[ofa-general] [Fwd: Re: NFS-RDMA backport question]

[Jeff Becker]

Hi. Any advice on backporting issue below? Thanks.

-jeff

-------- Original Message --------
Subject: 	Re: NFS-RDMA backport question
Date: 	Mon, 28 Jul 2008 10:11:07 -0400 (EDT)
From: 	James Lentini <jlentini at netapp.com>
To: 	Jeff Becker <Jeffrey.C.Becker at nasa.gov>
CC: 	Tom Tucker <tom at opengridcomputing.com>, Thomas Talpey 
<talpey at netapp.com>
References: 	<488A1DFD.8020301 at nasa.gov>



On Fri, 25 Jul 2008, Jeff Becker wrote:

> Hi. I was doing the backport to 2.6.16.60 (for sles 10 sp2), and I 
> hit a compile problem with launder_page in fs/nfs/file.c. I looked 
> it up in the git commits, and saw that Trond added it to 2.6.20 to 
> fix a race in nfs_release_page(). The problem for backports is that 
> the fix also adds code to mm/truncate.c. If my understanding is 
> correct, the whole point of the backports is to have modules that 
> will work with older kernels (e.g. from distros). However, if I 
> changed mm/truncate.c, the kernel would have to get recompiled, 
> which is a more drastic change than running a backported set of 
> NFS-RDMA modules. Thus, am I stuck leaving the race in kernels older 
> than 2.6.20? Please advise. Thanks.
> 
> -jeff

Jeff,

It sounds like you have a handle on this from an NFS perspective. For 
this sort of judgment call, I think you should bring this up with the 
OFED maintainers. I expect that they have a policy for handling these 
sorts of situations.

james