Re: Question about buffer usage

[Dave Kleikamp <shaggy@austin.ibm.com>]

On Wed, 2005-10-19 at 16:23 -0700, Dave Pifke wrote:
> Andrew Morton wrote:
> 
>  > It could be a JFS quirk - I don't know much about JFS.  It'd be 
> interesting
>  > to know if other filesystems behave in a similar manner.
> 
> I have two more machines on order, so perhaps I'll try a different 
> filesystem on them and report back if it makes a difference.
> 
>  > One thing you could do is to (re)mount the filesystems with `-o noatime'.
> 
> I probably should have mentioned that this is already the case.
> 
>  > That should release _some_ of the blockdev pagecache, but not a lot, I
>  > expect.  Maybe JFS is just metadata-intensive..
> 
> There appears to be a jfs-discussion mailing list on SourceForge; I'll 
> try asking there.

I answered this in jfs-discussion, but for curious on linux-kernel:

There was a pretty major memory leak in jfs in the 2.6.8 kernel.  This
one-line fix is in the 2.6.9 kernel.  I'm guessing that this would
explain the problem you are seeing.
http://www.kernel.org/git/gitweb.cgi?p=linux/kernel/git/tglx/history.git;a=commitdiff;h=fcc1fcc376bb1fe8ed7609b07931fc5bd774943a

Thanks,
Shaggy
-- 
David Kleikamp
IBM Linux Technology Center