Re: [NFS] New version of Linux NFSv3 client is out...

[Linus Torvalds]

On Mon, 24 Jan 2000, Eric Werme USG wrote:
> 
> I was delighted when someone at Sun sent me that code to try out several
> years ago and found it worked for us.  I'm not certain that Posix
> requires this work, someone here once broached the idea of changing it,
> but concluded we'd be better off with the status quo.

Linux doesn't do what Sun (or BSD) does for MS_INVALIDATE.

The reason Sun does what it does is for consistency reasons - they have
inconsistent caches with various problems, which means that MS_INVALIDATE
historically is used as a software cache consistency thing. 

Linux doesn't have inconsistent caches, so..

The only way to get reasonable benchmark numbers is to have a benchmark
that has a big enough working set that it doesn't fail over caching. The
same way "dhrystone" is not a good benchmark on modern chips, obviously
this is not a good benchmark on modern operating systems.

			Linus