Re: Q: NFSD readdir in linux-2.6.28

[David Woodhouse <dwmw2@infradead.org>]

On Fri, 2009-04-17 at 23:43 +0100, Al Viro wrote:
> On Fri, Apr 17, 2009 at 11:17:00PM +0100, David Woodhouse wrote:
> > It sounds like the better answer is to just make sure i_mutex is held
> > when nfsd_buffered_readdir() calls back into the provided filldir
> > function (we could do it in the various filldir functions themselves,
> > _if_ they call lookup_one_len(), but I think I prefer it this way --
> > it's simpler). Patch below for comment.
> 
> Umm...  I can live with that, assuming that we don't have callbacks
> that take i_mutex themselves.  AFAICS, everything we call there is
> either obviously not touching i_mutex or is already called while we
> hold i_mutex elsewhere,

More than that... until commit 14f7dd63, we were holding i_mutex when we
called back into those callbacks from nfsd_readdir() itself. We're only
reverting to a fairly recent behaviour, in that respect.

> > (While I'm staring at it, it looks like nfsd_buffered_readdir() should
> > be returning a __be32 not an int, and its 'return -ENOMEM' should be
> > 'return nfserrno(-ENOMEM)'. The first bug I inherited from the existing
> > nfsd_do_readdir() when I replaced it, but the second is all my own. I'll
> > send a patch to fix those shortly.)
> 
> Fold it into this one, please.

OK.

Ah, and see also commit 05f4f678b (Bruce).

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse@intel.com                              Intel Corporation