On Wed, Feb 26, 2003 at 01:55:25PM -0800, Jason R Thorpe wrote:
> On Wed, Feb 26, 2003 at 01:51:48PM -0800, Simon Gerraty wrote:
> 
>  > >Perhaps it makes sense to put a PID in the "mount arguments" handed to
>  > >the kernel, and have umount(8) check that.  This could be useful for
>  > >shutting down some other types of file systems as well (AFS and some
>  > >FTP translators come to mind).
>  > 
>  > Yes, that would probably do the trick - send it a SIGTERM and you're
>  > done.
> 
> Err, how does LFS do it?  It has to kill lfs_cleanerd when the file system
> is unmounted, right?
> 
Gah. I was going to try to answer your question - I have an LFS volume, and I
deleted a file (the partial dump that caused LFS to panic before) and then (with
nothing open on the LFS) got a panic:
panic: lfs_unmount: still dirty blocks on ifile vnode

I can make the core and kernel available if anyone wants it, but I don't think
it will be very helpful as the kernel wasn't compiled with DEBUG and ddb.onpanic
was set to 0. For now, I'm switching that disk back to FFS. LFS is simply too
buggy. (right now I'm hoping I'll be able to unmount the LFS volume without it
panicing.. sigh)

-Sean

-- 
/~\ The ASCII
\ / Ribbon Campaign                   Sean Davis
 X  Against HTML                       aka dive
/ \ Email!