On Sun, 17 Dec 2000, Petr Vandrovec wrote:
> my 2.4.0-test13-pre1 just stopped answering to my keystrokes.
> I've found that it is looping in tqueue_bh and flush_to_ldisc
> still again and again.
Thanks, I think you found the big problem with test12.
> Is current run_task_queue() behavior intentional, or is it
> only bug introduced by changing task queue to list based
> implementation?
It's a bug.
Ho humm. I'll have to check what the proper fix is. Right now the rule is
that nobody can _ever_ remove himself from a task queue, although there is
one bad user that does exactly that, and that means that it should be ok
to just cache the "next" pointer in run_task_queue(), and make it look
something like
static void run_task_queue(task_queue *list)
{
if (!list_empty(list)) {
unsigned long flags;
struct list_head *next;
spin_lock_irqsave(&tqueue_lock, flags);
next = list->next;
while (next != list) {
struct list_head *curr;
void *arg;
void (*f) (void *);
struct tq_struct *p;
curr = next;
next = next->next;
list_del(curr);
p = list_entry(curr, struct tq_struct, list);
arg = p->data;
f = p->routine;
p->sync = 0;
spin_unlock_irqrestore(&tqueue_lock, flags);
if (f)
f(arg);
spin_lock_irq(&tqueue_lock);
}
spin_unlock_irqrestore(&tqueue_lock, flags);
}
}
which basically will never re-start at the head if a tq re-inserts iself
(side note: new entires are inserted at the end, but if it was already the
last entry then re-inserting it will not re-execute it, should this should
avoid your problem).
Does this fix the problem for you?
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Dec 23 2000 - 21:00:18 EST