diff options
Diffstat (limited to 'src/plugins')
-rw-r--r-- | src/plugins/dev_octeon/CMakeLists.txt | 1 | ||||
-rw-r--r-- | src/plugins/dev_octeon/flow.c | 505 | ||||
-rw-r--r-- | src/plugins/dev_octeon/format.c | 19 | ||||
-rw-r--r-- | src/plugins/dev_octeon/init.c | 2 | ||||
-rw-r--r-- | src/plugins/dev_octeon/octeon.h | 26 | ||||
-rw-r--r-- | src/plugins/dev_octeon/port.c | 26 | ||||
-rw-r--r-- | src/plugins/dev_octeon/rx_node.c | 5 |
7 files changed, 581 insertions, 3 deletions
diff --git a/src/plugins/dev_octeon/CMakeLists.txt b/src/plugins/dev_octeon/CMakeLists.txt index b7c25fe0404..e8abf1a3389 100644 --- a/src/plugins/dev_octeon/CMakeLists.txt +++ b/src/plugins/dev_octeon/CMakeLists.txt @@ -30,6 +30,7 @@ add_vpp_plugin(dev_octeon roc_helper.c rx_node.c tx_node.c + flow.c MULTIARCH_SOURCES rx_node.c diff --git a/src/plugins/dev_octeon/flow.c b/src/plugins/dev_octeon/flow.c new file mode 100644 index 00000000000..1c367a036ab --- /dev/null +++ b/src/plugins/dev_octeon/flow.c @@ -0,0 +1,505 @@ +/* + * Copyright (c) 2024 Marvell. + * SPDX-License-Identifier: Apache-2.0 + * https://spdx.org/licenses/Apache-2.0.html + */ + +#include <dev_octeon/octeon.h> +#include <base/roc_npc_priv.h> + +VLIB_REGISTER_LOG_CLASS (oct_log, static) = { + .class_name = "octeon", + .subclass_name = "flow", +}; + +#define FLOW_IS_ETHERNET_CLASS(f) (f->type == VNET_FLOW_TYPE_ETHERNET) + +#define FLOW_IS_IPV4_CLASS(f) \ + ((f->type == VNET_FLOW_TYPE_IP4) || \ + (f->type == VNET_FLOW_TYPE_IP4_N_TUPLE) || \ + (f->type == VNET_FLOW_TYPE_IP4_N_TUPLE_TAGGED) || \ + (f->type == VNET_FLOW_TYPE_IP4_VXLAN) || \ + (f->type == VNET_FLOW_TYPE_IP4_GTPC) || \ + (f->type == VNET_FLOW_TYPE_IP4_GTPU) || \ + (f->type == VNET_FLOW_TYPE_IP4_L2TPV3OIP) || \ + (f->type == VNET_FLOW_TYPE_IP4_IPSEC_ESP) || \ + (f->type == VNET_FLOW_TYPE_IP4_IPSEC_AH)) + +#define FLOW_IS_IPV6_CLASS(f) \ + ((f->type == VNET_FLOW_TYPE_IP6) || \ + (f->type == VNET_FLOW_TYPE_IP6_N_TUPLE) || \ + (f->type == VNET_FLOW_TYPE_IP6_N_TUPLE_TAGGED) || \ + (f->type == VNET_FLOW_TYPE_IP6_VXLAN)) + +#define FLOW_IS_L3_TYPE(f) \ + ((f->type == VNET_FLOW_TYPE_IP4) || (f->type == VNET_FLOW_TYPE_IP6)) + +#define FLOW_IS_L4_TYPE(f) \ + ((f->type == VNET_FLOW_TYPE_IP4_N_TUPLE) || \ + (f->type == VNET_FLOW_TYPE_IP6_N_TUPLE) || \ + (f->type == VNET_FLOW_TYPE_IP4_N_TUPLE_TAGGED) || \ + (f->type == VNET_FLOW_TYPE_IP6_N_TUPLE_TAGGED)) + +#define FLOW_IS_L4_TUNNEL_TYPE(f) \ + ((f->type == VNET_FLOW_TYPE_IP4_VXLAN) || \ + (f->type == VNET_FLOW_TYPE_IP6_VXLAN) || \ + (f->type == VNET_FLOW_TYPE_IP4_GTPC) || \ + (f->type == VNET_FLOW_TYPE_IP4_GTPU)) + +#define OCT_FLOW_UNSUPPORTED_ACTIONS(f) \ + ((f->actions == VNET_FLOW_ACTION_BUFFER_ADVANCE) || \ + (f->actions == VNET_FLOW_ACTION_REDIRECT_TO_NODE)) + +/* Keep values in sync with vnet/flow.h */ +#define foreach_oct_flow_rss_types \ + _ (1, FLOW_KEY_TYPE_IPV4 | FLOW_KEY_TYPE_TCP, "ipv4-tcp") \ + _ (2, FLOW_KEY_TYPE_IPV4 | FLOW_KEY_TYPE_UDP, "ipv4-udp") \ + _ (3, FLOW_KEY_TYPE_IPV4 | FLOW_KEY_TYPE_SCTP, "ipv4-sctp") \ + _ (5, FLOW_KEY_TYPE_IPV4, "ipv4") \ + _ (9, FLOW_KEY_TYPE_IPV6 | FLOW_KEY_TYPE_TCP, "ipv6-tcp") \ + _ (10, FLOW_KEY_TYPE_IPV6 | FLOW_KEY_TYPE_UDP, "ipv6-udp") \ + _ (11, FLOW_KEY_TYPE_IPV6 | FLOW_KEY_TYPE_SCTP, "ipv6-sctp") \ + _ (13, FLOW_KEY_TYPE_IPV6_EXT, "ipv6-ex") \ + _ (14, FLOW_KEY_TYPE_IPV6, "ipv6") \ + _ (16, FLOW_KEY_TYPE_PORT, "port") \ + _ (17, FLOW_KEY_TYPE_VXLAN, "vxlan") \ + _ (18, FLOW_KEY_TYPE_GENEVE, "geneve") \ + _ (19, FLOW_KEY_TYPE_NVGRE, "nvgre") \ + _ (20, FLOW_KEY_TYPE_GTPU, "gtpu") \ + _ (60, FLOW_KEY_TYPE_L4_DST, "l4-dst-only") \ + _ (61, FLOW_KEY_TYPE_L4_SRC, "l4-src-only") \ + _ (62, FLOW_KEY_TYPE_L3_DST, "l3-dst-only") \ + _ (63, FLOW_KEY_TYPE_L3_SRC, "l3-src-only") + +typedef struct +{ + u16 src_port; + u16 dst_port; + u32 verification_tag; + u32 cksum; +} sctp_header_t; + +typedef struct +{ + u8 ver_flags; + u8 type; + u16 length; + u32 teid; +} gtpu_header_t; + +static void +oct_flow_convert_rss_types (u64 *key, u64 rss_types) +{ +#define _(a, b, c) \ + if (rss_types & (1UL << a)) \ + *key |= b; + + foreach_oct_flow_rss_types +#undef _ + + return; +} + +vnet_dev_rv_t +oct_flow_validate_params (vlib_main_t *vm, vnet_dev_port_t *port, + vnet_dev_port_cfg_type_t type, u32 flow_index, + uword *priv_data) +{ + vnet_flow_t *flow = vnet_get_flow (flow_index); + u32 last_queue; + u32 qid; + + if (type == VNET_DEV_PORT_CFG_GET_RX_FLOW_COUNTER || + type == VNET_DEV_PORT_CFG_RESET_RX_FLOW_COUNTER) + { + log_err (port->dev, "Unsupported request type"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + + if (OCT_FLOW_UNSUPPORTED_ACTIONS (flow)) + { + log_err (port->dev, "Unsupported flow action"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + + if (flow->actions & VNET_FLOW_ACTION_REDIRECT_TO_QUEUE) + { + qid = flow->redirect_queue; + if (qid > port->intf.num_rx_queues - 1 || qid < 0) + { + log_err (port->dev, + "Given Q(%d) is invalid, supported range is %d-%d", qid, 0, + port->intf.num_rx_queues - 1); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + } + + if (flow->actions & VNET_FLOW_ACTION_RSS) + { + last_queue = flow->queue_index + flow->queue_num; + if (last_queue > port->intf.num_rx_queues - 1) + { + log_err (port->dev, + "Given Q range(%d-%d) is invalid, supported range is %d-%d", + flow->queue_index, flow->queue_index + flow->queue_num, 0, + port->intf.num_rx_queues - 1); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + } + return VNET_DEV_OK; +} + +static vnet_dev_rv_t +oct_flow_rule_create (vnet_dev_port_t *port, struct roc_npc_action *actions, + struct roc_npc_item_info *item_info, vnet_flow_t *flow, + uword *private_data) +{ + oct_port_t *oct_port = vnet_dev_get_port_data (port); + struct roc_npc_attr attr = { .priority = 1, .ingress = 1 }; + struct roc_npc_flow *npc_flow; + oct_flow_entry_t *flow_entry; + struct roc_npc *npc; + int rv = 0; + + npc = &oct_port->npc; + + npc_flow = + roc_npc_flow_create (npc, &attr, item_info, actions, npc->pf_func, &rv); + if (rv) + { + log_err (port->dev, "roc_npc_flow_create failed with '%s' error", + roc_error_msg_get (rv)); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + roc_npc_mcam_clear_counter (npc, npc_flow->ctr_id); + + pool_get_zero (oct_port->flow_entries, flow_entry); + flow_entry->index = flow_entry - oct_port->flow_entries; + flow_entry->vnet_flow_index = flow->index; + flow_entry->npc_flow = npc_flow; + + *private_data = flow_entry->index; + + return VNET_DEV_OK; +} + +static vnet_dev_rv_t +oct_flow_add (vlib_main_t *vm, vnet_dev_port_t *port, vnet_flow_t *flow, + uword *private_data) +{ + struct roc_npc_item_info item_info[ROC_NPC_ITEM_TYPE_END] = {}; + struct roc_npc_action actions[ROC_NPC_ITEM_TYPE_END] = {}; + oct_port_t *oct_port = vnet_dev_get_port_data (port); + u16 l4_src_port = 0, l4_dst_port = 0; + u16 l4_src_mask = 0, l4_dst_mask = 0; + struct roc_npc_action_rss rss_conf = {}; + struct roc_npc_action_queue conf = {}; + struct roc_npc_action_mark mark = {}; + struct roc_npc *npc = &oct_port->npc; + vnet_dev_rv_t rv = VNET_DEV_OK; + int layer = 0, index = 0; + u16 *queues = NULL; + u64 flow_key = 0; + u8 proto = 0; + u16 action = 0; + + if (FLOW_IS_ETHERNET_CLASS (flow)) + { + ethernet_header_t eth_spec = { .type = clib_host_to_net_u16 ( + flow->ethernet.eth_hdr.type) }, + eth_mask = { .type = 0xFFFF }; + + item_info[layer].spec = (void *) ð_spec; + item_info[layer].mask = (void *) ð_mask; + item_info[layer].size = sizeof (ethernet_header_t); + item_info[layer].type = ROC_NPC_ITEM_TYPE_ETH; + layer++; + } + + else if (FLOW_IS_IPV4_CLASS (flow)) + { + vnet_flow_ip4_t *ip4_hdr = &flow->ip4; + proto = ip4_hdr->protocol.prot; + ip4_header_t ip4_spec = { .src_address = ip4_hdr->src_addr.addr, + .dst_address = ip4_hdr->dst_addr.addr }, + ip4_mask = { .src_address = ip4_hdr->src_addr.mask, + .dst_address = ip4_hdr->dst_addr.mask }; + + item_info[layer].spec = (void *) &ip4_spec; + item_info[layer].mask = (void *) &ip4_mask; + item_info[layer].size = sizeof (ip4_header_t); + item_info[layer].type = ROC_NPC_ITEM_TYPE_IPV4; + layer++; + + if (FLOW_IS_L4_TYPE (flow)) + { + vnet_flow_ip4_n_tuple_t *ip4_tuple_hdr = &flow->ip4_n_tuple; + + l4_src_port = clib_host_to_net_u16 (ip4_tuple_hdr->src_port.port); + l4_dst_port = clib_host_to_net_u16 (ip4_tuple_hdr->dst_port.port); + l4_src_mask = clib_host_to_net_u16 (ip4_tuple_hdr->src_port.mask); + l4_dst_mask = clib_host_to_net_u16 (ip4_tuple_hdr->dst_port.mask); + } + } + else if (FLOW_IS_IPV6_CLASS (flow)) + { + vnet_flow_ip6_t *ip6_hdr = &flow->ip6; + proto = ip6_hdr->protocol.prot; + ip6_header_t ip6_spec = { .src_address = ip6_hdr->src_addr.addr, + .dst_address = ip6_hdr->dst_addr.addr }, + ip6_mask = { .src_address = ip6_hdr->src_addr.mask, + .dst_address = ip6_hdr->dst_addr.mask }; + + item_info[layer].spec = (void *) &ip6_spec; + item_info[layer].mask = (void *) &ip6_mask; + item_info[layer].size = sizeof (ip6_header_t); + item_info[layer].type = ROC_NPC_ITEM_TYPE_IPV6; + layer++; + + if (FLOW_IS_L4_TYPE (flow)) + { + vnet_flow_ip6_n_tuple_t *ip6_tuple_hdr = &flow->ip6_n_tuple; + + l4_src_port = clib_host_to_net_u16 (ip6_tuple_hdr->src_port.port); + l4_dst_port = clib_host_to_net_u16 (ip6_tuple_hdr->dst_port.port); + l4_src_mask = clib_host_to_net_u16 (ip6_tuple_hdr->src_port.mask); + l4_dst_mask = clib_host_to_net_u16 (ip6_tuple_hdr->dst_port.mask); + } + } + + if (!proto) + goto end_item_info; + + switch (proto) + { + case IP_PROTOCOL_UDP: + item_info[layer].type = ROC_NPC_ITEM_TYPE_UDP; + + udp_header_t udp_spec = { .src_port = l4_src_port, + .dst_port = l4_dst_port }, + udp_mask = { .src_port = l4_src_mask, + .dst_port = l4_dst_mask }; + + item_info[layer].spec = (void *) &udp_spec; + item_info[layer].mask = (void *) &udp_mask; + item_info[layer].size = sizeof (udp_header_t); + layer++; + + if (FLOW_IS_L4_TUNNEL_TYPE (flow)) + { + switch (flow->type) + { + case VNET_FLOW_TYPE_IP4_GTPU: + item_info[layer].type = ROC_NPC_ITEM_TYPE_GTPU; + gtpu_header_t gtpu_spec = { .teid = clib_host_to_net_u32 ( + flow->ip4_gtpu.teid) }, + gtpu_mask = { .teid = 0XFFFFFFFF }; + + item_info[layer].spec = (void *) >pu_spec; + item_info[layer].mask = (void *) >pu_mask; + item_info[layer].size = sizeof (gtpu_header_t); + layer++; + break; + + default: + log_err (port->dev, "Unsupported L4 tunnel type"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + } /* FLOW_IS_L4_TUNNEL_TYPE */ + break; + + case IP_PROTOCOL_TCP: + item_info[layer].type = ROC_NPC_ITEM_TYPE_TCP; + + tcp_header_t tcp_spec = { .src_port = l4_src_port, + .dst_port = l4_dst_port }, + tcp_mask = { .src_port = l4_src_mask, + .dst_port = l4_dst_mask }; + + item_info[layer].spec = (void *) &tcp_spec; + item_info[layer].mask = (void *) &tcp_mask; + item_info[layer].size = sizeof (tcp_header_t); + layer++; + break; + + case IP_PROTOCOL_SCTP: + item_info[layer].type = ROC_NPC_ITEM_TYPE_SCTP; + + sctp_header_t sctp_spec = { .src_port = l4_src_port, + .dst_port = l4_dst_port }, + sctp_mask = { .src_port = l4_src_mask, + .dst_port = l4_dst_mask }; + + item_info[layer].spec = (void *) &sctp_spec; + item_info[layer].mask = (void *) &sctp_mask; + item_info[layer].size = sizeof (sctp_header_t); + layer++; + break; + + case IP_PROTOCOL_IPSEC_ESP: + item_info[layer].type = ROC_NPC_ITEM_TYPE_ESP; + esp_header_t esp_spec = { .spi = clib_host_to_net_u32 ( + flow->ip4_ipsec_esp.spi) }, + esp_mask = { .spi = 0xFFFFFFFF }; + + item_info[layer].spec = (void *) &esp_spec; + item_info[layer].mask = (void *) &esp_mask; + item_info[layer].size = sizeof (u32); + layer++; + break; + + default: + log_err (port->dev, "Unsupported IP protocol '%U'", format_ip_protocol, + proto); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + +end_item_info: + item_info[layer].type = ROC_NPC_ITEM_TYPE_END; + + if (flow->actions & VNET_FLOW_ACTION_REDIRECT_TO_QUEUE) + { + conf.index = flow->redirect_queue; + actions[action].type = ROC_NPC_ACTION_TYPE_QUEUE; + actions[action].conf = &conf; + action++; + } + + else if (flow->actions & VNET_FLOW_ACTION_DROP) + { + actions[action].type = ROC_NPC_ACTION_TYPE_DROP; + action++; + } + + else if (flow->actions & VNET_FLOW_ACTION_RSS) + { + if (!flow->queue_num) + { + log_err (port->dev, "RSS action has no queues"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + queues = clib_mem_alloc (sizeof (u16) * port->intf.num_rx_queues); + + for (index = 0; index < flow->queue_num; index++) + queues[index] = flow->queue_index++; + + oct_flow_convert_rss_types (&flow_key, flow->rss_types); + if (!flow_key) + { + log_err (port->dev, "Invalid RSS hash function"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + npc->flowkey_cfg_state = flow_key; + rss_conf.queue_num = flow->queue_num; + rss_conf.queue = queues; + + actions[action].type = ROC_NPC_ACTION_TYPE_RSS; + actions[action].conf = &rss_conf; + action++; + } + + if (flow->actions & VNET_FLOW_ACTION_MARK) + { + if (flow->mark_flow_id == 0 || + flow->mark_flow_id > (NPC_FLOW_FLAG_VAL - 2)) + { + log_err (port->dev, "mark flow id must be > 0 and < 0xfffe"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + /* RoC library adds 1 to id, so subtract 1 */ + mark.id = flow->mark_flow_id - 1; + actions[action].type = ROC_NPC_ACTION_TYPE_MARK; + actions[action].conf = &mark; + action++; + } + + /* make count as default action */ + actions[action].type = ROC_NPC_ACTION_TYPE_COUNT; + actions[action + 1].type = ROC_NPC_ACTION_TYPE_END; + + rv = oct_flow_rule_create (port, actions, item_info, flow, private_data); + + if (queues) + clib_mem_free (queues); + + return rv; +} + +static vnet_dev_rv_t +oct_flow_del (vlib_main_t *vm, vnet_dev_port_t *port, vnet_flow_t *flow, + uword *private_data) +{ + oct_port_t *oct_port = vnet_dev_get_port_data (port); + struct roc_npc *npc = &oct_port->npc; + struct roc_npc_flow *npc_flow; + oct_flow_entry_t *flow_entry; + int rv = 0, index; + + index = *private_data; + flow_entry = pool_elt_at_index (oct_port->flow_entries, index); + npc_flow = flow_entry->npc_flow; + rv = roc_npc_flow_destroy (npc, npc_flow); + if (rv) + { + log_err (port->dev, "roc_npc_flow_destroy failed with '%s' error", + roc_error_msg_get (rv)); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + pool_put (oct_port->flow_entries, flow_entry); + + return VNET_DEV_OK; +} + +vnet_dev_rv_t +oct_flow_query (vlib_main_t *vm, vnet_dev_port_t *port, u32 flow_index, + uword private_data, u64 *hits) +{ + oct_port_t *oct_port = vnet_dev_get_port_data (port); + struct roc_npc *npc = &oct_port->npc; + struct roc_npc_flow *npc_flow; + oct_flow_entry_t *flow_entry; + i32 flow_count; + int rv = 0; + + flow_count = pool_elts (oct_port->flow_entries); + if (!flow_count) + { + log_err (port->dev, "Flow entry pool is empty"); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + + flow_entry = pool_elt_at_index (oct_port->flow_entries, private_data); + npc_flow = flow_entry->npc_flow; + if (npc_flow->ctr_id == NPC_COUNTER_NONE) + { + log_err (port->dev, "Counters are not available for given flow id (%u)", + flow_index); + return VNET_DEV_ERR_NOT_SUPPORTED; + } + + rv = roc_npc_mcam_read_counter (npc, npc_flow->ctr_id, hits); + if (rv != 0) + { + log_err (port->dev, "Error reading flow counter for given flow id (%u)", + flow_index); + return VNET_DEV_ERR_INTERNAL; + } + + return VNET_DEV_OK; +} + +vnet_dev_rv_t +oct_flow_ops_fn (vlib_main_t *vm, vnet_dev_port_t *port, + vnet_dev_port_cfg_type_t type, u32 flow_index, + uword *priv_data) +{ + vnet_flow_t *flow = vnet_get_flow (flow_index); + + if (type == VNET_DEV_PORT_CFG_ADD_RX_FLOW) + return oct_flow_add (vm, port, flow, priv_data); + + if (type == VNET_DEV_PORT_CFG_DEL_RX_FLOW) + return oct_flow_del (vm, port, flow, priv_data); + + return VNET_DEV_ERR_NOT_SUPPORTED; +} diff --git a/src/plugins/dev_octeon/format.c b/src/plugins/dev_octeon/format.c index 5ee956ad4f4..e624b84f54e 100644 --- a/src/plugins/dev_octeon/format.c +++ b/src/plugins/dev_octeon/format.c @@ -162,3 +162,22 @@ format_oct_tx_trace (u8 *s, va_list *args) return s; } + +u8 * +format_oct_port_flow (u8 *s, va_list *args) +{ + vlib_main_t *vm = vlib_get_main (); + vnet_dev_port_t *port = va_arg (*args, vnet_dev_port_t *); + u32 flow_index = va_arg (*args, u32); + uword private_data = va_arg (*args, uword); + u64 hits; + + if (flow_index == ~0) + return s; + + if (oct_flow_query (vm, port, flow_index, private_data, &hits) == + VNET_DEV_OK) + s = format (s, "flow (%u) hit count: %lu", flow_index, hits); + + return s; +} diff --git a/src/plugins/dev_octeon/init.c b/src/plugins/dev_octeon/init.c index bee449f2123..d3b6814e507 100644 --- a/src/plugins/dev_octeon/init.c +++ b/src/plugins/dev_octeon/init.c @@ -135,7 +135,9 @@ oct_init_nix (vlib_main_t *vm, vnet_dev_t *dev) .start = oct_port_start, .stop = oct_port_stop, .config_change = oct_port_cfg_change, + .config_change_validate = oct_port_cfg_change_validate, .format_status = format_oct_port_status, + .format_flow = format_oct_port_flow, }, .data_size = sizeof (oct_port_t), .initial_data = &oct_port, diff --git a/src/plugins/dev_octeon/octeon.h b/src/plugins/dev_octeon/octeon.h index a87330c6f90..39501d04104 100644 --- a/src/plugins/dev_octeon/octeon.h +++ b/src/plugins/dev_octeon/octeon.h @@ -9,6 +9,9 @@ #include <vppinfra/format.h> #include <vnet/vnet.h> #include <vnet/dev/dev.h> +#include <vnet/flow/flow.h> +#include <vnet/udp/udp.h> +#include <vnet/ipsec/esp.h> #include <base/roc_api.h> #include <dev_octeon/hw_defs.h> @@ -33,10 +36,21 @@ typedef struct typedef struct { + /* vnet flow index */ + u32 vnet_flow_index; + + u32 index; + /* Internal flow object */ + struct roc_npc_flow *npc_flow; +} oct_flow_entry_t; + +typedef struct +{ u8 lf_allocated : 1; u8 tm_initialized : 1; u8 npc_initialized : 1; struct roc_npc npc; + oct_flow_entry_t *flow_entries; } oct_port_t; typedef struct @@ -87,6 +101,7 @@ typedef struct format_function_t format_oct_port_status; format_function_t format_oct_rx_trace; format_function_t format_oct_tx_trace; +format_function_t format_oct_port_flow; /* port.c */ vnet_dev_rv_t oct_port_init (vlib_main_t *, vnet_dev_port_t *); @@ -95,6 +110,8 @@ void oct_port_stop (vlib_main_t *, vnet_dev_port_t *); void oct_port_deinit (vlib_main_t *, vnet_dev_port_t *); vnet_dev_rv_t oct_port_cfg_change (vlib_main_t *, vnet_dev_port_t *, vnet_dev_port_cfg_change_req_t *); +vnet_dev_rv_t oct_port_cfg_change_validate (vlib_main_t *, vnet_dev_port_t *, + vnet_dev_port_cfg_change_req_t *); /* queue.c */ vnet_dev_rv_t oct_rx_queue_alloc (vlib_main_t *, vnet_dev_rx_queue_t *); @@ -108,6 +125,15 @@ void oct_txq_deinit (vlib_main_t *, vnet_dev_tx_queue_t *); format_function_t format_oct_rxq_info; format_function_t format_oct_txq_info; +/* flow.c */ +vnet_dev_rv_t oct_flow_ops_fn (vlib_main_t *, vnet_dev_port_t *, + vnet_dev_port_cfg_type_t, u32, uword *); +vnet_dev_rv_t oct_flow_validate_params (vlib_main_t *, vnet_dev_port_t *, + vnet_dev_port_cfg_type_t, u32, + uword *); +vnet_dev_rv_t oct_flow_query (vlib_main_t *, vnet_dev_port_t *, u32, uword, + u64 *); + #define log_debug(dev, f, ...) \ vlib_log (VLIB_LOG_LEVEL_DEBUG, oct_log.class, "%U: " f, \ format_vnet_dev_addr, (dev), ##__VA_ARGS__) diff --git a/src/plugins/dev_octeon/port.c b/src/plugins/dev_octeon/port.c index 5857bc15f77..f1ae15aadb1 100644 --- a/src/plugins/dev_octeon/port.c +++ b/src/plugins/dev_octeon/port.c @@ -10,6 +10,9 @@ #include <dev_octeon/common.h> #include <vnet/ethernet/ethernet.h> +#define OCT_FLOW_PREALLOC_SIZE 1 +#define OCT_FLOW_MAX_PRIORITY 7 + VLIB_REGISTER_LOG_CLASS (oct_log, static) = { .class_name = "octeon", .subclass_name = "port", @@ -95,6 +98,8 @@ oct_port_init (vlib_main_t *vm, vnet_dev_port_t *port) roc_nix_rss_key_set (nix, default_rss_key); cp->npc.roc_nix = nix; + cp->npc.flow_prealloc_size = OCT_FLOW_PREALLOC_SIZE; + cp->npc.flow_max_priority = OCT_FLOW_MAX_PRIORITY; if ((rrv = roc_npc_init (&cp->npc))) { oct_port_deinit (vm, port); @@ -360,7 +365,7 @@ oct_port_stop (vlib_main_t *vm, vnet_dev_port_t *port) } vnet_dev_rv_t -oct_port_cfg_change_precheck (vlib_main_t *vm, vnet_dev_port_t *port, +oct_port_cfg_change_validate (vlib_main_t *vm, vnet_dev_port_t *port, vnet_dev_port_cfg_change_req_t *req) { vnet_dev_rv_t rv = VNET_DEV_OK; @@ -378,6 +383,14 @@ oct_port_cfg_change_precheck (vlib_main_t *vm, vnet_dev_port_t *port, case VNET_DEV_PORT_CFG_REMOVE_SECONDARY_HW_ADDR: break; + case VNET_DEV_PORT_CFG_ADD_RX_FLOW: + case VNET_DEV_PORT_CFG_DEL_RX_FLOW: + case VNET_DEV_PORT_CFG_GET_RX_FLOW_COUNTER: + case VNET_DEV_PORT_CFG_RESET_RX_FLOW_COUNTER: + rv = oct_flow_validate_params (vm, port, req->type, req->flow_index, + req->private_data); + break; + default: rv = VNET_DEV_ERR_NOT_SUPPORTED; }; @@ -394,8 +407,6 @@ oct_port_cfg_change (vlib_main_t *vm, vnet_dev_port_t *port, switch (req->type) { case VNET_DEV_PORT_CFG_PROMISC_MODE: - { - } break; case VNET_DEV_PORT_CFG_CHANGE_PRIMARY_HW_ADDR: @@ -410,6 +421,15 @@ oct_port_cfg_change (vlib_main_t *vm, vnet_dev_port_t *port, case VNET_DEV_PORT_CFG_MAX_RX_FRAME_SIZE: break; + case VNET_DEV_PORT_CFG_ADD_RX_FLOW: + case VNET_DEV_PORT_CFG_DEL_RX_FLOW: + case VNET_DEV_PORT_CFG_GET_RX_FLOW_COUNTER: + case VNET_DEV_PORT_CFG_RESET_RX_FLOW_COUNTER: + rv = oct_flow_ops_fn (vm, port, req->type, req->flow_index, + req->private_data); + + break; + default: return VNET_DEV_ERR_NOT_SUPPORTED; }; diff --git a/src/plugins/dev_octeon/rx_node.c b/src/plugins/dev_octeon/rx_node.c index c1c4771795e..5f7e5a8469b 100644 --- a/src/plugins/dev_octeon/rx_node.c +++ b/src/plugins/dev_octeon/rx_node.c @@ -131,6 +131,10 @@ oct_rx_batch (vlib_main_t *vm, oct_rx_node_ctx_t *ctx, ctx->n_rx_bytes += b[1]->current_length = d[1].sg0.seg1_size; ctx->n_rx_bytes += b[2]->current_length = d[2].sg0.seg1_size; ctx->n_rx_bytes += b[3]->current_length = d[3].sg0.seg1_size; + b[0]->flow_id = d[0].parse.w[3] >> 48; + b[1]->flow_id = d[1].parse.w[3] >> 48; + b[2]->flow_id = d[2].parse.w[3] >> 48; + b[3]->flow_id = d[3].parse.w[3] >> 48; ctx->n_segs += 4; segs = d[0].sg0.segs + d[1].sg0.segs + d[2].sg0.segs + d[3].sg0.segs; @@ -149,6 +153,7 @@ oct_rx_batch (vlib_main_t *vm, oct_rx_node_ctx_t *ctx, ctx->to_next[0] = vlib_get_buffer_index (vm, b[0]); b[0]->template = bt; ctx->n_rx_bytes += b[0]->current_length = d[0].sg0.seg1_size; + b[0]->flow_id = d[0].parse.w[3] >> 48; ctx->n_segs += 1; if (d[0].sg0.segs > 1) oct_rx_attach_tail (vm, ctx, b[0], d + 0); |