On Tue, Oct 29, 2024 at 7:33 PM Leon Romanovsky <leon@xxxxxxxxxx> wrote:
>
> On Tue, Oct 22, 2024 at 03:11:53AM -0700, Selvin Xavier wrote:
> > From: Kashyap Desai <kashyap.desai@xxxxxxxxxxxx>
> >
> > Allow users to dump driver specific resource details when
> > queried through rdma tool. This supports the driver data
> > for QP, CQ, MR and SRQ.
> >
> > Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@xxxxxxxxxxxx>
> > Signed-off-by: Kashyap Desai <kashyap.desai@xxxxxxxxxxxx>
> > Signed-off-by: Selvin Xavier <selvin.xavier@xxxxxxxxxxxx>
> > ---
> > drivers/infiniband/hw/bnxt_re/main.c | 148 +++++++++++++++++++++++++++++++++++
> > 1 file changed, 148 insertions(+)
> >
> > diff --git a/drivers/infiniband/hw/bnxt_re/main.c b/drivers/infiniband/hw/bnxt_re/main.c
> > index 6715c96..5bed9af 100644
> > --- a/drivers/infiniband/hw/bnxt_re/main.c
> > +++ b/drivers/infiniband/hw/bnxt_re/main.c
> > @@ -882,6 +882,146 @@ static const struct attribute_group bnxt_re_dev_attr_group = {
> > .attrs = bnxt_re_attributes,
> > };
> >
> > +static int bnxt_re_fill_res_mr_entry(struct sk_buff *msg, struct ib_mr *ib_mr)
> > +{
> > + struct bnxt_qplib_hwq *mr_hwq;
> > + struct nlattr *table_attr;
> > + struct bnxt_re_mr *mr;
> > +
> > + table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_DRIVER);
> > + if (!table_attr)
> > + return -EMSGSIZE;
> > +
> > + mr = container_of(ib_mr, struct bnxt_re_mr, ib_mr);
> > + mr_hwq = &mr->qplib_mr.hwq;
> > +
> > + if (rdma_nl_put_driver_string(msg, "owner",
> > + mr_hwq->is_user ? "user" : "kernel"))
>
> Two comments:
> 1. There is already a helper function to decide if owner is user or kernel - rdma_is_kernel_res().
> 2. This print duplicates existing information. The difference between
> user and kernel can be easily seen by looking on the PID output.
Got it. I will remove this in the follow up patch.
>
> > + goto err;
> > + if (rdma_nl_put_driver_u32(msg, "page_size",
> > + mr_hwq->qe_ppg * mr_hwq->element_size))
> > + goto err;
> > + if (rdma_nl_put_driver_u32(msg, "max_elements", mr_hwq->max_elements))
> > + goto err;
> > + if (rdma_nl_put_driver_u32(msg, "element_size", mr_hwq->element_size))
> > + goto err;
> > + if (rdma_nl_put_driver_u64_hex(msg, "hwq", (unsigned long)mr_hwq))
> > + goto err;
> > + if (rdma_nl_put_driver_u64_hex(msg, "va", mr->qplib_mr.va))
> > + goto err;
>
> <...>
>
> > +static int bnxt_re_fill_res_qp_entry(struct sk_buff *msg, struct ib_qp *ib_qp)
> > +{
> > + struct bnxt_qplib_qp *qplib_qp;
> > + struct nlattr *table_attr;
> > + struct bnxt_re_qp *qp;
> > +
> > + table_attr = nla_nest_start(msg, RDMA_NLDEV_ATTR_DRIVER);
> > + if (!table_attr)
> > + return -EMSGSIZE;
> > +
> > + qp = container_of(ib_qp, struct bnxt_re_qp, ib_qp);
> > + qplib_qp = &qp->qplib_qp;
> > +
> > + if (rdma_nl_put_driver_string(msg, "owner",
> > + ib_qp->uobject ? "user" : "kernel"))
> > + goto err;
> > +
> > + if (rdma_nl_put_driver_u32(msg, "sq_max_wqe", qplib_qp->sq.max_wqe))
> > + goto err;
> > + if (rdma_nl_put_driver_u32(msg, "sq_max_sge", qplib_qp->sq.max_sge))
>
> Doesn't this information already exist in other places? devinfo?
device level max_sge that can be retrieved from devinfo.
This is a per QP value we are displaying here . It is important for
the latest GenP7 adapters as we support variable size
Work Queue entries (and variable size SGEs) now.
This information is available in the query_qp response. But it can be
handy to dump this while debugging.
>
> Thanks
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature
