From: Thomas Gleixner on
Stefan,

please use short and descriptive subject lines. "I attached a test
module to reproduce this bug ..." does not tell anything.

On Mon, 12 Apr 2010, Stefan Agner wrote:

> Unable to handle kernel paging request at kthread_should_stop

That would have been the real subject line :)

> I attached a test module to reproduce this bug. Does I miss something in my
> code or is there a bug?

There are several bugs in your code :)

> static int runloop(void *unused)
> {
>
> int ret;
>
> /* Daemonize */
> daemonize(MODULE_NAME);

Do not call daemonize() for a kthread created with kthread_create().
That was necessary for kthreads created with kernel_thread.

> /* Allow SIGTERM signal */
> allow_signal(SIGTERM);

Why do you want signals here ?

> /* Calculate first shot */
> now = ktime_get();
> thread_next = ktime_add_us(now, 500);
>
> do {
> set_current_state(TASK_UNINTERRUPTIBLE);
> ret = schedule_hrtimeout(&thread_next, HRTIMER_MODE_ABS);
> if(ret == -EINTR)
> {

That EINTR can never happen. You set the task to
TASK_UNINTERRUPTIBLE state, so it cannot be woken up by a signal.

> int __init init_test_module(void)
> {
> struct sched_param param = { .sched_priority = THREAD_PRIORITY };
>
> /* Kernel Thread erzeugen */
> if(thread == NULL)
> thread = kthread_create((void *)&runloop, NULL, "testthread");

The init code is called once, so no need to check for thread == NULL
before calling kthread_create().

But you miss to check thread _after_ the call ....

> // Realtime scheduler
> sched_setscheduler(thread, THREAD_SCHEDULER, &param);

Scheduler settings should happen in the thread.

> wake_up_process(thread);

Simply use kthread_run(....), which implies kthread_create() and
wake_up_process().

Thanks,

tglx
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/