Juergen Hannken-Illjes <hannken@eis.cs.tu-bs.de> writes:

> Btw.: there seems to be a queue error:
> 
>     TAILQ_REMOVE(&prio->bq_write, prio->bq_next, b_actq);
>     /*
>      * Advance the write pointer.
>      */
>     prio->bq_write_next =
> 	TAILQ_NEXT(prio->bq_write_next, b_actq);
>     if (prio->bq_write_next == NULL)
> 	    prio->bq_write_next =
> 		TAILQ_FIRST(&prio->bq_write);
> 
> should become
> 
>     /*
>      * Advance the write pointer.
>      */
>     prio->bq_write_next =
> 	TAILQ_NEXT(prio->bq_write_next, b_actq);
>     if (prio->bq_write_next == NULL)
> 	    prio->bq_write_next =
> 		TAILQ_FIRST(&prio->bq_write);
>     TAILQ_REMOVE(&prio->bq_write, prio->bq_next, b_actq);
> 

The latter doesn't work if there is only one element in the queue.
The former *usually* works, since TAILQ_REMOVE() leaves tqe_next of
removed entry unchanged.

> 	<bufq_prio_get+93> is the instruction generating the fault.
> 
> I cannnot see a zero dereference here, confused ...

One possiblility is an interrupt (but the sequence in question in
wdstragety is executed with disk interrupt blocked...).

enami.