xfs
[Top] [All Lists]

Re: more on NFS performance

To: Andi Kleen <ak@xxxxxxx>
Subject: Re: more on NFS performance
From: Steve Lord <lord@xxxxxxx>
Date: Tue, 06 Mar 2001 11:46:22 -0600
Cc: Robin Humble <rjh@xxxxxxxxxxxxxxxxxxxxxxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: Message from Andi Kleen <ak@suse.de> of "Tue, 06 Mar 2001 18:28:42 +0100." <20010306182842.A10901@gruyere.muc.suse.de>
Sender: owner-linux-xfs@xxxxxxxxxxx
> On Tue, Mar 06, 2001 at 11:00:41AM -0600, Steve Lord wrote:
> > 
> > Hi,
> > 
> > Is anyone on the list NFS savvy? I need a way from the filp or dentry of
> > telling that a write came from the NFS server, and I cannot really look
> > at the ops field in the dentry, since NFS built as a module will mess this 
> up.
> > 
> > I was hoping for some flag or something which I can use to test for this
> > from within xfs, currently I am doing a strcmp on the task name - which is
> > not something I want to do really.
> 
> There isn't a clean way currently AFAIK, but I guess a flag (O_MORE?) could 
> be set in nfsd_open() in the filp.   I guess it would be something similar 
> to MSG_MORE to turn off the early truncation? If you do that, don't you need 
> a 
> timer to do the truncation later?

The reference cache does its own cleanup internally, it is effectively
done out of write_super. Under Irix an internal open flag is the way
it is done, I just didn't want to dink with NFS to fix this, but maybe
I should.

Steve


> 
> 
> -Andi



<Prev in Thread] Current Thread [Next in Thread>