Driver porting: wait queue and preemptible code

Driver porting: wait queue and preemptible code

Posted Dec 21, 2003 15:12 UTC (Sun) by nblanc (guest, #17996)
Parent article: Driver porting: sleeping and waking up

With kernel 2.4, you can assume that when you set the process
into TASK_INTERRUPTIBLE mod, there will be no scheduling
(until you decide to do it explicitly).

But what happens with kernel 2.6 in the case below.

for (;;)
{
add_wait_queue(&queue, &wait);
<------------------------------------------ Event occured
set_current_state(TASK_INTERRUPTIBLE);
if (condition) break;
schedule();
remove_wait_queue(&queue, &wait);
if (signal_pending(current))return -ERESTARTSYS;
}
<------------------------------------------- Schedule
set_current_state(TASK_RUNNING);

I think code like the previous one is no more safe with kernel 2.6.
Is it right?

Does the new way to do take care of this?

Nicolas

---

(Log in to post comments)