xref: /freebsd/sys/dev/vnic/nic_main.c (revision 22cf89c938886d14f5796fc49f9f020c23ea8eaf)
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 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/bitset.h>
44 #include <sys/bitstring.h>
45 #include <sys/bus.h>
46 #include <sys/endian.h>
47 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/module.h>
50 #include <sys/rman.h>
51 #include <sys/pciio.h>
52 #include <sys/pcpu.h>
53 #include <sys/proc.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/cpuset.h>
57 #include <sys/lock.h>
58 #include <sys/mutex.h>
59 
60 #include <net/ethernet.h>
61 #include <net/if.h>
62 #include <net/if_media.h>
63 
64 #include <machine/bus.h>
65 #include <machine/_inttypes.h>
66 
67 #include <dev/pci/pcireg.h>
68 #include <dev/pci/pcivar.h>
69 
70 #include <sys/dnv.h>
71 #include <sys/nv.h>
72 #ifdef PCI_IOV
73 #include <sys/iov_schema.h>
74 #include <dev/pci/pci_iov.h>
75 #endif
76 
77 #include "thunder_bgx.h"
78 #include "nic_reg.h"
79 #include "nic.h"
80 #include "q_struct.h"
81 
82 #define	VNIC_PF_DEVSTR		"Cavium Thunder NIC Physical Function Driver"
83 
84 #define	VNIC_PF_REG_RID		PCIR_BAR(PCI_CFG_REG_BAR_NUM)
85 
86 #define	NIC_SET_VF_LMAC_MAP(bgx, lmac)		((((bgx) & 0xF) << 4) | ((lmac) & 0xF))
87 #define	NIC_GET_BGX_FROM_VF_LMAC_MAP(map)	(((map) >> 4) & 0xF)
88 #define	NIC_GET_LMAC_FROM_VF_LMAC_MAP(map)	((map) & 0xF)
89 
90 /* Structure to be used by the SR-IOV for VF configuration schemas */
91 struct nicvf_info {
92 	boolean_t		vf_enabled;
93 	int			vf_flags;
94 };
95 
96 struct nicpf {
97 	device_t		dev;
98 	uint8_t			node;
99 	u_int			flags;
100 	uint8_t			num_vf_en;      /* No of VF enabled */
101 	struct nicvf_info	vf_info[MAX_NUM_VFS_SUPPORTED];
102 	struct resource *	reg_base;       /* Register start address */
103 	struct pkind_cfg	pkind;
104 	uint8_t			vf_lmac_map[MAX_LMAC];
105 	boolean_t		mbx_lock[MAX_NUM_VFS_SUPPORTED];
106 
107 	struct callout		check_link;
108 	struct mtx		check_link_mtx;
109 
110 	uint8_t			link[MAX_LMAC];
111 	uint8_t			duplex[MAX_LMAC];
112 	uint32_t		speed[MAX_LMAC];
113 	uint16_t		cpi_base[MAX_NUM_VFS_SUPPORTED];
114 	uint16_t		rssi_base[MAX_NUM_VFS_SUPPORTED];
115 	uint16_t		rss_ind_tbl_size;
116 
117 	/* MSI-X */
118 	boolean_t		msix_enabled;
119 	uint8_t			num_vec;
120 	struct msix_entry	msix_entries[NIC_PF_MSIX_VECTORS];
121 	struct resource *	msix_table_res;
122 };
123 
124 static int nicpf_probe(device_t);
125 static int nicpf_attach(device_t);
126 static int nicpf_detach(device_t);
127 
128 #ifdef PCI_IOV
129 static int nicpf_iov_init(device_t, uint16_t, const nvlist_t *);
130 static void nicpf_iov_uninit(device_t);
131 static int nicpf_iov_add_vf(device_t, uint16_t, const nvlist_t *);
132 #endif
133 
134 static device_method_t nicpf_methods[] = {
135 	/* Device interface */
136 	DEVMETHOD(device_probe,		nicpf_probe),
137 	DEVMETHOD(device_attach,	nicpf_attach),
138 	DEVMETHOD(device_detach,	nicpf_detach),
139 	/* PCI SR-IOV interface */
140 #ifdef PCI_IOV
141 	DEVMETHOD(pci_iov_init,		nicpf_iov_init),
142 	DEVMETHOD(pci_iov_uninit,	nicpf_iov_uninit),
143 	DEVMETHOD(pci_iov_add_vf,	nicpf_iov_add_vf),
144 #endif
145 	DEVMETHOD_END,
146 };
147 
148 static driver_t vnicpf_driver = {
149 	"vnicpf",
150 	nicpf_methods,
151 	sizeof(struct nicpf),
152 };
153 
154 DRIVER_MODULE(vnicpf, pci, vnicpf_driver, 0, 0);
155 MODULE_VERSION(vnicpf, 1);
156 MODULE_DEPEND(vnicpf, pci, 1, 1, 1);
157 MODULE_DEPEND(vnicpf, ether, 1, 1, 1);
158 MODULE_DEPEND(vnicpf, thunder_bgx, 1, 1, 1);
159 
160 static int nicpf_alloc_res(struct nicpf *);
161 static void nicpf_free_res(struct nicpf *);
162 static void nic_set_lmac_vf_mapping(struct nicpf *);
163 static void nic_init_hw(struct nicpf *);
164 static int nic_sriov_init(device_t, struct nicpf *);
165 static void nic_poll_for_link(void *);
166 static int nic_register_interrupts(struct nicpf *);
167 static void nic_unregister_interrupts(struct nicpf *);
168 
169 /*
170  * Device interface
171  */
172 static int
173 nicpf_probe(device_t dev)
174 {
175 	uint16_t vendor_id;
176 	uint16_t device_id;
177 
178 	vendor_id = pci_get_vendor(dev);
179 	device_id = pci_get_device(dev);
180 
181 	if (vendor_id == PCI_VENDOR_ID_CAVIUM &&
182 	    device_id == PCI_DEVICE_ID_THUNDER_NIC_PF) {
183 		device_set_desc(dev, VNIC_PF_DEVSTR);
184 		return (BUS_PROBE_DEFAULT);
185 	}
186 
187 	return (ENXIO);
188 }
189 
190 static int
191 nicpf_attach(device_t dev)
192 {
193 	struct nicpf *nic;
194 	int err;
195 
196 	nic = device_get_softc(dev);
197 	nic->dev = dev;
198 
199 	/* Enable bus mastering */
200 	pci_enable_busmaster(dev);
201 
202 	/* Allocate PCI resources */
203 	err = nicpf_alloc_res(nic);
204 	if (err != 0) {
205 		device_printf(dev, "Could not allocate PCI resources\n");
206 		return (err);
207 	}
208 
209 	nic->node = nic_get_node_id(nic->reg_base);
210 
211 	/* Enable Traffic Network Switch (TNS) bypass mode by default */
212 	nic->flags &= ~NIC_TNS_ENABLED;
213 	nic_set_lmac_vf_mapping(nic);
214 
215 	/* Initialize hardware */
216 	nic_init_hw(nic);
217 
218 	/* Set RSS TBL size for each VF */
219 	nic->rss_ind_tbl_size = NIC_MAX_RSS_IDR_TBL_SIZE;
220 
221 	/* Setup interrupts */
222 	err = nic_register_interrupts(nic);
223 	if (err != 0)
224 		goto err_free_res;
225 
226 	/* Configure SRIOV */
227 	err = nic_sriov_init(dev, nic);
228 	if (err != 0)
229 		goto err_free_intr;
230 
231 	if (nic->flags & NIC_TNS_ENABLED)
232 		return (0);
233 
234 	mtx_init(&nic->check_link_mtx, "VNIC PF link poll", NULL, MTX_DEF);
235 	/* Register physical link status poll callout */
236 	callout_init_mtx(&nic->check_link, &nic->check_link_mtx, 0);
237 	mtx_lock(&nic->check_link_mtx);
238 	nic_poll_for_link(nic);
239 	mtx_unlock(&nic->check_link_mtx);
240 
241 	return (0);
242 
243 err_free_intr:
244 	nic_unregister_interrupts(nic);
245 err_free_res:
246 	nicpf_free_res(nic);
247 	pci_disable_busmaster(dev);
248 
249 	return (err);
250 }
251 
252 static int
253 nicpf_detach(device_t dev)
254 {
255 	struct nicpf *nic;
256 	int err;
257 
258 	err = 0;
259 	nic = device_get_softc(dev);
260 
261 	callout_drain(&nic->check_link);
262 	mtx_destroy(&nic->check_link_mtx);
263 
264 	nic_unregister_interrupts(nic);
265 	nicpf_free_res(nic);
266 	pci_disable_busmaster(dev);
267 
268 #ifdef PCI_IOV
269 	err = pci_iov_detach(dev);
270 	if (err != 0)
271 		device_printf(dev, "SR-IOV in use. Detach first.\n");
272 #endif
273 	return (err);
274 }
275 
276 /*
277  * SR-IOV interface
278  */
279 #ifdef PCI_IOV
280 static int
281 nicpf_iov_init(device_t dev, uint16_t num_vfs, const nvlist_t *params)
282 {
283 	struct nicpf *nic;
284 
285 	nic = device_get_softc(dev);
286 
287 	if (num_vfs == 0)
288 		return (ENXIO);
289 
290 	nic->flags |= NIC_SRIOV_ENABLED;
291 
292 	return (0);
293 }
294 
295 static void
296 nicpf_iov_uninit(device_t dev)
297 {
298 
299 	/* ARM64TODO: Implement this function */
300 }
301 
302 static int
303 nicpf_iov_add_vf(device_t dev, uint16_t vfnum, const nvlist_t *params)
304 {
305 	const void *mac;
306 	struct nicpf *nic;
307 	size_t size;
308 	int bgx, lmac;
309 
310 	nic = device_get_softc(dev);
311 
312 	if ((nic->flags & NIC_SRIOV_ENABLED) == 0)
313 		return (ENXIO);
314 
315 	if (vfnum > (nic->num_vf_en - 1))
316 		return (EINVAL);
317 
318 	if (nvlist_exists_binary(params, "mac-addr") != 0) {
319 		mac = nvlist_get_binary(params, "mac-addr", &size);
320 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
321 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vfnum]);
322 		bgx_set_lmac_mac(nic->node, bgx, lmac, mac);
323 	}
324 
325 	return (0);
326 }
327 #endif
328 
329 /*
330  * Helper routines
331  */
332 static int
333 nicpf_alloc_res(struct nicpf *nic)
334 {
335 	device_t dev;
336 	int rid;
337 
338 	dev = nic->dev;
339 
340 	rid = VNIC_PF_REG_RID;
341 	nic->reg_base = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
342 	    RF_ACTIVE);
343 	if (nic->reg_base == NULL) {
344 		/* For verbose output print some more details */
345 		if (bootverbose) {
346 			device_printf(dev,
347 			    "Could not allocate registers memory\n");
348 		}
349 		return (ENXIO);
350 	}
351 
352 	return (0);
353 }
354 
355 static void
356 nicpf_free_res(struct nicpf *nic)
357 {
358 	device_t dev;
359 
360 	dev = nic->dev;
361 
362 	if (nic->reg_base != NULL) {
363 		bus_release_resource(dev, SYS_RES_MEMORY,
364 		    rman_get_rid(nic->reg_base), nic->reg_base);
365 	}
366 }
367 
368 /* Register read/write APIs */
369 static __inline void
370 nic_reg_write(struct nicpf *nic, bus_space_handle_t offset,
371     uint64_t val)
372 {
373 
374 	bus_write_8(nic->reg_base, offset, val);
375 }
376 
377 static __inline uint64_t
378 nic_reg_read(struct nicpf *nic, uint64_t offset)
379 {
380 	uint64_t val;
381 
382 	val = bus_read_8(nic->reg_base, offset);
383 	return (val);
384 }
385 
386 /* PF -> VF mailbox communication APIs */
387 static void
388 nic_enable_mbx_intr(struct nicpf *nic)
389 {
390 
391 	/* Enable mailbox interrupt for all 128 VFs */
392 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S, ~0UL);
393 	nic_reg_write(nic, NIC_PF_MAILBOX_ENA_W1S + sizeof(uint64_t), ~0UL);
394 }
395 
396 static void
397 nic_clear_mbx_intr(struct nicpf *nic, int vf, int mbx_reg)
398 {
399 
400 	nic_reg_write(nic, NIC_PF_MAILBOX_INT + (mbx_reg << 3), (1UL << vf));
401 }
402 
403 static uint64_t
404 nic_get_mbx_addr(int vf)
405 {
406 
407 	return (NIC_PF_VF_0_127_MAILBOX_0_1 + (vf << NIC_VF_NUM_SHIFT));
408 }
409 
410 /*
411  * Send a mailbox message to VF
412  * @vf: vf to which this message to be sent
413  * @mbx: Message to be sent
414  */
415 static void
416 nic_send_msg_to_vf(struct nicpf *nic, int vf, union nic_mbx *mbx)
417 {
418 	bus_space_handle_t mbx_addr = nic_get_mbx_addr(vf);
419 	uint64_t *msg = (uint64_t *)mbx;
420 
421 	/*
422 	 * In first revision HW, mbox interrupt is triggerred
423 	 * when PF writes to MBOX(1), in next revisions when
424 	 * PF writes to MBOX(0)
425 	 */
426 	if (pass1_silicon(nic->dev)) {
427 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
428 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
429 	} else {
430 		nic_reg_write(nic, mbx_addr + 8, msg[1]);
431 		nic_reg_write(nic, mbx_addr + 0, msg[0]);
432 	}
433 }
434 
435 /*
436  * Responds to VF's READY message with VF's
437  * ID, node, MAC address e.t.c
438  * @vf: VF which sent READY message
439  */
440 static void
441 nic_mbx_send_ready(struct nicpf *nic, int vf)
442 {
443 	union nic_mbx mbx = {};
444 	int bgx_idx, lmac;
445 	const char *mac;
446 
447 	mbx.nic_cfg.msg = NIC_MBOX_MSG_READY;
448 	mbx.nic_cfg.vf_id = vf;
449 
450 	if (nic->flags & NIC_TNS_ENABLED)
451 		mbx.nic_cfg.tns_mode = NIC_TNS_MODE;
452 	else
453 		mbx.nic_cfg.tns_mode = NIC_TNS_BYPASS_MODE;
454 
455 	if (vf < MAX_LMAC) {
456 		bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
457 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
458 
459 		mac = bgx_get_lmac_mac(nic->node, bgx_idx, lmac);
460 		if (mac) {
461 			memcpy((uint8_t *)&mbx.nic_cfg.mac_addr, mac,
462 			    ETHER_ADDR_LEN);
463 		}
464 	}
465 	mbx.nic_cfg.node_id = nic->node;
466 
467 	mbx.nic_cfg.loopback_supported = vf < MAX_LMAC;
468 
469 	nic_send_msg_to_vf(nic, vf, &mbx);
470 }
471 
472 /*
473  * ACKs VF's mailbox message
474  * @vf: VF to which ACK to be sent
475  */
476 static void
477 nic_mbx_send_ack(struct nicpf *nic, int vf)
478 {
479 	union nic_mbx mbx = {};
480 
481 	mbx.msg.msg = NIC_MBOX_MSG_ACK;
482 	nic_send_msg_to_vf(nic, vf, &mbx);
483 }
484 
485 /*
486  * NACKs VF's mailbox message that PF is not able to
487  * complete the action
488  * @vf: VF to which ACK to be sent
489  */
490 static void
491 nic_mbx_send_nack(struct nicpf *nic, int vf)
492 {
493 	union nic_mbx mbx = {};
494 
495 	mbx.msg.msg = NIC_MBOX_MSG_NACK;
496 	nic_send_msg_to_vf(nic, vf, &mbx);
497 }
498 
499 /*
500  * Flush all in flight receive packets to memory and
501  * bring down an active RQ
502  */
503 static int
504 nic_rcv_queue_sw_sync(struct nicpf *nic)
505 {
506 	uint16_t timeout = ~0x00;
507 
508 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x01);
509 	/* Wait till sync cycle is finished */
510 	while (timeout) {
511 		if (nic_reg_read(nic, NIC_PF_SW_SYNC_RX_DONE) & 0x1)
512 			break;
513 		timeout--;
514 	}
515 	nic_reg_write(nic, NIC_PF_SW_SYNC_RX, 0x00);
516 	if (!timeout) {
517 		device_printf(nic->dev, "Receive queue software sync failed\n");
518 		return (ETIMEDOUT);
519 	}
520 	return (0);
521 }
522 
523 /* Get BGX Rx/Tx stats and respond to VF's request */
524 static void
525 nic_get_bgx_stats(struct nicpf *nic, struct bgx_stats_msg *bgx)
526 {
527 	int bgx_idx, lmac;
528 	union nic_mbx mbx = {};
529 
530 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
531 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[bgx->vf_id]);
532 
533 	mbx.bgx_stats.msg = NIC_MBOX_MSG_BGX_STATS;
534 	mbx.bgx_stats.vf_id = bgx->vf_id;
535 	mbx.bgx_stats.rx = bgx->rx;
536 	mbx.bgx_stats.idx = bgx->idx;
537 	if (bgx->rx != 0) {
538 		mbx.bgx_stats.stats =
539 		    bgx_get_rx_stats(nic->node, bgx_idx, lmac, bgx->idx);
540 	} else {
541 		mbx.bgx_stats.stats =
542 		    bgx_get_tx_stats(nic->node, bgx_idx, lmac, bgx->idx);
543 	}
544 	nic_send_msg_to_vf(nic, bgx->vf_id, &mbx);
545 }
546 
547 /* Update hardware min/max frame size */
548 static int
549 nic_update_hw_frs(struct nicpf *nic, int new_frs, int vf)
550 {
551 
552 	if ((new_frs > NIC_HW_MAX_FRS) || (new_frs < NIC_HW_MIN_FRS)) {
553 		device_printf(nic->dev,
554 		    "Invalid MTU setting from VF%d rejected, "
555 		    "should be between %d and %d\n",
556 		    vf, NIC_HW_MIN_FRS, NIC_HW_MAX_FRS);
557 		return (EINVAL);
558 	}
559 	new_frs += ETHER_HDR_LEN;
560 	if (new_frs <= nic->pkind.maxlen)
561 		return (0);
562 
563 	nic->pkind.maxlen = new_frs;
564 	nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG, *(uint64_t *)&nic->pkind);
565 	return (0);
566 }
567 
568 /* Set minimum transmit packet size */
569 static void
570 nic_set_tx_pkt_pad(struct nicpf *nic, int size)
571 {
572 	int lmac;
573 	uint64_t lmac_cfg;
574 
575 	/* Max value that can be set is 60 */
576 	if (size > 60)
577 		size = 60;
578 
579 	for (lmac = 0; lmac < (MAX_BGX_PER_CN88XX * MAX_LMAC_PER_BGX); lmac++) {
580 		lmac_cfg = nic_reg_read(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3));
581 		lmac_cfg &= ~(0xF << 2);
582 		lmac_cfg |= ((size / 4) << 2);
583 		nic_reg_write(nic, NIC_PF_LMAC_0_7_CFG | (lmac << 3), lmac_cfg);
584 	}
585 }
586 
587 /*
588  * Function to check number of LMACs present and set VF::LMAC mapping.
589  * Mapping will be used while initializing channels.
590  */
591 static void
592 nic_set_lmac_vf_mapping(struct nicpf *nic)
593 {
594 	unsigned bgx_map = bgx_get_map(nic->node);
595 	int bgx, next_bgx_lmac = 0;
596 	int lmac, lmac_cnt = 0;
597 	uint64_t lmac_credit;
598 
599 	nic->num_vf_en = 0;
600 	if (nic->flags & NIC_TNS_ENABLED) {
601 		nic->num_vf_en = DEFAULT_NUM_VF_ENABLED;
602 		return;
603 	}
604 
605 	for (bgx = 0; bgx < NIC_MAX_BGX; bgx++) {
606 		if ((bgx_map & (1 << bgx)) == 0)
607 			continue;
608 		lmac_cnt = bgx_get_lmac_count(nic->node, bgx);
609 		for (lmac = 0; lmac < lmac_cnt; lmac++)
610 			nic->vf_lmac_map[next_bgx_lmac++] =
611 						NIC_SET_VF_LMAC_MAP(bgx, lmac);
612 		nic->num_vf_en += lmac_cnt;
613 
614 		/* Program LMAC credits */
615 		lmac_credit = (1UL << 1); /* channel credit enable */
616 		lmac_credit |= (0x1ff << 2); /* Max outstanding pkt count */
617 		/* 48KB BGX Tx buffer size, each unit is of size 16bytes */
618 		lmac_credit |= (((((48 * 1024) / lmac_cnt) -
619 		    NIC_HW_MAX_FRS) / 16) << 12);
620 		lmac = bgx * MAX_LMAC_PER_BGX;
621 		for (; lmac < lmac_cnt + (bgx * MAX_LMAC_PER_BGX); lmac++) {
622 			nic_reg_write(nic, NIC_PF_LMAC_0_7_CREDIT + (lmac * 8),
623 			    lmac_credit);
624 		}
625 	}
626 }
627 
628 #define TNS_PORT0_BLOCK 6
629 #define TNS_PORT1_BLOCK 7
630 #define BGX0_BLOCK 8
631 #define BGX1_BLOCK 9
632 
633 static void
634 nic_init_hw(struct nicpf *nic)
635 {
636 	int i;
637 
638 	/* Enable NIC HW block */
639 	nic_reg_write(nic, NIC_PF_CFG, 0x3);
640 
641 	/* Enable backpressure */
642 	nic_reg_write(nic, NIC_PF_BP_CFG, (1UL << 6) | 0x03);
643 
644 	if (nic->flags & NIC_TNS_ENABLED) {
645 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
646 		    (NIC_TNS_MODE << 7) | TNS_PORT0_BLOCK);
647 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
648 		    (NIC_TNS_MODE << 7) | TNS_PORT1_BLOCK);
649 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
650 		    (1UL << 63) | TNS_PORT0_BLOCK);
651 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
652 		    (1UL << 63) | TNS_PORT1_BLOCK);
653 
654 	} else {
655 		/* Disable TNS mode on both interfaces */
656 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG,
657 		    (NIC_TNS_BYPASS_MODE << 7) | BGX0_BLOCK);
658 		nic_reg_write(nic, NIC_PF_INTF_0_1_SEND_CFG | (1 << 8),
659 		    (NIC_TNS_BYPASS_MODE << 7) | BGX1_BLOCK);
660 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG,
661 		    (1UL << 63) | BGX0_BLOCK);
662 		nic_reg_write(nic, NIC_PF_INTF_0_1_BP_CFG + (1 << 8),
663 		    (1UL << 63) | BGX1_BLOCK);
664 	}
665 
666 	/* PKIND configuration */
667 	nic->pkind.minlen = 0;
668 	nic->pkind.maxlen = NIC_HW_MAX_FRS + ETHER_HDR_LEN;
669 	nic->pkind.lenerr_en = 1;
670 	nic->pkind.rx_hdr = 0;
671 	nic->pkind.hdr_sl = 0;
672 
673 	for (i = 0; i < NIC_MAX_PKIND; i++) {
674 		nic_reg_write(nic, NIC_PF_PKIND_0_15_CFG | (i << 3),
675 		    *(uint64_t *)&nic->pkind);
676 	}
677 
678 	nic_set_tx_pkt_pad(nic, NIC_HW_MIN_FRS);
679 
680 	/* Timer config */
681 	nic_reg_write(nic, NIC_PF_INTR_TIMER_CFG, NICPF_CLK_PER_INT_TICK);
682 
683 	/* Enable VLAN ethertype matching and stripping */
684 	nic_reg_write(nic, NIC_PF_RX_ETYPE_0_7,
685 	    (2 << 19) | (ETYPE_ALG_VLAN_STRIP << 16) | ETHERTYPE_VLAN);
686 }
687 
688 /* Channel parse index configuration */
689 static void
690 nic_config_cpi(struct nicpf *nic, struct cpi_cfg_msg *cfg)
691 {
692 	uint32_t vnic, bgx, lmac, chan;
693 	uint32_t padd, cpi_count = 0;
694 	uint64_t cpi_base, cpi, rssi_base, rssi;
695 	uint8_t qset, rq_idx = 0;
696 
697 	vnic = cfg->vf_id;
698 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
699 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vnic]);
700 
701 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
702 	cpi_base = (lmac * NIC_MAX_CPI_PER_LMAC) + (bgx * NIC_CPI_PER_BGX);
703 	rssi_base = (lmac * nic->rss_ind_tbl_size) + (bgx * NIC_RSSI_PER_BGX);
704 
705 	/* Rx channel configuration */
706 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_BP_CFG | (chan << 3),
707 	    (1UL << 63) | (vnic << 0));
708 	nic_reg_write(nic, NIC_PF_CHAN_0_255_RX_CFG | (chan << 3),
709 	    ((uint64_t)cfg->cpi_alg << 62) | (cpi_base << 48));
710 
711 	if (cfg->cpi_alg == CPI_ALG_NONE)
712 		cpi_count = 1;
713 	else if (cfg->cpi_alg == CPI_ALG_VLAN) /* 3 bits of PCP */
714 		cpi_count = 8;
715 	else if (cfg->cpi_alg == CPI_ALG_VLAN16) /* 3 bits PCP + DEI */
716 		cpi_count = 16;
717 	else if (cfg->cpi_alg == CPI_ALG_DIFF) /* 6bits DSCP */
718 		cpi_count = NIC_MAX_CPI_PER_LMAC;
719 
720 	/* RSS Qset, Qidx mapping */
721 	qset = cfg->vf_id;
722 	rssi = rssi_base;
723 	for (; rssi < (rssi_base + cfg->rq_cnt); rssi++) {
724 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
725 		    (qset << 3) | rq_idx);
726 		rq_idx++;
727 	}
728 
729 	rssi = 0;
730 	cpi = cpi_base;
731 	for (; cpi < (cpi_base + cpi_count); cpi++) {
732 		/* Determine port to channel adder */
733 		if (cfg->cpi_alg != CPI_ALG_DIFF)
734 			padd = cpi % cpi_count;
735 		else
736 			padd = cpi % 8; /* 3 bits CS out of 6bits DSCP */
737 
738 		/* Leave RSS_SIZE as '0' to disable RSS */
739 		if (pass1_silicon(nic->dev)) {
740 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
741 			    (vnic << 24) | (padd << 16) | (rssi_base + rssi));
742 		} else {
743 			/* Set MPI_ALG to '0' to disable MCAM parsing */
744 			nic_reg_write(nic, NIC_PF_CPI_0_2047_CFG | (cpi << 3),
745 			    (padd << 16));
746 			/* MPI index is same as CPI if MPI_ALG is not enabled */
747 			nic_reg_write(nic, NIC_PF_MPI_0_2047_CFG | (cpi << 3),
748 			    (vnic << 24) | (rssi_base + rssi));
749 		}
750 
751 		if ((rssi + 1) >= cfg->rq_cnt)
752 			continue;
753 
754 		if (cfg->cpi_alg == CPI_ALG_VLAN)
755 			rssi++;
756 		else if (cfg->cpi_alg == CPI_ALG_VLAN16)
757 			rssi = ((cpi - cpi_base) & 0xe) >> 1;
758 		else if (cfg->cpi_alg == CPI_ALG_DIFF)
759 			rssi = ((cpi - cpi_base) & 0x38) >> 3;
760 	}
761 	nic->cpi_base[cfg->vf_id] = cpi_base;
762 	nic->rssi_base[cfg->vf_id] = rssi_base;
763 }
764 
765 /* Responsds to VF with its RSS indirection table size */
766 static void
767 nic_send_rss_size(struct nicpf *nic, int vf)
768 {
769 	union nic_mbx mbx = {};
770 
771 	mbx.rss_size.msg = NIC_MBOX_MSG_RSS_SIZE;
772 	mbx.rss_size.ind_tbl_size = nic->rss_ind_tbl_size;
773 	nic_send_msg_to_vf(nic, vf, &mbx);
774 }
775 
776 /*
777  * Receive side scaling configuration
778  * configure:
779  * - RSS index
780  * - indir table i.e hash::RQ mapping
781  * - no of hash bits to consider
782  */
783 static void
784 nic_config_rss(struct nicpf *nic, struct rss_cfg_msg *cfg)
785 {
786 	uint8_t qset, idx;
787 	uint64_t cpi_cfg, cpi_base, rssi_base, rssi;
788 	uint64_t idx_addr;
789 
790 	idx = 0;
791 	rssi_base = nic->rssi_base[cfg->vf_id] + cfg->tbl_offset;
792 
793 	rssi = rssi_base;
794 	qset = cfg->vf_id;
795 
796 	for (; rssi < (rssi_base + cfg->tbl_len); rssi++) {
797 		nic_reg_write(nic, NIC_PF_RSSI_0_4097_RQ | (rssi << 3),
798 		    (qset << 3) | (cfg->ind_tbl[idx] & 0x7));
799 		idx++;
800 	}
801 
802 	cpi_base = nic->cpi_base[cfg->vf_id];
803 	if (pass1_silicon(nic->dev))
804 		idx_addr = NIC_PF_CPI_0_2047_CFG;
805 	else
806 		idx_addr = NIC_PF_MPI_0_2047_CFG;
807 	cpi_cfg = nic_reg_read(nic, idx_addr | (cpi_base << 3));
808 	cpi_cfg &= ~(0xFUL << 20);
809 	cpi_cfg |= (cfg->hash_bits << 20);
810 	nic_reg_write(nic, idx_addr | (cpi_base << 3), cpi_cfg);
811 }
812 
813 /*
814  * 4 level transmit side scheduler configutation
815  * for TNS bypass mode
816  *
817  * Sample configuration for SQ0
818  * VNIC0-SQ0 -> TL4(0)   -> TL3[0]   -> TL2[0]  -> TL1[0] -> BGX0
819  * VNIC1-SQ0 -> TL4(8)   -> TL3[2]   -> TL2[0]  -> TL1[0] -> BGX0
820  * VNIC2-SQ0 -> TL4(16)  -> TL3[4]   -> TL2[1]  -> TL1[0] -> BGX0
821  * VNIC3-SQ0 -> TL4(24)  -> TL3[6]   -> TL2[1]  -> TL1[0] -> BGX0
822  * VNIC4-SQ0 -> TL4(512) -> TL3[128] -> TL2[32] -> TL1[1] -> BGX1
823  * VNIC5-SQ0 -> TL4(520) -> TL3[130] -> TL2[32] -> TL1[1] -> BGX1
824  * VNIC6-SQ0 -> TL4(528) -> TL3[132] -> TL2[33] -> TL1[1] -> BGX1
825  * VNIC7-SQ0 -> TL4(536) -> TL3[134] -> TL2[33] -> TL1[1] -> BGX1
826  */
827 static void
828 nic_tx_channel_cfg(struct nicpf *nic, uint8_t vnic, struct sq_cfg_msg *sq)
829 {
830 	uint32_t bgx, lmac, chan;
831 	uint32_t tl2, tl3, tl4;
832 	uint32_t rr_quantum;
833 	uint8_t sq_idx = sq->sq_num;
834 	uint8_t pqs_vnic;
835 
836 	pqs_vnic = vnic;
837 
838 	bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
839 	lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[pqs_vnic]);
840 
841 	/* 24 bytes for FCS, IPG and preamble */
842 	rr_quantum = ((NIC_HW_MAX_FRS + 24) / 4);
843 
844 	tl4 = (lmac * NIC_TL4_PER_LMAC) + (bgx * NIC_TL4_PER_BGX);
845 	tl4 += sq_idx;
846 
847 	tl3 = tl4 / (NIC_MAX_TL4 / NIC_MAX_TL3);
848 	nic_reg_write(nic, NIC_PF_QSET_0_127_SQ_0_7_CFG2 |
849 	    ((uint64_t)vnic << NIC_QS_ID_SHIFT) |
850 	    ((uint32_t)sq_idx << NIC_Q_NUM_SHIFT), tl4);
851 	nic_reg_write(nic, NIC_PF_TL4_0_1023_CFG | (tl4 << 3),
852 	    ((uint64_t)vnic << 27) | ((uint32_t)sq_idx << 24) | rr_quantum);
853 
854 	nic_reg_write(nic, NIC_PF_TL3_0_255_CFG | (tl3 << 3), rr_quantum);
855 	chan = (lmac * MAX_BGX_CHANS_PER_LMAC) + (bgx * NIC_CHANS_PER_INF);
856 	nic_reg_write(nic, NIC_PF_TL3_0_255_CHAN | (tl3 << 3), chan);
857 	/* Enable backpressure on the channel */
858 	nic_reg_write(nic, NIC_PF_CHAN_0_255_TX_CFG | (chan << 3), 1);
859 
860 	tl2 = tl3 >> 2;
861 	nic_reg_write(nic, NIC_PF_TL3A_0_63_CFG | (tl2 << 3), tl2);
862 	nic_reg_write(nic, NIC_PF_TL2_0_63_CFG | (tl2 << 3), rr_quantum);
863 	/* No priorities as of now */
864 	nic_reg_write(nic, NIC_PF_TL2_0_63_PRI | (tl2 << 3), 0x00);
865 }
866 
867 static int
868 nic_config_loopback(struct nicpf *nic, struct set_loopback *lbk)
869 {
870 	int bgx_idx, lmac_idx;
871 
872 	if (lbk->vf_id > MAX_LMAC)
873 		return (ENXIO);
874 
875 	bgx_idx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
876 	lmac_idx = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lbk->vf_id]);
877 
878 	bgx_lmac_internal_loopback(nic->node, bgx_idx, lmac_idx, lbk->enable);
879 
880 	return (0);
881 }
882 
883 /* Interrupt handler to handle mailbox messages from VFs */
884 static void
885 nic_handle_mbx_intr(struct nicpf *nic, int vf)
886 {
887 	union nic_mbx mbx = {};
888 	uint64_t *mbx_data;
889 	uint64_t mbx_addr;
890 	uint64_t reg_addr;
891 	uint64_t cfg;
892 	int bgx, lmac;
893 	int i;
894 	int ret = 0;
895 
896 	nic->mbx_lock[vf] = TRUE;
897 
898 	mbx_addr = nic_get_mbx_addr(vf);
899 	mbx_data = (uint64_t *)&mbx;
900 
901 	for (i = 0; i < NIC_PF_VF_MAILBOX_SIZE; i++) {
902 		*mbx_data = nic_reg_read(nic, mbx_addr);
903 		mbx_data++;
904 		mbx_addr += sizeof(uint64_t);
905 	}
906 
907 	switch (mbx.msg.msg) {
908 	case NIC_MBOX_MSG_READY:
909 		nic_mbx_send_ready(nic, vf);
910 		if (vf < MAX_LMAC) {
911 			nic->link[vf] = 0;
912 			nic->duplex[vf] = 0;
913 			nic->speed[vf] = 0;
914 		}
915 		ret = 1;
916 		break;
917 	case NIC_MBOX_MSG_QS_CFG:
918 		reg_addr = NIC_PF_QSET_0_127_CFG |
919 		    (mbx.qs.num << NIC_QS_ID_SHIFT);
920 		cfg = mbx.qs.cfg;
921 		nic_reg_write(nic, reg_addr, cfg);
922 		break;
923 	case NIC_MBOX_MSG_RQ_CFG:
924 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_CFG |
925 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
926 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
927 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
928 		break;
929 	case NIC_MBOX_MSG_RQ_BP_CFG:
930 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_BP_CFG |
931 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
932 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
933 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
934 		break;
935 	case NIC_MBOX_MSG_RQ_SW_SYNC:
936 		ret = nic_rcv_queue_sw_sync(nic);
937 		break;
938 	case NIC_MBOX_MSG_RQ_DROP_CFG:
939 		reg_addr = NIC_PF_QSET_0_127_RQ_0_7_DROP_CFG |
940 		    (mbx.rq.qs_num << NIC_QS_ID_SHIFT) |
941 		    (mbx.rq.rq_num << NIC_Q_NUM_SHIFT);
942 		nic_reg_write(nic, reg_addr, mbx.rq.cfg);
943 		break;
944 	case NIC_MBOX_MSG_SQ_CFG:
945 		reg_addr = NIC_PF_QSET_0_127_SQ_0_7_CFG |
946 		    (mbx.sq.qs_num << NIC_QS_ID_SHIFT) |
947 		    (mbx.sq.sq_num << NIC_Q_NUM_SHIFT);
948 		nic_reg_write(nic, reg_addr, mbx.sq.cfg);
949 		nic_tx_channel_cfg(nic, mbx.qs.num, &mbx.sq);
950 		break;
951 	case NIC_MBOX_MSG_SET_MAC:
952 		lmac = mbx.mac.vf_id;
953 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
954 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[lmac]);
955 		bgx_set_lmac_mac(nic->node, bgx, lmac, mbx.mac.mac_addr);
956 		break;
957 	case NIC_MBOX_MSG_SET_MAX_FRS:
958 		ret = nic_update_hw_frs(nic, mbx.frs.max_frs, mbx.frs.vf_id);
959 		break;
960 	case NIC_MBOX_MSG_CPI_CFG:
961 		nic_config_cpi(nic, &mbx.cpi_cfg);
962 		break;
963 	case NIC_MBOX_MSG_RSS_SIZE:
964 		nic_send_rss_size(nic, vf);
965 		goto unlock;
966 	case NIC_MBOX_MSG_RSS_CFG:
967 	case NIC_MBOX_MSG_RSS_CFG_CONT: /* fall through */
968 		nic_config_rss(nic, &mbx.rss_cfg);
969 		break;
970 	case NIC_MBOX_MSG_CFG_DONE:
971 		/* Last message of VF config msg sequence */
972 		nic->vf_info[vf].vf_enabled = TRUE;
973 		goto unlock;
974 	case NIC_MBOX_MSG_SHUTDOWN:
975 		/* First msg in VF teardown sequence */
976 		nic->vf_info[vf].vf_enabled = FALSE;
977 		break;
978 	case NIC_MBOX_MSG_BGX_STATS:
979 		nic_get_bgx_stats(nic, &mbx.bgx_stats);
980 		goto unlock;
981 	case NIC_MBOX_MSG_LOOPBACK:
982 		ret = nic_config_loopback(nic, &mbx.lbk);
983 		break;
984 	default:
985 		device_printf(nic->dev,
986 		    "Invalid msg from VF%d, msg 0x%x\n", vf, mbx.msg.msg);
987 		break;
988 	}
989 
990 	if (ret == 0)
991 		nic_mbx_send_ack(nic, vf);
992 	else if (mbx.msg.msg != NIC_MBOX_MSG_READY)
993 		nic_mbx_send_nack(nic, vf);
994 unlock:
995 	nic->mbx_lock[vf] = FALSE;
996 }
997 
998 static void
999 nic_mbx_intr_handler(struct nicpf *nic, int mbx)
1000 {
1001 	uint64_t intr;
1002 	uint8_t  vf, vf_per_mbx_reg = 64;
1003 
1004 	intr = nic_reg_read(nic, NIC_PF_MAILBOX_INT + (mbx << 3));
1005 	for (vf = 0; vf < vf_per_mbx_reg; vf++) {
1006 		if (intr & (1UL << vf)) {
1007 			nic_handle_mbx_intr(nic, vf + (mbx * vf_per_mbx_reg));
1008 			nic_clear_mbx_intr(nic, vf, mbx);
1009 		}
1010 	}
1011 }
1012 
1013 static void
1014 nic_mbx0_intr_handler (void *arg)
1015 {
1016 	struct nicpf *nic = (struct nicpf *)arg;
1017 
1018 	nic_mbx_intr_handler(nic, 0);
1019 }
1020 
1021 static void
1022 nic_mbx1_intr_handler (void *arg)
1023 {
1024 	struct nicpf *nic = (struct nicpf *)arg;
1025 
1026 	nic_mbx_intr_handler(nic, 1);
1027 }
1028 
1029 static int
1030 nic_enable_msix(struct nicpf *nic)
1031 {
1032 	struct pci_devinfo *dinfo;
1033 	int rid, count;
1034 	int ret;
1035 
1036 	dinfo = device_get_ivars(nic->dev);
1037 	rid = dinfo->cfg.msix.msix_table_bar;
1038 	nic->msix_table_res =
1039 	    bus_alloc_resource_any(nic->dev, SYS_RES_MEMORY, &rid, RF_ACTIVE);
1040 	if (nic->msix_table_res == NULL) {
1041 		device_printf(nic->dev,
1042 		    "Could not allocate memory for MSI-X table\n");
1043 		return (ENXIO);
1044 	}
1045 
1046 	count = nic->num_vec = NIC_PF_MSIX_VECTORS;
1047 
1048 	ret = pci_alloc_msix(nic->dev, &count);
1049 	if ((ret != 0) || (count != nic->num_vec)) {
1050 		device_printf(nic->dev,
1051 		    "Request for #%d msix vectors failed, error: %d\n",
1052 		    nic->num_vec, ret);
1053 		return (ret);
1054 	}
1055 
1056 	nic->msix_enabled = 1;
1057 	return (0);
1058 }
1059 
1060 static void
1061 nic_disable_msix(struct nicpf *nic)
1062 {
1063 	if (nic->msix_enabled) {
1064 		pci_release_msi(nic->dev);
1065 		nic->msix_enabled = 0;
1066 		nic->num_vec = 0;
1067 	}
1068 
1069 	bus_release_resource(nic->dev, SYS_RES_MEMORY,
1070 	    rman_get_rid(nic->msix_table_res), nic->msix_table_res);
1071 }
1072 
1073 static void
1074 nic_free_all_interrupts(struct nicpf *nic)
1075 {
1076 	int irq;
1077 
1078 	for (irq = 0; irq < nic->num_vec; irq++) {
1079 		if (nic->msix_entries[irq].irq_res == NULL)
1080 			continue;
1081 		if (nic->msix_entries[irq].handle != NULL) {
1082 			bus_teardown_intr(nic->dev,
1083 			    nic->msix_entries[irq].irq_res,
1084 			    nic->msix_entries[irq].handle);
1085 		}
1086 
1087 		bus_release_resource(nic->dev, SYS_RES_IRQ, irq + 1,
1088 		    nic->msix_entries[irq].irq_res);
1089 	}
1090 }
1091 
1092 static int
1093 nic_register_interrupts(struct nicpf *nic)
1094 {
1095 	int irq, rid;
1096 	int ret;
1097 
1098 	/* Enable MSI-X */
1099 	ret = nic_enable_msix(nic);
1100 	if (ret != 0)
1101 		return (ret);
1102 
1103 	/* Register mailbox interrupt handlers */
1104 	irq = NIC_PF_INTR_ID_MBOX0;
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_mbx0_intr_handler, nic,
1114 	    &nic->msix_entries[irq].handle);
1115 	if (ret != 0)
1116 		goto fail;
1117 
1118 	irq = NIC_PF_INTR_ID_MBOX1;
1119 	rid = irq + 1;
1120 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1121 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1122 	if (nic->msix_entries[irq].irq_res == NULL) {
1123 		ret = ENXIO;
1124 		goto fail;
1125 	}
1126 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1127 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nic_mbx1_intr_handler, nic,
1128 	    &nic->msix_entries[irq].handle);
1129 	if (ret != 0)
1130 		goto fail;
1131 
1132 	/* Enable mailbox interrupt */
1133 	nic_enable_mbx_intr(nic);
1134 	return (0);
1135 
1136 fail:
1137 	nic_free_all_interrupts(nic);
1138 	return (ret);
1139 }
1140 
1141 static void
1142 nic_unregister_interrupts(struct nicpf *nic)
1143 {
1144 
1145 	nic_free_all_interrupts(nic);
1146 	nic_disable_msix(nic);
1147 }
1148 
1149 static int nic_sriov_init(device_t dev, struct nicpf *nic)
1150 {
1151 #ifdef PCI_IOV
1152 	nvlist_t *pf_schema, *vf_schema;
1153 	int iov_pos;
1154 	int err;
1155 	uint16_t total_vf_cnt;
1156 
1157 	err = pci_find_extcap(dev, PCIZ_SRIOV, &iov_pos);
1158 	if (err != 0) {
1159 		device_printf(dev,
1160 		    "SR-IOV capability is not found in PCIe config space\n");
1161 		return (err);
1162 	}
1163 	/* Fix-up the number of enabled VFs */
1164 	total_vf_cnt = pci_read_config(dev, iov_pos + PCIR_SRIOV_TOTAL_VFS, 2);
1165 	if (total_vf_cnt == 0)
1166 		return (ENXIO);
1167 
1168 	/* Attach SR-IOV */
1169 	pf_schema = pci_iov_schema_alloc_node();
1170 	vf_schema = pci_iov_schema_alloc_node();
1171 	pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL);
1172 	/*
1173 	 * All VFs can change their MACs.
1174 	 * This flag will be ignored but we set it just for the record.
1175 	 */
1176 	pci_iov_schema_add_bool(vf_schema, "allow-set-mac",
1177 	    IOV_SCHEMA_HASDEFAULT, TRUE);
1178 
1179 	err = pci_iov_attach(dev, pf_schema, vf_schema);
1180 	if (err != 0) {
1181 		device_printf(dev,
1182 		    "Failed to initialize SR-IOV (error=%d)\n",
1183 		    err);
1184 		return (err);
1185 	}
1186 #endif
1187 	return (0);
1188 }
1189 
1190 /*
1191  * Poll for BGX LMAC link status and update corresponding VF
1192  * if there is a change, valid only if internal L2 switch
1193  * is not present otherwise VF link is always treated as up
1194  */
1195 static void
1196 nic_poll_for_link(void *arg)
1197 {
1198 	union nic_mbx mbx = {};
1199 	struct nicpf *nic;
1200 	struct bgx_link_status link;
1201 	uint8_t vf, bgx, lmac;
1202 
1203 	nic = (struct nicpf *)arg;
1204 
1205 	mbx.link_status.msg = NIC_MBOX_MSG_BGX_LINK_CHANGE;
1206 
1207 	for (vf = 0; vf < nic->num_vf_en; vf++) {
1208 		/* Poll only if VF is UP */
1209 		if (!nic->vf_info[vf].vf_enabled)
1210 			continue;
1211 
1212 		/* Get BGX, LMAC indices for the VF */
1213 		bgx = NIC_GET_BGX_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1214 		lmac = NIC_GET_LMAC_FROM_VF_LMAC_MAP(nic->vf_lmac_map[vf]);
1215 		/* Get interface link status */
1216 		bgx_get_lmac_link_state(nic->node, bgx, lmac, &link);
1217 
1218 		/* Inform VF only if link status changed */
1219 		if (nic->link[vf] == link.link_up)
1220 			continue;
1221 
1222 		if (!nic->mbx_lock[vf]) {
1223 			nic->link[vf] = link.link_up;
1224 			nic->duplex[vf] = link.duplex;
1225 			nic->speed[vf] = link.speed;
1226 
1227 			/* Send a mbox message to VF with current link status */
1228 			mbx.link_status.link_up = link.link_up;
1229 			mbx.link_status.duplex = link.duplex;
1230 			mbx.link_status.speed = link.speed;
1231 			nic_send_msg_to_vf(nic, vf, &mbx);
1232 		}
1233 	}
1234 	callout_reset(&nic->check_link, hz * 2, nic_poll_for_link, nic);
1235 }
1236