0. kernel 2-4.18-3
1. Is there a better way to sleep than sleep_on_timeout()?
2. What's the best way to achieve something like poll() (or select())
on a kernel socket.
Right now, I'm doing something like the following, which I cobbled together
by
combining sunrpc's _recvfrom function and khttpd's use of sleep_on_timeout.
for (;;) {
if (stop_cond)
break;
/* How could I exit if I blocked? */
msg = socket->ops->recvmsg(...NOBLOCK...);
process_msg();
/* sleep */
current->state = TASK_INTERRUPTIBLE;
(void) interruptible_sleep_on_timeout(&wq, HZ);
}
wake_up(&stop_wq);
Problem is, someone else is segv'ing while trying to read my thread's /proc
entry.
It's crashing in fs/proc/array.c, within a task_lock() block. This leads me
to
suspect I'm doing something I shouldn't be.
The thread is conceptually pretty simple. The killer thread sets stop_cond,
places itself on the stop_wq, and calls sys_wait4() to cleanup the deadwood.
That seems straightforward enough.
I've seen some dialog on the kernel list about sleep_on() being something to
avoid
if you can. But I don't see wait_queue_timeout() time of routine. Hence my
two Q's
above. Is there a better way to sleep than sleep_on_timeout() that I could
use as
a drop-in replacement? Or, is there a better way to set up the whole read
loop?
--
Gary Cote
gcote@transat-tech.com
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
