summaryrefslogtreecommitdiffstats
path: root/src/spdk/dpdk/lib/librte_sched
diff options
context:
space:
mode:
Diffstat (limited to 'src/spdk/dpdk/lib/librte_sched')
-rw-r--r--src/spdk/dpdk/lib/librte_sched/Makefile29
-rw-r--r--src/spdk/dpdk/lib/librte_sched/meson.build9
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_approx.c320
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_approx.h64
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_red.c129
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_red.h411
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_sched.c2755
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_sched.h514
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_sched_common.h95
-rw-r--r--src/spdk/dpdk/lib/librte_sched/rte_sched_version.map31
10 files changed, 4357 insertions, 0 deletions
diff --git a/src/spdk/dpdk/lib/librte_sched/Makefile b/src/spdk/dpdk/lib/librte_sched/Makefile
new file mode 100644
index 000000000..aee93a120
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/Makefile
@@ -0,0 +1,29 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2010-2014 Intel Corporation
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+#
+# library name
+#
+LIB = librte_sched.a
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+LDLIBS += -lm
+LDLIBS += -lrt
+LDLIBS += -lrte_eal -lrte_mempool -lrte_mbuf -lrte_net
+LDLIBS += -lrte_timer
+
+EXPORT_MAP := rte_sched_version.map
+
+#
+# all source are stored in SRCS-y
+#
+SRCS-$(CONFIG_RTE_LIBRTE_SCHED) += rte_sched.c rte_red.c rte_approx.c
+
+# install includes
+SYMLINK-$(CONFIG_RTE_LIBRTE_SCHED)-include := rte_sched.h rte_sched_common.h rte_red.h rte_approx.h
+
+include $(RTE_SDK)/mk/rte.lib.mk
diff --git a/src/spdk/dpdk/lib/librte_sched/meson.build b/src/spdk/dpdk/lib/librte_sched/meson.build
new file mode 100644
index 000000000..d15fd602b
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/meson.build
@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2017 Intel Corporation
+
+sources = files('rte_sched.c', 'rte_red.c', 'rte_approx.c')
+headers = files('rte_sched.h', 'rte_sched_common.h',
+ 'rte_red.h', 'rte_approx.h')
+deps += ['mbuf', 'meter']
+build = false
+reason = 'not needed by SPDK'
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_approx.c b/src/spdk/dpdk/lib/librte_sched/rte_approx.c
new file mode 100644
index 000000000..edc8d9113
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_approx.c
@@ -0,0 +1,320 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#include <stdlib.h>
+
+#include "rte_approx.h"
+
+/*
+ * Based on paper "Approximating Rational Numbers by Fractions" by Michal
+ * Forisek forisek@dcs.fmph.uniba.sk
+ *
+ * Given a rational number alpha with 0 < alpha < 1 and a precision d, the goal
+ * is to find positive integers p, q such that alpha - d < p/q < alpha + d, and
+ * q is minimal.
+ *
+ * http://people.ksp.sk/~misof/publications/2007approx.pdf
+ */
+
+/* fraction comparison: compare (a/b) and (c/d) */
+static inline uint32_t
+less(uint32_t a, uint32_t b, uint32_t c, uint32_t d)
+{
+ return a*d < b*c;
+}
+
+static inline uint32_t
+less_or_equal(uint32_t a, uint32_t b, uint32_t c, uint32_t d)
+{
+ return a*d <= b*c;
+}
+
+/* check whether a/b is a valid approximation */
+static inline uint32_t
+matches(uint32_t a, uint32_t b,
+ uint32_t alpha_num, uint32_t d_num, uint32_t denum)
+{
+ if (less_or_equal(a, b, alpha_num - d_num, denum))
+ return 0;
+
+ if (less(a ,b, alpha_num + d_num, denum))
+ return 1;
+
+ return 0;
+}
+
+static inline void
+find_exact_solution_left(uint32_t p_a, uint32_t q_a, uint32_t p_b, uint32_t q_b,
+ uint32_t alpha_num, uint32_t d_num, uint32_t denum, uint32_t *p, uint32_t *q)
+{
+ uint32_t k_num = denum * p_b - (alpha_num + d_num) * q_b;
+ uint32_t k_denum = (alpha_num + d_num) * q_a - denum * p_a;
+ uint32_t k = (k_num / k_denum) + 1;
+
+ *p = p_b + k * p_a;
+ *q = q_b + k * q_a;
+}
+
+static inline void
+find_exact_solution_right(uint32_t p_a, uint32_t q_a, uint32_t p_b, uint32_t q_b,
+ uint32_t alpha_num, uint32_t d_num, uint32_t denum, uint32_t *p, uint32_t *q)
+{
+ uint32_t k_num = - denum * p_b + (alpha_num - d_num) * q_b;
+ uint32_t k_denum = - (alpha_num - d_num) * q_a + denum * p_a;
+ uint32_t k = (k_num / k_denum) + 1;
+
+ *p = p_b + k * p_a;
+ *q = q_b + k * q_a;
+}
+
+static int
+find_best_rational_approximation(uint32_t alpha_num, uint32_t d_num, uint32_t denum, uint32_t *p, uint32_t *q)
+{
+ uint32_t p_a, q_a, p_b, q_b;
+
+ /* check assumptions on the inputs */
+ if (!((0 < d_num) && (d_num < alpha_num) && (alpha_num < denum) && (d_num + alpha_num < denum))) {
+ return -1;
+ }
+
+ /* set initial bounds for the search */
+ p_a = 0;
+ q_a = 1;
+ p_b = 1;
+ q_b = 1;
+
+ while (1) {
+ uint32_t new_p_a, new_q_a, new_p_b, new_q_b;
+ uint32_t x_num, x_denum, x;
+ int aa, bb;
+
+ /* compute the number of steps to the left */
+ x_num = denum * p_b - alpha_num * q_b;
+ x_denum = - denum * p_a + alpha_num * q_a;
+ x = (x_num + x_denum - 1) / x_denum; /* x = ceil(x_num / x_denum) */
+
+ /* check whether we have a valid approximation */
+ aa = matches(p_b + x * p_a, q_b + x * q_a, alpha_num, d_num, denum);
+ bb = matches(p_b + (x-1) * p_a, q_b + (x - 1) * q_a, alpha_num, d_num, denum);
+ if (aa || bb) {
+ find_exact_solution_left(p_a, q_a, p_b, q_b, alpha_num, d_num, denum, p, q);
+ return 0;
+ }
+
+ /* update the interval */
+ new_p_a = p_b + (x - 1) * p_a ;
+ new_q_a = q_b + (x - 1) * q_a;
+ new_p_b = p_b + x * p_a ;
+ new_q_b = q_b + x * q_a;
+
+ p_a = new_p_a ;
+ q_a = new_q_a;
+ p_b = new_p_b ;
+ q_b = new_q_b;
+
+ /* compute the number of steps to the right */
+ x_num = alpha_num * q_b - denum * p_b;
+ x_denum = - alpha_num * q_a + denum * p_a;
+ x = (x_num + x_denum - 1) / x_denum; /* x = ceil(x_num / x_denum) */
+
+ /* check whether we have a valid approximation */
+ aa = matches(p_b + x * p_a, q_b + x * q_a, alpha_num, d_num, denum);
+ bb = matches(p_b + (x - 1) * p_a, q_b + (x - 1) * q_a, alpha_num, d_num, denum);
+ if (aa || bb) {
+ find_exact_solution_right(p_a, q_a, p_b, q_b, alpha_num, d_num, denum, p, q);
+ return 0;
+ }
+
+ /* update the interval */
+ new_p_a = p_b + (x - 1) * p_a;
+ new_q_a = q_b + (x - 1) * q_a;
+ new_p_b = p_b + x * p_a;
+ new_q_b = q_b + x * q_a;
+
+ p_a = new_p_a;
+ q_a = new_q_a;
+ p_b = new_p_b;
+ q_b = new_q_b;
+ }
+}
+
+int rte_approx(double alpha, double d, uint32_t *p, uint32_t *q)
+{
+ uint32_t alpha_num, d_num, denum;
+
+ /* Check input arguments */
+ if (!((0.0 < d) && (d < alpha) && (alpha < 1.0))) {
+ return -1;
+ }
+
+ if ((p == NULL) || (q == NULL)) {
+ return -2;
+ }
+
+ /* Compute alpha_num, d_num and denum */
+ denum = 1;
+ while (d < 1) {
+ alpha *= 10;
+ d *= 10;
+ denum *= 10;
+ }
+ alpha_num = (uint32_t) alpha;
+ d_num = (uint32_t) d;
+
+ /* Perform approximation */
+ return find_best_rational_approximation(alpha_num, d_num, denum, p, q);
+}
+
+/* fraction comparison (64 bit version): compare (a/b) and (c/d) */
+static inline uint64_t
+less_64(uint64_t a, uint64_t b, uint64_t c, uint64_t d)
+{
+ return a*d < b*c;
+}
+
+static inline uint64_t
+less_or_equal_64(uint64_t a, uint64_t b, uint64_t c, uint64_t d)
+{
+ return a*d <= b*c;
+}
+
+/* check whether a/b is a valid approximation (64 bit version) */
+static inline uint64_t
+matches_64(uint64_t a, uint64_t b,
+ uint64_t alpha_num, uint64_t d_num, uint64_t denum)
+{
+ if (less_or_equal_64(a, b, alpha_num - d_num, denum))
+ return 0;
+
+ if (less_64(a, b, alpha_num + d_num, denum))
+ return 1;
+
+ return 0;
+}
+
+static inline void
+find_exact_solution_left_64(uint64_t p_a, uint64_t q_a, uint64_t p_b, uint64_t q_b,
+ uint64_t alpha_num, uint64_t d_num, uint64_t denum, uint64_t *p, uint64_t *q)
+{
+ uint64_t k_num = denum * p_b - (alpha_num + d_num) * q_b;
+ uint64_t k_denum = (alpha_num + d_num) * q_a - denum * p_a;
+ uint64_t k = (k_num / k_denum) + 1;
+
+ *p = p_b + k * p_a;
+ *q = q_b + k * q_a;
+}
+
+static inline void
+find_exact_solution_right_64(uint64_t p_a, uint64_t q_a, uint64_t p_b, uint64_t q_b,
+ uint64_t alpha_num, uint64_t d_num, uint64_t denum, uint64_t *p, uint64_t *q)
+{
+ uint64_t k_num = -denum * p_b + (alpha_num - d_num) * q_b;
+ uint64_t k_denum = -(alpha_num - d_num) * q_a + denum * p_a;
+ uint64_t k = (k_num / k_denum) + 1;
+
+ *p = p_b + k * p_a;
+ *q = q_b + k * q_a;
+}
+
+static int
+find_best_rational_approximation_64(uint64_t alpha_num, uint64_t d_num,
+ uint64_t denum, uint64_t *p, uint64_t *q)
+{
+ uint64_t p_a, q_a, p_b, q_b;
+
+ /* check assumptions on the inputs */
+ if (!((d_num > 0) && (d_num < alpha_num) &&
+ (alpha_num < denum) && (d_num + alpha_num < denum))) {
+ return -1;
+ }
+
+ /* set initial bounds for the search */
+ p_a = 0;
+ q_a = 1;
+ p_b = 1;
+ q_b = 1;
+
+ while (1) {
+ uint64_t new_p_a, new_q_a, new_p_b, new_q_b;
+ uint64_t x_num, x_denum, x;
+ int aa, bb;
+
+ /* compute the number of steps to the left */
+ x_num = denum * p_b - alpha_num * q_b;
+ x_denum = -denum * p_a + alpha_num * q_a;
+ x = (x_num + x_denum - 1) / x_denum; /* x = ceil(x_num / x_denum) */
+
+ /* check whether we have a valid approximation */
+ aa = matches_64(p_b + x * p_a, q_b + x * q_a, alpha_num, d_num, denum);
+ bb = matches_64(p_b + (x-1) * p_a, q_b + (x - 1) * q_a,
+ alpha_num, d_num, denum);
+ if (aa || bb) {
+ find_exact_solution_left_64(p_a, q_a, p_b, q_b,
+ alpha_num, d_num, denum, p, q);
+ return 0;
+ }
+
+ /* update the interval */
+ new_p_a = p_b + (x - 1) * p_a;
+ new_q_a = q_b + (x - 1) * q_a;
+ new_p_b = p_b + x * p_a;
+ new_q_b = q_b + x * q_a;
+
+ p_a = new_p_a;
+ q_a = new_q_a;
+ p_b = new_p_b;
+ q_b = new_q_b;
+
+ /* compute the number of steps to the right */
+ x_num = alpha_num * q_b - denum * p_b;
+ x_denum = -alpha_num * q_a + denum * p_a;
+ x = (x_num + x_denum - 1) / x_denum; /* x = ceil(x_num / x_denum) */
+
+ /* check whether we have a valid approximation */
+ aa = matches_64(p_b + x * p_a, q_b + x * q_a, alpha_num, d_num, denum);
+ bb = matches_64(p_b + (x - 1) * p_a, q_b + (x - 1) * q_a,
+ alpha_num, d_num, denum);
+ if (aa || bb) {
+ find_exact_solution_right_64(p_a, q_a, p_b, q_b,
+ alpha_num, d_num, denum, p, q);
+ return 0;
+ }
+
+ /* update the interval */
+ new_p_a = p_b + (x - 1) * p_a;
+ new_q_a = q_b + (x - 1) * q_a;
+ new_p_b = p_b + x * p_a;
+ new_q_b = q_b + x * q_a;
+
+ p_a = new_p_a;
+ q_a = new_q_a;
+ p_b = new_p_b;
+ q_b = new_q_b;
+ }
+}
+
+int rte_approx_64(double alpha, double d, uint64_t *p, uint64_t *q)
+{
+ uint64_t alpha_num, d_num, denum;
+
+ /* Check input arguments */
+ if (!((0.0 < d) && (d < alpha) && (alpha < 1.0)))
+ return -1;
+
+ if ((p == NULL) || (q == NULL))
+ return -2;
+
+ /* Compute alpha_num, d_num and denum */
+ denum = 1;
+ while (d < 1) {
+ alpha *= 10;
+ d *= 10;
+ denum *= 10;
+ }
+ alpha_num = (uint64_t) alpha;
+ d_num = (uint64_t) d;
+
+ /* Perform approximation */
+ return find_best_rational_approximation_64(alpha_num, d_num, denum, p, q);
+}
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_approx.h b/src/spdk/dpdk/lib/librte_sched/rte_approx.h
new file mode 100644
index 000000000..74d55f457
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_approx.h
@@ -0,0 +1,64 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#ifndef __INCLUDE_RTE_APPROX_H__
+#define __INCLUDE_RTE_APPROX_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @file
+ * RTE Rational Approximation
+ *
+ * Given a rational number alpha with 0 < alpha < 1 and a precision d, the goal
+ * is to find positive integers p, q such that alpha - d < p/q < alpha + d, and
+ * q is minimal.
+ *
+ ***/
+
+#include <stdint.h>
+
+/**
+ * Find best rational approximation
+ *
+ * @param alpha
+ * Rational number to approximate
+ * @param d
+ * Precision for the rational approximation
+ * @param p
+ * Pointer to pre-allocated space where the numerator of the rational
+ * approximation will be stored when operation is successful
+ * @param q
+ * Pointer to pre-allocated space where the denominator of the rational
+ * approximation will be stored when operation is successful
+ * @return
+ * 0 upon success, error code otherwise
+ */
+int rte_approx(double alpha, double d, uint32_t *p, uint32_t *q);
+
+/**
+ * Find best rational approximation (64 bit version)
+ *
+ * @param alpha
+ * Rational number to approximate
+ * @param d
+ * Precision for the rational approximation
+ * @param p
+ * Pointer to pre-allocated space where the numerator of the rational
+ * approximation will be stored when operation is successful
+ * @param q
+ * Pointer to pre-allocated space where the denominator of the rational
+ * approximation will be stored when operation is successful
+ * @return
+ * 0 upon success, error code otherwise
+ */
+int rte_approx_64(double alpha, double d, uint64_t *p, uint64_t *q);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __INCLUDE_RTE_APPROX_H__ */
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_red.c b/src/spdk/dpdk/lib/librte_sched/rte_red.c
new file mode 100644
index 000000000..45a452f68
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_red.c
@@ -0,0 +1,129 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#include <math.h>
+#include "rte_red.h"
+#include <rte_random.h>
+#include <rte_common.h>
+
+#ifdef __INTEL_COMPILER
+#pragma warning(disable:2259) /* conversion may lose significant bits */
+#endif
+
+static int rte_red_init_done = 0; /**< Flag to indicate that global initialisation is done */
+uint32_t rte_red_rand_val = 0; /**< Random value cache */
+uint32_t rte_red_rand_seed = 0; /**< Seed for random number generation */
+
+/**
+ * table[i] = log2(1-Wq) * Scale * -1
+ * Wq = 1/(2^i)
+ */
+uint16_t rte_red_log2_1_minus_Wq[RTE_RED_WQ_LOG2_NUM];
+
+/**
+ * table[i] = 2^(i/16) * Scale
+ */
+uint16_t rte_red_pow2_frac_inv[16];
+
+/**
+ * @brief Initialize tables used to compute average
+ * queue size when queue is empty.
+ */
+static void
+__rte_red_init_tables(void)
+{
+ uint32_t i = 0;
+ double scale = 0.0;
+ double table_size = 0.0;
+
+ scale = (double)(1 << RTE_RED_SCALING);
+ table_size = (double)(RTE_DIM(rte_red_pow2_frac_inv));
+
+ for (i = 0; i < RTE_DIM(rte_red_pow2_frac_inv); i++) {
+ double m = (double)i;
+
+ rte_red_pow2_frac_inv[i] = (uint16_t) round(scale / pow(2, m / table_size));
+ }
+
+ scale = 1024.0;
+
+ RTE_ASSERT(RTE_RED_WQ_LOG2_NUM == RTE_DIM(rte_red_log2_1_minus_Wq));
+
+ for (i = RTE_RED_WQ_LOG2_MIN; i <= RTE_RED_WQ_LOG2_MAX; i++) {
+ double n = (double)i;
+ double Wq = pow(2, -n);
+ uint32_t index = i - RTE_RED_WQ_LOG2_MIN;
+
+ rte_red_log2_1_minus_Wq[index] = (uint16_t) round(-1.0 * scale * log2(1.0 - Wq));
+ /**
+ * Table entry of zero, corresponds to a Wq of zero
+ * which is not valid (avg would remain constant no
+ * matter how long the queue is empty). So we have
+ * to check for zero and round up to one.
+ */
+ if (rte_red_log2_1_minus_Wq[index] == 0) {
+ rte_red_log2_1_minus_Wq[index] = 1;
+ }
+ }
+}
+
+int
+rte_red_rt_data_init(struct rte_red *red)
+{
+ if (red == NULL)
+ return -1;
+
+ red->avg = 0;
+ red->count = 0;
+ red->q_time = 0;
+ return 0;
+}
+
+int
+rte_red_config_init(struct rte_red_config *red_cfg,
+ const uint16_t wq_log2,
+ const uint16_t min_th,
+ const uint16_t max_th,
+ const uint16_t maxp_inv)
+{
+ if (red_cfg == NULL) {
+ return -1;
+ }
+ if (max_th > RTE_RED_MAX_TH_MAX) {
+ return -2;
+ }
+ if (min_th >= max_th) {
+ return -3;
+ }
+ if (wq_log2 > RTE_RED_WQ_LOG2_MAX) {
+ return -4;
+ }
+ if (wq_log2 < RTE_RED_WQ_LOG2_MIN) {
+ return -5;
+ }
+ if (maxp_inv < RTE_RED_MAXP_INV_MIN) {
+ return -6;
+ }
+ if (maxp_inv > RTE_RED_MAXP_INV_MAX) {
+ return -7;
+ }
+
+ /**
+ * Initialize the RED module if not already done
+ */
+ if (!rte_red_init_done) {
+ rte_red_rand_seed = rte_rand();
+ rte_red_rand_val = rte_fast_rand();
+ __rte_red_init_tables();
+ rte_red_init_done = 1;
+ }
+
+ red_cfg->min_th = ((uint32_t) min_th) << (wq_log2 + RTE_RED_SCALING);
+ red_cfg->max_th = ((uint32_t) max_th) << (wq_log2 + RTE_RED_SCALING);
+ red_cfg->pa_const = (2 * (max_th - min_th) * maxp_inv) << RTE_RED_SCALING;
+ red_cfg->maxp_inv = maxp_inv;
+ red_cfg->wq_log2 = wq_log2;
+
+ return 0;
+}
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_red.h b/src/spdk/dpdk/lib/librte_sched/rte_red.h
new file mode 100644
index 000000000..36273cac6
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_red.h
@@ -0,0 +1,411 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#ifndef __RTE_RED_H_INCLUDED__
+#define __RTE_RED_H_INCLUDED__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @file
+ * RTE Random Early Detection (RED)
+ *
+ *
+ ***/
+
+#include <stdint.h>
+#include <limits.h>
+#include <rte_common.h>
+#include <rte_debug.h>
+#include <rte_cycles.h>
+#include <rte_branch_prediction.h>
+
+#define RTE_RED_SCALING 10 /**< Fraction size for fixed-point */
+#define RTE_RED_S (1 << 22) /**< Packet size multiplied by number of leaf queues */
+#define RTE_RED_MAX_TH_MAX 1023 /**< Max threshold limit in fixed point format */
+#define RTE_RED_WQ_LOG2_MIN 1 /**< Min inverse filter weight value */
+#define RTE_RED_WQ_LOG2_MAX 12 /**< Max inverse filter weight value */
+#define RTE_RED_MAXP_INV_MIN 1 /**< Min inverse mark probability value */
+#define RTE_RED_MAXP_INV_MAX 255 /**< Max inverse mark probability value */
+#define RTE_RED_2POW16 (1<<16) /**< 2 power 16 */
+#define RTE_RED_INT16_NBITS (sizeof(uint16_t) * CHAR_BIT)
+#define RTE_RED_WQ_LOG2_NUM (RTE_RED_WQ_LOG2_MAX - RTE_RED_WQ_LOG2_MIN + 1)
+
+/**
+ * Externs
+ *
+ */
+extern uint32_t rte_red_rand_val;
+extern uint32_t rte_red_rand_seed;
+extern uint16_t rte_red_log2_1_minus_Wq[RTE_RED_WQ_LOG2_NUM];
+extern uint16_t rte_red_pow2_frac_inv[16];
+
+/**
+ * RED configuration parameters passed by user
+ *
+ */
+struct rte_red_params {
+ uint16_t min_th; /**< Minimum threshold for queue (max_th) */
+ uint16_t max_th; /**< Maximum threshold for queue (max_th) */
+ uint16_t maxp_inv; /**< Inverse of packet marking probability maximum value (maxp = 1 / maxp_inv) */
+ uint16_t wq_log2; /**< Negated log2 of queue weight (wq = 1 / (2 ^ wq_log2)) */
+};
+
+/**
+ * RED configuration parameters
+ */
+struct rte_red_config {
+ uint32_t min_th; /**< min_th scaled in fixed-point format */
+ uint32_t max_th; /**< max_th scaled in fixed-point format */
+ uint32_t pa_const; /**< Precomputed constant value used for pa calculation (scaled in fixed-point format) */
+ uint8_t maxp_inv; /**< maxp_inv */
+ uint8_t wq_log2; /**< wq_log2 */
+};
+
+/**
+ * RED run-time data
+ */
+struct rte_red {
+ uint32_t avg; /**< Average queue size (avg), scaled in fixed-point format */
+ uint32_t count; /**< Number of packets since last marked packet (count) */
+ uint64_t q_time; /**< Start of the queue idle time (q_time) */
+};
+
+/**
+ * @brief Initialises run-time data
+ *
+ * @param red [in,out] data pointer to RED runtime data
+ *
+ * @return Operation status
+ * @retval 0 success
+ * @retval !0 error
+ */
+int
+rte_red_rt_data_init(struct rte_red *red);
+
+/**
+ * @brief Configures a single RED configuration parameter structure.
+ *
+ * @param red_cfg [in,out] config pointer to a RED configuration parameter structure
+ * @param wq_log2 [in] log2 of the filter weight, valid range is:
+ * RTE_RED_WQ_LOG2_MIN <= wq_log2 <= RTE_RED_WQ_LOG2_MAX
+ * @param min_th [in] queue minimum threshold in number of packets
+ * @param max_th [in] queue maximum threshold in number of packets
+ * @param maxp_inv [in] inverse maximum mark probability
+ *
+ * @return Operation status
+ * @retval 0 success
+ * @retval !0 error
+ */
+int
+rte_red_config_init(struct rte_red_config *red_cfg,
+ const uint16_t wq_log2,
+ const uint16_t min_th,
+ const uint16_t max_th,
+ const uint16_t maxp_inv);
+
+/**
+ * @brief Generate random number for RED
+ *
+ * Implementation based on:
+ * http://software.intel.com/en-us/articles/fast-random-number-generator-on-the-intel-pentiumr-4-processor/
+ *
+ * 10 bit shift has been found through empirical tests (was 16).
+ *
+ * @return Random number between 0 and (2^22 - 1)
+ */
+static inline uint32_t
+rte_fast_rand(void)
+{
+ rte_red_rand_seed = (214013 * rte_red_rand_seed) + 2531011;
+ return rte_red_rand_seed >> 10;
+}
+
+/**
+ * @brief calculate factor to scale average queue size when queue
+ * becomes empty
+ *
+ * @param wq_log2 [in] where EWMA filter weight wq = 1/(2 ^ wq_log2)
+ * @param m [in] exponent in the computed value (1 - wq) ^ m
+ *
+ * @return computed value
+ * @retval ((1 - wq) ^ m) scaled in fixed-point format
+ */
+static inline uint16_t
+__rte_red_calc_qempty_factor(uint8_t wq_log2, uint16_t m)
+{
+ uint32_t n = 0;
+ uint32_t f = 0;
+
+ /**
+ * Basic math tells us that:
+ * a^b = 2^(b * log2(a) )
+ *
+ * in our case:
+ * a = (1-Wq)
+ * b = m
+ * Wq = 1/ (2^log2n)
+ *
+ * So we are computing this equation:
+ * factor = 2 ^ ( m * log2(1-Wq))
+ *
+ * First we are computing:
+ * n = m * log2(1-Wq)
+ *
+ * To avoid dealing with signed numbers log2 values are positive
+ * but they should be negative because (1-Wq) is always < 1.
+ * Contents of log2 table values are also scaled for precision.
+ */
+
+ n = m * rte_red_log2_1_minus_Wq[wq_log2 - RTE_RED_WQ_LOG2_MIN];
+
+ /**
+ * The tricky part is computing 2^n, for this I split n into
+ * integer part and fraction part.
+ * f - is fraction part of n
+ * n - is integer part of original n
+ *
+ * Now using basic math we compute 2^n:
+ * 2^(f+n) = 2^f * 2^n
+ * 2^f - we use lookup table
+ * 2^n - can be replaced with bit shift right operations
+ */
+
+ f = (n >> 6) & 0xf;
+ n >>= 10;
+
+ if (n < RTE_RED_SCALING)
+ return (uint16_t) ((rte_red_pow2_frac_inv[f] + (1 << (n - 1))) >> n);
+
+ return 0;
+}
+
+/**
+ * @brief Updates queue average in condition when queue is empty
+ *
+ * Note: packet is never dropped in this particular case.
+ *
+ * @param red_cfg [in] config pointer to a RED configuration parameter structure
+ * @param red [in,out] data pointer to RED runtime data
+ * @param time [in] current time stamp
+ *
+ * @return Operation status
+ * @retval 0 enqueue the packet
+ * @retval 1 drop the packet based on max threshold criterion
+ * @retval 2 drop the packet based on mark probability criterion
+ */
+static inline int
+rte_red_enqueue_empty(const struct rte_red_config *red_cfg,
+ struct rte_red *red,
+ const uint64_t time)
+{
+ uint64_t time_diff = 0, m = 0;
+
+ RTE_ASSERT(red_cfg != NULL);
+ RTE_ASSERT(red != NULL);
+
+ red->count ++;
+
+ /**
+ * We compute avg but we don't compare avg against
+ * min_th or max_th, nor calculate drop probability
+ */
+ time_diff = time - red->q_time;
+
+ /**
+ * m is the number of packets that might have arrived while the queue was empty.
+ * In this case we have time stamps provided by scheduler in byte units (bytes
+ * transmitted on network port). Such time stamp translates into time units as
+ * port speed is fixed but such approach simplifies the code.
+ */
+ m = time_diff / RTE_RED_S;
+
+ /**
+ * Check that m will fit into 16-bit unsigned integer
+ */
+ if (m >= RTE_RED_2POW16) {
+ red->avg = 0;
+ } else {
+ red->avg = (red->avg >> RTE_RED_SCALING) * __rte_red_calc_qempty_factor(red_cfg->wq_log2, (uint16_t) m);
+ }
+
+ return 0;
+}
+
+/**
+ * Drop probability (Sally Floyd and Van Jacobson):
+ *
+ * pb = (1 / maxp_inv) * (avg - min_th) / (max_th - min_th)
+ * pa = pb / (2 - count * pb)
+ *
+ *
+ * (1 / maxp_inv) * (avg - min_th)
+ * ---------------------------------
+ * max_th - min_th
+ * pa = -----------------------------------------------
+ * count * (1 / maxp_inv) * (avg - min_th)
+ * 2 - -----------------------------------------
+ * max_th - min_th
+ *
+ *
+ * avg - min_th
+ * pa = -----------------------------------------------------------
+ * 2 * (max_th - min_th) * maxp_inv - count * (avg - min_th)
+ *
+ *
+ * We define pa_const as: pa_const = 2 * (max_th - min_th) * maxp_inv. Then:
+ *
+ *
+ * avg - min_th
+ * pa = -----------------------------------
+ * pa_const - count * (avg - min_th)
+ */
+
+/**
+ * @brief make a decision to drop or enqueue a packet based on mark probability
+ * criteria
+ *
+ * @param red_cfg [in] config pointer to structure defining RED parameters
+ * @param red [in,out] data pointer to RED runtime data
+ *
+ * @return operation status
+ * @retval 0 enqueue the packet
+ * @retval 1 drop the packet
+ */
+static inline int
+__rte_red_drop(const struct rte_red_config *red_cfg, struct rte_red *red)
+{
+ uint32_t pa_num = 0; /* numerator of drop-probability */
+ uint32_t pa_den = 0; /* denominator of drop-probability */
+ uint32_t pa_num_count = 0;
+
+ pa_num = (red->avg - red_cfg->min_th) >> (red_cfg->wq_log2);
+
+ pa_num_count = red->count * pa_num;
+
+ if (red_cfg->pa_const <= pa_num_count)
+ return 1;
+
+ pa_den = red_cfg->pa_const - pa_num_count;
+
+ /* If drop, generate and save random number to be used next time */
+ if (unlikely((rte_red_rand_val % pa_den) < pa_num)) {
+ rte_red_rand_val = rte_fast_rand();
+
+ return 1;
+ }
+
+ /* No drop */
+ return 0;
+}
+
+/**
+ * @brief Decides if new packet should be enqeued or dropped in queue non-empty case
+ *
+ * @param red_cfg [in] config pointer to a RED configuration parameter structure
+ * @param red [in,out] data pointer to RED runtime data
+ * @param q [in] current queue size (measured in packets)
+ *
+ * @return Operation status
+ * @retval 0 enqueue the packet
+ * @retval 1 drop the packet based on max threshold criterion
+ * @retval 2 drop the packet based on mark probability criterion
+ */
+static inline int
+rte_red_enqueue_nonempty(const struct rte_red_config *red_cfg,
+ struct rte_red *red,
+ const unsigned q)
+{
+ RTE_ASSERT(red_cfg != NULL);
+ RTE_ASSERT(red != NULL);
+
+ /**
+ * EWMA filter (Sally Floyd and Van Jacobson):
+ * avg = (1 - wq) * avg + wq * q
+ * avg = avg + q * wq - avg * wq
+ *
+ * We select: wq = 2^(-n). Let scaled version of avg be: avg_s = avg * 2^(N+n). We get:
+ * avg_s = avg_s + q * 2^N - avg_s * 2^(-n)
+ *
+ * By using shift left/right operations, we get:
+ * avg_s = avg_s + (q << N) - (avg_s >> n)
+ * avg_s += (q << N) - (avg_s >> n)
+ */
+
+ /* avg update */
+ red->avg += (q << RTE_RED_SCALING) - (red->avg >> red_cfg->wq_log2);
+
+ /* avg < min_th: do not mark the packet */
+ if (red->avg < red_cfg->min_th) {
+ red->count ++;
+ return 0;
+ }
+
+ /* min_th <= avg < max_th: mark the packet with pa probability */
+ if (red->avg < red_cfg->max_th) {
+ if (!__rte_red_drop(red_cfg, red)) {
+ red->count ++;
+ return 0;
+ }
+
+ red->count = 0;
+ return 2;
+ }
+
+ /* max_th <= avg: always mark the packet */
+ red->count = 0;
+ return 1;
+}
+
+/**
+ * @brief Decides if new packet should be enqeued or dropped
+ * Updates run time data based on new queue size value.
+ * Based on new queue average and RED configuration parameters
+ * gives verdict whether to enqueue or drop the packet.
+ *
+ * @param red_cfg [in] config pointer to a RED configuration parameter structure
+ * @param red [in,out] data pointer to RED runtime data
+ * @param q [in] updated queue size in packets
+ * @param time [in] current time stamp
+ *
+ * @return Operation status
+ * @retval 0 enqueue the packet
+ * @retval 1 drop the packet based on max threshold criteria
+ * @retval 2 drop the packet based on mark probability criteria
+ */
+static inline int
+rte_red_enqueue(const struct rte_red_config *red_cfg,
+ struct rte_red *red,
+ const unsigned q,
+ const uint64_t time)
+{
+ RTE_ASSERT(red_cfg != NULL);
+ RTE_ASSERT(red != NULL);
+
+ if (q != 0) {
+ return rte_red_enqueue_nonempty(red_cfg, red, q);
+ } else {
+ return rte_red_enqueue_empty(red_cfg, red, time);
+ }
+}
+
+/**
+ * @brief Callback to records time that queue became empty
+ *
+ * @param red [in,out] data pointer to RED runtime data
+ * @param time [in] current time stamp
+ */
+static inline void
+rte_red_mark_queue_empty(struct rte_red *red, const uint64_t time)
+{
+ red->q_time = time;
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __RTE_RED_H_INCLUDED__ */
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_sched.c b/src/spdk/dpdk/lib/librte_sched/rte_sched.c
new file mode 100644
index 000000000..c0983ddda
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_sched.c
@@ -0,0 +1,2755 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#include <stdio.h>
+#include <string.h>
+
+#include <rte_common.h>
+#include <rte_log.h>
+#include <rte_memory.h>
+#include <rte_malloc.h>
+#include <rte_cycles.h>
+#include <rte_prefetch.h>
+#include <rte_branch_prediction.h>
+#include <rte_mbuf.h>
+#include <rte_bitmap.h>
+#include <rte_reciprocal.h>
+
+#include "rte_sched.h"
+#include "rte_sched_common.h"
+#include "rte_approx.h"
+
+#ifdef __INTEL_COMPILER
+#pragma warning(disable:2259) /* conversion may lose significant bits */
+#endif
+
+#ifdef RTE_SCHED_VECTOR
+#include <rte_vect.h>
+
+#ifdef RTE_ARCH_X86
+#define SCHED_VECTOR_SSE4
+#elif defined(RTE_MACHINE_CPUFLAG_NEON)
+#define SCHED_VECTOR_NEON
+#endif
+
+#endif
+
+#define RTE_SCHED_TB_RATE_CONFIG_ERR (1e-7)
+#define RTE_SCHED_WRR_SHIFT 3
+#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
+
+/* 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) */
+ uint64_t tb_period;
+ uint64_t tb_credits_per_period;
+ uint64_t tb_size;
+
+ /* Pipe traffic classes */
+ 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 */
+ uint64_t tb_credits;
+
+ /* Pipe profile and flags */
+ uint32_t profile;
+
+ /* Traffic classes (TCs) */
+ uint64_t tc_time; /* time of next update */
+ uint64_t tc_credits[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /* Weighted Round Robin (WRR) */
+ uint8_t wrr_tokens[RTE_SCHED_BE_QUEUES_PER_PIPE];
+
+ /* TC oversubscription */
+ uint64_t tc_ov_credits;
+ uint8_t tc_ov_period_id;
+} __rte_cache_aligned;
+
+struct rte_sched_queue {
+ uint16_t qw;
+ uint16_t qr;
+};
+
+struct rte_sched_queue_extra {
+ struct rte_sched_queue_stats stats;
+#ifdef RTE_SCHED_RED
+ struct rte_red red;
+#endif
+};
+
+enum grinder_state {
+ e_GRINDER_PREFETCH_PIPE = 0,
+ e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS,
+ e_GRINDER_PREFETCH_MBUF,
+ e_GRINDER_READ_MBUF
+};
+
+struct rte_sched_grinder {
+ /* Pipe cache */
+ uint16_t pcache_qmask[RTE_SCHED_GRINDER_PCACHE_SIZE];
+ 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;
+ uint32_t pindex;
+ struct rte_sched_subport *subport;
+ struct rte_sched_pipe *pipe;
+ struct rte_sched_pipe_profile *pipe_params;
+
+ /* TC cache */
+ 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_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;
+
+ /* 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_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][RTE_COLORS];
+#endif
+
+ /* Scheduling loop detection */
+ uint32_t pipe_loop;
+ uint32_t pipe_exhaustion;
+
+ /* 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;
+
+ /* Queue base calculation */
+ uint32_t qsize_add[RTE_SCHED_QUEUES_PER_PIPE];
+ uint32_t qsize_sum;
+
+ struct rte_sched_pipe *pipe;
+ struct rte_sched_queue *queue;
+ struct rte_sched_queue_extra *queue_extra;
+ struct rte_sched_pipe_profile *pipe_profiles;
+ uint8_t *bmp_array;
+ struct rte_mbuf **queue_array;
+ uint8_t memory[0] __rte_cache_aligned;
+} __rte_cache_aligned;
+
+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;
+
+ /* 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 */
+
+ /* 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_subport_pipe_queues(struct rte_sched_subport *subport)
+{
+ return RTE_SCHED_QUEUES_PER_PIPE * subport->n_pipes_per_subport_enabled;
+}
+
+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)
+{
+ 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
+pipe_profile_check(struct rte_sched_pipe_params *params,
+ uint32_t rate, uint16_t *qsize)
+{
+ uint32_t i;
+
+ /* Pipe parameters */
+ if (params == NULL) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for parameter params\n", __func__);
+ return -EINVAL;
+ }
+
+ /* 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;
+ }
+
+ /* TB size: non-zero */
+ if (params->tb_size == 0) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for tb size\n", __func__);
+ return -EINVAL;
+ }
+
+ /* 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;
+ }
+ }
+
+ 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;
+ }
+
+ /* TC period: non-zero */
+ if (params->tc_period == 0) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for tc period\n", __func__);
+ return -EINVAL;
+ }
+
+ /* 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;
+ }
+
+ /* 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;
+ }
+ }
+
+ 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;
+ }
+
+ /* socket */
+ if (params->socket < 0) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for socket id\n", __func__);
+ return -EINVAL;
+ }
+
+ /* rate */
+ if (params->rate == 0) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for rate\n", __func__);
+ return -EINVAL;
+ }
+
+ /* 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_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++) {
+ 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_subport * size_per_pipe_queue_array;
+
+ base = 0;
+
+ 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;
+}
+
+static void
+rte_sched_subport_config_qsize(struct rte_sched_subport *subport)
+{
+ 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_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_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]);
+}
+
+static inline uint64_t
+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_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 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, uint32_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;
+ 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;
+ }
+ }
+
+ 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;
+ }
+ }
+
+ 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, &params->qsize[0]);
+ if (status != 0) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Pipe profile check failed(%d)\n", __func__, status);
+ return -EINVAL;
+ }
+ }
+
+ 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;
+ }
+ }
+
+ 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;
+ 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 */
+ 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;
+ }
+
+ /* 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;
+
+ /* Timing */
+ port->time_cpu_cycles = rte_get_tsc_cycles();
+ port->time_cpu_bytes = 0;
+ port->time = 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);
+
+ /* Grinders */
+ port->pkts_out = NULL;
+ port->n_pkts_out = 0;
+ port->subport_id = 0;
+
+ return port;
+}
+
+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_bitmap_free(subport->bmp);
+}
+
+void
+rte_sched_port_free(struct rte_sched_port *port)
+{
+ uint32_t i;
+
+ /* Check user parameters */
+ if (port == NULL)
+ 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->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_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,
+ uint32_t subport_id,
+ struct rte_sched_subport_params *params)
+{
+ 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) {
+ RTE_LOG(ERR, SCHED,
+ "%s: Incorrect value for parameter port\n", __func__);
+ return 0;
+ }
+
+ 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;
+ }
+
+ 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;
+ }
+
+ /* 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;
+ }
+
+ n_subports++;
+
+ /* Port */
+ port->subports[subport_id] = s;
+
+ /* Token Bucket (TB) */
+ if (params->tb_rate == port->rate) {
+ s->tb_credits_per_period = 1;
+ s->tb_period = 1;
+ } else {
+ double tb_rate = ((double) params->tb_rate) / ((double) port->rate);
+ double d = RTE_SCHED_TB_RATE_CONFIG_ERR;
+
+ 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 = 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++)
+ 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_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;
+ 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 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) {
+ 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 = 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;
+
+ /* Unplug pipe from its subport */
+ 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);
+ }
+
+ /* Reset the pipe */
+ memset(p, 0, sizeof(struct rte_sched_pipe));
+ }
+
+ if (deactivate)
+ return 0;
+
+ /* Apply the new pipe configuration */
+ p->profile = 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++)
+ 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;
+ }
+
+ 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;
+ }
+
+ /* 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;
+ }
+
+ 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)
+{
+ struct rte_sched_subport *s;
+
+ /* Check user parameters */
+ 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;
+ }
+
+ 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 oversubscription status */
+ *tc_ov = s->tc_ov;
+
+ return 0;
+}
+
+int
+rte_sched_queue_read_stats(struct rte_sched_port *port,
+ uint32_t queue_id,
+ 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) {
+ 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;
+ }
+ 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;
+}
+
+#ifdef RTE_SCHED_DEBUG
+
+static inline int
+rte_sched_port_queue_is_empty(struct rte_sched_subport *subport,
+ uint32_t qindex)
+{
+ struct rte_sched_queue *queue = subport->queue + qindex;
+
+ return queue->qr == queue->qw;
+}
+
+#endif /* RTE_SCHED_DEBUG */
+
+#ifdef RTE_SCHED_COLLECT_STATS
+
+static inline void
+rte_sched_port_update_subport_stats(struct rte_sched_port *port,
+ struct rte_sched_subport *subport,
+ uint32_t qindex,
+ struct rte_mbuf *pkt)
+{
+ 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,
+ struct rte_sched_subport *subport,
+ uint32_t qindex,
+ struct rte_mbuf *pkt,
+ __rte_unused uint32_t red)
+#endif
+{
+ 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_subport *subport,
+ uint32_t qindex,
+ struct rte_mbuf *pkt)
+{
+ 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_subport *subport,
+ uint32_t qindex,
+ struct rte_mbuf *pkt,
+ __rte_unused uint32_t red)
+#endif
+{
+ 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 */
+
+#ifdef RTE_SCHED_RED
+
+static inline int
+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;
+ uint32_t tc_index;
+ enum rte_color color;
+
+ tc_index = rte_sched_port_pipe_tc(port, qindex);
+ color = rte_sched_port_pkt_read_color(pkt);
+ red_cfg = &subport->red_config[tc_index][color];
+
+ if ((red_cfg->min_th | red_cfg->max_th) == 0)
+ return 0;
+
+ 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,
+ struct rte_sched_subport *subport, uint32_t qindex)
+{
+ 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
+
+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, subport, qindex)
+
+#endif /* RTE_SCHED_RED */
+
+#ifdef RTE_SCHED_DEBUG
+
+static inline void
+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)
+ 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(subport, bmp_pos + i)) {
+ printf("Queue %u (slab offset %u) is empty\n", bmp_pos + i, i);
+ panic = 1;
+ }
+ }
+ }
+
+ 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_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 qindex = rte_mbuf_sched_queue_get(pkt);
+ uint32_t subport_queue_id = subport_qmask & qindex;
+
+ q = subport->queue + subport_queue_id;
+ rte_prefetch0(q);
+#ifdef RTE_SCHED_COLLECT_STATS
+ qe = subport->queue_extra + subport_queue_id;
+ rte_prefetch0(qe);
+#endif
+
+ return subport_queue_id;
+}
+
+static inline void
+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 = 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(subport->bmp, qindex);
+}
+
+static inline int
+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 = 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, subport, pkt, qindex, qlen) ||
+ (qlen >= qsize))) {
+ rte_pktmbuf_free(pkt);
+#ifdef RTE_SCHED_COLLECT_STATS
+ 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 subport bitmap */
+ rte_bitmap_set(subport->bmp, qindex);
+
+ /* Statistics */
+#ifdef RTE_SCHED_COLLECT_STATS
+ rte_sched_port_update_subport_stats(port, subport, qindex, pkt);
+ rte_sched_port_update_queue_stats(subport, qindex, pkt);
+#endif
+
+ return 1;
+}
+
+
+/*
+ * 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 |
+ * ----->|_______|----->|_______|----->|_______|----->|_______|----->
+ * p01 p11 p21 p31
+ *
+ */
+int
+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_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;
+ 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++)
+ 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(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_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, 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);
+
+ 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);
+
+ 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) {
+ /* Propagate stage inputs */
+ pkt30 = pkt20;
+ pkt31 = pkt21;
+ pkt20 = pkt10;
+ pkt21 = pkt11;
+ pkt10 = pkt00;
+ pkt11 = pkt01;
+ q30 = q20;
+ 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 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_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, 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.
+ */
+ pkt_last = pkts[n_pkts - 1];
+ rte_prefetch0(pkt_last);
+
+ 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;
+
+ 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_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, 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, subport_last,
+ q_last, q_last_base, pkt_last);
+ result += r_last;
+ }
+
+ return result;
+}
+
+#ifndef RTE_SCHED_SUBPORT_TC_OV
+
+static inline void
+grinder_credits_update(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ 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_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_MIN(pipe->tb_credits, params->tb_size);
+ pipe->tb_time += n_periods * params->tb_period;
+
+ /* Subport TCs */
+ if (unlikely(port->time >= subport->tc_time)) {
+ 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)) {
+ 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,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ 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_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_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_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)) {
+ 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)) {
+ 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,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ 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;
+ 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) &&
+ (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;
+ pipe->tb_credits -= pkt_len;
+ pipe->tc_credits[tc_index] -= pkt_len;
+
+ return 1;
+}
+
+#else
+
+static inline int
+grinder_credits_check(struct rte_sched_port *port,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ 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;
+ 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)
+ 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 -= 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,
+ struct rte_sched_subport *subport, uint32_t 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_len + port->frame_overhead;
+ uint32_t be_tc_active;
+
+ if (!grinder_credits_check(port, subport, pos))
+ return 0;
+
+ /* Advance port time */
+ port->time += pkt_len;
+
+ /* Send packet */
+ 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(subport->bmp, qindex);
+ grinder->qmask &= ~(1 << grinder->qpos);
+ if (be_tc_active)
+ grinder->wrr_mask[grinder->qpos] = 0;
+ rte_sched_port_set_queue_empty_timestamp(port, subport, qindex);
+ }
+
+ /* Reset pipe loop detection */
+ subport->pipe_loop = RTE_SCHED_PIPE_INVALID;
+ grinder->productive = 1;
+
+ return 1;
+}
+
+#ifdef SCHED_VECTOR_SSE4
+
+static inline int
+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 *)subport->grinder_base_bmp_pos);
+ __m128i res = _mm_cmpeq_epi32(pipes, index);
+
+ pipes = _mm_load_si128((__m128i *)(subport->grinder_base_bmp_pos + 4));
+ pipes = _mm_cmpeq_epi32(pipes, index);
+ res = _mm_or_si128(res, pipes);
+
+ if (_mm_testz_si128(res, res))
+ return 0;
+
+ 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_subport *subport, uint32_t base_pipe)
+{
+ uint32_t i;
+
+ 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_subport *subport,
+ uint32_t pos, uint32_t bmp_pos, uint64_t bmp_slab)
+{
+ 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_subport *subport,
+ uint32_t pos, uint32_t qindex, uint16_t qmask)
+{
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ uint8_t b, i;
+
+ grinder->tccache_w = 0;
+ grinder->tccache_r = 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,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_mbuf **qbase;
+ uint32_t qindex;
+ uint16_t qsize;
+
+ if (grinder->tccache_r == grinder->tccache_w)
+ return 0;
+
+ qindex = grinder->tccache_qindex[grinder->tccache_r];
+ qbase = rte_sched_subport_pipe_qbase(subport, qindex);
+ qsize = rte_sched_subport_pipe_qsize(port, subport, qindex);
+
+ grinder->tc_index = rte_sched_port_pipe_tc(port, qindex);
+ grinder->qmask = grinder->tccache_qmask[grinder->tccache_r];
+ grinder->qsize = qsize;
+
+ 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->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,
+ struct rte_sched_subport *subport, uint32_t 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++;
+ } else {
+ uint64_t bmp_slab = 0;
+ uint32_t bmp_pos = 0;
+
+ /* Get another non-empty pipe group */
+ if (unlikely(rte_bitmap_scan(subport->bmp, &bmp_pos, &bmp_slab) <= 0))
+ return 0;
+
+#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 */
+ subport->grinder_base_bmp_pos[pos] = RTE_SCHED_BMP_POS_INVALID;
+ if (unlikely(grinder_pipe_exists(subport, bmp_pos)))
+ return 0;
+
+ subport->grinder_base_bmp_pos[pos] = bmp_pos;
+
+ /* Install new pipe group into grinder's pipe cache */
+ 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 = 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(subport, pos, pipe_qindex, pipe_qmask);
+ grinder_next_tc(port, subport, pos);
+
+ /* Check for pipe exhaustion */
+ if (grinder->pindex == subport->pipe_loop) {
+ subport->pipe_exhaustion = 1;
+ subport->pipe_loop = RTE_SCHED_PIPE_INVALID;
+ }
+
+ return 1;
+}
+
+
+static inline void
+grinder_wrr_load(struct rte_sched_subport *subport, uint32_t 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 qmask = grinder->qmask;
+
+ 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[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_subport *subport, uint32_t pos)
+{
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+ struct rte_sched_pipe *pipe = grinder->pipe;
+
+ 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_subport *subport, uint32_t 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;
+}
+
+
+#define grinder_evict(subport, pos)
+
+static inline void
+grinder_prefetch_pipe(struct rte_sched_subport *subport, uint32_t 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_subport *subport, uint32_t pos)
+{
+ 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(subport, pos);
+ grinder_wrr(subport, pos);
+
+ rte_prefetch0(grinder->qbase[2] + qr[2]);
+ rte_prefetch0(grinder->qbase[3] + qr[3]);
+}
+
+static inline void
+grinder_prefetch_mbuf(struct rte_sched_subport *subport, uint32_t 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,
+ struct rte_sched_subport *subport, uint32_t pos)
+{
+ struct rte_sched_grinder *grinder = subport->grinder + pos;
+
+ switch (grinder->state) {
+ case e_GRINDER_PREFETCH_PIPE:
+ {
+ 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 = 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(subport, pos);
+
+ grinder->state = e_GRINDER_READ_MBUF;
+ return 0;
+ }
+
+ case e_GRINDER_READ_MBUF:
+ {
+ 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) {
+ if (wrr_active)
+ grinder_wrr(subport, pos);
+
+ grinder_prefetch_mbuf(subport, pos);
+
+ return 1;
+ }
+
+ if (wrr_active)
+ grinder_wrr_store(subport, pos);
+
+ /* Look for another active TC within same pipe */
+ 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 &&
+ 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, subport, pos)) {
+ grinder_prefetch_pipe(subport, pos);
+
+ grinder->state = e_GRINDER_PREFETCH_TC_QUEUE_ARRAYS;
+ return result;
+ }
+
+ /* No active pipe found */
+ subport->busy_grinders--;
+
+ grinder->state = e_GRINDER_PREFETCH_PIPE;
+ return result;
+ }
+
+ default:
+ rte_panic("Algorithmic error (invalid state)\n");
+ return 0;
+ }
+}
+
+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;
+ uint64_t bytes_diff;
+ uint32_t i;
+
+ /* 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 += bytes_diff;
+ if (port->time < port->time_cpu_bytes)
+ port->time = port->time_cpu_bytes;
+
+ /* Reset pipe loop detection */
+ 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_subport *subport, int second_pass)
+{
+ int exceptions;
+
+ /* Check if any exception flag is set */
+ exceptions = (second_pass && subport->busy_grinders == 0) ||
+ (subport->pipe_exhaustion == 1);
+
+ /* Clear exception flags */
+ subport->pipe_exhaustion = 0;
+
+ return exceptions;
+}
+
+int
+rte_sched_port_dequeue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts)
+{
+ 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++) {
+ 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;
+}
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_sched.h b/src/spdk/dpdk/lib/librte_sched/rte_sched.h
new file mode 100644
index 000000000..8a5a93c98
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_sched.h
@@ -0,0 +1,514 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#ifndef __INCLUDE_RTE_SCHED_H__
+#define __INCLUDE_RTE_SCHED_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @file
+ * RTE Hierarchical Scheduler
+ *
+ * The hierarchical scheduler prioritizes the transmission of packets
+ * from different users and traffic classes according to the Service
+ * Level Agreements (SLAs) defined for the current network node.
+ *
+ * The scheduler supports thousands of packet queues grouped under a
+ * 5-level hierarchy:
+ * 1. Port:
+ * - Typical usage: output Ethernet port;
+ * - Multiple ports are scheduled in round robin order with
+ * equal priority;
+ * 2. Subport:
+ * - Typical usage: group of users;
+ * - Traffic shaping using the token bucket algorithm
+ * (one bucket per subport);
+ * - Upper limit enforced per traffic class at subport level;
+ * - Lower priority traffic classes able to reuse subport
+ * bandwidth currently unused by higher priority traffic
+ * classes of the same subport;
+ * - When any subport traffic class is oversubscribed
+ * (configuration time event), the usage of subport member
+ * pipes with high demand for that traffic class pipes is
+ * truncated to a dynamically adjusted value with no
+ * impact to low demand pipes;
+ * 3. Pipe:
+ * - Typical usage: individual user/subscriber;
+ * - Traffic shaping using the token bucket algorithm
+ * (one bucket per pipe);
+ * 4. Traffic class:
+ * - Traffic classes of the same pipe handled in strict
+ * priority order;
+ * - Upper limit enforced per traffic class at the pipe level;
+ * - Lower priority traffic classes able to reuse pipe
+ * bandwidth currently unused by higher priority traffic
+ * classes of the same pipe;
+ * 5. Queue:
+ * - Typical usage: queue hosting packets from one or
+ * multiple connections of same traffic class belonging to
+ * the same user;
+ * - Weighted Round Robin (WRR) is used to service the
+ * queues within same pipe lowest priority traffic class (best-effort).
+ *
+ */
+
+#include <sys/types.h>
+#include <rte_compat.h>
+#include <rte_mbuf.h>
+#include <rte_meter.h>
+
+/** Random Early Detection (RED) */
+#ifdef RTE_SCHED_RED
+#include "rte_red.h"
+#endif
+
+/** Maximum number of queues per pipe.
+ * Note that the multiple queues (power of 2) can only be assigned to
+ * lowest priority (best-effort) traffic class. Other higher priority traffic
+ * classes can only have one queue.
+ * Can not change.
+ *
+ * @see struct rte_sched_port_params
+ */
+#define RTE_SCHED_QUEUES_PER_PIPE 16
+
+/** Number of WRR queues for best-effort traffic class per pipe.
+ *
+ * @see struct rte_sched_pipe_params
+ */
+#define RTE_SCHED_BE_QUEUES_PER_PIPE 4
+
+/** Number of traffic classes per pipe (as well as subport).
+ * @see struct rte_sched_subport_params
+ * @see struct rte_sched_pipe_params
+ */
+#define RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE \
+(RTE_SCHED_QUEUES_PER_PIPE - RTE_SCHED_BE_QUEUES_PER_PIPE + 1)
+
+/** Best-effort traffic class ID
+ * Can not change.
+ */
+#define RTE_SCHED_TRAFFIC_CLASS_BE (RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE - 1)
+
+/*
+ * Ethernet framing overhead. Overhead fields per Ethernet frame:
+ * 1. Preamble: 7 bytes;
+ * 2. Start of Frame Delimiter (SFD): 1 byte;
+ * 3. Frame Check Sequence (FCS): 4 bytes;
+ * 4. Inter Frame Gap (IFG): 12 bytes.
+ *
+ * The FCS is considered overhead only if not included in the packet
+ * length (field pkt_len of struct rte_mbuf).
+ *
+ * @see struct rte_sched_port_params
+ */
+#ifndef RTE_SCHED_FRAME_OVERHEAD_DEFAULT
+#define RTE_SCHED_FRAME_OVERHEAD_DEFAULT 24
+#endif
+
+/*
+ * Pipe configuration parameters. The period and credits_per_period
+ * parameters are measured in bytes, with one byte meaning the time
+ * duration associated with the transmission of one byte on the
+ * physical medium of the output port, with pipe or pipe traffic class
+ * rate (measured as percentage of output port rate) determined as
+ * credits_per_period divided by period. One credit represents one
+ * byte.
+ */
+struct rte_sched_pipe_params {
+ /** Token bucket rate (measured in bytes per second) */
+ uint64_t tb_rate;
+
+ /** Token bucket size (measured in credits) */
+ uint64_t tb_size;
+
+ /** Traffic class rates (measured in bytes per second) */
+ uint64_t tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Enforcement period (measured in milliseconds) */
+ uint64_t tc_period;
+
+ /** Best-effort traffic class oversubscription weight */
+ uint8_t tc_ov_weight;
+
+ /** WRR weights of best-effort traffic class queues */
+ uint8_t wrr_weights[RTE_SCHED_BE_QUEUES_PER_PIPE];
+};
+
+/*
+ * Subport configuration parameters. The period and credits_per_period
+ * parameters are measured in bytes, with one byte meaning the time
+ * duration associated with the transmission of one byte on the
+ * physical medium of the output port, with pipe or pipe traffic class
+ * rate (measured as percentage of output port rate) determined as
+ * credits_per_period divided by period. One credit represents one
+ * byte.
+ */
+struct rte_sched_subport_params {
+ /** Token bucket rate (measured in bytes per second) */
+ uint64_t tb_rate;
+
+ /** Token bucket size (measured in credits) */
+ uint64_t tb_size;
+
+ /** Traffic class rates (measured in bytes per second) */
+ uint64_t tc_rate[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Enforcement period for rates (measured in milliseconds) */
+ uint64_t tc_period;
+
+ /** Number of subport pipes.
+ * The subport can enable/allocate fewer pipes than the maximum
+ * number set through struct port_params::n_max_pipes_per_subport,
+ * as needed, to avoid memory allocation for the queues of the
+ * pipes that are not really needed.
+ */
+ uint32_t n_pipes_per_subport_enabled;
+
+ /** Packet queue size for each traffic class.
+ * All the pipes within the same subport share the similar
+ * configuration for the queues.
+ */
+ uint16_t qsize[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Pipe profile table.
+ * Every pipe is configured using one of the profiles from this table.
+ */
+ struct rte_sched_pipe_params *pipe_profiles;
+
+ /** Profiles in the pipe profile table */
+ uint32_t n_pipe_profiles;
+
+ /** Max allowed profiles in the pipe profile table */
+ uint32_t n_max_pipe_profiles;
+
+#ifdef RTE_SCHED_RED
+ /** RED parameters */
+ struct rte_red_params red_params[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE][RTE_COLORS];
+#endif
+};
+
+/** Subport statistics */
+struct rte_sched_subport_stats {
+ /** Number of packets successfully written */
+ uint64_t n_pkts_tc[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Number of packets dropped */
+ uint64_t n_pkts_tc_dropped[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Number of bytes successfully written for each traffic class */
+ uint64_t n_bytes_tc[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+ /** Number of bytes dropped for each traffic class */
+ uint64_t n_bytes_tc_dropped[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+
+#ifdef RTE_SCHED_RED
+ /** Number of packets dropped by red */
+ uint64_t n_pkts_red_dropped[RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE];
+#endif
+};
+
+/** Queue statistics */
+struct rte_sched_queue_stats {
+ /** Packets successfully written */
+ uint64_t n_pkts;
+
+ /** Packets dropped */
+ uint64_t n_pkts_dropped;
+
+#ifdef RTE_SCHED_RED
+ /** Packets dropped by RED */
+ uint64_t n_pkts_red_dropped;
+#endif
+
+ /** Bytes successfully written */
+ uint64_t n_bytes;
+
+ /** Bytes dropped */
+ uint64_t n_bytes_dropped;
+};
+
+/** Port configuration parameters. */
+struct rte_sched_port_params {
+ /** Name of the port to be associated */
+ const char *name;
+
+ /** CPU socket ID */
+ int socket;
+
+ /** Output port rate (measured in bytes per second) */
+ uint64_t rate;
+
+ /** Maximum Ethernet frame size (measured in bytes).
+ * Should not include the framing overhead.
+ */
+ uint32_t mtu;
+
+ /** Framing overhead per packet (measured in bytes) */
+ uint32_t frame_overhead;
+
+ /** Number of subports */
+ uint32_t n_subports_per_port;
+
+ /** Maximum number of subport pipes.
+ * This parameter is used to reserve a fixed number of bits
+ * in struct rte_mbuf::sched.queue_id for the pipe_id for all
+ * the subports of the same port.
+ */
+ uint32_t n_pipes_per_subport;
+};
+
+/*
+ * Configuration
+ *
+ ***/
+
+/**
+ * Hierarchical scheduler port configuration
+ *
+ * @param params
+ * Port scheduler configuration parameter structure
+ * @return
+ * Handle to port scheduler instance upon success or NULL otherwise.
+ */
+struct rte_sched_port *
+rte_sched_port_config(struct rte_sched_port_params *params);
+
+/**
+ * Hierarchical scheduler port free
+ *
+ * @param port
+ * Handle to port scheduler instance
+ */
+void
+rte_sched_port_free(struct rte_sched_port *port);
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Hierarchical scheduler pipe profile add
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param subport_id
+ * Subport ID
+ * @param params
+ * Pipe profile parameters
+ * @param pipe_profile_id
+ * Set to valid profile id when profile is added successfully.
+ * @return
+ * 0 upon success, error code otherwise
+ */
+__rte_experimental
+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);
+
+/**
+ * Hierarchical scheduler subport configuration
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param subport_id
+ * Subport ID
+ * @param params
+ * Subport configuration parameters
+ * @return
+ * 0 upon success, error code otherwise
+ */
+int
+rte_sched_subport_config(struct rte_sched_port *port,
+ uint32_t subport_id,
+ struct rte_sched_subport_params *params);
+
+/**
+ * Hierarchical scheduler pipe configuration
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param subport_id
+ * Subport ID
+ * @param pipe_id
+ * Pipe ID within subport
+ * @param pipe_profile
+ * ID of subport-level pre-configured pipe profile
+ * @return
+ * 0 upon success, error code otherwise
+ */
+int
+rte_sched_pipe_config(struct rte_sched_port *port,
+ uint32_t subport_id,
+ uint32_t pipe_id,
+ int32_t pipe_profile);
+
+/**
+ * Hierarchical scheduler memory footprint size per port
+ *
+ * @param port_params
+ * Port scheduler configuration parameter structure
+ * @param subport_params
+ * Array of subport parameter structures
+ * @return
+ * Memory footprint size in bytes upon success, 0 otherwise
+ */
+uint32_t
+rte_sched_port_get_memory_footprint(struct rte_sched_port_params *port_params,
+ struct rte_sched_subport_params **subport_params);
+/*
+ * Statistics
+ *
+ ***/
+
+/**
+ * Hierarchical scheduler subport statistics read
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param subport_id
+ * Subport ID
+ * @param stats
+ * Pointer to pre-allocated subport statistics structure where the statistics
+ * counters should be stored
+ * @param tc_ov
+ * Pointer to pre-allocated RTE_SCHED_TRAFFIC_CLASSES_PER_PIPE-entry array
+ * where the oversubscription status for each of the subport traffic classes
+ * should be stored.
+ * @return
+ * 0 upon success, error code otherwise
+ */
+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);
+
+/**
+ * Hierarchical scheduler queue statistics read
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param queue_id
+ * Queue ID within port scheduler
+ * @param stats
+ * Pointer to pre-allocated subport statistics structure where the statistics
+ * counters should be stored
+ * @param qlen
+ * Pointer to pre-allocated variable where the current queue length
+ * should be stored.
+ * @return
+ * 0 upon success, error code otherwise
+ */
+int
+rte_sched_queue_read_stats(struct rte_sched_port *port,
+ uint32_t queue_id,
+ struct rte_sched_queue_stats *stats,
+ uint16_t *qlen);
+
+/**
+ * Scheduler hierarchy path write to packet descriptor. Typically
+ * called by the packet classification stage.
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param pkt
+ * Packet descriptor handle
+ * @param subport
+ * Subport ID
+ * @param pipe
+ * Pipe ID within subport
+ * @param traffic_class
+ * Traffic class ID within pipe (0 .. RTE_SCHED_TRAFFIC_CLASS_BE)
+ * @param queue
+ * Queue ID within pipe traffic class, 0 for high priority TCs, and
+ * 0 .. (RTE_SCHED_BE_QUEUES_PER_PIPE - 1) for best-effort TC
+ * @param color
+ * Packet color set
+ */
+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);
+
+/**
+ * Scheduler hierarchy path read from packet descriptor (struct
+ * rte_mbuf). Typically called as part of the hierarchical scheduler
+ * enqueue operation. The subport, pipe, traffic class and queue
+ * parameters need to be pre-allocated by the caller.
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param pkt
+ * Packet descriptor handle
+ * @param subport
+ * Subport ID
+ * @param pipe
+ * Pipe ID within subport
+ * @param traffic_class
+ * Traffic class ID within pipe (0 .. RTE_SCHED_TRAFFIC_CLASS_BE)
+ * @param queue
+ * Queue ID within pipe traffic class, 0 for high priority TCs, and
+ * 0 .. (RTE_SCHED_BE_QUEUES_PER_PIPE - 1) for best-effort TC
+ */
+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);
+
+enum rte_color
+rte_sched_port_pkt_read_color(const struct rte_mbuf *pkt);
+
+/**
+ * Hierarchical scheduler port enqueue. Writes up to n_pkts to port
+ * scheduler and returns the number of packets actually written. For
+ * each packet, the port scheduler queue to write the packet to is
+ * identified by reading the hierarchy path from the packet
+ * descriptor; if the queue is full or congested and the packet is not
+ * written to the queue, then the packet is automatically dropped
+ * without any action required from the caller.
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param pkts
+ * Array storing the packet descriptor handles
+ * @param n_pkts
+ * Number of packets to enqueue from the pkts array into the port scheduler
+ * @return
+ * Number of packets successfully enqueued
+ */
+int
+rte_sched_port_enqueue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts);
+
+/**
+ * Hierarchical scheduler port dequeue. Reads up to n_pkts from the
+ * port scheduler and stores them in the pkts array and returns the
+ * number of packets actually read. The pkts array needs to be
+ * pre-allocated by the caller with at least n_pkts entries.
+ *
+ * @param port
+ * Handle to port scheduler instance
+ * @param pkts
+ * Pre-allocated packet descriptor array where the packets dequeued
+ * from the port
+ * scheduler should be stored
+ * @param n_pkts
+ * Number of packets to dequeue from the port scheduler
+ * @return
+ * Number of packets successfully dequeued and placed in the pkts array
+ */
+int
+rte_sched_port_dequeue(struct rte_sched_port *port, struct rte_mbuf **pkts, uint32_t n_pkts);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __INCLUDE_RTE_SCHED_H__ */
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_sched_common.h b/src/spdk/dpdk/lib/librte_sched/rte_sched_common.h
new file mode 100644
index 000000000..96706df7b
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_sched_common.h
@@ -0,0 +1,95 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
+ */
+
+#ifndef __INCLUDE_RTE_SCHED_COMMON_H__
+#define __INCLUDE_RTE_SCHED_COMMON_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdint.h>
+#include <sys/types.h>
+
+#define __rte_aligned_16 __rte_aligned(16)
+
+#if 0
+static inline uint32_t
+rte_min_pos_4_u16(uint16_t *x)
+{
+ uint32_t pos0, pos1;
+
+ pos0 = (x[0] <= x[1])? 0 : 1;
+ pos1 = (x[2] <= x[3])? 2 : 3;
+
+ return (x[pos0] <= x[pos1])? pos0 : pos1;
+}
+
+#else
+
+/* simplified version to remove branches with CMOV instruction */
+static inline uint32_t
+rte_min_pos_4_u16(uint16_t *x)
+{
+ uint32_t pos0 = 0;
+ uint32_t pos1 = 2;
+
+ if (x[1] <= x[0]) pos0 = 1;
+ if (x[3] <= x[2]) pos1 = 3;
+ if (x[pos1] <= x[pos0]) pos0 = pos1;
+
+ return pos0;
+}
+
+#endif
+
+/*
+ * Compute the Greatest Common Divisor (GCD) of two numbers.
+ * This implementation uses Euclid's algorithm:
+ * gcd(a, 0) = a
+ * gcd(a, b) = gcd(b, a mod b)
+ *
+ */
+static inline uint32_t
+rte_get_gcd(uint32_t a, uint32_t b)
+{
+ uint32_t c;
+
+ if (a == 0)
+ return b;
+ if (b == 0)
+ return a;
+
+ if (a < b) {
+ c = a;
+ a = b;
+ b = c;
+ }
+
+ while (b != 0) {
+ c = a % b;
+ a = b;
+ b = c;
+ }
+
+ return a;
+}
+
+/*
+ * Compute the Lowest Common Denominator (LCD) of two numbers.
+ * This implementation computes GCD first:
+ * LCD(a, b) = (a * b) / GCD(a, b)
+ *
+ */
+static inline uint32_t
+rte_get_lcd(uint32_t a, uint32_t b)
+{
+ return (a * b) / rte_get_gcd(a, b);
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __INCLUDE_RTE_SCHED_COMMON_H__ */
diff --git a/src/spdk/dpdk/lib/librte_sched/rte_sched_version.map b/src/spdk/dpdk/lib/librte_sched/rte_sched_version.map
new file mode 100644
index 000000000..cefd99036
--- /dev/null
+++ b/src/spdk/dpdk/lib/librte_sched/rte_sched_version.map
@@ -0,0 +1,31 @@
+DPDK_20.0 {
+ global:
+
+ rte_approx;
+ rte_red_config_init;
+ rte_red_log2_1_minus_Wq;
+ rte_red_pow2_frac_inv;
+ rte_red_rand_seed;
+ rte_red_rand_val;
+ rte_red_rt_data_init;
+ rte_sched_pipe_config;
+ rte_sched_port_config;
+ rte_sched_port_dequeue;
+ rte_sched_port_enqueue;
+ rte_sched_port_free;
+ rte_sched_port_get_memory_footprint;
+ rte_sched_port_pkt_read_color;
+ rte_sched_port_pkt_read_tree_path;
+ rte_sched_port_pkt_write;
+ rte_sched_queue_read_stats;
+ rte_sched_subport_config;
+ rte_sched_subport_read_stats;
+
+ local: *;
+};
+
+EXPERIMENTAL {
+ global:
+
+ rte_sched_subport_pipe_profile_add;
+};
generated by cgit v1.2.3 (git 2.39.1) at 2025-01-13 03:00:12 +0000