On Mon, 16 Feb 1998, Bill Hawes wrote:

> Steven N. Hirsch wrote:
> 
> > Since bringing up an installation of Coda, I have been getting a lot of
> > this in the logs on the server box:
> > 
> > Feb 12 18:55:18 cy kernel: iput: device 00:00 inode 28550 still has aliases!
> > Feb 12 18:55:18 cy kernel: iput: device 00:00 inode 28558 still has aliases!
> > Feb 12 18:55:18 cy kernel: iput: device 08:05 inode 127278 still has aliases!
> > Feb 12 18:55:20 cy kernel: VFS: bad count for in-use inode, count=243
> 
> > Bill, the 'VFS: bad count..' warning is from your d_cache debugging patch,
> > so I thought this might be of particular interest.  Do the messages
> > necessarily indicate filesystem problems?
> 
> Hi Steve,
> 
> The messages you're getting do indicate a serious problem, though the
> "bad count" message itself is more of a warning. For most filesystems
> inodes can only be used as many times as they have hard links, so the
> use count should be a small number. 
> 
> I took a quick look at coda and it appears to need a few changes. If
> Peter Braam (or someone familiar with Coda) has time to work on the
> code, I could suggest some changes. 
> 

Peter,

If you are listening, perhaps you and Bill should touch bases on this?

FWIW, since upgrading both client and server to kernel 2.1.86, things have
become a LOT more stable.  I actually built the kernel 5x in a row this
past weekend without corruption.  Only a couple of the above messages
popped up.

I would still love to know why "make" gets confused building on Coda.  It
always thinks that the entire kernel tree is out of date, and starts over
from the raw beginning every time.  Any guesses?

Steve