xref: /freebsd/sys/dev/vnic/nic_main.c (revision 9768746b)
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 /*
31  * Marvell/Cavium ThunderX vnic/bgx network controller
32  *
33  * UNIMPLEMENTED FEATURES
34  * ----------------------
35  * A number of features supported by the hardware are not yet implemented in
36  * this driver:
37  *
38  * - PR223573 multicast rx filter
39  * - PR223575 non-promiscuous mode (driver currently forces promisc)
40  */
41 
42 #include <sys/cdefs.h>
43 __FBSDID("$FreeBSD$");
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/bitset.h>
48 #include <sys/bitstring.h>
49 #include <sys/bus.h>
50 #include <sys/endian.h>
51 #include <sys/kernel.h>
52 #include <sys/malloc.h>
53 #include <sys/module.h>
54 #include <sys/rman.h>
55 #include <sys/pciio.h>
56 #include <sys/pcpu.h>
57 #include <sys/proc.h>
58 #include <sys/socket.h>
59 #include <sys/sockio.h>
60 #include <sys/cpuset.h>
61 #include <sys/lock.h>
62 #include <sys/mutex.h>
63 
64 #include <net/ethernet.h>
65 #include <net/if.h>
66 #include <net/if_media.h>
67 
68 #include <machine/bus.h>
69 #include <machine/_inttypes.h>
70 
71 #include <dev/pci/pcireg.h>
72 #include <dev/pci/pcivar.h>
73 
74 #include <sys/dnv.h>
75 #include <sys/nv.h>
76 #ifdef PCI_IOV
77 #include <sys/iov_schema.h>
78 #include <dev/pci/pci_iov.h>
79 #endif
80 
81 #include "thunder_bgx.h"
82 #include "nic_reg.h"
83 #include "nic.h"
84 #include "q_struct.h"
85 
86 #define	VNIC_PF_DEVSTR		"Cavium Thunder NIC Physical Function Driver"
87 
88 #define	VNIC_PF_REG_RID		PCIR_BAR(PCI_CFG_REG_BAR_NUM)
89 
90 #define	NIC_SET_VF_LMAC_MAP(bgx, lmac)		((((bgx) & 0xF) << 4) | ((lmac) & 0xF))
91 #define	NIC_GET_BGX_FROM_VF_LMAC_MAP(map)	(((map) >> 4) & 0xF)
92 #define	NIC_GET_LMAC_FROM_VF_LMAC_MAP(map)	((map) & 0xF)
93 
94 /* Structure to be used by the SR-IOV for VF configuration schemas */
95 struct nicvf_info {
96 	boolean_t		vf_enabled;
97 	int			vf_flags;
98 };
99 
100 struct nicpf {
101 	device_t		dev;
102 	uint8_t			node;
103 	u_int			flags;
104 	uint8_t			num_vf_en;      /* No of VF enabled */
105 	struct nicvf_info	vf_info[MAX_NUM_VFS_SUPPORTED];
106 	struct resource *	reg_base;       /* Register start address */
107 	struct pkind_cfg	pkind;
108 	uint8_t			vf_lmac_map[MAX_LMAC];
109 	boolean_t		mbx_lock[MAX_NUM_VFS_SUPPORTED];
110 
111 	struct callout		check_link;
112 	struct mtx		check_link_mtx;
113 
114 	uint8_t			link[MAX_LMAC];
115 	uint8_t			duplex[MAX_LMAC];
116 	uint32_t		speed[MAX_LMAC];
117 	uint16_t		cpi_base[MAX_NUM_VFS_SUPPORTED];
118 	uint16_t		rssi_base[MAX_NUM_VFS_SUPPORTED];
119 	uint16_t		rss_ind_tbl_size;
120 
121 	/* MSI-X */
122 	boolean_t		msix_enabled;
123 	uint8_t			num_vec;
124 	struct msix_entry	msix_entries[NIC_PF_MSIX_VECTORS];
125 	struct resource *	msix_table_res;
126 };
127 
128 static int nicpf_probe(device_t);
129 static int nicpf_attach(device_t);
130 static int nicpf_detach(device_t);
131 
132 #ifdef PCI_IOV
133 static int nicpf_iov_init(device_t, uint16_t, const nvlist_t *);
134 static void nicpf_iov_uninit(device_t);
135 static int nicpf_iov_add_vf(device_t, uint16_t, const nvlist_t *);
136 #endif
137 
138 static device_method_t nicpf_methods[] = {
139 	/* Device interface */
140 	DEVMETHOD(device_probe,		nicpf_probe),
141 	DEVMETHOD(device_attach,	nicpf_attach),
142 	DEVMETHOD(device_detach,	nicpf_detach),
143 	/* PCI SR-IOV interface */
144 #ifdef PCI_IOV
145 	DEVMETHOD(pci_iov_init,		nicpf_iov_init),
146 	DEVMETHOD(pci_iov_uninit,	nicpf_iov_uninit),
147 	DEVMETHOD(pci_iov_add_vf,	nicpf_iov_add_vf),
148 #endif
149 	DEVMETHOD_END,
150 };
151 
152 static driver_t vnicpf_driver = {
153 	"vnicpf",
154 	nicpf_methods,
155 	sizeof(struct nicpf),
156 };
157 
158 DRIVER_MODULE(vnicpf, pci, vnicpf_driver, 0, 0);
159 MODULE_VERSION(vnicpf, 1);
160 MODULE_DEPEND(vnicpf, pci, 1, 1, 1);
161 MODULE_DEPEND(vnicpf, ether, 1, 1, 1);
162 MODULE_DEPEND(vnicpf, thunder_bgx, 1, 1, 1);
163 
164 static int nicpf_alloc_res(struct nicpf *);
165 static void nicpf_free_res(struct nicpf *);
166 static void nic_set_lmac_vf_mapping(struct nicpf *);
167 static void nic_init_hw(struct nicpf *);
168 static int nic_sriov_init(device_t, struct nicpf *);
169 static void nic_poll_for_link(void *);
170 static int nic_register_interrupts(struct nicpf *);
171 static void nic_unregister_interrupts(struct nicpf *);
172 
173 /*
174  * Device interface
175  */
176 static int
177 nicpf_probe(device_t dev)
178 {
179 	uint16_t vendor_id;
180 	uint16_t device_id;
181 
182 	vendor_id = pci_get_vendor(dev);
183 	device_id = pci_get_device(dev);
184 
185 	if (vendor_id == PCI_VENDOR_ID_CAVIUM &&
186 	    device_id == PCI_DEVICE_ID_THUNDER_NIC_PF) {
187 		device_set_desc(dev, VNIC_PF_DEVSTR);
188 		return (BUS_PROBE_DEFAULT);
189 	}
190 
191 	return (ENXIO);
192 }
193 
194 static int
195 nicpf_attach(device_t dev)
196 {
197 	struct nicpf *nic;
198 	int err;
199 
200 	nic = device_get_softc(dev);
201 	nic->dev = dev;
202 
203 	/* Enable bus mastering */
204 	pci_enable_busmaster(dev);
205 
206 	/* Allocate PCI resources */
207 	err = nicpf_alloc_res(nic);
208 	if (err != 0) {
209 		device_printf(dev, "Could not allocate PCI resources\n");
210 		return (err);
211 	}
212 
213 	nic->node = nic_get_node_id(nic->reg_base);
214 
215 	/* Enable Traffic Network Switch (TNS) bypass mode by default */
216 	nic->flags &= ~NIC_TNS_ENABLED;
217 	nic_set_lmac_vf_mapping(nic);
218 
219 	/* Initialize hardware */
220 	nic_init_hw(nic);
221 
222 	/* Set RSS TBL size for each VF */
223 	nic->rss_ind_tbl_size = NIC_MAX_RSS_IDR_TBL_SIZE;
224 
225 	/* Setup interrupts */
226 	err = nic_register_interrupts(nic);
227 	if (err != 0)
228 		goto err_free_res;
229 
230 	/* Configure SRIOV */
231 	err = nic_sriov_init(dev, nic);
232 	if (err != 0)
233 		goto err_free_intr;
234 
235 	if (nic->flags & NIC_TNS_ENABLED)
236 		return (0);
237 
238 	mtx_init(&nic->check_link_mtx, "VNIC PF link poll", NULL, MTX_DEF);
239 	/* Register physical link status poll callout */
240 	callout_init_mtx(&nic->check_link, &nic->check_link_mtx, 0);
241 	mtx_lock(&nic->check_link_mtx);
242 	nic_poll_for_link(nic);
243 	mtx_unlock(&nic->check_link_mtx);
244 
245 	return (0);
246 
247 err_free_intr:
248 	nic_unregister_interrupts(nic);
249 err_free_res:
250 	nicpf_free_res(nic);
251 	pci_disable_busmaster(dev);
252 
253 	return (err);
254 }
255 
256 static int
257 nicpf_detach(device_t dev)
258 {
259 	struct nicpf *nic;
260 	int err;
261 
262 	err = 0;
263 	nic = device_get_softc(dev);
264 
265 	callout_drain(&nic->check_link);
266 	mtx_destroy(&nic->check_link_mtx);
267 
268 	nic_unregister_interrupts(nic);
269 	nicpf_free_res(nic);
270 	pci_disable_busmaster(dev);
271 
272 #ifdef PCI_IOV
273 	err = pci_iov_detach(dev);
274 	if (err != 0)
275 		device_printf(dev, "SR-IOV in use. Detach first.\n");
276 #endif
277 	return (err);
278 }
279 
280 /*
281  * SR-IOV interface
282  */
283 #ifdef PCI_IOV
284 static int
285 nicpf_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
286 {
287 	struct nicpf *nic;
288 
289 	nic = device_get_softc(dev);
290 
291 	if (num_vfs == 0)
292 		return (ENXIO);
293 
294 	nic->flags |= NIC_SRIOV_ENABLED;
295 
296 	return (0);
297 }
298 
299 static void
300 nicpf_iov_uninit(device_t dev)
301 {
302 
303 	/* ARM64TODO: Implement this function */
304 }
305 
306 static int
307 nicpf_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
308 {
309 	const void *mac;
310 	struct nicpf *nic;
311 	size_t size;
312 	int bgx, lmac;
313 
314 	nic = device_get_softc(dev);
315 
316 	if ((nic->flags & NIC_SRIOV_ENABLED) == 0)
317 		return (ENXIO);
318 
319 	if (vfnum > (nic->num_vf_en - 1))
320 		return (EINVAL);
321 
322 	if (nvlist_exists_binary(params, "mac-addr") != 0) {
323 		mac = nvlist_get_binary(params, "mac-addr", &size);
324 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
325 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
326 		bgx_set_lmac_mac(nic->node, bgx, lmac, mac);
327 	}
328 
329 	return (0);
330 }
331 #endif
332 
333 /*
334  * Helper routines
335  */
336 static int
337 nicpf_alloc_res(struct nicpf *nic)
338 {
339 	device_t dev;
340 	int rid;
341 
342 	dev = nic->dev;
343 
344 	rid = VNIC_PF_REG_RID;
345 	nic->reg_base = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
346 	    RF_ACTIVE);
347 	if (nic->reg_base == NULL) {
348 		/* For verbose output print some more details */
349 		if (bootverbose) {
350 			device_printf(dev,
351 			    "Could not allocate registers memory\n");
352 		}
353 		return (ENXIO);
354 	}
355 
356 	return (0);
357 }
358 
359 static void
360 nicpf_free_res(struct nicpf *nic)
361 {
362 	device_t dev;
363 
364 	dev = nic->dev;
365 
366 	if (nic->reg_base != NULL) {
367 		bus_release_resource(dev, SYS_RES_MEMORY,
368 		    rman_get_rid(nic->reg_base), nic->reg_base);
369 	}
370 }
371 
372 /* Register read/write APIs */
373 static __inline void
374 nic_reg_write(struct nicpf *nic, bus_space_handle_t offset,
375     uint64_t val)
376 {
377 
378 	bus_write_8(nic->reg_base, offset, val);
379 }
380 
381 static __inline uint64_t
382 nic_reg_read(struct nicpf *nic, uint64_t offset)
383 {
384 	uint64_t val;
385 
386 	val = bus_read_8(nic->reg_base, offset);
387 	return (val);
388 }
389 
390 /* PF -> VF mailbox communication APIs */
391 static void
392 nic_enable_mbx_intr(struct nicpf *nic)
393 {
394 
395 	/* Enable mailbox interrupt for all 128 VFs */
396 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S, ~0UL);
397 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S + sizeof(uint64_t), ~0UL);
398 }
399 
400 static void
401 nic_clear_mbx_intr(struct nicpf *nic, int vf, int mbx_reg)
402 {
403 
404 	nic_reg_write(nic, NIC_PF_MAILBOX_INT + (mbx_reg << 3), (1UL << vf));
405 }
406 
407 static uint64_t
408 nic_get_mbx_addr(int vf)
409 {
410 
411 	return (NIC_PF_VF_0_127_MAILBOX_0_1 + (vf << NIC_VF_NUM_SHIFT));
412 }
413 
414 /*
415  * Send a mailbox message to VF
416  * @vf: vf to which this message to be sent
417  * @mbx: Message to be sent
418  */
419 static void
420 nic_send_msg_to_vf(struct nicpf *nic, int vf, union nic_mbx *mbx)
421 {
422 	bus_space_handle_t mbx_addr = nic_get_mbx_addr(vf);
423 	uint64_t *msg = (uint64_t *)mbx;
424 
425 	/*
426 	 * In first revision HW, mbox interrupt is triggerred
427 	 * when PF writes to MBOX(1), in next revisions when
428 	 * PF writes to MBOX(0)
429 	 */
430 	if (pass1_silicon(nic->dev)) {
431 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
432 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
433 	} else {
434 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
435 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
436 	}
437 }
438 
439 /*
440  * Responds to VF's READY message with VF's
441  * ID, node, MAC address e.t.c
442  * @vf: VF which sent READY message
443  */
444 static void
445 nic_mbx_send_ready(struct nicpf *nic, int vf)
446 {
447 	union nic_mbx mbx = {};
448 	int bgx_idx, lmac;
449 	const char *mac;
450 
451 	mbx.nic_cfg.msg = NIC_MBOX_MSG_READY;
452 	mbx.nic_cfg.vf_id = vf;
453 
454 	if (nic->flags & NIC_TNS_ENABLED)
455 		mbx.nic_cfg.tns_mode = NIC_TNS_MODE;
456 	else
457 		mbx.nic_cfg.tns_mode = NIC_TNS_BYPASS_MODE;
458 
459 	if (vf < MAX_LMAC) {
460 		bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
461 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
462 
463 		mac = bgx_get_lmac_mac(nic->node, bgx_idx, lmac);
464 		if (mac) {
465 			memcpy((uint8_t *)&mbx.nic_cfg.mac_addr, mac,
466 			    ETHER_ADDR_LEN);
467 		}
468 	}
469 	mbx.nic_cfg.node_id = nic->node;
470 
471 	mbx.nic_cfg.loopback_supported = vf < MAX_LMAC;
472 
473 	nic_send_msg_to_vf(nic, vf, &mbx);
474 }
475 
476 /*
477  * ACKs VF's mailbox message
478  * @vf: VF to which ACK to be sent
479  */
480 static void
481 nic_mbx_send_ack(struct nicpf *nic, int vf)
482 {
483 	union nic_mbx mbx = {};
484 
485 	mbx.msg.msg = NIC_MBOX_MSG_ACK;
486 	nic_send_msg_to_vf(nic, vf, &mbx);
487 }
488 
489 /*
490  * NACKs VF's mailbox message that PF is not able to
491  * complete the action
492  * @vf: VF to which ACK to be sent
493  */
494 static void
495 nic_mbx_send_nack(struct nicpf *nic, int vf)
496 {
497 	union nic_mbx mbx = {};
498 
499 	mbx.msg.msg = NIC_MBOX_MSG_NACK;
500 	nic_send_msg_to_vf(nic, vf, &mbx);
501 }
502 
503 /*
504  * Flush all in flight receive packets to memory and
505  * bring down an active RQ
506  */
507 static int
508 nic_rcv_queue_sw_sync(struct nicpf *nic)
509 {
510 	uint16_t timeout = ~0x00;
511 
512 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x01);
513 	/* Wait till sync cycle is finished */
514 	while (timeout) {
515 		if (nic_reg_read(nic, NIC_PF_SW_SYNC_RX_DONE) & 0x1)
516 			break;
517 		timeout--;
518 	}
519 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x00);
520 	if (!timeout) {
521 		device_printf(nic->dev, "Receive queue software sync failed\n");
522 		return (ETIMEDOUT);
523 	}
524 	return (0);
525 }
526 
527 /* Get BGX Rx/Tx stats and respond to VF's request */
528 static void
529 nic_get_bgx_stats(struct nicpf *nic, struct bgx_stats_msg *bgx)
530 {
531 	int bgx_idx, lmac;
532 	union nic_mbx mbx = {};
533 
534 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
535 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
536 
537 	mbx.bgx_stats.msg = NIC_MBOX_MSG_BGX_STATS;
538 	mbx.bgx_stats.vf_id = bgx->vf_id;
539 	mbx.bgx_stats.rx = bgx->rx;
540 	mbx.bgx_stats.idx = bgx->idx;
541 	if (bgx->rx != 0) {
542 		mbx.bgx_stats.stats =
543 		    bgx_get_rx_stats(nic->node, bgx_idx, lmac, bgx->idx);
544 	} else {
545 		mbx.bgx_stats.stats =
546 		    bgx_get_tx_stats(nic->node, bgx_idx, lmac, bgx->idx);
547 	}
548 	nic_send_msg_to_vf(nic, bgx->vf_id, &mbx);
549 }
550 
551 /* Update hardware min/max frame size */
552 static int
553 nic_update_hw_frs(struct nicpf *nic, int new_frs, int vf)
554 {
555 
556 	if ((new_frs > NIC_HW_MAX_FRS) || (new_frs < NIC_HW_MIN_FRS)) {
557 		device_printf(nic->dev,
558 		    "Invalid MTU setting from VF%d rejected, "
559 		    "should be between %d and %d\n",
560 		    vf, NIC_HW_MIN_FRS, NIC_HW_MAX_FRS);
561 		return (EINVAL);
562 	}
563 	new_frs += ETHER_HDR_LEN;
564 	if (new_frs <= nic->pkind.maxlen)
565 		return (0);
566 
567 	nic->pkind.maxlen = new_frs;
568 	nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG, *(uint64_t *)&nic->pkind);
569 	return (0);
570 }
571 
572 /* Set minimum transmit packet size */
573 static void
574 nic_set_tx_pkt_pad(struct nicpf *nic, int size)
575 {
576 	int lmac;
577 	uint64_t lmac_cfg;
578 
579 	/* Max value that can be set is 60 */
580 	if (size > 60)
581 		size = 60;
582 
583 	for (lmac = 0; lmac < (MAX_BGX_PER_CN88XX * MAX_LMAC_PER_BGX); lmac++) {
584 		lmac_cfg = nic_reg_read(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3));
585 		lmac_cfg &= ~(0xF << 2);
586 		lmac_cfg |= ((size / 4) << 2);
587 		nic_reg_write(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3), lmac_cfg);
588 	}
589 }
590 
591 /*
592  * Function to check number of LMACs present and set VF::LMAC mapping.
593  * Mapping will be used while initializing channels.
594  */
595 static void
596 nic_set_lmac_vf_mapping(struct nicpf *nic)
597 {
598 	unsigned bgx_map = bgx_get_map(nic->node);
599 	int bgx, next_bgx_lmac = 0;
600 	int lmac, lmac_cnt = 0;
601 	uint64_t lmac_credit;
602 
603 	nic->num_vf_en = 0;
604 	if (nic->flags & NIC_TNS_ENABLED) {
605 		nic->num_vf_en = DEFAULT_NUM_VF_ENABLED;
606 		return;
607 	}
608 
609 	for (bgx = 0; bgx < NIC_MAX_BGX; bgx++) {
610 		if ((bgx_map & (1 << bgx)) == 0)
611 			continue;
612 		lmac_cnt = bgx_get_lmac_count(nic->node, bgx);
613 		for (lmac = 0; lmac < lmac_cnt; lmac++)
614 			nic->vf_lmac_map[next_bgx_lmac++] =
615 						NIC_SET_VF_LMAC_MAP(bgx, lmac);
616 		nic->num_vf_en += lmac_cnt;
617 
618 		/* Program LMAC credits */
619 		lmac_credit = (1UL << 1); /* channel credit enable */
620 		lmac_credit |= (0x1ff << 2); /* Max outstanding pkt count */
621 		/* 48KB BGX Tx buffer size, each unit is of size 16bytes */
622 		lmac_credit |= (((((48 * 1024) / lmac_cnt) -
623 		    NIC_HW_MAX_FRS) / 16) << 12);
624 		lmac = bgx * MAX_LMAC_PER_BGX;
625 		for (; lmac < lmac_cnt + (bgx * MAX_LMAC_PER_BGX); lmac++) {
626 			nic_reg_write(nic, NIC_PF_LMAC_0_7_CREDIT + (lmac * 8),
627 			    lmac_credit);
628 		}
629 	}
630 }
631 
632 #define TNS_PORT0_BLOCK 6
633 #define TNS_PORT1_BLOCK 7
634 #define BGX0_BLOCK 8
635 #define BGX1_BLOCK 9
636 
637 static void
638 nic_init_hw(struct nicpf *nic)
639 {
640 	int i;
641 
642 	/* Enable NIC HW block */
643 	nic_reg_write(nic, NIC_PF_CFG, 0x3);
644 
645 	/* Enable backpressure */
646 	nic_reg_write(nic, NIC_PF_BP_CFG, (1UL << 6) | 0x03);
647 
648 	if (nic->flags & NIC_TNS_ENABLED) {
649 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
650 		    (NIC_TNS_MODE << 7) | TNS_PORT0_BLOCK);
651 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
652 		    (NIC_TNS_MODE << 7) | TNS_PORT1_BLOCK);
653 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
654 		    (1UL << 63) | TNS_PORT0_BLOCK);
655 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
656 		    (1UL << 63) | TNS_PORT1_BLOCK);
657 
658 	} else {
659 		/* Disable TNS mode on both interfaces */
660 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
661 		    (NIC_TNS_BYPASS_MODE << 7) | BGX0_BLOCK);
662 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
663 		    (NIC_TNS_BYPASS_MODE << 7) | BGX1_BLOCK);
664 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
665 		    (1UL << 63) | BGX0_BLOCK);
666 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
667 		    (1UL << 63) | BGX1_BLOCK);
668 	}
669 
670 	/* PKIND configuration */
671 	nic->pkind.minlen = 0;
672 	nic->pkind.maxlen = NIC_HW_MAX_FRS + ETHER_HDR_LEN;
673 	nic->pkind.lenerr_en = 1;
674 	nic->pkind.rx_hdr = 0;
675 	nic->pkind.hdr_sl = 0;
676 
677 	for (i = 0; i < NIC_MAX_PKIND; i++) {
678 		nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG | (i << 3),
679 		    *(uint64_t *)&nic->pkind);
680 	}
681 
682 	nic_set_tx_pkt_pad(nic, NIC_HW_MIN_FRS);
683 
684 	/* Timer config */
685 	nic_reg_write(nic, NIC_PF_INTR_TIMER_CFG, NICPF_CLK_PER_INT_TICK);
686 
687 	/* Enable VLAN ethertype matching and stripping */
688 	nic_reg_write(nic, NIC_PF_RX_ETYPE_0_7,
689 	    (2 << 19) | (ETYPE_ALG_VLAN_STRIP << 16) | ETHERTYPE_VLAN);
690 }
691 
692 /* Channel parse index configuration */
693 static void
694 nic_config_cpi(struct nicpf *nic, struct cpi_cfg_msg *cfg)
695 {
696 	uint32_t vnic, bgx, lmac, chan;
697 	uint32_t padd, cpi_count = 0;
698 	uint64_t cpi_base, cpi, rssi_base, rssi;
699 	uint8_t qset, rq_idx = 0;
700 
701 	vnic = cfg->vf_id;
702 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
703 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
704 
705 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
706 	cpi_base = (lmac * NIC_MAX_CPI_PER_LMAC) + (bgx * NIC_CPI_PER_BGX);
707 	rssi_base = (lmac * nic->rss_ind_tbl_size) + (bgx * NIC_RSSI_PER_BGX);
708 
709 	/* Rx channel configuration */
710 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_BP_CFG | (chan << 3),
711 	    (1UL << 63) | (vnic << 0));
712 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_CFG | (chan << 3),
713 	    ((uint64_t)cfg->cpi_alg << 62) | (cpi_base << 48));
714 
715 	if (cfg->cpi_alg == CPI_ALG_NONE)
716 		cpi_count = 1;
717 	else if (cfg->cpi_alg == CPI_ALG_VLAN) /* 3 bits of PCP */
718 		cpi_count = 8;
719 	else if (cfg->cpi_alg == CPI_ALG_VLAN16) /* 3 bits PCP + DEI */
720 		cpi_count = 16;
721 	else if (cfg->cpi_alg == CPI_ALG_DIFF) /* 6bits DSCP */
722 		cpi_count = NIC_MAX_CPI_PER_LMAC;
723 
724 	/* RSS Qset, Qidx mapping */
725 	qset = cfg->vf_id;
726 	rssi = rssi_base;
727 	for (; rssi < (rssi_base + cfg->rq_cnt); rssi++) {
728 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
729 		    (qset << 3) | rq_idx);
730 		rq_idx++;
731 	}
732 
733 	rssi = 0;
734 	cpi = cpi_base;
735 	for (; cpi < (cpi_base + cpi_count); cpi++) {
736 		/* Determine port to channel adder */
737 		if (cfg->cpi_alg != CPI_ALG_DIFF)
738 			padd = cpi % cpi_count;
739 		else
740 			padd = cpi % 8; /* 3 bits CS out of 6bits DSCP */
741 
742 		/* Leave RSS_SIZE as '0' to disable RSS */
743 		if (pass1_silicon(nic->dev)) {
744 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
745 			    (vnic << 24) | (padd << 16) | (rssi_base + rssi));
746 		} else {
747 			/* Set MPI_ALG to '0' to disable MCAM parsing */
748 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
749 			    (padd << 16));
750 			/* MPI index is same as CPI if MPI_ALG is not enabled */
751 			nic_reg_write(nic, NIC_PF_MPI_0_2047_CFG | (cpi << 3),
752 			    (vnic << 24) | (rssi_base + rssi));
753 		}
754 
755 		if ((rssi + 1) >= cfg->rq_cnt)
756 			continue;
757 
758 		if (cfg->cpi_alg == CPI_ALG_VLAN)
759 			rssi++;
760 		else if (cfg->cpi_alg == CPI_ALG_VLAN16)
761 			rssi = ((cpi - cpi_base) & 0xe) >> 1;
762 		else if (cfg->cpi_alg == CPI_ALG_DIFF)
763 			rssi = ((cpi - cpi_base) & 0x38) >> 3;
764 	}
765 	nic->cpi_base[cfg->vf_id] = cpi_base;
766 	nic->rssi_base[cfg->vf_id] = rssi_base;
767 }
768 
769 /* Responsds to VF with its RSS indirection table size */
770 static void
771 nic_send_rss_size(struct nicpf *nic, int vf)
772 {
773 	union nic_mbx mbx = {};
774 
775 	mbx.rss_size.msg = NIC_MBOX_MSG_RSS_SIZE;
776 	mbx.rss_size.ind_tbl_size = nic->rss_ind_tbl_size;
777 	nic_send_msg_to_vf(nic, vf, &mbx);
778 }
779 
780 /*
781  * Receive side scaling configuration
782  * configure:
783  * - RSS index
784  * - indir table i.e hash::RQ mapping
785  * - no of hash bits to consider
786  */
787 static void
788 nic_config_rss(struct nicpf *nic, struct rss_cfg_msg *cfg)
789 {
790 	uint8_t qset, idx;
791 	uint64_t cpi_cfg, cpi_base, rssi_base, rssi;
792 	uint64_t idx_addr;
793 
794 	idx = 0;
795 	rssi_base = nic->rssi_base[cfg->vf_id] + cfg->tbl_offset;
796 
797 	rssi = rssi_base;
798 	qset = cfg->vf_id;
799 
800 	for (; rssi < (rssi_base + cfg->tbl_len); rssi++) {
801 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
802 		    (qset << 3) | (cfg->ind_tbl[idx] & 0x7));
803 		idx++;
804 	}
805 
806 	cpi_base = nic->cpi_base[cfg->vf_id];
807 	if (pass1_silicon(nic->dev))
808 		idx_addr = NIC_PF_CPI_0_2047_CFG;
809 	else
810 		idx_addr = NIC_PF_MPI_0_2047_CFG;
811 	cpi_cfg = nic_reg_read(nic, idx_addr | (cpi_base << 3));
812 	cpi_cfg &= ~(0xFUL << 20);
813 	cpi_cfg |= (cfg->hash_bits << 20);
814 	nic_reg_write(nic, idx_addr | (cpi_base << 3), cpi_cfg);
815 }
816 
817 /*
818  * 4 level transmit side scheduler configutation
819  * for TNS bypass mode
820  *
821  * Sample configuration for SQ0
822  * VNIC0-SQ0 -> TL4(0)   -> TL3[0]   -> TL2[0]  -> TL1[0] -> BGX0
823  * VNIC1-SQ0 -> TL4(8)   -> TL3[2]   -> TL2[0]  -> TL1[0] -> BGX0
824  * VNIC2-SQ0 -> TL4(16)  -> TL3[4]   -> TL2[1]  -> TL1[0] -> BGX0
825  * VNIC3-SQ0 -> TL4(24)  -> TL3[6]   -> TL2[1]  -> TL1[0] -> BGX0
826  * VNIC4-SQ0 -> TL4(512) -> TL3[128] -> TL2[32] -> TL1[1] -> BGX1
827  * VNIC5-SQ0 -> TL4(520) -> TL3[130] -> TL2[32] -> TL1[1] -> BGX1
828  * VNIC6-SQ0 -> TL4(528) -> TL3[132] -> TL2[33] -> TL1[1] -> BGX1
829  * VNIC7-SQ0 -> TL4(536) -> TL3[134] -> TL2[33] -> TL1[1] -> BGX1
830  */
831 static void
832 nic_tx_channel_cfg(struct nicpf *nic, uint8_t vnic, struct sq_cfg_msg *sq)
833 {
834 	uint32_t bgx, lmac, chan;
835 	uint32_t tl2, tl3, tl4;
836 	uint32_t rr_quantum;
837 	uint8_t sq_idx = sq->sq_num;
838 	uint8_t pqs_vnic;
839 
840 	pqs_vnic = vnic;
841 
842 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
843 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
844 
845 	/* 24 bytes for FCS, IPG and preamble */
846 	rr_quantum = ((NIC_HW_MAX_FRS + 24) / 4);
847 
848 	tl4 = (lmac * NIC_TL4_PER_LMAC) + (bgx * NIC_TL4_PER_BGX);
849 	tl4 += sq_idx;
850 
851 	tl3 = tl4 / (NIC_MAX_TL4 / NIC_MAX_TL3);
852 	nic_reg_write(nic, NIC_PF_QSET_0_127_SQ_0_7_CFG2 |
853 	    ((uint64_t)vnic << NIC_QS_ID_SHIFT) |
854 	    ((uint32_t)sq_idx << NIC_Q_NUM_SHIFT), tl4);
855 	nic_reg_write(nic, NIC_PF_TL4_0_1023_CFG | (tl4 << 3),
856 	    ((uint64_t)vnic << 27) | ((uint32_t)sq_idx << 24) | rr_quantum);
857 
858 	nic_reg_write(nic, NIC_PF_TL3_0_255_CFG | (tl3 << 3), rr_quantum);
859 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
860 	nic_reg_write(nic, NIC_PF_TL3_0_255_CHAN | (tl3 << 3), chan);
861 	/* Enable backpressure on the channel */
862 	nic_reg_write(nic, NIC_PF_CHAN_0_255_TX_CFG | (chan << 3), 1);
863 
864 	tl2 = tl3 >> 2;
865 	nic_reg_write(nic, NIC_PF_TL3A_0_63_CFG | (tl2 << 3), tl2);
866 	nic_reg_write(nic, NIC_PF_TL2_0_63_CFG | (tl2 << 3), rr_quantum);
867 	/* No priorities as of now */
868 	nic_reg_write(nic, NIC_PF_TL2_0_63_PRI | (tl2 << 3), 0x00);
869 }
870 
871 static int
872 nic_config_loopback(struct nicpf *nic, struct set_loopback *lbk)
873 {
874 	int bgx_idx, lmac_idx;
875 
876 	if (lbk->vf_id > MAX_LMAC)
877 		return (ENXIO);
878 
879 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
880 	lmac_idx = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
881 
882 	bgx_lmac_internal_loopback(nic->node, bgx_idx, lmac_idx, lbk->enable);
883 
884 	return (0);
885 }
886 
887 /* Interrupt handler to handle mailbox messages from VFs */
888 static void
889 nic_handle_mbx_intr(struct nicpf *nic, int vf)
890 {
891 	union nic_mbx mbx = {};
892 	uint64_t *mbx_data;
893 	uint64_t mbx_addr;
894 	uint64_t reg_addr;
895 	uint64_t cfg;
896 	int bgx, lmac;
897 	int i;
898 	int ret = 0;
899 
900 	nic->mbx_lock[vf] = TRUE;
901 
902 	mbx_addr = nic_get_mbx_addr(vf);
903 	mbx_data = (uint64_t *)&mbx;
904 
905 	for (i = 0; i < NIC_PF_VF_MAILBOX_SIZE; i++) {
906 		*mbx_data = nic_reg_read(nic, mbx_addr);
907 		mbx_data++;
908 		mbx_addr += sizeof(uint64_t);
909 	}
910 
911 	switch (mbx.msg.msg) {
912 	case NIC_MBOX_MSG_READY:
913 		nic_mbx_send_ready(nic, vf);
914 		if (vf < MAX_LMAC) {
915 			nic->link[vf] = 0;
916 			nic->duplex[vf] = 0;
917 			nic->speed[vf] = 0;
918 		}
919 		ret = 1;
920 		break;
921 	case NIC_MBOX_MSG_QS_CFG:
922 		reg_addr = NIC_PF_QSET_0_127_CFG |
923 		    (mbx.qs.num << NIC_QS_ID_SHIFT);
924 		cfg = mbx.qs.cfg;
925 		nic_reg_write(nic, reg_addr, cfg);
926 		break;
927 	case NIC_MBOX_MSG_RQ_CFG:
928 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_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_RQ_BP_CFG:
934 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_BP_CFG |
935 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
936 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
937 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
938 		break;
939 	case NIC_MBOX_MSG_RQ_SW_SYNC:
940 		ret = nic_rcv_queue_sw_sync(nic);
941 		break;
942 	case NIC_MBOX_MSG_RQ_DROP_CFG:
943 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_DROP_CFG |
944 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
945 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
946 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
947 		break;
948 	case NIC_MBOX_MSG_SQ_CFG:
949 		reg_addr = NIC_PF_QSET_0_127_SQ_0_7_CFG |
950 		    (mbx.sq.qs_num << NIC_QS_ID_SHIFT) |
951 		    (mbx.sq.sq_num << NIC_Q_NUM_SHIFT);
952 		nic_reg_write(nic, reg_addr, mbx.sq.cfg);
953 		nic_tx_channel_cfg(nic, mbx.qs.num, &mbx.sq);
954 		break;
955 	case NIC_MBOX_MSG_SET_MAC:
956 		lmac = mbx.mac.vf_id;
957 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
958 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
959 		bgx_set_lmac_mac(nic->node, bgx, lmac, mbx.mac.mac_addr);
960 		break;
961 	case NIC_MBOX_MSG_SET_MAX_FRS:
962 		ret = nic_update_hw_frs(nic, mbx.frs.max_frs, mbx.frs.vf_id);
963 		break;
964 	case NIC_MBOX_MSG_CPI_CFG:
965 		nic_config_cpi(nic, &mbx.cpi_cfg);
966 		break;
967 	case NIC_MBOX_MSG_RSS_SIZE:
968 		nic_send_rss_size(nic, vf);
969 		goto unlock;
970 	case NIC_MBOX_MSG_RSS_CFG:
971 	case NIC_MBOX_MSG_RSS_CFG_CONT: /* fall through */
972 		nic_config_rss(nic, &mbx.rss_cfg);
973 		break;
974 	case NIC_MBOX_MSG_CFG_DONE:
975 		/* Last message of VF config msg sequence */
976 		nic->vf_info[vf].vf_enabled = TRUE;
977 		goto unlock;
978 	case NIC_MBOX_MSG_SHUTDOWN:
979 		/* First msg in VF teardown sequence */
980 		nic->vf_info[vf].vf_enabled = FALSE;
981 		break;
982 	case NIC_MBOX_MSG_BGX_STATS:
983 		nic_get_bgx_stats(nic, &mbx.bgx_stats);
984 		goto unlock;
985 	case NIC_MBOX_MSG_LOOPBACK:
986 		ret = nic_config_loopback(nic, &mbx.lbk);
987 		break;
988 	default:
989 		device_printf(nic->dev,
990 		    "Invalid msg from VF%d, msg 0x%x\n", vf, mbx.msg.msg);
991 		break;
992 	}
993 
994 	if (ret == 0)
995 		nic_mbx_send_ack(nic, vf);
996 	else if (mbx.msg.msg != NIC_MBOX_MSG_READY)
997 		nic_mbx_send_nack(nic, vf);
998 unlock:
999 	nic->mbx_lock[vf] = FALSE;
1000 }
1001 
1002 static void
1003 nic_mbx_intr_handler(struct nicpf *nic, int mbx)
1004 {
1005 	uint64_t intr;
1006 	uint8_t  vf, vf_per_mbx_reg = 64;
1007 
1008 	intr = nic_reg_read(nic, NIC_PF_MAILBOX_INT + (mbx << 3));
1009 	for (vf = 0; vf < vf_per_mbx_reg; vf++) {
1010 		if (intr & (1UL << vf)) {
1011 			nic_handle_mbx_intr(nic, vf + (mbx * vf_per_mbx_reg));
1012 			nic_clear_mbx_intr(nic, vf, mbx);
1013 		}
1014 	}
1015 }
1016 
1017 static void
1018 nic_mbx0_intr_handler (void *arg)
1019 {
1020 	struct nicpf *nic = (struct nicpf *)arg;
1021 
1022 	nic_mbx_intr_handler(nic, 0);
1023 }
1024 
1025 static void
1026 nic_mbx1_intr_handler (void *arg)
1027 {
1028 	struct nicpf *nic = (struct nicpf *)arg;
1029 
1030 	nic_mbx_intr_handler(nic, 1);
1031 }
1032 
1033 static int
1034 nic_enable_msix(struct nicpf *nic)
1035 {
1036 	struct pci_devinfo *dinfo;
1037 	int rid, count;
1038 	int ret;
1039 
1040 	dinfo = device_get_ivars(nic->dev);
1041 	rid = dinfo->cfg.msix.msix_table_bar;
1042 	nic->msix_table_res =
1043 	    bus_alloc_resource_any(nic->dev, SYS_RES_MEMORY, &rid, RF_ACTIVE);
1044 	if (nic->msix_table_res == NULL) {
1045 		device_printf(nic->dev,
1046 		    "Could not allocate memory for MSI-X table\n");
1047 		return (ENXIO);
1048 	}
1049 
1050 	count = nic->num_vec = NIC_PF_MSIX_VECTORS;
1051 
1052 	ret = pci_alloc_msix(nic->dev, &count);
1053 	if ((ret != 0) || (count != nic->num_vec)) {
1054 		device_printf(nic->dev,
1055 		    "Request for #%d msix vectors failed, error: %d\n",
1056 		    nic->num_vec, ret);
1057 		return (ret);
1058 	}
1059 
1060 	nic->msix_enabled = 1;
1061 	return (0);
1062 }
1063 
1064 static void
1065 nic_disable_msix(struct nicpf *nic)
1066 {
1067 	if (nic->msix_enabled) {
1068 		pci_release_msi(nic->dev);
1069 		nic->msix_enabled = 0;
1070 		nic->num_vec = 0;
1071 	}
1072 
1073 	bus_release_resource(nic->dev, SYS_RES_MEMORY,
1074 	    rman_get_rid(nic->msix_table_res), nic->msix_table_res);
1075 }
1076 
1077 static void
1078 nic_free_all_interrupts(struct nicpf *nic)
1079 {
1080 	int irq;
1081 
1082 	for (irq = 0; irq < nic->num_vec; irq++) {
1083 		if (nic->msix_entries[irq].irq_res == NULL)
1084 			continue;
1085 		if (nic->msix_entries[irq].handle != NULL) {
1086 			bus_teardown_intr(nic->dev,
1087 			    nic->msix_entries[irq].irq_res,
1088 			    nic->msix_entries[irq].handle);
1089 		}
1090 
1091 		bus_release_resource(nic->dev, SYS_RES_IRQ, irq + 1,
1092 		    nic->msix_entries[irq].irq_res);
1093 	}
1094 }
1095 
1096 static int
1097 nic_register_interrupts(struct nicpf *nic)
1098 {
1099 	int irq, rid;
1100 	int ret;
1101 
1102 	/* Enable MSI-X */
1103 	ret = nic_enable_msix(nic);
1104 	if (ret != 0)
1105 		return (ret);
1106 
1107 	/* Register mailbox interrupt handlers */
1108 	irq = NIC_PF_INTR_ID_MBOX0;
1109 	rid = irq + 1;
1110 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1111 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1112 	if (nic->msix_entries[irq].irq_res == NULL) {
1113 		ret = ENXIO;
1114 		goto fail;
1115 	}
1116 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1117 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx0_intr_handler, nic,
1118 	    &nic->msix_entries[irq].handle);
1119 	if (ret != 0)
1120 		goto fail;
1121 
1122 	irq = NIC_PF_INTR_ID_MBOX1;
1123 	rid = irq + 1;
1124 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1125 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1126 	if (nic->msix_entries[irq].irq_res == NULL) {
1127 		ret = ENXIO;
1128 		goto fail;
1129 	}
1130 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1131 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx1_intr_handler, nic,
1132 	    &nic->msix_entries[irq].handle);
1133 	if (ret != 0)
1134 		goto fail;
1135 
1136 	/* Enable mailbox interrupt */
1137 	nic_enable_mbx_intr(nic);
1138 	return (0);
1139 
1140 fail:
1141 	nic_free_all_interrupts(nic);
1142 	return (ret);
1143 }
1144 
1145 static void
1146 nic_unregister_interrupts(struct nicpf *nic)
1147 {
1148 
1149 	nic_free_all_interrupts(nic);
1150 	nic_disable_msix(nic);
1151 }
1152 
1153 static int nic_sriov_init(device_t dev, struct nicpf *nic)
1154 {
1155 #ifdef PCI_IOV
1156 	nvlist_t *pf_schema, *vf_schema;
1157 	int iov_pos;
1158 	int err;
1159 	uint16_t total_vf_cnt;
1160 
1161 	err = pci_find_extcap(dev, PCIZ_SRIOV, &iov_pos);
1162 	if (err != 0) {
1163 		device_printf(dev,
1164 		    "SR-IOV capability is not found in PCIe config space\n");
1165 		return (err);
1166 	}
1167 	/* Fix-up the number of enabled VFs */
1168 	total_vf_cnt = pci_read_config(dev, iov_pos + PCIR_SRIOV_TOTAL_VFS, 2);
1169 	if (total_vf_cnt == 0)
1170 		return (ENXIO);
1171 
1172 	/* Attach SR-IOV */
1173 	pf_schema = pci_iov_schema_alloc_node();
1174 	vf_schema = pci_iov_schema_alloc_node();
1175 	pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL);
1176 	/*
1177 	 * All VFs can change their MACs.
1178 	 * This flag will be ignored but we set it just for the record.
1179 	 */
1180 	pci_iov_schema_add_bool(vf_schema, "allow-set-mac",
1181 	    IOV_SCHEMA_HASDEFAULT, TRUE);
1182 
1183 	err = pci_iov_attach(dev, pf_schema, vf_schema);
1184 	if (err != 0) {
1185 		device_printf(dev,
1186 		    "Failed to initialize SR-IOV (error=%d)\n",
1187 		    err);
1188 		return (err);
1189 	}
1190 #endif
1191 	return (0);
1192 }
1193 
1194 /*
1195  * Poll for BGX LMAC link status and update corresponding VF
1196  * if there is a change, valid only if internal L2 switch
1197  * is not present otherwise VF link is always treated as up
1198  */
1199 static void
1200 nic_poll_for_link(void *arg)
1201 {
1202 	union nic_mbx mbx = {};
1203 	struct nicpf *nic;
1204 	struct bgx_link_status link;
1205 	uint8_t vf, bgx, lmac;
1206 
1207 	nic = (struct nicpf *)arg;
1208 
1209 	mbx.link_status.msg = NIC_MBOX_MSG_BGX_LINK_CHANGE;
1210 
1211 	for (vf = 0; vf < nic->num_vf_en; vf++) {
1212 		/* Poll only if VF is UP */
1213 		if (!nic->vf_info[vf].vf_enabled)
1214 			continue;
1215 
1216 		/* Get BGX, LMAC indices for the VF */
1217 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1218 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1219 		/* Get interface link status */
1220 		bgx_get_lmac_link_state(nic->node, bgx, lmac, &link);
1221 
1222 		/* Inform VF only if link status changed */
1223 		if (nic->link[vf] == link.link_up)
1224 			continue;
1225 
1226 		if (!nic->mbx_lock[vf]) {
1227 			nic->link[vf] = link.link_up;
1228 			nic->duplex[vf] = link.duplex;
1229 			nic->speed[vf] = link.speed;
1230 
1231 			/* Send a mbox message to VF with current link status */
1232 			mbx.link_status.link_up = link.link_up;
1233 			mbx.link_status.duplex = link.duplex;
1234 			mbx.link_status.speed = link.speed;
1235 			nic_send_msg_to_vf(nic, vf, &mbx);
1236 		}
1237 	}
1238 	callout_reset(&nic->check_link, hz * 2, nic_poll_for_link, nic);
1239 }
1240