Coda File System

Re: "Assertion failed" in cfs mkmount is what leads to venus error -6

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Mon, 17 Sep 2001 13:21:38 -0400
On Mon, Sep 17, 2001 at 10:43:32PM +0800, Jeremy Malcolm wrote:
> 7. "cfs mkmount documents documents" (this is one of the volumes)
> 
> What happens is that server 1 crashes and instead of a nice mountpoint I
> get a broken symlink like this:
> 
> lrw-r--r--    1 root     65534   11 Sep 17 22:15 documents -> #documents
> 
> The log shows this:
> 
> 22:15:34 f_icreate: bitmap free at 0, inode 1 exists!

Ohh, you set up a new server on a system that wasn't clean. So there
still was a containerfile /vicepa/0/0/1, but the /vicepa/FTREEDB claimed
there wasn't a file there. We create new containerfiles with the O_EXCL
flag to avoid losing anything as a result of out-of-sync FTREEDB
information vs filesystem contents.

As this is a totally new server, with nothing on it. Redo the
vice-setup-rvm step to clear out server RVM and "rm -rf /vicepa/* ;
touch /vicepa/FTREEDB" .

> It's then that venus starts refusing to start with the "CHILD: mount
> system call failed. Killing parent" message and the "Failure of
> coda_cnode_make for root: error -6" (or on server 1 error -110) in the
> kernel.  See my previous postings for more information about that aspect
> of the problem.

Duh, there is no object that corresponds to the volume root on the
server, so the server isn't giving it and the client has no other choice
but to give up.

Jan
Received on 2001-09-17 13:23:18