Subject: Re: NFSv3 bug
To: Christos Zoulas <>
From: Manuel Bouyer <>
List: tech-kern
Date: 09/26/2002 20:44:31
On Thu, Sep 26, 2002 at 02:31:46PM -0400, Christos Zoulas wrote:
> On Sep 26,  8:23pm, (Manuel Bouyer) wrote:
> -- Subject: Re: NFSv3 bug
> | On Thu, Sep 26, 2002 at 02:18:57PM -0400, Christos Zoulas wrote:
> | > Or fix VOP_FSYNC for != 0 offset, and 0 count to flush till the end of file...
> | 
> | This would channge the interface, as currently a DIAGNOSTIC check considers
> | this as a bug from the caller. What do other people think ?
> the diagnostic could be changed to treat the count = 0 with offset < size
> correct.

Yes, of course. My point is just that this is an interface change, and I feel
it's important as, for other parts of the kernel a size of 0 could be
a bug. And it's not clear that other parts of the kernel have a need for
this: the param passed to FFS_FSYNC is end of region to flush, not size.
So size of 0 means start ==  end.
In the kernel it's as easy as to give the end of file as to give a size of 0.
I think it's just the NFS server which is a special case here (but I didn't

Manuel Bouyer, LIP6, Universite Paris VI.