xref: /freebsd/sys/dev/vnic/nic_main.c (revision 8191a879590630c1e82d132fc8f013ed35485aed)
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 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/bitset.h>
36 #include <sys/bitstring.h>
37 #include <sys/bus.h>
38 #include <sys/endian.h>
39 #include <sys/kernel.h>
40 #include <sys/malloc.h>
41 #include <sys/module.h>
42 #include <sys/rman.h>
43 #include <sys/pciio.h>
44 #include <sys/pcpu.h>
45 #include <sys/proc.h>
46 #include <sys/socket.h>
47 #include <sys/sockio.h>
48 #include <sys/cpuset.h>
49 #include <sys/lock.h>
50 #include <sys/mutex.h>
51 
52 #include <net/ethernet.h>
53 #include <net/if.h>
54 #include <net/if_media.h>
55 
56 #include <machine/bus.h>
57 #include <machine/_inttypes.h>
58 
59 #include <dev/pci/pcireg.h>
60 #include <dev/pci/pcivar.h>
61 
62 #include <sys/dnv.h>
63 #include <sys/nv.h>
64 #ifdef PCI_IOV
65 #include <sys/iov_schema.h>
66 #include <dev/pci/pci_iov.h>
67 #endif
68 
69 #include "thunder_bgx.h"
70 #include "nic_reg.h"
71 #include "nic.h"
72 #include "q_struct.h"
73 
74 #define	VNIC_PF_DEVSTR		"Cavium Thunder NIC Physical Function Driver"
75 
76 #define	VNIC_PF_REG_RID		PCIR_BAR(PCI_CFG_REG_BAR_NUM)
77 
78 #define	NIC_SET_VF_LMAC_MAP(bgx, lmac)		((((bgx) & 0xF) << 4) | ((lmac) & 0xF))
79 #define	NIC_GET_BGX_FROM_VF_LMAC_MAP(map)	(((map) >> 4) & 0xF)
80 #define	NIC_GET_LMAC_FROM_VF_LMAC_MAP(map)	((map) & 0xF)
81 
82 /* Structure to be used by the SR-IOV for VF configuration schemas */
83 struct nicvf_info {
84 	boolean_t		vf_enabled;
85 	int			vf_flags;
86 };
87 
88 struct nicpf {
89 	device_t		dev;
90 	uint8_t			node;
91 	u_int			flags;
92 	uint8_t			num_vf_en;      /* No of VF enabled */
93 	struct nicvf_info	vf_info[MAX_NUM_VFS_SUPPORTED];
94 	struct resource *	reg_base;       /* Register start address */
95 	struct pkind_cfg	pkind;
96 	uint8_t			vf_lmac_map[MAX_LMAC];
97 	boolean_t		mbx_lock[MAX_NUM_VFS_SUPPORTED];
98 
99 	struct callout		check_link;
100 	struct mtx		check_link_mtx;
101 
102 	uint8_t			link[MAX_LMAC];
103 	uint8_t			duplex[MAX_LMAC];
104 	uint32_t		speed[MAX_LMAC];
105 	uint16_t		cpi_base[MAX_NUM_VFS_SUPPORTED];
106 	uint16_t		rssi_base[MAX_NUM_VFS_SUPPORTED];
107 	uint16_t		rss_ind_tbl_size;
108 
109 	/* MSI-X */
110 	boolean_t		msix_enabled;
111 	uint8_t			num_vec;
112 	struct msix_entry	msix_entries[NIC_PF_MSIX_VECTORS];
113 	struct resource *	msix_table_res;
114 };
115 
116 static int nicpf_probe(device_t);
117 static int nicpf_attach(device_t);
118 static int nicpf_detach(device_t);
119 
120 #ifdef PCI_IOV
121 static int nicpf_iov_init(device_t, uint16_t, const nvlist_t *);
122 static void nicpf_iov_uninit(device_t);
123 static int nicpf_iov_add_vf(device_t, uint16_t, const nvlist_t *);
124 #endif
125 
126 static device_method_t nicpf_methods[] = {
127 	/* Device interface */
128 	DEVMETHOD(device_probe,		nicpf_probe),
129 	DEVMETHOD(device_attach,	nicpf_attach),
130 	DEVMETHOD(device_detach,	nicpf_detach),
131 	/* PCI SR-IOV interface */
132 #ifdef PCI_IOV
133 	DEVMETHOD(pci_iov_init,		nicpf_iov_init),
134 	DEVMETHOD(pci_iov_uninit,	nicpf_iov_uninit),
135 	DEVMETHOD(pci_iov_add_vf,	nicpf_iov_add_vf),
136 #endif
137 	DEVMETHOD_END,
138 };
139 
140 static driver_t nicpf_driver = {
141 	"vnicpf",
142 	nicpf_methods,
143 	sizeof(struct nicpf),
144 };
145 
146 static devclass_t nicpf_devclass;
147 
148 DRIVER_MODULE(nicpf, pci, nicpf_driver, nicpf_devclass, 0, 0);
149 MODULE_DEPEND(nicpf, pci, 1, 1, 1);
150 MODULE_DEPEND(nicpf, ether, 1, 1, 1);
151 MODULE_DEPEND(nicpf, thunder_bgx, 1, 1, 1);
152 
153 static int nicpf_alloc_res(struct nicpf *);
154 static void nicpf_free_res(struct nicpf *);
155 static void nic_set_lmac_vf_mapping(struct nicpf *);
156 static void nic_init_hw(struct nicpf *);
157 static int nic_sriov_init(device_t, struct nicpf *);
158 static void nic_poll_for_link(void *);
159 static int nic_register_interrupts(struct nicpf *);
160 static void nic_unregister_interrupts(struct nicpf *);
161 
162 /*
163  * Device interface
164  */
165 static int
166 nicpf_probe(device_t dev)
167 {
168 	uint16_t vendor_id;
169 	uint16_t device_id;
170 
171 	vendor_id = pci_get_vendor(dev);
172 	device_id = pci_get_device(dev);
173 
174 	if (vendor_id == PCI_VENDOR_ID_CAVIUM &&
175 	    device_id == PCI_DEVICE_ID_THUNDER_NIC_PF) {
176 		device_set_desc(dev, VNIC_PF_DEVSTR);
177 		return (BUS_PROBE_DEFAULT);
178 	}
179 
180 	return (ENXIO);
181 }
182 
183 static int
184 nicpf_attach(device_t dev)
185 {
186 	struct nicpf *nic;
187 	int err;
188 
189 	nic = device_get_softc(dev);
190 	nic->dev = dev;
191 
192 	/* Enable bus mastering */
193 	pci_enable_busmaster(dev);
194 
195 	/* Allocate PCI resources */
196 	err = nicpf_alloc_res(nic);
197 	if (err != 0) {
198 		device_printf(dev, "Could not allocate PCI resources\n");
199 		return (err);
200 	}
201 
202 	nic->node = nic_get_node_id(nic->reg_base);
203 
204 	/* Enable Traffic Network Switch (TNS) bypass mode by default */
205 	nic->flags &= ~NIC_TNS_ENABLED;
206 	nic_set_lmac_vf_mapping(nic);
207 
208 	/* Initialize hardware */
209 	nic_init_hw(nic);
210 
211 	/* Set RSS TBL size for each VF */
212 	nic->rss_ind_tbl_size = NIC_MAX_RSS_IDR_TBL_SIZE;
213 
214 	/* Setup interrupts */
215 	err = nic_register_interrupts(nic);
216 	if (err != 0)
217 		goto err_free_res;
218 
219 	/* Configure SRIOV */
220 	err = nic_sriov_init(dev, nic);
221 	if (err != 0)
222 		goto err_free_intr;
223 
224 	if (nic->flags & NIC_TNS_ENABLED)
225 		return (0);
226 
227 	mtx_init(&nic->check_link_mtx, "VNIC PF link poll", NULL, MTX_DEF);
228 	/* Register physical link status poll callout */
229 	callout_init_mtx(&nic->check_link, &nic->check_link_mtx, 0);
230 	mtx_lock(&nic->check_link_mtx);
231 	nic_poll_for_link(nic);
232 	mtx_unlock(&nic->check_link_mtx);
233 
234 	return (0);
235 
236 err_free_intr:
237 	nic_unregister_interrupts(nic);
238 err_free_res:
239 	nicpf_free_res(nic);
240 	pci_disable_busmaster(dev);
241 
242 	return (err);
243 }
244 
245 static int
246 nicpf_detach(device_t dev)
247 {
248 	struct nicpf *nic;
249 
250 	nic = device_get_softc(dev);
251 
252 	callout_drain(&nic->check_link);
253 	mtx_destroy(&nic->check_link_mtx);
254 
255 	nic_unregister_interrupts(nic);
256 	nicpf_free_res(nic);
257 	pci_disable_busmaster(dev);
258 
259 	return (0);
260 }
261 
262 /*
263  * SR-IOV interface
264  */
265 #ifdef PCI_IOV
266 static int
267 nicpf_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
268 {
269 	struct nicpf *nic;
270 
271 	nic = device_get_softc(dev);
272 
273 	if (num_vfs == 0)
274 		return (ENXIO);
275 
276 	nic->flags |= NIC_SRIOV_ENABLED;
277 
278 	return (0);
279 }
280 
281 static void
282 nicpf_iov_uninit(device_t dev)
283 {
284 
285 	/* ARM64TODO: Implement this function */
286 }
287 
288 static int
289 nicpf_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
290 {
291 	const void *mac;
292 	struct nicpf *nic;
293 	size_t size;
294 	int bgx, lmac;
295 
296 	nic = device_get_softc(dev);
297 
298 	if ((nic->flags & NIC_SRIOV_ENABLED) == 0)
299 		return (ENXIO);
300 
301 	if (vfnum > (nic->num_vf_en - 1))
302 		return (EINVAL);
303 
304 	if (nvlist_exists_binary(params, "mac-addr") != 0) {
305 		mac = nvlist_get_binary(params, "mac-addr", &size);
306 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
307 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
308 		bgx_set_lmac_mac(nic->node, bgx, lmac, mac);
309 	}
310 
311 	return (0);
312 }
313 #endif
314 
315 /*
316  * Helper routines
317  */
318 static int
319 nicpf_alloc_res(struct nicpf *nic)
320 {
321 	device_t dev;
322 	int rid;
323 
324 	dev = nic->dev;
325 
326 	rid = VNIC_PF_REG_RID;
327 	nic->reg_base = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
328 	    RF_ACTIVE);
329 	if (nic->reg_base == NULL) {
330 		/* For verbose output print some more details */
331 		if (bootverbose) {
332 			device_printf(dev,
333 			    "Could not allocate registers memory\n");
334 		}
335 		return (ENXIO);
336 	}
337 
338 	return (0);
339 }
340 
341 static void
342 nicpf_free_res(struct nicpf *nic)
343 {
344 	device_t dev;
345 
346 	dev = nic->dev;
347 
348 	if (nic->reg_base != NULL) {
349 		bus_release_resource(dev, SYS_RES_MEMORY,
350 		    rman_get_rid(nic->reg_base), nic->reg_base);
351 	}
352 }
353 
354 /* Register read/write APIs */
355 static __inline void
356 nic_reg_write(struct nicpf *nic, bus_space_handle_t offset,
357     uint64_t val)
358 {
359 
360 	bus_write_8(nic->reg_base, offset, val);
361 }
362 
363 static __inline uint64_t
364 nic_reg_read(struct nicpf *nic, uint64_t offset)
365 {
366 	uint64_t val;
367 
368 	val = bus_read_8(nic->reg_base, offset);
369 	return (val);
370 }
371 
372 /* PF -> VF mailbox communication APIs */
373 static void
374 nic_enable_mbx_intr(struct nicpf *nic)
375 {
376 
377 	/* Enable mailbox interrupt for all 128 VFs */
378 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S, ~0UL);
379 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S + sizeof(uint64_t), ~0UL);
380 }
381 
382 static void
383 nic_clear_mbx_intr(struct nicpf *nic, int vf, int mbx_reg)
384 {
385 
386 	nic_reg_write(nic, NIC_PF_MAILBOX_INT + (mbx_reg << 3), (1UL << vf));
387 }
388 
389 static uint64_t
390 nic_get_mbx_addr(int vf)
391 {
392 
393 	return (NIC_PF_VF_0_127_MAILBOX_0_1 + (vf << NIC_VF_NUM_SHIFT));
394 }
395 
396 /*
397  * Send a mailbox message to VF
398  * @vf: vf to which this message to be sent
399  * @mbx: Message to be sent
400  */
401 static void
402 nic_send_msg_to_vf(struct nicpf *nic, int vf, union nic_mbx *mbx)
403 {
404 	bus_space_handle_t mbx_addr = nic_get_mbx_addr(vf);
405 	uint64_t *msg = (uint64_t *)mbx;
406 
407 	/*
408 	 * In first revision HW, mbox interrupt is triggerred
409 	 * when PF writes to MBOX(1), in next revisions when
410 	 * PF writes to MBOX(0)
411 	 */
412 	if (pass1_silicon(nic->dev)) {
413 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
414 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
415 	} else {
416 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
417 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
418 	}
419 }
420 
421 /*
422  * Responds to VF's READY message with VF's
423  * ID, node, MAC address e.t.c
424  * @vf: VF which sent READY message
425  */
426 static void
427 nic_mbx_send_ready(struct nicpf *nic, int vf)
428 {
429 	union nic_mbx mbx = {};
430 	int bgx_idx, lmac;
431 	const char *mac;
432 
433 	mbx.nic_cfg.msg = NIC_MBOX_MSG_READY;
434 	mbx.nic_cfg.vf_id = vf;
435 
436 	if (nic->flags & NIC_TNS_ENABLED)
437 		mbx.nic_cfg.tns_mode = NIC_TNS_MODE;
438 	else
439 		mbx.nic_cfg.tns_mode = NIC_TNS_BYPASS_MODE;
440 
441 	if (vf < MAX_LMAC) {
442 		bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
443 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
444 
445 		mac = bgx_get_lmac_mac(nic->node, bgx_idx, lmac);
446 		if (mac) {
447 			memcpy((uint8_t *)&mbx.nic_cfg.mac_addr, mac,
448 			    ETHER_ADDR_LEN);
449 		}
450 	}
451 	mbx.nic_cfg.node_id = nic->node;
452 
453 	mbx.nic_cfg.loopback_supported = vf < MAX_LMAC;
454 
455 	nic_send_msg_to_vf(nic, vf, &mbx);
456 }
457 
458 /*
459  * ACKs VF's mailbox message
460  * @vf: VF to which ACK to be sent
461  */
462 static void
463 nic_mbx_send_ack(struct nicpf *nic, int vf)
464 {
465 	union nic_mbx mbx = {};
466 
467 	mbx.msg.msg = NIC_MBOX_MSG_ACK;
468 	nic_send_msg_to_vf(nic, vf, &mbx);
469 }
470 
471 /*
472  * NACKs VF's mailbox message that PF is not able to
473  * complete the action
474  * @vf: VF to which ACK to be sent
475  */
476 static void
477 nic_mbx_send_nack(struct nicpf *nic, int vf)
478 {
479 	union nic_mbx mbx = {};
480 
481 	mbx.msg.msg = NIC_MBOX_MSG_NACK;
482 	nic_send_msg_to_vf(nic, vf, &mbx);
483 }
484 
485 /*
486  * Flush all in flight receive packets to memory and
487  * bring down an active RQ
488  */
489 static int
490 nic_rcv_queue_sw_sync(struct nicpf *nic)
491 {
492 	uint16_t timeout = ~0x00;
493 
494 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x01);
495 	/* Wait till sync cycle is finished */
496 	while (timeout) {
497 		if (nic_reg_read(nic, NIC_PF_SW_SYNC_RX_DONE) & 0x1)
498 			break;
499 		timeout--;
500 	}
501 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x00);
502 	if (!timeout) {
503 		device_printf(nic->dev, "Receive queue software sync failed\n");
504 		return (ETIMEDOUT);
505 	}
506 	return (0);
507 }
508 
509 /* Get BGX Rx/Tx stats and respond to VF's request */
510 static void
511 nic_get_bgx_stats(struct nicpf *nic, struct bgx_stats_msg *bgx)
512 {
513 	int bgx_idx, lmac;
514 	union nic_mbx mbx = {};
515 
516 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
517 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
518 
519 	mbx.bgx_stats.msg = NIC_MBOX_MSG_BGX_STATS;
520 	mbx.bgx_stats.vf_id = bgx->vf_id;
521 	mbx.bgx_stats.rx = bgx->rx;
522 	mbx.bgx_stats.idx = bgx->idx;
523 	if (bgx->rx != 0) {
524 		mbx.bgx_stats.stats =
525 		    bgx_get_rx_stats(nic->node, bgx_idx, lmac, bgx->idx);
526 	} else {
527 		mbx.bgx_stats.stats =
528 		    bgx_get_tx_stats(nic->node, bgx_idx, lmac, bgx->idx);
529 	}
530 	nic_send_msg_to_vf(nic, bgx->vf_id, &mbx);
531 }
532 
533 /* Update hardware min/max frame size */
534 static int
535 nic_update_hw_frs(struct nicpf *nic, int new_frs, int vf)
536 {
537 
538 	if ((new_frs > NIC_HW_MAX_FRS) || (new_frs < NIC_HW_MIN_FRS)) {
539 		device_printf(nic->dev,
540 		    "Invalid MTU setting from VF%d rejected, "
541 		    "should be between %d and %d\n",
542 		    vf, NIC_HW_MIN_FRS, NIC_HW_MAX_FRS);
543 		return (EINVAL);
544 	}
545 	new_frs += ETHER_HDR_LEN;
546 	if (new_frs <= nic->pkind.maxlen)
547 		return (0);
548 
549 	nic->pkind.maxlen = new_frs;
550 	nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG, *(uint64_t *)&nic->pkind);
551 	return (0);
552 }
553 
554 /* Set minimum transmit packet size */
555 static void
556 nic_set_tx_pkt_pad(struct nicpf *nic, int size)
557 {
558 	int lmac;
559 	uint64_t lmac_cfg;
560 
561 	/* Max value that can be set is 60 */
562 	if (size > 60)
563 		size = 60;
564 
565 	for (lmac = 0; lmac < (MAX_BGX_PER_CN88XX * MAX_LMAC_PER_BGX); lmac++) {
566 		lmac_cfg = nic_reg_read(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3));
567 		lmac_cfg &= ~(0xF << 2);
568 		lmac_cfg |= ((size / 4) << 2);
569 		nic_reg_write(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3), lmac_cfg);
570 	}
571 }
572 
573 /*
574  * Function to check number of LMACs present and set VF::LMAC mapping.
575  * Mapping will be used while initializing channels.
576  */
577 static void
578 nic_set_lmac_vf_mapping(struct nicpf *nic)
579 {
580 	unsigned bgx_map = bgx_get_map(nic->node);
581 	int bgx, next_bgx_lmac = 0;
582 	int lmac, lmac_cnt = 0;
583 	uint64_t lmac_credit;
584 
585 	nic->num_vf_en = 0;
586 	if (nic->flags & NIC_TNS_ENABLED) {
587 		nic->num_vf_en = DEFAULT_NUM_VF_ENABLED;
588 		return;
589 	}
590 
591 	for (bgx = 0; bgx < NIC_MAX_BGX; bgx++) {
592 		if ((bgx_map & (1 << bgx)) == 0)
593 			continue;
594 		lmac_cnt = bgx_get_lmac_count(nic->node, bgx);
595 		for (lmac = 0; lmac < lmac_cnt; lmac++)
596 			nic->vf_lmac_map[next_bgx_lmac++] =
597 						NIC_SET_VF_LMAC_MAP(bgx, lmac);
598 		nic->num_vf_en += lmac_cnt;
599 
600 		/* Program LMAC credits */
601 		lmac_credit = (1UL << 1); /* channel credit enable */
602 		lmac_credit |= (0x1ff << 2); /* Max outstanding pkt count */
603 		/* 48KB BGX Tx buffer size, each unit is of size 16bytes */
604 		lmac_credit |= (((((48 * 1024) / lmac_cnt) -
605 		    NIC_HW_MAX_FRS) / 16) << 12);
606 		lmac = bgx * MAX_LMAC_PER_BGX;
607 		for (; lmac < lmac_cnt + (bgx * MAX_LMAC_PER_BGX); lmac++) {
608 			nic_reg_write(nic, NIC_PF_LMAC_0_7_CREDIT + (lmac * 8),
609 			    lmac_credit);
610 		}
611 	}
612 }
613 
614 #define TNS_PORT0_BLOCK 6
615 #define TNS_PORT1_BLOCK 7
616 #define BGX0_BLOCK 8
617 #define BGX1_BLOCK 9
618 
619 static void
620 nic_init_hw(struct nicpf *nic)
621 {
622 	int i;
623 
624 	/* Enable NIC HW block */
625 	nic_reg_write(nic, NIC_PF_CFG, 0x3);
626 
627 	/* Enable backpressure */
628 	nic_reg_write(nic, NIC_PF_BP_CFG, (1UL << 6) | 0x03);
629 
630 	if (nic->flags & NIC_TNS_ENABLED) {
631 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
632 		    (NIC_TNS_MODE << 7) | TNS_PORT0_BLOCK);
633 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
634 		    (NIC_TNS_MODE << 7) | TNS_PORT1_BLOCK);
635 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
636 		    (1UL << 63) | TNS_PORT0_BLOCK);
637 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
638 		    (1UL << 63) | TNS_PORT1_BLOCK);
639 
640 	} else {
641 		/* Disable TNS mode on both interfaces */
642 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
643 		    (NIC_TNS_BYPASS_MODE << 7) | BGX0_BLOCK);
644 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
645 		    (NIC_TNS_BYPASS_MODE << 7) | BGX1_BLOCK);
646 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
647 		    (1UL << 63) | BGX0_BLOCK);
648 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
649 		    (1UL << 63) | BGX1_BLOCK);
650 	}
651 
652 	/* PKIND configuration */
653 	nic->pkind.minlen = 0;
654 	nic->pkind.maxlen = NIC_HW_MAX_FRS + ETHER_HDR_LEN;
655 	nic->pkind.lenerr_en = 1;
656 	nic->pkind.rx_hdr = 0;
657 	nic->pkind.hdr_sl = 0;
658 
659 	for (i = 0; i < NIC_MAX_PKIND; i++) {
660 		nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG | (i << 3),
661 		    *(uint64_t *)&nic->pkind);
662 	}
663 
664 	nic_set_tx_pkt_pad(nic, NIC_HW_MIN_FRS);
665 
666 	/* Timer config */
667 	nic_reg_write(nic, NIC_PF_INTR_TIMER_CFG, NICPF_CLK_PER_INT_TICK);
668 
669 	/* Enable VLAN ethertype matching and stripping */
670 	nic_reg_write(nic, NIC_PF_RX_ETYPE_0_7,
671 	    (2 << 19) | (ETYPE_ALG_VLAN_STRIP << 16) | ETHERTYPE_VLAN);
672 }
673 
674 /* Channel parse index configuration */
675 static void
676 nic_config_cpi(struct nicpf *nic, struct cpi_cfg_msg *cfg)
677 {
678 	uint32_t vnic, bgx, lmac, chan;
679 	uint32_t padd, cpi_count = 0;
680 	uint64_t cpi_base, cpi, rssi_base, rssi;
681 	uint8_t qset, rq_idx = 0;
682 
683 	vnic = cfg->vf_id;
684 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
685 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
686 
687 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
688 	cpi_base = (lmac * NIC_MAX_CPI_PER_LMAC) + (bgx * NIC_CPI_PER_BGX);
689 	rssi_base = (lmac * nic->rss_ind_tbl_size) + (bgx * NIC_RSSI_PER_BGX);
690 
691 	/* Rx channel configuration */
692 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_BP_CFG | (chan << 3),
693 	    (1UL << 63) | (vnic << 0));
694 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_CFG | (chan << 3),
695 	    ((uint64_t)cfg->cpi_alg << 62) | (cpi_base << 48));
696 
697 	if (cfg->cpi_alg == CPI_ALG_NONE)
698 		cpi_count = 1;
699 	else if (cfg->cpi_alg == CPI_ALG_VLAN) /* 3 bits of PCP */
700 		cpi_count = 8;
701 	else if (cfg->cpi_alg == CPI_ALG_VLAN16) /* 3 bits PCP + DEI */
702 		cpi_count = 16;
703 	else if (cfg->cpi_alg == CPI_ALG_DIFF) /* 6bits DSCP */
704 		cpi_count = NIC_MAX_CPI_PER_LMAC;
705 
706 	/* RSS Qset, Qidx mapping */
707 	qset = cfg->vf_id;
708 	rssi = rssi_base;
709 	for (; rssi < (rssi_base + cfg->rq_cnt); rssi++) {
710 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
711 		    (qset << 3) | rq_idx);
712 		rq_idx++;
713 	}
714 
715 	rssi = 0;
716 	cpi = cpi_base;
717 	for (; cpi < (cpi_base + cpi_count); cpi++) {
718 		/* Determine port to channel adder */
719 		if (cfg->cpi_alg != CPI_ALG_DIFF)
720 			padd = cpi % cpi_count;
721 		else
722 			padd = cpi % 8; /* 3 bits CS out of 6bits DSCP */
723 
724 		/* Leave RSS_SIZE as '0' to disable RSS */
725 		if (pass1_silicon(nic->dev)) {
726 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
727 			    (vnic << 24) | (padd << 16) | (rssi_base + rssi));
728 		} else {
729 			/* Set MPI_ALG to '0' to disable MCAM parsing */
730 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
731 			    (padd << 16));
732 			/* MPI index is same as CPI if MPI_ALG is not enabled */
733 			nic_reg_write(nic, NIC_PF_MPI_0_2047_CFG | (cpi << 3),
734 			    (vnic << 24) | (rssi_base + rssi));
735 		}
736 
737 		if ((rssi + 1) >= cfg->rq_cnt)
738 			continue;
739 
740 		if (cfg->cpi_alg == CPI_ALG_VLAN)
741 			rssi++;
742 		else if (cfg->cpi_alg == CPI_ALG_VLAN16)
743 			rssi = ((cpi - cpi_base) & 0xe) >> 1;
744 		else if (cfg->cpi_alg == CPI_ALG_DIFF)
745 			rssi = ((cpi - cpi_base) & 0x38) >> 3;
746 	}
747 	nic->cpi_base[cfg->vf_id] = cpi_base;
748 	nic->rssi_base[cfg->vf_id] = rssi_base;
749 }
750 
751 /* Responsds to VF with its RSS indirection table size */
752 static void
753 nic_send_rss_size(struct nicpf *nic, int vf)
754 {
755 	union nic_mbx mbx = {};
756 	uint64_t  *msg;
757 
758 	msg = (uint64_t *)&mbx;
759 
760 	mbx.rss_size.msg = NIC_MBOX_MSG_RSS_SIZE;
761 	mbx.rss_size.ind_tbl_size = nic->rss_ind_tbl_size;
762 	nic_send_msg_to_vf(nic, vf, &mbx);
763 }
764 
765 /*
766  * Receive side scaling configuration
767  * configure:
768  * - RSS index
769  * - indir table i.e hash::RQ mapping
770  * - no of hash bits to consider
771  */
772 static void
773 nic_config_rss(struct nicpf *nic, struct rss_cfg_msg *cfg)
774 {
775 	uint8_t qset, idx;
776 	uint64_t cpi_cfg, cpi_base, rssi_base, rssi;
777 	uint64_t idx_addr;
778 
779 	idx = 0;
780 	rssi_base = nic->rssi_base[cfg->vf_id] + cfg->tbl_offset;
781 
782 	rssi = rssi_base;
783 	qset = cfg->vf_id;
784 
785 	for (; rssi < (rssi_base + cfg->tbl_len); rssi++) {
786 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
787 		    (qset << 3) | (cfg->ind_tbl[idx] & 0x7));
788 		idx++;
789 	}
790 
791 	cpi_base = nic->cpi_base[cfg->vf_id];
792 	if (pass1_silicon(nic->dev))
793 		idx_addr = NIC_PF_CPI_0_2047_CFG;
794 	else
795 		idx_addr = NIC_PF_MPI_0_2047_CFG;
796 	cpi_cfg = nic_reg_read(nic, idx_addr | (cpi_base << 3));
797 	cpi_cfg &= ~(0xFUL << 20);
798 	cpi_cfg |= (cfg->hash_bits << 20);
799 	nic_reg_write(nic, idx_addr | (cpi_base << 3), cpi_cfg);
800 }
801 
802 /*
803  * 4 level transmit side scheduler configutation
804  * for TNS bypass mode
805  *
806  * Sample configuration for SQ0
807  * VNIC0-SQ0 -> TL4(0)   -> TL3[0]   -> TL2[0]  -> TL1[0] -> BGX0
808  * VNIC1-SQ0 -> TL4(8)   -> TL3[2]   -> TL2[0]  -> TL1[0] -> BGX0
809  * VNIC2-SQ0 -> TL4(16)  -> TL3[4]   -> TL2[1]  -> TL1[0] -> BGX0
810  * VNIC3-SQ0 -> TL4(24)  -> TL3[6]   -> TL2[1]  -> TL1[0] -> BGX0
811  * VNIC4-SQ0 -> TL4(512) -> TL3[128] -> TL2[32] -> TL1[1] -> BGX1
812  * VNIC5-SQ0 -> TL4(520) -> TL3[130] -> TL2[32] -> TL1[1] -> BGX1
813  * VNIC6-SQ0 -> TL4(528) -> TL3[132] -> TL2[33] -> TL1[1] -> BGX1
814  * VNIC7-SQ0 -> TL4(536) -> TL3[134] -> TL2[33] -> TL1[1] -> BGX1
815  */
816 static void
817 nic_tx_channel_cfg(struct nicpf *nic, uint8_t vnic, struct sq_cfg_msg *sq)
818 {
819 	uint32_t bgx, lmac, chan;
820 	uint32_t tl2, tl3, tl4;
821 	uint32_t rr_quantum;
822 	uint8_t sq_idx = sq->sq_num;
823 	uint8_t pqs_vnic;
824 
825 	pqs_vnic = vnic;
826 
827 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
828 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
829 
830 	/* 24 bytes for FCS, IPG and preamble */
831 	rr_quantum = ((NIC_HW_MAX_FRS + 24) / 4);
832 
833 	tl4 = (lmac * NIC_TL4_PER_LMAC) + (bgx * NIC_TL4_PER_BGX);
834 	tl4 += sq_idx;
835 
836 	tl3 = tl4 / (NIC_MAX_TL4 / NIC_MAX_TL3);
837 	nic_reg_write(nic, NIC_PF_QSET_0_127_SQ_0_7_CFG2 |
838 	    ((uint64_t)vnic << NIC_QS_ID_SHIFT) |
839 	    ((uint32_t)sq_idx << NIC_Q_NUM_SHIFT), tl4);
840 	nic_reg_write(nic, NIC_PF_TL4_0_1023_CFG | (tl4 << 3),
841 	    ((uint64_t)vnic << 27) | ((uint32_t)sq_idx << 24) | rr_quantum);
842 
843 	nic_reg_write(nic, NIC_PF_TL3_0_255_CFG | (tl3 << 3), rr_quantum);
844 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
845 	nic_reg_write(nic, NIC_PF_TL3_0_255_CHAN | (tl3 << 3), chan);
846 	/* Enable backpressure on the channel */
847 	nic_reg_write(nic, NIC_PF_CHAN_0_255_TX_CFG | (chan << 3), 1);
848 
849 	tl2 = tl3 >> 2;
850 	nic_reg_write(nic, NIC_PF_TL3A_0_63_CFG | (tl2 << 3), tl2);
851 	nic_reg_write(nic, NIC_PF_TL2_0_63_CFG | (tl2 << 3), rr_quantum);
852 	/* No priorities as of now */
853 	nic_reg_write(nic, NIC_PF_TL2_0_63_PRI | (tl2 << 3), 0x00);
854 }
855 
856 static int
857 nic_config_loopback(struct nicpf *nic, struct set_loopback *lbk)
858 {
859 	int bgx_idx, lmac_idx;
860 
861 	if (lbk->vf_id > MAX_LMAC)
862 		return (ENXIO);
863 
864 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
865 	lmac_idx = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
866 
867 	bgx_lmac_internal_loopback(nic->node, bgx_idx, lmac_idx, lbk->enable);
868 
869 	return (0);
870 }
871 
872 /* Interrupt handler to handle mailbox messages from VFs */
873 static void
874 nic_handle_mbx_intr(struct nicpf *nic, int vf)
875 {
876 	union nic_mbx mbx = {};
877 	uint64_t *mbx_data;
878 	uint64_t mbx_addr;
879 	uint64_t reg_addr;
880 	uint64_t cfg;
881 	int bgx, lmac;
882 	int i;
883 	int ret = 0;
884 
885 	nic->mbx_lock[vf] = TRUE;
886 
887 	mbx_addr = nic_get_mbx_addr(vf);
888 	mbx_data = (uint64_t *)&mbx;
889 
890 	for (i = 0; i < NIC_PF_VF_MAILBOX_SIZE; i++) {
891 		*mbx_data = nic_reg_read(nic, mbx_addr);
892 		mbx_data++;
893 		mbx_addr += sizeof(uint64_t);
894 	}
895 
896 	switch (mbx.msg.msg) {
897 	case NIC_MBOX_MSG_READY:
898 		nic_mbx_send_ready(nic, vf);
899 		if (vf < MAX_LMAC) {
900 			nic->link[vf] = 0;
901 			nic->duplex[vf] = 0;
902 			nic->speed[vf] = 0;
903 		}
904 		ret = 1;
905 		break;
906 	case NIC_MBOX_MSG_QS_CFG:
907 		reg_addr = NIC_PF_QSET_0_127_CFG |
908 		    (mbx.qs.num << NIC_QS_ID_SHIFT);
909 		cfg = mbx.qs.cfg;
910 		nic_reg_write(nic, reg_addr, cfg);
911 		break;
912 	case NIC_MBOX_MSG_RQ_CFG:
913 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_CFG |
914 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
915 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
916 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
917 		break;
918 	case NIC_MBOX_MSG_RQ_BP_CFG:
919 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_BP_CFG |
920 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
921 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
922 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
923 		break;
924 	case NIC_MBOX_MSG_RQ_SW_SYNC:
925 		ret = nic_rcv_queue_sw_sync(nic);
926 		break;
927 	case NIC_MBOX_MSG_RQ_DROP_CFG:
928 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_DROP_CFG |
929 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
930 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
931 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
932 		break;
933 	case NIC_MBOX_MSG_SQ_CFG:
934 		reg_addr = NIC_PF_QSET_0_127_SQ_0_7_CFG |
935 		    (mbx.sq.qs_num << NIC_QS_ID_SHIFT) |
936 		    (mbx.sq.sq_num << NIC_Q_NUM_SHIFT);
937 		nic_reg_write(nic, reg_addr, mbx.sq.cfg);
938 		nic_tx_channel_cfg(nic, mbx.qs.num, &mbx.sq);
939 		break;
940 	case NIC_MBOX_MSG_SET_MAC:
941 		lmac = mbx.mac.vf_id;
942 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
943 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
944 		bgx_set_lmac_mac(nic->node, bgx, lmac, mbx.mac.mac_addr);
945 		break;
946 	case NIC_MBOX_MSG_SET_MAX_FRS:
947 		ret = nic_update_hw_frs(nic, mbx.frs.max_frs, mbx.frs.vf_id);
948 		break;
949 	case NIC_MBOX_MSG_CPI_CFG:
950 		nic_config_cpi(nic, &mbx.cpi_cfg);
951 		break;
952 	case NIC_MBOX_MSG_RSS_SIZE:
953 		nic_send_rss_size(nic, vf);
954 		goto unlock;
955 	case NIC_MBOX_MSG_RSS_CFG:
956 	case NIC_MBOX_MSG_RSS_CFG_CONT: /* fall through */
957 		nic_config_rss(nic, &mbx.rss_cfg);
958 		break;
959 	case NIC_MBOX_MSG_CFG_DONE:
960 		/* Last message of VF config msg sequence */
961 		nic->vf_info[vf].vf_enabled = TRUE;
962 		goto unlock;
963 	case NIC_MBOX_MSG_SHUTDOWN:
964 		/* First msg in VF teardown sequence */
965 		nic->vf_info[vf].vf_enabled = FALSE;
966 		break;
967 	case NIC_MBOX_MSG_BGX_STATS:
968 		nic_get_bgx_stats(nic, &mbx.bgx_stats);
969 		goto unlock;
970 	case NIC_MBOX_MSG_LOOPBACK:
971 		ret = nic_config_loopback(nic, &mbx.lbk);
972 		break;
973 	default:
974 		device_printf(nic->dev,
975 		    "Invalid msg from VF%d, msg 0x%x\n", vf, mbx.msg.msg);
976 		break;
977 	}
978 
979 	if (ret == 0)
980 		nic_mbx_send_ack(nic, vf);
981 	else if (mbx.msg.msg != NIC_MBOX_MSG_READY)
982 		nic_mbx_send_nack(nic, vf);
983 unlock:
984 	nic->mbx_lock[vf] = FALSE;
985 }
986 
987 static void
988 nic_mbx_intr_handler(struct nicpf *nic, int mbx)
989 {
990 	uint64_t intr;
991 	uint8_t  vf, vf_per_mbx_reg = 64;
992 
993 	intr = nic_reg_read(nic, NIC_PF_MAILBOX_INT + (mbx << 3));
994 	for (vf = 0; vf < vf_per_mbx_reg; vf++) {
995 		if (intr & (1UL << vf)) {
996 			nic_handle_mbx_intr(nic, vf + (mbx * vf_per_mbx_reg));
997 			nic_clear_mbx_intr(nic, vf, mbx);
998 		}
999 	}
1000 }
1001 
1002 static void
1003 nic_mbx0_intr_handler (void *arg)
1004 {
1005 	struct nicpf *nic = (struct nicpf *)arg;
1006 
1007 	nic_mbx_intr_handler(nic, 0);
1008 }
1009 
1010 static void
1011 nic_mbx1_intr_handler (void *arg)
1012 {
1013 	struct nicpf *nic = (struct nicpf *)arg;
1014 
1015 	nic_mbx_intr_handler(nic, 1);
1016 }
1017 
1018 static int
1019 nic_enable_msix(struct nicpf *nic)
1020 {
1021 	struct pci_devinfo *dinfo;
1022 	int rid, count;
1023 	int ret;
1024 
1025 	dinfo = device_get_ivars(nic->dev);
1026 	rid = dinfo->cfg.msix.msix_table_bar;
1027 	nic->msix_table_res =
1028 	    bus_alloc_resource_any(nic->dev, SYS_RES_MEMORY, &rid, RF_ACTIVE);
1029 	if (nic->msix_table_res == NULL) {
1030 		device_printf(nic->dev,
1031 		    "Could not allocate memory for MSI-X table\n");
1032 		return (ENXIO);
1033 	}
1034 
1035 	count = nic->num_vec = NIC_PF_MSIX_VECTORS;
1036 
1037 	ret = pci_alloc_msix(nic->dev, &count);
1038 	if ((ret != 0) || (count != nic->num_vec)) {
1039 		device_printf(nic->dev,
1040 		    "Request for #%d msix vectors failed, error: %d\n",
1041 		    nic->num_vec, ret);
1042 		return (ret);
1043 	}
1044 
1045 	nic->msix_enabled = 1;
1046 	return (0);
1047 }
1048 
1049 static void
1050 nic_disable_msix(struct nicpf *nic)
1051 {
1052 	if (nic->msix_enabled) {
1053 		pci_release_msi(nic->dev);
1054 		nic->msix_enabled = 0;
1055 		nic->num_vec = 0;
1056 	}
1057 }
1058 
1059 static void
1060 nic_free_all_interrupts(struct nicpf *nic)
1061 {
1062 	int irq;
1063 
1064 	for (irq = 0; irq < nic->num_vec; irq++) {
1065 		if (nic->msix_entries[irq].irq_res == NULL)
1066 			continue;
1067 		if (nic->msix_entries[irq].handle != NULL) {
1068 			bus_teardown_intr(nic->dev,
1069 			    nic->msix_entries[irq].irq_res,
1070 			    nic->msix_entries[irq].handle);
1071 		}
1072 
1073 		bus_release_resource(nic->dev, SYS_RES_IRQ, irq,
1074 		    nic->msix_entries[irq].irq_res);
1075 	}
1076 }
1077 
1078 static int
1079 nic_register_interrupts(struct nicpf *nic)
1080 {
1081 	int irq, rid;
1082 	int ret;
1083 
1084 	/* Enable MSI-X */
1085 	ret = nic_enable_msix(nic);
1086 	if (ret != 0)
1087 		return (ret);
1088 
1089 	/* Register mailbox interrupt handlers */
1090 	irq = NIC_PF_INTR_ID_MBOX0;
1091 	rid = irq + 1;
1092 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1093 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1094 	if (nic->msix_entries[irq].irq_res == NULL) {
1095 		ret = ENXIO;
1096 		goto fail;
1097 	}
1098 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1099 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx0_intr_handler, nic,
1100 	    &nic->msix_entries[irq].handle);
1101 	if (ret != 0)
1102 		goto fail;
1103 
1104 	irq = NIC_PF_INTR_ID_MBOX1;
1105 	rid = irq + 1;
1106 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1107 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1108 	if (nic->msix_entries[irq].irq_res == NULL) {
1109 		ret = ENXIO;
1110 		goto fail;
1111 	}
1112 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1113 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx1_intr_handler, nic,
1114 	    &nic->msix_entries[irq].handle);
1115 	if (ret != 0)
1116 		goto fail;
1117 
1118 	/* Enable mailbox interrupt */
1119 	nic_enable_mbx_intr(nic);
1120 	return (0);
1121 
1122 fail:
1123 	nic_free_all_interrupts(nic);
1124 	return (ret);
1125 }
1126 
1127 static void
1128 nic_unregister_interrupts(struct nicpf *nic)
1129 {
1130 
1131 	nic_free_all_interrupts(nic);
1132 	nic_disable_msix(nic);
1133 }
1134 
1135 static int nic_sriov_init(device_t dev, struct nicpf *nic)
1136 {
1137 #ifdef PCI_IOV
1138 	nvlist_t *pf_schema, *vf_schema;
1139 	int iov_pos;
1140 	int err;
1141 	uint16_t total_vf_cnt;
1142 
1143 	err = pci_find_extcap(dev, PCIZ_SRIOV, &iov_pos);
1144 	if (err != 0) {
1145 		device_printf(dev,
1146 		    "SR-IOV capability is not found in PCIe config space\n");
1147 		return (err);
1148 	}
1149 	/* Fix-up the number of enabled VFs */
1150 	total_vf_cnt = pci_read_config(dev, iov_pos + PCIR_SRIOV_TOTAL_VFS, 2);
1151 	if (total_vf_cnt == 0)
1152 		return (ENXIO);
1153 
1154 	/* Attach SR-IOV */
1155 	pf_schema = pci_iov_schema_alloc_node();
1156 	vf_schema = pci_iov_schema_alloc_node();
1157 	pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL);
1158 	/*
1159 	 * All VFs can change their MACs.
1160 	 * This flag will be ignored but we set it just for the record.
1161 	 */
1162 	pci_iov_schema_add_bool(vf_schema, "allow-set-mac",
1163 	    IOV_SCHEMA_HASDEFAULT, TRUE);
1164 
1165 	err = pci_iov_attach(dev, pf_schema, vf_schema);
1166 	if (err != 0) {
1167 		device_printf(dev,
1168 		    "Failed to initialize SR-IOV (error=%d)\n",
1169 		    err);
1170 		return (err);
1171 	}
1172 #endif
1173 	return (0);
1174 }
1175 
1176 /*
1177  * Poll for BGX LMAC link status and update corresponding VF
1178  * if there is a change, valid only if internal L2 switch
1179  * is not present otherwise VF link is always treated as up
1180  */
1181 static void
1182 nic_poll_for_link(void *arg)
1183 {
1184 	union nic_mbx mbx = {};
1185 	struct nicpf *nic;
1186 	struct bgx_link_status link;
1187 	uint8_t vf, bgx, lmac;
1188 
1189 	nic = (struct nicpf *)arg;
1190 
1191 	mbx.link_status.msg = NIC_MBOX_MSG_BGX_LINK_CHANGE;
1192 
1193 	for (vf = 0; vf < nic->num_vf_en; vf++) {
1194 		/* Poll only if VF is UP */
1195 		if (!nic->vf_info[vf].vf_enabled)
1196 			continue;
1197 
1198 		/* Get BGX, LMAC indices for the VF */
1199 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1200 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1201 		/* Get interface link status */
1202 		bgx_get_lmac_link_state(nic->node, bgx, lmac, &link);
1203 
1204 		/* Inform VF only if link status changed */
1205 		if (nic->link[vf] == link.link_up)
1206 			continue;
1207 
1208 		if (!nic->mbx_lock[vf]) {
1209 			nic->link[vf] = link.link_up;
1210 			nic->duplex[vf] = link.duplex;
1211 			nic->speed[vf] = link.speed;
1212 
1213 			/* Send a mbox message to VF with current link status */
1214 			mbx.link_status.link_up = link.link_up;
1215 			mbx.link_status.duplex = link.duplex;
1216 			mbx.link_status.speed = link.speed;
1217 			nic_send_msg_to_vf(nic, vf, &mbx);
1218 		}
1219 	}
1220 	callout_reset(&nic->check_link, hz * 2, nic_poll_for_link, nic);
1221 }
1222