Poor IDE cable causing a kernel lockup (PIIX)?

Poor IDE cable causing a kernel lockup (PIIX)?

[Russell Howe]

Before you go and read all of the following, I should probably
ask: Is the kernel/IDE driver expected to cope with CRC errors caused
by what appears to be bad cabling?

After getting a few (~70) entries like the following from the kernel,
the NMI watchdog takes over and locks the system up with a backtrace:

Mar 17 20:54:12 xiao kernel: hdc: dma_intr: status=0x51 { DriveReady SeekComplete Error }
Mar 17 20:54:12 xiao kernel: hdc: dma_intr: error=0x84 { DriveStatusError BadCRC }

I'm quite willing to blame the rather long, rounded cable this drive is
attached to (although no complaints about hdd, also on the same cable)

The drive itself is a Seagate:
hdc: ST380021A, ATA DISK drive

The controller is the onboard 440BX on an Abit BP6, running two Celeron
300A's with all clock speeds set to what they should be.

A copy of the bootup log is at:
http://siksai.co.uk/~rhowe/bootup.txt

Output of smartctl -a /dev/hdc is at:
http://siksai.co.uk/~rhowe/smart-hdc.txt

(currently running a SMART self test to see if the drive thinks it has
problems)

I have two slightly different kernel traces which I had to copy down
onto paper, so if they will be of any use, I can type them in. I'll be
hooking this box up with a serial console pretty soon, although I'll
need a new kernel build for that (note the broken serial support :)

The machine is just my home NFS server, so reboots and new kernels
aren't a problem.

-- 
Russell Howe     | Why be just another cog in the machine,
rhowe@wiss.co.uk | when you can be the spanner in the works?