So, I have a pair of ~6GB EIDE drives.

I have them under BSD/OS, running as a "splice" (like ccd, only a different
guy wrote it).

Unfortunately, BSD/OS has no DMA support.

So.

I move them.  I move them to an alpha, with a Promise plug-in EIDE card.

Initial state:  Nothing works.
I get a beta driver, and a friendly guy sends me a patch to make it work
right on my alpha.

I now see the disks.

If I ccd them, I end up with a disk.  Unfortunately, the superblock is about
32 blocks in.  If I build a label for it, I see the superblock, and I can
mount, but fsck dies horribly.  ls works, ls -l doesn't.

So, I rework the disk labels to use 32 block cylinders, and (thanks to Chris
Demetriou) manage to create new labels that show partitions starting "one
cylinder" in.

I ccd these.

The resulting partition fsck's almost-okay:

# fsck /dev/ccd0c
** /dev/rccd0c
** Last Mounted on /mnt
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
FREE BLK COUNT(S) WRONG IN SUPERBLK
SALVAGE? [yn] y

BLK(S) MISSING IN BIT MAPS
SALVAGE? [yn] y

SUMMARY INFORMATION BAD
SALVAGE? [yn] y

132202 files, 1854200 used, 1427180 free (19092 frags, 176011 blocks, 0.6%
fragmentation)

MARK FILE SYSTEM CLEAN? [yn] y


***** FILE SYSTEM MARKED CLEAN *****

***** FILE SYSTEM WAS MODIFIED *****


It mounts.

It lists.

Files pass checksum.

And it took me about 20 minutes less to figure this all out than it would have
to newfs and restore.  Yee-hah!

The depressing thing is that nfs-exporting this to my other box gets better
performance than I got having the drives physically in the other box.  Yow.

-s