Mark Hahn wrote:
>
> > if you can tolerate occasional few-millisecond delays, the plain
> > kernel may suffice.
>
> this is much too harsh; with the user-level app running RT, and no stupid
> misconfigurations (IDE in PIO, other RT tasks), there won't be "occasional"
> long delays.
Alas, the console code blocks interrupts for extended periods. Up to 3
msec on a 400 MHz CPU. See
http://www.uow.edu.au/~andrewm/linux/26-mar-00-ann3.txt
I believe that the console scrolling is the worst problem. If I recall
aright, the above figures didn't include a lot of scrolling. High
levels of console output have been correlated with transmit underruns
and receive overruns on high-performance network interfaces for this
reason. Also correlated with audio playback hiccups.
The bad IDE numbers are real. Enabling interrupt unmasking (hdparm -u
1) fixes this.
And while I'm being rude about the console driver, the kill_proc()
within complete_change_console() with console_lock held is a deadlock
waiting to happen. kill_proc() does a _lot_ of stuff, including calling
schedule(). If anything at all does a printk() it'll wedge...
-- -akpm-- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon May 15 2000 - 21:00:17 EST