softlockup: remove timestamp checking from hung_task
Subsystems (4)
PATCH 1/1 softlockup: remove timestamp checking from hung_task
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes per task_struct
By guaranteeing that sysctl_hung_task_timeout_secs have elapsed between tasklist scans we can avoid using timestamps.
--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)
include/linux/sched.h +0 -1
@@ -1241,7 +1241,6 @@ struct task_struct {#endif #ifdef CONFIG_DETECT_HUNG_TASK /* hung task detection */ -unsigned long last_switch_timestamp;unsigned long last_switch_count;#endif /* CPU-specific state of this task */
kernel/fork.c +3 -5
@@ -639,6 +639,9 @@ static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) tsk->min_flt = tsk->maj_flt =0 ; tsk->nvcsw = tsk->nivcsw =0 ; +#ifdef CONFIG_DETECT_HUNG_TASK + tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; +#endif tsk->mm =NULL ; tsk->active_mm =NULL ;
@@ -1041,11 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->default_timer_slack_ns = current->timer_slack_ns; -#ifdef CONFIG_DETECT_HUNG_TASK - p->last_switch_count =0 ; - p->last_switch_timestamp =0 ; -#endif -task_io_accounting_init (&p->ioac);acct_clear_integrals (p);
kernel/hung_task.c +9 -39
110 lines not on this page · raw message
Signature · 1 line
1.5.4.5
Hi Mandeep,
Quoted 9 lines · Mandeep Singh Baines
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes pertask_structBy guaranteeing that
sysctl_hung_task_timeout_secshave elapsed between tasklist scans we can avoid using timestamps.Signed-off-by Mandeep Singh Baines <[email protected]>
Good idea.
BTW, why haven't you put your name on top of this file?
That would help those who will send patches knowing to whom they have to route their mails.
I made some comments below about small things...
Quoted 23 lines
--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)include/linux/sched.h +0
-1@@ -1241,7 +1241,6 @@ struct task_struct {#endif #ifdef CONFIG_DETECT_HUNG_TASK /* hung task detection */ -unsigned long last_switch_timestamp;unsigned long last_switch_count;#endif /* CPU-specific state of this task */ kernel/fork.c +0
-0
kernel/fork.c static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) · line 644
Quoted 6 lines
kernel/fork.c +3
-0@@ -639,6 +639,9 @@ static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) tsk->min_flt = tsk->maj_flt =0 ; tsk->nvcsw = tsk->nivcsw =0 ; +#ifdef CONFIG_DETECT_HUNG_TASK + tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; +#endif
I think you can directly assign a zero here :-)
Or you want to let it as is to give some sense and explanation about the role of this field?
Why not, I guess gcc will optimize it anyway.
Quoted 47 lines
kernel/fork.c +0
-5tsk->mm =NULL ; tsk->active_mm =NULL ;@@ -1041,11 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->default_timer_slack_ns = current->timer_slack_ns; -#ifdef CONFIG_DETECT_HUNG_TASK - p->last_switch_count =0 ; - p->last_switch_timestamp =0 ; -#endif -task_io_accounting_init (&p->ioac);acct_clear_integrals (p);kernel/hung_task.c +1
-15@@ -34,7 +34,6 @@ unsigned long __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;* Zero means infinite timeout - no checking done: */ unsigned long __read_mostly sysctl_hung_task_timeout_secs =120 ; -static unsigned long __read_mostly hung_task_poll_jiffies;unsigned long __read_mostly sysctl_hung_task_warnings =10 ;@@ -69,33 +68,17 @@ static struct notifier_block panic_block = { .notifier_call = hung_task_panic, }; -/* -* Returns seconds, approximately. We don't need nanosecond -* resolution, and we don't need to waste time with a big divide when -* 2^30ns == 1.074s. -*/ -static unsigned long get_timestamp (void ) -{ -int this_cpu =raw_smp_processor_id (); - -return cpu_clock (this_cpu) >>30LL ;/* 2^30 ~= 10^9 */ -} - -static void check_hung_task (struct task_struct *t,unsigned long now, -unsigned long timeout) +static void check_hung_task (struct task_struct *t,unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw;
kernel/hung_task.c static struct notifier_block panic_block = { · line 81
Quoted 10 lines
kernel/hung_task.c +1
-2if (t->flags &PF_FROZEN )return ; -if (switch_count != t->last_switch_count || !t->last_switch_timestamp) { +if (switch_count != t->last_switch_count) { t->last_switch_count = switch_count; - t->last_switch_timestamp = now;return ; }
What happens here if khungtaskd is scheduled in after tsk is inserted on the task_list in copy_process() but before tsk has been scheduled once?
tsk->last_switch_count and tsk->nvcsw + tsk->nivcsw will still be equal to zero right?
Perhaps you could add another check such as
if (!switch_count)
return;
Quoted 45 lines
kernel/hung_task.c +3
-10-if ((long )(now - t->last_switch_timestamp) < timeout) -return ;if (!sysctl_hung_task_warnings)return ; sysctl_hung_task_warnings--;@@ -111,7 +94,6 @@ static void check_hung_task(struct task_struct *t, unsigned long now,sched_show_task (t);__debug_show_held_locks (t); - t->last_switch_timestamp = now;touch_nmi_watchdog ();if (sysctl_hung_task_panic)@@ -145,7 +127,6 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) {int max_count = sysctl_hung_task_check_count;int batch_count =HUNG_TASK_BATCHING ; -unsigned long now =get_timestamp ();struct task_struct *g, *t;/* @@ -168,19 +149,16 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) }/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */ if (t->state ==TASK_UNINTERRUPTIBLE ) -check_hung_task (t, now, timeout); +check_hung_task (t, timeout); }while_each_thread (g, t); unlock:rcu_read_unlock (); } -static void update_poll_jiffies (void ) +static unsigned long timeout_jiffies (unsigned long timeout) {/* timeout of 0 will disable the watchdog */ -if (sysctl_hung_task_timeout_secs ==0 ) - hung_task_poll_jiffies =MAX_SCHEDULE_TIMEOUT ; -else - hung_task_poll_jiffies = sysctl_hung_task_timeout_secs *HZ /2 ; +return (timeout ? timeout *HZ :MAX_SCHEDULE_TIMEOUT ); }/*
kernel/hung_task.c int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, · line 178
Quoted 6 lines
kernel/hung_task.c +0
-2@@ -197,8 +175,6 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,if (ret || !write) goto out; -update_poll_jiffies (); -wake_up_process (watchdog_task);
I'm not sure what does this function now that you dropped update_poll_jiffies() So if the user sets up a new timeout value, the only effect will be that khungtaskd will be awakened?
But actually the /sys file doesn't seem to be set up.
Other than these comments, that looks good!
Thanks.
Frederic.
Quoted 30 lines
kernel/hung_task.c +4
-10out:@@ -211,20 +187,14 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,static int watchdog (void *dummy) {set_user_nice (current,0 ); -update_poll_jiffies ();for ( ; ; ) { -unsigned long timeout; +unsigned long timeout = sysctl_hung_task_timeout_secs; -while (schedule_timeout_interruptible (hung_task_poll_jiffies)); +while (schedule_timeout_interruptible (timeout_jiffies (timeout))) + timeout = sysctl_hung_task_timeout_secs; -/* -* Need to cache timeout here to avoid timeout being set -* to 0 via sysctl while inside check_hung_*_tasks(). -*/ - timeout = sysctl_hung_task_timeout_secs; -if (timeout) -check_hung_uninterruptible_tasks (timeout); +check_hung_uninterruptible_tasks (timeout); }return 0 ;Signature · 1 line
1.5.4.5
Quoted 186 lines · Frederic Weisbecker
Hi Mandeep,
Quoted 9 lines · Mandeep Singh Baines
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes pertask_structBy guaranteeing that
sysctl_hung_task_timeout_secshave elapsed between tasklist scans we can avoid using timestamps.Signed-off-by Mandeep Singh Baines <[email protected]>Good idea.
BTW, why haven't you put your name on top of this file?
That would help those who will send patches knowing to whom they have to route their mails.I made some comments below about small things...
Quoted 23 lines
--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)include/linux/sched.h +0
-1@@ -1241,7 +1241,6 @@ struct task_struct {#endif #ifdef CONFIG_DETECT_HUNG_TASK /* hung task detection */ -unsigned long last_switch_timestamp;unsigned long last_switch_count;#endif /* CPU-specific state of this task */ kernel/fork.c +0
-0kernel/fork.c static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) · line 644
Quoted 6 lines
kernel/fork.c +3
-0@@ -639,6 +639,9 @@ static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) tsk->min_flt = tsk->maj_flt =0 ; tsk->nvcsw = tsk->nivcsw =0 ; +#ifdef CONFIG_DETECT_HUNG_TASK + tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; +#endif I think you can directly assign a zero here :-)
Or you want to let it as is to give some sense and explanation about the role of this field?
Why not, I guess gcc will optimize it anyway.Quoted 47 lines
kernel/fork.c +0
-5tsk->mm =NULL ; tsk->active_mm =NULL ;@@ -1041,11 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->default_timer_slack_ns = current->timer_slack_ns; -#ifdef CONFIG_DETECT_HUNG_TASK - p->last_switch_count =0 ; - p->last_switch_timestamp =0 ; -#endif -task_io_accounting_init (&p->ioac);acct_clear_integrals (p);kernel/hung_task.c +1
-15@@ -34,7 +34,6 @@ unsigned long __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;* Zero means infinite timeout - no checking done: */ unsigned long __read_mostly sysctl_hung_task_timeout_secs =120 ; -static unsigned long __read_mostly hung_task_poll_jiffies;unsigned long __read_mostly sysctl_hung_task_warnings =10 ;@@ -69,33 +68,17 @@ static struct notifier_block panic_block = { .notifier_call = hung_task_panic, }; -/* -* Returns seconds, approximately. We don't need nanosecond -* resolution, and we don't need to waste time with a big divide when -* 2^30ns == 1.074s. -*/ -static unsigned long get_timestamp (void ) -{ -int this_cpu =raw_smp_processor_id (); - -return cpu_clock (this_cpu) >>30LL ;/* 2^30 ~= 10^9 */ -} - -static void check_hung_task (struct task_struct *t,unsigned long now, -unsigned long timeout) +static void check_hung_task (struct task_struct *t,unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw;kernel/hung_task.c static struct notifier_block panic_block = { · line 81
Quoted 10 lines
kernel/hung_task.c +1
-2if (t->flags &PF_FROZEN )return ; -if (switch_count != t->last_switch_count || !t->last_switch_timestamp) { +if (switch_count != t->last_switch_count) { t->last_switch_count = switch_count; - t->last_switch_timestamp = now;return ; }What happens here if khungtaskd is scheduled in after tsk is inserted on the
task_listincopy_process()but before tsk has been scheduled once?
tsk->last_switch_countandtsk->nvcsw+tsk->nivcswwill still be equal to zero right?Perhaps you could add another check such as
if (!switch_count)
return;Quoted 45 lines
kernel/hung_task.c +3
-10-if ((long )(now - t->last_switch_timestamp) < timeout) -return ;if (!sysctl_hung_task_warnings)return ; sysctl_hung_task_warnings--;@@ -111,7 +94,6 @@ static void check_hung_task(struct task_struct *t, unsigned long now,sched_show_task (t);__debug_show_held_locks (t); - t->last_switch_timestamp = now;touch_nmi_watchdog ();if (sysctl_hung_task_panic)@@ -145,7 +127,6 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) {int max_count = sysctl_hung_task_check_count;int batch_count =HUNG_TASK_BATCHING ; -unsigned long now =get_timestamp ();struct task_struct *g, *t;/* @@ -168,19 +149,16 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) }/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */ if (t->state ==TASK_UNINTERRUPTIBLE ) -check_hung_task (t, now, timeout); +check_hung_task (t, timeout); }while_each_thread (g, t); unlock:rcu_read_unlock (); } -static void update_poll_jiffies (void ) +static unsigned long timeout_jiffies (unsigned long timeout) {/* timeout of 0 will disable the watchdog */ -if (sysctl_hung_task_timeout_secs ==0 ) - hung_task_poll_jiffies =MAX_SCHEDULE_TIMEOUT ; -else - hung_task_poll_jiffies = sysctl_hung_task_timeout_secs *HZ /2 ; +return (timeout ? timeout *HZ :MAX_SCHEDULE_TIMEOUT ); }/* kernel/hung_task.c int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, · line 178
Quoted 6 lines
kernel/hung_task.c +0
-2@@ -197,8 +175,6 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,if (ret || !write) goto out; -update_poll_jiffies (); -wake_up_process (watchdog_task);I'm not sure what does this function now that you dropped
update_poll_jiffies()So if the user sets up a new timeout value, the only effect will be that khungtaskd will be awakened?
Quoted 1 line
But actually the /sys file doesn't seem to be set up.
Oops, I should have grep on proc_dohung_task_timeout_secs which is set on kernel/sysctl. Sorry.
But still, sysctl_hung_task_timeout_secs doesn't seem to be set :-)
Quoted 37 lines
Other than these comments, that looks good!
Thanks.Frederic.
Quoted 30 lines
kernel/hung_task.c +4
-10out:@@ -211,20 +187,14 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,static int watchdog (void *dummy) {set_user_nice (current,0 ); -update_poll_jiffies ();for ( ; ; ) { -unsigned long timeout; +unsigned long timeout = sysctl_hung_task_timeout_secs; -while (schedule_timeout_interruptible (hung_task_poll_jiffies)); +while (schedule_timeout_interruptible (timeout_jiffies (timeout))) + timeout = sysctl_hung_task_timeout_secs; -/* -* Need to cache timeout here to avoid timeout being set -* to 0 via sysctl while inside check_hung_*_tasks(). -*/ - timeout = sysctl_hung_task_timeout_secs; -if (timeout) -check_hung_uninterruptible_tasks (timeout); +check_hung_uninterruptible_tasks (timeout); }return 0 ;Signature · 1 line
1.5.4.5
Quoted 187 lines · Frederic Weisbecker
Quoted 186 lines · Frederic Weisbecker
Hi Mandeep,
Quoted 9 lines · Mandeep Singh Baines
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes pertask_structBy guaranteeing that
sysctl_hung_task_timeout_secshave elapsed between tasklist scans we can avoid using timestamps.Signed-off-by Mandeep Singh Baines <[email protected]>Good idea.
BTW, why haven't you put your name on top of this file?
That would help those who will send patches knowing to whom they have to route their mails.I made some comments below about small things...
Quoted 23 lines
--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)include/linux/sched.h +0
-1@@ -1241,7 +1241,6 @@ struct task_struct {#endif #ifdef CONFIG_DETECT_HUNG_TASK /* hung task detection */ -unsigned long last_switch_timestamp;unsigned long last_switch_count;#endif /* CPU-specific state of this task */ kernel/fork.c +0
-0kernel/fork.c static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) · line 644
Quoted 6 lines
kernel/fork.c +3
-0@@ -639,6 +639,9 @@ static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) tsk->min_flt = tsk->maj_flt =0 ; tsk->nvcsw = tsk->nivcsw =0 ; +#ifdef CONFIG_DETECT_HUNG_TASK + tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; +#endif I think you can directly assign a zero here :-)
Or you want to let it as is to give some sense and explanation about the role of this field?
Why not, I guess gcc will optimize it anyway.Quoted 47 lines
kernel/fork.c +0
-5tsk->mm =NULL ; tsk->active_mm =NULL ;@@ -1041,11 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->default_timer_slack_ns = current->timer_slack_ns; -#ifdef CONFIG_DETECT_HUNG_TASK - p->last_switch_count =0 ; - p->last_switch_timestamp =0 ; -#endif -task_io_accounting_init (&p->ioac);acct_clear_integrals (p);kernel/hung_task.c +1
-15@@ -34,7 +34,6 @@ unsigned long __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;* Zero means infinite timeout - no checking done: */ unsigned long __read_mostly sysctl_hung_task_timeout_secs =120 ; -static unsigned long __read_mostly hung_task_poll_jiffies;unsigned long __read_mostly sysctl_hung_task_warnings =10 ;@@ -69,33 +68,17 @@ static struct notifier_block panic_block = { .notifier_call = hung_task_panic, }; -/* -* Returns seconds, approximately. We don't need nanosecond -* resolution, and we don't need to waste time with a big divide when -* 2^30ns == 1.074s. -*/ -static unsigned long get_timestamp (void ) -{ -int this_cpu =raw_smp_processor_id (); - -return cpu_clock (this_cpu) >>30LL ;/* 2^30 ~= 10^9 */ -} - -static void check_hung_task (struct task_struct *t,unsigned long now, -unsigned long timeout) +static void check_hung_task (struct task_struct *t,unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw;kernel/hung_task.c static struct notifier_block panic_block = { · line 81
Quoted 10 lines
kernel/hung_task.c +1
-2if (t->flags &PF_FROZEN )return ; -if (switch_count != t->last_switch_count || !t->last_switch_timestamp) { +if (switch_count != t->last_switch_count) { t->last_switch_count = switch_count; - t->last_switch_timestamp = now;return ; }What happens here if khungtaskd is scheduled in after tsk is inserted on the
task_listincopy_process()but before tsk has been scheduled once?
tsk->last_switch_countandtsk->nvcsw+tsk->nivcswwill still be equal to zero right?Perhaps you could add another check such as
if (!switch_count)
return;Quoted 45 lines
kernel/hung_task.c +3
-10-if ((long )(now - t->last_switch_timestamp) < timeout) -return ;if (!sysctl_hung_task_warnings)return ; sysctl_hung_task_warnings--;@@ -111,7 +94,6 @@ static void check_hung_task(struct task_struct *t, unsigned long now,sched_show_task (t);__debug_show_held_locks (t); - t->last_switch_timestamp = now;touch_nmi_watchdog ();if (sysctl_hung_task_panic)@@ -145,7 +127,6 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) {int max_count = sysctl_hung_task_check_count;int batch_count =HUNG_TASK_BATCHING ; -unsigned long now =get_timestamp ();struct task_struct *g, *t;/* @@ -168,19 +149,16 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) }/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */ if (t->state ==TASK_UNINTERRUPTIBLE ) -check_hung_task (t, now, timeout); +check_hung_task (t, timeout); }while_each_thread (g, t); unlock:rcu_read_unlock (); } -static void update_poll_jiffies (void ) +static unsigned long timeout_jiffies (unsigned long timeout) {/* timeout of 0 will disable the watchdog */ -if (sysctl_hung_task_timeout_secs ==0 ) - hung_task_poll_jiffies =MAX_SCHEDULE_TIMEOUT ; -else - hung_task_poll_jiffies = sysctl_hung_task_timeout_secs *HZ /2 ; +return (timeout ? timeout *HZ :MAX_SCHEDULE_TIMEOUT ); }/* kernel/hung_task.c int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, · line 178
Quoted 6 lines
kernel/hung_task.c +0
-2@@ -197,8 +175,6 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,if (ret || !write) goto out; -update_poll_jiffies (); -wake_up_process (watchdog_task);I'm not sure what does this function now that you dropped
update_poll_jiffies()So if the user sets up a new timeout value, the only effect will be that khungtaskd will be awakened?
Quoted 6 lines
Quoted 1 line
But actually the /sys file doesn't seem to be set up.
Oops, I should have grep on
proc_dohung_task_timeout_secswhich is set on kernel/sysctl. Sorry.
But still,sysctl_hung_task_timeout_secsdoesn't seem to be set :-)
And once again I'm wrong, I shoud read sysctl.c or at least read the sysfs documentation. Sorry for the noise.
BTW, here is a small fixlet on top of your patch about what I commented concerning the tasks than weren't yet scheduled once:
--
Quoted 1 line
From
e7120e424b031978e482b5fe311d90916ffb8b7eMon Sep 17 00:00:00 2001
From: Frederic Weisbecker <[email protected]>
Date: Sat, 7 Feb 2009 17:45:12 +0100
Subject: [PATCH] softlockup: ensure the task has been scheduled once
When we check if the task has been scheduled since the last scan, we might have a race condition if the task has been inserted on the task list but not yet scheduled once. So we just add a small check to ensure it has been switched in at least one time to avoid false positive.
--- kernel/hung_task.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)
kernel/hung_task.c +5 -1
@@ -72,7 +72,11 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw; -if (t->flags &PF_FROZEN ) +/* +* Ensure the task is not frozen and that it has been scheduled +* at least once. +*/ +if (t->flags &PF_FROZEN || !switch_count)return ;if (switch_count != t->last_switch_count) {
Signature · 1 line
1.6.1
Quoted 196 lines · Frederic Weisbecker
Quoted 187 lines · Frederic Weisbecker
Quoted 186 lines · Frederic Weisbecker
Hi Mandeep,
Quoted 9 lines · Mandeep Singh Baines
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes pertask_structBy guaranteeing that
sysctl_hung_task_timeout_secshave elapsed between tasklist scans we can avoid using timestamps.Signed-off-by Mandeep Singh Baines <[email protected]>Good idea.
BTW, why haven't you put your name on top of this file?
That would help those who will send patches knowing to whom they have to route their mails.I made some comments below about small things...
Quoted 23 lines
--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)include/linux/sched.h +0
-1@@ -1241,7 +1241,6 @@ struct task_struct {#endif #ifdef CONFIG_DETECT_HUNG_TASK /* hung task detection */ -unsigned long last_switch_timestamp;unsigned long last_switch_count;#endif /* CPU-specific state of this task */ kernel/fork.c +0
-0kernel/fork.c static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) · line 644
Quoted 6 lines
kernel/fork.c +3
-0@@ -639,6 +639,9 @@ static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) tsk->min_flt = tsk->maj_flt =0 ; tsk->nvcsw = tsk->nivcsw =0 ; +#ifdef CONFIG_DETECT_HUNG_TASK + tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; +#endif I think you can directly assign a zero here :-)
Or you want to let it as is to give some sense and explanation about the role of this field?
Why not, I guess gcc will optimize it anyway.Quoted 47 lines
kernel/fork.c +0
-5tsk->mm =NULL ; tsk->active_mm =NULL ;@@ -1041,11 +1044,6 @@ static struct task_struct *copy_process(unsigned long clone_flags, p->default_timer_slack_ns = current->timer_slack_ns; -#ifdef CONFIG_DETECT_HUNG_TASK - p->last_switch_count =0 ; - p->last_switch_timestamp =0 ; -#endif -task_io_accounting_init (&p->ioac);acct_clear_integrals (p);kernel/hung_task.c +1
-15@@ -34,7 +34,6 @@ unsigned long __read_mostly sysctl_hung_task_check_count = PID_MAX_LIMIT;* Zero means infinite timeout - no checking done: */ unsigned long __read_mostly sysctl_hung_task_timeout_secs =120 ; -static unsigned long __read_mostly hung_task_poll_jiffies;unsigned long __read_mostly sysctl_hung_task_warnings =10 ;@@ -69,33 +68,17 @@ static struct notifier_block panic_block = { .notifier_call = hung_task_panic, }; -/* -* Returns seconds, approximately. We don't need nanosecond -* resolution, and we don't need to waste time with a big divide when -* 2^30ns == 1.074s. -*/ -static unsigned long get_timestamp (void ) -{ -int this_cpu =raw_smp_processor_id (); - -return cpu_clock (this_cpu) >>30LL ;/* 2^30 ~= 10^9 */ -} - -static void check_hung_task (struct task_struct *t,unsigned long now, -unsigned long timeout) +static void check_hung_task (struct task_struct *t,unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw;kernel/hung_task.c static struct notifier_block panic_block = { · line 81
Quoted 10 lines
kernel/hung_task.c +1
-2if (t->flags &PF_FROZEN )return ; -if (switch_count != t->last_switch_count || !t->last_switch_timestamp) { +if (switch_count != t->last_switch_count) { t->last_switch_count = switch_count; - t->last_switch_timestamp = now;return ; }What happens here if khungtaskd is scheduled in after tsk is inserted on the
task_listincopy_process()but before tsk has been scheduled once?
tsk->last_switch_countandtsk->nvcsw+tsk->nivcswwill still be equal to zero right?Perhaps you could add another check such as
if (!switch_count)
return;Quoted 45 lines
kernel/hung_task.c +3
-10-if ((long )(now - t->last_switch_timestamp) < timeout) -return ;if (!sysctl_hung_task_warnings)return ; sysctl_hung_task_warnings--;@@ -111,7 +94,6 @@ static void check_hung_task(struct task_struct *t, unsigned long now,sched_show_task (t);__debug_show_held_locks (t); - t->last_switch_timestamp = now;touch_nmi_watchdog ();if (sysctl_hung_task_panic)@@ -145,7 +127,6 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) {int max_count = sysctl_hung_task_check_count;int batch_count =HUNG_TASK_BATCHING ; -unsigned long now =get_timestamp ();struct task_struct *g, *t;/* @@ -168,19 +149,16 @@ static void check_hung_uninterruptible_tasks(unsigned long timeout) }/* use "==" to skip the TASK_KILLABLE tasks waiting on NFS */ if (t->state ==TASK_UNINTERRUPTIBLE ) -check_hung_task (t, now, timeout); +check_hung_task (t, timeout); }while_each_thread (g, t); unlock:rcu_read_unlock (); } -static void update_poll_jiffies (void ) +static unsigned long timeout_jiffies (unsigned long timeout) {/* timeout of 0 will disable the watchdog */ -if (sysctl_hung_task_timeout_secs ==0 ) - hung_task_poll_jiffies =MAX_SCHEDULE_TIMEOUT ; -else - hung_task_poll_jiffies = sysctl_hung_task_timeout_secs *HZ /2 ; +return (timeout ? timeout *HZ :MAX_SCHEDULE_TIMEOUT ); }/* kernel/hung_task.c int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, · line 178
Quoted 6 lines
kernel/hung_task.c +0
-2@@ -197,8 +175,6 @@ int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,if (ret || !write) goto out; -update_poll_jiffies (); -wake_up_process (watchdog_task);I'm not sure what does this function now that you dropped
update_poll_jiffies()So if the user sets up a new timeout value, the only effect will be that khungtaskd will be awakened?Quoted 6 lines
Quoted 1 line
But actually the /sys file doesn't seem to be set up.
Oops, I should have grep on
proc_dohung_task_timeout_secswhich is set on kernel/sysctl. Sorry.
But still,sysctl_hung_task_timeout_secsdoesn't seem to be set :-)
Quoted 5 lines
And once again I'm wrong, I shoud read
sysctl.cor at least read the sysfs documentation. Sorry for the noise.BTW, here is a small fixlet on top of your patch about what I commented concerning the tasks than weren't yet scheduled once:
But it is not supposed to be TASK_UNINTERRUPTIBLE. Forget this fixlet.
So, again sorry for the noise, I'm going to hide somewhere...
Quoted 36 lines
--
Frome7120e424b031978e482b5fe311d90916ffb8b7eMon Sep 17 00:00:00 2001
From: Frederic Weisbecker <[email protected]>
Date: Sat, 7 Feb 2009 17:45:12 +0100
Subject: [PATCH] softlockup: ensure the task has been scheduled onceWhen we check if the task has been scheduled since the last scan, we might have a race condition if the task has been inserted on the task list but not yet scheduled once. So we just add a small check to ensure it has been switched in at least one time to avoid false positive.
Signed-off-by Frederic Weisbecker <[email protected]>--- kernel/hung_task.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)kernel/hung_task.c +5
-1@@ -72,7 +72,11 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw; -if (t->flags &PF_FROZEN ) +/* +* Ensure the task is not frozen and that it has been scheduled +* at least once. +*/ +if (t->flags &PF_FROZEN || !switch_count)return ;if (switch_count != t->last_switch_count) {Signature · 1 line
1.6.1
Ingo Molnar
AppliedQuoted 8 lines · Mandeep Singh Baines
Patch against tip/core/softlockup
---
Impact: saves sizeof(long) bytes pertask_structBy guaranteeing that
sysctl_hung_task_timeout_secshave elapsed between tasklist scans we can avoid using timestamps.
Quoted 6 lines
Signed-off-by Mandeep Singh Baines <[email protected]>--- include/linux/sched.h | 1 - kernel/fork.c | 8 +++----- kernel/hung_task.c | 48 +++++++++--------------------------------------- 3 files changed, 12 insertions(+), 45 deletions(-)
Nice simplification!
Applied to tip/core/softlockup, thanks Mandeep.
(i fixed the one scripts/checkpatch.pl error that your patch triggered)
Ingo
Quoted 39 lines · Frederic Weisbecker
BTW, here is a small fixlet on top of your patch about what I commented concerning the tasks than weren't yet scheduled once:
--
Frome7120e424b031978e482b5fe311d90916ffb8b7eMon Sep 17 00:00:00 2001
From: Frederic Weisbecker <[email protected]>
Date: Sat, 7 Feb 2009 17:45:12 +0100
Subject: [PATCH] softlockup: ensure the task has been scheduled onceWhen we check if the task has been scheduled since the last scan, we might have a race condition if the task has been inserted on the task list but not yet scheduled once. So we just add a small check to ensure it has been switched in at least one time to avoid false positive.
Signed-off-by Frederic Weisbecker <[email protected]>--- kernel/hung_task.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)kernel/hung_task.c +5
-1@@ -72,7 +72,11 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw; -if (t->flags &PF_FROZEN ) +/* +* Ensure the task is not frozen and that it has been scheduled +* at least once. +*/ +if (t->flags &PF_FROZEN || !switch_count)return ;if (switch_count != t->last_switch_count) {Signature · 1 line
1.6.1
Good catch! I would change the description though. The race is a little more subtle than your current description. For a newly forked process, the race can occur if check_hung_task() processes the task in the time between the task changing its state to UNINTERRUPTIBLE and the the scheduler updating the switch_count (nivcsw or nvcsw).
One minor change to the code comment. You want to ensure that the task has context switched at least once, NOT that the task has been scheduled at least once.
If you re-send the patch with the fixed comment and description, I'll ack it.
Quoted 47 lines · Mandeep Singh Baines
Quoted 39 lines · Frederic Weisbecker
BTW, here is a small fixlet on top of your patch about what I commented concerning the tasks than weren't yet scheduled once:
--
Frome7120e424b031978e482b5fe311d90916ffb8b7eMon Sep 17 00:00:00 2001
From: Frederic Weisbecker <[email protected]>
Date: Sat, 7 Feb 2009 17:45:12 +0100
Subject: [PATCH] softlockup: ensure the task has been scheduled onceWhen we check if the task has been scheduled since the last scan, we might have a race condition if the task has been inserted on the task list but not yet scheduled once. So we just add a small check to ensure it has been switched in at least one time to avoid false positive.
Signed-off-by Frederic Weisbecker <[email protected]>--- kernel/hung_task.c | 6 +++++- 1 files changed, 5 insertions(+), 1 deletions(-)kernel/hung_task.c +5
-1@@ -72,7 +72,11 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout) {unsigned long switch_count = t->nvcsw + t->nivcsw; -if (t->flags &PF_FROZEN ) +/* +* Ensure the task is not frozen and that it has been scheduled +* at least once. +*/ +if (t->flags &PF_FROZEN || !switch_count)return ;if (switch_count != t->last_switch_count) {Signature · 1 line
1.6.1Good catch! I would change the description though. The race is a little more subtle than your current description. For a newly forked process, the race can occur if
check_hung_task()processes the task in the time between the task changing its state to UNINTERRUPTIBLE and the the scheduler updating theswitch_count(nivcsw or nvcsw).
Quoted 3 lines
One minor change to the code comment. You want to ensure that the task has context switched at least once, NOT that the task has been scheduled at least once.
Ooh ok. I thought that the first schedule updated nivcsw/nvcsw, that's why I cancelled my comment.
Ok so I will resend the patch with the appropriate comment.
Quoted 3 lines
If you re-send the patch with the fixed comment and description, I'll ack it.
Thanks! :-)