linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] nohz: fix stale jiffies update in tick_nohz_restart()
@ 2012-03-27 19:09 Neal Cardwell
  2012-04-06 11:34 ` [tip:timers/urgent] nohz: Fix " tip-bot for Neal Cardwell
  0 siblings, 1 reply; 2+ messages in thread
From: Neal Cardwell @ 2012-03-27 19:09 UTC (permalink / raw)
  To: linux-kernel; +Cc: Neal Cardwell, Ben Segall, Ingo Molnar, Thomas Gleixner

Fix tick_nohz_restart() to not use a stale ktime_t "now" value when
calling tick_do_update_jiffies64(now).

If we reach this point in the loop it means that we crossed a tick
boundary since we grabbed the "now" timestamp, so at this point "now"
refers to a time in the old jiffy, so using the old value for "now" is
incorrect, and is likely to give us a stale jiffies value.

In particular, the first time through the loop the
tick_do_update_jiffies64(now) call is always a no-op, since the
caller, tick_nohz_restart_sched_tick(), will have already called
tick_do_update_jiffies64(now) with that "now" value.

Note that tick_nohz_stop_sched_tick() already uses the correct
approach: when we notice we cross a jiffy boundary, grab a new
timestamp with ktime_get(), and *then* update jiffies.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
CC: Ingo Molnar <mingo@elte.hu>
CC: Thomas Gleixner <tglx@linutronix.de>
---
 kernel/time/tick-sched.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 3526038..6a3a5b9 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -534,9 +534,9 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
 				hrtimer_get_expires(&ts->sched_timer), 0))
 				break;
 		}
-		/* Update jiffies and reread time */
-		tick_do_update_jiffies64(now);
+		/* Reread time and update jiffies */
 		now = ktime_get();
+		tick_do_update_jiffies64(now);
 	}
 }
 
-- 
1.7.7.3


^ permalink raw reply related	[flat|nested] 2+ messages in thread

* [tip:timers/urgent] nohz: Fix stale jiffies update in tick_nohz_restart()
  2012-03-27 19:09 [PATCH] nohz: fix stale jiffies update in tick_nohz_restart() Neal Cardwell
@ 2012-04-06 11:34 ` tip-bot for Neal Cardwell
  0 siblings, 0 replies; 2+ messages in thread
From: tip-bot for Neal Cardwell @ 2012-04-06 11:34 UTC (permalink / raw)
  To: linux-tip-commits
  Cc: linux-kernel, bsegall, ncardwell, hpa, mingo, tglx, mingo

Commit-ID:  6f103929f8979d2638e58d7f7fda0beefcb8ee7e
Gitweb:     http://git.kernel.org/tip/6f103929f8979d2638e58d7f7fda0beefcb8ee7e
Author:     Neal Cardwell <ncardwell@google.com>
AuthorDate: Tue, 27 Mar 2012 15:09:37 -0400
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Fri, 6 Apr 2012 13:24:17 +0200

nohz: Fix stale jiffies update in tick_nohz_restart()

Fix tick_nohz_restart() to not use a stale ktime_t "now" value when
calling tick_do_update_jiffies64(now).

If we reach this point in the loop it means that we crossed a tick
boundary since we grabbed the "now" timestamp, so at this point "now"
refers to a time in the old jiffy, so using the old value for "now" is
incorrect, and is likely to give us a stale jiffies value.

In particular, the first time through the loop the
tick_do_update_jiffies64(now) call is always a no-op, since the
caller, tick_nohz_restart_sched_tick(), will have already called
tick_do_update_jiffies64(now) with that "now" value.

Note that tick_nohz_stop_sched_tick() already uses the correct
approach: when we notice we cross a jiffy boundary, grab a new
timestamp with ktime_get(), and *then* update jiffies.

Signed-off-by: Neal Cardwell <ncardwell@google.com>
Cc: Ben Segall <bsegall@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/1332875377-23014-1-git-send-email-ncardwell@google.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 kernel/time/tick-sched.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 3526038..6a3a5b9 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -534,9 +534,9 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
 				hrtimer_get_expires(&ts->sched_timer), 0))
 				break;
 		}
-		/* Update jiffies and reread time */
-		tick_do_update_jiffies64(now);
+		/* Reread time and update jiffies */
 		now = ktime_get();
+		tick_do_update_jiffies64(now);
 	}
 }
 

^ permalink raw reply related	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2012-04-06 11:35 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-03-27 19:09 [PATCH] nohz: fix stale jiffies update in tick_nohz_restart() Neal Cardwell
2012-04-06 11:34 ` [tip:timers/urgent] nohz: Fix " tip-bot for Neal Cardwell

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).