stop_machine stalls for a considerable period on large cpu count machines.
PATCH 1/1 stop_machine stalls for a considerable period on large cpu count machines.
I forgot to include the linux-kernel mailing list.
Robin
----- Forwarded message from Robin Holt <[email protected]> -----
Date: Sat, 27 Jun 2009 05:53:11 -0500
From: Robin Holt <[email protected]>
To: Linus Torvalds <[email protected]>
Cc: Rusty Russell <[email protected]>, [email protected],
mike travis <[email protected]>
Subject: [Patch] stop_machine stalls for a considerable period on large cpu
count machines.
Mike Travis noted that a 2048 cpu machine booting would take hours to get through its modprobes. We would get numerous back traces from stop_cpu indicating they had not serviced interrupts.
A quick code review indicated we have a situation of heavy cacheline contention due to the 'state' (read-mostly) and 'thread_ack'
(write-mostly) variables being located in the same cacheline.
--- kernel/stop_machine.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
kernel/stop_machine.c +2 -2
@@ -36,8 +36,8 @@ struct stop_machine_data {/* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */ static unsigned int num_threads; -static atomic_t thread_ack; -static DEFINE_MUTEX (lock); +static atomic_t thread_ack____cacheline_aligned ; +static DEFINE_MUTEX (lock)____cacheline_aligned ;/* setup_lock protects refcount, stop_machine_wq and stop_machine_work. */ static DEFINE_MUTEX (setup_lock);/* Users of stop_machine. */
----- End forwarded message -----