Re: fbcon + scrolling = irq timeouts?

George (greerga@nidhogg.ham.muohio.edu)
Tue, 30 Nov 1999 10:18:19 -0500 (EST)


On Mon, 29 Nov 1999, Benno Senoner wrote:

>before claiming that this is a BIOS issue do the following:
>
>test this on a kernel with the low-latency patch applied with tuned EIDE
>disks, AND running the mp3 player at higher priority (SCHED_FIFO
>recommended) than the cat command.

I don't have the low-latency patch currently but I do have this on 2.2.13:

IBM IDE 7200RPM 512kB cache hard drive (U)DMA with:
/sbin/hdparm -c1 -d1 -u1 -X34 -W1 -k1 -K1 -S0 /dev/hd[ac]

hdparm benches it at:

/dev/hdc:
Timing buffer-cache reads: 128 MB in 4.39 seconds =29.16 MB/sec
Timing buffered disk reads: 64 MB in 6.25 seconds =10.24 MB/sec

My test consisted of opening a 14 megabyte mailbox in Pine
(/var/log/messages in my case, and it was cached).

11:56:41.122638 read(4, "Feb 17 00:02:03 moving syslogd 1"..., 14152466) = 14152466 <0.601153>

There wasn't any screen activity at the time. It happened before the
little spinning /-\| character.

I was running it as:

$ cat mail/foo > /dev/null;strace -tt -T -o x `which pine` -f foo -I q

total used free shared buffers cached
Mem: 128128 109796 18332 45072 2700 65456
-/+ buffers/cache: 41640 86488
Swap: 130532 23824 106708

In fact, this cheesy little program causes IRQ timeouts as well:

root@moving:~# cat read.c
#include <fcntl.h>

#define ALLOC 24*1024*1024
int main(void)
{
char *buf = (char *)malloc(ALLOC);
int fd;

fd = open("/var/log/messages", O_RDONLY);
read(fd, buf, ALLOC);
close(fd);
return 0;
}
root@moving:~# ls -l /var/log/messages
-rw-r--r-- 1 root root 14242408 Nov 30 12:11 /var/log/messages

Running this program 5 times in rapid succession just killed my
SoundBlaster. Now I get an endless stream of "Sound: DMA (output) timed out
- IRQ/DRQ config error?"

Nov 30 12:10:31 moving kernel: Sound: DMA (output) timed out - IRQ/DRQ config error?
Nov 30 12:11:02 moving last message repeated 61 times
Nov 30 12:12:03 moving last message repeated 122 times

Whee! Fortunately, a good kick in the pants (by rmmod/insmod) got the
sound working again.

-George Greer

-
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/