membarrier: Create Documentation/scheduler/membarrier.rst

To gather the architecture requirements of the "private/global
expedited" membarrier commands.  The file will be expanded to
integrate further information about the membarrier syscall (as
needed/desired in the future).  While at it, amend some related
inline comments in the membarrier codebase.

Suggested-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Andrea Parri <parri.andrea@gmail.com>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Link: https://lore.kernel.org/r/20240131144936.29190-3-parri.andrea@gmail.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

Documentation/scheduler/index.rst

@@ -7,6 +7,7 @@ Scheduler
 
 
     completion
+    membarrier
     sched-arch
     sched-bwc
     sched-deadline

Documentation/scheduler/membarrier.rst

@@ -0,0 +1,39 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+========================
+membarrier() System Call
+========================
+
+MEMBARRIER_CMD_{PRIVATE,GLOBAL}_EXPEDITED - Architecture requirements
+=====================================================================
+
+Memory barriers before updating rq->curr
+----------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after coming from
+user-space, before updating rq->curr.  This barrier is implied by the sequence
+rq_lock(); smp_mb__after_spinlock() in __schedule().  The barrier matches a full
+barrier in the proximity of the membarrier system call exit, cf.
+membarrier_{private,global}_expedited().
+
+Memory barriers after updating rq->curr
+---------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after updating rq->curr,
+before returning to user-space.  The schemes providing this barrier on the various
+architectures are as follows.
+
+ - alpha, arc, arm, hexagon, mips rely on the full barrier implied by
+   spin_unlock() in finish_lock_switch().
+
+ - arm64 relies on the full barrier implied by switch_to().
+
+ - powerpc, riscv, s390, sparc, x86 rely on the full barrier implied by
+   switch_mm(), if mm is not NULL; they rely on the full barrier implied
+   by mmdrop(), otherwise.  On powerpc and riscv, switch_mm() relies on
+   membarrier_arch_switch_mm().
+
+The barrier matches a full barrier in the proximity of the membarrier system call
+entry, cf. membarrier_{private,global}_expedited().

MAINTAINERS

@@ -14039,6 +14039,7 @@ M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 M:	"Paul E. McKenney" <paulmck@kernel.org>
 L:	linux-kernel@vger.kernel.org
 S:	Supported
+F:	Documentation/scheduler/membarrier.rst
 F:	arch/*/include/asm/membarrier.h
 F:	include/uapi/linux/membarrier.h
 F:	kernel/sched/membarrier.c

kernel/sched/core.c

@@ -6638,7 +6638,9 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 	 *     if (signal_pending_state())	    if (p->state & @state)
 	 *
 	 * Also, the membarrier system call requires a full memory barrier
-	 * after coming from user-space, before storing to rq->curr.
+	 * after coming from user-space, before storing to rq->curr; this
+	 * barrier matches a full barrier in the proximity of the membarrier
+	 * system call exit.
 	 */
 	rq_lock(rq, &rf);
 	smp_mb__after_spinlock();
@@ -6716,6 +6718,9 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 		 *   architectures where spin_unlock is a full barrier,
 		 * - switch_to() for arm64 (weakly-ordered, spin_unlock
 		 *   is a RELEASE barrier),
+		 *
+		 * The barrier matches a full barrier in the proximity of
+		 * the membarrier system call entry.
 		 */
 		++*switch_count;
 

kernel/sched/membarrier.c

@@ -251,7 +251,7 @@ static int membarrier_global_expedited(void)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
 	 */
 	smp_mb();	/* system call entry is not a mb. */
@@ -300,7 +300,7 @@ static int membarrier_global_expedited(void)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */
@@ -339,7 +339,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
 	 */
 	smp_mb();	/* system call entry is not a mb. */
@@ -415,7 +415,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */