On Thu, Sep 08, 2016 at 01:18:16 -0700, Yuval Shaia wrote:
> On Wed, Sep 07, 2016 at 05:25:30PM +0000, Adit Ranadive wrote:
> > > > +
> > > > +static int pvrdma_poll_one(struct pvrdma_cq *cq, struct pvrdma_qp
> > > **cur_qp,
> > > > + struct ib_wc *wc)
> > > > +{
> > > > + struct pvrdma_dev *dev = to_vdev(cq->ibcq.device);
> > > > + int has_data;
> > > > + unsigned int head;
> > > > + bool tried = false;
> > > > + struct pvrdma_cqe *cqe;
> > > > +
> > > > +retry:
> > > > + has_data = pvrdma_idx_ring_has_data(&cq->ring_state->rx,
> > > > + cq->ibcq.cqe, &head);
> > > > + if (has_data == 0) {
> > > > + u32 val;
> > > > +
> > > > + if (tried)
> > > > + return -EAGAIN;
> > > > +
> > > > + /* Pass down POLL to give physical HCA a chance to poll. */
> > > > + val = cq->cq_handle | PVRDMA_UAR_CQ_POLL;
> > > > + writel(cpu_to_le32(val),
> > > > + dev->driver_uar.map + PVRDMA_UAR_CQ_OFFSET);
> > > > +
> > > > + tried = true;
> > > > + goto retry;
> > > > + } else if (has_data == PVRDMA_INVALID_IDX) {
> > > > + /* Ring state could be corrupted. Treat as unrecoverable. */
> > > > + return -EINVAL;
> > >
> > > Please try your best not to return anything but 0 or -EAGAIN from
> > > this function.
> > > Returning -EINVAL will cause caller (pvrdma_poll_cq) to drop all
> > > successive CQE in tn batch.
> >
> > This reason I want to return -EINVAL here is that its possible that
> > the ring state is corrupted so any successive calls to poll_cq will likely not
> give any valid data.
> > If that means successive CQEs are dropped then that should be okay
> > since we have no idea now which CQEs were consumed.
>
> In this case you should handle the problematic CQ in driver, ULPs are not built
> to handle it.
Ok, sounds reasonable. I understand Jason's concerns. Thanks for pointing me those
links! I can return -EAGAIN here and add a dev warning to make sure we atleast log this
condition.
How would the ULP recover if the driver kept returning 0 for CQEs polled? Do ULPs
usually have some timeout notion when polling CQs?
Thanks,
Adit
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
