x86/ticketlock: Fix spin_unlock_wait() livelock
authorOleg Nesterov <oleg@redhat.com>
Mon, 1 Dec 2014 21:34:17 +0000 (22:34 +0100)
committerIngo Molnar <mingo@kernel.org>
Mon, 8 Dec 2014 10:36:44 +0000 (11:36 +0100)
commit78bff1c8684fb94f1ae7283688f90188b53fc433
tree48795bcb348947f7e213980570e7413dfc10fc3d
parent9fd7fc34cfcaf9f6c932ee1710cce83da3b7bd59
x86/ticketlock: Fix spin_unlock_wait() livelock

arch_spin_unlock_wait() looks very suboptimal, to the point I
think this is just wrong and can lead to livelock: if the lock
is heavily contended we can never see head == tail.

But we do not need to wait for arch_spin_is_locked() == F. If it
is locked we only need to wait until the current owner drops
this lock. So we could simply spin until old_head !=
lock->tickets.head in this case, but .head can overflow and thus
we can't check "unlocked" only once before the main loop.

Also, the "unlocked" check can ignore TICKET_SLOWPATH_FLAG bit.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Paul E.McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Waiman Long <Waiman.Long@hp.com>
Link: http://lkml.kernel.org/r/20141201213417.GA5842@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arch/x86/include/asm/spinlock.h