David A. Gatwood wrote:
> 
> Speaking of buffer cache, I've run into a weird bug.  A few months ago
> (mid april, I think?), Allen Briggs compiled an ncrscsi kernel for my
> laptop (PB145 from h*ll), with the hardware-based adb driver and the
> intvid code to handle its video.  Anyway, it greatly decreased, but didn't
> eliminate the fs damage.  I have found something that does, however.  When
> I shut down, if I manually sync several times and unmount each partition
> except /, then go to single-user, then sync several times, then mount -o
> ro /, then shutdown -h or -r now, I don't get any fs corruption.  However,
> when I just do a shutdown -h or -r now from multi-user mode, the root fs
> is seriously damaged.  (The others are so small, and so infrequently used
> that they don't usually get hurt.)  Does anyone have any idea what could
> cause this?

My only guess is that your PowerBook is managing to power off before
syncing the disks fully.  Is the full procedure you describe really
necessary?  In other words, I think that the following should accomplish
the same thing:

1) 'shutdown now' to go single user
2) 'sync'
3) 'umount -a'  (although I want to say this may not get the root
device...I'll have to read the man page to be sure).
4) 'shutdown -h now'

How does that work?

Later.

-- 
Colin Wood                                 cwood@ichips.intel.com
Component Design Engineer - MD6                 Intel Corporation
-----------------------------------------------------------------
I speak only on my own behalf, not for my employer.