xref: /freebsd/sys/dev/vnic/nic.h (revision a7623790fb345e6dc986dfd31df0ace115e6f2e4)
1 /*
2  * Copyright (C) 2015 Cavium Inc.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  *
14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24  * SUCH DAMAGE.
25  *
26  * $FreeBSD$
27  *
28  */
29 
30 #ifndef NIC_H
31 #define	NIC_H
32 
33 /* PCI vendor ID */
34 #define PCI_VENDOR_ID_CAVIUM			0x177D
35 /* PCI device IDs */
36 #define	PCI_DEVICE_ID_THUNDER_NIC_PF		0xA01E
37 #define	PCI_DEVICE_ID_THUNDER_PASS1_NIC_VF	0x0011
38 #define	PCI_DEVICE_ID_THUNDER_NIC_VF		0xA034
39 #define	PCI_DEVICE_ID_THUNDER_BGX		0xA026
40 
41 /* PCI BAR nos */
42 #define	PCI_CFG_REG_BAR_NUM		0
43 #define	PCI_MSIX_REG_BAR_NUM		4
44 
45 /* PCI revision IDs */
46 #define	PCI_REVID_PASS2			8
47 
48 /* NIC SRIOV VF count */
49 #define	MAX_NUM_VFS_SUPPORTED		128
50 #define	DEFAULT_NUM_VF_ENABLED		8
51 
52 #define	NIC_TNS_BYPASS_MODE		0
53 #define	NIC_TNS_MODE			1
54 
55 /* NIC priv flags */
56 #define	NIC_SRIOV_ENABLED		(1 << 0)
57 #define	NIC_TNS_ENABLED			(1 << 1)
58 
59 /* ARM64TODO */
60 #if 0
61 /* VNIC HW optimiation features */
62 #define VNIC_RSS_SUPPORT
63 #define VNIC_MULTI_QSET_SUPPORT
64 #endif
65 
66 /* Min/Max packet size */
67 #define	NIC_HW_MIN_FRS			64
68 #define	NIC_HW_MAX_FRS			9200 /* 9216 max packet including FCS */
69 
70 /* Max pkinds */
71 #define	NIC_MAX_PKIND			16
72 
73 /*
74  * Rx Channels */
75 /* Receive channel configuration in TNS bypass mode
76  * Below is configuration in TNS bypass mode
77  * BGX0-LMAC0-CHAN0 - VNIC CHAN0
78  * BGX0-LMAC1-CHAN0 - VNIC CHAN16
79  * ...
80  * BGX1-LMAC0-CHAN0 - VNIC CHAN128
81  * ...
82  * BGX1-LMAC3-CHAN0 - VNIC CHAN174
83  */
84 #define	NIC_INTF_COUNT			2  /* Interfaces btw VNIC and TNS/BGX */
85 #define	NIC_CHANS_PER_INF		128
86 #define	NIC_MAX_CHANS			(NIC_INTF_COUNT * NIC_CHANS_PER_INF)
87 #define	NIC_CPI_COUNT			2048 /* No of channel parse indices */
88 
89 /* TNS bypass mode: 1-1 mapping between VNIC and BGX:LMAC */
90 #define	NIC_MAX_BGX			MAX_BGX_PER_CN88XX
91 #define	NIC_CPI_PER_BGX			(NIC_CPI_COUNT / NIC_MAX_BGX)
92 #define	NIC_MAX_CPI_PER_LMAC		64 /* Max when CPI_ALG is IP diffserv */
93 #define	NIC_RSSI_PER_BGX		(NIC_RSSI_COUNT / NIC_MAX_BGX)
94 
95 /* Tx scheduling */
96 #define	NIC_MAX_TL4			1024
97 #define	NIC_MAX_TL4_SHAPERS		256 /* 1 shaper for 4 TL4s */
98 #define	NIC_MAX_TL3			256
99 #define	NIC_MAX_TL3_SHAPERS		64  /* 1 shaper for 4 TL3s */
100 #define	NIC_MAX_TL2			64
101 #define	NIC_MAX_TL2_SHAPERS		2  /* 1 shaper for 32 TL2s */
102 #define	NIC_MAX_TL1			2
103 
104 /* TNS bypass mode */
105 #define	NIC_TL2_PER_BGX			32
106 #define	NIC_TL4_PER_BGX			(NIC_MAX_TL4 / NIC_MAX_BGX)
107 #define	NIC_TL4_PER_LMAC		(NIC_MAX_TL4 / NIC_CHANS_PER_INF)
108 
109 /* NIC VF Interrupts */
110 #define	NICVF_INTR_CQ			0
111 #define	NICVF_INTR_SQ			1
112 #define	NICVF_INTR_RBDR			2
113 #define	NICVF_INTR_PKT_DROP		3
114 #define	NICVF_INTR_TCP_TIMER		4
115 #define	NICVF_INTR_MBOX			5
116 #define	NICVF_INTR_QS_ERR		6
117 
118 #define	NICVF_INTR_CQ_SHIFT		0
119 #define	NICVF_INTR_SQ_SHIFT		8
120 #define	NICVF_INTR_RBDR_SHIFT		16
121 #define	NICVF_INTR_PKT_DROP_SHIFT	20
122 #define	NICVF_INTR_TCP_TIMER_SHIFT	21
123 #define	NICVF_INTR_MBOX_SHIFT		22
124 #define	NICVF_INTR_QS_ERR_SHIFT		23
125 
126 #define	NICVF_INTR_CQ_MASK		(0xFF << NICVF_INTR_CQ_SHIFT)
127 #define	NICVF_INTR_SQ_MASK		(0xFF << NICVF_INTR_SQ_SHIFT)
128 #define	NICVF_INTR_RBDR_MASK		(0x03 << NICVF_INTR_RBDR_SHIFT)
129 #define	NICVF_INTR_PKT_DROP_MASK	(1 << NICVF_INTR_PKT_DROP_SHIFT)
130 #define	NICVF_INTR_TCP_TIMER_MASK	(1 << NICVF_INTR_TCP_TIMER_SHIFT)
131 #define	NICVF_INTR_MBOX_MASK		(1 << NICVF_INTR_MBOX_SHIFT)
132 #define	NICVF_INTR_QS_ERR_MASK		(1 << NICVF_INTR_QS_ERR_SHIFT)
133 
134 /* MSI-X interrupts */
135 #define	NIC_PF_MSIX_VECTORS		10
136 #define	NIC_VF_MSIX_VECTORS		20
137 
138 #define	NIC_PF_INTR_ID_ECC0_SBE		0
139 #define	NIC_PF_INTR_ID_ECC0_DBE		1
140 #define	NIC_PF_INTR_ID_ECC1_SBE		2
141 #define	NIC_PF_INTR_ID_ECC1_DBE		3
142 #define	NIC_PF_INTR_ID_ECC2_SBE		4
143 #define	NIC_PF_INTR_ID_ECC2_DBE		5
144 #define	NIC_PF_INTR_ID_ECC3_SBE		6
145 #define	NIC_PF_INTR_ID_ECC3_DBE		7
146 #define	NIC_PF_INTR_ID_MBOX0		8
147 #define	NIC_PF_INTR_ID_MBOX1		9
148 
149 struct msix_entry {
150 	struct resource *	irq_res;
151 	void *			handle;
152 };
153 
154 /*
155  * Global timer for CQ timer thresh interrupts
156  * Calculated for SCLK of 700Mhz
157  * value written should be a 1/16th of what is expected
158  *
159  * 1 tick per 0.05usec = value of 2.2
160  * This 10% would be covered in CQ timer thresh value
161  */
162 #define NICPF_CLK_PER_INT_TICK		2
163 
164 /*
165  * Time to wait before we decide that a SQ is stuck.
166  *
167  * Since both pkt rx and tx notifications are done with same CQ,
168  * when packets are being received at very high rate (eg: L2 forwarding)
169  * then freeing transmitted skbs will be delayed and watchdog
170  * will kick in, resetting interface. Hence keeping this value high.
171  */
172 #define	NICVF_TX_TIMEOUT		(50 * HZ)
173 
174 #define	NIC_RSSI_COUNT			4096 /* Total no of RSS indices */
175 #define	NIC_MAX_RSS_HASH_BITS		8
176 #define	NIC_MAX_RSS_IDR_TBL_SIZE	(1 << NIC_MAX_RSS_HASH_BITS)
177 #define	RSS_HASH_KEY_SIZE		5 /* 320 bit key */
178 
179 struct nicvf_rss_info {
180 	boolean_t enable;
181 #define	RSS_L2_EXTENDED_HASH_ENA	(1UL << 0)
182 #define	RSS_IP_HASH_ENA			(1UL << 1)
183 #define	RSS_TCP_HASH_ENA		(1UL << 2)
184 #define	RSS_TCP_SYN_DIS			(1UL << 3)
185 #define	RSS_UDP_HASH_ENA		(1UL << 4)
186 #define	RSS_L4_EXTENDED_HASH_ENA	(1UL << 5)
187 #define	RSS_ROCE_ENA			(1UL << 6)
188 #define	RSS_L3_BI_DIRECTION_ENA		(1UL << 7)
189 #define	RSS_L4_BI_DIRECTION_ENA		(1UL << 8)
190 	uint64_t cfg;
191 	uint8_t  hash_bits;
192 	uint16_t rss_size;
193 	uint8_t  ind_tbl[NIC_MAX_RSS_IDR_TBL_SIZE];
194 	uint64_t key[RSS_HASH_KEY_SIZE];
195 };
196 
197 enum rx_stats_reg_offset {
198 	RX_OCTS = 0x0,
199 	RX_UCAST = 0x1,
200 	RX_BCAST = 0x2,
201 	RX_MCAST = 0x3,
202 	RX_RED = 0x4,
203 	RX_RED_OCTS = 0x5,
204 	RX_ORUN = 0x6,
205 	RX_ORUN_OCTS = 0x7,
206 	RX_FCS = 0x8,
207 	RX_L2ERR = 0x9,
208 	RX_DRP_BCAST = 0xa,
209 	RX_DRP_MCAST = 0xb,
210 	RX_DRP_L3BCAST = 0xc,
211 	RX_DRP_L3MCAST = 0xd,
212 	RX_STATS_ENUM_LAST,
213 };
214 
215 enum tx_stats_reg_offset {
216 	TX_OCTS = 0x0,
217 	TX_UCAST = 0x1,
218 	TX_BCAST = 0x2,
219 	TX_MCAST = 0x3,
220 	TX_DROP = 0x4,
221 	TX_STATS_ENUM_LAST,
222 };
223 
224 struct nicvf_hw_stats {
225 	uint64_t rx_bytes;
226 	uint64_t rx_ucast_frames;
227 	uint64_t rx_bcast_frames;
228 	uint64_t rx_mcast_frames;
229 	uint64_t rx_fcs_errors;
230 	uint64_t rx_l2_errors;
231 	uint64_t rx_drop_red;
232 	uint64_t rx_drop_red_bytes;
233 	uint64_t rx_drop_overrun;
234 	uint64_t rx_drop_overrun_bytes;
235 	uint64_t rx_drop_bcast;
236 	uint64_t rx_drop_mcast;
237 	uint64_t rx_drop_l3_bcast;
238 	uint64_t rx_drop_l3_mcast;
239 	uint64_t rx_bgx_truncated_pkts;
240 	uint64_t rx_jabber_errs;
241 	uint64_t rx_fcs_errs;
242 	uint64_t rx_bgx_errs;
243 	uint64_t rx_prel2_errs;
244 	uint64_t rx_l2_hdr_malformed;
245 	uint64_t rx_oversize;
246 	uint64_t rx_undersize;
247 	uint64_t rx_l2_len_mismatch;
248 	uint64_t rx_l2_pclp;
249 	uint64_t rx_ip_ver_errs;
250 	uint64_t rx_ip_csum_errs;
251 	uint64_t rx_ip_hdr_malformed;
252 	uint64_t rx_ip_payload_malformed;
253 	uint64_t rx_ip_ttl_errs;
254 	uint64_t rx_l3_pclp;
255 	uint64_t rx_l4_malformed;
256 	uint64_t rx_l4_csum_errs;
257 	uint64_t rx_udp_len_errs;
258 	uint64_t rx_l4_port_errs;
259 	uint64_t rx_tcp_flag_errs;
260 	uint64_t rx_tcp_offset_errs;
261 	uint64_t rx_l4_pclp;
262 	uint64_t rx_truncated_pkts;
263 
264 	uint64_t tx_bytes_ok;
265 	uint64_t tx_ucast_frames_ok;
266 	uint64_t tx_bcast_frames_ok;
267 	uint64_t tx_mcast_frames_ok;
268 	uint64_t tx_drops;
269 };
270 
271 struct nicvf_drv_stats {
272 	/* Rx */
273 	uint64_t rx_frames_ok;
274 	uint64_t rx_frames_64;
275 	uint64_t rx_frames_127;
276 	uint64_t rx_frames_255;
277 	uint64_t rx_frames_511;
278 	uint64_t rx_frames_1023;
279 	uint64_t rx_frames_1518;
280 	uint64_t rx_frames_jumbo;
281 	uint64_t rx_drops;
282 
283 	/* Tx */
284 	uint64_t tx_frames_ok;
285 	uint64_t tx_drops;
286 	uint64_t tx_tso;
287 	uint64_t txq_stop;
288 	uint64_t txq_wake;
289 };
290 
291 struct nicvf {
292 	struct nicvf		*pnicvf;
293 	device_t		dev;
294 
295 	struct ifnet *		ifp;
296 	struct sx		core_sx;
297 	struct ifmedia		if_media;
298 	uint32_t		if_flags;
299 
300 	uint8_t			hwaddr[ETHER_ADDR_LEN];
301 	uint8_t			vf_id;
302 	uint8_t			node;
303 	boolean_t		tns_mode:1;
304 	boolean_t		sqs_mode:1;
305 	bool			loopback_supported:1;
306 	struct nicvf_rss_info	rss_info;
307 	uint16_t		mtu;
308 	struct queue_set	*qs;
309 	uint8_t			rx_queues;
310 	uint8_t			tx_queues;
311 	uint8_t			max_queues;
312 	struct resource		*reg_base;
313 	boolean_t		link_up;
314 	boolean_t		hw_tso;
315 	uint8_t			duplex;
316 	uint32_t		speed;
317 	uint8_t			cpi_alg;
318 	/* Interrupt coalescing settings */
319 	uint32_t		cq_coalesce_usecs;
320 
321 	uint32_t		msg_enable;
322 	struct nicvf_hw_stats	hw_stats;
323 	struct nicvf_drv_stats	drv_stats;
324 	struct bgx_stats	bgx_stats;
325 
326 	/* Interface statistics */
327 	struct callout		stats_callout;
328 	struct mtx		stats_mtx;
329 
330 	/* MSI-X  */
331 	boolean_t		msix_enabled;
332 	uint8_t			num_vec;
333 	struct msix_entry	msix_entries[NIC_VF_MSIX_VECTORS];
334 	struct resource *	msix_table_res;
335 	char			irq_name[NIC_VF_MSIX_VECTORS][20];
336 	boolean_t		irq_allocated[NIC_VF_MSIX_VECTORS];
337 
338 	/* VF <-> PF mailbox communication */
339 	boolean_t		pf_acked;
340 	boolean_t		pf_nacked;
341 } __aligned(CACHE_LINE_SIZE);
342 
343 /*
344  * PF <--> VF Mailbox communication
345  * Eight 64bit registers are shared between PF and VF.
346  * Separate set for each VF.
347  * Writing '1' into last register mbx7 means end of message.
348  */
349 
350 /* PF <--> VF mailbox communication */
351 #define	NIC_PF_VF_MAILBOX_SIZE		2
352 #define	NIC_MBOX_MSG_TIMEOUT		2000 /* ms */
353 
354 /* Mailbox message types */
355 #define	NIC_MBOX_MSG_READY		0x01	/* Is PF ready to rcv msgs */
356 #define	NIC_MBOX_MSG_ACK		0x02	/* ACK the message received */
357 #define	NIC_MBOX_MSG_NACK		0x03	/* NACK the message received */
358 #define	NIC_MBOX_MSG_QS_CFG		0x04	/* Configure Qset */
359 #define	NIC_MBOX_MSG_RQ_CFG		0x05	/* Configure receive queue */
360 #define	NIC_MBOX_MSG_SQ_CFG		0x06	/* Configure Send queue */
361 #define	NIC_MBOX_MSG_RQ_DROP_CFG	0x07	/* Configure receive queue */
362 #define	NIC_MBOX_MSG_SET_MAC		0x08	/* Add MAC ID to DMAC filter */
363 #define	NIC_MBOX_MSG_SET_MAX_FRS	0x09	/* Set max frame size */
364 #define	NIC_MBOX_MSG_CPI_CFG		0x0A	/* Config CPI, RSSI */
365 #define	NIC_MBOX_MSG_RSS_SIZE		0x0B	/* Get RSS indir_tbl size */
366 #define	NIC_MBOX_MSG_RSS_CFG		0x0C	/* Config RSS table */
367 #define	NIC_MBOX_MSG_RSS_CFG_CONT	0x0D	/* RSS config continuation */
368 #define	NIC_MBOX_MSG_RQ_BP_CFG		0x0E	/* RQ backpressure config */
369 #define	NIC_MBOX_MSG_RQ_SW_SYNC		0x0F	/* Flush inflight pkts to RQ */
370 #define	NIC_MBOX_MSG_BGX_STATS		0x10	/* Get stats from BGX */
371 #define	NIC_MBOX_MSG_BGX_LINK_CHANGE	0x11	/* BGX:LMAC link status */
372 #define	NIC_MBOX_MSG_ALLOC_SQS		0x12	/* Allocate secondary Qset */
373 #define	NIC_MBOX_MSG_NICVF_PTR		0x13	/* Send nicvf ptr to PF */
374 #define	NIC_MBOX_MSG_PNICVF_PTR		0x14	/* Get primary qset nicvf ptr */
375 #define	NIC_MBOX_MSG_SNICVF_PTR		0x15	/* Send sqet nicvf ptr to PVF */
376 #define	NIC_MBOX_MSG_LOOPBACK		0x16	/* Set interface in loopback */
377 #define	NIC_MBOX_MSG_CFG_DONE		0xF0	/* VF configuration done */
378 #define	NIC_MBOX_MSG_SHUTDOWN		0xF1	/* VF is being shutdown */
379 
380 struct nic_cfg_msg {
381 	uint8_t		msg;
382 	uint8_t		vf_id;
383 	uint8_t		node_id;
384 	boolean_t	tns_mode:1;
385 	boolean_t	sqs_mode:1;
386 	boolean_t	loopback_supported:1;
387 	uint8_t	mac_addr[ETHER_ADDR_LEN];
388 };
389 
390 /* Qset configuration */
391 struct qs_cfg_msg {
392 	uint8_t		msg;
393 	uint8_t		num;
394 	uint8_t		sqs_count;
395 	uint64_t	cfg;
396 };
397 
398 /* Receive queue configuration */
399 struct rq_cfg_msg {
400 	uint8_t		msg;
401 	uint8_t		qs_num;
402 	uint8_t		rq_num;
403 	uint64_t	cfg;
404 };
405 
406 /* Send queue configuration */
407 struct sq_cfg_msg {
408 	uint8_t		msg;
409 	uint8_t		qs_num;
410 	uint8_t		sq_num;
411 	boolean_t	sqs_mode;
412 	uint64_t	cfg;
413 };
414 
415 /* Set VF's MAC address */
416 struct set_mac_msg {
417 	uint8_t		msg;
418 	uint8_t		vf_id;
419 	uint8_t		mac_addr[ETHER_ADDR_LEN];
420 };
421 
422 /* Set Maximum frame size */
423 struct set_frs_msg {
424 	uint8_t		msg;
425 	uint8_t		vf_id;
426 	uint16_t	max_frs;
427 };
428 
429 /* Set CPI algorithm type */
430 struct cpi_cfg_msg {
431 	uint8_t		msg;
432 	uint8_t		vf_id;
433 	uint8_t		rq_cnt;
434 	uint8_t		cpi_alg;
435 };
436 
437 /* Get RSS table size */
438 struct rss_sz_msg {
439 	uint8_t		msg;
440 	uint8_t		vf_id;
441 	uint16_t	ind_tbl_size;
442 };
443 
444 /* Set RSS configuration */
445 struct rss_cfg_msg {
446 	uint8_t		msg;
447 	uint8_t		vf_id;
448 	uint8_t		hash_bits;
449 	uint8_t		tbl_len;
450 	uint8_t		tbl_offset;
451 #define	RSS_IND_TBL_LEN_PER_MBX_MSG	8
452 	uint8_t		ind_tbl[RSS_IND_TBL_LEN_PER_MBX_MSG];
453 };
454 
455 struct bgx_stats_msg {
456 	uint8_t		msg;
457 	uint8_t		vf_id;
458 	uint8_t		rx;
459 	uint8_t		idx;
460 	uint64_t	stats;
461 };
462 
463 /* Physical interface link status */
464 struct bgx_link_status {
465 	uint8_t		msg;
466 	uint8_t		link_up;
467 	uint8_t		duplex;
468 	uint32_t	speed;
469 };
470 
471 /* Set interface in loopback mode */
472 struct set_loopback {
473 	uint8_t		msg;
474 	uint8_t		vf_id;
475 	boolean_t	enable;
476 };
477 
478 /* 128 bit shared memory between PF and each VF */
479 union nic_mbx {
480 	struct {
481 		uint8_t msg;
482 	} msg;
483 	struct nic_cfg_msg	nic_cfg;
484 	struct qs_cfg_msg	qs;
485 	struct rq_cfg_msg	rq;
486 	struct sq_cfg_msg	sq;
487 	struct set_mac_msg	mac;
488 	struct set_frs_msg	frs;
489 	struct cpi_cfg_msg	cpi_cfg;
490 	struct rss_sz_msg	rss_size;
491 	struct rss_cfg_msg	rss_cfg;
492 	struct bgx_stats_msg	bgx_stats;
493 	struct bgx_link_status	link_status;
494 	struct set_loopback	lbk;
495 };
496 
497 #define	NIC_NODE_ID_MASK	0x03
498 #define	NIC_NODE_ID_SHIFT	44
499 
500 static __inline int
501 nic_get_node_id(struct resource *res)
502 {
503 	pci_addr_t addr;
504 
505 	addr = rman_get_start(res);
506 	return ((addr >> NIC_NODE_ID_SHIFT) & NIC_NODE_ID_MASK);
507 }
508 
509 static __inline boolean_t
510 pass1_silicon(device_t dev)
511 {
512 
513 	/* Check if the chip revision is < Pass2 */
514 	return (pci_get_revid(dev) < PCI_REVID_PASS2);
515 }
516 
517 int nicvf_send_msg_to_pf(struct nicvf *vf, union nic_mbx *mbx);
518 
519 #endif /* NIC_H */
520