Add a device capability field csum_cap to denote IPv4 checksum offload support. Devices should configure this field if they support insertion/verification of IPv4, TCP and UDP checksums on outgoing/incoming IPv4 packets according link layer and QP types.
Flags IBV_SEND_IP_CSUM and IBV_WC_IP_CSUM_OK are added for utilizing this capability for send and receive separately. Signed-off-by: Bodong Wang <[email protected]> --- examples/devinfo.c | 33 +++++++++++++++++++++++++++++++++ include/infiniband/kern-abi.h | 7 +++++++ include/infiniband/verbs.h | 22 ++++++++++++++++++++-- man/ibv_poll_cq.3 | 5 +++++ man/ibv_post_send.3 | 4 ++++ src/cmd.c | 13 +++++++++++++ 6 files changed, 82 insertions(+), 2 deletions(-) diff --git a/examples/devinfo.c b/examples/devinfo.c index a8de982..46d4614 100644 --- a/examples/devinfo.c +++ b/examples/devinfo.c @@ -253,6 +253,38 @@ void print_odp_caps(const struct ibv_odp_caps *caps) print_odp_trans_caps(caps->per_transport_caps.ud_odp_caps); } +void print_csum_caps(const struct ibv_csum_cap_per_link *caps) +{ + uint32_t unknown_csum_caps = ~(IBV_CSUM_SUPPORT_RAW | + IBV_CSUM_SUPPORT_UD); + + printf("\teth_csum_cap:\n"); + if (!caps->eth_csum_cap) { + printf("\t\t\t\t\tNO_SUPPORT\n"); + } else { + if (caps->eth_csum_cap & IBV_CSUM_SUPPORT_RAW) + printf("\t\t\t\t\tRAW_QP_SUPPORT\n"); + if (caps->eth_csum_cap & IBV_CSUM_SUPPORT_UD) + printf("\t\t\t\t\tUD_QP_SUPPORT\n"); + if (caps->eth_csum_cap & unknown_csum_caps) + printf("\t\t\t\t\tUnknown flags: 0x%" PRIX32 "\n", + caps->eth_csum_cap & unknown_csum_caps); + } + + printf("\tib_csum_cap:\n"); + if (!caps->ib_csum_cap) { + printf("\t\t\t\t\tNO_SUPPORT\n"); + } else { + if (caps->ib_csum_cap & IBV_CSUM_SUPPORT_RAW) + printf("\t\t\t\t\tRAW_QP_SUPPORT\n"); + if (caps->ib_csum_cap & IBV_CSUM_SUPPORT_UD) + printf("\t\t\t\t\tUD_QP_SUPPORT\n"); + if (caps->ib_csum_cap & unknown_csum_caps) + printf("\t\t\t\t\tUnknown flags: 0x%" PRIX32 "\n", + caps->ib_csum_cap & unknown_csum_caps); + } +} + static int print_hca_cap(struct ibv_device *ib_dev, uint8_t ib_port) { struct ibv_context *ctx; @@ -339,6 +371,7 @@ static int print_hca_cap(struct ibv_device *ib_dev, uint8_t ib_port) printf("\tlocal_ca_ack_delay:\t\t%d\n", device_attr.orig_attr.local_ca_ack_delay); print_odp_caps(&device_attr.odp_caps); + print_csum_caps(&device_attr.csum_cap); } for (port = 1; port <= device_attr.orig_attr.phys_port_cnt; ++port) { diff --git a/include/infiniband/kern-abi.h b/include/infiniband/kern-abi.h index 800c5ab..51d4fb0 100644 --- a/include/infiniband/kern-abi.h +++ b/include/infiniband/kern-abi.h @@ -262,11 +262,18 @@ struct ibv_odp_caps_resp { __u32 reserved; }; +struct ibv_csum_cap_per_link_resp { + __u32 eth_csum_cap; + __u32 ib_csum_cap; +}; + struct ibv_query_device_resp_ex { struct ibv_query_device_resp base; __u32 comp_mask; __u32 response_length; struct ibv_odp_caps_resp odp_caps; + __u64 reserved0[2]; + struct ibv_csum_cap_per_link_resp csum_cap; }; struct ibv_query_port { diff --git a/include/infiniband/verbs.h b/include/infiniband/verbs.h index 1ff5265..134359f 100644 --- a/include/infiniband/verbs.h +++ b/include/infiniband/verbs.h @@ -196,10 +196,16 @@ enum ibv_odp_general_caps { IBV_ODP_SUPPORT = 1 << 0, }; +struct ibv_csum_cap_per_link { + uint32_t eth_csum_cap; + uint32_t ib_csum_cap; +}; + struct ibv_device_attr_ex { struct ibv_device_attr orig_attr; uint32_t comp_mask; struct ibv_odp_caps odp_caps; + struct ibv_csum_cap_per_link csum_cap; }; enum ibv_mtu { @@ -348,9 +354,14 @@ enum ibv_wc_opcode { IBV_WC_RECV_RDMA_WITH_IMM }; +enum { + IBV_WC_IP_CSUM_OK_SHIFT = 2 +}; + enum ibv_wc_flags { IBV_WC_GRH = 1 << 0, - IBV_WC_WITH_IMM = 1 << 1 + IBV_WC_WITH_IMM = 1 << 1, + IBV_WC_IP_CSUM_OK = 1 << IBV_WC_IP_CSUM_OK_SHIFT }; struct ibv_wc { @@ -646,6 +657,11 @@ enum ibv_mig_state { IBV_MIG_ARMED }; +enum ibv_csum_cap_flags { + IBV_CSUM_SUPPORT_UD = 1 << IBV_QPT_UD, + IBV_CSUM_SUPPORT_RAW = 1 << IBV_QPT_RAW_PACKET, +}; + struct ibv_qp_attr { enum ibv_qp_state qp_state; enum ibv_qp_state cur_qp_state; @@ -688,7 +704,8 @@ enum ibv_send_flags { IBV_SEND_FENCE = 1 << 0, IBV_SEND_SIGNALED = 1 << 1, IBV_SEND_SOLICITED = 1 << 2, - IBV_SEND_INLINE = 1 << 3 + IBV_SEND_INLINE = 1 << 3, + IBV_SEND_IP_CSUM = 1 << 4 }; struct ibv_sge { @@ -1459,6 +1476,7 @@ ibv_query_device_ex(struct ibv_context *context, legacy: memset(attr, 0, sizeof(*attr)); ret = ibv_query_device(context, &attr->orig_attr); + return ret; } diff --git a/man/ibv_poll_cq.3 b/man/ibv_poll_cq.3 index 57c6daa..ba5d2ef 100644 --- a/man/ibv_poll_cq.3 +++ b/man/ibv_poll_cq.3 @@ -50,6 +50,11 @@ It is either 0 or the bitwise OR of one or more of the following flags: .B IBV_WC_GRH \fR GRH is present (valid only for UD QPs) .TP .B IBV_WC_WITH_IMM \fR Immediate data value is valid +.TP +.B IBV_WC_IP_CSUM_OK \fR TCP/UDP checksum over IPv4 and IPv4 header checksum are +verified. +This feature is supported only when \fBcsum_cap\fR in device_attr indicates +current QP under current link layer is supported. .PP Not all .I wc diff --git a/man/ibv_post_send.3 b/man/ibv_post_send.3 index 33fbb50..00afd66 100644 --- a/man/ibv_post_send.3 +++ b/man/ibv_post_send.3 @@ -98,6 +98,10 @@ The attribute send_flags describes the properties of the \s-1WR\s0. It is either .TP .B IBV_SEND_INLINE \fR Send data in given gather list as inline data in a send WQE. Valid only for Send and RDMA Write. The L_Key will not be checked. +.TP +.B IBV_SEND_IP_CSUM \fR Offload the IPv4 and TCP/UDP checksum calculation. +This feature is supported only when \fBcsum_cap\fR in device_attr indicates +current QP under current link layer is supported. .SH "RETURN VALUE" .B ibv_post_send() returns 0 on success, or the value of errno on failure (which indicates the failure reason). diff --git a/src/cmd.c b/src/cmd.c index e1914e9..17fc386 100644 --- a/src/cmd.c +++ b/src/cmd.c @@ -160,6 +160,7 @@ int ibv_cmd_query_device_ex(struct ibv_context *context, IBV_INIT_CMD_RESP_EX_V(cmd, cmd_core_size, cmd_size, QUERY_DEVICE_EX, resp, resp_core_size, resp_size); + cmd->comp_mask = 0; cmd->reserved = 0; memset(attr->orig_attr.fw_ver, 0, sizeof(attr->orig_attr.fw_ver)); @@ -189,6 +190,18 @@ int ibv_cmd_query_device_ex(struct ibv_context *context, } } + if (attr_size >= offsetof(struct ibv_device_attr_ex, csum_cap) + + sizeof(attr->csum_cap)) { + if (resp->response_length >= + offsetof(struct ibv_query_device_resp_ex, csum_cap) + + sizeof(resp->csum_cap)) { + attr->csum_cap.eth_csum_cap = resp->csum_cap.eth_csum_cap; + attr->csum_cap.ib_csum_cap = resp->csum_cap.ib_csum_cap; + } else { + memset(&attr->csum_cap, 0, sizeof(attr->csum_cap)); + } + } + return 0; } -- 1.8.3.1 -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
