Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1284344

[PATCH tip/core/rcu 0/2] Enforce RCU grace-period transitivity

From "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Newsgroups linux.kernel
Subject [PATCH tip/core/rcu 0/2] Enforce RCU grace-period transitivity
Date 2015-12-05 00:50 +0100
Message-ID <qC8gV-Xi-7@gated-at.bofh.it> (permalink)
Organization linux.* mail to news gateway

Show all headers | View raw


Hello!

Within RCU, acquisitions of the rcu_node structure's ->lock field must
be followed by smp_mb__after_unlock_lock() in order to enforce RCU's
grace-period memory-ordering guarantees.  In theory, acquisitions for
debug purposes need not have smp_mb__after_unlock_lock(), but in practice,
anything other than fully consistent enforcement of this rule is a recipe
for disaster.  This series therefore adds smp_mb__after_unlock_lock()
for consistency:

1.	Create wrapper functions to minimize the number of these
	smp_mb__after_unlock_lock() calls, and apply these wrapper
	functions throughout.  Courtesy of Peter Zijlstra.

2.	Catch up wrapper-function application to patches that were
	in flight at the time #1 was created.

							Thanx, Paul

------------------------------------------------------------------------

 b/kernel/rcu/tree.c        |  102 ++++++++++++++++-----------------------------
 b/kernel/rcu/tree.h        |   39 +++++++++++++++++
 b/kernel/rcu/tree_plugin.h |   20 +++-----
 b/kernel/rcu/tree_trace.c  |    2 
 4 files changed, 85 insertions(+), 78 deletions(-)

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Back to linux.kernel | Previous | Next | Find similar | Unroll thread


Thread

[PATCH tip/core/rcu 0/2] Enforce RCU grace-period transitivity "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> - 2015-12-05 00:50 +0100

csiph-web