18 Sep
2007
18 Sep
'07
2:40 p.m.
At Tue, 18 Sep 2007 13:54:23 +0200, Rene Herman wrote:
Hi.
While looking through ALSA for schedule_timeout() calls, I ran into:
int snd_mixart_send_msg(...) {
[ ... ]
set_current_state(TASK_UNINTERRUPTIBLE); add_wait_queue(&mgr->msg_sleep, &wait); spin_unlock_irq(&mgr->msg_lock); timeout = schedule_timeout(MSG_TIMEOUT_JIFFIES); remove_wait_queue(&mgr->msg_sleep, &wait); if (! timeout) { /* error - no ack */ mutex_unlock(&mgr->msg_mutex); snd_printk(KERN_ERR "error: no reponse on msg %x\n",
msg_frame); return -EIO; }
[ ... ]
}
and the same in snd_mixart_send_msg_wait_notif().
I believe there to be something wrong with this code. A schedule_timeout() in TASK_UNINTERRUPTIBLE is always going to return 0.
No, it returns the time left at the point it's woken up. *_uninterruptible() ignores signals but not wake_up() itself (otherwise it makes no sense). The description in the kernel comment is misleading indeed.
Takashi