xref: /dpdk/drivers/crypto/scheduler/scheduler_multicore.c (revision b3bbd9e5f265937164ed8c9c61d12f7543a745ea)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2017 Intel Corporation. All rights reserved.
5  *
6  *   Redistribution and use in source and binary forms, with or without
7  *   modification, are permitted provided that the following conditions
8  *   are met:
9  *
10  *     * Redistributions of source code must retain the above copyright
11  *       notice, this list of conditions and the following disclaimer.
12  *     * Redistributions in binary form must reproduce the above copyright
13  *       notice, this list of conditions and the following disclaimer in
14  *       the documentation and/or other materials provided with the
15  *       distribution.
16  *     * Neither the name of Intel Corporation nor the names of its
17  *       contributors may be used to endorse or promote products derived
18  *       from this software without specific prior written permission.
19  *
20  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
23  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
24  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
25  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
26  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
30  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31  */
32 #include <unistd.h>
33 
34 #include <rte_cryptodev.h>
35 #include <rte_malloc.h>
36 
37 #include "rte_cryptodev_scheduler_operations.h"
38 #include "scheduler_pmd_private.h"
39 
40 #define MC_SCHED_ENQ_RING_NAME_PREFIX	"MCS_ENQR_"
41 #define MC_SCHED_DEQ_RING_NAME_PREFIX	"MCS_DEQR_"
42 
43 #define MC_SCHED_BUFFER_SIZE 32
44 
45 /** multi-core scheduler context */
46 struct mc_scheduler_ctx {
47 	uint32_t num_workers;             /**< Number of workers polling */
48 	uint32_t stop_signal;
49 
50 	struct rte_ring *sched_enq_ring[RTE_CRYPTODEV_SCHEDULER_MAX_NB_WORKER_CORES];
51 	struct rte_ring *sched_deq_ring[RTE_CRYPTODEV_SCHEDULER_MAX_NB_WORKER_CORES];
52 };
53 
54 struct mc_scheduler_qp_ctx {
55 	struct scheduler_slave slaves[RTE_CRYPTODEV_SCHEDULER_MAX_NB_SLAVES];
56 	uint32_t nb_slaves;
57 
58 	uint32_t last_enq_worker_idx;
59 	uint32_t last_deq_worker_idx;
60 
61 	struct mc_scheduler_ctx *mc_private_ctx;
62 };
63 
64 static uint16_t
65 schedule_enqueue(void *qp, struct rte_crypto_op **ops, uint16_t nb_ops)
66 {
67 	struct mc_scheduler_qp_ctx *mc_qp_ctx =
68 			((struct scheduler_qp_ctx *)qp)->private_qp_ctx;
69 	struct mc_scheduler_ctx *mc_ctx = mc_qp_ctx->mc_private_ctx;
70 	uint32_t worker_idx = mc_qp_ctx->last_enq_worker_idx;
71 	uint16_t i, processed_ops = 0;
72 
73 	if (unlikely(nb_ops == 0))
74 		return 0;
75 
76 	for (i = 0; i <  mc_ctx->num_workers && nb_ops != 0; i++) {
77 		struct rte_ring *enq_ring = mc_ctx->sched_enq_ring[worker_idx];
78 		uint16_t nb_queue_ops = rte_ring_enqueue_burst(enq_ring,
79 			(void *)(&ops[processed_ops]), nb_ops, NULL);
80 
81 		nb_ops -= nb_queue_ops;
82 		processed_ops += nb_queue_ops;
83 
84 		if (++worker_idx == mc_ctx->num_workers)
85 			worker_idx = 0;
86 	}
87 	mc_qp_ctx->last_enq_worker_idx = worker_idx;
88 
89 	return processed_ops;
90 }
91 
92 static uint16_t
93 schedule_enqueue_ordering(void *qp, struct rte_crypto_op **ops,
94 		uint16_t nb_ops)
95 {
96 	struct rte_ring *order_ring =
97 			((struct scheduler_qp_ctx *)qp)->order_ring;
98 	uint16_t nb_ops_to_enq = get_max_enqueue_order_count(order_ring,
99 			nb_ops);
100 	uint16_t nb_ops_enqd = schedule_enqueue(qp, ops,
101 			nb_ops_to_enq);
102 
103 	scheduler_order_insert(order_ring, ops, nb_ops_enqd);
104 
105 	return nb_ops_enqd;
106 }
107 
108 
109 static uint16_t
110 schedule_dequeue(void *qp, struct rte_crypto_op **ops, uint16_t nb_ops)
111 {
112 	struct mc_scheduler_qp_ctx *mc_qp_ctx =
113 			((struct scheduler_qp_ctx *)qp)->private_qp_ctx;
114 	struct mc_scheduler_ctx *mc_ctx = mc_qp_ctx->mc_private_ctx;
115 	uint32_t worker_idx = mc_qp_ctx->last_deq_worker_idx;
116 	uint16_t i, processed_ops = 0;
117 
118 	for (i = 0; i < mc_ctx->num_workers && nb_ops != 0; i++) {
119 		struct rte_ring *deq_ring = mc_ctx->sched_deq_ring[worker_idx];
120 		uint16_t nb_deq_ops = rte_ring_dequeue_burst(deq_ring,
121 			(void *)(&ops[processed_ops]), nb_ops, NULL);
122 
123 		nb_ops -= nb_deq_ops;
124 		processed_ops += nb_deq_ops;
125 		if (++worker_idx == mc_ctx->num_workers)
126 			worker_idx = 0;
127 	}
128 
129 	mc_qp_ctx->last_deq_worker_idx = worker_idx;
130 
131 	return processed_ops;
132 
133 }
134 
135 static uint16_t
136 schedule_dequeue_ordering(void *qp, struct rte_crypto_op **ops,
137 		uint16_t nb_ops)
138 {
139 	struct rte_ring *order_ring =
140 			((struct scheduler_qp_ctx *)qp)->order_ring;
141 
142 	return scheduler_order_drain(order_ring, ops, nb_ops);
143 }
144 
145 static int
146 slave_attach(__rte_unused struct rte_cryptodev *dev,
147 		__rte_unused uint8_t slave_id)
148 {
149 	return 0;
150 }
151 
152 static int
153 slave_detach(__rte_unused struct rte_cryptodev *dev,
154 		__rte_unused uint8_t slave_id)
155 {
156 	return 0;
157 }
158 
159 static int
160 mc_scheduler_worker(struct rte_cryptodev *dev)
161 {
162 	struct scheduler_ctx *sched_ctx = dev->data->dev_private;
163 	struct mc_scheduler_ctx *mc_ctx = sched_ctx->private_ctx;
164 	struct rte_ring *enq_ring;
165 	struct rte_ring *deq_ring;
166 	uint32_t core_id = rte_lcore_id();
167 	int i, worker_idx = -1;
168 	struct scheduler_slave *slave;
169 	struct rte_crypto_op *enq_ops[MC_SCHED_BUFFER_SIZE];
170 	struct rte_crypto_op *deq_ops[MC_SCHED_BUFFER_SIZE];
171 	uint16_t processed_ops;
172 	uint16_t left_op = 0;
173 	uint16_t left_op_idx = 0;
174 	uint16_t inflight_ops = 0;
175 
176 	for (i = 0; i < (int)sched_ctx->nb_wc; i++) {
177 		if (sched_ctx->wc_pool[i] == core_id) {
178 			worker_idx = i;
179 			break;
180 		}
181 	}
182 	if (worker_idx == -1) {
183 		CS_LOG_ERR("worker on core %u:cannot find worker index!\n", core_id);
184 		return -1;
185 	}
186 
187 	slave = &sched_ctx->slaves[worker_idx];
188 	enq_ring = mc_ctx->sched_enq_ring[worker_idx];
189 	deq_ring = mc_ctx->sched_deq_ring[worker_idx];
190 
191 	while (!mc_ctx->stop_signal) {
192 		if (left_op) {
193 			processed_ops =
194 				rte_cryptodev_enqueue_burst(slave->dev_id,
195 						slave->qp_id,
196 						&enq_ops[left_op_idx], left_op);
197 
198 			left_op -= processed_ops;
199 			left_op_idx += processed_ops;
200 		} else {
201 			uint16_t nb_deq_ops = rte_ring_dequeue_burst(enq_ring,
202 				(void *)enq_ops, MC_SCHED_BUFFER_SIZE, NULL);
203 			if (nb_deq_ops) {
204 				processed_ops = rte_cryptodev_enqueue_burst(slave->dev_id,
205 						slave->qp_id, enq_ops, nb_deq_ops);
206 
207 				if (unlikely(processed_ops < nb_deq_ops)) {
208 					left_op = nb_deq_ops - processed_ops;
209 					left_op_idx = processed_ops;
210 				}
211 
212 				inflight_ops += processed_ops;
213 			}
214 		}
215 
216 		if (inflight_ops > 0) {
217 			processed_ops = rte_cryptodev_dequeue_burst(slave->dev_id,
218 					slave->qp_id, deq_ops, MC_SCHED_BUFFER_SIZE);
219 			if (processed_ops) {
220 				uint16_t nb_enq_ops = rte_ring_enqueue_burst(deq_ring,
221 					(void *)deq_ops, processed_ops, NULL);
222 				inflight_ops -= nb_enq_ops;
223 			}
224 		}
225 
226 		rte_pause();
227 	}
228 
229 	return 0;
230 }
231 
232 static int
233 scheduler_start(struct rte_cryptodev *dev)
234 {
235 	struct scheduler_ctx *sched_ctx = dev->data->dev_private;
236 	struct mc_scheduler_ctx *mc_ctx = sched_ctx->private_ctx;
237 	uint16_t i;
238 
239 	mc_ctx->stop_signal = 0;
240 
241 	for (i = 0; i < sched_ctx->nb_wc; i++)
242 		rte_eal_remote_launch(
243 			(lcore_function_t *)mc_scheduler_worker, dev,
244 					sched_ctx->wc_pool[i]);
245 
246 	if (sched_ctx->reordering_enabled) {
247 		dev->enqueue_burst = &schedule_enqueue_ordering;
248 		dev->dequeue_burst = &schedule_dequeue_ordering;
249 	} else {
250 		dev->enqueue_burst = &schedule_enqueue;
251 		dev->dequeue_burst = &schedule_dequeue;
252 	}
253 
254 	for (i = 0; i < dev->data->nb_queue_pairs; i++) {
255 		struct scheduler_qp_ctx *qp_ctx = dev->data->queue_pairs[i];
256 		struct mc_scheduler_qp_ctx *mc_qp_ctx =
257 				qp_ctx->private_qp_ctx;
258 		uint32_t j;
259 
260 		memset(mc_qp_ctx->slaves, 0,
261 				RTE_CRYPTODEV_SCHEDULER_MAX_NB_SLAVES *
262 				sizeof(struct scheduler_slave));
263 		for (j = 0; j < sched_ctx->nb_slaves; j++) {
264 			mc_qp_ctx->slaves[j].dev_id =
265 					sched_ctx->slaves[j].dev_id;
266 			mc_qp_ctx->slaves[j].qp_id = i;
267 		}
268 
269 		mc_qp_ctx->nb_slaves = sched_ctx->nb_slaves;
270 
271 		mc_qp_ctx->last_enq_worker_idx = 0;
272 		mc_qp_ctx->last_deq_worker_idx = 0;
273 	}
274 
275 	return 0;
276 }
277 
278 static int
279 scheduler_stop(struct rte_cryptodev *dev)
280 {
281 	struct scheduler_ctx *sched_ctx = dev->data->dev_private;
282 	struct mc_scheduler_ctx *mc_ctx = sched_ctx->private_ctx;
283 
284 	mc_ctx->stop_signal = 1;
285 
286 	for (uint16_t i = 0; i < sched_ctx->nb_wc; i++)
287 		rte_eal_wait_lcore(sched_ctx->wc_pool[i]);
288 
289 	return 0;
290 }
291 
292 static int
293 scheduler_config_qp(struct rte_cryptodev *dev, uint16_t qp_id)
294 {
295 	struct scheduler_qp_ctx *qp_ctx = dev->data->queue_pairs[qp_id];
296 	struct mc_scheduler_qp_ctx *mc_qp_ctx;
297 	struct scheduler_ctx *sched_ctx = dev->data->dev_private;
298 	struct mc_scheduler_ctx *mc_ctx = sched_ctx->private_ctx;
299 
300 	mc_qp_ctx = rte_zmalloc_socket(NULL, sizeof(*mc_qp_ctx), 0,
301 			rte_socket_id());
302 	if (!mc_qp_ctx) {
303 		CS_LOG_ERR("failed allocate memory for private queue pair");
304 		return -ENOMEM;
305 	}
306 
307 	mc_qp_ctx->mc_private_ctx = mc_ctx;
308 	qp_ctx->private_qp_ctx = (void *)mc_qp_ctx;
309 
310 
311 	return 0;
312 }
313 
314 static int
315 scheduler_create_private_ctx(struct rte_cryptodev *dev)
316 {
317 	struct scheduler_ctx *sched_ctx = dev->data->dev_private;
318 	struct mc_scheduler_ctx *mc_ctx;
319 
320 	if (sched_ctx->private_ctx)
321 		rte_free(sched_ctx->private_ctx);
322 
323 	mc_ctx = rte_zmalloc_socket(NULL, sizeof(struct mc_scheduler_ctx), 0,
324 			rte_socket_id());
325 	if (!mc_ctx) {
326 		CS_LOG_ERR("failed allocate memory");
327 		return -ENOMEM;
328 	}
329 
330 	mc_ctx->num_workers = sched_ctx->nb_wc;
331 	for (uint16_t i = 0; i < sched_ctx->nb_wc; i++) {
332 		char r_name[16];
333 
334 		snprintf(r_name, sizeof(r_name), MC_SCHED_ENQ_RING_NAME_PREFIX "%u", i);
335 		mc_ctx->sched_enq_ring[i] = rte_ring_create(r_name, PER_SLAVE_BUFF_SIZE,
336 					rte_socket_id(), RING_F_SC_DEQ | RING_F_SP_ENQ);
337 		if (!mc_ctx->sched_enq_ring[i]) {
338 			CS_LOG_ERR("Cannot create ring for worker %u", i);
339 			return -1;
340 		}
341 		snprintf(r_name, sizeof(r_name), MC_SCHED_DEQ_RING_NAME_PREFIX "%u", i);
342 		mc_ctx->sched_deq_ring[i] = rte_ring_create(r_name, PER_SLAVE_BUFF_SIZE,
343 					rte_socket_id(), RING_F_SC_DEQ | RING_F_SP_ENQ);
344 		if (!mc_ctx->sched_deq_ring[i]) {
345 			CS_LOG_ERR("Cannot create ring for worker %u", i);
346 			return -1;
347 		}
348 	}
349 
350 	sched_ctx->private_ctx = (void *)mc_ctx;
351 
352 	return 0;
353 }
354 
355 struct rte_cryptodev_scheduler_ops scheduler_mc_ops = {
356 	slave_attach,
357 	slave_detach,
358 	scheduler_start,
359 	scheduler_stop,
360 	scheduler_config_qp,
361 	scheduler_create_private_ctx,
362 	NULL,	/* option_set */
363 	NULL	/* option_get */
364 };
365 
366 struct rte_cryptodev_scheduler mc_scheduler = {
367 		.name = "multicore-scheduler",
368 		.description = "scheduler which will run burst across multiple cpu cores",
369 		.mode = CDEV_SCHED_MODE_MULTICORE,
370 		.ops = &scheduler_mc_ops
371 };
372 
373 struct rte_cryptodev_scheduler *multicore_scheduler = &mc_scheduler;
374