doc: Update memory-barriers.txt for read-to-write dependencies

The memory-barriers.txt document contains an obsolete passage stating that
smp_read_barrier_depends() is required to force ordering for read-to-write
dependencies.  We now know that this is not required, even for DEC Alpha.
This commit therefore updates this passage to state that read-to-write
dependencies are respected even without smp_read_barrier_depends().

Reported-by: Lance Roy <ldr709@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Andrea Parri <parri.andrea@gmail.com>
Cc: Jade Alglave <j.alglave@ucl.ac.uk>
Cc: Luc Maranget <luc.maranget@inria.fr>
[ paulmck: Reference control-dependencies sections and use WRITE_ONCE()
  per Will Deacon.  Correctly place split-cache paragraph while there. ]
Acked-by: Will Deacon <will.deacon@arm.com>
This commit is contained in:
Paul E. McKenney 2017-06-30 16:18:28 -07:00
parent 4de5f89ef8
commit 66ce3a4dcb
1 changed files with 30 additions and 23 deletions

View File

@ -594,29 +594,6 @@ between the address load and the data load:
This enforces the occurrence of one of the two implications, and prevents the
third possibility from arising.
A data-dependency barrier must also order against dependent writes:
CPU 1 CPU 2
=============== ===============
{ A == 1, B == 2, C = 3, P == &A, Q == &C }
B = 4;
<write barrier>
WRITE_ONCE(P, &B);
Q = READ_ONCE(P);
<data dependency barrier>
*Q = 5;
The data-dependency barrier must order the read into Q with the store
into *Q. This prohibits this outcome:
(Q == &B) && (B == 4)
Please note that this pattern should be rare. After all, the whole point
of dependency ordering is to -prevent- writes to the data structure, along
with the expensive cache misses associated with those writes. This pattern
can be used to record rare error conditions and the like, and the ordering
prevents such records from being lost.
[!] Note that this extremely counterintuitive situation arises most easily on
machines with split caches, so that, for example, one cache bank processes
@ -628,6 +605,36 @@ odd-numbered bank is idle, one can see the new value of the pointer P (&B),
but the old value of the variable B (2).
A data-dependency barrier is not required to order dependent writes
because the CPUs that the Linux kernel supports don't do writes
until they are certain (1) that the write will actually happen, (2)
of the location of the write, and (3) of the value to be written.
But please carefully read the "CONTROL DEPENDENCIES" section and the
Documentation/RCU/rcu_dereference.txt file: The compiler can and does
break dependencies in a great many highly creative ways.
CPU 1 CPU 2
=============== ===============
{ A == 1, B == 2, C = 3, P == &A, Q == &C }
B = 4;
<write barrier>
WRITE_ONCE(P, &B);
Q = READ_ONCE(P);
WRITE_ONCE(*Q, 5);
Therefore, no data-dependency barrier is required to order the read into
Q with the store into *Q. In other words, this outcome is prohibited,
even without a data-dependency barrier:
(Q == &B) && (B == 4)
Please note that this pattern should be rare. After all, the whole point
of dependency ordering is to -prevent- writes to the data structure, along
with the expensive cache misses associated with those writes. This pattern
can be used to record rare error conditions and the like, and the CPUs'
naturally occurring ordering prevents such records from being lost.
The data dependency barrier is very important to the RCU system,
for example. See rcu_assign_pointer() and rcu_dereference() in
include/linux/rcupdate.h. This permits the current target of an RCU'd