net: unix: fix inflight counting bug in garbage collector
authorMiklos Szeredi <mszeredi@suse.cz>
Sun, 9 Nov 2008 14:23:57 +0000 (15:23 +0100)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sun, 9 Nov 2008 19:17:33 +0000 (11:17 -0800)
commit6209344f5a3795d34b7f2c0061f49802283b6bdd
tree5c037ddbb8caac17b0c6101c9ab86387df106d41
parent058e3739f6b0753696db1952378de9e8d2a11735
net: unix: fix inflight counting bug in garbage collector

Previously I assumed that the receive queues of candidates don't
change during the GC.  This is only half true, nothing can be received
from the queues (see comment in unix_gc()), but buffers could be added
through the other half of the socket pair, which may still have file
descriptors referring to it.

This can result in inc_inflight_move_tail() erronously increasing the
"inflight" counter for a unix socket for which dec_inflight() wasn't
previously called.  This in turn can trigger the "BUG_ON(total_refs <
inflight_refs)" in a later garbage collection run.

Fix this by only manipulating the "inflight" counter for sockets which
are candidates themselves.  Duplicating the file references in
unix_attach_fds() is also needed to prevent a socket becoming a
candidate for GC while the skb that contains it is not yet queued.

Reported-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/net/af_unix.h
net/unix/af_unix.c
net/unix/garbage.c