Re: [External] - Re: EVL Memory

[Philippe Gerum <rpm@xenomai.org>]

Philippe Gerum <rpm@xenomai.org> writes:

> Russell Johnson <russell.johnson@kratosdefense.com> writes:
>
>> [[S/MIME Signed Part:Undecided]]
>> I do already have that enabled and have confirmed the proper commit shows up
>> in uname. I will disable LOCKDEP and re-run.
>>
>> Thanks,
>>
>> Russell
>>
>> [[End of S/MIME Signed Part]]
>
> I realized that the line information from the kernel splats was matching
> the latest sources anyway, so you are definitely running the right
> kernel. Can you observe that the threads are indeed stuck with a boosted
> priority that never changes, using evl ps?


The patch below should capture a trace snapshot when the assertion
trips, with a backtrace of the most recently executed function calls
before this happens. This works with the "evl trace" command [1],

e.g.

# evl trace -e
... wait for kernel splats
...
# evl trace > tracelog.txt

[1] https://evlproject.org/core/commands/#evl-trace-command



diff --git a/kernel/evl/mutex.c b/kernel/evl/mutex.c
index 16750fd193377..c3c1576975c7e 100644
--- a/kernel/evl/mutex.c
+++ b/kernel/evl/mutex.c
@@ -653,6 +653,7 @@ void __evl_unlock_mutex(struct evl_mutex *mutex)
 	 */
 	if (IS_ENABLED(CONFIG_EVL_DEBUG_CORE)) {
 		bool bad;
+		trace_evl_trigger("BOOST1");
 		raw_spin_lock_irqsave(&curr->lock, flags);
 		bad = curr->state & EVL_T_BOOST && list_empty(&curr->boosters);
 		raw_spin_unlock_irqrestore(&curr->lock, flags);
diff --git a/kernel/evl/sched/core.c b/kernel/evl/sched/core.c
index 3bfbe37b1eb7b..d1307fb1b4fd5 100644
--- a/kernel/evl/sched/core.c
+++ b/kernel/evl/sched/core.c
@@ -985,6 +985,7 @@ void __evl_schedule(void) /* oob or/and hard irqs off (CPU migration-safe) */
 	 */
 	if (IS_ENABLED(CONFIG_EVL_DEBUG_CORE) &&
 		curr->state & EVL_T_BOOST && list_empty(&curr->boosters)) {
+		trace_evl_trigger("BOOST2");
 		raw_spin_unlock(&curr->lock);
 		EVL_WARN_ON_ONCE(CORE, 1);
 		raw_spin_lock(&curr->lock);

-- 
Philippe.