xref: /dpdk/drivers/net/hns3/hns3_rxtx.h (revision 68a03efeed657e6e05f281479b33b51102797e15)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2018-2019 Hisilicon Limited.
3  */
4 
5 #ifndef _HNS3_RXTX_H_
6 #define _HNS3_RXTX_H_
7 
8 #include <stdint.h>
9 #include <rte_mbuf_core.h>
10 
11 #define	HNS3_MIN_RING_DESC	64
12 #define	HNS3_MAX_RING_DESC	32768
13 #define HNS3_DEFAULT_RING_DESC  1024
14 #define	HNS3_ALIGN_RING_DESC	32
15 #define HNS3_RING_BASE_ALIGN	128
16 #define HNS3_BULK_ALLOC_MBUF_NUM	32
17 
18 #define HNS3_DEFAULT_RX_FREE_THRESH	32
19 #define HNS3_DEFAULT_TX_FREE_THRESH	32
20 #define HNS3_DEFAULT_TX_RS_THRESH	32
21 #define HNS3_TX_FAST_FREE_AHEAD		64
22 
23 #define HNS3_DEFAULT_RX_BURST		32
24 #if (HNS3_DEFAULT_RX_BURST > 64)
25 #error "PMD HNS3: HNS3_DEFAULT_RX_BURST must <= 64\n"
26 #endif
27 #define HNS3_DEFAULT_DESCS_PER_LOOP	4
28 #define HNS3_SVE_DEFAULT_DESCS_PER_LOOP	8
29 #if (HNS3_DEFAULT_DESCS_PER_LOOP > HNS3_SVE_DEFAULT_DESCS_PER_LOOP)
30 #define HNS3_VECTOR_RX_OFFSET_TABLE_LEN	HNS3_DEFAULT_DESCS_PER_LOOP
31 #else
32 #define HNS3_VECTOR_RX_OFFSET_TABLE_LEN	HNS3_SVE_DEFAULT_DESCS_PER_LOOP
33 #endif
34 #define HNS3_DEFAULT_RXQ_REARM_THRESH	64
35 #define HNS3_UINT8_BIT			8
36 #define HNS3_UINT16_BIT			16
37 #define HNS3_UINT32_BIT			32
38 
39 #define HNS3_512_BD_BUF_SIZE	512
40 #define HNS3_1K_BD_BUF_SIZE	1024
41 #define HNS3_2K_BD_BUF_SIZE	2048
42 #define HNS3_4K_BD_BUF_SIZE	4096
43 
44 #define HNS3_MIN_BD_BUF_SIZE	HNS3_512_BD_BUF_SIZE
45 #define HNS3_MAX_BD_BUF_SIZE	HNS3_4K_BD_BUF_SIZE
46 
47 #define HNS3_BD_SIZE_512_TYPE			0
48 #define HNS3_BD_SIZE_1024_TYPE			1
49 #define HNS3_BD_SIZE_2048_TYPE			2
50 #define HNS3_BD_SIZE_4096_TYPE			3
51 
52 #define HNS3_RX_FLAG_VLAN_PRESENT		0x1
53 #define HNS3_RX_FLAG_L3ID_IPV4			0x0
54 #define HNS3_RX_FLAG_L3ID_IPV6			0x1
55 #define HNS3_RX_FLAG_L4ID_UDP			0x0
56 #define HNS3_RX_FLAG_L4ID_TCP			0x1
57 
58 #define HNS3_RXD_DMAC_S				0
59 #define HNS3_RXD_DMAC_M				(0x3 << HNS3_RXD_DMAC_S)
60 #define HNS3_RXD_VLAN_S				2
61 #define HNS3_RXD_VLAN_M				(0x3 << HNS3_RXD_VLAN_S)
62 #define HNS3_RXD_L3ID_S				4
63 #define HNS3_RXD_L3ID_M				(0xf << HNS3_RXD_L3ID_S)
64 #define HNS3_RXD_L4ID_S				8
65 #define HNS3_RXD_L4ID_M				(0xf << HNS3_RXD_L4ID_S)
66 #define HNS3_RXD_FRAG_B				12
67 #define HNS3_RXD_STRP_TAGP_S			13
68 #define HNS3_RXD_STRP_TAGP_M			(0x3 << HNS3_RXD_STRP_TAGP_S)
69 
70 #define HNS3_RXD_L2E_B				16
71 #define HNS3_RXD_L3E_B				17
72 #define HNS3_RXD_L4E_B				18
73 #define HNS3_RXD_TRUNCATE_B			19
74 #define HNS3_RXD_HOI_B				20
75 #define HNS3_RXD_DOI_B				21
76 #define HNS3_RXD_OL3E_B				22
77 #define HNS3_RXD_OL4E_B				23
78 #define HNS3_RXD_GRO_COUNT_S			24
79 #define HNS3_RXD_GRO_COUNT_M			(0x3f << HNS3_RXD_GRO_COUNT_S)
80 #define HNS3_RXD_GRO_FIXID_B			30
81 #define HNS3_RXD_GRO_ECN_B			31
82 
83 #define HNS3_RXD_ODMAC_S			0
84 #define HNS3_RXD_ODMAC_M			(0x3 << HNS3_RXD_ODMAC_S)
85 #define HNS3_RXD_OVLAN_S			2
86 #define HNS3_RXD_OVLAN_M			(0x3 << HNS3_RXD_OVLAN_S)
87 #define HNS3_RXD_OL3ID_S			4
88 #define HNS3_RXD_OL3ID_M			(0xf << HNS3_RXD_OL3ID_S)
89 #define HNS3_RXD_OL4ID_S			8
90 #define HNS3_RXD_OL4ID_M			(0xf << HNS3_RXD_OL4ID_S)
91 #define HNS3_RXD_PTYPE_S			4
92 #define HNS3_RXD_PTYPE_M			(0xff << HNS3_RXD_PTYPE_S)
93 #define HNS3_RXD_FBHI_S				12
94 #define HNS3_RXD_FBHI_M				(0x3 << HNS3_RXD_FBHI_S)
95 #define HNS3_RXD_FBLI_S				14
96 #define HNS3_RXD_FBLI_M				(0x3 << HNS3_RXD_FBLI_S)
97 
98 #define HNS3_RXD_BDTYPE_S			0
99 #define HNS3_RXD_BDTYPE_M			(0xf << HNS3_RXD_BDTYPE_S)
100 #define HNS3_RXD_VLD_B				4
101 #define HNS3_RXD_UDP0_B				5
102 #define HNS3_RXD_EXTEND_B			7
103 #define HNS3_RXD_FE_B				8
104 #define HNS3_RXD_LUM_B				9
105 #define HNS3_RXD_CRCP_B				10
106 #define HNS3_RXD_L3L4P_B			11
107 #define HNS3_RXD_TSIND_S			12
108 #define HNS3_RXD_TSIND_M			(0x7 << HNS3_RXD_TSIND_S)
109 #define HNS3_RXD_LKBK_B				15
110 #define HNS3_RXD_GRO_SIZE_S			16
111 #define HNS3_RXD_GRO_SIZE_M			(0x3fff << HNS3_RXD_GRO_SIZE_S)
112 
113 #define HNS3_TXD_L3T_S				0
114 #define HNS3_TXD_L3T_M				(0x3 << HNS3_TXD_L3T_S)
115 #define HNS3_TXD_L4T_S				2
116 #define HNS3_TXD_L4T_M				(0x3 << HNS3_TXD_L4T_S)
117 #define HNS3_TXD_L3CS_B				4
118 #define HNS3_TXD_L4CS_B				5
119 #define HNS3_TXD_VLAN_B				6
120 #define HNS3_TXD_TSO_B				7
121 
122 #define HNS3_TXD_L2LEN_S			8
123 #define HNS3_TXD_L2LEN_M			(0xff << HNS3_TXD_L2LEN_S)
124 #define HNS3_TXD_L3LEN_S			16
125 #define HNS3_TXD_L3LEN_M			(0xff << HNS3_TXD_L3LEN_S)
126 #define HNS3_TXD_L4LEN_S			24
127 #define HNS3_TXD_L4LEN_M			(0xffUL << HNS3_TXD_L4LEN_S)
128 
129 #define HNS3_TXD_OL3T_S				0
130 #define HNS3_TXD_OL3T_M				(0x3 << HNS3_TXD_OL3T_S)
131 #define HNS3_TXD_OVLAN_B			2
132 #define HNS3_TXD_MACSEC_B			3
133 #define HNS3_TXD_TUNTYPE_S			4
134 #define HNS3_TXD_TUNTYPE_M			(0xf << HNS3_TXD_TUNTYPE_S)
135 
136 #define HNS3_TXD_BDTYPE_S			0
137 #define HNS3_TXD_BDTYPE_M			(0xf << HNS3_TXD_BDTYPE_S)
138 #define HNS3_TXD_FE_B				4
139 #define HNS3_TXD_SC_S				5
140 #define HNS3_TXD_SC_M				(0x3 << HNS3_TXD_SC_S)
141 #define HNS3_TXD_EXTEND_B			7
142 #define HNS3_TXD_VLD_B				8
143 #define HNS3_TXD_RI_B				9
144 #define HNS3_TXD_RA_B				10
145 #define HNS3_TXD_TSYN_B				11
146 #define HNS3_TXD_DECTTL_S			12
147 #define HNS3_TXD_DECTTL_M			(0xf << HNS3_TXD_DECTTL_S)
148 
149 #define HNS3_TXD_MSS_S				0
150 #define HNS3_TXD_MSS_M				(0x3fff << HNS3_TXD_MSS_S)
151 
152 #define HNS3_TXD_OL4CS_B			22
153 #define HNS3_L2_LEN_UNIT			1UL
154 #define HNS3_L3_LEN_UNIT			2UL
155 #define HNS3_L4_LEN_UNIT			2UL
156 
157 #define HNS3_TXD_DEFAULT_BDTYPE		0
158 #define HNS3_TXD_VLD_CMD		(0x1 << HNS3_TXD_VLD_B)
159 #define HNS3_TXD_FE_CMD			(0x1 << HNS3_TXD_FE_B)
160 #define HNS3_TXD_DEFAULT_VLD_FE_BDTYPE		\
161 		(HNS3_TXD_VLD_CMD | HNS3_TXD_FE_CMD | HNS3_TXD_DEFAULT_BDTYPE)
162 #define HNS3_TXD_SEND_SIZE_SHIFT	16
163 
164 enum hns3_pkt_l2t_type {
165 	HNS3_L2_TYPE_UNICAST,
166 	HNS3_L2_TYPE_MULTICAST,
167 	HNS3_L2_TYPE_BROADCAST,
168 	HNS3_L2_TYPE_INVALID,
169 };
170 
171 enum hns3_pkt_l3t_type {
172 	HNS3_L3T_NONE,
173 	HNS3_L3T_IPV6,
174 	HNS3_L3T_IPV4,
175 	HNS3_L3T_RESERVED
176 };
177 
178 enum hns3_pkt_l4t_type {
179 	HNS3_L4T_UNKNOWN,
180 	HNS3_L4T_TCP,
181 	HNS3_L4T_UDP,
182 	HNS3_L4T_SCTP
183 };
184 
185 enum hns3_pkt_ol3t_type {
186 	HNS3_OL3T_NONE,
187 	HNS3_OL3T_IPV6,
188 	HNS3_OL3T_IPV4_NO_CSUM,
189 	HNS3_OL3T_IPV4_CSUM
190 };
191 
192 enum hns3_pkt_tun_type {
193 	HNS3_TUN_NONE,
194 	HNS3_TUN_MAC_IN_UDP,
195 	HNS3_TUN_NVGRE,
196 	HNS3_TUN_OTHER
197 };
198 
199 /* hardware spec ring buffer format */
200 struct hns3_desc {
201 	union {
202 		uint64_t addr;
203 		struct {
204 			uint32_t addr0;
205 			uint32_t addr1;
206 		};
207 	};
208 	union {
209 		struct {
210 			uint16_t vlan_tag;
211 			uint16_t send_size;
212 			union {
213 				/*
214 				 * L3T | L4T | L3CS | L4CS | VLAN | TSO |
215 				 * L2_LEN
216 				 */
217 				uint32_t type_cs_vlan_tso_len;
218 				struct {
219 					uint8_t type_cs_vlan_tso;
220 					uint8_t l2_len;
221 					uint8_t l3_len;
222 					uint8_t l4_len;
223 				};
224 			};
225 			uint16_t outer_vlan_tag;
226 			uint16_t tv;
227 			union {
228 				/* OL3T | OVALAN | MACSEC */
229 				uint32_t ol_type_vlan_len_msec;
230 				struct {
231 					uint8_t ol_type_vlan_msec;
232 					uint8_t ol2_len;
233 					uint8_t ol3_len;
234 					uint8_t ol4_len;
235 				};
236 			};
237 
238 			uint32_t paylen_fd_dop_ol4cs;
239 			uint16_t tp_fe_sc_vld_ra_ri;
240 			uint16_t mss;
241 		} tx;
242 
243 		struct {
244 			uint32_t l234_info;
245 			uint16_t pkt_len;
246 			uint16_t size;
247 			uint32_t rss_hash;
248 			uint16_t fd_id;
249 			uint16_t vlan_tag;
250 			union {
251 				uint32_t ol_info;
252 				struct {
253 					uint16_t o_dm_vlan_id_fb;
254 					uint16_t ot_vlan_tag;
255 				};
256 			};
257 			union {
258 				uint32_t bd_base_info;
259 				struct {
260 					uint16_t bdtype_vld_udp0;
261 					uint16_t fe_lum_crcp_l3l4p;
262 				};
263 			};
264 		} rx;
265 	};
266 } __rte_packed;
267 
268 struct hns3_entry {
269 	struct rte_mbuf *mbuf;
270 };
271 
272 struct hns3_rx_basic_stats {
273 	uint64_t packets;
274 	uint64_t bytes;
275 	uint64_t errors;
276 };
277 
278 struct hns3_rx_dfx_stats {
279 	uint64_t l3_csum_errors;
280 	uint64_t l4_csum_errors;
281 	uint64_t ol3_csum_errors;
282 	uint64_t ol4_csum_errors;
283 };
284 
285 struct hns3_rx_bd_errors_stats {
286 	uint64_t l2_errors;
287 	uint64_t pkt_len_errors;
288 };
289 
290 struct hns3_rx_queue {
291 	void *io_base;
292 	volatile void *io_head_reg;
293 	struct hns3_adapter *hns;
294 	struct hns3_ptype_table *ptype_tbl;
295 	struct rte_mempool *mb_pool;
296 	struct hns3_desc *rx_ring;
297 	uint64_t rx_ring_phys_addr; /* RX ring DMA address */
298 	const struct rte_memzone *mz;
299 	struct hns3_entry *sw_ring;
300 	struct rte_mbuf *pkt_first_seg;
301 	struct rte_mbuf *pkt_last_seg;
302 
303 	uint16_t queue_id;
304 	uint16_t port_id;
305 	uint16_t nb_rx_desc;
306 	uint16_t rx_buf_len;
307 	/*
308 	 * threshold for the number of BDs waited to passed to hardware. If the
309 	 * number exceeds the threshold, driver will pass these BDs to hardware.
310 	 */
311 	uint16_t rx_free_thresh;
312 	uint16_t next_to_use;    /* index of next BD to be polled */
313 	uint16_t rx_free_hold;   /* num of BDs waited to passed to hardware */
314 	uint16_t rx_rearm_start; /* index of BD that driver re-arming from */
315 	uint16_t rx_rearm_nb;    /* number of remaining BDs to be re-armed */
316 
317 	/* 4 if DEV_RX_OFFLOAD_KEEP_CRC offload set, 0 otherwise */
318 	uint8_t crc_len;
319 
320 	bool rx_deferred_start; /* don't start this queue in dev start */
321 	bool configured;        /* indicate if rx queue has been configured */
322 	/*
323 	 * Indicate whether ignore the outer VLAN field in the Rx BD reported
324 	 * by the Hardware. Because the outer VLAN is the PVID if the PVID is
325 	 * set for some version of hardware network engine whose vlan mode is
326 	 * HNS3_SW_SHIFT_AND_DISCARD_MODE, such as kunpeng 920. And this VLAN
327 	 * should not be transitted to the upper-layer application. For hardware
328 	 * network engine whose vlan mode is HNS3_HW_SHIFT_AND_DISCARD_MODE,
329 	 * such as kunpeng 930, PVID will not be reported to the BDs. So, PMD
330 	 * driver does not need to perform PVID-related operation in Rx. At this
331 	 * point, the pvid_sw_discard_en will be false.
332 	 */
333 	bool pvid_sw_discard_en;
334 	bool ptype_en;          /* indicate if the ptype field enabled */
335 	bool enabled;           /* indicate if Rx queue has been enabled */
336 
337 	struct hns3_rx_basic_stats basic_stats;
338 	/* DFX statistics that driver does not need to discard packets */
339 	struct hns3_rx_dfx_stats dfx_stats;
340 	/* Error statistics that driver needs to discard packets */
341 	struct hns3_rx_bd_errors_stats err_stats;
342 
343 	struct rte_mbuf *bulk_mbuf[HNS3_BULK_ALLOC_MBUF_NUM];
344 	uint16_t bulk_mbuf_num;
345 
346 	/* offset_table: used for vector, to solve execute re-order problem */
347 	uint8_t offset_table[HNS3_VECTOR_RX_OFFSET_TABLE_LEN + 1];
348 	uint64_t mbuf_initializer; /* value to init mbufs used with vector rx */
349 	struct rte_mbuf fake_mbuf; /* fake mbuf used with vector rx */
350 };
351 
352 struct hns3_tx_basic_stats {
353 	uint64_t packets;
354 	uint64_t bytes;
355 };
356 
357 /*
358  * The following items are used for the abnormal errors statistics in
359  * the Tx datapath. When upper level application calls the
360  * rte_eth_tx_burst API function to send multiple packets at a time with
361  * burst mode based on hns3 network engine, there are some abnormal
362  * conditions that cause the driver to fail to operate the hardware to
363  * send packets correctly.
364  * Note: When using burst mode to call the rte_eth_tx_burst API function
365  * to send multiple packets at a time. When the first abnormal error is
366  * detected, add one to the relevant error statistics item, and then
367  * exit the loop of sending multiple packets of the function. That is to
368  * say, even if there are multiple packets in which abnormal errors may
369  * be detected in the burst, the relevant error statistics in the driver
370  * will only be increased by one.
371  * The detail description of the Tx abnormal errors statistic items as
372  * below:
373  *  - over_length_pkt_cnt
374  *     Total number of greater than HNS3_MAX_FRAME_LEN the driver
375  *     supported.
376  *
377  * - exceed_limit_bd_pkt_cnt
378  *     Total number of exceeding the hardware limited bd which process
379  *     a packet needed bd numbers.
380  *
381  * - exceed_limit_bd_reassem_fail
382  *     Total number of exceeding the hardware limited bd fail which
383  *     process a packet needed bd numbers and reassemble fail.
384  *
385  * - unsupported_tunnel_pkt_cnt
386  *     Total number of unsupported tunnel packet. The unsupported tunnel
387  *     type: vxlan_gpe, gtp, ipip and MPLSINUDP, MPLSINUDP is a packet
388  *     with MPLS-in-UDP RFC 7510 header.
389  *
390  * - queue_full_cnt
391  *     Total count which the available bd numbers in current bd queue is
392  *     less than the bd numbers with the pkt process needed.
393  *
394  * - pkt_padding_fail_cnt
395  *     Total count which the packet length is less than minimum packet
396  *     length(struct hns3_tx_queue::min_tx_pkt_len) supported by
397  *     hardware in Tx direction and fail to be appended with 0.
398  */
399 struct hns3_tx_dfx_stats {
400 	uint64_t over_length_pkt_cnt;
401 	uint64_t exceed_limit_bd_pkt_cnt;
402 	uint64_t exceed_limit_bd_reassem_fail;
403 	uint64_t unsupported_tunnel_pkt_cnt;
404 	uint64_t queue_full_cnt;
405 	uint64_t pkt_padding_fail_cnt;
406 };
407 
408 struct hns3_tx_queue {
409 	void *io_base;
410 	volatile void *io_tail_reg;
411 	struct hns3_adapter *hns;
412 	struct hns3_desc *tx_ring;
413 	uint64_t tx_ring_phys_addr; /* TX ring DMA address */
414 	const struct rte_memzone *mz;
415 	struct hns3_entry *sw_ring;
416 
417 	uint16_t queue_id;
418 	uint16_t port_id;
419 	uint16_t nb_tx_desc;
420 	/*
421 	 * index of next BD whose corresponding rte_mbuf can be released by
422 	 * driver.
423 	 */
424 	uint16_t next_to_clean;
425 	/* index of next BD to be filled by driver to send packet */
426 	uint16_t next_to_use;
427 	/* num of remaining BDs ready to be filled by driver to send packet */
428 	uint16_t tx_bd_ready;
429 
430 	/* threshold for free tx buffer if available BDs less than this value */
431 	uint16_t tx_free_thresh;
432 
433 	/*
434 	 * For better performance in tx datapath, releasing mbuf in batches is
435 	 * required.
436 	 * Only checking the VLD bit of the last descriptor in a batch of the
437 	 * thresh descriptors does not mean that these descriptors are all sent
438 	 * by hardware successfully. So we need to check that the VLD bits of
439 	 * all descriptors are cleared. and then free all mbufs in the batch.
440 	 * - tx_rs_thresh
441 	 *   Number of mbufs released at a time.
442 	 *
443 	 * - free
444 	 *   Tx mbuf free array used for preserving temporarily address of mbuf
445 	 *   released back to mempool, when releasing mbuf in batches.
446 	 */
447 	uint16_t tx_rs_thresh;
448 	struct rte_mbuf **free;
449 
450 	/*
451 	 * tso mode.
452 	 * value range:
453 	 *      HNS3_TSO_SW_CAL_PSEUDO_H_CSUM/HNS3_TSO_HW_CAL_PSEUDO_H_CSUM
454 	 *
455 	 *  - HNS3_TSO_SW_CAL_PSEUDO_H_CSUM
456 	 *     In this mode, because of the hardware constraint, network driver
457 	 *     software need erase the L4 len value of the TCP pseudo header
458 	 *     and recalculate the TCP pseudo header checksum of packets that
459 	 *     need TSO.
460 	 *
461 	 *  - HNS3_TSO_HW_CAL_PSEUDO_H_CSUM
462 	 *     In this mode, hardware support recalculate the TCP pseudo header
463 	 *     checksum of packets that need TSO, so network driver software
464 	 *     not need to recalculate it.
465 	 */
466 	uint8_t tso_mode;
467 	/*
468 	 * udp checksum mode.
469 	 * value range:
470 	 *      HNS3_SPECIAL_PORT_HW_CKSUM_MODE/HNS3_SPECIAL_PORT_SW_CKSUM_MODE
471 	 *
472 	 *  - HNS3_SPECIAL_PORT_SW_CKSUM_MODE
473 	 *     In this mode, HW can not do checksum for special UDP port like
474 	 *     4789, 4790, 6081 for non-tunnel UDP packets and UDP tunnel
475 	 *     packets without the PKT_TX_TUNEL_MASK in the mbuf. So, PMD need
476 	 *     do the checksum for these packets to avoid a checksum error.
477 	 *
478 	 *  - HNS3_SPECIAL_PORT_HW_CKSUM_MODE
479 	 *     In this mode, HW does not have the preceding problems and can
480 	 *     directly calculate the checksum of these UDP packets.
481 	 */
482 	uint8_t udp_cksum_mode;
483 	/*
484 	 * The minimum length of the packet supported by hardware in the Tx
485 	 * direction.
486 	 */
487 	uint32_t min_tx_pkt_len;
488 
489 	uint8_t max_non_tso_bd_num; /* max BD number of one non-TSO packet */
490 	bool tx_deferred_start; /* don't start this queue in dev start */
491 	bool configured;        /* indicate if tx queue has been configured */
492 	/*
493 	 * Indicate whether add the vlan_tci of the mbuf to the inner VLAN field
494 	 * of Tx BD. Because the outer VLAN will always be the PVID when the
495 	 * PVID is set and for some version of hardware network engine whose
496 	 * vlan mode is HNS3_SW_SHIFT_AND_DISCARD_MODE, such as kunpeng 920, the
497 	 * PVID will overwrite the outer VLAN field of Tx BD. For the hardware
498 	 * network engine whose vlan mode is HNS3_HW_SHIFT_AND_DISCARD_MODE,
499 	 * such as kunpeng 930, if the PVID is set, the hardware will shift the
500 	 * VLAN field automatically. So, PMD driver does not need to do
501 	 * PVID-related operations in Tx. And pvid_sw_shift_en will be false at
502 	 * this point.
503 	 */
504 	bool pvid_sw_shift_en;
505 	bool enabled;           /* indicate if Tx queue has been enabled */
506 
507 	struct hns3_tx_basic_stats basic_stats;
508 	struct hns3_tx_dfx_stats dfx_stats;
509 };
510 
511 #define HNS3_GET_TX_QUEUE_PEND_BD_NUM(txq) \
512 		((txq)->nb_tx_desc - 1 - (txq)->tx_bd_ready)
513 
514 struct hns3_queue_info {
515 	const char *type;   /* point to queue memory name */
516 	const char *ring_name;  /* point to hardware ring name */
517 	uint16_t idx;
518 	uint16_t nb_desc;
519 	unsigned int socket_id;
520 };
521 
522 #define HNS3_TX_CKSUM_OFFLOAD_MASK ( \
523 	PKT_TX_OUTER_UDP_CKSUM | \
524 	PKT_TX_OUTER_IP_CKSUM | \
525 	PKT_TX_IP_CKSUM | \
526 	PKT_TX_TCP_SEG | \
527 	PKT_TX_L4_MASK)
528 
529 enum hns3_cksum_status {
530 	HNS3_CKSUM_NONE = 0,
531 	HNS3_L3_CKSUM_ERR = 1,
532 	HNS3_L4_CKSUM_ERR = 2,
533 	HNS3_OUTER_L3_CKSUM_ERR = 4,
534 	HNS3_OUTER_L4_CKSUM_ERR = 8
535 };
536 
537 static inline int
538 hns3_handle_bdinfo(struct hns3_rx_queue *rxq, struct rte_mbuf *rxm,
539 		   uint32_t bd_base_info, uint32_t l234_info,
540 		   uint32_t *cksum_err)
541 {
542 #define L2E_TRUNC_ERR_FLAG	(BIT(HNS3_RXD_L2E_B) | \
543 				 BIT(HNS3_RXD_TRUNCATE_B))
544 #define CHECKSUM_ERR_FLAG	(BIT(HNS3_RXD_L3E_B) | \
545 				 BIT(HNS3_RXD_L4E_B) | \
546 				 BIT(HNS3_RXD_OL3E_B) | \
547 				 BIT(HNS3_RXD_OL4E_B))
548 
549 	uint32_t tmp = 0;
550 
551 	/*
552 	 * If packet len bigger than mtu when recv with no-scattered algorithm,
553 	 * the first n bd will without FE bit, we need process this sisution.
554 	 * Note: we don't need add statistic counter because latest BD which
555 	 *       with FE bit will mark HNS3_RXD_L2E_B bit.
556 	 */
557 	if (unlikely((bd_base_info & BIT(HNS3_RXD_FE_B)) == 0))
558 		return -EINVAL;
559 
560 	if (unlikely((l234_info & L2E_TRUNC_ERR_FLAG) || rxm->pkt_len == 0)) {
561 		if (l234_info & BIT(HNS3_RXD_L2E_B))
562 			rxq->err_stats.l2_errors++;
563 		else
564 			rxq->err_stats.pkt_len_errors++;
565 		return -EINVAL;
566 	}
567 
568 	if (bd_base_info & BIT(HNS3_RXD_L3L4P_B)) {
569 		if (likely((l234_info & CHECKSUM_ERR_FLAG) == 0)) {
570 			*cksum_err = 0;
571 			return 0;
572 		}
573 
574 		if (unlikely(l234_info & BIT(HNS3_RXD_L3E_B))) {
575 			rxm->ol_flags |= PKT_RX_IP_CKSUM_BAD;
576 			rxq->dfx_stats.l3_csum_errors++;
577 			tmp |= HNS3_L3_CKSUM_ERR;
578 		}
579 
580 		if (unlikely(l234_info & BIT(HNS3_RXD_L4E_B))) {
581 			rxm->ol_flags |= PKT_RX_L4_CKSUM_BAD;
582 			rxq->dfx_stats.l4_csum_errors++;
583 			tmp |= HNS3_L4_CKSUM_ERR;
584 		}
585 
586 		if (unlikely(l234_info & BIT(HNS3_RXD_OL3E_B))) {
587 			rxq->dfx_stats.ol3_csum_errors++;
588 			tmp |= HNS3_OUTER_L3_CKSUM_ERR;
589 		}
590 
591 		if (unlikely(l234_info & BIT(HNS3_RXD_OL4E_B))) {
592 			rxm->ol_flags |= PKT_RX_OUTER_L4_CKSUM_BAD;
593 			rxq->dfx_stats.ol4_csum_errors++;
594 			tmp |= HNS3_OUTER_L4_CKSUM_ERR;
595 		}
596 	}
597 	*cksum_err = tmp;
598 
599 	return 0;
600 }
601 
602 static inline void
603 hns3_rx_set_cksum_flag(struct rte_mbuf *rxm, const uint64_t packet_type,
604 		       const uint32_t cksum_err)
605 {
606 	if (unlikely((packet_type & RTE_PTYPE_TUNNEL_MASK))) {
607 		if (likely(packet_type & RTE_PTYPE_INNER_L3_MASK) &&
608 		    (cksum_err & HNS3_L3_CKSUM_ERR) == 0)
609 			rxm->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
610 		if (likely(packet_type & RTE_PTYPE_INNER_L4_MASK) &&
611 		    (cksum_err & HNS3_L4_CKSUM_ERR) == 0)
612 			rxm->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
613 		if (likely(packet_type & RTE_PTYPE_L4_MASK) &&
614 		    (cksum_err & HNS3_OUTER_L4_CKSUM_ERR) == 0)
615 			rxm->ol_flags |= PKT_RX_OUTER_L4_CKSUM_GOOD;
616 	} else {
617 		if (likely(packet_type & RTE_PTYPE_L3_MASK) &&
618 		    (cksum_err & HNS3_L3_CKSUM_ERR) == 0)
619 			rxm->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
620 		if (likely(packet_type & RTE_PTYPE_L4_MASK) &&
621 		    (cksum_err & HNS3_L4_CKSUM_ERR) == 0)
622 			rxm->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
623 	}
624 }
625 
626 static inline uint32_t
627 hns3_rx_calc_ptype(struct hns3_rx_queue *rxq, const uint32_t l234_info,
628 		   const uint32_t ol_info)
629 {
630 	const struct hns3_ptype_table * const ptype_tbl = rxq->ptype_tbl;
631 	uint32_t l2id, l3id, l4id;
632 	uint32_t ol3id, ol4id, ol2id;
633 	uint32_t ptype;
634 
635 	if (rxq->ptype_en) {
636 		ptype = hns3_get_field(ol_info, HNS3_RXD_PTYPE_M,
637 				       HNS3_RXD_PTYPE_S);
638 		return ptype_tbl->ptype[ptype];
639 	}
640 
641 	ol4id = hns3_get_field(ol_info, HNS3_RXD_OL4ID_M, HNS3_RXD_OL4ID_S);
642 	ol3id = hns3_get_field(ol_info, HNS3_RXD_OL3ID_M, HNS3_RXD_OL3ID_S);
643 	ol2id = hns3_get_field(ol_info, HNS3_RXD_OVLAN_M, HNS3_RXD_OVLAN_S);
644 	l2id = hns3_get_field(l234_info, HNS3_RXD_VLAN_M, HNS3_RXD_VLAN_S);
645 	l3id = hns3_get_field(l234_info, HNS3_RXD_L3ID_M, HNS3_RXD_L3ID_S);
646 	l4id = hns3_get_field(l234_info, HNS3_RXD_L4ID_M, HNS3_RXD_L4ID_S);
647 
648 	if (unlikely(ptype_tbl->ol4table[ol4id]))
649 		return ptype_tbl->inner_l2table[l2id] |
650 			ptype_tbl->inner_l3table[l3id] |
651 			ptype_tbl->inner_l4table[l4id] |
652 			ptype_tbl->ol3table[ol3id] |
653 			ptype_tbl->ol4table[ol4id] | ptype_tbl->ol2table[ol2id];
654 	else
655 		return ptype_tbl->l2l3table[l2id][l3id] |
656 			ptype_tbl->l4table[l4id];
657 }
658 
659 void hns3_dev_rx_queue_release(void *queue);
660 void hns3_dev_tx_queue_release(void *queue);
661 void hns3_free_all_queues(struct rte_eth_dev *dev);
662 int hns3_reset_all_tqps(struct hns3_adapter *hns);
663 void hns3_dev_all_rx_queue_intr_enable(struct hns3_hw *hw, bool en);
664 int hns3_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
665 int hns3_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
666 void hns3_enable_all_queues(struct hns3_hw *hw, bool en);
667 int hns3_init_queues(struct hns3_adapter *hns, bool reset_queue);
668 void hns3_start_tqps(struct hns3_hw *hw);
669 void hns3_stop_tqps(struct hns3_hw *hw);
670 int hns3_rxq_iterate(struct rte_eth_dev *dev,
671 		 int (*callback)(struct hns3_rx_queue *, void *), void *arg);
672 void hns3_dev_release_mbufs(struct hns3_adapter *hns);
673 int hns3_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t nb_desc,
674 			unsigned int socket, const struct rte_eth_rxconf *conf,
675 			struct rte_mempool *mp);
676 int hns3_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t nb_desc,
677 			unsigned int socket, const struct rte_eth_txconf *conf);
678 uint32_t hns3_rx_queue_count(struct rte_eth_dev *dev, uint16_t rx_queue_id);
679 int hns3_dev_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id);
680 int hns3_dev_rx_queue_stop(struct rte_eth_dev *dev, uint16_t rx_queue_id);
681 int hns3_dev_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id);
682 int hns3_dev_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id);
683 uint16_t hns3_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
684 			uint16_t nb_pkts);
685 uint16_t hns3_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
686 				  uint16_t nb_pkts);
687 uint16_t hns3_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
688 			    uint16_t nb_pkts);
689 uint16_t hns3_recv_pkts_vec_sve(void *rx_queue, struct rte_mbuf **rx_pkts,
690 				uint16_t nb_pkts);
691 int hns3_rx_burst_mode_get(struct rte_eth_dev *dev,
692 			   __rte_unused uint16_t queue_id,
693 			   struct rte_eth_burst_mode *mode);
694 int hns3_rx_check_vec_support(struct rte_eth_dev *dev);
695 uint16_t hns3_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
696 			uint16_t nb_pkts);
697 uint16_t hns3_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts,
698 			       uint16_t nb_pkts);
699 uint16_t hns3_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
700 			uint16_t nb_pkts);
701 uint16_t hns3_xmit_pkts_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
702 							uint16_t nb_pkts);
703 uint16_t hns3_xmit_pkts_vec_sve(void *tx_queue, struct rte_mbuf **tx_pkts,
704 				uint16_t nb_pkts);
705 int hns3_tx_burst_mode_get(struct rte_eth_dev *dev,
706 			   __rte_unused uint16_t queue_id,
707 			   struct rte_eth_burst_mode *mode);
708 const uint32_t *hns3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
709 void hns3_init_rx_ptype_tble(struct rte_eth_dev *dev);
710 void hns3_set_rxtx_function(struct rte_eth_dev *eth_dev);
711 uint32_t hns3_get_tqp_intr_reg_offset(uint16_t tqp_intr_id);
712 void hns3_set_queue_intr_gl(struct hns3_hw *hw, uint16_t queue_id,
713 			    uint8_t gl_idx, uint16_t gl_value);
714 void hns3_set_queue_intr_rl(struct hns3_hw *hw, uint16_t queue_id,
715 			    uint16_t rl_value);
716 void hns3_set_queue_intr_ql(struct hns3_hw *hw, uint16_t queue_id,
717 			    uint16_t ql_value);
718 int hns3_set_fake_rx_or_tx_queues(struct rte_eth_dev *dev, uint16_t nb_rx_q,
719 				  uint16_t nb_tx_q);
720 int hns3_config_gro(struct hns3_hw *hw, bool en);
721 int hns3_restore_gro_conf(struct hns3_hw *hw);
722 void hns3_update_all_queues_pvid_proc_en(struct hns3_hw *hw);
723 void hns3_rx_scattered_reset(struct rte_eth_dev *dev);
724 void hns3_rx_scattered_calc(struct rte_eth_dev *dev);
725 int hns3_rx_check_vec_support(struct rte_eth_dev *dev);
726 int hns3_tx_check_vec_support(struct rte_eth_dev *dev);
727 void hns3_rxq_vec_setup(struct hns3_rx_queue *rxq);
728 void hns3_rxq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
729 		       struct rte_eth_rxq_info *qinfo);
730 void hns3_txq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
731 		       struct rte_eth_txq_info *qinfo);
732 uint32_t hns3_get_tqp_reg_offset(uint16_t idx);
733 int hns3_start_all_txqs(struct rte_eth_dev *dev);
734 int hns3_start_all_rxqs(struct rte_eth_dev *dev);
735 void hns3_stop_all_txqs(struct rte_eth_dev *dev);
736 void hns3_restore_tqp_enable_state(struct hns3_hw *hw);
737 int hns3_tx_done_cleanup(void *txq, uint32_t free_cnt);
738 void hns3_enable_rxd_adv_layout(struct hns3_hw *hw);
739 int hns3_dev_rx_descriptor_status(void *rx_queue, uint16_t offset);
740 int hns3_dev_tx_descriptor_status(void *tx_queue, uint16_t offset);
741 
742 #endif /* _HNS3_RXTX_H_ */
743