[PATCH] Remove down_write() from taskstats code invoked on the exit() path
authorShailabh Nagar <nagar@watson.ibm.com>
Fri, 14 Jul 2006 07:24:47 +0000 (00:24 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Sat, 15 Jul 2006 04:53:57 +0000 (21:53 -0700)
commitbb129994c3bff9c5e8df91f05d7e9b6402fbd83f
tree8d8c4e27af727c9de42ac6aeae04f70537e6bd01
parentf9fd8914c1acca0d98b69d831b128d5b52f03c51
[PATCH] Remove down_write() from taskstats code invoked on the exit() path

In send_cpu_listeners(), which is called on the exit path, a down_write()
was protecting operations like skb_clone() and genlmsg_unicast() that do
GFP_KERNEL allocations.  If the oom-killer decides to kill tasks to satisfy
the allocations,the exit of those tasks could block on the same semphore.

The down_write() was only needed to allow removal of invalid listeners from
the listener list.  The patch converts the down_write to a down_read and
defers the removal to a separate critical region.  This ensures that even
if the oom-killer is called, no other task's exit is blocked as it can
still acquire another down_read.

Thanks to Andrew Morton & Herbert Xu for pointing out the oom related
pitfalls, and to Chandra Seetharaman for suggesting this fix instead of
using something more complex like RCU.

Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Shailabh Nagar <nagar@watson.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/taskstats.c