From: David Lesher (wb8foz@skybridge.SCL.CWRU.Edu)
Date: 04/14/93


From: wb8foz@skybridge.SCL.CWRU.Edu (David Lesher)
Subject: efs fun...
Date: 15 Apr 1993 03:57:37 GMT

I've been having fun with filesystems today. ;-}

Mount and others complained about my /dev/sda6 partition. That's a
274432-block extended partition, with the efs filesystem on it. (Yes, I
*know* I should have efs2, or xiafs, but I'm waiting for the new SLS..)

Anyhow when I looked at it with fdisk, it claimed the partition was
-257394672983 or thereabouts, blocks long. But since the only file (a
BIG tar called SLS.tar) on it was safe on tape, I used fdisk to delete
the partition, and rebuilt it. All was fine. Get out the tape, recover
the file.

Then I played around with iozone for a while. Iozone builds long
sequential files & reads them, then tells you how long it took. Not
being one for half measures, I let it build a 200 meg test file. Gee,
the response sure slows while THAT runs ;_}

Later, guess what, more garbage. What a bore. Unmount, efsck said
every block was not empty when it was, or the other way around. -a the
thing, and found the result was no file. fdisk, delete, new, exit,
reboot, mkefs again, remount. Guess what? The tar file was INTACT.
Through the partition, & the mkefs. Hmm.

q1) Is the efs, or extended partitions on SCSI, noted for this
corruption on large files? Yes, I sync before booting, and brush after
every meal.

q2) How'd that file survive?

q3) When will the all-new SLS, with me-proof filesystems, be available
in a store nearby ;-?