Another "Me Too" post about the SCSI problem --
I have a Conner 1060S with the firmware patch and a NCR 53c810 board
(made by NexStor) running on a Triton based PCI mb (I have 24 MB of RAM
for the record, maybe the buffer cache causes extra timing
constraints?). It causes the machine to lock solid, and the machine
won't sync() so I never could get a syslog dump, though it is similar to
what's posted earlier. The problem is reproducable but not always
consistent. Here was my results with the release 17 driver on kernel
1.3.5_ (I forget which, it was a while ago -- and I was fed up and
reverted to the aha152x driver which seems stable)
750 meg ext2fs partition, at end of disk. Copying (with cp -av file1
file2) 24 meg file causes it to crash almost immediately (about 1 meg
into the copy). cd, ls, df, hdparm -t /dev/sda, du, dd if=/dev/sda3
of=/dev/null, all work without any trouble. (destination of data was to
my IDE primary disk) hdparm reports a transfer rate of about 3.6
MB/second with the disk.
30 meg ext2fs partition (on 1GB disk at beginning of disk), copying 30 meg
file from the disk to the IDE. No problems whatsoever.
1GB MSDOS/VFAT/UMSDOS partition, copying 24 meg file. Occasionally
crashes, mean time between crashes increased about fourfold. Sometimes
I can copy the file, sometimes not.
I've tried my CPU clock speed and playing with termination; it didn't
seem to help.
Win95/MSDOS/Win31 works fine with the subsystem (which is why I
reluctantly decided to format the disk under FAT :(
Hope this leads to some ideas of what's wrong. Some people I've
corresponded to have claimed defective motherboards, etc., and I'm not
too sure why it works fine under win95 using its protected mode drivers?