IDE corruption during heavy bt878-induced interrupt load
Quoted 4 lines
I am today replacing the IDE RAID with a 3WARE IDE-RAID that emulates SCSI on the premise that the problem has to do with hardware bus arbitration. But as of yet I don't have any data to support that or show improvement.
Did you use hdparm to enable unmaskirq and 32-bit IO on normal IDE?
With lots of activity I find 8 sector IO to work better than 16,
and 32-bit IO with sync to be no slower than 32-bit IO without it, so I put a line like this in /etc/rc.local for each disk on HPT370:
hdparm -c 3 -u 1 -m 8 -a 8 /dev/hde
------
Chuck