Coda File System

Re: venus startup fails; ramfs and coda

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Tue, 19 Sep 2000 10:19:37 -0400
On Tue, Sep 19, 2000 at 12:47:19AM -0400, Christopher J Mason wrote:
> Howdy.
> 
> So call me crazy but I'm trying to get coda to work as a client on a 
> diskless machine (it happens to be an ARM based handheld called the iPAQ). 
> I've got everything compiled and installed, and libraries seem to be setup 
> in the right places.  I've run venus-setup and it creates and edits some 
> files (actually that's a lie, I've edited most of them by hand to have the 
> right stuff in them because they're read only, see below).

Nice. But if you are using ram backed storage for RVM you don't need to
really use RVM at all (remember _Recoverable_ virtual memory). There is
a commandline flag for venus to use malloc/free instead of
rvm_malloc/free,

    venus -rvmt 3

Ofcourse this hasn't been tested in at least... 6 or more years ;)

> / is a read-only file system (cramfs if you must know)
> /coda is just a directory in a read only file system
> /usr is also a read only file system.
> /usr/coda -> /mnt/ramfs/coda which is a ramfs file system
> 
> /etc also happens to live in ramfs.
> 
> /dev also happens to live in ramfs and
> /dev/cfs0 exists.

You could use devfs, then the device would show up as /dev/coda/0. You
do need to update /etc/coda/venus.conf in that case.

Coda's kernel module relies a lot on the container files being stored in
an ext2 filesystem. I just looked at ramfs, and it does use
generic_file_read/write, so it might actually work.

> I'd appreciate any help you could give.  I think coda would be perfect for 
> this kind of device, and I'm willing to try to debug this.

We believe so as well. We already got Coda working on a couple of Itsy's. ;)

Jan
Received on 2000-09-19 10:22:10