X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=lib%2Flibrte_sched%2Frte_sched.c;h=0fa074166469705f987167b415e45ac814623371;hb=83415d4fd88c925002655aa755601998a3cdef2c;hp=44b6f11ca09aba01edeb8659809949ef79757748;hpb=a91c3cadb8c33804d4b5f194a47837f699fd9af0;p=dpdk.git diff --git a/lib/librte_sched/rte_sched.c b/lib/librte_sched/rte_sched.c index 44b6f11ca0..0fa0741664 100644 --- a/lib/librte_sched/rte_sched.c +++ b/lib/librte_sched/rte_sched.c @@ -1,35 +1,5 @@ -/*- - * BSD LICENSE - * - * Copyright(c) 2010-2013 Intel Corporation. All rights reserved. - * All rights reserved. - * - * Redistribution and use in source and binary forms, with or without - * modification, are permitted provided that the following conditions - * are met: - * - * * Redistributions of source code must retain the above copyright - * notice, this list of conditions and the following disclaimer. - * * Redistributions in binary form must reproduce the above copyright - * notice, this list of conditions and the following disclaimer in - * the documentation and/or other materials provided with the - * distribution. - * * Neither the name of Intel Corporation nor the names of its - * contributors may be used to endorse or promote products derived - * from this software without specific prior written permission. - * - * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS - * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT - * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR - * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT - * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, - * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT - * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, - * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY - * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT - * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE - * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. - * +/* SPDX-License-Identifier: BSD-3-Clause + * Copyright(c) 2010-2014 Intel Corporation */ #include @@ -38,14 +8,15 @@ #include #include #include -#include +#include #include #include #include #include +#include +#include #include "rte_sched.h" -#include "rte_bitmap.h" #include "rte_sched_common.h" #include "rte_approx.h" @@ -53,120 +24,62 @@ #pragma warning(disable:2259) /* conversion may lose significant bits */ #endif -#ifndef RTE_SCHED_DEBUG -#define RTE_SCHED_DEBUG 0 -#endif - -#ifndef RTE_SCHED_OPTIMIZATIONS -#define RTE_SCHED_OPTIMIZATIONS 0 -#endif +#ifdef RTE_SCHED_VECTOR +#include -#if RTE_SCHED_OPTIMIZATIONS -#include +#ifdef RTE_ARCH_X86 +#define SCHED_VECTOR_SSE4 +#elif defined(RTE_MACHINE_CPUFLAG_NEON) +#define SCHED_VECTOR_NEON #endif -#define RTE_SCHED_ENQUEUE 1 - -#define RTE_SCHED_TS 1 - -#if RTE_SCHED_TS == 0 /* Infinite credits. Traffic shaping disabled. */ -#define RTE_SCHED_TS_CREDITS_UPDATE 0 -#define RTE_SCHED_TS_CREDITS_CHECK 0 -#else /* Real Credits. Full traffic shaping implemented. */ -#define RTE_SCHED_TS_CREDITS_UPDATE 1 -#define RTE_SCHED_TS_CREDITS_CHECK 1 #endif -#ifndef RTE_SCHED_TB_RATE_CONFIG_ERR #define RTE_SCHED_TB_RATE_CONFIG_ERR (1e-7) -#endif - -#define RTE_SCHED_WRR 1 - -#ifndef RTE_SCHED_WRR_SHIFT #define RTE_SCHED_WRR_SHIFT 3 -#endif - -#ifndef RTE_SCHED_PORT_N_GRINDERS -#define RTE_SCHED_PORT_N_GRINDERS 8 -#endif -#if (RTE_SCHED_PORT_N_GRINDERS == 0) || (RTE_SCHED_PORT_N_GRINDERS & (RTE_SCHED_PORT_N_GRINDERS - 1)) -#error Number of grinders must be non-zero and a power of 2 -#endif -#if (RTE_SCHED_OPTIMIZATIONS && (RTE_SCHED_PORT_N_GRINDERS != 8)) -#error Number of grinders must be 8 when RTE_SCHED_OPTIMIZATIONS is set -#endif - +#define RTE_SCHED_MAX_QUEUES_PER_TC RTE_SCHED_BE_QUEUES_PER_PIPE #define RTE_SCHED_GRINDER_PCACHE_SIZE (64 / RTE_SCHED_QUEUES_PER_PIPE) - #define RTE_SCHED_PIPE_INVALID UINT32_MAX - #define RTE_SCHED_BMP_POS_INVALID UINT32_MAX -struct rte_sched_subport { - /* Token bucket (TB) */ - uint64_t tb_time; /* time of last update */ - uint32_t tb_period; - uint32_t tb_credits_per_period; - uint32_t tb_size; - uint32_t tb_credits; - - /* Traffic classes (TCs) */ - uint64_t tc_time; /* time of next update */ - uint32_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t tc_period; - - /* TC oversubscription */ - uint32_t tc_ov_period; - uint64_t tc_ov_time; - uint32_t tc_ov_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint8_t tc_ov_period_id; - uint8_t tc_ov[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t tc_ov_n[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - double tc_ov_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - - /* Statistics */ - struct rte_sched_subport_stats stats; -}; +/* Scaling for cycles_per_byte calculation + * Chosen so that minimum rate is 480 bit/sec + */ +#define RTE_SCHED_TIME_SHIFT 8 struct rte_sched_pipe_profile { /* Token bucket (TB) */ - uint32_t tb_period; - uint32_t tb_credits_per_period; - uint32_t tb_size; - + uint64_t tb_period; + uint64_t tb_credits_per_period; + uint64_t tb_size; + /* Pipe traffic classes */ - uint32_t tc_period; - uint32_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint8_t tc_ov_weight[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - - /* Pipe queues */ - uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_PIPE]; + uint64_t tc_period; + uint64_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint8_t tc_ov_weight; + + /* Pipe best-effort traffic class queues */ + uint8_t wrr_cost[RTE_SCHED_BE_QUEUES_PER_PIPE]; }; struct rte_sched_pipe { /* Token bucket (TB) */ uint64_t tb_time; /* time of last update */ - uint32_t tb_credits; + uint64_t tb_credits; /* Pipe profile and flags */ uint32_t profile; - + /* Traffic classes (TCs) */ uint64_t tc_time; /* time of next update */ - uint32_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - + uint64_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + /* Weighted Round Robin (WRR) */ - uint8_t wrr_tokens[RTE_SCHED_QUEUES_PER_PIPE]; - + uint8_t wrr_tokens[RTE_SCHED_BE_QUEUES_PER_PIPE]; + /* TC oversubscription */ -#ifdef RTE_SCHED_SUBPORT_TC_OV - uint32_t tc_ov_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint64_t tc_ov_credits; uint8_t tc_ov_period_id; -#else - uint64_t reserved; -#endif } __rte_cache_aligned; struct rte_sched_queue { @@ -194,7 +107,7 @@ struct rte_sched_grinder { uint32_t pcache_qindex[RTE_SCHED_GRINDER_PCACHE_SIZE]; uint32_t pcache_w; uint32_t pcache_r; - + /* Current pipe */ enum grinder_state state; uint32_t productive; @@ -204,47 +117,68 @@ struct rte_sched_grinder { struct rte_sched_pipe_profile *pipe_params; /* TC cache */ - uint8_t tccache_qmask[4]; - uint32_t tccache_qindex[4]; + uint8_t tccache_qmask[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint32_t tccache_qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; uint32_t tccache_w; uint32_t tccache_r; - + /* Current TC */ uint32_t tc_index; - struct rte_sched_queue *queue[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - struct rte_mbuf **qbase[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t qindex[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + struct rte_sched_queue *queue[RTE_SCHED_MAX_QUEUES_PER_TC]; + struct rte_mbuf **qbase[RTE_SCHED_MAX_QUEUES_PER_TC]; + uint32_t qindex[RTE_SCHED_MAX_QUEUES_PER_TC]; uint16_t qsize; uint32_t qmask; uint32_t qpos; struct rte_mbuf *pkt; - - double ov_coef; - - uint16_t wrr_tokens[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; - uint16_t wrr_mask[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; - uint8_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; + + /* WRR */ + uint16_t wrr_tokens[RTE_SCHED_BE_QUEUES_PER_PIPE]; + uint16_t wrr_mask[RTE_SCHED_BE_QUEUES_PER_PIPE]; + uint8_t wrr_cost[RTE_SCHED_BE_QUEUES_PER_PIPE]; }; -struct rte_sched_port { - /* User parameters */ - uint32_t n_subports_per_port; - uint32_t n_pipes_per_subport; - uint32_t rate; - uint32_t mtu; - uint32_t frame_overhead; - uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; +struct rte_sched_subport { + /* Token bucket (TB) */ + uint64_t tb_time; /* time of last update */ + uint64_t tb_period; + uint64_t tb_credits_per_period; + uint64_t tb_size; + uint64_t tb_credits; + + /* Traffic classes (TCs) */ + uint64_t tc_time; /* time of next update */ + uint64_t tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint64_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint64_t tc_period; + + /* TC oversubscription */ + uint64_t tc_ov_wm; + uint64_t tc_ov_wm_min; + uint64_t tc_ov_wm_max; + uint8_t tc_ov_period_id; + uint8_t tc_ov; + uint32_t tc_ov_n; + double tc_ov_rate; + + /* Statistics */ + struct rte_sched_subport_stats stats __rte_cache_aligned; + + /* Subport pipes */ + uint32_t n_pipes_per_subport_enabled; uint32_t n_pipe_profiles; + uint32_t n_max_pipe_profiles; + + /* Pipe best-effort TC rate */ + uint64_t pipe_tc_be_rate_max; + + /* Pipe queues size */ + uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + #ifdef RTE_SCHED_RED - struct rte_red_config red_config[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][e_RTE_METER_COLORS]; + struct rte_red_config red_config[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][RTE_COLORS]; #endif - /* Timing */ - uint64_t time_cpu_cycles; /* Current CPU time measured in CPU cyles */ - uint64_t time_cpu_bytes; /* Current CPU time measured in bytes */ - uint64_t time; /* Current NIC TX time measured in bytes */ - double cycles_per_byte; /* CPU cycles per byte */ - /* Scheduling loop detection */ uint32_t pipe_loop; uint32_t pipe_exhaustion; @@ -252,19 +186,15 @@ struct rte_sched_port { /* Bitmap */ struct rte_bitmap *bmp; uint32_t grinder_base_bmp_pos[RTE_SCHED_PORT_N_GRINDERS] __rte_aligned_16; - + /* Grinders */ struct rte_sched_grinder grinder[RTE_SCHED_PORT_N_GRINDERS]; uint32_t busy_grinders; - struct rte_mbuf **pkts_out; - uint32_t n_pkts_out; - + /* Queue base calculation */ uint32_t qsize_add[RTE_SCHED_QUEUES_PER_PIPE]; uint32_t qsize_sum; - - /* Large data structures */ - struct rte_sched_subport *subport; + struct rte_sched_pipe *pipe; struct rte_sched_queue *queue; struct rte_sched_queue_extra *queue_extra; @@ -274,516 +204,825 @@ struct rte_sched_port { uint8_t memory[0] __rte_cache_aligned; } __rte_cache_aligned; -enum rte_sched_port_array { - e_RTE_SCHED_PORT_ARRAY_SUBPORT = 0, - e_RTE_SCHED_PORT_ARRAY_PIPE, - e_RTE_SCHED_PORT_ARRAY_QUEUE, - e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA, - e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES, - e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY, - e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY, - e_RTE_SCHED_PORT_ARRAY_TOTAL, -}; +struct rte_sched_port { + /* User parameters */ + uint32_t n_subports_per_port; + uint32_t n_pipes_per_subport; + uint32_t n_pipes_per_subport_log2; + uint16_t pipe_queue[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint8_t pipe_tc[RTE_SCHED_QUEUES_PER_PIPE]; + uint8_t tc_queue[RTE_SCHED_QUEUES_PER_PIPE]; + uint64_t rate; + uint32_t mtu; + uint32_t frame_overhead; + int socket; -#ifdef RTE_SCHED_COLLECT_STATS + /* Timing */ + uint64_t time_cpu_cycles; /* Current CPU time measured in CPU cyles */ + uint64_t time_cpu_bytes; /* Current CPU time measured in bytes */ + uint64_t time; /* Current NIC TX time measured in bytes */ + struct rte_reciprocal inv_cycles_per_byte; /* CPU cycles per byte */ + uint64_t cycles_per_byte; + + /* Grinders */ + struct rte_mbuf **pkts_out; + uint32_t n_pkts_out; + uint32_t subport_id; + + /* Large data structures */ + struct rte_sched_subport *subports[0] __rte_cache_aligned; +} __rte_cache_aligned; + +enum rte_sched_subport_array { + e_RTE_SCHED_SUBPORT_ARRAY_PIPE = 0, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_EXTRA, + e_RTE_SCHED_SUBPORT_ARRAY_PIPE_PROFILES, + e_RTE_SCHED_SUBPORT_ARRAY_BMP_ARRAY, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_ARRAY, + e_RTE_SCHED_SUBPORT_ARRAY_TOTAL, +}; static inline uint32_t -rte_sched_port_queues_per_subport(struct rte_sched_port *port) +rte_sched_subport_pipe_queues(struct rte_sched_subport *subport) { - return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport; + return RTE_SCHED_QUEUES_PER_PIPE * subport->n_pipes_per_subport_enabled; } -#endif +static inline struct rte_mbuf ** +rte_sched_subport_pipe_qbase(struct rte_sched_subport *subport, uint32_t qindex) +{ + uint32_t pindex = qindex >> 4; + uint32_t qpos = qindex & (RTE_SCHED_QUEUES_PER_PIPE - 1); + + return (subport->queue_array + pindex * + subport->qsize_sum + subport->qsize_add[qpos]); +} + +static inline uint16_t +rte_sched_subport_pipe_qsize(struct rte_sched_port *port, +struct rte_sched_subport *subport, uint32_t qindex) +{ + uint32_t tc = port->pipe_tc[qindex & (RTE_SCHED_QUEUES_PER_PIPE - 1)]; + + return subport->qsize[tc]; +} static inline uint32_t rte_sched_port_queues_per_port(struct rte_sched_port *port) { - return RTE_SCHED_QUEUES_PER_PIPE * port->n_pipes_per_subport * port->n_subports_per_port; + uint32_t n_queues = 0, i; + + for (i = 0; i < port->n_subports_per_port; i++) + n_queues += rte_sched_subport_pipe_queues(port->subports[i]); + + return n_queues; +} + +static inline uint16_t +rte_sched_port_pipe_queue(struct rte_sched_port *port, uint32_t traffic_class) +{ + uint16_t pipe_queue = port->pipe_queue[traffic_class]; + + return pipe_queue; +} + +static inline uint8_t +rte_sched_port_pipe_tc(struct rte_sched_port *port, uint32_t qindex) +{ + uint8_t pipe_tc = port->pipe_tc[qindex & (RTE_SCHED_QUEUES_PER_PIPE - 1)]; + + return pipe_tc; +} + +static inline uint8_t +rte_sched_port_tc_queue(struct rte_sched_port *port, uint32_t qindex) +{ + uint8_t tc_queue = port->tc_queue[qindex & (RTE_SCHED_QUEUES_PER_PIPE - 1)]; + + return tc_queue; } static int -rte_sched_port_check_params(struct rte_sched_port_params *params) +pipe_profile_check(struct rte_sched_pipe_params *params, + uint64_t rate, uint16_t *qsize) { - uint32_t i, j; - + uint32_t i; + + /* Pipe parameters */ if (params == NULL) { - return -1; + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter params\n", __func__); + return -EINVAL; } - - /* name */ - if (params->name == NULL) { - return -2; + + /* TB rate: non-zero, not greater than port rate */ + if (params->tb_rate == 0 || + params->tb_rate > rate) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tb rate\n", __func__); + return -EINVAL; } - - /* socket */ - if ((params->socket < 0) || (params->socket >= RTE_MAX_NUMA_NODES)) { - return -3; + + /* TB size: non-zero */ + if (params->tb_size == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tb size\n", __func__); + return -EINVAL; } - - /* rate */ - if (params->rate == 0) { - return -4; + + /* TC rate: non-zero if qsize non-zero, less than pipe rate */ + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + if ((qsize[i] == 0 && params->tc_rate[i] != 0) || + (qsize[i] != 0 && (params->tc_rate[i] == 0 || + params->tc_rate[i] > params->tb_rate))) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for qsize or tc_rate\n", __func__); + return -EINVAL; + } } - - /* mtu */ - if (params->mtu == 0) { - return -5; + + if (params->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE] == 0 || + qsize[RTE_SCHED_TRAFFIC_CLASS_BE] == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for be traffic class rate\n", __func__); + return -EINVAL; } - - /* n_subports_per_port: non-zero, power of 2 */ - if ((params->n_subports_per_port == 0) || (!rte_is_power_of_2(params->n_subports_per_port))) { - return -6; + + /* TC period: non-zero */ + if (params->tc_period == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tc period\n", __func__); + return -EINVAL; } - /* n_pipes_per_subport: non-zero, power of 2 */ - if ((params->n_pipes_per_subport == 0) || (!rte_is_power_of_2(params->n_pipes_per_subport))) { - return -7; + /* Best effort tc oversubscription weight: non-zero */ + if (params->tc_ov_weight == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tc ov weight\n", __func__); + return -EINVAL; } - - /* qsize: non-zero, power of 2, no bigger than 32K (due to 16-bit read/write pointers) */ - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - uint16_t qsize = params->qsize[i]; - - if ((qsize == 0) || (!rte_is_power_of_2(qsize))) { - return -8; + + /* Queue WRR weights: non-zero */ + for (i = 0; i < RTE_SCHED_BE_QUEUES_PER_PIPE; i++) { + if (params->wrr_weights[i] == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for wrr weight\n", __func__); + return -EINVAL; } } - - /* pipe_profiles and n_pipe_profiles */ - if ((params->pipe_profiles == NULL) || - (params->n_pipe_profiles == 0) || - (params->n_pipe_profiles > RTE_SCHED_PIPE_PROFILES_PER_PORT)) { - return -9; + + return 0; +} + +static int +rte_sched_port_check_params(struct rte_sched_port_params *params) +{ + if (params == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter params\n", __func__); + return -EINVAL; } - - for (i = 0; i < params->n_pipe_profiles; i ++) { - struct rte_sched_pipe_params *p = params->pipe_profiles + i; - - /* TB rate: non-zero, not greater than port rate */ - if ((p->tb_rate == 0) || (p->tb_rate > params->rate)) { - return -10; - } - - /* TB size: non-zero */ - if (p->tb_size == 0) { - return -11; - } - /* TC rate: non-zero, less than pipe rate */ - for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j ++) { - if ((p->tc_rate[j] == 0) || (p->tc_rate[j] > p->tb_rate)) { - return -12; - } - } - - /* TC period: non-zero */ - if (p->tc_period == 0) { - return -13; - } + /* socket */ + if (params->socket < 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for socket id\n", __func__); + return -EINVAL; + } -#ifdef RTE_SCHED_SUBPORT_TC_OV - /* TC oversubscription weights: non-zero */ - for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j ++) { - if (p->tc_ov_weight[j] == 0) { - return -13; - } - } -#endif + /* rate */ + if (params->rate == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for rate\n", __func__); + return -EINVAL; + } - /* Queue WRR weights: non-zero */ - for (j = 0; j < RTE_SCHED_QUEUES_PER_PIPE; j ++) { - if (p->wrr_weights[j] == 0) { - return -15; - } - } + /* mtu */ + if (params->mtu == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for mtu\n", __func__); + return -EINVAL; } - + + /* n_subports_per_port: non-zero, limited to 16 bits, power of 2 */ + if (params->n_subports_per_port == 0 || + params->n_subports_per_port > 1u << 16 || + !rte_is_power_of_2(params->n_subports_per_port)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for number of subports\n", __func__); + return -EINVAL; + } + + /* n_pipes_per_subport: non-zero, power of 2 */ + if (params->n_pipes_per_subport == 0 || + !rte_is_power_of_2(params->n_pipes_per_subport)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for maximum pipes number\n", __func__); + return -EINVAL; + } + return 0; } static uint32_t -rte_sched_port_get_array_base(struct rte_sched_port_params *params, enum rte_sched_port_array array) -{ - uint32_t n_subports_per_port = params->n_subports_per_port; - uint32_t n_pipes_per_subport = params->n_pipes_per_subport; - uint32_t n_pipes_per_port = n_pipes_per_subport * n_subports_per_port; - uint32_t n_queues_per_port = RTE_SCHED_QUEUES_PER_PIPE * n_pipes_per_subport * n_subports_per_port; - - uint32_t size_subport = n_subports_per_port * sizeof(struct rte_sched_subport); - uint32_t size_pipe = n_pipes_per_port * sizeof(struct rte_sched_pipe); - uint32_t size_queue = n_queues_per_port * sizeof(struct rte_sched_queue); - uint32_t size_queue_extra = n_queues_per_port * sizeof(struct rte_sched_queue_extra); - uint32_t size_pipe_profiles = RTE_SCHED_PIPE_PROFILES_PER_PORT * sizeof(struct rte_sched_pipe_profile); - uint32_t size_bmp_array = rte_bitmap_get_memory_footprint(n_queues_per_port); +rte_sched_subport_get_array_base(struct rte_sched_subport_params *params, + enum rte_sched_subport_array array) +{ + uint32_t n_pipes_per_subport = params->n_pipes_per_subport_enabled; + uint32_t n_subport_pipe_queues = + RTE_SCHED_QUEUES_PER_PIPE * n_pipes_per_subport; + + uint32_t size_pipe = n_pipes_per_subport * sizeof(struct rte_sched_pipe); + uint32_t size_queue = + n_subport_pipe_queues * sizeof(struct rte_sched_queue); + uint32_t size_queue_extra + = n_subport_pipe_queues * sizeof(struct rte_sched_queue_extra); + uint32_t size_pipe_profiles = params->n_max_pipe_profiles * + sizeof(struct rte_sched_pipe_profile); + uint32_t size_bmp_array = + rte_bitmap_get_memory_footprint(n_subport_pipe_queues); uint32_t size_per_pipe_queue_array, size_queue_array; - + uint32_t base, i; - + size_per_pipe_queue_array = 0; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - size_per_pipe_queue_array += RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS * params->qsize[i] * sizeof(struct rte_mbuf *); + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + if (i < RTE_SCHED_TRAFFIC_CLASS_BE) + size_per_pipe_queue_array += + params->qsize[i] * sizeof(struct rte_mbuf *); + else + size_per_pipe_queue_array += RTE_SCHED_MAX_QUEUES_PER_TC * + params->qsize[i] * sizeof(struct rte_mbuf *); } - size_queue_array = n_pipes_per_port * size_per_pipe_queue_array; - + size_queue_array = n_pipes_per_subport * size_per_pipe_queue_array; + base = 0; - - if (array == e_RTE_SCHED_PORT_ARRAY_SUBPORT) return base; - base += CACHE_LINE_ROUNDUP(size_subport); - - if (array == e_RTE_SCHED_PORT_ARRAY_PIPE) return base; - base += CACHE_LINE_ROUNDUP(size_pipe); - - if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE) return base; - base += CACHE_LINE_ROUNDUP(size_queue); - - if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA) return base; - base += CACHE_LINE_ROUNDUP(size_queue_extra); - - if (array == e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES) return base; - base += CACHE_LINE_ROUNDUP(size_pipe_profiles); - - if (array == e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY) return base; - base += CACHE_LINE_ROUNDUP(size_bmp_array); - - if (array == e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY) return base; - base += CACHE_LINE_ROUNDUP(size_queue_array); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_PIPE) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_pipe); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_queue); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_EXTRA) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_queue_extra); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_PIPE_PROFILES) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_pipe_profiles); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_BMP_ARRAY) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_bmp_array); + + if (array == e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_ARRAY) + return base; + base += RTE_CACHE_LINE_ROUNDUP(size_queue_array); return base; } -uint32_t -rte_sched_port_get_memory_footprint(struct rte_sched_port_params *params) +static void +rte_sched_subport_config_qsize(struct rte_sched_subport *subport) { - uint32_t size0, size1; - int status; - - status = rte_sched_port_check_params(params); - if (status != 0) { - RTE_LOG(INFO, SCHED, "Port scheduler params check failed (%d)\n", status); - - return 0; - } - - size0 = sizeof(struct rte_sched_port); - size1 = rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_TOTAL); - - return (size0 + size1); + uint32_t i; + + subport->qsize_add[0] = 0; + + /* Strict prority traffic class */ + for (i = 1; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + subport->qsize_add[i] = subport->qsize_add[i-1] + subport->qsize[i-1]; + + /* Best-effort traffic class */ + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 1] = + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE] + + subport->qsize[RTE_SCHED_TRAFFIC_CLASS_BE]; + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 2] = + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 1] + + subport->qsize[RTE_SCHED_TRAFFIC_CLASS_BE]; + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 3] = + subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 2] + + subport->qsize[RTE_SCHED_TRAFFIC_CLASS_BE]; + + subport->qsize_sum = subport->qsize_add[RTE_SCHED_TRAFFIC_CLASS_BE + 3] + + subport->qsize[RTE_SCHED_TRAFFIC_CLASS_BE]; } static void -rte_sched_port_config_qsize(struct rte_sched_port *port) -{ - /* TC 0 */ - port->qsize_add[0] = 0; - port->qsize_add[1] = port->qsize_add[0] + port->qsize[0]; - port->qsize_add[2] = port->qsize_add[1] + port->qsize[0]; - port->qsize_add[3] = port->qsize_add[2] + port->qsize[0]; - - /* TC 1 */ - port->qsize_add[4] = port->qsize_add[3] + port->qsize[0]; - port->qsize_add[5] = port->qsize_add[4] + port->qsize[1]; - port->qsize_add[6] = port->qsize_add[5] + port->qsize[1]; - port->qsize_add[7] = port->qsize_add[6] + port->qsize[1]; - - /* TC 2 */ - port->qsize_add[8] = port->qsize_add[7] + port->qsize[1]; - port->qsize_add[9] = port->qsize_add[8] + port->qsize[2]; - port->qsize_add[10] = port->qsize_add[9] + port->qsize[2]; - port->qsize_add[11] = port->qsize_add[10] + port->qsize[2]; - - /* TC 3 */ - port->qsize_add[12] = port->qsize_add[11] + port->qsize[2]; - port->qsize_add[13] = port->qsize_add[12] + port->qsize[3]; - port->qsize_add[14] = port->qsize_add[13] + port->qsize[3]; - port->qsize_add[15] = port->qsize_add[14] + port->qsize[3]; - - port->qsize_sum = port->qsize_add[15] + port->qsize[3]; -} - -static void -rte_sched_port_log_pipe_profile(struct rte_sched_port *port, uint32_t i) -{ - struct rte_sched_pipe_profile *p = port->pipe_profiles + i; - - RTE_LOG(INFO, SCHED, "Low level config for pipe profile %u:\n" - "\tToken bucket: period = %u, credits per period = %u, size = %u\n" - "\tTraffic classes: period = %u, credits per period = [%u, %u, %u, %u], ov weights = [%hhu, %hhu, %hhu, %hhu]\n" - "\tWRR cost: [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu], [%hhu, %hhu, %hhu, %hhu]\n", +rte_sched_port_log_pipe_profile(struct rte_sched_subport *subport, uint32_t i) +{ + struct rte_sched_pipe_profile *p = subport->pipe_profiles + i; + + RTE_LOG(DEBUG, SCHED, "Low level config for pipe profile %u:\n" + " Token bucket: period = %"PRIu64", credits per period = %"PRIu64", size = %"PRIu64"\n" + " Traffic classes: period = %"PRIu64",\n" + " credits per period = [%"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64 + ", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64 + ", %"PRIu64", %"PRIu64", %"PRIu64"]\n" + " Best-effort traffic class oversubscription: weight = %hhu\n" + " WRR cost: [%hhu, %hhu, %hhu, %hhu]\n", i, - + /* Token bucket */ p->tb_period, p->tb_credits_per_period, p->tb_size, - + /* Traffic classes */ p->tc_period, p->tc_credits_per_period[0], p->tc_credits_per_period[1], p->tc_credits_per_period[2], p->tc_credits_per_period[3], - p->tc_ov_weight[0], - p->tc_ov_weight[1], - p->tc_ov_weight[2], - p->tc_ov_weight[3], - + p->tc_credits_per_period[4], + p->tc_credits_per_period[5], + p->tc_credits_per_period[6], + p->tc_credits_per_period[7], + p->tc_credits_per_period[8], + p->tc_credits_per_period[9], + p->tc_credits_per_period[10], + p->tc_credits_per_period[11], + p->tc_credits_per_period[12], + + /* Best-effort traffic class oversubscription */ + p->tc_ov_weight, + /* WRR */ - p->wrr_cost[ 0], p->wrr_cost[ 1], p->wrr_cost[ 2], p->wrr_cost[ 3], - p->wrr_cost[ 4], p->wrr_cost[ 5], p->wrr_cost[ 6], p->wrr_cost[ 7], - p->wrr_cost[ 8], p->wrr_cost[ 9], p->wrr_cost[10], p->wrr_cost[11], - p->wrr_cost[12], p->wrr_cost[13], p->wrr_cost[14], p->wrr_cost[15]); + p->wrr_cost[0], p->wrr_cost[1], p->wrr_cost[2], p->wrr_cost[3]); } static inline uint64_t -rte_sched_time_ms_to_bytes(uint32_t time_ms, uint32_t rate) +rte_sched_time_ms_to_bytes(uint64_t time_ms, uint64_t rate) { uint64_t time = time_ms; + time = (time * rate) / 1000; - + return time; } static void -rte_sched_port_config_pipe_profile_table(struct rte_sched_port *port, struct rte_sched_port_params *params) +rte_sched_pipe_profile_convert(struct rte_sched_subport *subport, + struct rte_sched_pipe_params *src, + struct rte_sched_pipe_profile *dst, + uint64_t rate) { - uint32_t i, j; - - for (i = 0; i < port->n_pipe_profiles; i ++) { + uint32_t wrr_cost[RTE_SCHED_BE_QUEUES_PER_PIPE]; + uint32_t lcd1, lcd2, lcd; + uint32_t i; + + /* Token Bucket */ + if (src->tb_rate == rate) { + dst->tb_credits_per_period = 1; + dst->tb_period = 1; + } else { + double tb_rate = (double) src->tb_rate + / (double) rate; + double d = RTE_SCHED_TB_RATE_CONFIG_ERR; + + rte_approx_64(tb_rate, d, &dst->tb_credits_per_period, + &dst->tb_period); + } + + dst->tb_size = src->tb_size; + + /* Traffic Classes */ + dst->tc_period = rte_sched_time_ms_to_bytes(src->tc_period, + rate); + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + if (subport->qsize[i]) + dst->tc_credits_per_period[i] + = rte_sched_time_ms_to_bytes(src->tc_period, + src->tc_rate[i]); + + dst->tc_ov_weight = src->tc_ov_weight; + + /* WRR queues */ + wrr_cost[0] = src->wrr_weights[0]; + wrr_cost[1] = src->wrr_weights[1]; + wrr_cost[2] = src->wrr_weights[2]; + wrr_cost[3] = src->wrr_weights[3]; + + lcd1 = rte_get_lcd(wrr_cost[0], wrr_cost[1]); + lcd2 = rte_get_lcd(wrr_cost[2], wrr_cost[3]); + lcd = rte_get_lcd(lcd1, lcd2); + + wrr_cost[0] = lcd / wrr_cost[0]; + wrr_cost[1] = lcd / wrr_cost[1]; + wrr_cost[2] = lcd / wrr_cost[2]; + wrr_cost[3] = lcd / wrr_cost[3]; + + dst->wrr_cost[0] = (uint8_t) wrr_cost[0]; + dst->wrr_cost[1] = (uint8_t) wrr_cost[1]; + dst->wrr_cost[2] = (uint8_t) wrr_cost[2]; + dst->wrr_cost[3] = (uint8_t) wrr_cost[3]; +} + +static void +rte_sched_subport_config_pipe_profile_table(struct rte_sched_subport *subport, + struct rte_sched_subport_params *params, uint64_t rate) +{ + uint32_t i; + + for (i = 0; i < subport->n_pipe_profiles; i++) { + struct rte_sched_pipe_params *src = params->pipe_profiles + i; + struct rte_sched_pipe_profile *dst = subport->pipe_profiles + i; + + rte_sched_pipe_profile_convert(subport, src, dst, rate); + rte_sched_port_log_pipe_profile(subport, i); + } + + subport->pipe_tc_be_rate_max = 0; + for (i = 0; i < subport->n_pipe_profiles; i++) { struct rte_sched_pipe_params *src = params->pipe_profiles + i; - struct rte_sched_pipe_profile *dst = port->pipe_profiles + i; - - /* Token Bucket */ - if (src->tb_rate == params->rate) { - dst->tb_credits_per_period = 1; - dst->tb_period = 1; - } else { - double tb_rate = ((double) src->tb_rate) / ((double) params->rate); - double d = RTE_SCHED_TB_RATE_CONFIG_ERR; - - rte_approx(tb_rate, d, &dst->tb_credits_per_period, &dst->tb_period); + uint64_t pipe_tc_be_rate = src->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE]; + + if (subport->pipe_tc_be_rate_max < pipe_tc_be_rate) + subport->pipe_tc_be_rate_max = pipe_tc_be_rate; + } +} + +static int +rte_sched_subport_check_params(struct rte_sched_subport_params *params, + uint32_t n_max_pipes_per_subport, + uint64_t rate) +{ + uint32_t i; + + /* Check user parameters */ + if (params == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter params\n", __func__); + return -EINVAL; + } + + if (params->tb_rate == 0 || params->tb_rate > rate) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tb rate\n", __func__); + return -EINVAL; + } + + if (params->tb_size == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tb size\n", __func__); + return -EINVAL; + } + + /* qsize: if non-zero, power of 2, + * no bigger than 32K (due to 16-bit read/write pointers) + */ + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + uint16_t qsize = params->qsize[i]; + + if (qsize != 0 && !rte_is_power_of_2(qsize)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for qsize\n", __func__); + return -EINVAL; } - dst->tb_size = src->tb_size; - - /* Traffic Classes */ - dst->tc_period = (uint32_t) rte_sched_time_ms_to_bytes(src->tc_period, params->rate); - for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j ++) { - dst->tc_credits_per_period[j] = (uint32_t) rte_sched_time_ms_to_bytes(src->tc_period, src->tc_rate[j]); + } + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + uint64_t tc_rate = params->tc_rate[i]; + uint16_t qsize = params->qsize[i]; + + if ((qsize == 0 && tc_rate != 0) || + (qsize != 0 && tc_rate == 0) || + (tc_rate > params->tb_rate)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tc rate\n", __func__); + return -EINVAL; } -#ifdef RTE_SCHED_SUBPORT_TC_OV - for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j ++) { - dst->tc_ov_weight[j] = src->tc_ov_weight[j]; + } + + if (params->qsize[RTE_SCHED_TRAFFIC_CLASS_BE] == 0 || + params->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE] == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect qsize or tc rate(best effort)\n", __func__); + return -EINVAL; + } + + if (params->tc_period == 0) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tc period\n", __func__); + return -EINVAL; + } + + /* n_pipes_per_subport: non-zero, power of 2 */ + if (params->n_pipes_per_subport_enabled == 0 || + params->n_pipes_per_subport_enabled > n_max_pipes_per_subport || + !rte_is_power_of_2(params->n_pipes_per_subport_enabled)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for pipes number\n", __func__); + return -EINVAL; + } + + /* pipe_profiles and n_pipe_profiles */ + if (params->pipe_profiles == NULL || + params->n_pipe_profiles == 0 || + params->n_max_pipe_profiles == 0 || + params->n_pipe_profiles > params->n_max_pipe_profiles) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for pipe profiles\n", __func__); + return -EINVAL; + } + + for (i = 0; i < params->n_pipe_profiles; i++) { + struct rte_sched_pipe_params *p = params->pipe_profiles + i; + int status; + + status = pipe_profile_check(p, rate, ¶ms->qsize[0]); + if (status != 0) { + RTE_LOG(ERR, SCHED, + "%s: Pipe profile check failed(%d)\n", __func__, status); + return -EINVAL; } -#endif - - /* WRR */ - for (j = 0; j < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; j ++) { - uint32_t wrr_cost[RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS]; - uint32_t lcd, lcd1, lcd2; - uint32_t qindex; - - qindex = j * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS; - - wrr_cost[0] = src->wrr_weights[qindex]; - wrr_cost[1] = src->wrr_weights[qindex + 1]; - wrr_cost[2] = src->wrr_weights[qindex + 2]; - wrr_cost[3] = src->wrr_weights[qindex + 3]; - - lcd1 = rte_get_lcd(wrr_cost[0], wrr_cost[1]); - lcd2 = rte_get_lcd(wrr_cost[2], wrr_cost[3]); - lcd = rte_get_lcd(lcd1, lcd2); - - wrr_cost[0] = lcd / wrr_cost[0]; - wrr_cost[1] = lcd / wrr_cost[1]; - wrr_cost[2] = lcd / wrr_cost[2]; - wrr_cost[3] = lcd / wrr_cost[3]; - - dst->wrr_cost[qindex] = (uint8_t) wrr_cost[0]; - dst->wrr_cost[qindex + 1] = (uint8_t) wrr_cost[1]; - dst->wrr_cost[qindex + 2] = (uint8_t) wrr_cost[2]; - dst->wrr_cost[qindex + 3] = (uint8_t) wrr_cost[3]; + } + + return 0; +} + +uint32_t +rte_sched_port_get_memory_footprint(struct rte_sched_port_params *port_params, + struct rte_sched_subport_params **subport_params) +{ + uint32_t size0 = 0, size1 = 0, i; + int status; + + status = rte_sched_port_check_params(port_params); + if (status != 0) { + RTE_LOG(ERR, SCHED, + "%s: Port scheduler port params check failed (%d)\n", + __func__, status); + + return 0; + } + + for (i = 0; i < port_params->n_subports_per_port; i++) { + struct rte_sched_subport_params *sp = subport_params[i]; + + status = rte_sched_subport_check_params(sp, + port_params->n_pipes_per_subport, + port_params->rate); + if (status != 0) { + RTE_LOG(ERR, SCHED, + "%s: Port scheduler subport params check failed (%d)\n", + __func__, status); + + return 0; } - - rte_sched_port_log_pipe_profile(port, i); } + + size0 = sizeof(struct rte_sched_port); + + for (i = 0; i < port_params->n_subports_per_port; i++) { + struct rte_sched_subport_params *sp = subport_params[i]; + + size1 += rte_sched_subport_get_array_base(sp, + e_RTE_SCHED_SUBPORT_ARRAY_TOTAL); + } + + return size0 + size1; } struct rte_sched_port * rte_sched_port_config(struct rte_sched_port_params *params) { struct rte_sched_port *port = NULL; - const struct rte_memzone *mz = NULL; - uint32_t mem_size, bmp_mem_size, n_queues_per_port, i; - - /* Check user parameters. Determine the amount of memory to allocate */ - mem_size = rte_sched_port_get_memory_footprint(params); - if (mem_size == 0) { + uint32_t size0, size1; + uint32_t cycles_per_byte; + uint32_t i, j; + int status; + + status = rte_sched_port_check_params(params); + if (status != 0) { + RTE_LOG(ERR, SCHED, + "%s: Port scheduler params check failed (%d)\n", + __func__, status); return NULL; } - + + size0 = sizeof(struct rte_sched_port); + size1 = params->n_subports_per_port * sizeof(struct rte_sched_subport *); + /* Allocate memory to store the data structures */ - mz = rte_memzone_lookup(params->name); - if (mz) { - /* Use existing memzone, provided that its size is big enough */ - if (mz->len < mem_size) { - return NULL; - } - } else { - /* Create new memzone */ - mz = rte_memzone_reserve(params->name, mem_size, params->socket, 0); - if (mz == NULL) { - return NULL; - } + port = rte_zmalloc_socket("qos_params", size0 + size1, RTE_CACHE_LINE_SIZE, + params->socket); + if (port == NULL) { + RTE_LOG(ERR, SCHED, "%s: Memory allocation fails\n", __func__); + + return NULL; } - memset(mz->addr, 0, mem_size); - port = (struct rte_sched_port *) mz->addr; /* User parameters */ port->n_subports_per_port = params->n_subports_per_port; port->n_pipes_per_subport = params->n_pipes_per_subport; + port->n_pipes_per_subport_log2 = + __builtin_ctz(params->n_pipes_per_subport); + port->socket = params->socket; + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + port->pipe_queue[i] = i; + + for (i = 0, j = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++) { + port->pipe_tc[i] = j; + + if (j < RTE_SCHED_TRAFFIC_CLASS_BE) + j++; + } + + for (i = 0, j = 0; i < RTE_SCHED_QUEUES_PER_PIPE; i++) { + port->tc_queue[i] = j; + + if (i >= RTE_SCHED_TRAFFIC_CLASS_BE) + j++; + } port->rate = params->rate; port->mtu = params->mtu + params->frame_overhead; port->frame_overhead = params->frame_overhead; - memcpy(port->qsize, params->qsize, sizeof(params->qsize)); - port->n_pipe_profiles = params->n_pipe_profiles; - -#ifdef RTE_SCHED_RED - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { - uint32_t j; - - for (j = 0; j < e_RTE_METER_COLORS; j++) { - if (rte_red_config_init(&port->red_config[i][j], - params->red_params[i][j].wq_log2, - params->red_params[i][j].min_th, - params->red_params[i][j].max_th, - params->red_params[i][j].maxp_inv) != 0) { - return NULL; - } - } - } -#endif /* Timing */ port->time_cpu_cycles = rte_get_tsc_cycles(); port->time_cpu_bytes = 0; port->time = 0; - port->cycles_per_byte = ((double) rte_get_tsc_hz()) / ((double) params->rate); - /* Scheduling loop detection */ - port->pipe_loop = RTE_SCHED_PIPE_INVALID; - port->pipe_exhaustion = 0; + cycles_per_byte = (rte_get_tsc_hz() << RTE_SCHED_TIME_SHIFT) + / params->rate; + port->inv_cycles_per_byte = rte_reciprocal_value(cycles_per_byte); + port->cycles_per_byte = cycles_per_byte; /* Grinders */ - port->busy_grinders = 0; port->pkts_out = NULL; port->n_pkts_out = 0; - - /* Queue base calculation */ - rte_sched_port_config_qsize(port); - - /* Large data structures */ - port->subport = (struct rte_sched_subport *) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_SUBPORT)); - port->pipe = (struct rte_sched_pipe *) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_PIPE)); - port->queue = (struct rte_sched_queue *) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_QUEUE)); - port->queue_extra = (struct rte_sched_queue_extra *) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_QUEUE_EXTRA)); - port->pipe_profiles = (struct rte_sched_pipe_profile *) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_PIPE_PROFILES)); - port->bmp_array = port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_BMP_ARRAY); - port->queue_array = (struct rte_mbuf **) (port->memory + rte_sched_port_get_array_base(params, e_RTE_SCHED_PORT_ARRAY_QUEUE_ARRAY)); + port->subport_id = 0; - /* Pipe profile table */ - rte_sched_port_config_pipe_profile_table(port, params); - - /* Bitmap */ - n_queues_per_port = rte_sched_port_queues_per_port(port); - bmp_mem_size = rte_bitmap_get_memory_footprint(n_queues_per_port); - port->bmp = rte_bitmap_init(n_queues_per_port, port->bmp_array, bmp_mem_size); - if (port->bmp == NULL) { - RTE_LOG(INFO, SCHED, "Bitmap init error\n"); - return NULL; - } - for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i ++) { - port->grinder_base_bmp_pos[i] = RTE_SCHED_PIPE_INVALID; - } - return port; } -void +static inline void +rte_sched_subport_free(struct rte_sched_port *port, + struct rte_sched_subport *subport) +{ + uint32_t n_subport_pipe_queues; + uint32_t qindex; + + if (subport == NULL) + return; + + n_subport_pipe_queues = rte_sched_subport_pipe_queues(subport); + + /* Free enqueued mbufs */ + for (qindex = 0; qindex < n_subport_pipe_queues; qindex++) { + struct rte_mbuf **mbufs = + rte_sched_subport_pipe_qbase(subport, qindex); + uint16_t qsize = rte_sched_subport_pipe_qsize(port, subport, qindex); + if (qsize != 0) { + struct rte_sched_queue *queue = subport->queue + qindex; + uint16_t qr = queue->qr & (qsize - 1); + uint16_t qw = queue->qw & (qsize - 1); + + for (; qr != qw; qr = (qr + 1) & (qsize - 1)) + rte_pktmbuf_free(mbufs[qr]); + } + } + + rte_free(subport); +} + +void rte_sched_port_free(struct rte_sched_port *port) { + uint32_t i; + /* Check user parameters */ - if (port == NULL){ + if (port == NULL) return; - } - rte_bitmap_free(port->bmp); - - return; + + for (i = 0; i < port->n_subports_per_port; i++) + rte_sched_subport_free(port, port->subports[i]); + + rte_free(port); } static void rte_sched_port_log_subport_config(struct rte_sched_port *port, uint32_t i) { - struct rte_sched_subport *s = port->subport + i; - - RTE_LOG(INFO, SCHED, "Low level config for subport %u:\n" - "\tToken bucket: period = %u, credits per period = %u, size = %u\n" - "\tTraffic classes: period = %u, credits per period = [%u, %u, %u, %u], ov period = %u\n", + struct rte_sched_subport *s = port->subports[i]; + + RTE_LOG(DEBUG, SCHED, "Low level config for subport %u:\n" + " Token bucket: period = %"PRIu64", credits per period = %"PRIu64 + ", size = %"PRIu64"\n" + " Traffic classes: period = %"PRIu64"\n" + " credits per period = [%"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64 + ", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64", %"PRIu64 + ", %"PRIu64", %"PRIu64", %"PRIu64"]\n" + " Best effort traffic class oversubscription: wm min = %"PRIu64 + ", wm max = %"PRIu64"\n", i, - + /* Token bucket */ s->tb_period, s->tb_credits_per_period, s->tb_size, - + /* Traffic classes */ s->tc_period, s->tc_credits_per_period[0], s->tc_credits_per_period[1], s->tc_credits_per_period[2], s->tc_credits_per_period[3], - s->tc_ov_period); + s->tc_credits_per_period[4], + s->tc_credits_per_period[5], + s->tc_credits_per_period[6], + s->tc_credits_per_period[7], + s->tc_credits_per_period[8], + s->tc_credits_per_period[9], + s->tc_credits_per_period[10], + s->tc_credits_per_period[11], + s->tc_credits_per_period[12], + + /* Best effort traffic class oversubscription */ + s->tc_ov_wm_min, + s->tc_ov_wm_max); +} + +static void +rte_sched_free_memory(struct rte_sched_port *port, uint32_t n_subports) +{ + uint32_t i; + + for (i = 0; i < n_subports; i++) { + struct rte_sched_subport *subport = port->subports[i]; + + rte_sched_subport_free(port, subport); + } + + rte_free(port); } int -rte_sched_subport_config(struct rte_sched_port *port, +rte_sched_subport_config(struct rte_sched_port *port, uint32_t subport_id, struct rte_sched_subport_params *params) { - struct rte_sched_subport *s; - uint32_t i; - + struct rte_sched_subport *s = NULL; + uint32_t n_subports = subport_id; + uint32_t n_subport_pipe_queues, i; + uint32_t size0, size1, bmp_mem_size; + int status; + /* Check user parameters */ - if ((port == NULL) || - (subport_id >= port->n_subports_per_port) || - (params == NULL)) { - return -1; - } - - if ((params->tb_rate == 0) || (params->tb_rate > port->rate)) { - return -2; - } - - if (params->tb_size == 0) { - return -3; + if (port == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter port\n", __func__); + return 0; } - - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - if ((params->tc_rate[i] == 0) || (params->tc_rate[i] > params->tb_rate)) { - return -4; - } + + if (subport_id >= port->n_subports_per_port) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for subport id\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; } - - if (params->tc_period == 0) { - return -5; + + status = rte_sched_subport_check_params(params, + port->n_pipes_per_subport, + port->rate); + if (status != 0) { + RTE_LOG(NOTICE, SCHED, + "%s: Port scheduler params check failed (%d)\n", + __func__, status); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; } -#ifdef RTE_SCHED_SUBPORT_TC_OV - if ((params->tc_ov_period == 0) || (params->tc_ov_period > params->tc_period)) { - return -6; + /* Determine the amount of memory to allocate */ + size0 = sizeof(struct rte_sched_subport); + size1 = rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_TOTAL); + + /* Allocate memory to store the data structures */ + s = rte_zmalloc_socket("subport_params", size0 + size1, + RTE_CACHE_LINE_SIZE, port->socket); + if (s == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Memory allocation fails\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -ENOMEM; } -#endif - - s = port->subport + subport_id; - + + n_subports++; + + /* Port */ + port->subports[subport_id] = s; + /* Token Bucket (TB) */ if (params->tb_rate == port->rate) { s->tb_credits_per_period = 1; @@ -791,150 +1030,404 @@ rte_sched_subport_config(struct rte_sched_port *port, } else { double tb_rate = ((double) params->tb_rate) / ((double) port->rate); double d = RTE_SCHED_TB_RATE_CONFIG_ERR; - - rte_approx(tb_rate, d, &s->tb_credits_per_period, &s->tb_period); + + rte_approx_64(tb_rate, d, &s->tb_credits_per_period, &s->tb_period); } + s->tb_size = params->tb_size; s->tb_time = port->time; s->tb_credits = s->tb_size / 2; - + /* Traffic Classes (TCs) */ - s->tc_period = (uint32_t) rte_sched_time_ms_to_bytes(params->tc_period, port->rate); - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - s->tc_credits_per_period[i] = (uint32_t) rte_sched_time_ms_to_bytes(params->tc_period, params->tc_rate[i]); + s->tc_period = rte_sched_time_ms_to_bytes(params->tc_period, port->rate); + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + if (params->qsize[i]) + s->tc_credits_per_period[i] + = rte_sched_time_ms_to_bytes(params->tc_period, + params->tc_rate[i]); } s->tc_time = port->time + s->tc_period; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - s->tc_credits[i] = s->tc_credits_per_period[i]; + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + if (params->qsize[i]) + s->tc_credits[i] = s->tc_credits_per_period[i]; + + /* compile time checks */ + RTE_BUILD_BUG_ON(RTE_SCHED_PORT_N_GRINDERS == 0); + RTE_BUILD_BUG_ON(RTE_SCHED_PORT_N_GRINDERS & + (RTE_SCHED_PORT_N_GRINDERS - 1)); + + /* User parameters */ + s->n_pipes_per_subport_enabled = params->n_pipes_per_subport_enabled; + memcpy(s->qsize, params->qsize, sizeof(params->qsize)); + s->n_pipe_profiles = params->n_pipe_profiles; + s->n_max_pipe_profiles = params->n_max_pipe_profiles; + +#ifdef RTE_SCHED_RED + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) { + uint32_t j; + + for (j = 0; j < RTE_COLORS; j++) { + /* if min/max are both zero, then RED is disabled */ + if ((params->red_params[i][j].min_th | + params->red_params[i][j].max_th) == 0) { + continue; + } + + if (rte_red_config_init(&s->red_config[i][j], + params->red_params[i][j].wq_log2, + params->red_params[i][j].min_th, + params->red_params[i][j].max_th, + params->red_params[i][j].maxp_inv) != 0) { + rte_sched_free_memory(port, n_subports); + + RTE_LOG(NOTICE, SCHED, + "%s: RED configuration init fails\n", __func__); + return -EINVAL; + } + } + } +#endif + + /* Scheduling loop detection */ + s->pipe_loop = RTE_SCHED_PIPE_INVALID; + s->pipe_exhaustion = 0; + + /* Grinders */ + s->busy_grinders = 0; + + /* Queue base calculation */ + rte_sched_subport_config_qsize(s); + + /* Large data structures */ + s->pipe = (struct rte_sched_pipe *) + (s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_PIPE)); + s->queue = (struct rte_sched_queue *) + (s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE)); + s->queue_extra = (struct rte_sched_queue_extra *) + (s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_EXTRA)); + s->pipe_profiles = (struct rte_sched_pipe_profile *) + (s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_PIPE_PROFILES)); + s->bmp_array = s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_BMP_ARRAY); + s->queue_array = (struct rte_mbuf **) + (s->memory + rte_sched_subport_get_array_base(params, + e_RTE_SCHED_SUBPORT_ARRAY_QUEUE_ARRAY)); + + /* Pipe profile table */ + rte_sched_subport_config_pipe_profile_table(s, params, port->rate); + + /* Bitmap */ + n_subport_pipe_queues = rte_sched_subport_pipe_queues(s); + bmp_mem_size = rte_bitmap_get_memory_footprint(n_subport_pipe_queues); + s->bmp = rte_bitmap_init(n_subport_pipe_queues, s->bmp_array, + bmp_mem_size); + if (s->bmp == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Subport bitmap init error\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; } - + + for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i++) + s->grinder_base_bmp_pos[i] = RTE_SCHED_PIPE_INVALID; + #ifdef RTE_SCHED_SUBPORT_TC_OV /* TC oversubscription */ - s->tc_ov_period = (uint32_t) rte_sched_time_ms_to_bytes(params->tc_ov_period, port->rate); - s->tc_ov_time = port->time + s->tc_ov_period; + s->tc_ov_wm_min = port->mtu; + s->tc_ov_wm_max = rte_sched_time_ms_to_bytes(params->tc_period, + s->pipe_tc_be_rate_max); + s->tc_ov_wm = s->tc_ov_wm_max; s->tc_ov_period_id = 0; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - s->tc_ov[i] = 0; - s->tc_ov_n[i] = 0; - s->tc_ov_rate[i] = 0; - s->tc_ov_credits[i] = 0; - } + s->tc_ov = 0; + s->tc_ov_n = 0; + s->tc_ov_rate = 0; #endif - + rte_sched_port_log_subport_config(port, subport_id); - + return 0; } int rte_sched_pipe_config(struct rte_sched_port *port, - uint32_t subport_id, + uint32_t subport_id, uint32_t pipe_id, int32_t pipe_profile) { struct rte_sched_subport *s; struct rte_sched_pipe *p; struct rte_sched_pipe_profile *params; + uint32_t n_subports = subport_id + 1; uint32_t deactivate, profile, i; - + /* Check user parameters */ profile = (uint32_t) pipe_profile; deactivate = (pipe_profile < 0); - if ((port == NULL) || - (subport_id >= port->n_subports_per_port) || - (pipe_id >= port->n_pipes_per_subport) || - ((!deactivate) && (profile >= port->n_pipe_profiles))) { - return -1; - } - - /* Check that subport configuration is valid */ - s = port->subport + subport_id; - if (s->tb_period == 0) { - return -2; - } - - p = port->pipe + (subport_id * port->n_pipes_per_subport + pipe_id); - + + if (port == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter port\n", __func__); + return -EINVAL; + } + + if (subport_id >= port->n_subports_per_port) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter subport id\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; + } + + s = port->subports[subport_id]; + if (pipe_id >= s->n_pipes_per_subport_enabled) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter pipe id\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; + } + + if (!deactivate && profile >= s->n_pipe_profiles) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter pipe profile\n", __func__); + + rte_sched_free_memory(port, n_subports); + return -EINVAL; + } + /* Handle the case when pipe already has a valid configuration */ + p = s->pipe + pipe_id; if (p->tb_time) { - params = port->pipe_profiles + p->profile; + params = s->pipe_profiles + p->profile; + + double subport_tc_be_rate = + (double) s->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] + / (double) s->tc_period; + double pipe_tc_be_rate = + (double) params->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] + / (double) params->tc_period; + uint32_t tc_be_ov = s->tc_ov; -#ifdef RTE_SCHED_SUBPORT_TC_OV /* Unplug pipe from its subport */ - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - s->tc_ov_n[i] -= params->tc_ov_weight[i]; - s->tc_ov_rate[i] -= ((double) params->tc_credits_per_period[i]) / ((double) params->tc_period); - s->tc_ov[i] = s->tc_ov_rate[i] > (((double) s->tc_credits_per_period[i]) / ((double) s->tc_period)); + s->tc_ov_n -= params->tc_ov_weight; + s->tc_ov_rate -= pipe_tc_be_rate; + s->tc_ov = s->tc_ov_rate > subport_tc_be_rate; + + if (s->tc_ov != tc_be_ov) { + RTE_LOG(DEBUG, SCHED, + "Subport %u Best-effort TC oversubscription is OFF (%.4lf >= %.4lf)\n", + subport_id, subport_tc_be_rate, s->tc_ov_rate); } -#endif - + /* Reset the pipe */ memset(p, 0, sizeof(struct rte_sched_pipe)); } - - if (deactivate) { + + if (deactivate) return 0; - } - + /* Apply the new pipe configuration */ p->profile = profile; - params = port->pipe_profiles + p->profile; + params = s->pipe_profiles + p->profile; /* Token Bucket (TB) */ p->tb_time = port->time; p->tb_credits = params->tb_size / 2; - + /* Traffic Classes (TCs) */ p->tc_time = port->time + params->tc_period; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - p->tc_credits[i] = params->tc_credits_per_period[i]; + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + if (s->qsize[i]) + p->tc_credits[i] = params->tc_credits_per_period[i]; + + { + /* Subport best effort tc oversubscription */ + double subport_tc_be_rate = + (double) s->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] + / (double) s->tc_period; + double pipe_tc_be_rate = + (double) params->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] + / (double) params->tc_period; + uint32_t tc_be_ov = s->tc_ov; + + s->tc_ov_n += params->tc_ov_weight; + s->tc_ov_rate += pipe_tc_be_rate; + s->tc_ov = s->tc_ov_rate > subport_tc_be_rate; + + if (s->tc_ov != tc_be_ov) { + RTE_LOG(DEBUG, SCHED, + "Subport %u Best effort TC oversubscription is ON (%.4lf < %.4lf)\n", + subport_id, subport_tc_be_rate, s->tc_ov_rate); + } + p->tc_ov_period_id = s->tc_ov_period_id; + p->tc_ov_credits = s->tc_ov_wm; } - -#ifdef RTE_SCHED_SUBPORT_TC_OV - /* Subport TC oversubscription */ - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - s->tc_ov_n[i] += params->tc_ov_weight[i]; - s->tc_ov_rate[i] += ((double) params->tc_credits_per_period[i]) / ((double) params->tc_period); - s->tc_ov[i] = s->tc_ov_rate[i] > (((double) s->tc_credits_per_period[i]) / ((double) s->tc_period)); + + return 0; +} + +int +rte_sched_subport_pipe_profile_add(struct rte_sched_port *port, + uint32_t subport_id, + struct rte_sched_pipe_params *params, + uint32_t *pipe_profile_id) +{ + struct rte_sched_subport *s; + struct rte_sched_pipe_profile *pp; + uint32_t i; + int status; + + /* Port */ + if (port == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter port\n", __func__); + return -EINVAL; } - p->tc_ov_period_id = s->tc_ov_period_id; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - p->tc_ov_credits[i] = 0; + + /* Subport id not exceeds the max limit */ + if (subport_id > port->n_subports_per_port) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for subport id\n", __func__); + return -EINVAL; } -#endif - + + s = port->subports[subport_id]; + + /* Pipe profiles exceeds the max limit */ + if (s->n_pipe_profiles >= s->n_max_pipe_profiles) { + RTE_LOG(ERR, SCHED, + "%s: Number of pipe profiles exceeds the max limit\n", __func__); + return -EINVAL; + } + + /* Pipe params */ + status = pipe_profile_check(params, port->rate, &s->qsize[0]); + if (status != 0) { + RTE_LOG(ERR, SCHED, + "%s: Pipe profile check failed(%d)\n", __func__, status); + return -EINVAL; + } + + pp = &s->pipe_profiles[s->n_pipe_profiles]; + rte_sched_pipe_profile_convert(s, params, pp, port->rate); + + /* Pipe profile should not exists */ + for (i = 0; i < s->n_pipe_profiles; i++) + if (memcmp(s->pipe_profiles + i, pp, sizeof(*pp)) == 0) { + RTE_LOG(ERR, SCHED, + "%s: Pipe profile exists\n", __func__); + return -EINVAL; + } + + /* Pipe profile commit */ + *pipe_profile_id = s->n_pipe_profiles; + s->n_pipe_profiles++; + + if (s->pipe_tc_be_rate_max < params->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE]) + s->pipe_tc_be_rate_max = params->tc_rate[RTE_SCHED_TRAFFIC_CLASS_BE]; + + rte_sched_port_log_pipe_profile(s, *pipe_profile_id); + return 0; } +static inline uint32_t +rte_sched_port_qindex(struct rte_sched_port *port, + uint32_t subport, + uint32_t pipe, + uint32_t traffic_class, + uint32_t queue) +{ + return ((subport & (port->n_subports_per_port - 1)) << + (port->n_pipes_per_subport_log2 + 4)) | + ((pipe & + (port->subports[subport]->n_pipes_per_subport_enabled - 1)) << 4) | + ((rte_sched_port_pipe_queue(port, traffic_class) + queue) & + (RTE_SCHED_QUEUES_PER_PIPE - 1)); +} + +void +rte_sched_port_pkt_write(struct rte_sched_port *port, + struct rte_mbuf *pkt, + uint32_t subport, uint32_t pipe, + uint32_t traffic_class, + uint32_t queue, enum rte_color color) +{ + uint32_t queue_id = + rte_sched_port_qindex(port, subport, pipe, traffic_class, queue); + + rte_mbuf_sched_set(pkt, queue_id, traffic_class, (uint8_t)color); +} + +void +rte_sched_port_pkt_read_tree_path(struct rte_sched_port *port, + const struct rte_mbuf *pkt, + uint32_t *subport, uint32_t *pipe, + uint32_t *traffic_class, uint32_t *queue) +{ + uint32_t queue_id = rte_mbuf_sched_queue_get(pkt); + + *subport = queue_id >> (port->n_pipes_per_subport_log2 + 4); + *pipe = (queue_id >> 4) & + (port->subports[*subport]->n_pipes_per_subport_enabled - 1); + *traffic_class = rte_sched_port_pipe_tc(port, queue_id); + *queue = rte_sched_port_tc_queue(port, queue_id); +} + +enum rte_color +rte_sched_port_pkt_read_color(const struct rte_mbuf *pkt) +{ + return (enum rte_color)rte_mbuf_sched_color_get(pkt); +} + int rte_sched_subport_read_stats(struct rte_sched_port *port, - uint32_t subport_id, - struct rte_sched_subport_stats *stats, - uint32_t *tc_ov) + uint32_t subport_id, + struct rte_sched_subport_stats *stats, + uint32_t *tc_ov) { struct rte_sched_subport *s; - uint32_t mask, i; - + /* Check user parameters */ - if ((port == NULL) || - (subport_id >= port->n_subports_per_port) || - (stats == NULL) || - (tc_ov == NULL)) { - return -1; + if (port == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter port\n", __func__); + return -EINVAL; + } + + if (subport_id >= port->n_subports_per_port) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for subport id\n", __func__); + return -EINVAL; + } + + if (stats == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter stats\n", __func__); + return -EINVAL; } - s = port->subport + subport_id; + + if (tc_ov == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for tc_ov\n", __func__); + return -EINVAL; + } + + s = port->subports[subport_id]; /* Copy subport stats and clear */ memcpy(stats, &s->stats, sizeof(struct rte_sched_subport_stats)); memset(&s->stats, 0, sizeof(struct rte_sched_subport_stats)); - - /* Subport TC ovesubscription status */ - mask = 0; - for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i ++) { - mask |= ((uint32_t) s->tc_ov[i]) << i; - } - *tc_ov = mask; - + + /* Subport TC oversubscription status */ + *tc_ov = s->tc_ov; + return 0; } @@ -944,76 +1437,62 @@ rte_sched_queue_read_stats(struct rte_sched_port *port, struct rte_sched_queue_stats *stats, uint16_t *qlen) { + struct rte_sched_subport *s; struct rte_sched_queue *q; struct rte_sched_queue_extra *qe; - + uint32_t subport_id, subport_qmask, subport_qindex; + /* Check user parameters */ - if ((port == NULL) || - (queue_id >= rte_sched_port_queues_per_port(port)) || - (stats == NULL) || - (qlen == NULL)) { - return -1; + if (port == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter port\n", __func__); + return -EINVAL; + } + + if (queue_id >= rte_sched_port_queues_per_port(port)) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for queue id\n", __func__); + return -EINVAL; + } + + if (stats == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter stats\n", __func__); + return -EINVAL; + } + + if (qlen == NULL) { + RTE_LOG(ERR, SCHED, + "%s: Incorrect value for parameter qlen\n", __func__); + return -EINVAL; } - q = port->queue + queue_id; - qe = port->queue_extra + queue_id; + subport_qmask = port->n_pipes_per_subport_log2 + 4; + subport_id = (queue_id >> subport_qmask) & (port->n_subports_per_port - 1); + + s = port->subports[subport_id]; + subport_qindex = ((1 << subport_qmask) - 1) & queue_id; + q = s->queue + subport_qindex; + qe = s->queue_extra + subport_qindex; /* Copy queue stats and clear */ memcpy(stats, &qe->stats, sizeof(struct rte_sched_queue_stats)); memset(&qe->stats, 0, sizeof(struct rte_sched_queue_stats)); - + /* Queue length */ *qlen = q->qw - q->qr; - - return 0; -} - -static inline uint32_t -rte_sched_port_qindex(struct rte_sched_port *port, uint32_t subport, uint32_t pipe, uint32_t traffic_class, uint32_t queue) -{ - uint32_t result; - - result = subport * port->n_pipes_per_subport + pipe; - result = result * RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE + traffic_class; - result = result * RTE_SCHED_QUEUES_PER_TRAFFIC_CLASS + queue; - - return result; -} - -static inline struct rte_mbuf ** -rte_sched_port_qbase(struct rte_sched_port *port, uint32_t qindex) -{ - uint32_t pindex = qindex >> 4; - uint32_t qpos = qindex & 0xF; - - return (port->queue_array + pindex * port->qsize_sum + port->qsize_add[qpos]); -} -static inline uint16_t -rte_sched_port_qsize(struct rte_sched_port *port, uint32_t qindex) -{ - uint32_t tc = (qindex >> 2) & 0x3; - - return port->qsize[tc]; + return 0; } -#if RTE_SCHED_DEBUG +#ifdef RTE_SCHED_DEBUG static inline int -rte_sched_port_queue_is_empty(struct rte_sched_port *port, uint32_t qindex) +rte_sched_port_queue_is_empty(struct rte_sched_subport *subport, + uint32_t qindex) { - struct rte_sched_queue *queue = port->queue + qindex; - - return (queue->qr == queue->qw); -} + struct rte_sched_queue *queue = subport->queue + qindex; -static inline int -rte_sched_port_queue_is_full(struct rte_sched_port *port, uint32_t qindex) -{ - struct rte_sched_queue *queue = port->queue + qindex; - uint16_t qsize = rte_sched_port_qsize(port, qindex); - uint16_t qlen = q->qw - q->qr; - - return (qlen >= qsize); + return queue->qr == queue->qw; } #endif /* RTE_SCHED_DEBUG */ @@ -1021,45 +1500,78 @@ rte_sched_port_queue_is_full(struct rte_sched_port *port, uint32_t qindex) #ifdef RTE_SCHED_COLLECT_STATS static inline void -rte_sched_port_update_subport_stats(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt) +rte_sched_port_update_subport_stats(struct rte_sched_port *port, + struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt) { - struct rte_sched_subport *s = port->subport + (qindex / rte_sched_port_queues_per_subport(port)); - uint32_t tc_index = (qindex >> 2) & 0x3; - uint32_t pkt_len = pkt->pkt.pkt_len; - - s->stats.n_pkts_tc[tc_index] += 1; - s->stats.n_bytes_tc[tc_index] += pkt_len; + uint32_t tc_index = rte_sched_port_pipe_tc(port, qindex); + uint32_t pkt_len = pkt->pkt_len; + + subport->stats.n_pkts_tc[tc_index] += 1; + subport->stats.n_bytes_tc[tc_index] += pkt_len; } +#ifdef RTE_SCHED_RED +static inline void +rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port, + struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt, + uint32_t red) +#else static inline void -rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt) +rte_sched_port_update_subport_stats_on_drop(struct rte_sched_port *port, + struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt, + __rte_unused uint32_t red) +#endif { - struct rte_sched_subport *s = port->subport + (qindex / rte_sched_port_queues_per_subport(port)); - uint32_t tc_index = (qindex >> 2) & 0x3; - uint32_t pkt_len = pkt->pkt.pkt_len; - - s->stats.n_pkts_tc_dropped[tc_index] += 1; - s->stats.n_bytes_tc_dropped[tc_index] += pkt_len; + uint32_t tc_index = rte_sched_port_pipe_tc(port, qindex); + uint32_t pkt_len = pkt->pkt_len; + + subport->stats.n_pkts_tc_dropped[tc_index] += 1; + subport->stats.n_bytes_tc_dropped[tc_index] += pkt_len; +#ifdef RTE_SCHED_RED + subport->stats.n_pkts_red_dropped[tc_index] += red; +#endif } static inline void -rte_sched_port_update_queue_stats(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt) +rte_sched_port_update_queue_stats(struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt) { - struct rte_sched_queue_extra *qe = port->queue_extra + qindex; - uint32_t pkt_len = pkt->pkt.pkt_len; - + struct rte_sched_queue_extra *qe = subport->queue_extra + qindex; + uint32_t pkt_len = pkt->pkt_len; + qe->stats.n_pkts += 1; qe->stats.n_bytes += pkt_len; } +#ifdef RTE_SCHED_RED +static inline void +rte_sched_port_update_queue_stats_on_drop(struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt, + uint32_t red) +#else static inline void -rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf *pkt) +rte_sched_port_update_queue_stats_on_drop(struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf *pkt, + __rte_unused uint32_t red) +#endif { - struct rte_sched_queue_extra *qe = port->queue_extra + qindex; - uint32_t pkt_len = pkt->pkt.pkt_len; - + struct rte_sched_queue_extra *qe = subport->queue_extra + qindex; + uint32_t pkt_len = pkt->pkt_len; + qe->stats.n_pkts_dropped += 1; qe->stats.n_bytes_dropped += pkt_len; +#ifdef RTE_SCHED_RED + qe->stats.n_pkts_red_dropped += red; +#endif } #endif /* RTE_SCHED_COLLECT_STATS */ @@ -1067,282 +1579,284 @@ rte_sched_port_update_queue_stats_on_drop(struct rte_sched_port *port, uint32_t #ifdef RTE_SCHED_RED static inline int -rte_sched_port_red_drop(struct rte_sched_port *port, struct rte_mbuf *pkt, uint32_t qindex, uint16_t qlen) +rte_sched_port_red_drop(struct rte_sched_port *port, + struct rte_sched_subport *subport, + struct rte_mbuf *pkt, + uint32_t qindex, + uint16_t qlen) { struct rte_sched_queue_extra *qe; struct rte_red_config *red_cfg; - struct rte_red *red; + struct rte_red *red; uint32_t tc_index; - enum rte_meter_color color; - - tc_index = (qindex >> 2) & 0x3; + enum rte_color color; + + tc_index = rte_sched_port_pipe_tc(port, qindex); color = rte_sched_port_pkt_read_color(pkt); - red_cfg = &port->red_config[tc_index][color]; + red_cfg = &subport->red_config[tc_index][color]; + + if ((red_cfg->min_th | red_cfg->max_th) == 0) + return 0; - qe = port->queue_extra + qindex; + qe = subport->queue_extra + qindex; red = &qe->red; return rte_red_enqueue(red_cfg, red, qlen, port->time); } static inline void -rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port, uint32_t qindex) +rte_sched_port_set_queue_empty_timestamp(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t qindex) { - struct rte_sched_queue_extra *qe; - struct rte_red *red; - - qe = port->queue_extra + qindex; - red = &qe->red; + struct rte_sched_queue_extra *qe = subport->queue_extra + qindex; + struct rte_red *red = &qe->red; rte_red_mark_queue_empty(red, port->time); } #else -#define rte_sched_port_red_drop(port, pkt, qindex, qlen) 0 +static inline int rte_sched_port_red_drop(struct rte_sched_port *port __rte_unused, + struct rte_sched_subport *subport __rte_unused, + struct rte_mbuf *pkt __rte_unused, + uint32_t qindex __rte_unused, + uint16_t qlen __rte_unused) +{ + return 0; +} -#define rte_sched_port_set_queue_empty_timestamp(port, qindex) +#define rte_sched_port_set_queue_empty_timestamp(port, subport, qindex) #endif /* RTE_SCHED_RED */ -#if RTE_SCHED_DEBUG - -static inline int -debug_pipe_is_empty(struct rte_sched_port *port, uint32_t pindex) -{ - uint32_t qindex, i; - - qindex = pindex << 4; - - for (i = 0; i < 16; i ++){ - uint32_t queue_empty = rte_sched_port_queue_is_empty(port, qindex + i); - uint32_t bmp_bit_clear = (rte_bitmap_get(port->bmp, qindex + i) == 0); - - if (queue_empty != bmp_bit_clear){ - rte_panic("Queue status mismatch for queue %u of pipe %u\n", i, pindex); - } - - if (!queue_empty){ - return 0; - } - } - - return 1; -} +#ifdef RTE_SCHED_DEBUG static inline void -debug_check_queue_slab(struct rte_sched_port *port, uint32_t bmp_pos, uint64_t bmp_slab) +debug_check_queue_slab(struct rte_sched_subport *subport, uint32_t bmp_pos, + uint64_t bmp_slab) { uint64_t mask; uint32_t i, panic; - - if (bmp_slab == 0){ + + if (bmp_slab == 0) rte_panic("Empty slab at position %u\n", bmp_pos); - } - + panic = 0; - for (i = 0, mask = 1; i < 64; i ++, mask <<= 1) { - if (mask & bmp_slab){ - if (rte_sched_port_queue_is_empty(port, bmp_pos + i)) { + for (i = 0, mask = 1; i < 64; i++, mask <<= 1) { + if (mask & bmp_slab) { + if (rte_sched_port_queue_is_empty(subport, bmp_pos + i)) { printf("Queue %u (slab offset %u) is empty\n", bmp_pos + i, i); panic = 1; } } } - - if (panic){ + + if (panic) rte_panic("Empty queues in slab 0x%" PRIx64 "starting at position %u\n", bmp_slab, bmp_pos); - } } #endif /* RTE_SCHED_DEBUG */ +static inline struct rte_sched_subport * +rte_sched_port_subport(struct rte_sched_port *port, + struct rte_mbuf *pkt) +{ + uint32_t queue_id = rte_mbuf_sched_queue_get(pkt); + uint32_t subport_id = queue_id >> (port->n_pipes_per_subport_log2 + 4); + + return port->subports[subport_id]; +} + static inline uint32_t -rte_sched_port_enqueue_qptrs_prefetch0(struct rte_sched_port *port, struct rte_mbuf *pkt) +rte_sched_port_enqueue_qptrs_prefetch0(struct rte_sched_subport *subport, + struct rte_mbuf *pkt, uint32_t subport_qmask) { struct rte_sched_queue *q; #ifdef RTE_SCHED_COLLECT_STATS struct rte_sched_queue_extra *qe; #endif - uint32_t subport, pipe, traffic_class, queue, qindex; + uint32_t qindex = rte_mbuf_sched_queue_get(pkt); + uint32_t subport_queue_id = subport_qmask & qindex; - rte_sched_port_pkt_read_tree_path(pkt, &subport, &pipe, &traffic_class, &queue); - - qindex = rte_sched_port_qindex(port, subport, pipe, traffic_class, queue); - q = port->queue + qindex; + q = subport->queue + subport_queue_id; rte_prefetch0(q); #ifdef RTE_SCHED_COLLECT_STATS - qe = port->queue_extra + qindex; + qe = subport->queue_extra + subport_queue_id; rte_prefetch0(qe); #endif - - return qindex; + + return subport_queue_id; } static inline void -rte_sched_port_enqueue_qwa_prefetch0(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf **qbase) -{ +rte_sched_port_enqueue_qwa_prefetch0(struct rte_sched_port *port, + struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf **qbase) +{ struct rte_sched_queue *q; struct rte_mbuf **q_qw; - uint16_t qsize; - - q = port->queue + qindex; - qsize = rte_sched_port_qsize(port, qindex); + uint16_t qsize; + + q = subport->queue + qindex; + qsize = rte_sched_subport_pipe_qsize(port, subport, qindex); q_qw = qbase + (q->qw & (qsize - 1)); - + rte_prefetch0(q_qw); - rte_bitmap_prefetch0(port->bmp, qindex); + rte_bitmap_prefetch0(subport->bmp, qindex); } static inline int -rte_sched_port_enqueue_qwa(struct rte_sched_port *port, uint32_t qindex, struct rte_mbuf **qbase, struct rte_mbuf *pkt) +rte_sched_port_enqueue_qwa(struct rte_sched_port *port, + struct rte_sched_subport *subport, + uint32_t qindex, + struct rte_mbuf **qbase, + struct rte_mbuf *pkt) { struct rte_sched_queue *q; uint16_t qsize; uint16_t qlen; - q = port->queue + qindex; - qsize = rte_sched_port_qsize(port, qindex); + q = subport->queue + qindex; + qsize = rte_sched_subport_pipe_qsize(port, subport, qindex); qlen = q->qw - q->qr; /* Drop the packet (and update drop stats) when queue is full */ - if (unlikely(rte_sched_port_red_drop(port, pkt, qindex, qlen) || (qlen >= qsize))) { + if (unlikely(rte_sched_port_red_drop(port, subport, pkt, qindex, qlen) || + (qlen >= qsize))) { rte_pktmbuf_free(pkt); #ifdef RTE_SCHED_COLLECT_STATS - rte_sched_port_update_subport_stats_on_drop(port, qindex, pkt); - rte_sched_port_update_queue_stats_on_drop(port, qindex, pkt); + rte_sched_port_update_subport_stats_on_drop(port, subport, + qindex, pkt, qlen < qsize); + rte_sched_port_update_queue_stats_on_drop(subport, qindex, pkt, + qlen < qsize); #endif return 0; } - + /* Enqueue packet */ qbase[q->qw & (qsize - 1)] = pkt; - q->qw ++; - - /* Activate queue in the port bitmap */ - rte_bitmap_set(port->bmp, qindex); - + q->qw++; + + /* Activate queue in the subport bitmap */ + rte_bitmap_set(subport->bmp, qindex); + /* Statistics */ #ifdef RTE_SCHED_COLLECT_STATS - rte_sched_port_update_subport_stats(port, qindex, pkt); - rte_sched_port_update_queue_stats(port, qindex, pkt); + rte_sched_port_update_subport_stats(port, subport, qindex, pkt); + rte_sched_port_update_queue_stats(subport, qindex, pkt); #endif return 1; } -#if RTE_SCHED_ENQUEUE == 0 - -int -rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts) -{ - uint32_t result, i; - - result = 0; - - for (i = 0; i < n_pkts; i ++) { - struct rte_mbuf *pkt; - struct rte_mbuf **q_base; - uint32_t subport, pipe, traffic_class, queue, qindex; - - pkt = pkts[i]; - - rte_sched_port_pkt_read_tree_path(pkt, &subport, &pipe, &traffic_class, &queue); - - qindex = rte_sched_port_qindex(port, subport, pipe, traffic_class, queue); - - q_base = rte_sched_port_qbase(port, qindex); - - result += rte_sched_port_enqueue_qwa(port, qindex, q_base, pkt); - } - - return result; -} - -#else -/* The enqueue function implements a 4-level pipeline with each stage processing - * two different packets. The purpose of using a pipeline is to hide the latency - * of prefetching the data structures. The naming convention is presented in the - * diagram below: - * - * p00 _______ p10 _______ p20 _______ p30 _______ +/* + * The enqueue function implements a 4-level pipeline with each stage + * processing two different packets. The purpose of using a pipeline + * is to hide the latency of prefetching the data structures. The + * naming convention is presented in the diagram below: + * + * p00 _______ p10 _______ p20 _______ p30 _______ * ----->| |----->| |----->| |----->| |-----> - * | 0 | | 1 | | 2 | | 3 | + * | 0 | | 1 | | 2 | | 3 | * ----->|_______|----->|_______|----->|_______|----->|_______|-----> - * p01 p11 p21 p31 + * p01 p11 p21 p31 * - ***/ + */ int -rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts) +rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts, + uint32_t n_pkts) { - struct rte_mbuf *pkt00, *pkt01, *pkt10, *pkt11, *pkt20, *pkt21, *pkt30, *pkt31, *pkt_last; - struct rte_mbuf **q00_base, **q01_base, **q10_base, **q11_base, **q20_base, **q21_base, **q30_base, **q31_base, **q_last_base; + struct rte_mbuf *pkt00, *pkt01, *pkt10, *pkt11, *pkt20, *pkt21, + *pkt30, *pkt31, *pkt_last; + struct rte_mbuf **q00_base, **q01_base, **q10_base, **q11_base, + **q20_base, **q21_base, **q30_base, **q31_base, **q_last_base; + struct rte_sched_subport *subport00, *subport01, *subport10, *subport11, + *subport20, *subport21, *subport30, *subport31, *subport_last; uint32_t q00, q01, q10, q11, q20, q21, q30, q31, q_last; uint32_t r00, r01, r10, r11, r20, r21, r30, r31, r_last; + uint32_t subport_qmask; uint32_t result, i; - + result = 0; - - /* Less then 6 input packets available, which is not enough to feed the pipeline */ + subport_qmask = (1 << (port->n_pipes_per_subport_log2 + 4)) - 1; + + /* + * Less then 6 input packets available, which is not enough to + * feed the pipeline + */ if (unlikely(n_pkts < 6)) { + struct rte_sched_subport *subports[5]; struct rte_mbuf **q_base[5]; uint32_t q[5]; - + /* Prefetch the mbuf structure of each packet */ - for (i = 0; i < n_pkts; i ++) { + for (i = 0; i < n_pkts; i++) rte_prefetch0(pkts[i]); - } - + + /* Prefetch the subport structure for each packet */ + for (i = 0; i < n_pkts; i++) + subports[i] = rte_sched_port_subport(port, pkts[i]); + /* Prefetch the queue structure for each queue */ - for (i = 0; i < n_pkts; i ++) { - q[i] = rte_sched_port_enqueue_qptrs_prefetch0(port, pkts[i]); - } - + for (i = 0; i < n_pkts; i++) + q[i] = rte_sched_port_enqueue_qptrs_prefetch0(subports[i], + pkts[i], subport_qmask); + /* Prefetch the write pointer location of each queue */ - for (i = 0; i < n_pkts; i ++) { - q_base[i] = rte_sched_port_qbase(port, q[i]); - rte_sched_port_enqueue_qwa_prefetch0(port, q[i], q_base[i]); + for (i = 0; i < n_pkts; i++) { + q_base[i] = rte_sched_subport_pipe_qbase(subports[i], q[i]); + rte_sched_port_enqueue_qwa_prefetch0(port, subports[i], + q[i], q_base[i]); } - + /* Write each packet to its queue */ - for (i = 0; i < n_pkts; i ++) { - result += rte_sched_port_enqueue_qwa(port, q[i], q_base[i], pkts[i]); - } - + for (i = 0; i < n_pkts; i++) + result += rte_sched_port_enqueue_qwa(port, subports[i], + q[i], q_base[i], pkts[i]); + return result; } - + /* Feed the first 3 stages of the pipeline (6 packets needed) */ pkt20 = pkts[0]; pkt21 = pkts[1]; rte_prefetch0(pkt20); rte_prefetch0(pkt21); - + pkt10 = pkts[2]; pkt11 = pkts[3]; rte_prefetch0(pkt10); rte_prefetch0(pkt11); - q20 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt20); - q21 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt21); + subport20 = rte_sched_port_subport(port, pkt20); + subport21 = rte_sched_port_subport(port, pkt21); + q20 = rte_sched_port_enqueue_qptrs_prefetch0(subport20, + pkt20, subport_qmask); + q21 = rte_sched_port_enqueue_qptrs_prefetch0(subport21, + pkt21, subport_qmask); pkt00 = pkts[4]; pkt01 = pkts[5]; rte_prefetch0(pkt00); rte_prefetch0(pkt01); - - q10 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt10); - q11 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt11); - - q20_base = rte_sched_port_qbase(port, q20); - q21_base = rte_sched_port_qbase(port, q21); - rte_sched_port_enqueue_qwa_prefetch0(port, q20, q20_base); - rte_sched_port_enqueue_qwa_prefetch0(port, q21, q21_base); - + + subport10 = rte_sched_port_subport(port, pkt10); + subport11 = rte_sched_port_subport(port, pkt11); + q10 = rte_sched_port_enqueue_qptrs_prefetch0(subport10, + pkt10, subport_qmask); + q11 = rte_sched_port_enqueue_qptrs_prefetch0(subport11, + pkt11, subport_qmask); + + q20_base = rte_sched_subport_pipe_qbase(subport20, q20); + q21_base = rte_sched_subport_pipe_qbase(subport21, q21); + rte_sched_port_enqueue_qwa_prefetch0(port, subport20, q20, q20_base); + rte_sched_port_enqueue_qwa_prefetch0(port, subport21, q21, q21_base); + /* Run the pipeline */ - for (i = 6; i < (n_pkts & (~1)); i += 2) { + for (i = 6; i < (n_pkts & (~1)); i += 2) { /* Propagate stage inputs */ pkt30 = pkt20; pkt31 = pkt21; @@ -1354,226 +1868,263 @@ rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint q31 = q21; q20 = q10; q21 = q11; + subport30 = subport20; + subport31 = subport21; + subport20 = subport10; + subport21 = subport11; q30_base = q20_base; q31_base = q21_base; - + /* Stage 0: Get packets in */ pkt00 = pkts[i]; pkt01 = pkts[i + 1]; rte_prefetch0(pkt00); rte_prefetch0(pkt01); - - /* Stage 1: Prefetch queue structure storing queue pointers */ - q10 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt10); - q11 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt11); - + + /* Stage 1: Prefetch subport and queue structure storing queue pointers */ + subport10 = rte_sched_port_subport(port, pkt10); + subport11 = rte_sched_port_subport(port, pkt11); + q10 = rte_sched_port_enqueue_qptrs_prefetch0(subport10, + pkt10, subport_qmask); + q11 = rte_sched_port_enqueue_qptrs_prefetch0(subport11, + pkt11, subport_qmask); + /* Stage 2: Prefetch queue write location */ - q20_base = rte_sched_port_qbase(port, q20); - q21_base = rte_sched_port_qbase(port, q21); - rte_sched_port_enqueue_qwa_prefetch0(port, q20, q20_base); - rte_sched_port_enqueue_qwa_prefetch0(port, q21, q21_base); - + q20_base = rte_sched_subport_pipe_qbase(subport20, q20); + q21_base = rte_sched_subport_pipe_qbase(subport21, q21); + rte_sched_port_enqueue_qwa_prefetch0(port, subport20, q20, q20_base); + rte_sched_port_enqueue_qwa_prefetch0(port, subport21, q21, q21_base); + /* Stage 3: Write packet to queue and activate queue */ - r30 = rte_sched_port_enqueue_qwa(port, q30, q30_base, pkt30); - r31 = rte_sched_port_enqueue_qwa(port, q31, q31_base, pkt31); + r30 = rte_sched_port_enqueue_qwa(port, subport30, + q30, q30_base, pkt30); + r31 = rte_sched_port_enqueue_qwa(port, subport31, + q31, q31_base, pkt31); result += r30 + r31; } - - /* Drain the pipeline (exactly 6 packets). Handle the last packet in the case - of an odd number of input packets. */ + + /* + * Drain the pipeline (exactly 6 packets). + * Handle the last packet in the case + * of an odd number of input packets. + */ pkt_last = pkts[n_pkts - 1]; rte_prefetch0(pkt_last); - - q00 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt00); - q01 = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt01); - - q10_base = rte_sched_port_qbase(port, q10); - q11_base = rte_sched_port_qbase(port, q11); - rte_sched_port_enqueue_qwa_prefetch0(port, q10, q10_base); - rte_sched_port_enqueue_qwa_prefetch0(port, q11, q11_base); - - r20 = rte_sched_port_enqueue_qwa(port, q20, q20_base, pkt20); - r21 = rte_sched_port_enqueue_qwa(port, q21, q21_base, pkt21); + + subport00 = rte_sched_port_subport(port, pkt00); + subport01 = rte_sched_port_subport(port, pkt01); + q00 = rte_sched_port_enqueue_qptrs_prefetch0(subport00, + pkt00, subport_qmask); + q01 = rte_sched_port_enqueue_qptrs_prefetch0(subport01, + pkt01, subport_qmask); + + q10_base = rte_sched_subport_pipe_qbase(subport10, q10); + q11_base = rte_sched_subport_pipe_qbase(subport11, q11); + rte_sched_port_enqueue_qwa_prefetch0(port, subport10, q10, q10_base); + rte_sched_port_enqueue_qwa_prefetch0(port, subport11, q11, q11_base); + + r20 = rte_sched_port_enqueue_qwa(port, subport20, + q20, q20_base, pkt20); + r21 = rte_sched_port_enqueue_qwa(port, subport21, + q21, q21_base, pkt21); result += r20 + r21; - - q_last = rte_sched_port_enqueue_qptrs_prefetch0(port, pkt_last); - - q00_base = rte_sched_port_qbase(port, q00); - q01_base = rte_sched_port_qbase(port, q01); - rte_sched_port_enqueue_qwa_prefetch0(port, q00, q00_base); - rte_sched_port_enqueue_qwa_prefetch0(port, q01, q01_base); - - r10 = rte_sched_port_enqueue_qwa(port, q10, q10_base, pkt10); - r11 = rte_sched_port_enqueue_qwa(port, q11, q11_base, pkt11); + + subport_last = rte_sched_port_subport(port, pkt_last); + q_last = rte_sched_port_enqueue_qptrs_prefetch0(subport_last, + pkt_last, subport_qmask); + + q00_base = rte_sched_subport_pipe_qbase(subport00, q00); + q01_base = rte_sched_subport_pipe_qbase(subport01, q01); + rte_sched_port_enqueue_qwa_prefetch0(port, subport00, q00, q00_base); + rte_sched_port_enqueue_qwa_prefetch0(port, subport01, q01, q01_base); + + r10 = rte_sched_port_enqueue_qwa(port, subport10, q10, + q10_base, pkt10); + r11 = rte_sched_port_enqueue_qwa(port, subport11, q11, + q11_base, pkt11); result += r10 + r11; - q_last_base = rte_sched_port_qbase(port, q_last); - rte_sched_port_enqueue_qwa_prefetch0(port, q_last, q_last_base); + q_last_base = rte_sched_subport_pipe_qbase(subport_last, q_last); + rte_sched_port_enqueue_qwa_prefetch0(port, subport_last, + q_last, q_last_base); - r00 = rte_sched_port_enqueue_qwa(port, q00, q00_base, pkt00); - r01 = rte_sched_port_enqueue_qwa(port, q01, q01_base, pkt01); + r00 = rte_sched_port_enqueue_qwa(port, subport00, q00, + q00_base, pkt00); + r01 = rte_sched_port_enqueue_qwa(port, subport01, q01, + q01_base, pkt01); result += r00 + r01; if (n_pkts & 1) { - r_last = rte_sched_port_enqueue_qwa(port, q_last, q_last_base, pkt_last); + r_last = rte_sched_port_enqueue_qwa(port, subport_last, + q_last, q_last_base, pkt_last); result += r_last; } - + return result; } -#endif /* RTE_SCHED_ENQUEUE */ - -#if RTE_SCHED_TS_CREDITS_UPDATE == 0 - -#define grinder_credits_update(port, pos) - -#elif !defined(RTE_SCHED_SUBPORT_TC_OV) +#ifndef RTE_SCHED_SUBPORT_TC_OV static inline void -grinder_credits_update(struct rte_sched_port *port, uint32_t pos) +grinder_credits_update(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - struct rte_sched_subport *subport = grinder->subport; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; struct rte_sched_pipe_profile *params = grinder->pipe_params; uint64_t n_periods; - + uint32_t i; + /* Subport TB */ n_periods = (port->time - subport->tb_time) / subport->tb_period; subport->tb_credits += n_periods * subport->tb_credits_per_period; - subport->tb_credits = rte_sched_min_val_2_u32(subport->tb_credits, subport->tb_size); + subport->tb_credits = RTE_MIN(subport->tb_credits, subport->tb_size); subport->tb_time += n_periods * subport->tb_period; - + /* Pipe TB */ n_periods = (port->time - pipe->tb_time) / params->tb_period; pipe->tb_credits += n_periods * params->tb_credits_per_period; - pipe->tb_credits = rte_sched_min_val_2_u32(pipe->tb_credits, params->tb_size); + pipe->tb_credits = RTE_MIN(pipe->tb_credits, params->tb_size); pipe->tb_time += n_periods * params->tb_period; /* Subport TCs */ if (unlikely(port->time >= subport->tc_time)) { - subport->tc_credits[0] = subport->tc_credits_per_period[0]; - subport->tc_credits[1] = subport->tc_credits_per_period[1]; - subport->tc_credits[2] = subport->tc_credits_per_period[2]; - subport->tc_credits[3] = subport->tc_credits_per_period[3]; + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + subport->tc_credits[i] = subport->tc_credits_per_period[i]; + subport->tc_time = port->time + subport->tc_period; } - + /* Pipe TCs */ if (unlikely(port->time >= pipe->tc_time)) { - pipe->tc_credits[0] = params->tc_credits_per_period[0]; - pipe->tc_credits[1] = params->tc_credits_per_period[1]; - pipe->tc_credits[2] = params->tc_credits_per_period[2]; - pipe->tc_credits[3] = params->tc_credits_per_period[3]; + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + pipe->tc_credits[i] = params->tc_credits_per_period[i]; + pipe->tc_time = port->time + params->tc_period; } } #else +static inline uint64_t +grinder_tc_ov_credits_update(struct rte_sched_port *port, + struct rte_sched_subport *subport) +{ + uint64_t tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint64_t tc_consumption = 0, tc_ov_consumption_max; + uint64_t tc_ov_wm = subport->tc_ov_wm; + uint32_t i; + + if (subport->tc_ov == 0) + return subport->tc_ov_wm_max; + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASS_BE; i++) { + tc_ov_consumption[i] = + subport->tc_credits_per_period[i] - subport->tc_credits[i]; + tc_consumption += tc_ov_consumption[i]; + } + + tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASS_BE] = + subport->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] - + subport->tc_credits[RTE_SCHED_TRAFFIC_CLASS_BE]; + + tc_ov_consumption_max = + subport->tc_credits_per_period[RTE_SCHED_TRAFFIC_CLASS_BE] - + tc_consumption; + + if (tc_ov_consumption[RTE_SCHED_TRAFFIC_CLASS_BE] > + (tc_ov_consumption_max - port->mtu)) { + tc_ov_wm -= tc_ov_wm >> 7; + if (tc_ov_wm < subport->tc_ov_wm_min) + tc_ov_wm = subport->tc_ov_wm_min; + + return tc_ov_wm; + } + + tc_ov_wm += (tc_ov_wm >> 7) + 1; + if (tc_ov_wm > subport->tc_ov_wm_max) + tc_ov_wm = subport->tc_ov_wm_max; + + return tc_ov_wm; +} + static inline void -grinder_credits_update(struct rte_sched_port *port, uint32_t pos) +grinder_credits_update(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - struct rte_sched_subport *subport = grinder->subport; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; struct rte_sched_pipe_profile *params = grinder->pipe_params; uint64_t n_periods; - + uint32_t i; + /* Subport TB */ n_periods = (port->time - subport->tb_time) / subport->tb_period; subport->tb_credits += n_periods * subport->tb_credits_per_period; - subport->tb_credits = rte_sched_min_val_2_u32(subport->tb_credits, subport->tb_size); + subport->tb_credits = RTE_MIN(subport->tb_credits, subport->tb_size); subport->tb_time += n_periods * subport->tb_period; - + /* Pipe TB */ n_periods = (port->time - pipe->tb_time) / params->tb_period; pipe->tb_credits += n_periods * params->tb_credits_per_period; - pipe->tb_credits = rte_sched_min_val_2_u32(pipe->tb_credits, params->tb_size); + pipe->tb_credits = RTE_MIN(pipe->tb_credits, params->tb_size); pipe->tb_time += n_periods * params->tb_period; /* Subport TCs */ - if (unlikely(port->time >= subport->tc_ov_time)) { - uint64_t n_ov_periods; - - if (unlikely(port->time >= subport->tc_time)) { - subport->tc_credits[0] = subport->tc_credits_per_period[0]; - subport->tc_credits[1] = subport->tc_credits_per_period[1]; - subport->tc_credits[2] = subport->tc_credits_per_period[2]; - subport->tc_credits[3] = subport->tc_credits_per_period[3]; - - subport->tc_time = port->time + subport->tc_period; - } - - n_ov_periods = (subport->tc_time - port->time + subport->tc_ov_period - 1) / subport->tc_ov_period; - - subport->tc_ov_credits[0] = subport->tc_credits[0] / (n_ov_periods * subport->tc_ov_n[0]); - subport->tc_ov_credits[1] = subport->tc_credits[1] / (n_ov_periods * subport->tc_ov_n[1]); - subport->tc_ov_credits[2] = subport->tc_credits[2] / (n_ov_periods * subport->tc_ov_n[2]); - subport->tc_ov_credits[3] = subport->tc_credits[3] / (n_ov_periods * subport->tc_ov_n[3]); - - subport->tc_ov_time = port->time + subport->tc_ov_period; - subport->tc_ov_period_id ++; - } - + if (unlikely(port->time >= subport->tc_time)) { + subport->tc_ov_wm = grinder_tc_ov_credits_update(port, subport); + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + subport->tc_credits[i] = subport->tc_credits_per_period[i]; + + subport->tc_time = port->time + subport->tc_period; + subport->tc_ov_period_id++; + } + /* Pipe TCs */ if (unlikely(port->time >= pipe->tc_time)) { - pipe->tc_credits[0] = params->tc_credits_per_period[0]; - pipe->tc_credits[1] = params->tc_credits_per_period[1]; - pipe->tc_credits[2] = params->tc_credits_per_period[2]; - pipe->tc_credits[3] = params->tc_credits_per_period[3]; + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + pipe->tc_credits[i] = params->tc_credits_per_period[i]; pipe->tc_time = port->time + params->tc_period; } + + /* Pipe TCs - Oversubscription */ if (unlikely(pipe->tc_ov_period_id != subport->tc_ov_period_id)) { - uint32_t pipe_tc_ov_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t tc_mask[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; - uint32_t mask[] = {UINT32_MAX, 0}; - - tc_mask[0] = mask[subport->tc_ov[0]]; - tc_mask[1] = mask[subport->tc_ov[1]]; - tc_mask[2] = mask[subport->tc_ov[2]]; - tc_mask[3] = mask[subport->tc_ov[3]]; - - pipe_tc_ov_credits[0] = subport->tc_ov_credits[0] * params->tc_ov_weight[0]; - pipe_tc_ov_credits[1] = subport->tc_ov_credits[1] * params->tc_ov_weight[1]; - pipe_tc_ov_credits[2] = subport->tc_ov_credits[2] * params->tc_ov_weight[2]; - pipe_tc_ov_credits[3] = subport->tc_ov_credits[3] * params->tc_ov_weight[3]; - - pipe->tc_ov_credits[0] = (tc_mask[0] & pipe->tc_credits[0]) | ((~ tc_mask[0]) & pipe_tc_ov_credits[0]); - pipe->tc_ov_credits[1] = (tc_mask[1] & pipe->tc_credits[1]) | ((~ tc_mask[1]) & pipe_tc_ov_credits[1]); - pipe->tc_ov_credits[2] = (tc_mask[2] & pipe->tc_credits[2]) | ((~ tc_mask[2]) & pipe_tc_ov_credits[2]); - pipe->tc_ov_credits[3] = (tc_mask[3] & pipe->tc_credits[3]) | ((~ tc_mask[3]) & pipe_tc_ov_credits[3]); - + pipe->tc_ov_credits = subport->tc_ov_wm * params->tc_ov_weight; + pipe->tc_ov_period_id = subport->tc_ov_period_id; } } #endif /* RTE_SCHED_TS_CREDITS_UPDATE, RTE_SCHED_SUBPORT_TC_OV */ + #ifndef RTE_SCHED_SUBPORT_TC_OV static inline int -grinder_credits_check(struct rte_sched_port *port, uint32_t pos) +grinder_credits_check(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - struct rte_sched_subport *subport = grinder->subport; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; struct rte_mbuf *pkt = grinder->pkt; uint32_t tc_index = grinder->tc_index; - uint32_t pkt_len = pkt->pkt.pkt_len + port->frame_overhead; + uint64_t pkt_len = pkt->pkt_len + port->frame_overhead; + uint64_t subport_tb_credits = subport->tb_credits; + uint64_t subport_tc_credits = subport->tc_credits[tc_index]; + uint64_t pipe_tb_credits = pipe->tb_credits; + uint64_t pipe_tc_credits = pipe->tc_credits[tc_index]; int enough_credits; /* Check queue credits */ - enough_credits = (pkt_len <= subport->tb_credits) && - (pkt_len <= subport->tc_credits[tc_index]) && - (pkt_len <= pipe->tb_credits) && - (pkt_len <= pipe->tc_credits[tc_index]); - - if (!enough_credits) { + enough_credits = (pkt_len <= subport_tb_credits) && + (pkt_len <= subport_tc_credits) && + (pkt_len <= pipe_tb_credits) && + (pkt_len <= pipe_tc_credits); + + if (!enough_credits) return 0; - } - + /* Update port credits */ subport->tb_credits -= pkt_len; subport->tc_credits[tc_index] -= pkt_len; @@ -1586,90 +2137,105 @@ grinder_credits_check(struct rte_sched_port *port, uint32_t pos) #else static inline int -grinder_credits_check(struct rte_sched_port *port, uint32_t pos) +grinder_credits_check(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - struct rte_sched_subport *subport = grinder->subport; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; struct rte_mbuf *pkt = grinder->pkt; uint32_t tc_index = grinder->tc_index; - uint32_t pkt_len = pkt->pkt.pkt_len + port->frame_overhead; - uint32_t subport_tb_credits = subport->tb_credits; - uint32_t subport_tc_credits = subport->tc_credits[tc_index]; - uint32_t pipe_tb_credits = pipe->tb_credits; - uint32_t pipe_tc_credits = pipe->tc_credits[tc_index]; - uint32_t pipe_tc_ov_credits = pipe->tc_ov_credits[tc_index]; + uint64_t pkt_len = pkt->pkt_len + port->frame_overhead; + uint64_t subport_tb_credits = subport->tb_credits; + uint64_t subport_tc_credits = subport->tc_credits[tc_index]; + uint64_t pipe_tb_credits = pipe->tb_credits; + uint64_t pipe_tc_credits = pipe->tc_credits[tc_index]; + uint64_t pipe_tc_ov_mask1[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE]; + uint64_t pipe_tc_ov_mask2[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE] = {0}; + uint64_t pipe_tc_ov_credits; + uint32_t i; int enough_credits; - + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE; i++) + pipe_tc_ov_mask1[i] = ~0LLU; + + pipe_tc_ov_mask1[RTE_SCHED_TRAFFIC_CLASS_BE] = pipe->tc_ov_credits; + pipe_tc_ov_mask2[RTE_SCHED_TRAFFIC_CLASS_BE] = ~0LLU; + pipe_tc_ov_credits = pipe_tc_ov_mask1[tc_index]; + /* Check pipe and subport credits */ enough_credits = (pkt_len <= subport_tb_credits) && (pkt_len <= subport_tc_credits) && (pkt_len <= pipe_tb_credits) && (pkt_len <= pipe_tc_credits) && (pkt_len <= pipe_tc_ov_credits); - - if (!enough_credits) { + + if (!enough_credits) return 0; - } - + /* Update pipe and subport credits */ subport->tb_credits -= pkt_len; subport->tc_credits[tc_index] -= pkt_len; pipe->tb_credits -= pkt_len; pipe->tc_credits[tc_index] -= pkt_len; - pipe->tc_ov_credits[tc_index] -= pkt_len; - + pipe->tc_ov_credits -= pipe_tc_ov_mask2[tc_index] & pkt_len; + return 1; } #endif /* RTE_SCHED_SUBPORT_TC_OV */ -static inline int -grinder_schedule(struct rte_sched_port *port, uint32_t pos) + +static inline int +grinder_schedule(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_queue *queue = grinder->queue[grinder->qpos]; struct rte_mbuf *pkt = grinder->pkt; - uint32_t pkt_len = pkt->pkt.pkt_len + port->frame_overhead; + uint32_t pkt_len = pkt->pkt_len + port->frame_overhead; + uint32_t be_tc_active; -#if RTE_SCHED_TS_CREDITS_CHECK - if (!grinder_credits_check(port, pos)) { + if (!grinder_credits_check(port, subport, pos)) return 0; - } -#endif /* Advance port time */ port->time += pkt_len; - + /* Send packet */ - port->pkts_out[port->n_pkts_out ++] = pkt; - queue->qr ++; - grinder->wrr_tokens[grinder->qpos] += pkt_len * grinder->wrr_cost[grinder->qpos]; + port->pkts_out[port->n_pkts_out++] = pkt; + queue->qr++; + + be_tc_active = (grinder->tc_index == RTE_SCHED_TRAFFIC_CLASS_BE) ? ~0x0 : 0x0; + grinder->wrr_tokens[grinder->qpos] += + (pkt_len * grinder->wrr_cost[grinder->qpos]) & be_tc_active; + if (queue->qr == queue->qw) { uint32_t qindex = grinder->qindex[grinder->qpos]; - rte_bitmap_clear(port->bmp, qindex); + rte_bitmap_clear(subport->bmp, qindex); grinder->qmask &= ~(1 << grinder->qpos); - grinder->wrr_mask[grinder->qpos] = 0; - rte_sched_port_set_queue_empty_timestamp(port, qindex); + if (be_tc_active) + grinder->wrr_mask[grinder->qpos] = 0; + rte_sched_port_set_queue_empty_timestamp(port, subport, qindex); } - + /* Reset pipe loop detection */ - port->pipe_loop = RTE_SCHED_PIPE_INVALID; + subport->pipe_loop = RTE_SCHED_PIPE_INVALID; grinder->productive = 1; - + return 1; } -#if RTE_SCHED_OPTIMIZATIONS +#ifdef SCHED_VECTOR_SSE4 static inline int -grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe) +grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe) { - __m128i index = _mm_set1_epi32 (base_pipe); - __m128i pipes = _mm_load_si128((__m128i *)port->grinder_base_bmp_pos); + __m128i index = _mm_set1_epi32(base_pipe); + __m128i pipes = _mm_load_si128((__m128i *)subport->grinder_base_bmp_pos); __m128i res = _mm_cmpeq_epi32(pipes, index); - pipes = _mm_load_si128((__m128i *)(port->grinder_base_bmp_pos + 4)); + + pipes = _mm_load_si128((__m128i *)(subport->grinder_base_bmp_pos + 4)); pipes = _mm_cmpeq_epi32(pipes, index); res = _mm_or_si128(res, pipes); @@ -1679,461 +2245,518 @@ grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe) return 1; } +#elif defined(SCHED_VECTOR_NEON) + +static inline int +grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe) +{ + uint32x4_t index, pipes; + uint32_t *pos = (uint32_t *)subport->grinder_base_bmp_pos; + + index = vmovq_n_u32(base_pipe); + pipes = vld1q_u32(pos); + if (!vminvq_u32(veorq_u32(pipes, index))) + return 1; + + pipes = vld1q_u32(pos + 4); + if (!vminvq_u32(veorq_u32(pipes, index))) + return 1; + + return 0; +} + #else static inline int -grinder_pipe_exists(struct rte_sched_port *port, uint32_t base_pipe) +grinder_pipe_exists(struct rte_sched_subport *subport, uint32_t base_pipe) { uint32_t i; - - for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i ++) { - if (port->grinder_base_bmp_pos[i] == base_pipe) { + + for (i = 0; i < RTE_SCHED_PORT_N_GRINDERS; i++) { + if (subport->grinder_base_bmp_pos[i] == base_pipe) return 1; - } } - + return 0; } #endif /* RTE_SCHED_OPTIMIZATIONS */ static inline void -grinder_pcache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t bmp_pos, uint64_t bmp_slab) +grinder_pcache_populate(struct rte_sched_subport *subport, + uint32_t pos, uint32_t bmp_pos, uint64_t bmp_slab) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; uint16_t w[4]; grinder->pcache_w = 0; grinder->pcache_r = 0; - + w[0] = (uint16_t) bmp_slab; w[1] = (uint16_t) (bmp_slab >> 16); w[2] = (uint16_t) (bmp_slab >> 32); w[3] = (uint16_t) (bmp_slab >> 48); - + grinder->pcache_qmask[grinder->pcache_w] = w[0]; grinder->pcache_qindex[grinder->pcache_w] = bmp_pos; grinder->pcache_w += (w[0] != 0); - + grinder->pcache_qmask[grinder->pcache_w] = w[1]; grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 16; grinder->pcache_w += (w[1] != 0); - + grinder->pcache_qmask[grinder->pcache_w] = w[2]; grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 32; grinder->pcache_w += (w[2] != 0); - + grinder->pcache_qmask[grinder->pcache_w] = w[3]; grinder->pcache_qindex[grinder->pcache_w] = bmp_pos + 48; grinder->pcache_w += (w[3] != 0); } static inline void -grinder_tccache_populate(struct rte_sched_port *port, uint32_t pos, uint32_t qindex, uint16_t qmask) +grinder_tccache_populate(struct rte_sched_subport *subport, + uint32_t pos, uint32_t qindex, uint16_t qmask) { - struct rte_sched_grinder *grinder = port->grinder + pos; - uint8_t b[4]; - + struct rte_sched_grinder *grinder = subport->grinder + pos; + uint8_t b, i; + grinder->tccache_w = 0; grinder->tccache_r = 0; - - b[0] = (uint8_t) (qmask & 0xF); - b[1] = (uint8_t) ((qmask >> 4) & 0xF); - b[2] = (uint8_t) ((qmask >> 8) & 0xF); - b[3] = (uint8_t) ((qmask >> 12) & 0xF); - - grinder->tccache_qmask[grinder->tccache_w] = b[0]; - grinder->tccache_qindex[grinder->tccache_w] = qindex; - grinder->tccache_w += (b[0] != 0); - - grinder->tccache_qmask[grinder->tccache_w] = b[1]; - grinder->tccache_qindex[grinder->tccache_w] = qindex + 4; - grinder->tccache_w += (b[1] != 0); - - grinder->tccache_qmask[grinder->tccache_w] = b[2]; - grinder->tccache_qindex[grinder->tccache_w] = qindex + 8; - grinder->tccache_w += (b[2] != 0); - - grinder->tccache_qmask[grinder->tccache_w] = b[3]; - grinder->tccache_qindex[grinder->tccache_w] = qindex + 12; - grinder->tccache_w += (b[3] != 0); + + for (i = 0; i < RTE_SCHED_TRAFFIC_CLASS_BE; i++) { + b = (uint8_t) ((qmask >> i) & 0x1); + grinder->tccache_qmask[grinder->tccache_w] = b; + grinder->tccache_qindex[grinder->tccache_w] = qindex + i; + grinder->tccache_w += (b != 0); + } + + b = (uint8_t) (qmask >> (RTE_SCHED_TRAFFIC_CLASS_BE)); + grinder->tccache_qmask[grinder->tccache_w] = b; + grinder->tccache_qindex[grinder->tccache_w] = qindex + + RTE_SCHED_TRAFFIC_CLASS_BE; + grinder->tccache_w += (b != 0); } static inline int -grinder_next_tc(struct rte_sched_port *port, uint32_t pos) +grinder_next_tc(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_mbuf **qbase; - uint32_t qindex; - uint16_t qsize; + uint32_t qindex; + uint16_t qsize; - if (grinder->tccache_r == grinder->tccache_w) { + if (grinder->tccache_r == grinder->tccache_w) return 0; - } qindex = grinder->tccache_qindex[grinder->tccache_r]; - qbase = rte_sched_port_qbase(port, qindex); - qsize = rte_sched_port_qsize(port, qindex); + qbase = rte_sched_subport_pipe_qbase(subport, qindex); + qsize = rte_sched_subport_pipe_qsize(port, subport, qindex); - grinder->tc_index = (qindex >> 2) & 0x3; + grinder->tc_index = rte_sched_port_pipe_tc(port, qindex); grinder->qmask = grinder->tccache_qmask[grinder->tccache_r]; grinder->qsize = qsize; - - grinder->qindex[0] = qindex; - grinder->qindex[1] = qindex + 1; - grinder->qindex[2] = qindex + 2; - grinder->qindex[3] = qindex + 3; - grinder->queue[0] = port->queue + qindex; - grinder->queue[1] = port->queue + qindex + 1; - grinder->queue[2] = port->queue + qindex + 2; - grinder->queue[3] = port->queue + qindex + 3; + if (grinder->tc_index < RTE_SCHED_TRAFFIC_CLASS_BE) { + grinder->queue[0] = subport->queue + qindex; + grinder->qbase[0] = qbase; + grinder->qindex[0] = qindex; + grinder->tccache_r++; + + return 1; + } + + grinder->queue[0] = subport->queue + qindex; + grinder->queue[1] = subport->queue + qindex + 1; + grinder->queue[2] = subport->queue + qindex + 2; + grinder->queue[3] = subport->queue + qindex + 3; grinder->qbase[0] = qbase; grinder->qbase[1] = qbase + qsize; grinder->qbase[2] = qbase + 2 * qsize; grinder->qbase[3] = qbase + 3 * qsize; - - grinder->tccache_r ++; + + grinder->qindex[0] = qindex; + grinder->qindex[1] = qindex + 1; + grinder->qindex[2] = qindex + 2; + grinder->qindex[3] = qindex + 3; + + grinder->tccache_r++; return 1; } static inline int -grinder_next_pipe(struct rte_sched_port *port, uint32_t pos) +grinder_next_pipe(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; uint32_t pipe_qindex; uint16_t pipe_qmask; if (grinder->pcache_r < grinder->pcache_w) { pipe_qmask = grinder->pcache_qmask[grinder->pcache_r]; pipe_qindex = grinder->pcache_qindex[grinder->pcache_r]; - grinder->pcache_r ++; + grinder->pcache_r++; } else { uint64_t bmp_slab = 0; uint32_t bmp_pos = 0; - - /* Get another non-empty pipe group */ - if (unlikely(rte_bitmap_scan(port->bmp, &bmp_pos, &bmp_slab) <= 0)) { + + /* Get another non-empty pipe group */ + if (unlikely(rte_bitmap_scan(subport->bmp, &bmp_pos, &bmp_slab) <= 0)) return 0; - } - -#if RTE_SCHED_DEBUG - debug_check_queue_slab(port, bmp_pos, bmp_slab); -#endif + +#ifdef RTE_SCHED_DEBUG + debug_check_queue_slab(subport, bmp_pos, bmp_slab); +#endif /* Return if pipe group already in one of the other grinders */ - port->grinder_base_bmp_pos[pos] = RTE_SCHED_BMP_POS_INVALID; - if (unlikely(grinder_pipe_exists(port, bmp_pos))) { + subport->grinder_base_bmp_pos[pos] = RTE_SCHED_BMP_POS_INVALID; + if (unlikely(grinder_pipe_exists(subport, bmp_pos))) return 0; - } - port->grinder_base_bmp_pos[pos] = bmp_pos; - + + subport->grinder_base_bmp_pos[pos] = bmp_pos; + /* Install new pipe group into grinder's pipe cache */ - grinder_pcache_populate(port, pos, bmp_pos, bmp_slab); + grinder_pcache_populate(subport, pos, bmp_pos, bmp_slab); pipe_qmask = grinder->pcache_qmask[0]; pipe_qindex = grinder->pcache_qindex[0]; grinder->pcache_r = 1; } - + /* Install new pipe in the grinder */ grinder->pindex = pipe_qindex >> 4; - grinder->subport = port->subport + (grinder->pindex / port->n_pipes_per_subport); - grinder->pipe = port->pipe + grinder->pindex; + grinder->subport = subport; + grinder->pipe = subport->pipe + grinder->pindex; grinder->pipe_params = NULL; /* to be set after the pipe structure is prefetched */ grinder->productive = 0; - grinder_tccache_populate(port, pos, pipe_qindex, pipe_qmask); - grinder_next_tc(port, pos); - + grinder_tccache_populate(subport, pos, pipe_qindex, pipe_qmask); + grinder_next_tc(port, subport, pos); + /* Check for pipe exhaustion */ - if (grinder->pindex == port->pipe_loop) { - port->pipe_exhaustion = 1; - port->pipe_loop = RTE_SCHED_PIPE_INVALID; + if (grinder->pindex == subport->pipe_loop) { + subport->pipe_exhaustion = 1; + subport->pipe_loop = RTE_SCHED_PIPE_INVALID; } - - return 1; -} - -#if RTE_SCHED_WRR == 0 - -#define grinder_wrr_load(a,b) -#define grinder_wrr_store(a,b) - -static inline void -grinder_wrr(struct rte_sched_port *port, uint32_t pos) -{ - struct rte_sched_grinder *grinder = port->grinder + pos; - uint64_t slab = grinder->qmask; - - if (rte_bsf64(slab, &grinder->qpos) == 0) { - rte_panic("grinder wrr\n"); - } + return 1; } -#elif RTE_SCHED_WRR == 1 static inline void -grinder_wrr_load(struct rte_sched_port *port, uint32_t pos) +grinder_wrr_load(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; struct rte_sched_pipe_profile *pipe_params = grinder->pipe_params; - uint32_t tc_index = grinder->tc_index; uint32_t qmask = grinder->qmask; - uint32_t qindex; - - qindex = tc_index * 4; - - grinder->wrr_tokens[0] = ((uint16_t) pipe->wrr_tokens[qindex]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[1] = ((uint16_t) pipe->wrr_tokens[qindex + 1]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[2] = ((uint16_t) pipe->wrr_tokens[qindex + 2]) << RTE_SCHED_WRR_SHIFT; - grinder->wrr_tokens[3] = ((uint16_t) pipe->wrr_tokens[qindex + 3]) << RTE_SCHED_WRR_SHIFT; - + + grinder->wrr_tokens[0] = + ((uint16_t) pipe->wrr_tokens[0]) << RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[1] = + ((uint16_t) pipe->wrr_tokens[1]) << RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[2] = + ((uint16_t) pipe->wrr_tokens[2]) << RTE_SCHED_WRR_SHIFT; + grinder->wrr_tokens[3] = + ((uint16_t) pipe->wrr_tokens[3]) << RTE_SCHED_WRR_SHIFT; + grinder->wrr_mask[0] = (qmask & 0x1) * 0xFFFF; grinder->wrr_mask[1] = ((qmask >> 1) & 0x1) * 0xFFFF; grinder->wrr_mask[2] = ((qmask >> 2) & 0x1) * 0xFFFF; grinder->wrr_mask[3] = ((qmask >> 3) & 0x1) * 0xFFFF; - - grinder->wrr_cost[0] = pipe_params->wrr_cost[qindex]; - grinder->wrr_cost[1] = pipe_params->wrr_cost[qindex + 1]; - grinder->wrr_cost[2] = pipe_params->wrr_cost[qindex + 2]; - grinder->wrr_cost[3] = pipe_params->wrr_cost[qindex + 3]; + + grinder->wrr_cost[0] = pipe_params->wrr_cost[0]; + grinder->wrr_cost[1] = pipe_params->wrr_cost[1]; + grinder->wrr_cost[2] = pipe_params->wrr_cost[2]; + grinder->wrr_cost[3] = pipe_params->wrr_cost[3]; } static inline void -grinder_wrr_store(struct rte_sched_port *port, uint32_t pos) +grinder_wrr_store(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; struct rte_sched_pipe *pipe = grinder->pipe; - uint32_t tc_index = grinder->tc_index; - uint32_t qindex; - - qindex = tc_index * 4; - - pipe->wrr_tokens[qindex] = (uint8_t) ((grinder->wrr_tokens[0] & grinder->wrr_mask[0]) >> RTE_SCHED_WRR_SHIFT); - pipe->wrr_tokens[qindex + 1] = (uint8_t) ((grinder->wrr_tokens[1] & grinder->wrr_mask[1]) >> RTE_SCHED_WRR_SHIFT); - pipe->wrr_tokens[qindex + 2] = (uint8_t) ((grinder->wrr_tokens[2] & grinder->wrr_mask[2]) >> RTE_SCHED_WRR_SHIFT); - pipe->wrr_tokens[qindex + 3] = (uint8_t) ((grinder->wrr_tokens[3] & grinder->wrr_mask[3]) >> RTE_SCHED_WRR_SHIFT); + + pipe->wrr_tokens[0] = + (grinder->wrr_tokens[0] & grinder->wrr_mask[0]) >> + RTE_SCHED_WRR_SHIFT; + pipe->wrr_tokens[1] = + (grinder->wrr_tokens[1] & grinder->wrr_mask[1]) >> + RTE_SCHED_WRR_SHIFT; + pipe->wrr_tokens[2] = + (grinder->wrr_tokens[2] & grinder->wrr_mask[2]) >> + RTE_SCHED_WRR_SHIFT; + pipe->wrr_tokens[3] = + (grinder->wrr_tokens[3] & grinder->wrr_mask[3]) >> + RTE_SCHED_WRR_SHIFT; } static inline void -grinder_wrr(struct rte_sched_port *port, uint32_t pos) +grinder_wrr(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; uint16_t wrr_tokens_min; grinder->wrr_tokens[0] |= ~grinder->wrr_mask[0]; grinder->wrr_tokens[1] |= ~grinder->wrr_mask[1]; grinder->wrr_tokens[2] |= ~grinder->wrr_mask[2]; grinder->wrr_tokens[3] |= ~grinder->wrr_mask[3]; - + grinder->qpos = rte_min_pos_4_u16(grinder->wrr_tokens); wrr_tokens_min = grinder->wrr_tokens[grinder->qpos]; - + grinder->wrr_tokens[0] -= wrr_tokens_min; grinder->wrr_tokens[1] -= wrr_tokens_min; grinder->wrr_tokens[2] -= wrr_tokens_min; grinder->wrr_tokens[3] -= wrr_tokens_min; } -#else - -#error Invalid value for RTE_SCHED_WRR - -#endif /* RTE_SCHED_WRR */ -#define grinder_evict(port, pos) +#define grinder_evict(subport, pos) static inline void -grinder_prefetch_pipe(struct rte_sched_port *port, uint32_t pos) +grinder_prefetch_pipe(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - + struct rte_sched_grinder *grinder = subport->grinder + pos; + rte_prefetch0(grinder->pipe); rte_prefetch0(grinder->queue[0]); } static inline void -grinder_prefetch_tc_queue_arrays(struct rte_sched_port *port, uint32_t pos) +grinder_prefetch_tc_queue_arrays(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - uint16_t qsize, qr[4]; - + struct rte_sched_grinder *grinder = subport->grinder + pos; + uint16_t qsize, qr[RTE_SCHED_MAX_QUEUES_PER_TC]; + qsize = grinder->qsize; + grinder->qpos = 0; + + if (grinder->tc_index < RTE_SCHED_TRAFFIC_CLASS_BE) { + qr[0] = grinder->queue[0]->qr & (qsize - 1); + + rte_prefetch0(grinder->qbase[0] + qr[0]); + return; + } + qr[0] = grinder->queue[0]->qr & (qsize - 1); qr[1] = grinder->queue[1]->qr & (qsize - 1); qr[2] = grinder->queue[2]->qr & (qsize - 1); qr[3] = grinder->queue[3]->qr & (qsize - 1); - + rte_prefetch0(grinder->qbase[0] + qr[0]); rte_prefetch0(grinder->qbase[1] + qr[1]); - grinder_wrr_load(port, pos); - grinder_wrr(port, pos); - + grinder_wrr_load(subport, pos); + grinder_wrr(subport, pos); + rte_prefetch0(grinder->qbase[2] + qr[2]); - rte_prefetch0(grinder->qbase[3] + qr[3]); + rte_prefetch0(grinder->qbase[3] + qr[3]); } static inline void -grinder_prefetch_mbuf(struct rte_sched_port *port, uint32_t pos) +grinder_prefetch_mbuf(struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; + struct rte_sched_grinder *grinder = subport->grinder + pos; uint32_t qpos = grinder->qpos; struct rte_mbuf **qbase = grinder->qbase[qpos]; uint16_t qsize = grinder->qsize; uint16_t qr = grinder->queue[qpos]->qr & (qsize - 1); - + grinder->pkt = qbase[qr]; rte_prefetch0(grinder->pkt); - + if (unlikely((qr & 0x7) == 7)) { uint16_t qr_next = (grinder->queue[qpos]->qr + 1) & (qsize - 1); - + rte_prefetch0(qbase + qr_next); } } static inline uint32_t -grinder_handle(struct rte_sched_port *port, uint32_t pos) +grinder_handle(struct rte_sched_port *port, + struct rte_sched_subport *subport, uint32_t pos) { - struct rte_sched_grinder *grinder = port->grinder + pos; - + struct rte_sched_grinder *grinder = subport->grinder + pos; + switch (grinder->state) { case e_GRINDER_PREFETCH_PIPE: { - if (grinder_next_pipe(port, pos)) { - grinder_prefetch_pipe(port, pos); - port->busy_grinders ++; - + if (grinder_next_pipe(port, subport, pos)) { + grinder_prefetch_pipe(subport, pos); + subport->busy_grinders++; + grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS; return 0; } - + return 0; } case e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS: { struct rte_sched_pipe *pipe = grinder->pipe; - - grinder->pipe_params = port->pipe_profiles + pipe->profile; - grinder_prefetch_tc_queue_arrays(port, pos); - grinder_credits_update(port, pos); - + + grinder->pipe_params = subport->pipe_profiles + pipe->profile; + grinder_prefetch_tc_queue_arrays(subport, pos); + grinder_credits_update(port, subport, pos); + grinder->state = e_GRINDER_PREFETCH_MBUF; return 0; } - + case e_GRINDER_PREFETCH_MBUF: { - grinder_prefetch_mbuf(port, pos); - + grinder_prefetch_mbuf(subport, pos); + grinder->state = e_GRINDER_READ_MBUF; return 0; } - + case e_GRINDER_READ_MBUF: { - uint32_t result = 0; - - result = grinder_schedule(port, pos); - + uint32_t wrr_active, result = 0; + + result = grinder_schedule(port, subport, pos); + + wrr_active = (grinder->tc_index == RTE_SCHED_TRAFFIC_CLASS_BE); + /* Look for next packet within the same TC */ if (result && grinder->qmask) { - grinder_wrr(port, pos); - grinder_prefetch_mbuf(port, pos); - + if (wrr_active) + grinder_wrr(subport, pos); + + grinder_prefetch_mbuf(subport, pos); + return 1; } - grinder_wrr_store(port, pos); - + + if (wrr_active) + grinder_wrr_store(subport, pos); + /* Look for another active TC within same pipe */ - if (grinder_next_tc(port, pos)) { - grinder_prefetch_tc_queue_arrays(port, pos); - + if (grinder_next_tc(port, subport, pos)) { + grinder_prefetch_tc_queue_arrays(subport, pos); + grinder->state = e_GRINDER_PREFETCH_MBUF; return result; - } - if ((grinder->productive == 0) && (port->pipe_loop == RTE_SCHED_PIPE_INVALID)) { - port->pipe_loop = grinder->pindex; } - grinder_evict(port, pos); - + + if (grinder->productive == 0 && + subport->pipe_loop == RTE_SCHED_PIPE_INVALID) + subport->pipe_loop = grinder->pindex; + + grinder_evict(subport, pos); + /* Look for another active pipe */ - if (grinder_next_pipe(port, pos)) { - grinder_prefetch_pipe(port, pos); - + if (grinder_next_pipe(port, subport, pos)) { + grinder_prefetch_pipe(subport, pos); + grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS; return result; } - + /* No active pipe found */ - port->busy_grinders --; - + subport->busy_grinders--; + grinder->state = e_GRINDER_PREFETCH_PIPE; return result; } - + default: rte_panic("Algorithmic error (invalid state)\n"); return 0; } } -static inline void +static inline void rte_sched_port_time_resync(struct rte_sched_port *port) { uint64_t cycles = rte_get_tsc_cycles(); - uint64_t cycles_diff = cycles - port->time_cpu_cycles; - double bytes_diff = ((double) cycles_diff) / port->cycles_per_byte; - + uint64_t cycles_diff; + uint64_t bytes_diff; + uint32_t i; + + if (cycles < port->time_cpu_cycles) + port->time_cpu_cycles = 0; + + cycles_diff = cycles - port->time_cpu_cycles; + /* Compute elapsed time in bytes */ + bytes_diff = rte_reciprocal_divide(cycles_diff << RTE_SCHED_TIME_SHIFT, + port->inv_cycles_per_byte); + /* Advance port time */ - port->time_cpu_cycles = cycles; - port->time_cpu_bytes += (uint64_t) bytes_diff; - if (port->time < port->time_cpu_bytes) { + port->time_cpu_cycles += + (bytes_diff * port->cycles_per_byte) >> RTE_SCHED_TIME_SHIFT; + port->time_cpu_bytes += bytes_diff; + if (port->time < port->time_cpu_bytes) port->time = port->time_cpu_bytes; - } /* Reset pipe loop detection */ - port->pipe_loop = RTE_SCHED_PIPE_INVALID; + for (i = 0; i < port->n_subports_per_port; i++) + port->subports[i]->pipe_loop = RTE_SCHED_PIPE_INVALID; } static inline int -rte_sched_port_exceptions(struct rte_sched_port *port) +rte_sched_port_exceptions(struct rte_sched_subport *subport, int second_pass) { int exceptions; /* Check if any exception flag is set */ - exceptions = (port->busy_grinders == 0) || - (port->pipe_exhaustion == 1); - + exceptions = (second_pass && subport->busy_grinders == 0) || + (subport->pipe_exhaustion == 1); + /* Clear exception flags */ - port->pipe_exhaustion = 0; - + subport->pipe_exhaustion = 0; + return exceptions; } int rte_sched_port_dequeue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts) { - uint32_t i, count; - + struct rte_sched_subport *subport; + uint32_t subport_id = port->subport_id; + uint32_t i, n_subports = 0, count; + port->pkts_out = pkts; port->n_pkts_out = 0; - + rte_sched_port_time_resync(port); - + /* Take each queue in the grinder one step further */ - for (i = 0, count = 0; ; i ++) { - count += grinder_handle(port, i & (RTE_SCHED_PORT_N_GRINDERS - 1)); - if ((count == n_pkts) || rte_sched_port_exceptions(port)) { + for (i = 0, count = 0; ; i++) { + subport = port->subports[subport_id]; + + count += grinder_handle(port, subport, + i & (RTE_SCHED_PORT_N_GRINDERS - 1)); + + if (count == n_pkts) { + subport_id++; + + if (subport_id == port->n_subports_per_port) + subport_id = 0; + + port->subport_id = subport_id; + break; + } + + if (rte_sched_port_exceptions(subport, i >= RTE_SCHED_PORT_N_GRINDERS)) { + i = 0; + subport_id++; + n_subports++; + } + + if (subport_id == port->n_subports_per_port) + subport_id = 0; + + if (n_subports == port->n_subports_per_port) { + port->subport_id = subport_id; break; } } - + return count; }