On Tue, Apr 21, 2020 at 12:29:43AM +0000, Saleem, Shiraz wrote:
> > Subject: Re: [RFC PATCH v5 09/16] RDMA/irdma: Implement device supported
> > verb APIs
> >
> > On Fri, Apr 17, 2020 at 10:12:44AM -0700, Jeff Kirsher wrote:
> > > From: Mustafa Ismail <mustafa.ismail@xxxxxxxxx>
> > >
> > > Implement device supported verb APIs. The supported APIs vary based on
> > > the underlying transport the ibdev is registered as (i.e. iWARP or
> > > RoCEv2).
> > >
> > > Signed-off-by: Mustafa Ismail <mustafa.ismail@xxxxxxxxx>
> > > Signed-off-by: Shiraz Saleem <shiraz.saleem@xxxxxxxxx>
> > > ---
> > > drivers/infiniband/hw/irdma/verbs.c | 4555 +++++++++++++++++++++++
> > > drivers/infiniband/hw/irdma/verbs.h | 213 ++
> > > include/uapi/rdma/ib_user_ioctl_verbs.h | 1 +
> > > 3 files changed, 4769 insertions(+)
> > > create mode 100644 drivers/infiniband/hw/irdma/verbs.c
> > > create mode 100644 drivers/infiniband/hw/irdma/verbs.h
> >
> > <...>
> >
> > > +static int irdma_destroy_qp(struct ib_qp *ibqp, struct ib_udata
> > > +*udata) {
> > > + struct irdma_qp *iwqp = to_iwqp(ibqp);
> > > +
> > > + iwqp->destroyed = 1;
> > > + if (iwqp->ibqp_state >= IB_QPS_INIT && iwqp->ibqp_state <
> > IB_QPS_RTS)
> > > + irdma_next_iw_state(iwqp, IRDMA_QP_STATE_ERROR, 0, 0, 0);
> > > +
> > > + if (!iwqp->user_mode) {
> > > + if (iwqp->iwscq) {
> > > + irdma_clean_cqes(iwqp, iwqp->iwscq);
> > > + if (iwqp->iwrcq != iwqp->iwscq)
> > > + irdma_clean_cqes(iwqp, iwqp->iwrcq);
> > > + }
> > > + }
> > > +
> > > + irdma_remove_push_mmap_entries(iwqp);
> > > + irdma_free_lsmm_rsrc(iwqp);
> > > + irdma_rem_ref(&iwqp->ibqp);
> >
> > No, please ensure that call to destroy_qp is kfree QP without any need in reference
> > counting. We need this to move QP allocation to be IB/core responsibility. I hope
> > that all other verbs objects (with MR as
> > exception) follow the same pattern: create->kzalloc->destroy>kfree.
>
> Yes. I did see the other verb objects allocation move to IB core
> responsibility but not QP. Since we are headed in that direction,
> I do think it's a reasonable expectation to make destroy QP
> synchronous in providers. We ll look to change it in next rev.
Thanks
>
> Thank you Leon for taking the time to review and provide
> feedback.
>
> Shiraz
