event/dlb2: add token pop API

The PMD uses a public interface to allow applications to
control the token pop mode. Supported token pop modes are
as follows, and they impact core scheduling affinity for
ldb ports.

AUTO_POP: Pop the CQ tokens immediately after dequeueing.
DELAYED_POP: Pop CQ tokens after (dequeue_depth - 1) events
             are released. Supported on load-balanced ports
             only.
DEFERRED_POP: Pop the CQ tokens during next dequeue operation.

Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
Reviewed-by: Gage Eads <gage.eads@intel.com>

doc/api/doxy-api-index.md

@@ -52,7 +52,8 @@ The public API headers are grouped by topics:
   [dpaa2_mempool]      (@ref rte_dpaa2_mempool.h),
   [dpaa2_cmdif]        (@ref rte_pmd_dpaa2_cmdif.h),
   [dpaa2_qdma]         (@ref rte_pmd_dpaa2_qdma.h),
-  [crypto_scheduler]   (@ref rte_cryptodev_scheduler.h)
+  [crypto_scheduler]   (@ref rte_cryptodev_scheduler.h),
+  [dlb2]               (@ref rte_pmd_dlb2.h)
 
 - **memory**:
   [memseg]             (@ref rte_memory.h),

doc/api/doxy-api.conf.in

@@ -7,6 +7,7 @@ USE_MDFILE_AS_MAINPAGE  = @TOPDIR@/doc/api/doxy-api-index.md
 INPUT                   = @TOPDIR@/doc/api/doxy-api-index.md \
                           @TOPDIR@/drivers/bus/vdev \
                           @TOPDIR@/drivers/crypto/scheduler \
+                          @TOPDIR@/drivers/event/dlb2 \
                           @TOPDIR@/drivers/mempool/dpaa2 \
                           @TOPDIR@/drivers/net/ark \
                           @TOPDIR@/drivers/net/bnxt \

drivers/event/dlb2/dlb2.c

@@ -1197,7 +1197,7 @@ dlb2_hw_create_ldb_port(struct dlb2_eventdev *dlb2,
 	qm_port->qid_mappings = &dlb2->qm_ldb_to_ev_queue_id[0];
 
 	qm_port->dequeue_depth = dequeue_depth;
-
+	qm_port->token_pop_thresh = dequeue_depth;
 	qm_port->owed_tokens = 0;
 	qm_port->issued_releases = 0;
 
@@ -1365,6 +1365,8 @@ dlb2_hw_create_dir_port(struct dlb2_eventdev *dlb2,
 
 	qm_port->dequeue_depth = dequeue_depth;
 
+	/* Directed ports are auto-pop, by default. */
+	qm_port->token_pop_mode = AUTO_POP;
 	qm_port->owed_tokens = 0;
 	qm_port->issued_releases = 0;
 
@@ -2252,6 +2254,18 @@ dlb2_hw_do_enqueue(struct dlb2_port *qm_port,
 	dlb2_pp_write(qm_port->qe4, port_data);
 }
 
+static inline void
+dlb2_construct_token_pop_qe(struct dlb2_port *qm_port, int idx)
+{
+	struct dlb2_cq_pop_qe *qe = (void *)qm_port->qe4;
+	int num = qm_port->owed_tokens;
+
+	qe[idx].cmd_byte = DLB2_POP_CMD_BYTE;
+	qe[idx].tokens = num - 1;
+
+	qm_port->owed_tokens = 0;
+}
+
 static inline void
 dlb2_event_build_hcws(struct dlb2_port *qm_port,
 		      const struct rte_event ev[],
@@ -2600,6 +2614,14 @@ dlb2_event_enqueue_burst(void *event_port,
 		dlb2_event_build_hcws(qm_port, &events[i], j,
 				      sched_types, queue_ids);
 
+		if (qm_port->token_pop_mode == DELAYED_POP && j < 4 &&
+		    qm_port->issued_releases >= qm_port->token_pop_thresh - 1) {
+			dlb2_construct_token_pop_qe(qm_port, j);
+
+			/* Reset the releases counter for the next QE batch */
+			qm_port->issued_releases -= qm_port->token_pop_thresh;
+		}
+
 		dlb2_hw_do_enqueue(qm_port, i == 0, port_data);
 
 		cnt += j;
@@ -2608,6 +2630,11 @@ dlb2_event_enqueue_burst(void *event_port,
 			break;
 	}
 
+	if (qm_port->token_pop_mode == DELAYED_POP &&
+	    qm_port->issued_releases >= qm_port->token_pop_thresh - 1) {
+		dlb2_consume_qe_immediate(qm_port, qm_port->owed_tokens);
+		qm_port->issued_releases -= qm_port->token_pop_thresh;
+	}
 	return cnt;
 }
 
@@ -3085,11 +3112,25 @@ dlb2_event_release(struct dlb2_eventdev *dlb2,
 		if (j == 0)
 			break;
 
+		if (qm_port->token_pop_mode == DELAYED_POP && j < 4 &&
+		    qm_port->issued_releases >= qm_port->token_pop_thresh - 1) {
+			dlb2_construct_token_pop_qe(qm_port, j);
+
+			/* Reset the releases counter for the next QE batch */
+			qm_port->issued_releases -= qm_port->token_pop_thresh;
+		}
+
 		dlb2_hw_do_enqueue(qm_port, i == 0, port_data);
 
 		cnt += j;
 	}
 
+	if (qm_port->token_pop_mode == DELAYED_POP &&
+	    qm_port->issued_releases >= qm_port->token_pop_thresh - 1) {
+		dlb2_consume_qe_immediate(qm_port, qm_port->owed_tokens);
+		qm_port->issued_releases -= qm_port->token_pop_thresh;
+	}
+
 sw_credit_update:
 	/* each release returns one credit */
 	if (!ev_port->outstanding_releases) {
@@ -3173,7 +3214,7 @@ dlb2_hw_dequeue_sparse(struct dlb2_eventdev *dlb2,
 	qm_port->owed_tokens += num;
 
 	if (num) {
-
+		if (qm_port->token_pop_mode == AUTO_POP)
 			dlb2_consume_qe_immediate(qm_port, num);
 
 		ev_port->outstanding_releases += num;
@@ -3300,7 +3341,7 @@ dlb2_hw_dequeue(struct dlb2_eventdev *dlb2,
 	qm_port->owed_tokens += num;
 
 	if (num) {
-
+		if (qm_port->token_pop_mode == AUTO_POP)
 			dlb2_consume_qe_immediate(qm_port, num);
 
 		ev_port->outstanding_releases += num;
@@ -3316,6 +3357,7 @@ dlb2_event_dequeue_burst(void *event_port, struct rte_event *ev, uint16_t num,
 			 uint64_t wait)
 {
 	struct dlb2_eventdev_port *ev_port = event_port;
+	struct dlb2_port *qm_port = &ev_port->qm_port;
 	struct dlb2_eventdev *dlb2 = ev_port->dlb2;
 	uint16_t cnt;
 
@@ -3331,6 +3373,9 @@ dlb2_event_dequeue_burst(void *event_port, struct rte_event *ev, uint16_t num,
 		DLB2_INC_STAT(ev_port->stats.tx_implicit_rel, out_rels);
 	}
 
+	if (qm_port->token_pop_mode == DEFERRED_POP && qm_port->owed_tokens)
+		dlb2_consume_qe_immediate(qm_port, qm_port->owed_tokens);
+
 	cnt = dlb2_hw_dequeue(dlb2, ev_port, ev, num, wait);
 
 	DLB2_INC_STAT(ev_port->stats.traffic.total_polls, 1);
@@ -3350,6 +3395,7 @@ dlb2_event_dequeue_burst_sparse(void *event_port, struct rte_event *ev,
 				uint16_t num, uint64_t wait)
 {
 	struct dlb2_eventdev_port *ev_port = event_port;
+	struct dlb2_port *qm_port = &ev_port->qm_port;
 	struct dlb2_eventdev *dlb2 = ev_port->dlb2;
 	uint16_t cnt;
 
@@ -3365,6 +3411,9 @@ dlb2_event_dequeue_burst_sparse(void *event_port, struct rte_event *ev,
 		DLB2_INC_STAT(ev_port->stats.tx_implicit_rel, out_rels);
 	}
 
+	if (qm_port->token_pop_mode == DEFERRED_POP && qm_port->owed_tokens)
+		dlb2_consume_qe_immediate(qm_port, qm_port->owed_tokens);
+
 	cnt = dlb2_hw_dequeue_sparse(dlb2, ev_port, ev, num, wait);
 
 	DLB2_INC_STAT(ev_port->stats.traffic.total_polls, 1);
@@ -3669,7 +3718,7 @@ dlb2_primary_eventdev_probe(struct rte_eventdev *dev,
 			    struct dlb2_devargs *dlb2_args)
 {
 	struct dlb2_eventdev *dlb2;
-	int err;
+	int err, i;
 
 	dlb2 = dev->data->dev_private;
 
@@ -3719,6 +3768,10 @@ dlb2_primary_eventdev_probe(struct rte_eventdev *dev,
 		return err;
 	}
 
+	/* Initialize each port's token pop mode */
+	for (i = 0; i < DLB2_MAX_NUM_PORTS; i++)
+		dlb2->ev_ports[i].qm_port.token_pop_mode = AUTO_POP;
+
 	rte_spinlock_init(&dlb2->qm_instance.resource_lock);
 
 	dlb2_iface_low_level_io_init();

drivers/event/dlb2/dlb2_priv.h

@@ -12,6 +12,7 @@
 #include <rte_config.h>
 #include "dlb2_user.h"
 #include "dlb2_log.h"
+#include "rte_pmd_dlb2.h"
 
 #ifndef RTE_LIBRTE_PMD_DLB2_QUELL_STATS
 #define DLB2_INC_STAT(_stat, _incr_val) ((_stat) += _incr_val)
@@ -290,6 +291,7 @@ struct dlb2_port {
 	bool gen_bit;
 	uint16_t dir_credits;
 	uint32_t dequeue_depth;
+	enum dlb2_token_pop_mode token_pop_mode;
 	union dlb2_port_config cfg;
 	uint32_t *credit_pool[DLB2_NUM_QUEUE_TYPES]; /* use __atomic builtins */
 	uint16_t cached_ldb_credits;
@@ -298,6 +300,7 @@ struct dlb2_port {
 	bool int_armed;
 	uint16_t owed_tokens;
 	int16_t issued_releases;
+	int16_t token_pop_thresh;
 	int cq_depth;
 	uint16_t cq_idx;
 	uint16_t cq_idx_unmasked;

drivers/event/dlb2/meson.build

@@ -1,3 +1,4 @@
+
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2019-2020 Intel Corporation
 
@@ -12,9 +13,10 @@ sources = files('dlb2.c',
 		'dlb2_xstats.c',
 		'pf/dlb2_main.c',
 		'pf/dlb2_pf.c',
-		'pf/base/dlb2_resource.c'
+		'pf/base/dlb2_resource.c',
+		'rte_pmd_dlb2.c'
 )
 
-headers = files()
+headers = files('rte_pmd_dlb2.h')
 
 deps += ['mbuf', 'mempool', 'ring', 'pci', 'bus_pci']

drivers/event/dlb2/rte_pmd_dlb2.c

@@ -0,0 +1,39 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Intel Corporation
+ */
+
+#include <rte_eventdev.h>
+#include <rte_eventdev_pmd.h>
+
+#include "rte_pmd_dlb2.h"
+#include "dlb2_priv.h"
+#include "dlb2_inline_fns.h"
+
+int
+rte_pmd_dlb2_set_token_pop_mode(uint8_t dev_id,
+				uint8_t port_id,
+				enum dlb2_token_pop_mode mode)
+{
+	struct dlb2_eventdev *dlb2;
+	struct rte_eventdev *dev;
+
+	RTE_EVENTDEV_VALID_DEVID_OR_ERR_RET(dev_id, -EINVAL);
+	dev = &rte_eventdevs[dev_id];
+
+	dlb2 = dlb2_pmd_priv(dev);
+
+	if (mode >= NUM_TOKEN_POP_MODES)
+		return -EINVAL;
+
+	/* The event device must be configured, but not yet started */
+	if (!dlb2->configured || dlb2->run_state != DLB2_RUN_STATE_STOPPED)
+		return -EINVAL;
+
+	/* The token pop mode must be set before configuring the port */
+	if (port_id >= dlb2->num_ports || dlb2->ev_ports[port_id].setup_done)
+		return -EINVAL;
+
+	dlb2->ev_ports[port_id].qm_port.token_pop_mode = mode;
+
+	return 0;
+}

drivers/event/dlb2/rte_pmd_dlb2.h

@@ -0,0 +1,72 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Intel Corporation
+ */
+
+/*!
+ *  @file      rte_pmd_dlb2.h
+ *
+ *  @brief     DLB PMD-specific functions
+ */
+
+#ifndef _RTE_PMD_DLB2_H_
+#define _RTE_PMD_DLB2_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdint.h>
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
+ *
+ * Selects the token pop mode for a DLB2 port.
+ */
+enum dlb2_token_pop_mode {
+	/* Pop the CQ tokens immediately after dequeueing. */
+	AUTO_POP,
+	/* Pop CQ tokens after (dequeue_depth - 1) events are released.
+	 * Supported on load-balanced ports only.
+	 */
+	DELAYED_POP,
+	/* Pop the CQ tokens during next dequeue operation. */
+	DEFERRED_POP,
+
+	/* NUM_TOKEN_POP_MODES must be last */
+	NUM_TOKEN_POP_MODES
+};
+
+/*!
+ * @warning
+ * @b EXPERIMENTAL: this API may change, or be removed, without prior notice
+ *
+ * Configure the token pop mode for a DLB2 port. By default, all ports use
+ * AUTO_POP. This function must be called before calling rte_event_port_setup()
+ * for the port, but after calling rte_event_dev_configure().
+ *
+ * @param dev_id
+ *    The identifier of the event device.
+ * @param port_id
+ *    The identifier of the event port.
+ * @param mode
+ *    The token pop mode.
+ *
+ * @return
+ * - 0: Success
+ * - EINVAL: Invalid dev_id, port_id, or mode
+ * - EINVAL: The DLB2 is not configured, is already running, or the port is
+ *   already setup
+ */
+
+__rte_experimental
+int
+rte_pmd_dlb2_set_token_pop_mode(uint8_t dev_id,
+				uint8_t port_id,
+				enum dlb2_token_pop_mode mode);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _RTE_PMD_DLB2_H_ */

drivers/event/dlb2/version.map

@@ -1,3 +1,9 @@
 DPDK_21 {
 	local: *;
 };
+
+EXPERIMENTAL {
+	global:
+
+	rte_pmd_dlb2_set_token_pop_mode;
+};