2005-11-26 21:07:17

by Paul E. McKenney

Subject: [PATCH] Make RCU task_struct safe for oprofile

Hello!

Applying RCU to the task structure broke oprofile, because
free_task_notify() can now be called from softirq. This means that the
task_mortuary lock must be acquired with irq disabled in order to avoid
intermittent self-deadlock. Since irq is now disabled, the critical
section within process_task_mortuary() has been restructured to be O(1) in
order to maximize scalability and minimize realtime latency degradation.

Kudos to Wu Fengguang for finding this problem!

CC: <[email protected]>
CC: <[email protected]>
Signed-off-by: <[email protected]>

---

buffer_sync.c | 30 +++++++++++++++---------------
1 files changed, 15 insertions(+), 15 deletions(-)

diff -urpNa -X dontdiff linux-2.6.14-mm2/drivers/oprofile/buffer_sync.c linux-2.6.14-mm2-fixmortuary/drivers/oprofile/buffer_sync.c
--- linux-2.6.14-mm2/drivers/oprofile/buffer_sync.c 2005-10-27 17:02:08.000000000 -0700
+++ linux-2.6.14-mm2-fixmortuary/drivers/oprofile/buffer_sync.c 2005-11-26 07:51:08.000000000 -0800
@@ -43,13 +43,16 @@ static void process_task_mortuary(void);
* list for processing. Only after two full buffer syncs
* does the task eventually get freed, because by then
* we are sure we will not reference it again.
+ * Can be invoked from softirq via RCU callback due to
+ * call_rcu() of the task struct, hence the _irqsave.
*/
static int task_free_notify(struct notifier_block * self, unsigned long val, void * data)
{
+ unsigned long flags;
struct task_struct * task = data;
- spin_lock(&task_mortuary);
+ spin_lock_irqsave(&task_mortuary, flags);
list_add(&task->tasks, &dying_tasks);
- spin_unlock(&task_mortuary);
+ spin_unlock_irqrestore(&task_mortuary, flags);
return NOTIFY_OK;
}

@@ -431,25 +434,22 @@ static void increment_tail(struct oprofi
*/
static void process_task_mortuary(void)
{
- struct list_head * pos;
- struct list_head * pos2;
+ unsigned long flags;
+ LIST_HEAD(local_dead_tasks);
struct task_struct * task;
+ struct task_struct * ttask;

- spin_lock(&task_mortuary);
+ spin_lock_irqsave(&task_mortuary, flags);

- list_for_each_safe(pos, pos2, &dead_tasks) {
- task = list_entry(pos, struct task_struct, tasks);
- list_del(&task->tasks);
- free_task(task);
- }
+ list_splice_init(&dead_tasks, &local_dead_tasks);
+ list_splice_init(&dying_tasks, &dead_tasks);

- list_for_each_safe(pos, pos2, &dying_tasks) {
- task = list_entry(pos, struct task_struct, tasks);
+ spin_unlock_irqrestore(&task_mortuary, flags);
+
+ list_for_each_entry_safe(task, ttask, &local_dead_tasks, tasks) {
list_del(&task->tasks);
- list_add_tail(&task->tasks, &dead_tasks);
+ free_task(task);
}
-
- spin_unlock(&task_mortuary);
}

2005-11-27 12:28:59

by Ingo Molnar

Subject: Re: [PATCH] Make RCU task_struct safe for oprofile

* Paul E. McKenney <[email protected]> wrote:

> Hello!
>
> Applying RCU to the task structure broke oprofile, because
> free_task_notify() can now be called from softirq. This means that
> the task_mortuary lock must be acquired with irq disabled in order to
> avoid intermittent self-deadlock. Since irq is now disabled, the
> critical section within process_task_mortuary() has been restructured
> to be O(1) in order to maximize scalability and minimize realtime
> latency degradation.
>
> Kudos to Wu Fengguang for finding this problem!

thanks, i applied this to the -rt tree.

Ingo