Re: RedHat Disk Dangers? [was: Linux and ADSL]

["SN_Diamond" <Norman.Diamond@example.com>]

Sorry I've been too busy to read this list for a while.  Last week,
Stephen J. Turnbull wrote:
>
> >>>>> "Scott" == Scott Stone <SStone@example.com> writes:
>     Scott> why?  7.0 did something that I've never, in over 6 years of
>     Scott> using Linux, come across.  For no reason at all, just
>     Scott> COMPLETELY trashed my ext2fs file system to the point where
>     Scott> /usr/lib was no longer a directory and /lost+found
>     Scott> contained only spotty remnants of what used to be in there.
>
> the ext2fs would
> happily write past physical capacity and start reusing in-use blocks.
> And trashing directories.  Including the /lost+found directory.  Oops.
>
> Didn't matter whether it was
> mke2fs /dev/sdc or fdisk /dev/sdc1, mke2fs /dev/sdc1.

Neato.  Now Linux is Windows 95 compatible.  Windows 95 FDISK does the same
thing on SCSI hard disks.  After getting slaughtered and spending weeks
going back to manufacturers and a retail store, I figured out how to
reproduce it rather quickly and prove that it's Microsoft's doing.  I'm not
quite sure of this detail but think it probably sizes the extended partition
to end one cylinder past the end of the media, then the last logical drive
in the extended partition thinks it occupies that space but actually
overlaps the second-to-last logical drive.

> As you say, I've never seen this on a fixed disk,

I have on one, and followed the details better but am not sure if it's
easily reproduced.  On a 16GB IDE drive, Windows 95 FDISK sized the extended
partition to end at the end of the media, but positioned the last logical
drive to end one cylinder past the end of the extended partition.  Then I
did something a bit different on the same disk, used Windows NT to put an
NTFS primary partition in the last 8GB and then used Windows 95 FDISK to put
the extended partition in between the C: partition and the NTFS partition.
Again on this disk, Windows 95 positioned the extended partition correctly
but positioned the last logical drive to end one cylinder past the end of
the extended partition, so it overlapped with the NTFS partition.  After
that, Windows NT wouldn't boot, and the repair mode of the Windows NT boot
floppies couldn't repair it.  But I didn't get slaughtered by this because I
had already learned my lesson and these operations were tests.

It is really really sad to see Red Hat copy from Windows 95.  Looks like I'd
better figure out a way to test...  Ah, but you say you haven't seen Red Hat
do that on IDE disks, it only does that to SCSI disks?  Then my IDE disks
might be safe.  But, um, maybe I'd better figure out a way to test anyway.

-- Norman Diamond