xref: /freebsd/sys/dev/vnic/nicvf_main.c (revision a5ff72cb0e51a7675d4e2b5810a2b6dad5b91960)
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 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
31 
32 #include "opt_inet.h"
33 #include "opt_inet6.h"
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/bitset.h>
38 #include <sys/bitstring.h>
39 #include <sys/bus.h>
40 #include <sys/endian.h>
41 #include <sys/kernel.h>
42 #include <sys/malloc.h>
43 #include <sys/mbuf.h>
44 #include <sys/module.h>
45 #include <sys/rman.h>
46 #include <sys/pciio.h>
47 #include <sys/pcpu.h>
48 #include <sys/proc.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/stdatomic.h>
52 #include <sys/cpuset.h>
53 #include <sys/lock.h>
54 #include <sys/mutex.h>
55 #include <sys/smp.h>
56 #include <sys/taskqueue.h>
57 
58 #include <net/bpf.h>
59 #include <net/ethernet.h>
60 #include <net/if.h>
61 #include <net/if_var.h>
62 #include <net/if_arp.h>
63 #include <net/if_dl.h>
64 #include <net/if_media.h>
65 #include <net/if_types.h>
66 #include <net/if_vlan_var.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/ip.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/tcp_lro.h>
72 
73 #include <dev/pci/pcireg.h>
74 #include <dev/pci/pcivar.h>
75 
76 #include <sys/dnv.h>
77 #include <sys/nv.h>
78 #include <sys/iov_schema.h>
79 
80 #include <machine/bus.h>
81 
82 #include "thunder_bgx.h"
83 #include "nic_reg.h"
84 #include "nic.h"
85 #include "nicvf_queues.h"
86 
87 #define	VNIC_VF_DEVSTR		"Cavium Thunder NIC Virtual Function Driver"
88 
89 #define	VNIC_VF_REG_RID		PCIR_BAR(PCI_CFG_REG_BAR_NUM)
90 
91 /* Lock for core interface settings */
92 #define	NICVF_CORE_LOCK_INIT(nic)				\
93     sx_init(&(nic)->core_sx, device_get_nameunit((nic)->dev))
94 
95 #define	NICVF_CORE_LOCK_DESTROY(nic)				\
96     sx_destroy(&(nic)->core_sx)
97 
98 #define	NICVF_CORE_LOCK(nic)		sx_xlock(&(nic)->core_sx)
99 #define	NICVF_CORE_UNLOCK(nic)		sx_xunlock(&(nic)->core_sx)
100 
101 #define	NICVF_CORE_LOCK_ASSERT(nic)	sx_assert(&(nic)->core_sx, SA_XLOCKED)
102 
103 #define	SPEED_10	10
104 #define	SPEED_100	100
105 #define	SPEED_1000	1000
106 #define	SPEED_10000	10000
107 #define	SPEED_40000	40000
108 
109 MALLOC_DEFINE(M_NICVF, "nicvf", "ThunderX VNIC VF dynamic memory");
110 
111 static int nicvf_probe(device_t);
112 static int nicvf_attach(device_t);
113 static int nicvf_detach(device_t);
114 
115 static device_method_t nicvf_methods[] = {
116 	/* Device interface */
117 	DEVMETHOD(device_probe,		nicvf_probe),
118 	DEVMETHOD(device_attach,	nicvf_attach),
119 	DEVMETHOD(device_detach,	nicvf_detach),
120 
121 	DEVMETHOD_END,
122 };
123 
124 static driver_t nicvf_driver = {
125 	"vnic",
126 	nicvf_methods,
127 	sizeof(struct nicvf),
128 };
129 
130 static devclass_t nicvf_devclass;
131 
132 DRIVER_MODULE(nicvf, pci, nicvf_driver, nicvf_devclass, 0, 0);
133 MODULE_DEPEND(nicvf, pci, 1, 1, 1);
134 MODULE_DEPEND(nicvf, ether, 1, 1, 1);
135 MODULE_DEPEND(nicvf, vnic_pf, 1, 1, 1);
136 
137 static int nicvf_allocate_misc_interrupt(struct nicvf *);
138 static int nicvf_enable_misc_interrupt(struct nicvf *);
139 static int nicvf_allocate_net_interrupts(struct nicvf *);
140 static void nicvf_release_all_interrupts(struct nicvf *);
141 static int nicvf_hw_set_mac_addr(struct nicvf *, uint8_t *);
142 static void nicvf_config_cpi(struct nicvf *);
143 static int nicvf_init_resources(struct nicvf *);
144 
145 static int nicvf_setup_ifnet(struct nicvf *);
146 static int nicvf_setup_ifmedia(struct nicvf *);
147 static void nicvf_hw_addr_random(uint8_t *);
148 
149 static int nicvf_if_ioctl(struct ifnet *, u_long, caddr_t);
150 static void nicvf_if_init(void *);
151 static void nicvf_if_init_locked(struct nicvf *);
152 static int nicvf_if_transmit(struct ifnet *, struct mbuf *);
153 static void nicvf_if_qflush(struct ifnet *);
154 static uint64_t nicvf_if_getcounter(struct ifnet *, ift_counter);
155 
156 static int nicvf_stop_locked(struct nicvf *);
157 
158 static void nicvf_media_status(struct ifnet *, struct ifmediareq *);
159 static int nicvf_media_change(struct ifnet *);
160 
161 static void nicvf_tick_stats(void *);
162 
163 static int
164 nicvf_probe(device_t dev)
165 {
166 	uint16_t vendor_id;
167 	uint16_t device_id;
168 
169 	vendor_id = pci_get_vendor(dev);
170 	device_id = pci_get_device(dev);
171 
172 	if (vendor_id != PCI_VENDOR_ID_CAVIUM)
173 		return (ENXIO);
174 
175 	if (device_id == PCI_DEVICE_ID_THUNDER_NIC_VF ||
176 	    device_id == PCI_DEVICE_ID_THUNDER_PASS1_NIC_VF) {
177 		device_set_desc(dev, VNIC_VF_DEVSTR);
178 		return (BUS_PROBE_DEFAULT);
179 	}
180 
181 	return (ENXIO);
182 }
183 
184 static int
185 nicvf_attach(device_t dev)
186 {
187 	struct nicvf *nic;
188 	int rid, qcount;
189 	int err = 0;
190 	uint8_t hwaddr[ETHER_ADDR_LEN];
191 	uint8_t zeromac[] = {[0 ... (ETHER_ADDR_LEN - 1)] = 0};
192 
193 	nic = device_get_softc(dev);
194 	nic->dev = dev;
195 	nic->pnicvf = nic;
196 
197 	NICVF_CORE_LOCK_INIT(nic);
198 	/* Enable HW TSO on Pass2 */
199 	if (!pass1_silicon(dev))
200 		nic->hw_tso = TRUE;
201 
202 	rid = VNIC_VF_REG_RID;
203 	nic->reg_base = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid,
204 	    RF_ACTIVE);
205 	if (nic->reg_base == NULL) {
206 		device_printf(dev, "Could not allocate registers memory\n");
207 		return (ENXIO);
208 	}
209 
210 	qcount = MAX_CMP_QUEUES_PER_QS;
211 	nic->max_queues = qcount;
212 
213 	err = nicvf_set_qset_resources(nic);
214 	if (err != 0)
215 		goto err_free_res;
216 
217 	/* Check if PF is alive and get MAC address for this VF */
218 	err = nicvf_allocate_misc_interrupt(nic);
219 	if (err != 0)
220 		goto err_free_res;
221 
222 	NICVF_CORE_LOCK(nic);
223 	err = nicvf_enable_misc_interrupt(nic);
224 	NICVF_CORE_UNLOCK(nic);
225 	if (err != 0)
226 		goto err_release_intr;
227 
228 	err = nicvf_allocate_net_interrupts(nic);
229 	if (err != 0) {
230 		device_printf(dev,
231 		    "Could not allocate network interface interrupts\n");
232 		goto err_free_ifnet;
233 	}
234 
235 	/* If no MAC address was obtained we generate random one */
236 	if (memcmp(nic->hwaddr, zeromac, ETHER_ADDR_LEN) == 0) {
237 		nicvf_hw_addr_random(hwaddr);
238 		memcpy(nic->hwaddr, hwaddr, ETHER_ADDR_LEN);
239 		NICVF_CORE_LOCK(nic);
240 		nicvf_hw_set_mac_addr(nic, hwaddr);
241 		NICVF_CORE_UNLOCK(nic);
242 	}
243 
244 	/* Configure CPI alorithm */
245 	nic->cpi_alg = CPI_ALG_NONE;
246 	NICVF_CORE_LOCK(nic);
247 	nicvf_config_cpi(nic);
248 	NICVF_CORE_UNLOCK(nic);
249 
250 	err = nicvf_setup_ifnet(nic);
251 	if (err != 0) {
252 		device_printf(dev, "Could not set-up ifnet\n");
253 		goto err_release_intr;
254 	}
255 
256 	err = nicvf_setup_ifmedia(nic);
257 	if (err != 0) {
258 		device_printf(dev, "Could not set-up ifmedia\n");
259 		goto err_free_ifnet;
260 	}
261 
262 	mtx_init(&nic->stats_mtx, "VNIC stats", NULL, MTX_DEF);
263 	callout_init_mtx(&nic->stats_callout, &nic->stats_mtx, 0);
264 
265 	ether_ifattach(nic->ifp, nic->hwaddr);
266 
267 	return (0);
268 
269 err_free_ifnet:
270 	if_free(nic->ifp);
271 err_release_intr:
272 	nicvf_release_all_interrupts(nic);
273 err_free_res:
274 	bus_release_resource(dev, SYS_RES_MEMORY, rman_get_rid(nic->reg_base),
275 	    nic->reg_base);
276 
277 	return (err);
278 }
279 
280 static int
281 nicvf_detach(device_t dev)
282 {
283 	struct nicvf *nic;
284 
285 	nic = device_get_softc(dev);
286 
287 	NICVF_CORE_LOCK(nic);
288 	/* Shut down the port and release ring resources */
289 	nicvf_stop_locked(nic);
290 	/* Release stats lock */
291 	mtx_destroy(&nic->stats_mtx);
292 	/* Release interrupts */
293 	nicvf_release_all_interrupts(nic);
294 	/* Release memory resource */
295 	if (nic->reg_base != NULL) {
296 		bus_release_resource(dev, SYS_RES_MEMORY,
297 		    rman_get_rid(nic->reg_base), nic->reg_base);
298 	}
299 
300 	/* Remove all ifmedia configurations */
301 	ifmedia_removeall(&nic->if_media);
302 	/* Free this ifnet */
303 	if_free(nic->ifp);
304 	NICVF_CORE_UNLOCK(nic);
305 	/* Finally destroy the lock */
306 	NICVF_CORE_LOCK_DESTROY(nic);
307 
308 	return (0);
309 }
310 
311 static void
312 nicvf_hw_addr_random(uint8_t *hwaddr)
313 {
314 	uint32_t rnd;
315 	uint8_t addr[ETHER_ADDR_LEN];
316 
317 	/*
318 	 * Create randomized MAC address.
319 	 * Set 'bsd' + random 24 low-order bits.
320 	 */
321 	rnd = arc4random() & 0x00ffffff;
322 	addr[0] = 'b';
323 	addr[1] = 's';
324 	addr[2] = 'd';
325 	addr[3] = rnd >> 16;
326 	addr[4] = rnd >> 8;
327 	addr[5] = rnd >> 0;
328 
329 	memcpy(hwaddr, addr, ETHER_ADDR_LEN);
330 }
331 
332 static int
333 nicvf_setup_ifnet(struct nicvf *nic)
334 {
335 	struct ifnet *ifp;
336 
337 	ifp = if_alloc(IFT_ETHER);
338 	if (ifp == NULL) {
339 		device_printf(nic->dev, "Could not allocate ifnet structure\n");
340 		return (ENOMEM);
341 	}
342 
343 	nic->ifp = ifp;
344 
345 	if_setsoftc(ifp, nic);
346 	if_initname(ifp, device_get_name(nic->dev), device_get_unit(nic->dev));
347 	if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX);
348 
349 	if_settransmitfn(ifp, nicvf_if_transmit);
350 	if_setqflushfn(ifp, nicvf_if_qflush);
351 	if_setioctlfn(ifp, nicvf_if_ioctl);
352 	if_setinitfn(ifp, nicvf_if_init);
353 	if_setgetcounterfn(ifp, nicvf_if_getcounter);
354 
355 	if_setmtu(ifp, ETHERMTU);
356 
357 	/* Reset caps */
358 	if_setcapabilities(ifp, 0);
359 
360 	/* Set the default values */
361 	if_setcapabilitiesbit(ifp, IFCAP_VLAN_MTU, 0);
362 	if_setcapabilitiesbit(ifp, IFCAP_LRO, 0);
363 	if (nic->hw_tso) {
364 		/* TSO */
365 		if_setcapabilitiesbit(ifp, IFCAP_TSO4, 0);
366 		/* TSO parameters */
367 		ifp->if_hw_tsomax = NICVF_TSO_MAXSIZE;
368 		ifp->if_hw_tsomaxsegcount = NICVF_TSO_NSEGS;
369 		ifp->if_hw_tsomaxsegsize = MCLBYTES;
370 	}
371 	/* IP/TCP/UDP HW checksums */
372 	if_setcapabilitiesbit(ifp, IFCAP_HWCSUM, 0);
373 	if_setcapabilitiesbit(ifp, IFCAP_HWSTATS, 0);
374 	/*
375 	 * HW offload enable
376 	 */
377 	if_clearhwassist(ifp);
378 	if_sethwassistbits(ifp, (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP), 0);
379 	if (nic->hw_tso)
380 		if_sethwassistbits(ifp, (CSUM_TSO), 0);
381 	if_setcapenable(ifp, if_getcapabilities(ifp));
382 
383 	return (0);
384 }
385 
386 static int
387 nicvf_setup_ifmedia(struct nicvf *nic)
388 {
389 
390 	ifmedia_init(&nic->if_media, IFM_IMASK, nicvf_media_change,
391 	    nicvf_media_status);
392 
393 	/*
394 	 * Advertise availability of all possible connection types,
395 	 * even though not all are possible at the same time.
396 	 */
397 
398 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_10_T | IFM_FDX),
399 	    0, NULL);
400 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_100_TX | IFM_FDX),
401 	    0, NULL);
402 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_1000_T | IFM_FDX),
403 	    0, NULL);
404 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_10G_SR | IFM_FDX),
405 	    0, NULL);
406 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_40G_CR4 | IFM_FDX),
407 	    0, NULL);
408 	ifmedia_add(&nic->if_media, (IFM_ETHER | IFM_AUTO | IFM_FDX),
409 	    0, NULL);
410 
411 	ifmedia_set(&nic->if_media, (IFM_ETHER | IFM_AUTO | IFM_FDX));
412 
413 	return (0);
414 }
415 
416 static int
417 nicvf_if_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
418 {
419 	struct nicvf *nic;
420 	struct rcv_queue *rq;
421 	struct ifreq *ifr;
422 	uint32_t flags;
423 	int mask, err;
424 	int rq_idx;
425 #if defined(INET) || defined(INET6)
426 	struct ifaddr *ifa;
427 	boolean_t avoid_reset = FALSE;
428 #endif
429 
430 	nic = if_getsoftc(ifp);
431 	ifr = (struct ifreq *)data;
432 #if defined(INET) || defined(INET6)
433 	ifa = (struct ifaddr *)data;
434 #endif
435 	err = 0;
436 	switch (cmd) {
437 	case SIOCSIFADDR:
438 #ifdef INET
439 		if (ifa->ifa_addr->sa_family == AF_INET)
440 			avoid_reset = TRUE;
441 #endif
442 #ifdef INET6
443 		if (ifa->ifa_addr->sa_family == AF_INET6)
444 			avoid_reset = TRUE;
445 #endif
446 
447 #if defined(INET) || defined(INET6)
448 		/* Avoid reinitialization unless it's necessary */
449 		if (avoid_reset) {
450 			ifp->if_flags |= IFF_UP;
451 			if (!(if_getdrvflags(ifp) & IFF_DRV_RUNNING))
452 				nicvf_if_init(nic);
453 #ifdef INET
454 			if (!(if_getflags(ifp) & IFF_NOARP))
455 				arp_ifinit(ifp, ifa);
456 #endif
457 
458 			return (0);
459 		}
460 #endif
461 		err = ether_ioctl(ifp, cmd, data);
462 		break;
463 	case SIOCSIFMTU:
464 		/*
465 		 * ARM64TODO: Needs to be implemented.
466 		 * Currently ETHERMTU is set by default.
467 		 */
468 		err = ether_ioctl(ifp, cmd, data);
469 		break;
470 	case SIOCSIFFLAGS:
471 		NICVF_CORE_LOCK(nic);
472 		if (if_getflags(ifp) & IFF_UP) {
473 			if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
474 				flags = ifp->if_flags ^ nic->if_flags;
475 				if ((nic->if_flags & ifp->if_flags) &
476 				    IFF_PROMISC) {
477 					/* Change promiscous mode */
478 #if 0
479 					/* ARM64TODO */
480 					nicvf_set_promiscous(nic);
481 #endif
482 				}
483 
484 				if ((nic->if_flags ^ ifp->if_flags) &
485 				    IFF_ALLMULTI) {
486 					/* Change multicasting settings */
487 #if 0
488 					/* ARM64TODO */
489 					nicvf_set_multicast(nic);
490 #endif
491 				}
492 			} else {
493 				nicvf_if_init_locked(nic);
494 			}
495 		} else if (if_getdrvflags(ifp) & IFF_DRV_RUNNING)
496 			nicvf_stop_locked(nic);
497 
498 		nic->if_flags = ifp->if_flags;
499 		NICVF_CORE_UNLOCK(nic);
500 		break;
501 
502 	case SIOCADDMULTI:
503 	case SIOCDELMULTI:
504 		if (if_getdrvflags(ifp) & IFF_DRV_RUNNING) {
505 #if 0
506 			NICVF_CORE_LOCK(nic);
507 			/* ARM64TODO */
508 			nicvf_set_multicast(nic);
509 			NICVF_CORE_UNLOCK(nic);
510 #endif
511 		}
512 		break;
513 
514 	case SIOCSIFMEDIA:
515 	case SIOCGIFMEDIA:
516 		err = ifmedia_ioctl(ifp, ifr, &nic->if_media, cmd);
517 		break;
518 
519 	case SIOCSIFCAP:
520 		mask = ifp->if_capenable ^ ifr->ifr_reqcap;
521 		if (mask & IFCAP_VLAN_MTU) {
522 			/* No work to do except acknowledge the change took. */
523 			ifp->if_capenable ^= IFCAP_VLAN_MTU;
524 		}
525 		if (mask & IFCAP_TXCSUM)
526 			ifp->if_capenable ^= IFCAP_TXCSUM;
527 		if (mask & IFCAP_RXCSUM)
528 			ifp->if_capenable ^= IFCAP_RXCSUM;
529 		if ((mask & IFCAP_TSO4) && nic->hw_tso)
530 			ifp->if_capenable ^= IFCAP_TSO4;
531 		if (mask & IFCAP_LRO) {
532 			/*
533 			 * Lock the driver for a moment to avoid
534 			 * mismatch in per-queue settings.
535 			 */
536 			NICVF_CORE_LOCK(nic);
537 			ifp->if_capenable ^= IFCAP_LRO;
538 			if ((if_getdrvflags(nic->ifp) & IFF_DRV_RUNNING) != 0) {
539 				/*
540 				 * Now disable LRO for subsequent packets.
541 				 * Atomicity of this change is not necessary
542 				 * as we don't need precise toggle of this
543 				 * feature for all threads processing the
544 				 * completion queue.
545 				 */
546 				for (rq_idx = 0;
547 				    rq_idx < nic->qs->rq_cnt; rq_idx++) {
548 					rq = &nic->qs->rq[rq_idx];
549 					rq->lro_enabled = !rq->lro_enabled;
550 				}
551 			}
552 			NICVF_CORE_UNLOCK(nic);
553 		}
554 
555 		break;
556 
557 	default:
558 		err = ether_ioctl(ifp, cmd, data);
559 		break;
560 	}
561 
562 	return (err);
563 }
564 
565 static void
566 nicvf_if_init_locked(struct nicvf *nic)
567 {
568 	struct queue_set *qs = nic->qs;
569 	struct ifnet *ifp;
570 	int qidx;
571 	int err;
572 	caddr_t if_addr;
573 
574 	NICVF_CORE_LOCK_ASSERT(nic);
575 	ifp = nic->ifp;
576 
577 	if ((if_getdrvflags(ifp) & IFF_DRV_RUNNING) != 0)
578 		nicvf_stop_locked(nic);
579 
580 	err = nicvf_enable_misc_interrupt(nic);
581 	if (err != 0) {
582 		if_printf(ifp, "Could not reenable Mbox interrupt\n");
583 		return;
584 	}
585 
586 	/* Get the latest MAC address */
587 	if_addr = if_getlladdr(ifp);
588 	/* Update MAC address if changed */
589 	if (memcmp(nic->hwaddr, if_addr, ETHER_ADDR_LEN) != 0) {
590 		memcpy(nic->hwaddr, if_addr, ETHER_ADDR_LEN);
591 		nicvf_hw_set_mac_addr(nic, if_addr);
592 	}
593 
594 	/* Initialize the queues */
595 	err = nicvf_init_resources(nic);
596 	if (err != 0)
597 		goto error;
598 
599 	/* Make sure queue initialization is written */
600 	wmb();
601 
602 	nicvf_reg_write(nic, NIC_VF_INT, ~0UL);
603 	/* Enable Qset err interrupt */
604 	nicvf_enable_intr(nic, NICVF_INTR_QS_ERR, 0);
605 
606 	/* Enable completion queue interrupt */
607 	for (qidx = 0; qidx < qs->cq_cnt; qidx++)
608 		nicvf_enable_intr(nic, NICVF_INTR_CQ, qidx);
609 
610 	/* Enable RBDR threshold interrupt */
611 	for (qidx = 0; qidx < qs->rbdr_cnt; qidx++)
612 		nicvf_enable_intr(nic, NICVF_INTR_RBDR, qidx);
613 
614 	nic->drv_stats.txq_stop = 0;
615 	nic->drv_stats.txq_wake = 0;
616 
617 	/* Activate network interface */
618 	if_setdrvflagbits(ifp, IFF_DRV_RUNNING, IFF_DRV_OACTIVE);
619 
620 	/* Schedule callout to update stats */
621 	callout_reset(&nic->stats_callout, hz, nicvf_tick_stats, nic);
622 
623 	return;
624 
625 error:
626 	/* Something went very wrong. Disable this ifnet for good */
627 	if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
628 }
629 
630 static void
631 nicvf_if_init(void *if_softc)
632 {
633 	struct nicvf *nic = if_softc;
634 
635 	NICVF_CORE_LOCK(nic);
636 	nicvf_if_init_locked(nic);
637 	NICVF_CORE_UNLOCK(nic);
638 }
639 
640 static int
641 nicvf_if_transmit(struct ifnet *ifp, struct mbuf *mbuf)
642 {
643 	struct nicvf *nic = if_getsoftc(ifp);
644 	struct queue_set *qs = nic->qs;
645 	struct snd_queue *sq;
646 	struct mbuf *mtmp;
647 	int qidx;
648 	int err = 0;
649 
650 
651 	if (__predict_false(qs == NULL)) {
652 		panic("%s: missing queue set for %s", __func__,
653 		    device_get_nameunit(nic->dev));
654 	}
655 
656 	/* Select queue */
657 	if (M_HASHTYPE_GET(mbuf) != M_HASHTYPE_NONE)
658 		qidx = mbuf->m_pkthdr.flowid % qs->sq_cnt;
659 	else
660 		qidx = curcpu % qs->sq_cnt;
661 
662 	sq = &qs->sq[qidx];
663 
664 	if (mbuf->m_next != NULL &&
665 	    (mbuf->m_pkthdr.csum_flags &
666 	    (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_SCTP)) != 0) {
667 		if (M_WRITABLE(mbuf) == 0) {
668 			mtmp = m_dup(mbuf, M_NOWAIT);
669 			m_freem(mbuf);
670 			if (mtmp == NULL)
671 				return (ENOBUFS);
672 			mbuf = mtmp;
673 		}
674 	}
675 
676 	err = drbr_enqueue(ifp, sq->br, mbuf);
677 	if (((if_getdrvflags(ifp) & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) !=
678 	    IFF_DRV_RUNNING) || !nic->link_up || (err != 0)) {
679 		/*
680 		 * Try to enqueue packet to the ring buffer.
681 		 * If the driver is not active, link down or enqueue operation
682 		 * failed, return with the appropriate error code.
683 		 */
684 		return (err);
685 	}
686 
687 	if (NICVF_TX_TRYLOCK(sq) != 0) {
688 		err = nicvf_xmit_locked(sq);
689 		NICVF_TX_UNLOCK(sq);
690 		return (err);
691 	} else
692 		taskqueue_enqueue(sq->snd_taskq, &sq->snd_task);
693 
694 	return (0);
695 }
696 
697 static void
698 nicvf_if_qflush(struct ifnet *ifp)
699 {
700 	struct nicvf *nic;
701 	struct queue_set *qs;
702 	struct snd_queue *sq;
703 	struct mbuf *mbuf;
704 	size_t idx;
705 
706 	nic = if_getsoftc(ifp);
707 	qs = nic->qs;
708 
709 	for (idx = 0; idx < qs->sq_cnt; idx++) {
710 		sq = &qs->sq[idx];
711 		NICVF_TX_LOCK(sq);
712 		while ((mbuf = buf_ring_dequeue_sc(sq->br)) != NULL)
713 			m_freem(mbuf);
714 		NICVF_TX_UNLOCK(sq);
715 	}
716 	if_qflush(ifp);
717 }
718 
719 static uint64_t
720 nicvf_if_getcounter(struct ifnet *ifp, ift_counter cnt)
721 {
722 	struct nicvf *nic;
723 	struct nicvf_hw_stats *hw_stats;
724 	struct nicvf_drv_stats *drv_stats;
725 
726 	nic = if_getsoftc(ifp);
727 	hw_stats = &nic->hw_stats;
728 	drv_stats = &nic->drv_stats;
729 
730 	switch (cnt) {
731 	case IFCOUNTER_IPACKETS:
732 		return (drv_stats->rx_frames_ok);
733 	case IFCOUNTER_OPACKETS:
734 		return (drv_stats->tx_frames_ok);
735 	case IFCOUNTER_IBYTES:
736 		return (hw_stats->rx_bytes);
737 	case IFCOUNTER_OBYTES:
738 		return (hw_stats->tx_bytes_ok);
739 	case IFCOUNTER_IMCASTS:
740 		return (hw_stats->rx_mcast_frames);
741 	case IFCOUNTER_COLLISIONS:
742 		return (0);
743 	case IFCOUNTER_IQDROPS:
744 		return (drv_stats->rx_drops);
745 	case IFCOUNTER_OQDROPS:
746 		return (drv_stats->tx_drops);
747 	default:
748 		return (if_get_counter_default(ifp, cnt));
749 	}
750 
751 }
752 
753 static void
754 nicvf_media_status(struct ifnet *ifp, struct ifmediareq *ifmr)
755 {
756 	struct nicvf *nic = if_getsoftc(ifp);
757 
758 	NICVF_CORE_LOCK(nic);
759 
760 	ifmr->ifm_status = IFM_AVALID;
761 	ifmr->ifm_active = IFM_ETHER;
762 
763 	if (nic->link_up) {
764 		/* Device attached to working network */
765 		ifmr->ifm_status |= IFM_ACTIVE;
766 	}
767 
768 	switch (nic->speed) {
769 	case SPEED_10:
770 		ifmr->ifm_active |= IFM_10_T;
771 		break;
772 	case SPEED_100:
773 		ifmr->ifm_active |= IFM_100_TX;
774 		break;
775 	case SPEED_1000:
776 		ifmr->ifm_active |= IFM_1000_T;
777 		break;
778 	case SPEED_10000:
779 		ifmr->ifm_active |= IFM_10G_SR;
780 		break;
781 	case SPEED_40000:
782 		ifmr->ifm_active |= IFM_40G_CR4;
783 		break;
784 	default:
785 		ifmr->ifm_active |= IFM_AUTO;
786 		break;
787 	}
788 
789 	if (nic->duplex)
790 		ifmr->ifm_active |= IFM_FDX;
791 	else
792 		ifmr->ifm_active |= IFM_HDX;
793 
794 	NICVF_CORE_UNLOCK(nic);
795 }
796 
797 static int
798 nicvf_media_change(struct ifnet *ifp __unused)
799 {
800 
801 	return (0);
802 }
803 
804 /* Register read/write APIs */
805 void
806 nicvf_reg_write(struct nicvf *nic, bus_space_handle_t offset, uint64_t val)
807 {
808 
809 	bus_write_8(nic->reg_base, offset, val);
810 }
811 
812 uint64_t
813 nicvf_reg_read(struct nicvf *nic, uint64_t offset)
814 {
815 
816 	return (bus_read_8(nic->reg_base, offset));
817 }
818 
819 void
820 nicvf_queue_reg_write(struct nicvf *nic, bus_space_handle_t offset,
821     uint64_t qidx, uint64_t val)
822 {
823 
824 	bus_write_8(nic->reg_base, offset + (qidx << NIC_Q_NUM_SHIFT), val);
825 }
826 
827 uint64_t
828 nicvf_queue_reg_read(struct nicvf *nic, bus_space_handle_t offset,
829     uint64_t qidx)
830 {
831 
832 	return (bus_read_8(nic->reg_base, offset + (qidx << NIC_Q_NUM_SHIFT)));
833 }
834 
835 /* VF -> PF mailbox communication */
836 static void
837 nicvf_write_to_mbx(struct nicvf *nic, union nic_mbx *mbx)
838 {
839 	uint64_t *msg = (uint64_t *)mbx;
840 
841 	nicvf_reg_write(nic, NIC_VF_PF_MAILBOX_0_1 + 0, msg[0]);
842 	nicvf_reg_write(nic, NIC_VF_PF_MAILBOX_0_1 + 8, msg[1]);
843 }
844 
845 int
846 nicvf_send_msg_to_pf(struct nicvf *nic, union nic_mbx *mbx)
847 {
848 	int timeout = NIC_MBOX_MSG_TIMEOUT * 10;
849 	int sleep = 2;
850 
851 	NICVF_CORE_LOCK_ASSERT(nic);
852 
853 	nic->pf_acked = FALSE;
854 	nic->pf_nacked = FALSE;
855 
856 	nicvf_write_to_mbx(nic, mbx);
857 
858 	/* Wait for previous message to be acked, timeout 2sec */
859 	while (!nic->pf_acked) {
860 		if (nic->pf_nacked)
861 			return (EINVAL);
862 
863 		DELAY(sleep * 1000);
864 
865 		if (nic->pf_acked)
866 			break;
867 		timeout -= sleep;
868 		if (!timeout) {
869 			device_printf(nic->dev,
870 				   "PF didn't ack to mbox msg %d from VF%d\n",
871 				   (mbx->msg.msg & 0xFF), nic->vf_id);
872 
873 			return (EBUSY);
874 		}
875 	}
876 	return (0);
877 }
878 
879 /*
880  * Checks if VF is able to comminicate with PF
881  * and also gets the VNIC number this VF is associated to.
882  */
883 static int
884 nicvf_check_pf_ready(struct nicvf *nic)
885 {
886 	union nic_mbx mbx = {};
887 
888 	mbx.msg.msg = NIC_MBOX_MSG_READY;
889 	if (nicvf_send_msg_to_pf(nic, &mbx)) {
890 		device_printf(nic->dev,
891 			   "PF didn't respond to READY msg\n");
892 		return 0;
893 	}
894 
895 	return 1;
896 }
897 
898 static void
899 nicvf_read_bgx_stats(struct nicvf *nic, struct bgx_stats_msg *bgx)
900 {
901 
902 	if (bgx->rx)
903 		nic->bgx_stats.rx_stats[bgx->idx] = bgx->stats;
904 	else
905 		nic->bgx_stats.tx_stats[bgx->idx] = bgx->stats;
906 }
907 
908 static void
909 nicvf_handle_mbx_intr(struct nicvf *nic)
910 {
911 	union nic_mbx mbx = {};
912 	uint64_t *mbx_data;
913 	uint64_t mbx_addr;
914 	int i;
915 
916 	mbx_addr = NIC_VF_PF_MAILBOX_0_1;
917 	mbx_data = (uint64_t *)&mbx;
918 
919 	for (i = 0; i < NIC_PF_VF_MAILBOX_SIZE; i++) {
920 		*mbx_data = nicvf_reg_read(nic, mbx_addr);
921 		mbx_data++;
922 		mbx_addr += sizeof(uint64_t);
923 	}
924 
925 	switch (mbx.msg.msg) {
926 	case NIC_MBOX_MSG_READY:
927 		nic->pf_acked = TRUE;
928 		nic->vf_id = mbx.nic_cfg.vf_id & 0x7F;
929 		nic->tns_mode = mbx.nic_cfg.tns_mode & 0x7F;
930 		nic->node = mbx.nic_cfg.node_id;
931 		memcpy(nic->hwaddr, mbx.nic_cfg.mac_addr, ETHER_ADDR_LEN);
932 		nic->loopback_supported = mbx.nic_cfg.loopback_supported;
933 		nic->link_up = FALSE;
934 		nic->duplex = 0;
935 		nic->speed = 0;
936 		break;
937 	case NIC_MBOX_MSG_ACK:
938 		nic->pf_acked = TRUE;
939 		break;
940 	case NIC_MBOX_MSG_NACK:
941 		nic->pf_nacked = TRUE;
942 		break;
943 	case NIC_MBOX_MSG_BGX_STATS:
944 		nicvf_read_bgx_stats(nic, &mbx.bgx_stats);
945 		nic->pf_acked = TRUE;
946 		break;
947 	case NIC_MBOX_MSG_BGX_LINK_CHANGE:
948 		nic->pf_acked = TRUE;
949 		nic->link_up = mbx.link_status.link_up;
950 		nic->duplex = mbx.link_status.duplex;
951 		nic->speed = mbx.link_status.speed;
952 		if (nic->link_up) {
953 			if_setbaudrate(nic->ifp, nic->speed * 1000000);
954 			if_link_state_change(nic->ifp, LINK_STATE_UP);
955 		} else {
956 			if_setbaudrate(nic->ifp, 0);
957 			if_link_state_change(nic->ifp, LINK_STATE_DOWN);
958 		}
959 		break;
960 	default:
961 		device_printf(nic->dev,
962 			   "Invalid message from PF, msg 0x%x\n", mbx.msg.msg);
963 		break;
964 	}
965 	nicvf_clear_intr(nic, NICVF_INTR_MBOX, 0);
966 }
967 
968 static int
969 nicvf_hw_set_mac_addr(struct nicvf *nic, uint8_t *hwaddr)
970 {
971 	union nic_mbx mbx = {};
972 
973 	mbx.mac.msg = NIC_MBOX_MSG_SET_MAC;
974 	mbx.mac.vf_id = nic->vf_id;
975 	memcpy(mbx.mac.mac_addr, hwaddr, ETHER_ADDR_LEN);
976 
977 	return (nicvf_send_msg_to_pf(nic, &mbx));
978 }
979 
980 static void
981 nicvf_config_cpi(struct nicvf *nic)
982 {
983 	union nic_mbx mbx = {};
984 
985 	mbx.cpi_cfg.msg = NIC_MBOX_MSG_CPI_CFG;
986 	mbx.cpi_cfg.vf_id = nic->vf_id;
987 	mbx.cpi_cfg.cpi_alg = nic->cpi_alg;
988 	mbx.cpi_cfg.rq_cnt = nic->qs->rq_cnt;
989 
990 	nicvf_send_msg_to_pf(nic, &mbx);
991 }
992 
993 static int
994 nicvf_init_resources(struct nicvf *nic)
995 {
996 	int err;
997 	union nic_mbx mbx = {};
998 
999 	mbx.msg.msg = NIC_MBOX_MSG_CFG_DONE;
1000 
1001 	/* Enable Qset */
1002 	nicvf_qset_config(nic, TRUE);
1003 
1004 	/* Initialize queues and HW for data transfer */
1005 	err = nicvf_config_data_transfer(nic, TRUE);
1006 	if (err) {
1007 		device_printf(nic->dev,
1008 		    "Failed to alloc/config VF's QSet resources\n");
1009 		return (err);
1010 	}
1011 
1012 	/* Send VF config done msg to PF */
1013 	nicvf_write_to_mbx(nic, &mbx);
1014 
1015 	return (0);
1016 }
1017 
1018 static void
1019 nicvf_misc_intr_handler(void *arg)
1020 {
1021 	struct nicvf *nic = (struct nicvf *)arg;
1022 	uint64_t intr;
1023 
1024 	intr = nicvf_reg_read(nic, NIC_VF_INT);
1025 	/* Check for spurious interrupt */
1026 	if (!(intr & NICVF_INTR_MBOX_MASK))
1027 		return;
1028 
1029 	nicvf_handle_mbx_intr(nic);
1030 }
1031 
1032 static int
1033 nicvf_intr_handler(void *arg)
1034 {
1035 	struct nicvf *nic;
1036 	struct cmp_queue *cq;
1037 	int qidx;
1038 
1039 	cq = (struct cmp_queue *)arg;
1040 	nic = cq->nic;
1041 	qidx = cq->idx;
1042 
1043 	/* Disable interrupts */
1044 	nicvf_disable_intr(nic, NICVF_INTR_CQ, qidx);
1045 
1046 	taskqueue_enqueue(cq->cmp_taskq, &cq->cmp_task);
1047 
1048 	/* Clear interrupt */
1049 	nicvf_clear_intr(nic, NICVF_INTR_CQ, qidx);
1050 
1051 	return (FILTER_HANDLED);
1052 }
1053 
1054 static void
1055 nicvf_rbdr_intr_handler(void *arg)
1056 {
1057 	struct nicvf *nic;
1058 	struct queue_set *qs;
1059 	struct rbdr *rbdr;
1060 	int qidx;
1061 
1062 	nic = (struct nicvf *)arg;
1063 
1064 	/* Disable RBDR interrupt and schedule softirq */
1065 	for (qidx = 0; qidx < nic->qs->rbdr_cnt; qidx++) {
1066 		if (!nicvf_is_intr_enabled(nic, NICVF_INTR_RBDR, qidx))
1067 			continue;
1068 		nicvf_disable_intr(nic, NICVF_INTR_RBDR, qidx);
1069 
1070 		qs = nic->qs;
1071 		rbdr = &qs->rbdr[qidx];
1072 		taskqueue_enqueue(rbdr->rbdr_taskq, &rbdr->rbdr_task_nowait);
1073 		/* Clear interrupt */
1074 		nicvf_clear_intr(nic, NICVF_INTR_RBDR, qidx);
1075 	}
1076 }
1077 
1078 static void
1079 nicvf_qs_err_intr_handler(void *arg)
1080 {
1081 	struct nicvf *nic = (struct nicvf *)arg;
1082 	struct queue_set *qs = nic->qs;
1083 
1084 	/* Disable Qset err interrupt and schedule softirq */
1085 	nicvf_disable_intr(nic, NICVF_INTR_QS_ERR, 0);
1086 	taskqueue_enqueue(qs->qs_err_taskq, &qs->qs_err_task);
1087 	nicvf_clear_intr(nic, NICVF_INTR_QS_ERR, 0);
1088 
1089 }
1090 
1091 static int
1092 nicvf_enable_msix(struct nicvf *nic)
1093 {
1094 	struct pci_devinfo *dinfo;
1095 	int rid, count;
1096 	int ret;
1097 
1098 	dinfo = device_get_ivars(nic->dev);
1099 	rid = dinfo->cfg.msix.msix_table_bar;
1100 	nic->msix_table_res =
1101 	    bus_alloc_resource_any(nic->dev, SYS_RES_MEMORY, &rid, RF_ACTIVE);
1102 	if (nic->msix_table_res == NULL) {
1103 		device_printf(nic->dev,
1104 		    "Could not allocate memory for MSI-X table\n");
1105 		return (ENXIO);
1106 	}
1107 
1108 	count = nic->num_vec = NIC_VF_MSIX_VECTORS;
1109 
1110 	ret = pci_alloc_msix(nic->dev, &count);
1111 	if ((ret != 0) || (count != nic->num_vec)) {
1112 		device_printf(nic->dev,
1113 		    "Request for #%d msix vectors failed, error: %d\n",
1114 		    nic->num_vec, ret);
1115 		return (ret);
1116 	}
1117 
1118 	nic->msix_enabled = 1;
1119 	return (0);
1120 }
1121 
1122 static void
1123 nicvf_disable_msix(struct nicvf *nic)
1124 {
1125 
1126 	if (nic->msix_enabled) {
1127 		pci_release_msi(nic->dev);
1128 		nic->msix_enabled = 0;
1129 		nic->num_vec = 0;
1130 	}
1131 }
1132 
1133 static void
1134 nicvf_release_all_interrupts(struct nicvf *nic)
1135 {
1136 	struct resource *res;
1137 	int irq;
1138 	int err;
1139 
1140 	/* Free registered interrupts */
1141 	for (irq = 0; irq < nic->num_vec; irq++) {
1142 		res = nic->msix_entries[irq].irq_res;
1143 		if (res == NULL)
1144 			continue;
1145 		/* Teardown interrupt first */
1146 		if (nic->msix_entries[irq].handle != NULL) {
1147 			err = bus_teardown_intr(nic->dev,
1148 			    nic->msix_entries[irq].irq_res,
1149 			    nic->msix_entries[irq].handle);
1150 			KASSERT(err == 0,
1151 			    ("ERROR: Unable to teardown interrupt %d", irq));
1152 			nic->msix_entries[irq].handle = NULL;
1153 		}
1154 
1155 		bus_release_resource(nic->dev, SYS_RES_IRQ,
1156 			    rman_get_rid(res), nic->msix_entries[irq].irq_res);
1157 		nic->msix_entries[irq].irq_res = NULL;
1158 	}
1159 	/* Disable MSI-X */
1160 	nicvf_disable_msix(nic);
1161 }
1162 
1163 /*
1164  * Initialize MSIX vectors and register MISC interrupt.
1165  * Send READY message to PF to check if its alive
1166  */
1167 static int
1168 nicvf_allocate_misc_interrupt(struct nicvf *nic)
1169 {
1170 	struct resource *res;
1171 	int irq, rid;
1172 	int ret = 0;
1173 
1174 	/* Return if mailbox interrupt is already registered */
1175 	if (nic->msix_enabled)
1176 		return (0);
1177 
1178 	/* Enable MSI-X */
1179 	if (nicvf_enable_msix(nic) != 0)
1180 		return (ENXIO);
1181 
1182 	irq = NICVF_INTR_ID_MISC;
1183 	rid = irq + 1;
1184 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1185 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1186 	if (nic->msix_entries[irq].irq_res == NULL) {
1187 		device_printf(nic->dev,
1188 		    "Could not allocate Mbox interrupt for VF%d\n",
1189 		    device_get_unit(nic->dev));
1190 		return (ENXIO);
1191 	}
1192 
1193 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1194 	    (INTR_MPSAFE | INTR_TYPE_MISC), NULL, nicvf_misc_intr_handler, nic,
1195 	    &nic->msix_entries[irq].handle);
1196 	if (ret != 0) {
1197 		res = nic->msix_entries[irq].irq_res;
1198 		bus_release_resource(nic->dev, SYS_RES_IRQ,
1199 			    rman_get_rid(res), res);
1200 		nic->msix_entries[irq].irq_res = NULL;
1201 		return (ret);
1202 	}
1203 
1204 	return (0);
1205 }
1206 
1207 static int
1208 nicvf_enable_misc_interrupt(struct nicvf *nic)
1209 {
1210 
1211 	/* Enable mailbox interrupt */
1212 	nicvf_enable_intr(nic, NICVF_INTR_MBOX, 0);
1213 
1214 	/* Check if VF is able to communicate with PF */
1215 	if (!nicvf_check_pf_ready(nic)) {
1216 		nicvf_disable_intr(nic, NICVF_INTR_MBOX, 0);
1217 		return (ENXIO);
1218 	}
1219 
1220 	return (0);
1221 }
1222 
1223 static void
1224 nicvf_release_net_interrupts(struct nicvf *nic)
1225 {
1226 	struct resource *res;
1227 	int irq;
1228 	int err;
1229 
1230 	for_each_cq_irq(irq) {
1231 		res = nic->msix_entries[irq].irq_res;
1232 		if (res == NULL)
1233 			continue;
1234 		/* Teardown active interrupts first */
1235 		if (nic->msix_entries[irq].handle != NULL) {
1236 			err = bus_teardown_intr(nic->dev,
1237 			    nic->msix_entries[irq].irq_res,
1238 			    nic->msix_entries[irq].handle);
1239 			KASSERT(err == 0,
1240 			    ("ERROR: Unable to teardown CQ interrupt %d",
1241 			    (irq - NICVF_INTR_ID_CQ)));
1242 			if (err != 0)
1243 				continue;
1244 		}
1245 
1246 		/* Release resource */
1247 		bus_release_resource(nic->dev, SYS_RES_IRQ, rman_get_rid(res),
1248 		    res);
1249 		nic->msix_entries[irq].irq_res = NULL;
1250 	}
1251 
1252 	for_each_rbdr_irq(irq) {
1253 		res = nic->msix_entries[irq].irq_res;
1254 		if (res == NULL)
1255 			continue;
1256 		/* Teardown active interrupts first */
1257 		if (nic->msix_entries[irq].handle != NULL) {
1258 			err = bus_teardown_intr(nic->dev,
1259 			    nic->msix_entries[irq].irq_res,
1260 			    nic->msix_entries[irq].handle);
1261 			KASSERT(err == 0,
1262 			    ("ERROR: Unable to teardown RDBR interrupt %d",
1263 			    (irq - NICVF_INTR_ID_RBDR)));
1264 			if (err != 0)
1265 				continue;
1266 		}
1267 
1268 		/* Release resource */
1269 		bus_release_resource(nic->dev, SYS_RES_IRQ, rman_get_rid(res),
1270 		    res);
1271 		nic->msix_entries[irq].irq_res = NULL;
1272 	}
1273 
1274 	irq = NICVF_INTR_ID_QS_ERR;
1275 	res = nic->msix_entries[irq].irq_res;
1276 	if (res != NULL) {
1277 		/* Teardown active interrupts first */
1278 		if (nic->msix_entries[irq].handle != NULL) {
1279 			err = bus_teardown_intr(nic->dev,
1280 			    nic->msix_entries[irq].irq_res,
1281 			    nic->msix_entries[irq].handle);
1282 			KASSERT(err == 0,
1283 			    ("ERROR: Unable to teardown QS Error interrupt %d",
1284 			    irq));
1285 			if (err != 0)
1286 				return;
1287 		}
1288 
1289 		/* Release resource */
1290 		bus_release_resource(nic->dev, SYS_RES_IRQ, rman_get_rid(res),
1291 		    res);
1292 		nic->msix_entries[irq].irq_res = NULL;
1293 	}
1294 }
1295 
1296 static int
1297 nicvf_allocate_net_interrupts(struct nicvf *nic)
1298 {
1299 	int irq, rid;
1300 	int qidx;
1301 	int ret = 0;
1302 
1303 	/* MSI-X must be configured by now */
1304 	if (!nic->msix_enabled) {
1305 		device_printf(nic->dev, "Cannot alloacte queue interrups. "
1306 		    "MSI-X interrupts disabled.\n");
1307 		return (ENXIO);
1308 	}
1309 
1310 	/* Register CQ interrupts */
1311 	for_each_cq_irq(irq) {
1312 		if (irq >= (NICVF_INTR_ID_CQ + nic->qs->cq_cnt))
1313 			break;
1314 
1315 		qidx = irq - NICVF_INTR_ID_CQ;
1316 		rid = irq + 1;
1317 		nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1318 		    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1319 		if (nic->msix_entries[irq].irq_res == NULL) {
1320 			device_printf(nic->dev,
1321 			    "Could not allocate CQ interrupt %d for VF%d\n",
1322 			    (irq - NICVF_INTR_ID_CQ), device_get_unit(nic->dev));
1323 			ret = ENXIO;
1324 			goto error;
1325 		}
1326 		ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1327 		    (INTR_MPSAFE | INTR_TYPE_NET), nicvf_intr_handler,
1328 		    NULL, &nic->qs->cq[qidx], &nic->msix_entries[irq].handle);
1329 		if (ret != 0) {
1330 			device_printf(nic->dev,
1331 			    "Could not setup CQ interrupt %d for VF%d\n",
1332 			    (irq - NICVF_INTR_ID_CQ), device_get_unit(nic->dev));
1333 			goto error;
1334 		}
1335 	}
1336 
1337 	/* Register RBDR interrupt */
1338 	for_each_rbdr_irq(irq) {
1339 		if (irq >= (NICVF_INTR_ID_RBDR + nic->qs->rbdr_cnt))
1340 			break;
1341 
1342 		rid = irq + 1;
1343 		nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1344 		    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1345 		if (nic->msix_entries[irq].irq_res == NULL) {
1346 			device_printf(nic->dev,
1347 			    "Could not allocate RBDR interrupt %d for VF%d\n",
1348 			    (irq - NICVF_INTR_ID_RBDR),
1349 			    device_get_unit(nic->dev));
1350 			ret = ENXIO;
1351 			goto error;
1352 		}
1353 		ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1354 		    (INTR_MPSAFE | INTR_TYPE_NET), NULL,
1355 		    nicvf_rbdr_intr_handler, nic,
1356 		    &nic->msix_entries[irq].handle);
1357 		if (ret != 0) {
1358 			device_printf(nic->dev,
1359 			    "Could not setup RBDR interrupt %d for VF%d\n",
1360 			    (irq - NICVF_INTR_ID_RBDR),
1361 			    device_get_unit(nic->dev));
1362 			goto error;
1363 		}
1364 	}
1365 
1366 	/* Register QS error interrupt */
1367 	irq = NICVF_INTR_ID_QS_ERR;
1368 	rid = irq + 1;
1369 	nic->msix_entries[irq].irq_res = bus_alloc_resource_any(nic->dev,
1370 	    SYS_RES_IRQ, &rid, (RF_SHAREABLE | RF_ACTIVE));
1371 	if (nic->msix_entries[irq].irq_res == NULL) {
1372 		device_printf(nic->dev,
1373 		    "Could not allocate QS Error interrupt for VF%d\n",
1374 		    device_get_unit(nic->dev));
1375 		ret = ENXIO;
1376 		goto error;
1377 	}
1378 	ret = bus_setup_intr(nic->dev, nic->msix_entries[irq].irq_res,
1379 	    (INTR_MPSAFE | INTR_TYPE_NET), NULL, nicvf_qs_err_intr_handler,
1380 	    nic, &nic->msix_entries[irq].handle);
1381 	if (ret != 0) {
1382 		device_printf(nic->dev,
1383 		    "Could not setup QS Error interrupt for VF%d\n",
1384 		    device_get_unit(nic->dev));
1385 		goto error;
1386 	}
1387 
1388 	return (0);
1389 error:
1390 	nicvf_release_net_interrupts(nic);
1391 	return (ret);
1392 }
1393 
1394 static int
1395 nicvf_stop_locked(struct nicvf *nic)
1396 {
1397 	struct ifnet *ifp;
1398 	int qidx;
1399 	struct queue_set *qs = nic->qs;
1400 	union nic_mbx mbx = {};
1401 
1402 	NICVF_CORE_LOCK_ASSERT(nic);
1403 	/* Stop callout. Can block here since holding SX lock */
1404 	callout_drain(&nic->stats_callout);
1405 
1406 	ifp = nic->ifp;
1407 
1408 	mbx.msg.msg = NIC_MBOX_MSG_SHUTDOWN;
1409 	nicvf_send_msg_to_pf(nic, &mbx);
1410 
1411 	/* Disable RBDR & QS error interrupts */
1412 	for (qidx = 0; qidx < qs->rbdr_cnt; qidx++) {
1413 		nicvf_disable_intr(nic, NICVF_INTR_RBDR, qidx);
1414 		nicvf_clear_intr(nic, NICVF_INTR_RBDR, qidx);
1415 	}
1416 	nicvf_disable_intr(nic, NICVF_INTR_QS_ERR, 0);
1417 	nicvf_clear_intr(nic, NICVF_INTR_QS_ERR, 0);
1418 
1419 	/* Deactivate network interface */
1420 	if_setdrvflagbits(ifp, IFF_DRV_OACTIVE, IFF_DRV_RUNNING);
1421 
1422 	/* Free resources */
1423 	nicvf_config_data_transfer(nic, FALSE);
1424 
1425 	/* Disable HW Qset */
1426 	nicvf_qset_config(nic, FALSE);
1427 
1428 	/* disable mailbox interrupt */
1429 	nicvf_disable_intr(nic, NICVF_INTR_MBOX, 0);
1430 
1431 	return (0);
1432 }
1433 
1434 static void
1435 nicvf_update_stats(struct nicvf *nic)
1436 {
1437 	int qidx;
1438 	struct nicvf_hw_stats *stats = &nic->hw_stats;
1439 	struct nicvf_drv_stats *drv_stats = &nic->drv_stats;
1440 	struct queue_set *qs = nic->qs;
1441 
1442 #define	GET_RX_STATS(reg) \
1443     nicvf_reg_read(nic, NIC_VNIC_RX_STAT_0_13 | ((reg) << 3))
1444 #define GET_TX_STATS(reg) \
1445     nicvf_reg_read(nic, NIC_VNIC_TX_STAT_0_4 | ((reg) << 3))
1446 
1447 	stats->rx_bytes = GET_RX_STATS(RX_OCTS);
1448 	stats->rx_ucast_frames = GET_RX_STATS(RX_UCAST);
1449 	stats->rx_bcast_frames = GET_RX_STATS(RX_BCAST);
1450 	stats->rx_mcast_frames = GET_RX_STATS(RX_MCAST);
1451 	stats->rx_fcs_errors = GET_RX_STATS(RX_FCS);
1452 	stats->rx_l2_errors = GET_RX_STATS(RX_L2ERR);
1453 	stats->rx_drop_red = GET_RX_STATS(RX_RED);
1454 	stats->rx_drop_red_bytes = GET_RX_STATS(RX_RED_OCTS);
1455 	stats->rx_drop_overrun = GET_RX_STATS(RX_ORUN);
1456 	stats->rx_drop_overrun_bytes = GET_RX_STATS(RX_ORUN_OCTS);
1457 	stats->rx_drop_bcast = GET_RX_STATS(RX_DRP_BCAST);
1458 	stats->rx_drop_mcast = GET_RX_STATS(RX_DRP_MCAST);
1459 	stats->rx_drop_l3_bcast = GET_RX_STATS(RX_DRP_L3BCAST);
1460 	stats->rx_drop_l3_mcast = GET_RX_STATS(RX_DRP_L3MCAST);
1461 
1462 	stats->tx_bytes_ok = GET_TX_STATS(TX_OCTS);
1463 	stats->tx_ucast_frames_ok = GET_TX_STATS(TX_UCAST);
1464 	stats->tx_bcast_frames_ok = GET_TX_STATS(TX_BCAST);
1465 	stats->tx_mcast_frames_ok = GET_TX_STATS(TX_MCAST);
1466 	stats->tx_drops = GET_TX_STATS(TX_DROP);
1467 
1468 	drv_stats->tx_frames_ok = stats->tx_ucast_frames_ok +
1469 	    stats->tx_bcast_frames_ok + stats->tx_mcast_frames_ok;
1470 	drv_stats->rx_drops = stats->rx_drop_red + stats->rx_drop_overrun;
1471 	drv_stats->tx_drops = stats->tx_drops;
1472 
1473 	/* Update RQ and SQ stats */
1474 	for (qidx = 0; qidx < qs->rq_cnt; qidx++)
1475 		nicvf_update_rq_stats(nic, qidx);
1476 	for (qidx = 0; qidx < qs->sq_cnt; qidx++)
1477 		nicvf_update_sq_stats(nic, qidx);
1478 }
1479 
1480 static void
1481 nicvf_tick_stats(void *arg)
1482 {
1483 	struct nicvf *nic;
1484 
1485 	nic = (struct nicvf *)arg;
1486 
1487 	/* Read the statistics */
1488 	nicvf_update_stats(nic);
1489 
1490 	callout_reset(&nic->stats_callout, hz, nicvf_tick_stats, nic);
1491 }
1492