On Mon, Aug 28, 2017 at 05:19:35PM -0700, Adit Ranadive wrote:
> From: Aditya Sarwade <asarwade@xxxxxxxxxx>
>
> We should report the network header type in the work completion so that
> the kernel can infer the right RoCE type headers.
>
> Reviewed-by: Bryan Tan <bryantan@xxxxxxxxxx>
> Signed-off-by: Aditya Sarwade <asarwade@xxxxxxxxxx>
> Signed-off-by: Adit Ranadive <aditr@xxxxxxxxxx>
> ---
> drivers/infiniband/hw/vmw_pvrdma/pvrdma.h | 6 ++++++
> drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c | 2 ++
> include/uapi/rdma/vmw_pvrdma-abi.h | 13 +++++++++++--
> 3 files changed, 19 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
> index 663a0c3..99b2c97 100644
> --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
> +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
> @@ -426,6 +426,12 @@ static inline int pvrdma_wc_flags_to_ib(int flags)
> return flags;
> }
>
> +static inline enum rdma_network_type pvrdma_wc_network_hdr_to_ib(
> + enum pvrdma_network_type type)
> +{
> + return (enum rdma_network_type)type;
> +}
> +
> static inline int ib_send_flags_to_pvrdma(int flags)
> {
> return flags & PVRDMA_MASK(PVRDMA_SEND_FLAGS_MAX);
> diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c
> index 90aa326..34727f6 100644
> --- a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c
> +++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c
> @@ -389,6 +389,8 @@ static int pvrdma_poll_one(struct pvrdma_cq *cq, struct pvrdma_qp **cur_qp,
> wc->dlid_path_bits = cqe->dlid_path_bits;
> wc->port_num = cqe->port_num;
> wc->vendor_err = cqe->vendor_err;
> + wc->network_hdr_type =
> + pvrdma_wc_network_hdr_to_ib(cqe->network_hdr_type);
>
> /* Update shared ring state */
> pvrdma_idx_ring_inc(&cq->ring_state->rx.cons_head, cq->ibcq.cqe);
> diff --git a/include/uapi/rdma/vmw_pvrdma-abi.h b/include/uapi/rdma/vmw_pvrdma-abi.h
> index c8c1d2d..6a87806 100644
> --- a/include/uapi/rdma/vmw_pvrdma-abi.h
> +++ b/include/uapi/rdma/vmw_pvrdma-abi.h
> @@ -58,6 +58,13 @@
> #define PVRDMA_UAR_CQ_ARM BIT(30) /* Arm bit. */
> #define PVRDMA_UAR_CQ_POLL BIT(31) /* Poll bit. */
>
> +enum pvrdma_network_type {
> + PVRDMA_NETWORK_IB,
> + PVRDMA_NETWORK_ROCE_V1 = PVRDMA_NETWORK_IB,
> + PVRDMA_NETWORK_IPV4,
> + PVRDMA_NETWORK_IPV6
> +};
Doug,
I see that you are already merged this patch, but it is problematic patch.
They defined in uapi file, the new enum which is equal to already existed
128 enum rdma_network_type {
129 RDMA_NETWORK_IB,
130 RDMA_NETWORK_ROCE_V1 = RDMA_NETWORK_IB,
131 RDMA_NETWORK_IPV4,
132 RDMA_NETWORK_IPV6
133 };
And the more important they are doing direct casting from rdma_network_type to
pvrdma_network_type as is in the same patch.
The proper way to do it is to return rdma_network_type directly without obfuscation
and without creating new supported forever UAPI enum.
> +
> enum pvrdma_wr_opcode {
> PVRDMA_WR_RDMA_WRITE,
> PVRDMA_WR_RDMA_WRITE_WITH_IMM,
> @@ -125,7 +132,8 @@ enum pvrdma_wc_flags {
> PVRDMA_WC_IP_CSUM_OK = 1 << 3,
> PVRDMA_WC_WITH_SMAC = 1 << 4,
> PVRDMA_WC_WITH_VLAN = 1 << 5,
> - PVRDMA_WC_FLAGS_MAX = PVRDMA_WC_WITH_VLAN,
> + PVRDMA_WC_WITH_NETWORK_HDR_TYPE = 1 << 6,
> + PVRDMA_WC_FLAGS_MAX = PVRDMA_WC_WITH_NETWORK_HDR_TYPE,
> };
>
> struct pvrdma_alloc_ucontext_resp {
> @@ -283,7 +291,8 @@ struct pvrdma_cqe {
> __u8 dlid_path_bits;
> __u8 port_num;
> __u8 smac[6];
> - __u8 reserved2[7]; /* Pad to next power of 2 (64). */
> + __u8 network_hdr_type;
> + __u8 reserved2[6]; /* Pad to next power of 2 (64). */
> };
>
> #endif /* __VMW_PVRDMA_ABI_H__ */
> --
> 2.7.4
>
> --
> 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
Attachment:
signature.asc
Description: PGP signature
