locking/mcs: Fix mcs_spin_lock() ordering
authorPeter Zijlstra <peterz@infradead.org>
Mon, 1 Feb 2016 14:11:28 +0000 (15:11 +0100)
committerIngo Molnar <mingo@kernel.org>
Mon, 29 Feb 2016 09:02:41 +0000 (10:02 +0100)
commit920c720aa5aa3900a7f1689228fdfc2580a91e7e
treede04ffb2172a814c84fbbb1fd906bb9c2b139ae5
parent39a1142dbba04d2e08259bd10a369465c932126b
locking/mcs: Fix mcs_spin_lock() ordering

Similar to commit b4b29f94856a ("locking/osq: Fix ordering of node
initialisation in osq_lock") the use of xchg_acquire() is
fundamentally broken with MCS like constructs.

Furthermore, it turns out we rely on the global transitivity of this
operation because the unlock path observes the pointer with a
READ_ONCE(), not an smp_load_acquire().

This is non-critical because the MCS code isn't actually used and
mostly serves as documentation, a stepping stone to the more complex
things we've build on top of the idea.

Reported-by: Andrea Parri <parri.andrea@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: 3552a07a9c4a ("locking/mcs: Use acquire/release semantics")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/locking/mcs_spinlock.h