Yep Andre, basically just a load test. The following is a script I wrote to do the 'doctest'. It kills my ata system every time, usually by the 6'th loop. It has never made it to 10. (it fails with or without the hdparm, fyi)
-Doc (doc@lame.org)
#!/bin/sh
cd /tmp
umount /mnt/test1
umount /mnt/test2
umount /mnt/test3
umount /mnt/test4
/usr/src/hdparm-3.9/hdparm -c 1 -d 1 -m 16 -W 1 /dev/hde
/usr/src/hdparm-3.9/hdparm -c 1 -d 1 -m 16 -W 1 /dev/hdg
/usr/src/hdparm-3.9/hdparm -c 1 -d 1 -m 16 -W 1 /dev/hdi
/usr/src/hdparm-3.9/hdparm -c 1 -d 1 -m 16 -W 1 /dev/hdk
mke2fs -vm 0 /dev/hde &
mke2fs -vm 0 /dev/hdg &
mke2fs -vm 0 /dev/hdi &
mke2fs -vm 0 /dev/hdk
wait
wait
wait
mount -t ext2 /dev/hde /mnt/test1
mount -t ext2 /dev/hdg /mnt/test2
mount -t ext2 /dev/hdi /mnt/test3
mount -t ext2 /dev/hdk /mnt/test4
awk 'BEGIN {
x = 1
while(x<200) {
blah = sprintf("cat /usr/src/linux-2.4.0-test1.tar.bz2 >>/mnt/test1/blah")
system(blah)
x = x +1
}
}'
cp -v /mnt/test1/blah /mnt/test3/blah
cp -v /mnt/test1/blah /mnt/test4/blah &
cp -v /mnt/test3/blah /mnt/test2/blah
wait
COUNTER=1
while true; do
cp -v /mnt/test1/blah /mnt/test4/blah.${COUNTER} &
cp -v /mnt/test2/blah /mnt/test3/blah.${COUNTER} &
cp -v /mnt/test3/blah /mnt/test2/blah.${COUNTER} &
cp -v /mnt/test4/blah /mnt/test1/blah.${COUNTER}
wait
wait
wait
COUNTER=$((${COUNTER}+1))
echo Loop: ${COUNTER}
done
>
> On Thu, 15 Jun 2000, Christopher Zimmerman wrote:
>
> > Andre Hedrick wrote:
> >
> > > Which storage sub-system is in the torcher test again.....
> > >
> > > On Thu, 15 Jun 2000, Christopher Zimmerman wrote:
> > >
> > > > Hey,
> > > >
> > > > I just thought I should let every one know that I can still kill
> > > > ac17-18 by running 5 concurrent sessions of bonnie with a file size of
> > > > 2000 MB. However, I can know longer kill them by sicking 1000 threads
> > > > on a 10 GB file.
>
> Doc,
>
> This is the same kind of test that you are killing me with correct?
>
> > I've been testing on a scsi system and a system with an ide raid array using an
> > 8 port 3ware card.
>
> Is it a clean messageless deadlock?
>
> Andre Hedrick
> The Linux ATA/IDE guy
>
-
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 : Thu Jun 15 2000 - 21:00:36 EST