Re: [PATCH] RDMA/core: reduce IB_POLL_BATCH constant

[Max Gurtovoy]

On 2/28/2018 8:55 PM, Doug Ledford wrote:
> On Wed, 2018-02-28 at 11:50 +0200, Max Gurtovoy wrote:
> > On 2/28/2018 2:21 AM, Bart Van Assche wrote:
> > > On 02/27/18 14:15, Max Gurtovoy wrote:
> > > > -static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc
> > > > *poll_wc)
> > > > +static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc
> > > > *poll_wc,
> > > > +                          int batch)
> > > >    {
> > > > -       int i, n, completed = 0;
> > > > -       struct ib_wc *wcs = poll_wc ? : cq->wc;
> > > > +       int i, n, ib_poll_batch, completed = 0;
> > > > +       struct ib_wc *wcs;
> > > > +
> > > > +       if (poll_wc) {
> > > > +               wcs = poll_wc;
> > > > +               ib_poll_batch = batch;
> > > > +       } else {
> > > > +               wcs = cq->wc;
> > > > +               ib_poll_batch = IB_POLL_BATCH;
> > > > +       }
> > >
> > > Since this code has to be touched I think that we can use this
> > > opportunity to get rid of the "poll_wc ? : cq->wc" conditional and
> > > instead use what the caller passes. That will require to update all
> > > __ib_process_cq(..., ..., NULL) calls. I also propose to let the caller
> > > pass ib_poll_batch instead of figuring it out in this function.
> > > Otherwise the approach of this patch looks fine to me.
> >
> > Thanks Bart.
> > I'll make these changes and submit.
>
> That sounds reasonable to me too, thanks for reworking and resubmitting.

Sure, NP.
We've run NVMe-oF and SRP with the new patch.
I'll send it through Mellanox maintainers pull request.

Thanks for reporting and reviewing.

-Max.