[nfsv4] 2.6.16 to 2.6.17 read performance change

[Bryce Harrington]

On Mon, Aug 14, 2006 at 05:44:47PM -0400, J. Bruce Fields wrote:
> On Tue, Aug 08, 2006 at 10:27:18AM -0700, Bryce Harrington wrote:
> > We did a bit more investigation into the cause of the big performance
> > jump that I presented about at OLS, that occurred between 2.6.16 and
> > 2.6.17.  It looks like there was an NFS patch submitted by Trond a long
> > time ago to fix the performance issue, that only recently was
> > incorporated into the mainline kernel.
> 
> OK, neat, makes sense.
> 
> So, out of curiosity--did you just do a binary search for that patch?
> Were you able to automate the search, or did someone have to kick off
> each build and test?

Actually Jason deduced it after googling up an old post from Trond on
the subject.  He noticed that the patch hadn't gotten applied until the
time in question, and so tried it out and found it was indeed the cause.
(Jason, feel free to fill in details I missed...)

Bryce