RT and pi_test

RT and pi_test

[Daniel Walker]

	I've run the pi_test a few times recently with some strange 
numbers. I'm getting abnormally high task latency if I run 
"./test --tasks=10 ./hist" . Depending on how long I run it, I've seen 
task latency as high as 3 milliseconds, should be 0.1ms or less.

The first time I saw these latencies I has the PI abstraction applied, and 
the most recent time I had the attached patch applied only. This patch is 
small so I'm not sure if it could have that type of effect on task 
latency.

btw, 'm not incrementing the RTC priority.

Daniel


Index: linux-2.6.11/include/linux/plist.h
===================================================================
--- linux-2.6.11.orig/include/linux/plist.h	2005-05-27 22:04:12.000000000 +0000
+++ linux-2.6.11/include/linux/plist.h	2005-06-01 13:12:37.000000000 +0000
@@ -83,7 +83,7 @@ struct plist {
  * @member:     the name of the list_struct within the struct.
  */
 #define plist_entry(ptr, type, member) \
-        container_of(plist_first(ptr), type, member)
+        container_of(ptr, type, member)
 /**
  * plist_for_each  -       iterate over the plist
  * @pos1:        the type * to use as a loop counter.
Index: linux-2.6.11/kernel/rt.c
===================================================================
--- linux-2.6.11.orig/kernel/rt.c	2005-06-01 13:06:21.000000000 +0000
+++ linux-2.6.11/kernel/rt.c	2005-06-01 13:50:08.000000000 +0000
@@ -773,7 +773,7 @@ static inline struct task_struct * pick_
 	 *
 	 * (same-prio RT tasks go FIFO)
 	 */
-	waiter = plist_entry(&lock->wait_list, struct rt_mutex_waiter, list);
+	waiter = plist_entry(plist_first(&lock->wait_list), struct rt_mutex_waiter, list);
 
 	trace_special_pid(waiter->task->pid, waiter->task->prio, 0);
 
@@ -1351,7 +1351,7 @@ static void __up_mutex(struct rt_mutex *
 	 */
 	prio = mutex_getprio(old_owner);
 	if (!plist_empty(&old_owner->pi_waiters)) {
-		w = plist_entry(&old_owner->pi_waiters, struct rt_mutex_waiter, pi_list);
+		w = plist_entry(plist_first(&old_owner->pi_waiters), struct rt_mutex_waiter, pi_list);
 		if (w->task->prio < prio)
 			prio = w->task->prio;
 	}

Re: RT and pi_test

[Ingo Molnar]

* Daniel Walker <dwalker@mvista.com> wrote:

> 	I've run the pi_test a few times recently with some strange 
> numbers. I'm getting abnormally high task latency if I run "./test 
> --tasks=10 ./hist" . Depending on how long I run it, I've seen task 
> latency as high as 3 milliseconds, should be 0.1ms or less.
> 
> The first time I saw these latencies I has the PI abstraction applied, 
> and the most recent time I had the attached patch applied only. This 
> patch is small so I'm not sure if it could have that type of effect on 
> task latency.

this patch now clashes with Thomas' changes. Did you mean me to apply 
this patch, or is it now moot? (The way i read your mail originally what 
that it was describing a problem with latencies, and including a patch 
as the only thing that you had applied. It didnt too inviting to apply
:-| )

	Ingo