My disk has shrunk?

[Lars Nordlund <lars.nordlund%hem.utfors.se@localhost>]

Hello list

I am using NetBSD amd64 4.99.72.

My messages log says that one of my disks had this amount of sectors:

Aug 15 23:22:30 laphroaig /netbsd: wd0 at atabus6 drive 0: <SAMSUNG HD300LJ> 
Aug 15 23:22:30 laphroaig /netbsd: wd0: drive supports 16-sector PIO transfers, 
LBA48 addressing
Aug 15 23:22:30 laphroaig /netbsd: wd0: 279 GB, 581421 cyl, 16 head, 63 sec, 
512 bytes/sect x 586072368 sectors 
Aug 15 23:22:30 laphroaig /netbsd: wd0: 32-bit data port 
Aug 15 23:22:30 laphroaig /netbsd: wd0: drive supports PIO mode 4, DMA mode 2, 
Ultra-DMA mode 7 
Aug 15 23:22:30 laphroaig /netbsd: wd0(piixide1:0:0):using PIO mode 4, Ultra-DM 
A mode 6 (Ultra/133) (using DMA)

This weekend I discovered that "extreme instability" when writing DVDRs
has come back to haunt me (something I have been trying to debug for
some time..). After the second hang, the wd0f partition could not be
mounted. I quickly noticed that the partition table was zeroed.. After
putting back the netbsd slice at the old start point I noticed that
I could no longer make it as big as it used to...?? I had to make it
2113 sectors shorter..

Knowing that the disklabel accepts pretty much everything..?, I went 
on and created an f-partition on the same sectors as it used to be
(start=40980143, size=545092225). This partition is accepted by fsck
without complaints, but it cannot be mounted:

wd0f: error reading fsbn 545092220 of 545092220-545092223 (wd0 bn 586072363; cn 
581420 tn 15 sn 58), retrying 
wd0: (id not found)

Trying to create a slightly smaller partition (which fits on the now
smaller disk..) with a the same start sector and size=545090112 is also
ok with fsck, but not mount:

mount_ffs: /dev/wd0f on /mnt/dumbo: incorrect super block

Looking closely at the messages log reveals that the disk is nowadays
slightly smaller:

Aug 22 22:08:13 laphroaig /netbsd: wd0 at atabus6 drive 0: <SAMSUNG HD300LJ> 
Aug 22 22:08:13 laphroaig /netbsd: wd0: drive supports 16-sector PIO transfers, 
LBA48 addressing 
Aug 22 22:08:13 laphroaig /netbsd: wd0: 279 GB, 581418 cyl, 16 head, 63 sec, 
512 bytes/sect x 586070255 sectors
Aug 22 22:08:13 laphroaig /netbsd: wd0: 32-bit data port
Aug 22 22:08:13 laphroaig /netbsd: wd0: drive supports PIO mode 4, DMA mode 2, 
U ltra-DMA mode 7 
Aug 22 22:08:13 laphroaig /netbsd: wd0(piixide1:0:0): using PIO mode 4, 
Ultra-DM A mode 6 (Ultra/133) (using DMA)

I have upgraded the kernel slightly (4.99.72 -> 4.99.72) before the 
first crash. I did "make install" but did not reboot on the new kernel 
(until after the crash of course). Going back to the older kernel does 
not fix anything.

Any ideas? The data on the disk is not terribly important. It is just a
setback in time to recreate it. It is also slightly worrysome that
a big disk suddenly may loose a couple of sectors "just like that"?


Best regards,
        Lars Nordlund