerGreg Kroah-Hartman <gregkh@suse.de>
Mon, 9 May 2011 22:55:36 +0000 (9 15:55 -0700)
treefe8cba40bbbd0643fefd1e17c7bac5c095390a99
parent6019f3837946cd5872ed473cd492d90c49228ee3
af_unix: limit unix_tot_inflight

commit 9915672d41273f5b77f1b3c29b391ffb7732b84b upstream.

Vegard Nossum found a unix socket OOM was possible, posting an exploit
program.

My analysis is we can eat all LOWMEM memory before unix_gc() being
called from unix_release_sock(). Moreover, the thread blocked in
unix_gc() can consume huge amount of time to perform cleanup because of
huge working set.

One way to handle this is to have a sensible limit on unix_tot_inflight,
tested from wait_for_unix_gc() and to force a call to unix_gc() if this
limit is hit.

This solves the OOM and also reduce overall latencies, and should not
slowdown normal workloads.

Reported-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
net/unix/garbage.c