[RFC PATCH]nohz: Use raw_smp_processor_id() in tick_nohz_task_switch()

[RFC PATCH]nohz: Use raw_smp_processor_id() in tick_nohz_task_switch()

[Li Zhong]

I saw following error when testing the latest nohz code on Power:

[   85.295384] BUG: using smp_processor_id() in preemptible [00000000] code: rsyslogd/3493
[   85.295396] caller is .tick_nohz_task_switch+0x1c/0xb8
[   85.295402] Call Trace:
[   85.295408] [c0000001fababab0] [c000000000012dc4] .show_stack+0x110/0x25c (unreliable)
[   85.295420] [c0000001fababba0] [c0000000007c4b54] .dump_stack+0x20/0x30
[   85.295430] [c0000001fababc10] [c00000000044eb74] .debug_smp_processor_id+0xf4/0x124
[   85.295438] [c0000001fababca0] [c0000000000d7594] .tick_nohz_task_switch+0x1c/0xb8
[   85.295447] [c0000001fababd20] [c0000000000b9748] .finish_task_switch+0x13c/0x160
[   85.295455] [c0000001fababdb0] [c0000000000bbe50] .schedule_tail+0x50/0x124
[   85.295463] [c0000001fababe30] [c000000000009dc8] .ret_from_fork+0x4/0x54

It seems to me that we could just use raw_smp_processor_id() here. Even
if the tick_nohz_full_cpu() check is done on a !nohz_full cpu, then the
task is moved to another nohz_full cpu, it seems the context switching
because of the task moving would call tick_nohz_task_switch() again to
evaluate the need for tick.

I don't know whether I missed something here.

Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
---
 kernel/time/tick-sched.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index da53c8f..0aa575b 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -251,7 +251,7 @@ void tick_nohz_task_switch(struct task_struct *tsk)
 {
 	unsigned long flags;
 
-	if (!tick_nohz_full_cpu(smp_processor_id()))
+	if (!tick_nohz_full_cpu(raw_smp_processor_id()))
 		return;
 
 	local_irq_save(flags);
-- 
1.7.1

Re: [RFC PATCH]nohz: Use raw_smp_processor_id() in tick_nohz_task_switch()

[Frederic Weisbecker]

2013/4/27 Li Zhong <zhong@linux.vnet.ibm.com>:
> I saw following error when testing the latest nohz code on Power:
>
> [   85.295384] BUG: using smp_processor_id() in preemptible [00000000] code: rsyslogd/3493
> [   85.295396] caller is .tick_nohz_task_switch+0x1c/0xb8
> [   85.295402] Call Trace:
> [   85.295408] [c0000001fababab0] [c000000000012dc4] .show_stack+0x110/0x25c (unreliable)
> [   85.295420] [c0000001fababba0] [c0000000007c4b54] .dump_stack+0x20/0x30
> [   85.295430] [c0000001fababc10] [c00000000044eb74] .debug_smp_processor_id+0xf4/0x124
> [   85.295438] [c0000001fababca0] [c0000000000d7594] .tick_nohz_task_switch+0x1c/0xb8
> [   85.295447] [c0000001fababd20] [c0000000000b9748] .finish_task_switch+0x13c/0x160
> [   85.295455] [c0000001fababdb0] [c0000000000bbe50] .schedule_tail+0x50/0x124
> [   85.295463] [c0000001fababe30] [c000000000009dc8] .ret_from_fork+0x4/0x54
>
> It seems to me that we could just use raw_smp_processor_id() here. Even
> if the tick_nohz_full_cpu() check is done on a !nohz_full cpu, then the
> task is moved to another nohz_full cpu, it seems the context switching
> because of the task moving would call tick_nohz_task_switch() again to
> evaluate the need for tick.
>
> I don't know whether I missed something here.

You're right it looks safe to do so. But I suggest we rather move the
test inside local_irq_save()/restore section to avoid confusion on
reviewers minds.

Thanks!

Re: [RFC PATCH]nohz: Use raw_smp_processor_id() in tick_nohz_task_switch()

[Li Zhong]

On Sat, 2013-04-27 at 15:40 +0200, Frederic Weisbecker wrote:
> 2013/4/27 Li Zhong <zhong@linux.vnet.ibm.com>:
> > I saw following error when testing the latest nohz code on Power:
> >
> > [   85.295384] BUG: using smp_processor_id() in preemptible [00000000] code: rsyslogd/3493
> > [   85.295396] caller is .tick_nohz_task_switch+0x1c/0xb8
> > [   85.295402] Call Trace:
> > [   85.295408] [c0000001fababab0] [c000000000012dc4] .show_stack+0x110/0x25c (unreliable)
> > [   85.295420] [c0000001fababba0] [c0000000007c4b54] .dump_stack+0x20/0x30
> > [   85.295430] [c0000001fababc10] [c00000000044eb74] .debug_smp_processor_id+0xf4/0x124
> > [   85.295438] [c0000001fababca0] [c0000000000d7594] .tick_nohz_task_switch+0x1c/0xb8
> > [   85.295447] [c0000001fababd20] [c0000000000b9748] .finish_task_switch+0x13c/0x160
> > [   85.295455] [c0000001fababdb0] [c0000000000bbe50] .schedule_tail+0x50/0x124
> > [   85.295463] [c0000001fababe30] [c000000000009dc8] .ret_from_fork+0x4/0x54
> >
> > It seems to me that we could just use raw_smp_processor_id() here. Even
> > if the tick_nohz_full_cpu() check is done on a !nohz_full cpu, then the
> > task is moved to another nohz_full cpu, it seems the context switching
> > because of the task moving would call tick_nohz_task_switch() again to
> > evaluate the need for tick.
> >
> > I don't know whether I missed something here.
> 
> You're right it looks safe to do so. But I suggest we rather move the
> test inside local_irq_save()/restore section to avoid confusion on
> reviewers minds.

OK, I'll send an updated version, using local_irq_save() to protect it. 
I tried using raw_* because seems it could avoid some unnecessary irq
disabling...

Thanks, Zhong

> Thanks!
> 

[RFC PATCH v2]nohz: protect smp_processor_id() in tick_nohz_task_switch()

[Li Zhong]

I saw following error when testing the latest nohz code on Power:

[   85.295384] BUG: using smp_processor_id() in preemptible [00000000] code: rsyslogd/3493
[   85.295396] caller is .tick_nohz_task_switch+0x1c/0xb8
[   85.295402] Call Trace:
[   85.295408] [c0000001fababab0] [c000000000012dc4] .show_stack+0x110/0x25c (unreliable)
[   85.295420] [c0000001fababba0] [c0000000007c4b54] .dump_stack+0x20/0x30
[   85.295430] [c0000001fababc10] [c00000000044eb74] .debug_smp_processor_id+0xf4/0x124
[   85.295438] [c0000001fababca0] [c0000000000d7594] .tick_nohz_task_switch+0x1c/0xb8
[   85.295447] [c0000001fababd20] [c0000000000b9748] .finish_task_switch+0x13c/0x160
[   85.295455] [c0000001fababdb0] [c0000000000bbe50] .schedule_tail+0x50/0x124
[   85.295463] [c0000001fababe30] [c000000000009dc8] .ret_from_fork+0x4/0x54

The code below moves the test into local_irq_save/restore section to
avoid the above complaint.

Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
---
 kernel/time/tick-sched.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index da53c8f..1c9f53b 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -251,14 +251,15 @@ void tick_nohz_task_switch(struct task_struct *tsk)
 {
 	unsigned long flags;
 
-	if (!tick_nohz_full_cpu(smp_processor_id()))
-		return;
-
 	local_irq_save(flags);
 
+	if (!tick_nohz_full_cpu(smp_processor_id()))
+		goto out;
+
 	if (tick_nohz_tick_stopped() && !can_stop_full_tick())
 		tick_nohz_full_kick();
 
+out:
 	local_irq_restore(flags);
 }
 
-- 
1.7.1

[tip:timers/nohz] nohz: Protect smp_processor_id() in tick_nohz_task_switch()

[tip-bot for Li Zhong]

Commit-ID:  6296ace467c8640317414ba589b124323806f7ce
Gitweb:     http://git.kernel.org/tip/6296ace467c8640317414ba589b124323806f7ce
Author:     Li Zhong <zhong@linux.vnet.ibm.com>
AuthorDate: Sun, 28 Apr 2013 11:25:58 +0800
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Mon, 29 Apr 2013 13:17:33 +0200

nohz: Protect smp_processor_id() in tick_nohz_task_switch()

I saw following error when testing the latest nohz code on
Power:

[   85.295384] BUG: using smp_processor_id() in preemptible [00000000] code: rsyslogd/3493
[   85.295396] caller is .tick_nohz_task_switch+0x1c/0xb8
[   85.295402] Call Trace:
[   85.295408] [c0000001fababab0] [c000000000012dc4] .show_stack+0x110/0x25c (unreliable)
[   85.295420] [c0000001fababba0] [c0000000007c4b54] .dump_stack+0x20/0x30
[   85.295430] [c0000001fababc10] [c00000000044eb74] .debug_smp_processor_id+0xf4/0x124
[   85.295438] [c0000001fababca0] [c0000000000d7594] .tick_nohz_task_switch+0x1c/0xb8
[   85.295447] [c0000001fababd20] [c0000000000b9748] .finish_task_switch+0x13c/0x160
[   85.295455] [c0000001fababdb0] [c0000000000bbe50] .schedule_tail+0x50/0x124
[   85.295463] [c0000001fababe30] [c000000000009dc8] .ret_from_fork+0x4/0x54

The code below moves the test into local_irq_save/restore
section to avoid the above complaint.

Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1367119558.6391.34.camel@ThinkPad-T5421.cn.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/time/tick-sched.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index da53c8f..1c9f53b 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -251,14 +251,15 @@ void tick_nohz_task_switch(struct task_struct *tsk)
 {
 	unsigned long flags;
 
-	if (!tick_nohz_full_cpu(smp_processor_id()))
-		return;
-
 	local_irq_save(flags);
 
+	if (!tick_nohz_full_cpu(smp_processor_id()))
+		goto out;
+
 	if (tick_nohz_tick_stopped() && !can_stop_full_tick())
 		tick_nohz_full_kick();
 
+out:
 	local_irq_restore(flags);
 }