Coda File System

Re: Reproduceable Oops with linux 2.2.x

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Tue, 22 Jun 1999 14:49:58 -0400
On Tue, Jun 22, 1999 at 07:04:22PM +0200, Miklos Szeredi wrote:
>
> I'm not including the whole log of this, because that is very long,
> only the end. But if you need it I can send the whole log (36k
> compressed).
> 
> Note, that the Oops is not produced by the 'ls' process, but by the
> venus-like client, doing a CODA_LOOKUP operation, and calling stat()
> for a normal file. 
> 
> I tried to investigate why this Oops happens, and here's where I got:
> 
> coda_cnremove() is called with a NULL pointer from coda_cache_clear_inode()
> 
> And I don't see how that can happen, unless the c_cnhead field is
> uninitialized, or the memory is corrupted. But I'm not an expert on
> kernel debugging (this is the first time I do this).
> 
> Can this make any sense? 

Yeah, I think I found the problem. Next time however, run the oops through
ksymoops which is in /usr/src/linux/tools/ksymoops, as it will decode the
Code: part and helps pinpointing the exact part of the code where the crash
occurs.

It seems to be the classic case of removing a listentry we need later on to
get to the next item, does the following patch help? (ps. I haven't tested
this myself, but then again, I haven't had such an oops yet).

Jan



Received on 1999-06-22 14:51:04