rcu: Fix batch-limit size problem
authorEric Dumazet <edumazet@google.com>
Thu, 18 Oct 2012 11:55:36 +0000 (04:55 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Mon, 17 Dec 2012 17:27:17 +0000 (09:27 -0800)
commit2989161c3a4079f4b3340464a3a83af7bdb026ea
treea5e6f4f2b0fe6e89fc620ec2515a68d3d53d7e77
parentf063c3cae7e16c5d45941ab358aa8063349724f8
rcu: Fix batch-limit size problem

commit 878d7439d0f45a95869e417576774673d1fa243f upstream.

Commit 29c00b4a1d9e27 (rcu: Add event-tracing for RCU callback
invocation) added a regression in rcu_do_batch()

Under stress, RCU is supposed to allow to process all items in queue,
instead of a batch of 10 items (blimit), but an integer overflow makes
the effective limit being 1.  So, unless there is frequent idle periods
(during which RCU ignores batch limits), RCU can be forced into a
state where it cannot keep up with the callback-generation rate,
eventually resulting in OOM.

This commit therefore converts a few variables in rcu_do_batch() from
int to long to fix this problem, along with the module parameters
controlling the batch limits.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/rcutree.c