On Wed, Mar 08, 2006 at 12:53:11PM +0000, Christoph Hellwig wrote:
> Please don't push this to mainline. The static kiocb will go away soon and
> this just causes unessecary churn.
Oh, you mean from the AIO cleanup patches sent out in the last couple
of days? OK. Those patches will not apply to top of tree anyway now,
due to other unrelated changes - I'll work with the IBM guys to get
that sorted out though.
> > Dynamically allocate vattr in places it makes sense to do so, to reduce
> > stack use. Also re-use vattr in some places so that multiple copies are
> > not held on-stack.
>
> Please don't. The vattr is small enough to allocate it on the stack. If
> that's
Hmm, it was surprisingly big to me when I was checkstack'ing those
routines. There were also places where we'd have two of these on-
stack (at different call depths) at one time.
> not enough I have some patches lying around to decrease it's size further.
Yes please. I really can't see it getting much smaller though.
But youre probably right in that we should undo the dynamic-alloc
part of this one, but I we should keep the part that allows just
one vattr onstack at a time.
> That's much better than requiring a dynamic allocation in all these places.
Ayup, naturally.
cheers.
--
Nathan
|