rcu: Decrease memory-barrier usage based on semi-formal proof

Commit d09b62d fixed grace-period synchronization, but left some smp_mb()
invocations in rcu_process_callbacks() that are no longer needed, but
sheer paranoia prevented them from being removed.  This commit removes
them and provides a proof of correctness in their absence.  It also adds
a memory barrier to rcu_report_qs_rsp() immediately before the update to
rsp->completed in order to handle the theoretical possibility that the
compiler or CPU might move massive quantities of code into a lock-based
critical section.  This also proves that the sheer paranoia was not
entirely unjustified, at least from a theoretical point of view.

In addition, the old dyntick-idle synchronization depended on the fact
that grace periods were many milliseconds in duration, so that it could
be assumed that no dyntick-idle CPU could reorder a memory reference
across an entire grace period.  Unfortunately for this design, the
addition of expedited grace periods breaks this assumption, which has
the unfortunate side-effect of requiring atomic operations in the
functions that track dyntick-idle state for RCU.  (There is some hope
that the algorithms used in user-level RCU might be applied here, but
some work is required to handle the NMIs that user-space applications
can happily ignore.  For the short term, better safe than sorry.)

This proof assumes that neither compiler nor CPU will allow a lock
acquisition and release to be reordered, as doing so can result in
deadlock.  The proof is as follows:

1.	A given CPU declares a quiescent state under the protection of
	its leaf rcu_node's lock.

2.	If there is more than one level of rcu_node hierarchy, the
	last CPU to declare a quiescent state will also acquire the
	->lock of the next rcu_node up in the hierarchy,  but only
	after releasing the lower level's lock.  The acquisition of this
	lock clearly cannot occur prior to the acquisition of the leaf
	node's lock.

3.	Step 2 repeats until we reach the root rcu_node structure.
	Please note again that only one lock is held at a time through
	this process.  The acquisition of the root rcu_node's ->lock
	must occur after the release of that of the leaf rcu_node.

4.	At this point, we set the ->completed field in the rcu_state
	structure in rcu_report_qs_rsp().  However, if the rcu_node
	hierarchy contains only one rcu_node, then in theory the code
	preceding the quiescent state could leak into the critical
	section.  We therefore precede the update of ->completed with a
	memory barrier.  All CPUs will therefore agree that any updates
	preceding any report of a quiescent state will have happened
	before the update of ->completed.

5.	Regardless of whether a new grace period is needed, rcu_start_gp()
	will propagate the new value of ->completed to all of the leaf
	rcu_node structures, under the protection of each rcu_node's ->lock.
	If a new grace period is needed immediately, this propagation
	will occur in the same critical section that ->completed was
	set in, but courtesy of the memory barrier in #4 above, is still
	seen to follow any pre-quiescent-state activity.

6.	When a given CPU invokes __rcu_process_gp_end(), it becomes
	aware of the end of the old grace period and therefore makes
	any RCU callbacks that were waiting on that grace period eligible
	for invocation.

	If this CPU is the same one that detected the end of the grace
	period, and if there is but a single rcu_node in the hierarchy,
	we will still be in the single critical section.  In this case,
	the memory barrier in step #4 guarantees that all callbacks will
	be seen to execute after each CPU's quiescent state.

	On the other hand, if this is a different CPU, it will acquire
	the leaf rcu_node's ->lock, and will again be serialized after
	each CPU's quiescent state for the old grace period.

On the strength of this proof, this commit therefore removes the memory
barriers from rcu_process_callbacks() and adds one to rcu_report_qs_rsp().
The effect is to reduce the number of memory barriers by one and to
reduce the frequency of execution from about once per scheduling tick
per CPU to once per grace period.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>

1 files changed, 20 insertions, 28 deletions

diff --git a/Documentation/RCU/trace.txt b/Documentation/RCU/trace.txt
index 6a8c73f..e731ad2 100644
--- a/Documentation/RCU/trace.txt
+++ b/Documentation/RCU/trace.txt
@@ -21,23 +21,23 @@ rcu_pending() function decided that there was core RCU work to do).
 The output of "cat rcu/rcudata" looks as follows:
 
 rcu_sched:
-  0 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=10951/1 dn=0 df=1101 of=0 ri=36 ql=0 b=10
-  1 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=16117/1 dn=0 df=1015 of=0 ri=0 ql=0 b=10
-  2 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=1445/1 dn=0 df=1839 of=0 ri=0 ql=0 b=10
-  3 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=6681/1 dn=0 df=1545 of=0 ri=0 ql=0 b=10
-  4 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=1003/1 dn=0 df=1992 of=0 ri=0 ql=0 b=10
-  5 c=17829 g=17830 pq=1 pqc=17829 qp=1 dt=3887/1 dn=0 df=3331 of=0 ri=4 ql=2 b=10
-  6 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=859/1 dn=0 df=3224 of=0 ri=0 ql=0 b=10
-  7 c=17829 g=17830 pq=0 pqc=17829 qp=1 dt=3761/1 dn=0 df=1818 of=0 ri=0 ql=2 b=10
+  0 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=10951/1/0 df=1101 of=0 ri=36 ql=0 b=10
+  1 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=16117/1/0 df=1015 of=0 ri=0 ql=0 b=10
+  2 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=1445/1/0 df=1839 of=0 ri=0 ql=0 b=10
+  3 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=6681/1/0 df=1545 of=0 ri=0 ql=0 b=10
+  4 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=1003/1/0 df=1992 of=0 ri=0 ql=0 b=10
+  5 c=17829 g=17830 pq=1 pqc=17829 qp=1 dt=3887/1/0 df=3331 of=0 ri=4 ql=2 b=10
+  6 c=17829 g=17829 pq=1 pqc=17829 qp=0 dt=859/1/0 df=3224 of=0 ri=0 ql=0 b=10
+  7 c=17829 g=17830 pq=0 pqc=17829 qp=1 dt=3761/1/0 df=1818 of=0 ri=0 ql=2 b=10
 rcu_bh:
-  0 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=10951/1 dn=0 df=0 of=0 ri=0 ql=0 b=10
-  1 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=16117/1 dn=0 df=13 of=0 ri=0 ql=0 b=10
-  2 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=1445/1 dn=0 df=15 of=0 ri=0 ql=0 b=10
-  3 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=6681/1 dn=0 df=9 of=0 ri=0 ql=0 b=10
-  4 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=1003/1 dn=0 df=15 of=0 ri=0 ql=0 b=10
-  5 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=3887/1 dn=0 df=15 of=0 ri=0 ql=0 b=10
-  6 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=859/1 dn=0 df=15 of=0 ri=0 ql=0 b=10
-  7 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=3761/1 dn=0 df=15 of=0 ri=0 ql=0 b=10
+  0 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=10951/1/0 df=0 of=0 ri=0 ql=0 b=10
+  1 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=16117/1/0 df=13 of=0 ri=0 ql=0 b=10
+  2 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=1445/1/0 df=15 of=0 ri=0 ql=0 b=10
+  3 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=6681/1/0 df=9 of=0 ri=0 ql=0 b=10
+  4 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=1003/1/0 df=15 of=0 ri=0 ql=0 b=10
+  5 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=3887/1/0 df=15 of=0 ri=0 ql=0 b=10
+  6 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=859/1/0 df=15 of=0 ri=0 ql=0 b=10
+  7 c=-275 g=-275 pq=1 pqc=-275 qp=0 dt=3761/1/0 df=15 of=0 ri=0 ql=0 b=10
 
 The first section lists the rcu_data structures for rcu_sched, the second
 for rcu_bh.  Note that CONFIG_TREE_PREEMPT_RCU kernels will have an
@@ -85,18 +85,10 @@ o	"qp" indicates that RCU still expects a quiescent state from
 
 o	"dt" is the current value of the dyntick counter that is incremented
 	when entering or leaving dynticks idle state, either by the
-	scheduler or by irq.  The number after the "/" is the interrupt
-	nesting depth when in dyntick-idle state, or one greater than
-	the interrupt-nesting depth otherwise.
-
-	This field is displayed only for CONFIG_NO_HZ kernels.
-
-o	"dn" is the current value of the dyntick counter that is incremented
-	when entering or leaving dynticks idle state via NMI.  If both
-	the "dt" and "dn" values are even, then this CPU is in dynticks
-	idle mode and may be ignored by RCU.  If either of these two
-	counters is odd, then RCU must be alert to the possibility of
-	an RCU read-side critical section running on this CPU.
+	scheduler or by irq.  The number after the first "/" is the
+	interrupt nesting depth when in dyntick-idle state, or one
+	greater than the interrupt-nesting depth otherwise.  The number
+	after the second "/" is the NMI nesting depth.
 
 	This field is displayed only for CONFIG_NO_HZ kernels.