Coda File System

Re: Coda-client-setup 0.5 released

From: Greg Troxel <gdt_at_ir.bbn.com>
Date: 08 Mar 2005 10:45:53 -0500
Ivan Popov <pin_at_konvalo.org> writes:

> Unfortunately, I had to conclude that the Coda kernel module has
> problems.  After some Konvalo usage (populating the cache via
> strictly readonly, tokenless operations) processes begin to hang on
> access to some file(s?), it is possible to kill -9 such process, but
> the relevant file is inaccessible, any retry hangs forever.

What version, and what arch?  I have used coda on 1.5.x, 1.6.x, 2.0
and current with fairly little trouble.  I think there is a vnode
locking bug someplace that needs to be tracked down, but I rarely lose.

> Then the system becomes unstable and hangs totally.

I have never seen this.  Just panics due to vnode lock counts.

> I have observed crashes after reboot (without venus reinit),
> very soon after the system is up it falls to the kernel debugger
> complaining about inconsitent locking of a parent and a child.

Can you post the backtrace, or the place in the source where this
happens?

> What is the status of the Coda kernel module on NetBSD?
> How current/maintained is it?

Well, it's me and Brett....

> I can guess that *BSD kernel modules may share some common bugs
> from long ago... FreeBSD showed me problems while reintegrating, and I think
> Christer observes similar pattern on Darwin (he used FreeBSD code as the
> template). Though the NetBSD problem is different, as it shows up
> on readonly operations.

I think reintegration problems are separate from this issue.

> I was delighted to run on a compact NetBSD installation, at the same time
> enjoying the full range of my favourite programs directly from Coda.
> Let us make it stable?

If you can get a crash dump, and run gdb that might help.

http://netbsd.org/Documentation/kernel/#debugging_a_kernel_crash_dump


-- 
        Greg Troxel <gdt_at_ir.bbn.com>
Received on 2005-03-08 10:49:28