The impact of this patch is that the RCU read-side is now marked "notrace" in the atomic call chain notifier. It also switch this atomic notifier call chain from RCU to "sched RCU". Disabling preemption in the atomic notifier does not have any significant side-effect, because atomic notifiers already expect to execute in atomic context. The notifier unregistration is adapted to use "sched RCU".
The "Generic Ring Buffer Library" needs to be able to use the atomic notifier from cpu idle entry. The aim is to ensure the ring buffer makes the last events ready for reading before the CPU goes idle. This is required by the Generic Ring Buffer Library because it uses deferrable timers for trace streaming. This is also useful to the die notifier, which makes breakpoint insertion valid in more code paths. I submit this patch for inclusion in -tip. Meanwhile, I will add this patch to the tip-pull-queue in the ringbuffer git tree: git://git.kernel.org/pub/scm/linux/kernel/git/compudj/linux-2.6-ringbuffer.git branch: tip-pull-queue Signed-off-by: Mathieu Desnoyers <[email protected]> Reviewed-by: Paul E. McKenney <[email protected]> Reviewed-by: Masami Hiramatsu <[email protected]> CC: Jason Baron <[email protected]> CC: Ingo Molnar <[email protected]> CC: Thomas Gleixner <[email protected]> --- kernel/notifier.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) Index: linux-2.6-lttng/kernel/notifier.c =================================================================== --- linux-2.6-lttng.orig/kernel/notifier.c 2009-11-12 17:58:56.000000000 -0500 +++ linux-2.6-lttng/kernel/notifier.c 2009-11-12 18:03:28.000000000 -0500 @@ -148,7 +148,7 @@ int atomic_notifier_chain_unregister(str spin_lock_irqsave(&nh->lock, flags); ret = notifier_chain_unregister(&nh->head, n); spin_unlock_irqrestore(&nh->lock, flags); - synchronize_rcu(); + synchronize_sched(); return ret; } EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister); @@ -178,9 +178,9 @@ int __kprobes __atomic_notifier_call_cha { int ret; - rcu_read_lock(); + rcu_read_lock_sched_notrace(); ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls); - rcu_read_unlock(); + rcu_read_unlock_sched_notrace(); return ret; } EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain); -- Mathieu Desnoyers Operating System Efficiency R&D Consultant EfficiOS Inc. http://www.efficios.com _______________________________________________ ltt-dev mailing list [email protected] http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
