xref: /dpdk/drivers/net/dpaa2/dpaa2_rxtx.c (revision 4e30ead5e7ca886535e2b30632b2948d2aac1681)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright (c) 2016 Freescale Semiconductor, Inc. All rights reserved.
5  *   Copyright (c) 2016 NXP. All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Freescale Semiconductor, Inc nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <time.h>
35 #include <net/if.h>
36 
37 #include <rte_mbuf.h>
38 #include <rte_ethdev.h>
39 #include <rte_malloc.h>
40 #include <rte_memcpy.h>
41 #include <rte_string_fns.h>
42 #include <rte_dev.h>
43 #include <rte_ethdev.h>
44 
45 #include <fslmc_logs.h>
46 #include <fslmc_vfio.h>
47 #include <dpaa2_hw_pvt.h>
48 #include <dpaa2_hw_dpio.h>
49 #include <dpaa2_hw_mempool.h>
50 
51 #include "dpaa2_ethdev.h"
52 #include "base/dpaa2_hw_dpni_annot.h"
53 
54 static inline uint32_t __attribute__((hot))
55 dpaa2_dev_rx_parse(uint64_t hw_annot_addr)
56 {
57 	uint32_t pkt_type = RTE_PTYPE_UNKNOWN;
58 	struct dpaa2_annot_hdr *annotation =
59 			(struct dpaa2_annot_hdr *)hw_annot_addr;
60 
61 	PMD_RX_LOG(DEBUG, "annotation = 0x%lx   ", annotation->word4);
62 
63 	if (BIT_ISSET_AT_POS(annotation->word3, L2_ARP_PRESENT)) {
64 		pkt_type = RTE_PTYPE_L2_ETHER_ARP;
65 		goto parse_done;
66 	} else if (BIT_ISSET_AT_POS(annotation->word3, L2_ETH_MAC_PRESENT)) {
67 		pkt_type = RTE_PTYPE_L2_ETHER;
68 	} else {
69 		goto parse_done;
70 	}
71 
72 	if (BIT_ISSET_AT_POS(annotation->word4, L3_IPV4_1_PRESENT |
73 			     L3_IPV4_N_PRESENT)) {
74 		pkt_type |= RTE_PTYPE_L3_IPV4;
75 		if (BIT_ISSET_AT_POS(annotation->word4, L3_IP_1_OPT_PRESENT |
76 			L3_IP_N_OPT_PRESENT))
77 			pkt_type |= RTE_PTYPE_L3_IPV4_EXT;
78 
79 	} else if (BIT_ISSET_AT_POS(annotation->word4, L3_IPV6_1_PRESENT |
80 		  L3_IPV6_N_PRESENT)) {
81 		pkt_type |= RTE_PTYPE_L3_IPV6;
82 		if (BIT_ISSET_AT_POS(annotation->word4, L3_IP_1_OPT_PRESENT |
83 		    L3_IP_N_OPT_PRESENT))
84 			pkt_type |= RTE_PTYPE_L3_IPV6_EXT;
85 	} else {
86 		goto parse_done;
87 	}
88 
89 	if (BIT_ISSET_AT_POS(annotation->word4, L3_IP_1_FIRST_FRAGMENT |
90 	    L3_IP_1_MORE_FRAGMENT |
91 	    L3_IP_N_FIRST_FRAGMENT |
92 	    L3_IP_N_MORE_FRAGMENT)) {
93 		pkt_type |= RTE_PTYPE_L4_FRAG;
94 		goto parse_done;
95 	} else {
96 		pkt_type |= RTE_PTYPE_L4_NONFRAG;
97 	}
98 
99 	if (BIT_ISSET_AT_POS(annotation->word4, L3_PROTO_UDP_PRESENT))
100 		pkt_type |= RTE_PTYPE_L4_UDP;
101 
102 	else if (BIT_ISSET_AT_POS(annotation->word4, L3_PROTO_TCP_PRESENT))
103 		pkt_type |= RTE_PTYPE_L4_TCP;
104 
105 	else if (BIT_ISSET_AT_POS(annotation->word4, L3_PROTO_SCTP_PRESENT))
106 		pkt_type |= RTE_PTYPE_L4_SCTP;
107 
108 	else if (BIT_ISSET_AT_POS(annotation->word4, L3_PROTO_ICMP_PRESENT))
109 		pkt_type |= RTE_PTYPE_L4_ICMP;
110 
111 	else if (BIT_ISSET_AT_POS(annotation->word4, L3_IP_UNKNOWN_PROTOCOL))
112 		pkt_type |= RTE_PTYPE_UNKNOWN;
113 
114 parse_done:
115 	return pkt_type;
116 }
117 
118 static inline void __attribute__((hot))
119 dpaa2_dev_rx_offload(uint64_t hw_annot_addr, struct rte_mbuf *mbuf)
120 {
121 	struct dpaa2_annot_hdr *annotation =
122 		(struct dpaa2_annot_hdr *)hw_annot_addr;
123 
124 	if (BIT_ISSET_AT_POS(annotation->word3,
125 			     L2_VLAN_1_PRESENT | L2_VLAN_N_PRESENT))
126 		mbuf->ol_flags |= PKT_RX_VLAN_PKT;
127 
128 	if (BIT_ISSET_AT_POS(annotation->word8, DPAA2_ETH_FAS_L3CE))
129 		mbuf->ol_flags |= PKT_RX_IP_CKSUM_BAD;
130 
131 	if (BIT_ISSET_AT_POS(annotation->word8, DPAA2_ETH_FAS_L4CE))
132 		mbuf->ol_flags |= PKT_RX_L4_CKSUM_BAD;
133 }
134 
135 static inline struct rte_mbuf *__attribute__((hot))
136 eth_fd_to_mbuf(const struct qbman_fd *fd)
137 {
138 	struct rte_mbuf *mbuf = DPAA2_INLINE_MBUF_FROM_BUF(
139 		DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd)),
140 		     rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size);
141 
142 	/* need to repopulated some of the fields,
143 	 * as they may have changed in last transmission
144 	 */
145 	mbuf->nb_segs = 1;
146 	mbuf->ol_flags = 0;
147 	mbuf->data_off = DPAA2_GET_FD_OFFSET(fd);
148 	mbuf->data_len = DPAA2_GET_FD_LEN(fd);
149 	mbuf->pkt_len = mbuf->data_len;
150 
151 	/* Parse the packet */
152 	/* parse results are after the private - sw annotation area */
153 	mbuf->packet_type = dpaa2_dev_rx_parse(
154 			(uint64_t)DPAA2_IOVA_TO_VADDR(DPAA2_GET_FD_ADDR(fd))
155 			 + DPAA2_FD_PTA_SIZE);
156 
157 	dpaa2_dev_rx_offload((uint64_t)DPAA2_IOVA_TO_VADDR(
158 			     DPAA2_GET_FD_ADDR(fd)) +
159 			     DPAA2_FD_PTA_SIZE, mbuf);
160 
161 	mbuf->next = NULL;
162 	rte_mbuf_refcnt_set(mbuf, 1);
163 
164 	PMD_RX_LOG(DEBUG, "to mbuf - mbuf =%p, mbuf->buf_addr =%p, off = %d,"
165 		"fd_off=%d fd =%lx, meta = %d  bpid =%d, len=%d\n",
166 		mbuf, mbuf->buf_addr, mbuf->data_off,
167 		DPAA2_GET_FD_OFFSET(fd), DPAA2_GET_FD_ADDR(fd),
168 		rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size,
169 		DPAA2_GET_FD_BPID(fd), DPAA2_GET_FD_LEN(fd));
170 
171 	return mbuf;
172 }
173 
174 static void __attribute__ ((noinline)) __attribute__((hot))
175 eth_mbuf_to_fd(struct rte_mbuf *mbuf,
176 	       struct qbman_fd *fd, uint16_t bpid)
177 {
178 	/*Resetting the buffer pool id and offset field*/
179 	fd->simple.bpid_offset = 0;
180 
181 	DPAA2_SET_FD_ADDR(fd, DPAA2_MBUF_VADDR_TO_IOVA(mbuf));
182 	DPAA2_SET_FD_LEN(fd, mbuf->data_len);
183 	DPAA2_SET_FD_BPID(fd, bpid);
184 	DPAA2_SET_FD_OFFSET(fd, mbuf->data_off);
185 	DPAA2_SET_FD_ASAL(fd, DPAA2_ASAL_VAL);
186 
187 	PMD_TX_LOG(DEBUG, "mbuf =%p, mbuf->buf_addr =%p, off = %d,"
188 		"fd_off=%d fd =%lx, meta = %d  bpid =%d, len=%d\n",
189 		mbuf, mbuf->buf_addr, mbuf->data_off,
190 		DPAA2_GET_FD_OFFSET(fd), DPAA2_GET_FD_ADDR(fd),
191 		rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size,
192 		DPAA2_GET_FD_BPID(fd), DPAA2_GET_FD_LEN(fd));
193 }
194 
195 
196 static inline int __attribute__((hot))
197 eth_copy_mbuf_to_fd(struct rte_mbuf *mbuf,
198 		    struct qbman_fd *fd, uint16_t bpid)
199 {
200 	struct rte_mbuf *m;
201 	void *mb = NULL;
202 
203 	if (rte_dpaa2_mbuf_alloc_bulk(
204 		rte_dpaa2_bpid_info[bpid].bp_list->mp, &mb, 1)) {
205 		PMD_TX_LOG(WARNING, "Unable to allocated DPAA2 buffer");
206 		rte_pktmbuf_free(mbuf);
207 		return -1;
208 	}
209 	m = (struct rte_mbuf *)mb;
210 	memcpy((char *)m->buf_addr + mbuf->data_off,
211 	       (void *)((char *)mbuf->buf_addr + mbuf->data_off),
212 		mbuf->pkt_len);
213 
214 	/* Copy required fields */
215 	m->data_off = mbuf->data_off;
216 	m->ol_flags = mbuf->ol_flags;
217 	m->packet_type = mbuf->packet_type;
218 	m->tx_offload = mbuf->tx_offload;
219 
220 	/*Resetting the buffer pool id and offset field*/
221 	fd->simple.bpid_offset = 0;
222 
223 	DPAA2_SET_FD_ADDR(fd, DPAA2_MBUF_VADDR_TO_IOVA(m));
224 	DPAA2_SET_FD_LEN(fd, mbuf->data_len);
225 	DPAA2_SET_FD_BPID(fd, bpid);
226 	DPAA2_SET_FD_OFFSET(fd, mbuf->data_off);
227 	DPAA2_SET_FD_ASAL(fd, DPAA2_ASAL_VAL);
228 
229 	PMD_TX_LOG(DEBUG, " mbuf %p BMAN buf addr %p",
230 		   (void *)mbuf, mbuf->buf_addr);
231 
232 	PMD_TX_LOG(DEBUG, " fdaddr =%lx bpid =%d meta =%d off =%d, len =%d",
233 		   DPAA2_GET_FD_ADDR(fd),
234 		DPAA2_GET_FD_BPID(fd),
235 		rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size,
236 		DPAA2_GET_FD_OFFSET(fd),
237 		DPAA2_GET_FD_LEN(fd));
238 	/*free the original packet */
239 	rte_pktmbuf_free(mbuf);
240 
241 	return 0;
242 }
243 
244 uint16_t
245 dpaa2_dev_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
246 {
247 	/* Function is responsible to receive frames for a given device and VQ*/
248 	struct dpaa2_queue *dpaa2_q = (struct dpaa2_queue *)queue;
249 	struct qbman_result *dq_storage;
250 	uint32_t fqid = dpaa2_q->fqid;
251 	int ret, num_rx = 0;
252 	uint8_t is_last = 0, status;
253 	struct qbman_swp *swp;
254 	const struct qbman_fd *fd;
255 	struct qbman_pull_desc pulldesc;
256 	struct rte_eth_dev *dev = dpaa2_q->dev;
257 
258 	if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
259 		ret = dpaa2_affine_qbman_swp();
260 		if (ret) {
261 			RTE_LOG(ERR, PMD, "Failure in affining portal\n");
262 			return 0;
263 		}
264 	}
265 	swp = DPAA2_PER_LCORE_PORTAL;
266 	dq_storage = dpaa2_q->q_storage->dq_storage[0];
267 
268 	qbman_pull_desc_clear(&pulldesc);
269 	qbman_pull_desc_set_numframes(&pulldesc,
270 				      (nb_pkts > DPAA2_DQRR_RING_SIZE) ?
271 				       DPAA2_DQRR_RING_SIZE : nb_pkts);
272 	qbman_pull_desc_set_fq(&pulldesc, fqid);
273 	/* todo optimization - we can have dq_storage_phys available*/
274 	qbman_pull_desc_set_storage(&pulldesc, dq_storage,
275 			(dma_addr_t)(DPAA2_VADDR_TO_IOVA(dq_storage)), 1);
276 
277 	/*Issue a volatile dequeue command. */
278 	while (1) {
279 		if (qbman_swp_pull(swp, &pulldesc)) {
280 			PMD_RX_LOG(ERR, "VDQ command is not issued."
281 				   "QBMAN is busy\n");
282 			/* Portal was busy, try again */
283 			continue;
284 		}
285 		break;
286 	};
287 
288 	/* Receive the packets till Last Dequeue entry is found with
289 	 * respect to the above issues PULL command.
290 	 */
291 	while (!is_last) {
292 		struct rte_mbuf *mbuf;
293 		/*Check if the previous issued command is completed.
294 		 * Also seems like the SWP is shared between the
295 		 * Ethernet Driver and the SEC driver.
296 		 */
297 		while (!qbman_check_command_complete(swp, dq_storage))
298 			;
299 		/* Loop until the dq_storage is updated with
300 		 * new token by QBMAN
301 		 */
302 		while (!qbman_result_has_new_result(swp, dq_storage))
303 			;
304 		/* Check whether Last Pull command is Expired and
305 		 * setting Condition for Loop termination
306 		 */
307 		if (qbman_result_DQ_is_pull_complete(dq_storage)) {
308 			is_last = 1;
309 			/* Check for valid frame. */
310 			status = (uint8_t)qbman_result_DQ_flags(dq_storage);
311 			if (unlikely((status & QBMAN_DQ_STAT_VALIDFRAME) == 0))
312 				continue;
313 		}
314 
315 		fd = qbman_result_DQ_fd(dq_storage);
316 		mbuf = (struct rte_mbuf *)DPAA2_IOVA_TO_VADDR(
317 		   DPAA2_GET_FD_ADDR(fd)
318 		   - rte_dpaa2_bpid_info[DPAA2_GET_FD_BPID(fd)].meta_data_size);
319 		/* Prefeth mbuf */
320 		rte_prefetch0(mbuf);
321 		/* Prefetch Annotation address for the parse results */
322 		rte_prefetch0((void *)((uint64_t)DPAA2_GET_FD_ADDR(fd)
323 						+ DPAA2_FD_PTA_SIZE + 16));
324 
325 		bufs[num_rx] = eth_fd_to_mbuf(fd);
326 		bufs[num_rx]->port = dev->data->port_id;
327 
328 		num_rx++;
329 		dq_storage++;
330 	} /* End of Packet Rx loop */
331 
332 	dpaa2_q->rx_pkts += num_rx;
333 
334 	/*Return the total number of packets received to DPAA2 app*/
335 	return num_rx;
336 }
337 
338 /*
339  * Callback to handle sending packets through WRIOP based interface
340  */
341 uint16_t
342 dpaa2_dev_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
343 {
344 	/* Function to transmit the frames to given device and VQ*/
345 	uint32_t loop;
346 	int32_t ret;
347 	struct qbman_fd fd_arr[MAX_TX_RING_SLOTS];
348 	uint32_t frames_to_send;
349 	struct rte_mempool *mp;
350 	struct qbman_eq_desc eqdesc;
351 	struct dpaa2_queue *dpaa2_q = (struct dpaa2_queue *)queue;
352 	struct qbman_swp *swp;
353 	uint16_t num_tx = 0;
354 	uint16_t bpid;
355 	struct rte_eth_dev *dev = dpaa2_q->dev;
356 	struct dpaa2_dev_priv *priv = dev->data->dev_private;
357 
358 	if (unlikely(!DPAA2_PER_LCORE_DPIO)) {
359 		ret = dpaa2_affine_qbman_swp();
360 		if (ret) {
361 			RTE_LOG(ERR, PMD, "Failure in affining portal\n");
362 			return 0;
363 		}
364 	}
365 	swp = DPAA2_PER_LCORE_PORTAL;
366 
367 	PMD_TX_LOG(DEBUG, "===> dev =%p, fqid =%d", dev, dpaa2_q->fqid);
368 
369 	/*Prepare enqueue descriptor*/
370 	qbman_eq_desc_clear(&eqdesc);
371 	qbman_eq_desc_set_no_orp(&eqdesc, DPAA2_EQ_RESP_ERR_FQ);
372 	qbman_eq_desc_set_response(&eqdesc, 0, 0);
373 	qbman_eq_desc_set_qd(&eqdesc, priv->qdid,
374 			     dpaa2_q->flow_id, dpaa2_q->tc_index);
375 
376 	/*Clear the unused FD fields before sending*/
377 	while (nb_pkts) {
378 		frames_to_send = (nb_pkts >> 3) ? MAX_TX_RING_SLOTS : nb_pkts;
379 
380 		for (loop = 0; loop < frames_to_send; loop++) {
381 			fd_arr[loop].simple.frc = 0;
382 			DPAA2_RESET_FD_CTRL((&fd_arr[loop]));
383 			DPAA2_SET_FD_FLC((&fd_arr[loop]), NULL);
384 			mp = (*bufs)->pool;
385 			/* Not a hw_pkt pool allocated frame */
386 			if (mp->ops_index != priv->bp_list->dpaa2_ops_index) {
387 				PMD_TX_LOG(ERR, "non hw offload bufffer ");
388 				/* alloc should be from the default buffer pool
389 				 * attached to this interface
390 				 */
391 				if (priv->bp_list) {
392 					bpid = priv->bp_list->buf_pool.bpid;
393 				} else {
394 					PMD_TX_LOG(ERR, "errr: why no bpool"
395 						   " attached");
396 					num_tx = 0;
397 					goto skip_tx;
398 				}
399 				if (eth_copy_mbuf_to_fd(*bufs,
400 							&fd_arr[loop], bpid)) {
401 					bufs++;
402 					continue;
403 				}
404 			} else {
405 				bpid = mempool_to_bpid(mp);
406 				eth_mbuf_to_fd(*bufs, &fd_arr[loop], bpid);
407 			}
408 			bufs++;
409 		}
410 		loop = 0;
411 		while (loop < frames_to_send) {
412 			loop += qbman_swp_send_multiple(swp, &eqdesc,
413 					&fd_arr[loop], frames_to_send - loop);
414 		}
415 
416 		num_tx += frames_to_send;
417 		dpaa2_q->tx_pkts += frames_to_send;
418 		nb_pkts -= frames_to_send;
419 	}
420 skip_tx:
421 	return num_tx;
422 }
423