Fix: Disable sys_membarrier when nohz_full is enabled
authorMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Thu, 3 Nov 2016 16:29:28 +0000 (10:29 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 12 Mar 2017 05:37:26 +0000 (06:37 +0100)
commit 907565337ebf998a68cb5c5b2174ce5e5da065eb upstream.

Userspace applications should be allowed to expect the membarrier system
call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
nohz_full CPUs, but synchronize_sched() does not take those into
account.

Given that we do not want unrelated processes to be able to affect
real-time sensitive nohz_full CPUs, simply return ENOSYS when membarrier
is invoked on a kernel with enabled nohz_full CPUs.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
CC: Josh Triplett <josh@joshtriplett.org>
CC: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Chris Metcalf <cmetcalf@mellanox.com>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: Lai Jiangshan <jiangshanlai@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
kernel/membarrier.c

index 536c727a56e975e2e310fb3fb82474f07a1b916f..9f9284f37f8d8b6d3149ecc22634c613123e792d 100644 (file)
@@ -16,6 +16,7 @@
 
 #include <linux/syscalls.h>
 #include <linux/membarrier.h>
+#include <linux/tick.h>
 
 /*
  * Bitmask made from a "or" of all commands within enum membarrier_cmd,
@@ -51,6 +52,9 @@
  */
 SYSCALL_DEFINE2(membarrier, int, cmd, int, flags)
 {
+       /* MEMBARRIER_CMD_SHARED is not compatible with nohz_full. */
+       if (tick_nohz_full_enabled())
+               return -ENOSYS;
        if (unlikely(flags))
                return -EINVAL;
        switch (cmd) {