xref: /freebsd/sys/dev/vnic/nic_main.c (revision b3e7694832e81d7a904a10f525f8797b753bf0d3)
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  */
27 
28 /*
29  * Marvell/Cavium ThunderX vnic/bgx network controller
30  *
31  * UNIMPLEMENTED FEATURES
32  * ----------------------
33  * A number of features supported by the hardware are not yet implemented in
34  * this driver:
35  *
36  * - PR223573 multicast rx filter
37  * - PR223575 non-promiscuous mode (driver currently forces promisc)
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/bitset.h>
46 #include <sys/bitstring.h>
47 #include <sys/bus.h>
48 #include <sys/endian.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/module.h>
52 #include <sys/rman.h>
53 #include <sys/pciio.h>
54 #include <sys/pcpu.h>
55 #include <sys/proc.h>
56 #include <sys/socket.h>
57 #include <sys/sockio.h>
58 #include <sys/cpuset.h>
59 #include <sys/lock.h>
60 #include <sys/mutex.h>
61 
62 #include <net/ethernet.h>
63 #include <net/if.h>
64 #include <net/if_media.h>
65 
66 #include <machine/bus.h>
67 #include <machine/_inttypes.h>
68 
69 #include <dev/pci/pcireg.h>
70 #include <dev/pci/pcivar.h>
71 
72 #include <sys/dnv.h>
73 #include <sys/nv.h>
74 #ifdef PCI_IOV
75 #include <sys/iov_schema.h>
76 #include <dev/pci/pci_iov.h>
77 #endif
78 
79 #include "thunder_bgx.h"
80 #include "nic_reg.h"
81 #include "nic.h"
82 #include "q_struct.h"
83 
84 #define	VNIC_PF_DEVSTR		"Cavium Thunder NIC Physical Function Driver"
85 
86 #define	VNIC_PF_REG_RID		PCIR_BAR(PCI_CFG_REG_BAR_NUM)
87 
88 #define	NIC_SET_VF_LMAC_MAP(bgx, lmac)		((((bgx) & 0xF) << 4) | ((lmac) & 0xF))
89 #define	NIC_GET_BGX_FROM_VF_LMAC_MAP(map)	(((map) >> 4) & 0xF)
90 #define	NIC_GET_LMAC_FROM_VF_LMAC_MAP(map)	((map) & 0xF)
91 
92 /* Structure to be used by the SR-IOV for VF configuration schemas */
93 struct nicvf_info {
94 	boolean_t		vf_enabled;
95 	int			vf_flags;
96 };
97 
98 struct nicpf {
99 	device_t		dev;
100 	uint8_t			node;
101 	u_int			flags;
102 	uint8_t			num_vf_en;      /* No of VF enabled */
103 	struct nicvf_info	vf_info[MAX_NUM_VFS_SUPPORTED];
104 	struct resource *	reg_base;       /* Register start address */
105 	struct pkind_cfg	pkind;
106 	uint8_t			vf_lmac_map[MAX_LMAC];
107 	boolean_t		mbx_lock[MAX_NUM_VFS_SUPPORTED];
108 
109 	struct callout		check_link;
110 	struct mtx		check_link_mtx;
111 
112 	uint8_t			link[MAX_LMAC];
113 	uint8_t			duplex[MAX_LMAC];
114 	uint32_t		speed[MAX_LMAC];
115 	uint16_t		cpi_base[MAX_NUM_VFS_SUPPORTED];
116 	uint16_t		rssi_base[MAX_NUM_VFS_SUPPORTED];
117 	uint16_t		rss_ind_tbl_size;
118 
119 	/* MSI-X */
120 	boolean_t		msix_enabled;
121 	uint8_t			num_vec;
122 	struct msix_entry	msix_entries[NIC_PF_MSIX_VECTORS];
123 	struct resource *	msix_table_res;
124 };
125 
126 static int nicpf_probe(device_t);
127 static int nicpf_attach(device_t);
128 static int nicpf_detach(device_t);
129 
130 #ifdef PCI_IOV
131 static int nicpf_iov_init(device_t, uint16_t, const nvlist_t *);
132 static void nicpf_iov_uninit(device_t);
133 static int nicpf_iov_add_vf(device_t, uint16_t, const nvlist_t *);
134 #endif
135 
136 static device_method_t nicpf_methods[] = {
137 	/* Device interface */
138 	DEVMETHOD(device_probe,		nicpf_probe),
139 	DEVMETHOD(device_attach,	nicpf_attach),
140 	DEVMETHOD(device_detach,	nicpf_detach),
141 	/* PCI SR-IOV interface */
142 #ifdef PCI_IOV
143 	DEVMETHOD(pci_iov_init,		nicpf_iov_init),
144 	DEVMETHOD(pci_iov_uninit,	nicpf_iov_uninit),
145 	DEVMETHOD(pci_iov_add_vf,	nicpf_iov_add_vf),
146 #endif
147 	DEVMETHOD_END,
148 };
149 
150 static driver_t vnicpf_driver = {
151 	"vnicpf",
152 	nicpf_methods,
153 	sizeof(struct nicpf),
154 };
155 
156 DRIVER_MODULE(vnicpf, pci, vnicpf_driver, 0, 0);
157 MODULE_VERSION(vnicpf, 1);
158 MODULE_DEPEND(vnicpf, pci, 1, 1, 1);
159 MODULE_DEPEND(vnicpf, ether, 1, 1, 1);
160 MODULE_DEPEND(vnicpf, thunder_bgx, 1, 1, 1);
161 
162 static int nicpf_alloc_res(struct nicpf *);
163 static void nicpf_free_res(struct nicpf *);
164 static void nic_set_lmac_vf_mapping(struct nicpf *);
165 static void nic_init_hw(struct nicpf *);
166 static int nic_sriov_init(device_t, struct nicpf *);
167 static void nic_poll_for_link(void *);
168 static int nic_register_interrupts(struct nicpf *);
169 static void nic_unregister_interrupts(struct nicpf *);
170 
171 /*
172  * Device interface
173  */
174 static int
175 nicpf_probe(device_t dev)
176 {
177 	uint16_t vendor_id;
178 	uint16_t device_id;
179 
180 	vendor_id = pci_get_vendor(dev);
181 	device_id = pci_get_device(dev);
182 
183 	if (vendor_id == PCI_VENDOR_ID_CAVIUM &&
184 	    device_id == PCI_DEVICE_ID_THUNDER_NIC_PF) {
185 		device_set_desc(dev, VNIC_PF_DEVSTR);
186 		return (BUS_PROBE_DEFAULT);
187 	}
188 
189 	return (ENXIO);
190 }
191 
192 static int
193 nicpf_attach(device_t dev)
194 {
195 	struct nicpf *nic;
196 	int err;
197 
198 	nic = device_get_softc(dev);
199 	nic->dev = dev;
200 
201 	/* Enable bus mastering */
202 	pci_enable_busmaster(dev);
203 
204 	/* Allocate PCI resources */
205 	err = nicpf_alloc_res(nic);
206 	if (err != 0) {
207 		device_printf(dev, "Could not allocate PCI resources\n");
208 		return (err);
209 	}
210 
211 	nic->node = nic_get_node_id(nic->reg_base);
212 
213 	/* Enable Traffic Network Switch (TNS) bypass mode by default */
214 	nic->flags &= ~NIC_TNS_ENABLED;
215 	nic_set_lmac_vf_mapping(nic);
216 
217 	/* Initialize hardware */
218 	nic_init_hw(nic);
219 
220 	/* Set RSS TBL size for each VF */
221 	nic->rss_ind_tbl_size = NIC_MAX_RSS_IDR_TBL_SIZE;
222 
223 	/* Setup interrupts */
224 	err = nic_register_interrupts(nic);
225 	if (err != 0)
226 		goto err_free_res;
227 
228 	/* Configure SRIOV */
229 	err = nic_sriov_init(dev, nic);
230 	if (err != 0)
231 		goto err_free_intr;
232 
233 	if (nic->flags & NIC_TNS_ENABLED)
234 		return (0);
235 
236 	mtx_init(&nic->check_link_mtx, "VNIC PF link poll", NULL, MTX_DEF);
237 	/* Register physical link status poll callout */
238 	callout_init_mtx(&nic->check_link, &nic->check_link_mtx, 0);
239 	mtx_lock(&nic->check_link_mtx);
240 	nic_poll_for_link(nic);
241 	mtx_unlock(&nic->check_link_mtx);
242 
243 	return (0);
244 
245 err_free_intr:
246 	nic_unregister_interrupts(nic);
247 err_free_res:
248 	nicpf_free_res(nic);
249 	pci_disable_busmaster(dev);
250 
251 	return (err);
252 }
253 
254 static int
255 nicpf_detach(device_t dev)
256 {
257 	struct nicpf *nic;
258 	int err;
259 
260 	err = 0;
261 	nic = device_get_softc(dev);
262 
263 	callout_drain(&nic->check_link);
264 	mtx_destroy(&nic->check_link_mtx);
265 
266 	nic_unregister_interrupts(nic);
267 	nicpf_free_res(nic);
268 	pci_disable_busmaster(dev);
269 
270 #ifdef PCI_IOV
271 	err = pci_iov_detach(dev);
272 	if (err != 0)
273 		device_printf(dev, "SR-IOV in use. Detach first.\n");
274 #endif
275 	return (err);
276 }
277 
278 /*
279  * SR-IOV interface
280  */
281 #ifdef PCI_IOV
282 static int
283 nicpf_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
284 {
285 	struct nicpf *nic;
286 
287 	nic = device_get_softc(dev);
288 
289 	if (num_vfs == 0)
290 		return (ENXIO);
291 
292 	nic->flags |= NIC_SRIOV_ENABLED;
293 
294 	return (0);
295 }
296 
297 static void
298 nicpf_iov_uninit(device_t dev)
299 {
300 
301 	/* ARM64TODO: Implement this function */
302 }
303 
304 static int
305 nicpf_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
306 {
307 	const void *mac;
308 	struct nicpf *nic;
309 	size_t size;
310 	int bgx, lmac;
311 
312 	nic = device_get_softc(dev);
313 
314 	if ((nic->flags & NIC_SRIOV_ENABLED) == 0)
315 		return (ENXIO);
316 
317 	if (vfnum > (nic->num_vf_en - 1))
318 		return (EINVAL);
319 
320 	if (nvlist_exists_binary(params, "mac-addr") != 0) {
321 		mac = nvlist_get_binary(params, "mac-addr", &size);
322 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
323 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
324 		bgx_set_lmac_mac(nic->node, bgx, lmac, mac);
325 	}
326 
327 	return (0);
328 }
329 #endif
330 
331 /*
332  * Helper routines
333  */
334 static int
335 nicpf_alloc_res(struct nicpf *nic)
336 {
337 	device_t dev;
338 	int rid;
339 
340 	dev = nic->dev;
341 
342 	rid = VNIC_PF_REG_RID;
343 	nic->reg_base = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
344 	    RF_ACTIVE);
345 	if (nic->reg_base == NULL) {
346 		/* For verbose output print some more details */
347 		if (bootverbose) {
348 			device_printf(dev,
349 			    "Could not allocate registers memory\n");
350 		}
351 		return (ENXIO);
352 	}
353 
354 	return (0);
355 }
356 
357 static void
358 nicpf_free_res(struct nicpf *nic)
359 {
360 	device_t dev;
361 
362 	dev = nic->dev;
363 
364 	if (nic->reg_base != NULL) {
365 		bus_release_resource(dev, SYS_RES_MEMORY,
366 		    rman_get_rid(nic->reg_base), nic->reg_base);
367 	}
368 }
369 
370 /* Register read/write APIs */
371 static __inline void
372 nic_reg_write(struct nicpf *nic, bus_space_handle_t offset,
373     uint64_t val)
374 {
375 
376 	bus_write_8(nic->reg_base, offset, val);
377 }
378 
379 static __inline uint64_t
380 nic_reg_read(struct nicpf *nic, uint64_t offset)
381 {
382 	uint64_t val;
383 
384 	val = bus_read_8(nic->reg_base, offset);
385 	return (val);
386 }
387 
388 /* PF -> VF mailbox communication APIs */
389 static void
390 nic_enable_mbx_intr(struct nicpf *nic)
391 {
392 
393 	/* Enable mailbox interrupt for all 128 VFs */
394 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S, ~0UL);
395 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S + sizeof(uint64_t), ~0UL);
396 }
397 
398 static void
399 nic_clear_mbx_intr(struct nicpf *nic, int vf, int mbx_reg)
400 {
401 
402 	nic_reg_write(nic, NIC_PF_MAILBOX_INT + (mbx_reg << 3), (1UL << vf));
403 }
404 
405 static uint64_t
406 nic_get_mbx_addr(int vf)
407 {
408 
409 	return (NIC_PF_VF_0_127_MAILBOX_0_1 + (vf << NIC_VF_NUM_SHIFT));
410 }
411 
412 /*
413  * Send a mailbox message to VF
414  * @vf: vf to which this message to be sent
415  * @mbx: Message to be sent
416  */
417 static void
418 nic_send_msg_to_vf(struct nicpf *nic, int vf, union nic_mbx *mbx)
419 {
420 	bus_space_handle_t mbx_addr = nic_get_mbx_addr(vf);
421 	uint64_t *msg = (uint64_t *)mbx;
422 
423 	/*
424 	 * In first revision HW, mbox interrupt is triggerred
425 	 * when PF writes to MBOX(1), in next revisions when
426 	 * PF writes to MBOX(0)
427 	 */
428 	if (pass1_silicon(nic->dev)) {
429 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
430 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
431 	} else {
432 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
433 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
434 	}
435 }
436 
437 /*
438  * Responds to VF's READY message with VF's
439  * ID, node, MAC address e.t.c
440  * @vf: VF which sent READY message
441  */
442 static void
443 nic_mbx_send_ready(struct nicpf *nic, int vf)
444 {
445 	union nic_mbx mbx = {};
446 	int bgx_idx, lmac;
447 	const char *mac;
448 
449 	mbx.nic_cfg.msg = NIC_MBOX_MSG_READY;
450 	mbx.nic_cfg.vf_id = vf;
451 
452 	if (nic->flags & NIC_TNS_ENABLED)
453 		mbx.nic_cfg.tns_mode = NIC_TNS_MODE;
454 	else
455 		mbx.nic_cfg.tns_mode = NIC_TNS_BYPASS_MODE;
456 
457 	if (vf < MAX_LMAC) {
458 		bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
459 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
460 
461 		mac = bgx_get_lmac_mac(nic->node, bgx_idx, lmac);
462 		if (mac) {
463 			memcpy((uint8_t *)&mbx.nic_cfg.mac_addr, mac,
464 			    ETHER_ADDR_LEN);
465 		}
466 	}
467 	mbx.nic_cfg.node_id = nic->node;
468 
469 	mbx.nic_cfg.loopback_supported = vf < MAX_LMAC;
470 
471 	nic_send_msg_to_vf(nic, vf, &mbx);
472 }
473 
474 /*
475  * ACKs VF's mailbox message
476  * @vf: VF to which ACK to be sent
477  */
478 static void
479 nic_mbx_send_ack(struct nicpf *nic, int vf)
480 {
481 	union nic_mbx mbx = {};
482 
483 	mbx.msg.msg = NIC_MBOX_MSG_ACK;
484 	nic_send_msg_to_vf(nic, vf, &mbx);
485 }
486 
487 /*
488  * NACKs VF's mailbox message that PF is not able to
489  * complete the action
490  * @vf: VF to which ACK to be sent
491  */
492 static void
493 nic_mbx_send_nack(struct nicpf *nic, int vf)
494 {
495 	union nic_mbx mbx = {};
496 
497 	mbx.msg.msg = NIC_MBOX_MSG_NACK;
498 	nic_send_msg_to_vf(nic, vf, &mbx);
499 }
500 
501 /*
502  * Flush all in flight receive packets to memory and
503  * bring down an active RQ
504  */
505 static int
506 nic_rcv_queue_sw_sync(struct nicpf *nic)
507 {
508 	uint16_t timeout = ~0x00;
509 
510 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x01);
511 	/* Wait till sync cycle is finished */
512 	while (timeout) {
513 		if (nic_reg_read(nic, NIC_PF_SW_SYNC_RX_DONE) & 0x1)
514 			break;
515 		timeout--;
516 	}
517 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x00);
518 	if (!timeout) {
519 		device_printf(nic->dev, "Receive queue software sync failed\n");
520 		return (ETIMEDOUT);
521 	}
522 	return (0);
523 }
524 
525 /* Get BGX Rx/Tx stats and respond to VF's request */
526 static void
527 nic_get_bgx_stats(struct nicpf *nic, struct bgx_stats_msg *bgx)
528 {
529 	int bgx_idx, lmac;
530 	union nic_mbx mbx = {};
531 
532 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
533 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
534 
535 	mbx.bgx_stats.msg = NIC_MBOX_MSG_BGX_STATS;
536 	mbx.bgx_stats.vf_id = bgx->vf_id;
537 	mbx.bgx_stats.rx = bgx->rx;
538 	mbx.bgx_stats.idx = bgx->idx;
539 	if (bgx->rx != 0) {
540 		mbx.bgx_stats.stats =
541 		    bgx_get_rx_stats(nic->node, bgx_idx, lmac, bgx->idx);
542 	} else {
543 		mbx.bgx_stats.stats =
544 		    bgx_get_tx_stats(nic->node, bgx_idx, lmac, bgx->idx);
545 	}
546 	nic_send_msg_to_vf(nic, bgx->vf_id, &mbx);
547 }
548 
549 /* Update hardware min/max frame size */
550 static int
551 nic_update_hw_frs(struct nicpf *nic, int new_frs, int vf)
552 {
553 
554 	if ((new_frs > NIC_HW_MAX_FRS) || (new_frs < NIC_HW_MIN_FRS)) {
555 		device_printf(nic->dev,
556 		    "Invalid MTU setting from VF%d rejected, "
557 		    "should be between %d and %d\n",
558 		    vf, NIC_HW_MIN_FRS, NIC_HW_MAX_FRS);
559 		return (EINVAL);
560 	}
561 	new_frs += ETHER_HDR_LEN;
562 	if (new_frs <= nic->pkind.maxlen)
563 		return (0);
564 
565 	nic->pkind.maxlen = new_frs;
566 	nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG, *(uint64_t *)&nic->pkind);
567 	return (0);
568 }
569 
570 /* Set minimum transmit packet size */
571 static void
572 nic_set_tx_pkt_pad(struct nicpf *nic, int size)
573 {
574 	int lmac;
575 	uint64_t lmac_cfg;
576 
577 	/* Max value that can be set is 60 */
578 	if (size > 60)
579 		size = 60;
580 
581 	for (lmac = 0; lmac < (MAX_BGX_PER_CN88XX * MAX_LMAC_PER_BGX); lmac++) {
582 		lmac_cfg = nic_reg_read(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3));
583 		lmac_cfg &= ~(0xF << 2);
584 		lmac_cfg |= ((size / 4) << 2);
585 		nic_reg_write(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3), lmac_cfg);
586 	}
587 }
588 
589 /*
590  * Function to check number of LMACs present and set VF::LMAC mapping.
591  * Mapping will be used while initializing channels.
592  */
593 static void
594 nic_set_lmac_vf_mapping(struct nicpf *nic)
595 {
596 	unsigned bgx_map = bgx_get_map(nic->node);
597 	int bgx, next_bgx_lmac = 0;
598 	int lmac, lmac_cnt = 0;
599 	uint64_t lmac_credit;
600 
601 	nic->num_vf_en = 0;
602 	if (nic->flags & NIC_TNS_ENABLED) {
603 		nic->num_vf_en = DEFAULT_NUM_VF_ENABLED;
604 		return;
605 	}
606 
607 	for (bgx = 0; bgx < NIC_MAX_BGX; bgx++) {
608 		if ((bgx_map & (1 << bgx)) == 0)
609 			continue;
610 		lmac_cnt = bgx_get_lmac_count(nic->node, bgx);
611 		for (lmac = 0; lmac < lmac_cnt; lmac++)
612 			nic->vf_lmac_map[next_bgx_lmac++] =
613 						NIC_SET_VF_LMAC_MAP(bgx, lmac);
614 		nic->num_vf_en += lmac_cnt;
615 
616 		/* Program LMAC credits */
617 		lmac_credit = (1UL << 1); /* channel credit enable */
618 		lmac_credit |= (0x1ff << 2); /* Max outstanding pkt count */
619 		/* 48KB BGX Tx buffer size, each unit is of size 16bytes */
620 		lmac_credit |= (((((48 * 1024) / lmac_cnt) -
621 		    NIC_HW_MAX_FRS) / 16) << 12);
622 		lmac = bgx * MAX_LMAC_PER_BGX;
623 		for (; lmac < lmac_cnt + (bgx * MAX_LMAC_PER_BGX); lmac++) {
624 			nic_reg_write(nic, NIC_PF_LMAC_0_7_CREDIT + (lmac * 8),
625 			    lmac_credit);
626 		}
627 	}
628 }
629 
630 #define TNS_PORT0_BLOCK 6
631 #define TNS_PORT1_BLOCK 7
632 #define BGX0_BLOCK 8
633 #define BGX1_BLOCK 9
634 
635 static void
636 nic_init_hw(struct nicpf *nic)
637 {
638 	int i;
639 
640 	/* Enable NIC HW block */
641 	nic_reg_write(nic, NIC_PF_CFG, 0x3);
642 
643 	/* Enable backpressure */
644 	nic_reg_write(nic, NIC_PF_BP_CFG, (1UL << 6) | 0x03);
645 
646 	if (nic->flags & NIC_TNS_ENABLED) {
647 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
648 		    (NIC_TNS_MODE << 7) | TNS_PORT0_BLOCK);
649 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
650 		    (NIC_TNS_MODE << 7) | TNS_PORT1_BLOCK);
651 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
652 		    (1UL << 63) | TNS_PORT0_BLOCK);
653 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
654 		    (1UL << 63) | TNS_PORT1_BLOCK);
655 
656 	} else {
657 		/* Disable TNS mode on both interfaces */
658 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
659 		    (NIC_TNS_BYPASS_MODE << 7) | BGX0_BLOCK);
660 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
661 		    (NIC_TNS_BYPASS_MODE << 7) | BGX1_BLOCK);
662 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
663 		    (1UL << 63) | BGX0_BLOCK);
664 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
665 		    (1UL << 63) | BGX1_BLOCK);
666 	}
667 
668 	/* PKIND configuration */
669 	nic->pkind.minlen = 0;
670 	nic->pkind.maxlen = NIC_HW_MAX_FRS + ETHER_HDR_LEN;
671 	nic->pkind.lenerr_en = 1;
672 	nic->pkind.rx_hdr = 0;
673 	nic->pkind.hdr_sl = 0;
674 
675 	for (i = 0; i < NIC_MAX_PKIND; i++) {
676 		nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG | (i << 3),
677 		    *(uint64_t *)&nic->pkind);
678 	}
679 
680 	nic_set_tx_pkt_pad(nic, NIC_HW_MIN_FRS);
681 
682 	/* Timer config */
683 	nic_reg_write(nic, NIC_PF_INTR_TIMER_CFG, NICPF_CLK_PER_INT_TICK);
684 
685 	/* Enable VLAN ethertype matching and stripping */
686 	nic_reg_write(nic, NIC_PF_RX_ETYPE_0_7,
687 	    (2 << 19) | (ETYPE_ALG_VLAN_STRIP << 16) | ETHERTYPE_VLAN);
688 }
689 
690 /* Channel parse index configuration */
691 static void
692 nic_config_cpi(struct nicpf *nic, struct cpi_cfg_msg *cfg)
693 {
694 	uint32_t vnic, bgx, lmac, chan;
695 	uint32_t padd, cpi_count = 0;
696 	uint64_t cpi_base, cpi, rssi_base, rssi;
697 	uint8_t qset, rq_idx = 0;
698 
699 	vnic = cfg->vf_id;
700 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
701 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
702 
703 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
704 	cpi_base = (lmac * NIC_MAX_CPI_PER_LMAC) + (bgx * NIC_CPI_PER_BGX);
705 	rssi_base = (lmac * nic->rss_ind_tbl_size) + (bgx * NIC_RSSI_PER_BGX);
706 
707 	/* Rx channel configuration */
708 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_BP_CFG | (chan << 3),
709 	    (1UL << 63) | (vnic << 0));
710 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_CFG | (chan << 3),
711 	    ((uint64_t)cfg->cpi_alg << 62) | (cpi_base << 48));
712 
713 	if (cfg->cpi_alg == CPI_ALG_NONE)
714 		cpi_count = 1;
715 	else if (cfg->cpi_alg == CPI_ALG_VLAN) /* 3 bits of PCP */
716 		cpi_count = 8;
717 	else if (cfg->cpi_alg == CPI_ALG_VLAN16) /* 3 bits PCP + DEI */
718 		cpi_count = 16;
719 	else if (cfg->cpi_alg == CPI_ALG_DIFF) /* 6bits DSCP */
720 		cpi_count = NIC_MAX_CPI_PER_LMAC;
721 
722 	/* RSS Qset, Qidx mapping */
723 	qset = cfg->vf_id;
724 	rssi = rssi_base;
725 	for (; rssi < (rssi_base + cfg->rq_cnt); rssi++) {
726 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
727 		    (qset << 3) | rq_idx);
728 		rq_idx++;
729 	}
730 
731 	rssi = 0;
732 	cpi = cpi_base;
733 	for (; cpi < (cpi_base + cpi_count); cpi++) {
734 		/* Determine port to channel adder */
735 		if (cfg->cpi_alg != CPI_ALG_DIFF)
736 			padd = cpi % cpi_count;
737 		else
738 			padd = cpi % 8; /* 3 bits CS out of 6bits DSCP */
739 
740 		/* Leave RSS_SIZE as '0' to disable RSS */
741 		if (pass1_silicon(nic->dev)) {
742 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
743 			    (vnic << 24) | (padd << 16) | (rssi_base + rssi));
744 		} else {
745 			/* Set MPI_ALG to '0' to disable MCAM parsing */
746 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
747 			    (padd << 16));
748 			/* MPI index is same as CPI if MPI_ALG is not enabled */
749 			nic_reg_write(nic, NIC_PF_MPI_0_2047_CFG | (cpi << 3),
750 			    (vnic << 24) | (rssi_base + rssi));
751 		}
752 
753 		if ((rssi + 1) >= cfg->rq_cnt)
754 			continue;
755 
756 		if (cfg->cpi_alg == CPI_ALG_VLAN)
757 			rssi++;
758 		else if (cfg->cpi_alg == CPI_ALG_VLAN16)
759 			rssi = ((cpi - cpi_base) & 0xe) >> 1;
760 		else if (cfg->cpi_alg == CPI_ALG_DIFF)
761 			rssi = ((cpi - cpi_base) & 0x38) >> 3;
762 	}
763 	nic->cpi_base[cfg->vf_id] = cpi_base;
764 	nic->rssi_base[cfg->vf_id] = rssi_base;
765 }
766 
767 /* Responsds to VF with its RSS indirection table size */
768 static void
769 nic_send_rss_size(struct nicpf *nic, int vf)
770 {
771 	union nic_mbx mbx = {};
772 
773 	mbx.rss_size.msg = NIC_MBOX_MSG_RSS_SIZE;
774 	mbx.rss_size.ind_tbl_size = nic->rss_ind_tbl_size;
775 	nic_send_msg_to_vf(nic, vf, &mbx);
776 }
777 
778 /*
779  * Receive side scaling configuration
780  * configure:
781  * - RSS index
782  * - indir table i.e hash::RQ mapping
783  * - no of hash bits to consider
784  */
785 static void
786 nic_config_rss(struct nicpf *nic, struct rss_cfg_msg *cfg)
787 {
788 	uint8_t qset, idx;
789 	uint64_t cpi_cfg, cpi_base, rssi_base, rssi;
790 	uint64_t idx_addr;
791 
792 	idx = 0;
793 	rssi_base = nic->rssi_base[cfg->vf_id] + cfg->tbl_offset;
794 
795 	rssi = rssi_base;
796 	qset = cfg->vf_id;
797 
798 	for (; rssi < (rssi_base + cfg->tbl_len); rssi++) {
799 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
800 		    (qset << 3) | (cfg->ind_tbl[idx] & 0x7));
801 		idx++;
802 	}
803 
804 	cpi_base = nic->cpi_base[cfg->vf_id];
805 	if (pass1_silicon(nic->dev))
806 		idx_addr = NIC_PF_CPI_0_2047_CFG;
807 	else
808 		idx_addr = NIC_PF_MPI_0_2047_CFG;
809 	cpi_cfg = nic_reg_read(nic, idx_addr | (cpi_base << 3));
810 	cpi_cfg &= ~(0xFUL << 20);
811 	cpi_cfg |= (cfg->hash_bits << 20);
812 	nic_reg_write(nic, idx_addr | (cpi_base << 3), cpi_cfg);
813 }
814 
815 /*
816  * 4 level transmit side scheduler configutation
817  * for TNS bypass mode
818  *
819  * Sample configuration for SQ0
820  * VNIC0-SQ0 -> TL4(0)   -> TL3[0]   -> TL2[0]  -> TL1[0] -> BGX0
821  * VNIC1-SQ0 -> TL4(8)   -> TL3[2]   -> TL2[0]  -> TL1[0] -> BGX0
822  * VNIC2-SQ0 -> TL4(16)  -> TL3[4]   -> TL2[1]  -> TL1[0] -> BGX0
823  * VNIC3-SQ0 -> TL4(24)  -> TL3[6]   -> TL2[1]  -> TL1[0] -> BGX0
824  * VNIC4-SQ0 -> TL4(512) -> TL3[128] -> TL2[32] -> TL1[1] -> BGX1
825  * VNIC5-SQ0 -> TL4(520) -> TL3[130] -> TL2[32] -> TL1[1] -> BGX1
826  * VNIC6-SQ0 -> TL4(528) -> TL3[132] -> TL2[33] -> TL1[1] -> BGX1
827  * VNIC7-SQ0 -> TL4(536) -> TL3[134] -> TL2[33] -> TL1[1] -> BGX1
828  */
829 static void
830 nic_tx_channel_cfg(struct nicpf *nic, uint8_t vnic, struct sq_cfg_msg *sq)
831 {
832 	uint32_t bgx, lmac, chan;
833 	uint32_t tl2, tl3, tl4;
834 	uint32_t rr_quantum;
835 	uint8_t sq_idx = sq->sq_num;
836 	uint8_t pqs_vnic;
837 
838 	pqs_vnic = vnic;
839 
840 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
841 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
842 
843 	/* 24 bytes for FCS, IPG and preamble */
844 	rr_quantum = ((NIC_HW_MAX_FRS + 24) / 4);
845 
846 	tl4 = (lmac * NIC_TL4_PER_LMAC) + (bgx * NIC_TL4_PER_BGX);
847 	tl4 += sq_idx;
848 
849 	tl3 = tl4 / (NIC_MAX_TL4 / NIC_MAX_TL3);
850 	nic_reg_write(nic, NIC_PF_QSET_0_127_SQ_0_7_CFG2 |
851 	    ((uint64_t)vnic << NIC_QS_ID_SHIFT) |
852 	    ((uint32_t)sq_idx << NIC_Q_NUM_SHIFT), tl4);
853 	nic_reg_write(nic, NIC_PF_TL4_0_1023_CFG | (tl4 << 3),
854 	    ((uint64_t)vnic << 27) | ((uint32_t)sq_idx << 24) | rr_quantum);
855 
856 	nic_reg_write(nic, NIC_PF_TL3_0_255_CFG | (tl3 << 3), rr_quantum);
857 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
858 	nic_reg_write(nic, NIC_PF_TL3_0_255_CHAN | (tl3 << 3), chan);
859 	/* Enable backpressure on the channel */
860 	nic_reg_write(nic, NIC_PF_CHAN_0_255_TX_CFG | (chan << 3), 1);
861 
862 	tl2 = tl3 >> 2;
863 	nic_reg_write(nic, NIC_PF_TL3A_0_63_CFG | (tl2 << 3), tl2);
864 	nic_reg_write(nic, NIC_PF_TL2_0_63_CFG | (tl2 << 3), rr_quantum);
865 	/* No priorities as of now */
866 	nic_reg_write(nic, NIC_PF_TL2_0_63_PRI | (tl2 << 3), 0x00);
867 }
868 
869 static int
870 nic_config_loopback(struct nicpf *nic, struct set_loopback *lbk)
871 {
872 	int bgx_idx, lmac_idx;
873 
874 	if (lbk->vf_id > MAX_LMAC)
875 		return (ENXIO);
876 
877 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
878 	lmac_idx = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
879 
880 	bgx_lmac_internal_loopback(nic->node, bgx_idx, lmac_idx, lbk->enable);
881 
882 	return (0);
883 }
884 
885 /* Interrupt handler to handle mailbox messages from VFs */
886 static void
887 nic_handle_mbx_intr(struct nicpf *nic, int vf)
888 {
889 	union nic_mbx mbx = {};
890 	uint64_t *mbx_data;
891 	uint64_t mbx_addr;
892 	uint64_t reg_addr;
893 	uint64_t cfg;
894 	int bgx, lmac;
895 	int i;
896 	int ret = 0;
897 
898 	nic->mbx_lock[vf] = TRUE;
899 
900 	mbx_addr = nic_get_mbx_addr(vf);
901 	mbx_data = (uint64_t *)&mbx;
902 
903 	for (i = 0; i < NIC_PF_VF_MAILBOX_SIZE; i++) {
904 		*mbx_data = nic_reg_read(nic, mbx_addr);
905 		mbx_data++;
906 		mbx_addr += sizeof(uint64_t);
907 	}
908 
909 	switch (mbx.msg.msg) {
910 	case NIC_MBOX_MSG_READY:
911 		nic_mbx_send_ready(nic, vf);
912 		if (vf < MAX_LMAC) {
913 			nic->link[vf] = 0;
914 			nic->duplex[vf] = 0;
915 			nic->speed[vf] = 0;
916 		}
917 		ret = 1;
918 		break;
919 	case NIC_MBOX_MSG_QS_CFG:
920 		reg_addr = NIC_PF_QSET_0_127_CFG |
921 		    (mbx.qs.num << NIC_QS_ID_SHIFT);
922 		cfg = mbx.qs.cfg;
923 		nic_reg_write(nic, reg_addr, cfg);
924 		break;
925 	case NIC_MBOX_MSG_RQ_CFG:
926 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_CFG |
927 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
928 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
929 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
930 		break;
931 	case NIC_MBOX_MSG_RQ_BP_CFG:
932 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_BP_CFG |
933 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
934 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
935 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
936 		break;
937 	case NIC_MBOX_MSG_RQ_SW_SYNC:
938 		ret = nic_rcv_queue_sw_sync(nic);
939 		break;
940 	case NIC_MBOX_MSG_RQ_DROP_CFG:
941 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_DROP_CFG |
942 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
943 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
944 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
945 		break;
946 	case NIC_MBOX_MSG_SQ_CFG:
947 		reg_addr = NIC_PF_QSET_0_127_SQ_0_7_CFG |
948 		    (mbx.sq.qs_num << NIC_QS_ID_SHIFT) |
949 		    (mbx.sq.sq_num << NIC_Q_NUM_SHIFT);
950 		nic_reg_write(nic, reg_addr, mbx.sq.cfg);
951 		nic_tx_channel_cfg(nic, mbx.qs.num, &mbx.sq);
952 		break;
953 	case NIC_MBOX_MSG_SET_MAC:
954 		lmac = mbx.mac.vf_id;
955 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
956 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
957 		bgx_set_lmac_mac(nic->node, bgx, lmac, mbx.mac.mac_addr);
958 		break;
959 	case NIC_MBOX_MSG_SET_MAX_FRS:
960 		ret = nic_update_hw_frs(nic, mbx.frs.max_frs, mbx.frs.vf_id);
961 		break;
962 	case NIC_MBOX_MSG_CPI_CFG:
963 		nic_config_cpi(nic, &mbx.cpi_cfg);
964 		break;
965 	case NIC_MBOX_MSG_RSS_SIZE:
966 		nic_send_rss_size(nic, vf);
967 		goto unlock;
968 	case NIC_MBOX_MSG_RSS_CFG:
969 	case NIC_MBOX_MSG_RSS_CFG_CONT: /* fall through */
970 		nic_config_rss(nic, &mbx.rss_cfg);
971 		break;
972 	case NIC_MBOX_MSG_CFG_DONE:
973 		/* Last message of VF config msg sequence */
974 		nic->vf_info[vf].vf_enabled = TRUE;
975 		goto unlock;
976 	case NIC_MBOX_MSG_SHUTDOWN:
977 		/* First msg in VF teardown sequence */
978 		nic->vf_info[vf].vf_enabled = FALSE;
979 		break;
980 	case NIC_MBOX_MSG_BGX_STATS:
981 		nic_get_bgx_stats(nic, &mbx.bgx_stats);
982 		goto unlock;
983 	case NIC_MBOX_MSG_LOOPBACK:
984 		ret = nic_config_loopback(nic, &mbx.lbk);
985 		break;
986 	default:
987 		device_printf(nic->dev,
988 		    "Invalid msg from VF%d, msg 0x%x\n", vf, mbx.msg.msg);
989 		break;
990 	}
991 
992 	if (ret == 0)
993 		nic_mbx_send_ack(nic, vf);
994 	else if (mbx.msg.msg != NIC_MBOX_MSG_READY)
995 		nic_mbx_send_nack(nic, vf);
996 unlock:
997 	nic->mbx_lock[vf] = FALSE;
998 }
999 
1000 static void
1001 nic_mbx_intr_handler(struct nicpf *nic, int mbx)
1002 {
1003 	uint64_t intr;
1004 	uint8_t  vf, vf_per_mbx_reg = 64;
1005 
1006 	intr = nic_reg_read(nic, NIC_PF_MAILBOX_INT + (mbx << 3));
1007 	for (vf = 0; vf < vf_per_mbx_reg; vf++) {
1008 		if (intr & (1UL << vf)) {
1009 			nic_handle_mbx_intr(nic, vf + (mbx * vf_per_mbx_reg));
1010 			nic_clear_mbx_intr(nic, vf, mbx);
1011 		}
1012 	}
1013 }
1014 
1015 static void
1016 nic_mbx0_intr_handler (void *arg)
1017 {
1018 	struct nicpf *nic = (struct nicpf *)arg;
1019 
1020 	nic_mbx_intr_handler(nic, 0);
1021 }
1022 
1023 static void
1024 nic_mbx1_intr_handler (void *arg)
1025 {
1026 	struct nicpf *nic = (struct nicpf *)arg;
1027 
1028 	nic_mbx_intr_handler(nic, 1);
1029 }
1030 
1031 static int
1032 nic_enable_msix(struct nicpf *nic)
1033 {
1034 	struct pci_devinfo *dinfo;
1035 	int rid, count;
1036 	int ret;
1037 
1038 	dinfo = device_get_ivars(nic->dev);
1039 	rid = dinfo->cfg.msix.msix_table_bar;
1040 	nic->msix_table_res =
1041 	    bus_alloc_resource_any(nic->dev, SYS_RES_MEMORY, &rid, RF_ACTIVE);
1042 	if (nic->msix_table_res == NULL) {
1043 		device_printf(nic->dev,
1044 		    "Could not allocate memory for MSI-X table\n");
1045 		return (ENXIO);
1046 	}
1047 
1048 	count = nic->num_vec = NIC_PF_MSIX_VECTORS;
1049 
1050 	ret = pci_alloc_msix(nic->dev, &count);
1051 	if ((ret != 0) || (count != nic->num_vec)) {
1052 		device_printf(nic->dev,
1053 		    "Request for #%d msix vectors failed, error: %d\n",
1054 		    nic->num_vec, ret);
1055 		return (ret);
1056 	}
1057 
1058 	nic->msix_enabled = 1;
1059 	return (0);
1060 }
1061 
1062 static void
1063 nic_disable_msix(struct nicpf *nic)
1064 {
1065 	if (nic->msix_enabled) {
1066 		pci_release_msi(nic->dev);
1067 		nic->msix_enabled = 0;
1068 		nic->num_vec = 0;
1069 	}
1070 
1071 	bus_release_resource(nic->dev, SYS_RES_MEMORY,
1072 	    rman_get_rid(nic->msix_table_res), nic->msix_table_res);
1073 }
1074 
1075 static void
1076 nic_free_all_interrupts(struct nicpf *nic)
1077 {
1078 	int irq;
1079 
1080 	for (irq = 0; irq < nic->num_vec; irq++) {
1081 		if (nic->msix_entries[irq].irq_res == NULL)
1082 			continue;
1083 		if (nic->msix_entries[irq].handle != NULL) {
1084 			bus_teardown_intr(nic->dev,
1085 			    nic->msix_entries[irq].irq_res,
1086 			    nic->msix_entries[irq].handle);
1087 		}
1088 
1089 		bus_release_resource(nic->dev, SYS_RES_IRQ, irq + 1,
1090 		    nic->msix_entries[irq].irq_res);
1091 	}
1092 }
1093 
1094 static int
1095 nic_register_interrupts(struct nicpf *nic)
1096 {
1097 	int irq, rid;
1098 	int ret;
1099 
1100 	/* Enable MSI-X */
1101 	ret = nic_enable_msix(nic);
1102 	if (ret != 0)
1103 		return (ret);
1104 
1105 	/* Register mailbox interrupt handlers */
1106 	irq = NIC_PF_INTR_ID_MBOX0;
1107 	rid = irq + 1;
1108 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1109 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1110 	if (nic->msix_entries[irq].irq_res == NULL) {
1111 		ret = ENXIO;
1112 		goto fail;
1113 	}
1114 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1115 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx0_intr_handler, nic,
1116 	    &nic->msix_entries[irq].handle);
1117 	if (ret != 0)
1118 		goto fail;
1119 
1120 	irq = NIC_PF_INTR_ID_MBOX1;
1121 	rid = irq + 1;
1122 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1123 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1124 	if (nic->msix_entries[irq].irq_res == NULL) {
1125 		ret = ENXIO;
1126 		goto fail;
1127 	}
1128 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1129 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx1_intr_handler, nic,
1130 	    &nic->msix_entries[irq].handle);
1131 	if (ret != 0)
1132 		goto fail;
1133 
1134 	/* Enable mailbox interrupt */
1135 	nic_enable_mbx_intr(nic);
1136 	return (0);
1137 
1138 fail:
1139 	nic_free_all_interrupts(nic);
1140 	return (ret);
1141 }
1142 
1143 static void
1144 nic_unregister_interrupts(struct nicpf *nic)
1145 {
1146 
1147 	nic_free_all_interrupts(nic);
1148 	nic_disable_msix(nic);
1149 }
1150 
1151 static int nic_sriov_init(device_t dev, struct nicpf *nic)
1152 {
1153 #ifdef PCI_IOV
1154 	nvlist_t *pf_schema, *vf_schema;
1155 	int iov_pos;
1156 	int err;
1157 	uint16_t total_vf_cnt;
1158 
1159 	err = pci_find_extcap(dev, PCIZ_SRIOV, &iov_pos);
1160 	if (err != 0) {
1161 		device_printf(dev,
1162 		    "SR-IOV capability is not found in PCIe config space\n");
1163 		return (err);
1164 	}
1165 	/* Fix-up the number of enabled VFs */
1166 	total_vf_cnt = pci_read_config(dev, iov_pos + PCIR_SRIOV_TOTAL_VFS, 2);
1167 	if (total_vf_cnt == 0)
1168 		return (ENXIO);
1169 
1170 	/* Attach SR-IOV */
1171 	pf_schema = pci_iov_schema_alloc_node();
1172 	vf_schema = pci_iov_schema_alloc_node();
1173 	pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL);
1174 	/*
1175 	 * All VFs can change their MACs.
1176 	 * This flag will be ignored but we set it just for the record.
1177 	 */
1178 	pci_iov_schema_add_bool(vf_schema, "allow-set-mac",
1179 	    IOV_SCHEMA_HASDEFAULT, TRUE);
1180 
1181 	err = pci_iov_attach(dev, pf_schema, vf_schema);
1182 	if (err != 0) {
1183 		device_printf(dev,
1184 		    "Failed to initialize SR-IOV (error=%d)\n",
1185 		    err);
1186 		return (err);
1187 	}
1188 #endif
1189 	return (0);
1190 }
1191 
1192 /*
1193  * Poll for BGX LMAC link status and update corresponding VF
1194  * if there is a change, valid only if internal L2 switch
1195  * is not present otherwise VF link is always treated as up
1196  */
1197 static void
1198 nic_poll_for_link(void *arg)
1199 {
1200 	union nic_mbx mbx = {};
1201 	struct nicpf *nic;
1202 	struct bgx_link_status link;
1203 	uint8_t vf, bgx, lmac;
1204 
1205 	nic = (struct nicpf *)arg;
1206 
1207 	mbx.link_status.msg = NIC_MBOX_MSG_BGX_LINK_CHANGE;
1208 
1209 	for (vf = 0; vf < nic->num_vf_en; vf++) {
1210 		/* Poll only if VF is UP */
1211 		if (!nic->vf_info[vf].vf_enabled)
1212 			continue;
1213 
1214 		/* Get BGX, LMAC indices for the VF */
1215 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1216 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1217 		/* Get interface link status */
1218 		bgx_get_lmac_link_state(nic->node, bgx, lmac, &link);
1219 
1220 		/* Inform VF only if link status changed */
1221 		if (nic->link[vf] == link.link_up)
1222 			continue;
1223 
1224 		if (!nic->mbx_lock[vf]) {
1225 			nic->link[vf] = link.link_up;
1226 			nic->duplex[vf] = link.duplex;
1227 			nic->speed[vf] = link.speed;
1228 
1229 			/* Send a mbox message to VF with current link status */
1230 			mbx.link_status.link_up = link.link_up;
1231 			mbx.link_status.duplex = link.duplex;
1232 			mbx.link_status.speed = link.speed;
1233 			nic_send_msg_to_vf(nic, vf, &mbx);
1234 		}
1235 	}
1236 	callout_reset(&nic->check_link, hz * 2, nic_poll_for_link, nic);
1237 }
1238