[PATCH] inotify: fix race between the kernel and user space
authorJohn McCutchan <ttb@tentacle.dhs.org>
Mon, 1 Aug 2005 15:00:45 +0000 (11:00 -0400)
committerLinus Torvalds <torvalds@g5.osdl.org>
Mon, 1 Aug 2005 16:16:53 +0000 (09:16 -0700)
commitb9c55d29e9fced1eb1b4c252b2efd4b55a0c3c7f
tree3fc1fe1e0f40d673d9675a3db5b84ca00b81eaea
parent7544953685859875b5ac0260b6b1856066c092d6
[PATCH] inotify: fix race between the kernel and user space

When you rm a watch, an IN_IGNORED event is sent down the event queue
with the watch descriptor that you just rm'd.

If you then add a watch you could get the ignored watch's wd and if you
haven't read the entire event queue, user space will think that it's
newly created watch was just ignored.

To avoid this problem we just use idr_get_new_above instead of
idr_get_new.

Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
fs/inotify.c