xref: /titanic_44/usr/src/uts/common/io/mac/mac_client.c (revision 349b53dd4e695e3d833b5380540385145b2d3ae8)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 /*
28  * - General Introduction:
29  *
30  * This file contains the implementation of the MAC client kernel
31  * API and related code. The MAC client API allows a kernel module
32  * to gain access to a MAC instance (physical NIC, link aggregation, etc).
33  * It allows a MAC client to associate itself with a MAC address,
34  * VLANs, callback functions for data traffic and for promiscuous mode.
35  * The MAC client API is also used to specify the properties associated
36  * with a MAC client, such as bandwidth limits, priority, CPUS, etc.
37  * These properties are further used to determine the hardware resources
38  * to allocate to the various MAC clients.
39  *
40  * - Primary MAC clients:
41  *
42  * The MAC client API refers to "primary MAC clients". A primary MAC
43  * client is a client which "owns" the primary MAC address of
44  * the underlying MAC instance. The primary MAC address is called out
45  * since it is associated with specific semantics: the primary MAC
46  * address is the MAC address which is assigned to the IP interface
47  * when it is plumbed, and the primary MAC address is assigned
48  * to VLAN data-links. The primary address of a MAC instance can
49  * also change dynamically from under the MAC client, for example
50  * as a result of a change of state of a link aggregation. In that
51  * case the MAC layer automatically updates all data-structures which
52  * refer to the current value of the primary MAC address. Typical
53  * primary MAC clients are dls, aggr, and xnb. A typical non-primary
54  * MAC client is the vnic driver.
55  *
56  * - Virtual Switching:
57  *
58  * The MAC layer implements a virtual switch between the MAC clients
59  * (primary and non-primary) defined on top of the same underlying
60  * NIC (physical, link aggregation, etc). The virtual switch is
61  * VLAN-aware, i.e. it allows multiple MAC clients to be member
62  * of one or more VLANs, and the virtual switch will distribute
63  * multicast tagged packets only to the member of the corresponding
64  * VLANs.
65  *
66  * - Upper vs Lower MAC:
67  *
68  * Creating a VNIC on top of a MAC instance effectively causes
69  * two MAC instances to be layered on top of each other, one for
70  * the VNIC(s), one for the underlying MAC instance (physical NIC,
71  * link aggregation, etc). In the code below we refer to the
72  * underlying NIC as the "lower MAC", and we refer to VNICs as
73  * the "upper MAC".
74  *
75  * - Pass-through for VNICs:
76  *
77  * When VNICs are created on top of an underlying MAC, this causes
78  * a layering of two MAC instances. Since the lower MAC already
79  * does the switching and demultiplexing to its MAC clients, the
80  * upper MAC would simply have to pass packets to the layer below
81  * or above it, which would introduce overhead. In order to avoid
82  * this overhead, the MAC layer implements a pass-through mechanism
83  * for VNICs. When a VNIC opens the lower MAC instance, it saves
84  * the MAC client handle it optains from the MAC layer. When a MAC
85  * client opens a VNIC (upper MAC), the MAC layer detects that
86  * the MAC being opened is a VNIC, and gets the MAC client handle
87  * that the VNIC driver obtained from the lower MAC. This exchange
88  * is doing through a private capability between the MAC layer
89  * and the VNIC driver. The upper MAC then returns that handle
90  * directly to its MAC client. Any operation done by the upper
91  * MAC client is now done on the lower MAC client handle, which
92  * allows the VNIC driver to be completely bypassed for the
93  * performance sensitive data-path.
94  *
95  */
96 
97 #include <sys/types.h>
98 #include <sys/conf.h>
99 #include <sys/id_space.h>
100 #include <sys/esunddi.h>
101 #include <sys/stat.h>
102 #include <sys/mkdev.h>
103 #include <sys/stream.h>
104 #include <sys/strsun.h>
105 #include <sys/strsubr.h>
106 #include <sys/dlpi.h>
107 #include <sys/modhash.h>
108 #include <sys/mac_impl.h>
109 #include <sys/mac_client_impl.h>
110 #include <sys/mac_soft_ring.h>
111 #include <sys/dls.h>
112 #include <sys/dld.h>
113 #include <sys/modctl.h>
114 #include <sys/fs/dv_node.h>
115 #include <sys/thread.h>
116 #include <sys/proc.h>
117 #include <sys/callb.h>
118 #include <sys/cpuvar.h>
119 #include <sys/atomic.h>
120 #include <sys/sdt.h>
121 #include <sys/mac_flow.h>
122 #include <sys/ddi_intr_impl.h>
123 #include <sys/disp.h>
124 #include <sys/sdt.h>
125 #include <sys/vnic.h>
126 #include <sys/vnic_impl.h>
127 #include <sys/vlan.h>
128 #include <inet/ip.h>
129 #include <inet/ip6.h>
130 #include <sys/exacct.h>
131 #include <sys/exacct_impl.h>
132 #include <inet/nd.h>
133 #include <sys/ethernet.h>
134 
135 kmem_cache_t	*mac_client_impl_cache;
136 kmem_cache_t	*mac_promisc_impl_cache;
137 
138 static boolean_t mac_client_single_rcvr(mac_client_impl_t *);
139 static flow_entry_t *mac_client_swap_mciflent(mac_client_impl_t *);
140 static flow_entry_t *mac_client_get_flow(mac_client_impl_t *,
141     mac_unicast_impl_t *);
142 static void mac_client_remove_flow_from_list(mac_client_impl_t *,
143     flow_entry_t *);
144 static void mac_client_add_to_flow_list(mac_client_impl_t *, flow_entry_t *);
145 static void mac_rename_flow_names(mac_client_impl_t *, const char *);
146 static void mac_virtual_link_update(mac_impl_t *);
147 
148 /* ARGSUSED */
149 static int
150 i_mac_client_impl_ctor(void *buf, void *arg, int kmflag)
151 {
152 	int	i;
153 	mac_client_impl_t	*mcip = buf;
154 
155 	bzero(buf, MAC_CLIENT_IMPL_SIZE);
156 	mutex_init(&mcip->mci_tx_cb_lock, NULL, MUTEX_DRIVER, NULL);
157 	mcip->mci_tx_notify_cb_info.mcbi_lockp = &mcip->mci_tx_cb_lock;
158 
159 	ASSERT(mac_tx_percpu_cnt >= 0);
160 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
161 		mutex_init(&mcip->mci_tx_pcpu[i].pcpu_tx_lock, NULL,
162 		    MUTEX_DRIVER, NULL);
163 	}
164 	cv_init(&mcip->mci_tx_cv, NULL, CV_DRIVER, NULL);
165 
166 	return (0);
167 }
168 
169 /* ARGSUSED */
170 static void
171 i_mac_client_impl_dtor(void *buf, void *arg)
172 {
173 	int	i;
174 	mac_client_impl_t *mcip = buf;
175 
176 	ASSERT(mcip->mci_promisc_list == NULL);
177 	ASSERT(mcip->mci_unicast_list == NULL);
178 	ASSERT(mcip->mci_state_flags == 0);
179 	ASSERT(mcip->mci_tx_flag == 0);
180 
181 	mutex_destroy(&mcip->mci_tx_cb_lock);
182 
183 	ASSERT(mac_tx_percpu_cnt >= 0);
184 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
185 		ASSERT(mcip->mci_tx_pcpu[i].pcpu_tx_refcnt == 0);
186 		mutex_destroy(&mcip->mci_tx_pcpu[i].pcpu_tx_lock);
187 	}
188 	cv_destroy(&mcip->mci_tx_cv);
189 }
190 
191 /* ARGSUSED */
192 static int
193 i_mac_promisc_impl_ctor(void *buf, void *arg, int kmflag)
194 {
195 	mac_promisc_impl_t	*mpip = buf;
196 
197 	bzero(buf, sizeof (mac_promisc_impl_t));
198 	mpip->mpi_mci_link.mcb_objp = buf;
199 	mpip->mpi_mci_link.mcb_objsize = sizeof (mac_promisc_impl_t);
200 	mpip->mpi_mi_link.mcb_objp = buf;
201 	mpip->mpi_mi_link.mcb_objsize = sizeof (mac_promisc_impl_t);
202 	return (0);
203 }
204 
205 /* ARGSUSED */
206 static void
207 i_mac_promisc_impl_dtor(void *buf, void *arg)
208 {
209 	mac_promisc_impl_t	*mpip = buf;
210 
211 	ASSERT(mpip->mpi_mci_link.mcb_objp != NULL);
212 	ASSERT(mpip->mpi_mci_link.mcb_objsize == sizeof (mac_promisc_impl_t));
213 	ASSERT(mpip->mpi_mi_link.mcb_objp == mpip->mpi_mci_link.mcb_objp);
214 	ASSERT(mpip->mpi_mi_link.mcb_objsize == sizeof (mac_promisc_impl_t));
215 
216 	mpip->mpi_mci_link.mcb_objp = NULL;
217 	mpip->mpi_mci_link.mcb_objsize = 0;
218 	mpip->mpi_mi_link.mcb_objp = NULL;
219 	mpip->mpi_mi_link.mcb_objsize = 0;
220 
221 	ASSERT(mpip->mpi_mci_link.mcb_flags == 0);
222 	mpip->mpi_mci_link.mcb_objsize = 0;
223 }
224 
225 void
226 mac_client_init(void)
227 {
228 	ASSERT(mac_tx_percpu_cnt >= 0);
229 
230 	mac_client_impl_cache = kmem_cache_create("mac_client_impl_cache",
231 	    MAC_CLIENT_IMPL_SIZE, 0, i_mac_client_impl_ctor,
232 	    i_mac_client_impl_dtor, NULL, NULL, NULL, 0);
233 	ASSERT(mac_client_impl_cache != NULL);
234 
235 	mac_promisc_impl_cache = kmem_cache_create("mac_promisc_impl_cache",
236 	    sizeof (mac_promisc_impl_t), 0, i_mac_promisc_impl_ctor,
237 	    i_mac_promisc_impl_dtor, NULL, NULL, NULL, 0);
238 	ASSERT(mac_promisc_impl_cache != NULL);
239 }
240 
241 void
242 mac_client_fini(void)
243 {
244 	kmem_cache_destroy(mac_client_impl_cache);
245 	kmem_cache_destroy(mac_promisc_impl_cache);
246 }
247 
248 /*
249  * Return the lower MAC client handle from the VNIC driver for the
250  * specified VNIC MAC instance.
251  */
252 mac_client_impl_t *
253 mac_vnic_lower(mac_impl_t *mip)
254 {
255 	mac_capab_vnic_t cap;
256 	mac_client_impl_t *mcip;
257 
258 	VERIFY(i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_VNIC, &cap));
259 	mcip = cap.mcv_mac_client_handle(cap.mcv_arg);
260 
261 	return (mcip);
262 }
263 
264 /*
265  * Return the MAC client handle of the primary MAC client for the
266  * specified MAC instance, or NULL otherwise.
267  */
268 mac_client_impl_t *
269 mac_primary_client_handle(mac_impl_t *mip)
270 {
271 	mac_client_impl_t *mcip;
272 
273 	if (mip->mi_state_flags & MIS_IS_VNIC)
274 		return (mac_vnic_lower(mip));
275 
276 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
277 
278 	for (mcip = mip->mi_clients_list; mcip != NULL;
279 	    mcip = mcip->mci_client_next) {
280 		if (MCIP_DATAPATH_SETUP(mcip) && mac_is_primary_client(mcip))
281 			return (mcip);
282 	}
283 	return (NULL);
284 }
285 
286 /*
287  * Open a MAC specified by its MAC name.
288  */
289 int
290 mac_open(const char *macname, mac_handle_t *mhp)
291 {
292 	mac_impl_t	*mip;
293 	int		err;
294 
295 	/*
296 	 * Look up its entry in the global hash table.
297 	 */
298 	if ((err = mac_hold(macname, &mip)) != 0)
299 		return (err);
300 
301 	/*
302 	 * Hold the dip associated to the MAC to prevent it from being
303 	 * detached. For a softmac, its underlying dip is held by the
304 	 * mi_open() callback.
305 	 *
306 	 * This is done to be more tolerant with some defective drivers,
307 	 * which incorrectly handle mac_unregister() failure in their
308 	 * xxx_detach() routine. For example, some drivers ignore the
309 	 * failure of mac_unregister() and free all resources that
310 	 * that are needed for data transmition.
311 	 */
312 	e_ddi_hold_devi(mip->mi_dip);
313 
314 	if (!(mip->mi_callbacks->mc_callbacks & MC_OPEN)) {
315 		*mhp = (mac_handle_t)mip;
316 		return (0);
317 	}
318 
319 	/*
320 	 * The mac perimeter is used in both mac_open and mac_close by the
321 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
322 	 */
323 	i_mac_perim_enter(mip);
324 	mip->mi_oref++;
325 	if (mip->mi_oref != 1 || ((err = mip->mi_open(mip->mi_driver)) == 0)) {
326 		*mhp = (mac_handle_t)mip;
327 		i_mac_perim_exit(mip);
328 		return (0);
329 	}
330 	mip->mi_oref--;
331 	ddi_release_devi(mip->mi_dip);
332 	mac_rele(mip);
333 	i_mac_perim_exit(mip);
334 	return (err);
335 }
336 
337 /*
338  * Open a MAC specified by its linkid.
339  */
340 int
341 mac_open_by_linkid(datalink_id_t linkid, mac_handle_t *mhp)
342 {
343 	dls_dl_handle_t	dlh;
344 	int		err;
345 
346 	if ((err = dls_devnet_hold_tmp(linkid, &dlh)) != 0)
347 		return (err);
348 
349 	dls_devnet_prop_task_wait(dlh);
350 
351 	err = mac_open(dls_devnet_mac(dlh), mhp);
352 
353 	dls_devnet_rele_tmp(dlh);
354 	return (err);
355 }
356 
357 /*
358  * Open a MAC specified by its link name.
359  */
360 int
361 mac_open_by_linkname(const char *link, mac_handle_t *mhp)
362 {
363 	datalink_id_t	linkid;
364 	int		err;
365 
366 	if ((err = dls_mgmt_get_linkid(link, &linkid)) != 0)
367 		return (err);
368 	return (mac_open_by_linkid(linkid, mhp));
369 }
370 
371 /*
372  * Close the specified MAC.
373  */
374 void
375 mac_close(mac_handle_t mh)
376 {
377 	mac_impl_t	*mip = (mac_impl_t *)mh;
378 
379 	i_mac_perim_enter(mip);
380 	/*
381 	 * The mac perimeter is used in both mac_open and mac_close by the
382 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
383 	 */
384 	if (mip->mi_callbacks->mc_callbacks & MC_OPEN) {
385 		ASSERT(mip->mi_oref != 0);
386 		if (--mip->mi_oref == 0) {
387 			if ((mip->mi_callbacks->mc_callbacks & MC_CLOSE))
388 				mip->mi_close(mip->mi_driver);
389 		}
390 	}
391 	i_mac_perim_exit(mip);
392 	ddi_release_devi(mip->mi_dip);
393 	mac_rele(mip);
394 }
395 
396 /*
397  * Misc utility functions to retrieve various information about a MAC
398  * instance or a MAC client.
399  */
400 
401 const mac_info_t *
402 mac_info(mac_handle_t mh)
403 {
404 	return (&((mac_impl_t *)mh)->mi_info);
405 }
406 
407 dev_info_t *
408 mac_devinfo_get(mac_handle_t mh)
409 {
410 	return (((mac_impl_t *)mh)->mi_dip);
411 }
412 
413 void *
414 mac_driver(mac_handle_t mh)
415 {
416 	return (((mac_impl_t *)mh)->mi_driver);
417 }
418 
419 const char *
420 mac_name(mac_handle_t mh)
421 {
422 	return (((mac_impl_t *)mh)->mi_name);
423 }
424 
425 char *
426 mac_client_name(mac_client_handle_t mch)
427 {
428 	return (((mac_client_impl_t *)mch)->mci_name);
429 }
430 
431 minor_t
432 mac_minor(mac_handle_t mh)
433 {
434 	return (((mac_impl_t *)mh)->mi_minor);
435 }
436 
437 /*
438  * Return the VID associated with a MAC client. This function should
439  * be called for clients which are associated with only one VID.
440  */
441 uint16_t
442 mac_client_vid(mac_client_handle_t mch)
443 {
444 	uint16_t		vid = VLAN_ID_NONE;
445 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
446 	flow_desc_t		flow_desc;
447 
448 	if (mcip->mci_nflents == 0)
449 		return (vid);
450 
451 	ASSERT(MCIP_DATAPATH_SETUP(mcip) && mac_client_single_rcvr(mcip));
452 
453 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
454 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
455 		vid = flow_desc.fd_vid;
456 
457 	return (vid);
458 }
459 
460 /*
461  * Return whether the specified MAC client corresponds to a VLAN VNIC.
462  */
463 boolean_t
464 mac_client_is_vlan_vnic(mac_client_handle_t mch)
465 {
466 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
467 
468 	return (((mcip->mci_state_flags & MCIS_IS_VNIC) != 0) &&
469 	    ((mcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) != 0));
470 }
471 
472 /*
473  * Return the link speed associated with the specified MAC client.
474  *
475  * The link speed of a MAC client is equal to the smallest value of
476  * 1) the current link speed of the underlying NIC, or
477  * 2) the bandwidth limit set for the MAC client.
478  *
479  * Note that the bandwidth limit can be higher than the speed
480  * of the underlying NIC. This is allowed to avoid spurious
481  * administration action failures or artifically lowering the
482  * bandwidth limit of a link that may  have temporarily lowered
483  * its link speed due to hardware problem or administrator action.
484  */
485 static uint64_t
486 mac_client_ifspeed(mac_client_impl_t *mcip)
487 {
488 	mac_impl_t *mip = mcip->mci_mip;
489 	uint64_t nic_speed;
490 
491 	nic_speed = mac_stat_get((mac_handle_t)mip, MAC_STAT_IFSPEED);
492 
493 	if (nic_speed == 0) {
494 		return (0);
495 	} else {
496 		uint64_t policy_limit = (uint64_t)-1;
497 
498 		if (MCIP_RESOURCE_PROPS_MASK(mcip) & MRP_MAXBW)
499 			policy_limit = MCIP_RESOURCE_PROPS_MAXBW(mcip);
500 
501 		return (MIN(policy_limit, nic_speed));
502 	}
503 }
504 
505 /*
506  * Return the link state of the specified client. If here are more
507  * than one clients of the underying mac_impl_t, the link state
508  * will always be UP regardless of the link state of the underlying
509  * mac_impl_t. This is needed to allow the MAC clients to continue
510  * to communicate with each other even when the physical link of
511  * their mac_impl_t is down.
512  */
513 static uint64_t
514 mac_client_link_state(mac_client_impl_t *mcip)
515 {
516 	mac_impl_t *mip = mcip->mci_mip;
517 	uint16_t vid;
518 	mac_client_impl_t *mci_list;
519 	mac_unicast_impl_t *mui_list, *oth_mui_list;
520 
521 	/*
522 	 * Returns LINK_STATE_UP if there are other MAC clients defined on
523 	 * mac_impl_t which share same VLAN ID as that of mcip. Note that
524 	 * if 'mcip' has more than one VID's then we match ANY one of the
525 	 * VID's with other MAC client's VID's and return LINK_STATE_UP.
526 	 */
527 	rw_enter(&mcip->mci_rw_lock, RW_READER);
528 	for (mui_list = mcip->mci_unicast_list; mui_list != NULL;
529 	    mui_list = mui_list->mui_next) {
530 		vid = mui_list->mui_vid;
531 		for (mci_list = mip->mi_clients_list; mci_list != NULL;
532 		    mci_list = mci_list->mci_client_next) {
533 			if (mci_list == mcip)
534 				continue;
535 			for (oth_mui_list = mci_list->mci_unicast_list;
536 			    oth_mui_list != NULL; oth_mui_list = oth_mui_list->
537 			    mui_next) {
538 				if (vid == oth_mui_list->mui_vid) {
539 					rw_exit(&mcip->mci_rw_lock);
540 					return (LINK_STATE_UP);
541 				}
542 			}
543 		}
544 	}
545 	rw_exit(&mcip->mci_rw_lock);
546 
547 	return (mac_stat_get((mac_handle_t)mip, MAC_STAT_LINK_STATE));
548 }
549 
550 /*
551  * Return the statistics of a MAC client. These statistics are different
552  * then the statistics of the underlying MAC which are returned by
553  * mac_stat_get().
554  */
555 uint64_t
556 mac_client_stat_get(mac_client_handle_t mch, uint_t stat)
557 {
558 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
559 	mac_impl_t *mip = mcip->mci_mip;
560 	uint64_t val;
561 
562 	switch (stat) {
563 	case MAC_STAT_LINK_STATE:
564 		val = mac_client_link_state(mcip);
565 		break;
566 	case MAC_STAT_LINK_UP:
567 		val = (mac_client_link_state(mcip) == LINK_STATE_UP);
568 		break;
569 	case MAC_STAT_PROMISC:
570 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_PROMISC);
571 		break;
572 	case MAC_STAT_IFSPEED:
573 		val = mac_client_ifspeed(mcip);
574 		break;
575 	case MAC_STAT_MULTIRCV:
576 		val = mcip->mci_stat_multircv;
577 		break;
578 	case MAC_STAT_BRDCSTRCV:
579 		val = mcip->mci_stat_brdcstrcv;
580 		break;
581 	case MAC_STAT_MULTIXMT:
582 		val = mcip->mci_stat_multixmt;
583 		break;
584 	case MAC_STAT_BRDCSTXMT:
585 		val = mcip->mci_stat_brdcstxmt;
586 		break;
587 	case MAC_STAT_OBYTES:
588 		val = mcip->mci_stat_obytes;
589 		break;
590 	case MAC_STAT_OPACKETS:
591 		val = mcip->mci_stat_opackets;
592 		break;
593 	case MAC_STAT_OERRORS:
594 		val = mcip->mci_stat_oerrors;
595 		break;
596 	case MAC_STAT_IPACKETS:
597 		val = mcip->mci_stat_ipackets;
598 		break;
599 	case MAC_STAT_RBYTES:
600 		val = mcip->mci_stat_ibytes;
601 		break;
602 	case MAC_STAT_IERRORS:
603 		val = mcip->mci_stat_ierrors;
604 		break;
605 	default:
606 		val = mac_stat_default(mip, stat);
607 		break;
608 	}
609 
610 	return (val);
611 }
612 
613 /*
614  * Return the statistics of the specified MAC instance.
615  */
616 uint64_t
617 mac_stat_get(mac_handle_t mh, uint_t stat)
618 {
619 	mac_impl_t	*mip = (mac_impl_t *)mh;
620 	uint64_t	val;
621 	int		ret;
622 
623 	/*
624 	 * The range of stat determines where it is maintained.  Stat
625 	 * values from 0 up to (but not including) MAC_STAT_MIN are
626 	 * mainteined by the mac module itself.  Everything else is
627 	 * maintained by the driver.
628 	 *
629 	 * If the mac_impl_t being queried corresponds to a VNIC,
630 	 * the stats need to be queried from the lower MAC client
631 	 * corresponding to the VNIC. (The mac_link_update()
632 	 * invoked by the driver to the lower MAC causes the *lower
633 	 * MAC* to update its mi_linkstate, and send a notification
634 	 * to its MAC clients. Due to the VNIC passthrough,
635 	 * these notifications are sent to the upper MAC clients
636 	 * of the VNIC directly, and the upper mac_impl_t of the VNIC
637 	 * does not have a valid mi_linkstate.
638 	 */
639 	if (stat < MAC_STAT_MIN && !(mip->mi_state_flags & MIS_IS_VNIC)) {
640 		/* these stats are maintained by the mac module itself */
641 		switch (stat) {
642 		case MAC_STAT_LINK_STATE:
643 			return (mip->mi_linkstate);
644 		case MAC_STAT_LINK_UP:
645 			return (mip->mi_linkstate == LINK_STATE_UP);
646 		case MAC_STAT_PROMISC:
647 			return (mip->mi_devpromisc != 0);
648 		default:
649 			ASSERT(B_FALSE);
650 		}
651 	}
652 
653 	/*
654 	 * Call the driver to get the given statistic.
655 	 */
656 	ret = mip->mi_getstat(mip->mi_driver, stat, &val);
657 	if (ret != 0) {
658 		/*
659 		 * The driver doesn't support this statistic.  Get the
660 		 * statistic's default value.
661 		 */
662 		val = mac_stat_default(mip, stat);
663 	}
664 	return (val);
665 }
666 
667 /*
668  * Utility function which returns the VID associated with a flow entry.
669  */
670 uint16_t
671 i_mac_flow_vid(flow_entry_t *flent)
672 {
673 	flow_desc_t	flow_desc;
674 
675 	mac_flow_get_desc(flent, &flow_desc);
676 
677 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
678 		return (flow_desc.fd_vid);
679 	return (VLAN_ID_NONE);
680 }
681 
682 /*
683  * Verify the validity of the specified unicast MAC address. Returns B_TRUE
684  * if the address is valid, B_FALSE otherwise (multicast address, or incorrect
685  * length.
686  */
687 boolean_t
688 mac_unicst_verify(mac_handle_t mh, const uint8_t *addr, uint_t len)
689 {
690 	mac_impl_t	*mip = (mac_impl_t *)mh;
691 
692 	/*
693 	 * Verify the address. No lock is needed since mi_type and plugin
694 	 * details don't change after mac_register().
695 	 */
696 	if ((len != mip->mi_type->mt_addr_length) ||
697 	    (mip->mi_type->mt_ops.mtops_unicst_verify(addr,
698 	    mip->mi_pdata)) != 0) {
699 		return (B_FALSE);
700 	} else {
701 		return (B_TRUE);
702 	}
703 }
704 
705 void
706 mac_sdu_get(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu)
707 {
708 	mac_impl_t	*mip = (mac_impl_t *)mh;
709 
710 	if (min_sdu != NULL)
711 		*min_sdu = mip->mi_sdu_min;
712 	if (max_sdu != NULL)
713 		*max_sdu = mip->mi_sdu_max;
714 }
715 
716 /*
717  * Update the MAC unicast address of the specified client's flows. Currently
718  * only one unicast MAC unicast address is allowed per client.
719  */
720 static void
721 mac_unicast_update_client_flow(mac_client_impl_t *mcip)
722 {
723 	mac_impl_t *mip = mcip->mci_mip;
724 	flow_entry_t *flent = mcip->mci_flent;
725 	mac_address_t *map = mcip->mci_unicast;
726 	flow_desc_t flow_desc;
727 
728 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
729 	ASSERT(flent != NULL);
730 
731 	mac_flow_get_desc(flent, &flow_desc);
732 	ASSERT(flow_desc.fd_mask & FLOW_LINK_DST);
733 
734 	bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
735 	mac_flow_set_desc(flent, &flow_desc);
736 
737 	/*
738 	 * A MAC client could have one MAC address but multiple
739 	 * VLANs. In that case update the flow entries corresponding
740 	 * to all VLANs of the MAC client.
741 	 */
742 	for (flent = mcip->mci_flent_list; flent != NULL;
743 	    flent = flent->fe_client_next) {
744 		mac_flow_get_desc(flent, &flow_desc);
745 		if (!(flent->fe_type & FLOW_PRIMARY_MAC ||
746 		    flent->fe_type & FLOW_VNIC_MAC))
747 			continue;
748 
749 		bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
750 		mac_flow_set_desc(flent, &flow_desc);
751 	}
752 }
753 
754 /*
755  * Update all clients that share the same unicast address.
756  */
757 void
758 mac_unicast_update_clients(mac_impl_t *mip, mac_address_t *map)
759 {
760 	mac_client_impl_t *mcip;
761 
762 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
763 
764 	/*
765 	 * Find all clients that share the same unicast MAC address and update
766 	 * them appropriately.
767 	 */
768 	for (mcip = mip->mi_clients_list; mcip != NULL;
769 	    mcip = mcip->mci_client_next) {
770 		/*
771 		 * Ignore clients that don't share this MAC address.
772 		 */
773 		if (map != mcip->mci_unicast)
774 			continue;
775 
776 		/*
777 		 * Update those clients with same old unicast MAC address.
778 		 */
779 		mac_unicast_update_client_flow(mcip);
780 	}
781 }
782 
783 /*
784  * Update the unicast MAC address of the specified VNIC MAC client.
785  *
786  * Check whether the operation is valid. Any of following cases should fail:
787  *
788  * 1. It's a VLAN type of VNIC.
789  * 2. The new value is current "primary" MAC address.
790  * 3. The current MAC address is shared with other clients.
791  * 4. The new MAC address has been used. This case will be valid when
792  *    client migration is fully supported.
793  */
794 int
795 mac_vnic_unicast_set(mac_client_handle_t mch, const uint8_t *addr)
796 {
797 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
798 	mac_impl_t *mip = mcip->mci_mip;
799 	mac_address_t *map = mcip->mci_unicast;
800 	int err;
801 
802 	ASSERT(!(mip->mi_state_flags & MIS_IS_VNIC));
803 	ASSERT(mcip->mci_state_flags & MCIS_IS_VNIC);
804 	ASSERT(mcip->mci_flags != MAC_CLIENT_FLAGS_PRIMARY);
805 
806 	i_mac_perim_enter(mip);
807 
808 	/*
809 	 * If this is a VLAN type of VNIC, it's using "primary" MAC address
810 	 * of the underlying interface. Must fail here. Refer to case 1 above.
811 	 */
812 	if (bcmp(map->ma_addr, mip->mi_addr, map->ma_len) == 0) {
813 		i_mac_perim_exit(mip);
814 		return (ENOTSUP);
815 	}
816 
817 	/*
818 	 * If the new address is the "primary" one, must fail. Refer to
819 	 * case 2 above.
820 	 */
821 	if (bcmp(addr, mip->mi_addr, map->ma_len) == 0) {
822 		i_mac_perim_exit(mip);
823 		return (EACCES);
824 	}
825 
826 	/*
827 	 * If the address is shared by multiple clients, must fail. Refer
828 	 * to case 3 above.
829 	 */
830 	if (mac_check_macaddr_shared(map)) {
831 		i_mac_perim_exit(mip);
832 		return (EBUSY);
833 	}
834 
835 	/*
836 	 * If the new address has been used, must fail for now. Refer to
837 	 * case 4 above.
838 	 */
839 	if (mac_find_macaddr(mip, (uint8_t *)addr) != NULL) {
840 		i_mac_perim_exit(mip);
841 		return (ENOTSUP);
842 	}
843 
844 	/*
845 	 * Update the MAC address.
846 	 */
847 	err = mac_update_macaddr(map, (uint8_t *)addr);
848 
849 	if (err != 0) {
850 		i_mac_perim_exit(mip);
851 		return (err);
852 	}
853 
854 	/*
855 	 * Update all flows of this MAC client.
856 	 */
857 	mac_unicast_update_client_flow(mcip);
858 
859 	i_mac_perim_exit(mip);
860 	return (0);
861 }
862 
863 /*
864  * Program the new primary unicast address of the specified MAC.
865  *
866  * Function mac_update_macaddr() takes care different types of underlying
867  * MAC. If the underlying MAC is VNIC, the VNIC driver must have registerd
868  * mi_unicst() entry point, that indirectly calls mac_vnic_unicast_set()
869  * which will take care of updating the MAC address of the corresponding
870  * MAC client.
871  *
872  * This is the only interface that allow the client to update the "primary"
873  * MAC address of the underlying MAC. The new value must have not been
874  * used by other clients.
875  */
876 int
877 mac_unicast_primary_set(mac_handle_t mh, const uint8_t *addr)
878 {
879 	mac_impl_t *mip = (mac_impl_t *)mh;
880 	mac_address_t *map;
881 	int err;
882 
883 	/* verify the address validity */
884 	if (!mac_unicst_verify(mh, addr, mip->mi_type->mt_addr_length))
885 		return (EINVAL);
886 
887 	i_mac_perim_enter(mip);
888 
889 	/*
890 	 * If the new value is the same as the current primary address value,
891 	 * there's nothing to do.
892 	 */
893 	if (bcmp(addr, mip->mi_addr, mip->mi_type->mt_addr_length) == 0) {
894 		i_mac_perim_exit(mip);
895 		return (0);
896 	}
897 
898 	if (mac_find_macaddr(mip, (uint8_t *)addr) != 0) {
899 		i_mac_perim_exit(mip);
900 		return (EBUSY);
901 	}
902 
903 	map = mac_find_macaddr(mip, mip->mi_addr);
904 	ASSERT(map != NULL);
905 
906 	/*
907 	 * Update the MAC address.
908 	 */
909 	if (mip->mi_state_flags & MIS_IS_AGGR) {
910 		mac_capab_aggr_t aggr_cap;
911 
912 		/*
913 		 * If the mac is an aggregation, other than the unicast
914 		 * addresses programming, aggr must be informed about this
915 		 * primary unicst address change to change its mac address
916 		 * policy to be user-specified.
917 		 */
918 		ASSERT(map->ma_type == MAC_ADDRESS_TYPE_UNICAST_CLASSIFIED);
919 		VERIFY(i_mac_capab_get(mh, MAC_CAPAB_AGGR, &aggr_cap));
920 		err = aggr_cap.mca_unicst(mip->mi_driver, addr);
921 		if (err == 0)
922 			bcopy(addr, map->ma_addr, map->ma_len);
923 	} else {
924 		err = mac_update_macaddr(map, (uint8_t *)addr);
925 	}
926 
927 	if (err != 0) {
928 		i_mac_perim_exit(mip);
929 		return (err);
930 	}
931 
932 	mac_unicast_update_clients(mip, map);
933 
934 	/*
935 	 * Save the new primary MAC address in mac_impl_t.
936 	 */
937 	bcopy(addr, mip->mi_addr, mip->mi_type->mt_addr_length);
938 
939 	i_mac_perim_exit(mip);
940 
941 	if (err == 0)
942 		i_mac_notify(mip, MAC_NOTE_UNICST);
943 
944 	return (err);
945 }
946 
947 /*
948  * Return the current primary MAC address of the specified MAC.
949  */
950 void
951 mac_unicast_primary_get(mac_handle_t mh, uint8_t *addr)
952 {
953 	mac_impl_t *mip = (mac_impl_t *)mh;
954 
955 	rw_enter(&mip->mi_rw_lock, RW_READER);
956 	bcopy(mip->mi_addr, addr, mip->mi_type->mt_addr_length);
957 	rw_exit(&mip->mi_rw_lock);
958 }
959 
960 /*
961  * Return information about the use of the primary MAC address of the
962  * specified MAC instance:
963  *
964  * - if client_name is non-NULL, it must point to a string of at
965  *   least MAXNAMELEN bytes, and will be set to the name of the MAC
966  *   client which uses the primary MAC address.
967  *
968  * - if in_use is non-NULL, used to return whether the primary MAC
969  *   address is currently in use.
970  */
971 void
972 mac_unicast_primary_info(mac_handle_t mh, char *client_name, boolean_t *in_use)
973 {
974 	mac_impl_t *mip = (mac_impl_t *)mh;
975 	mac_client_impl_t *cur_client;
976 
977 	if (in_use != NULL)
978 		*in_use = B_FALSE;
979 	if (client_name != NULL)
980 		bzero(client_name, MAXNAMELEN);
981 
982 	/*
983 	 * The mi_rw_lock is used to protect threads that don't hold the
984 	 * mac perimeter to get a consistent view of the mi_clients_list.
985 	 * Threads that modify the list must hold both the mac perimeter and
986 	 * mi_rw_lock(RW_WRITER)
987 	 */
988 	rw_enter(&mip->mi_rw_lock, RW_READER);
989 	for (cur_client = mip->mi_clients_list; cur_client != NULL;
990 	    cur_client = cur_client->mci_client_next) {
991 		if (mac_is_primary_client(cur_client) ||
992 		    (mip->mi_state_flags & MIS_IS_VNIC)) {
993 			rw_exit(&mip->mi_rw_lock);
994 			if (in_use != NULL)
995 				*in_use = B_TRUE;
996 			if (client_name != NULL) {
997 				bcopy(cur_client->mci_name, client_name,
998 				    MAXNAMELEN);
999 			}
1000 			return;
1001 		}
1002 	}
1003 	rw_exit(&mip->mi_rw_lock);
1004 }
1005 
1006 /*
1007  * Add the specified MAC client to the list of clients which opened
1008  * the specified MAC.
1009  */
1010 static void
1011 mac_client_add(mac_client_impl_t *mcip)
1012 {
1013 	mac_impl_t *mip = mcip->mci_mip;
1014 
1015 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1016 
1017 	/* add VNIC to the front of the list */
1018 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1019 	mcip->mci_client_next = mip->mi_clients_list;
1020 	mip->mi_clients_list = mcip;
1021 	mip->mi_nclients++;
1022 	rw_exit(&mip->mi_rw_lock);
1023 }
1024 
1025 /*
1026  * Remove the specified MAC client from the list of clients which opened
1027  * the specified MAC.
1028  */
1029 static void
1030 mac_client_remove(mac_client_impl_t *mcip)
1031 {
1032 	mac_impl_t *mip = mcip->mci_mip;
1033 	mac_client_impl_t **prev, *cclient;
1034 
1035 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1036 
1037 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1038 	prev = &mip->mi_clients_list;
1039 	cclient = *prev;
1040 	while (cclient != NULL && cclient != mcip) {
1041 		prev = &cclient->mci_client_next;
1042 		cclient = *prev;
1043 	}
1044 	ASSERT(cclient != NULL);
1045 	*prev = cclient->mci_client_next;
1046 	mip->mi_nclients--;
1047 	rw_exit(&mip->mi_rw_lock);
1048 }
1049 
1050 static mac_unicast_impl_t *
1051 mac_client_find_vid(mac_client_impl_t *mcip, uint16_t vid)
1052 {
1053 	mac_unicast_impl_t *muip = mcip->mci_unicast_list;
1054 
1055 	while ((muip != NULL) && (muip->mui_vid != vid))
1056 		muip = muip->mui_next;
1057 
1058 	return (muip);
1059 }
1060 
1061 /*
1062  * Return whether the specified (MAC address, VID) tuple is already used by
1063  * one of the MAC clients associated with the specified MAC.
1064  */
1065 static boolean_t
1066 mac_addr_in_use(mac_impl_t *mip, uint8_t *mac_addr, uint16_t vid)
1067 {
1068 	mac_client_impl_t *client;
1069 	mac_address_t *map;
1070 
1071 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1072 
1073 	for (client = mip->mi_clients_list; client != NULL;
1074 	    client = client->mci_client_next) {
1075 
1076 		/*
1077 		 * Ignore clients that don't have unicast address.
1078 		 */
1079 		if (client->mci_unicast_list == NULL)
1080 			continue;
1081 
1082 		map = client->mci_unicast;
1083 
1084 		if ((bcmp(mac_addr, map->ma_addr, map->ma_len) == 0) &&
1085 		    (mac_client_find_vid(client, vid) != NULL)) {
1086 			return (B_TRUE);
1087 		}
1088 	}
1089 
1090 	return (B_FALSE);
1091 }
1092 
1093 /*
1094  * Generate a random MAC address. The MAC address prefix is
1095  * stored in the array pointed to by mac_addr, and its length, in bytes,
1096  * is specified by prefix_len. The least significant bits
1097  * after prefix_len bytes are generated, and stored after the prefix
1098  * in the mac_addr array.
1099  */
1100 int
1101 mac_addr_random(mac_client_handle_t mch, uint_t prefix_len,
1102     uint8_t *mac_addr, mac_diag_t *diag)
1103 {
1104 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1105 	mac_impl_t *mip = mcip->mci_mip;
1106 	size_t addr_len = mip->mi_type->mt_addr_length;
1107 
1108 	if (prefix_len >= addr_len) {
1109 		*diag = MAC_DIAG_MACPREFIXLEN_INVALID;
1110 		return (EINVAL);
1111 	}
1112 
1113 	/* check the prefix value */
1114 	if (prefix_len > 0) {
1115 		bzero(mac_addr + prefix_len, addr_len - prefix_len);
1116 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr,
1117 		    addr_len)) {
1118 			*diag = MAC_DIAG_MACPREFIX_INVALID;
1119 			return (EINVAL);
1120 		}
1121 	}
1122 
1123 	/* generate the MAC address */
1124 	if (prefix_len < addr_len) {
1125 		(void) random_get_pseudo_bytes(mac_addr +
1126 		    prefix_len, addr_len - prefix_len);
1127 	}
1128 
1129 	*diag = 0;
1130 	return (0);
1131 }
1132 
1133 /*
1134  * Set the priority range for this MAC client. This will be used to
1135  * determine the absolute priority for the threads created for this
1136  * MAC client using the specified "low", "medium" and "high" level.
1137  * This will also be used for any subflows on this MAC client.
1138  */
1139 #define	MAC_CLIENT_SET_PRIORITY_RANGE(mcip, pri) {			\
1140 	(mcip)->mci_min_pri = FLOW_MIN_PRIORITY(MINCLSYSPRI,	\
1141 	    MAXCLSYSPRI, (pri));					\
1142 	(mcip)->mci_max_pri = FLOW_MAX_PRIORITY(MINCLSYSPRI,	\
1143 	    MAXCLSYSPRI, (mcip)->mci_min_pri);				\
1144 	}
1145 
1146 /*
1147  * MAC client open entry point. Return a new MAC client handle. Each
1148  * MAC client is associated with a name, specified through the 'name'
1149  * argument.
1150  */
1151 int
1152 mac_client_open(mac_handle_t mh, mac_client_handle_t *mchp, char *name,
1153     uint16_t flags)
1154 {
1155 	mac_impl_t *mip = (mac_impl_t *)mh;
1156 	mac_client_impl_t *mcip;
1157 	int err = 0;
1158 	boolean_t share_desired =
1159 	    ((flags & MAC_OPEN_FLAGS_SHARES_DESIRED) != 0);
1160 	boolean_t no_hwrings = ((flags & MAC_OPEN_FLAGS_NO_HWRINGS) != 0);
1161 	boolean_t req_hwrings = ((flags & MAC_OPEN_FLAGS_REQ_HWRINGS) != 0);
1162 	flow_entry_t	*flent = NULL;
1163 
1164 	*mchp = NULL;
1165 	if (share_desired && no_hwrings) {
1166 		/* can't have shares but no hardware rings */
1167 		return (EINVAL);
1168 	}
1169 
1170 	i_mac_perim_enter(mip);
1171 
1172 	if (mip->mi_state_flags & MIS_IS_VNIC) {
1173 		/*
1174 		 * The underlying MAC is a VNIC. Return the MAC client
1175 		 * handle of the lower MAC which was obtained by
1176 		 * the VNIC driver when it did its mac_client_open().
1177 		 */
1178 
1179 		mcip = mac_vnic_lower(mip);
1180 
1181 		/*
1182 		 * Note that multiple mac clients share the same mcip in
1183 		 * this case.
1184 		 */
1185 		if (flags & MAC_OPEN_FLAGS_EXCLUSIVE)
1186 			mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1187 
1188 		if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1189 			mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1190 
1191 		mip->mi_clients_list = mcip;
1192 		i_mac_perim_exit(mip);
1193 		*mchp = (mac_client_handle_t)mcip;
1194 		return (err);
1195 	}
1196 
1197 	mcip = kmem_cache_alloc(mac_client_impl_cache, KM_SLEEP);
1198 
1199 	mcip->mci_mip = mip;
1200 	mcip->mci_upper_mip = NULL;
1201 	mcip->mci_rx_fn = mac_pkt_drop;
1202 	mcip->mci_rx_arg = NULL;
1203 	mcip->mci_rx_p_fn = NULL;
1204 	mcip->mci_rx_p_arg = NULL;
1205 	mcip->mci_p_unicast_list = NULL;
1206 	mcip->mci_direct_rx_fn = NULL;
1207 	mcip->mci_direct_rx_arg = NULL;
1208 
1209 	mcip->mci_unicast_list = NULL;
1210 
1211 	if ((flags & MAC_OPEN_FLAGS_IS_VNIC) != 0)
1212 		mcip->mci_state_flags |= MCIS_IS_VNIC;
1213 
1214 	if ((flags & MAC_OPEN_FLAGS_EXCLUSIVE) != 0)
1215 		mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1216 
1217 	if ((flags & MAC_OPEN_FLAGS_IS_AGGR_PORT) != 0)
1218 		mcip->mci_state_flags |= MCIS_IS_AGGR_PORT;
1219 
1220 	if ((flags & MAC_OPEN_FLAGS_USE_DATALINK_NAME) != 0) {
1221 		datalink_id_t	linkid;
1222 
1223 		ASSERT(name == NULL);
1224 		if ((err = dls_devnet_macname2linkid(mip->mi_name,
1225 		    &linkid)) != 0) {
1226 			goto done;
1227 		}
1228 		if ((err = dls_mgmt_get_linkinfo(linkid, mcip->mci_name, NULL,
1229 		    NULL, NULL)) != 0) {
1230 			/*
1231 			 * Use mac name if dlmgmtd is not available.
1232 			 */
1233 			if (err == EBADF) {
1234 				(void) strlcpy(mcip->mci_name, mip->mi_name,
1235 				    sizeof (mcip->mci_name));
1236 				err = 0;
1237 			} else {
1238 				goto done;
1239 			}
1240 		}
1241 		mcip->mci_state_flags |= MCIS_USE_DATALINK_NAME;
1242 	} else {
1243 		ASSERT(name != NULL);
1244 		if (strlen(name) > MAXNAMELEN) {
1245 			err = EINVAL;
1246 			goto done;
1247 		}
1248 		(void) strlcpy(mcip->mci_name, name, sizeof (mcip->mci_name));
1249 	}
1250 
1251 	if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1252 		mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1253 
1254 	/* the subflow table will be created dynamically */
1255 	mcip->mci_subflow_tab = NULL;
1256 	mcip->mci_stat_multircv = 0;
1257 	mcip->mci_stat_brdcstrcv = 0;
1258 	mcip->mci_stat_multixmt = 0;
1259 	mcip->mci_stat_brdcstxmt = 0;
1260 
1261 	mcip->mci_stat_obytes = 0;
1262 	mcip->mci_stat_opackets = 0;
1263 	mcip->mci_stat_oerrors = 0;
1264 	mcip->mci_stat_ibytes = 0;
1265 	mcip->mci_stat_ipackets = 0;
1266 	mcip->mci_stat_ierrors = 0;
1267 
1268 	/* Create an initial flow */
1269 
1270 	err = mac_flow_create(NULL, NULL, mcip->mci_name, NULL,
1271 	    mcip->mci_state_flags & MCIS_IS_VNIC ? FLOW_VNIC_MAC :
1272 	    FLOW_PRIMARY_MAC, &flent);
1273 	if (err != 0)
1274 		goto done;
1275 	mcip->mci_flent = flent;
1276 	FLOW_MARK(flent, FE_MC_NO_DATAPATH);
1277 	flent->fe_mcip = mcip;
1278 	/*
1279 	 * Place initial creation reference on the flow. This reference
1280 	 * is released in the corresponding delete action viz.
1281 	 * mac_unicast_remove after waiting for all transient refs to
1282 	 * to go away. The wait happens in mac_flow_wait.
1283 	 */
1284 	FLOW_REFHOLD(flent);
1285 
1286 	/*
1287 	 * Do this ahead of the mac_bcast_add() below so that the mi_nclients
1288 	 * will have the right value for mac_rx_srs_setup().
1289 	 */
1290 	mac_client_add(mcip);
1291 
1292 	if (no_hwrings)
1293 		mcip->mci_state_flags |= MCIS_NO_HWRINGS;
1294 	if (req_hwrings)
1295 		mcip->mci_state_flags |= MCIS_REQ_HWRINGS;
1296 	mcip->mci_share = NULL;
1297 	if (share_desired) {
1298 		ASSERT(!no_hwrings);
1299 		i_mac_share_alloc(mcip);
1300 	}
1301 
1302 	DTRACE_PROBE2(mac__client__open__allocated, mac_impl_t *,
1303 	    mcip->mci_mip, mac_client_impl_t *, mcip);
1304 	*mchp = (mac_client_handle_t)mcip;
1305 
1306 	i_mac_perim_exit(mip);
1307 	return (0);
1308 
1309 done:
1310 	i_mac_perim_exit(mip);
1311 	mcip->mci_state_flags = 0;
1312 	mcip->mci_tx_flag = 0;
1313 	kmem_cache_free(mac_client_impl_cache, mcip);
1314 	return (err);
1315 }
1316 
1317 /*
1318  * Close the specified MAC client handle.
1319  */
1320 void
1321 mac_client_close(mac_client_handle_t mch, uint16_t flags)
1322 {
1323 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1324 	mac_impl_t		*mip = mcip->mci_mip;
1325 	flow_entry_t		*flent;
1326 
1327 	i_mac_perim_enter(mip);
1328 
1329 	if (flags & MAC_CLOSE_FLAGS_EXCLUSIVE)
1330 		mcip->mci_state_flags &= ~MCIS_EXCLUSIVE;
1331 
1332 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
1333 	    !(flags & MAC_CLOSE_FLAGS_IS_VNIC)) {
1334 		/*
1335 		 * This is an upper VNIC client initiated operation.
1336 		 * The lower MAC client will be closed by the VNIC driver
1337 		 * when the VNIC is deleted.
1338 		 */
1339 
1340 		i_mac_perim_exit(mip);
1341 		return;
1342 	}
1343 
1344 	/*
1345 	 * Remove the flent associated with the MAC client
1346 	 */
1347 	flent = mcip->mci_flent;
1348 	mcip->mci_flent = NULL;
1349 	FLOW_FINAL_REFRELE(flent);
1350 
1351 	/*
1352 	 * MAC clients must remove the unicast addresses and promisc callbacks
1353 	 * they added before issuing a mac_client_close().
1354 	 */
1355 	ASSERT(mcip->mci_unicast_list == NULL);
1356 	ASSERT(mcip->mci_promisc_list == NULL);
1357 	ASSERT(mcip->mci_tx_notify_cb_list == NULL);
1358 
1359 	i_mac_share_free(mcip);
1360 
1361 	mac_client_remove(mcip);
1362 
1363 	i_mac_perim_exit(mip);
1364 	mcip->mci_subflow_tab = NULL;
1365 	mcip->mci_state_flags = 0;
1366 	mcip->mci_tx_flag = 0;
1367 	kmem_cache_free(mac_client_impl_cache, mch);
1368 }
1369 
1370 /*
1371  * Enable bypass for the specified MAC client.
1372  */
1373 boolean_t
1374 mac_rx_bypass_set(mac_client_handle_t mch, mac_direct_rx_t rx_fn, void *arg1)
1375 {
1376 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1377 	mac_impl_t		*mip = mcip->mci_mip;
1378 
1379 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1380 
1381 	/*
1382 	 * If the mac_client is a VLAN, we should not do DLS bypass and
1383 	 * instead let the packets come up via mac_rx_deliver so the vlan
1384 	 * header can be stripped.
1385 	 */
1386 	if (mcip->mci_nvids > 0)
1387 		return (B_FALSE);
1388 
1389 	/*
1390 	 * These are not accessed directly in the data path, and hence
1391 	 * don't need any protection
1392 	 */
1393 	mcip->mci_direct_rx_fn = rx_fn;
1394 	mcip->mci_direct_rx_arg = arg1;
1395 	mcip->mci_state_flags |= MCIS_CLIENT_POLL_CAPABLE;
1396 	return (B_TRUE);
1397 }
1398 
1399 /*
1400  * Set the receive callback for the specified MAC client. There can be
1401  * at most one such callback per MAC client.
1402  */
1403 void
1404 mac_rx_set(mac_client_handle_t mch, mac_rx_t rx_fn, void *arg)
1405 {
1406 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1407 	mac_impl_t	*mip = mcip->mci_mip;
1408 
1409 	/*
1410 	 * Instead of adding an extra set of locks and refcnts in
1411 	 * the datapath at the mac client boundary, we temporarily quiesce
1412 	 * the SRS and related entities. We then change the receive function
1413 	 * without interference from any receive data thread and then reenable
1414 	 * the data flow subsequently.
1415 	 */
1416 	i_mac_perim_enter(mip);
1417 	mac_rx_client_quiesce(mch);
1418 
1419 	mcip->mci_rx_fn = rx_fn;
1420 	mcip->mci_rx_arg = arg;
1421 	mac_rx_client_restart(mch);
1422 	i_mac_perim_exit(mip);
1423 }
1424 
1425 /*
1426  * Reset the receive callback for the specified MAC client.
1427  */
1428 void
1429 mac_rx_clear(mac_client_handle_t mch)
1430 {
1431 	mac_rx_set(mch, mac_pkt_drop, NULL);
1432 }
1433 
1434 /*
1435  * Walk the MAC client subflow table and updates their priority values.
1436  */
1437 static int
1438 mac_update_subflow_priority_cb(flow_entry_t *flent, void *arg)
1439 {
1440 	mac_flow_update_priority(arg, flent);
1441 	return (0);
1442 }
1443 
1444 void
1445 mac_update_subflow_priority(mac_client_impl_t *mcip)
1446 {
1447 	(void) mac_flow_walk(mcip->mci_subflow_tab,
1448 	    mac_update_subflow_priority_cb, mcip);
1449 }
1450 
1451 /*
1452  * When the MAC client is being brought up (i.e. we do a unicast_add) we need
1453  * to initialize the cpu and resource control structure in the
1454  * mac_client_impl_t from the mac_impl_t (i.e if there are any cached
1455  * properties before the flow entry for the unicast address was created).
1456  */
1457 int
1458 mac_resource_ctl_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
1459 {
1460 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
1461 	mac_impl_t		*mip = (mac_impl_t *)mcip->mci_mip;
1462 	int			err = 0;
1463 
1464 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1465 
1466 	err = mac_validate_props(mrp);
1467 	if (err != 0)
1468 		return (err);
1469 
1470 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
1471 	if (MCIP_DATAPATH_SETUP(mcip)) {
1472 		/*
1473 		 * We have to set this prior to calling mac_flow_modify.
1474 		 */
1475 		if (mrp->mrp_mask & MRP_PRIORITY) {
1476 			if (mrp->mrp_priority == MPL_RESET) {
1477 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1478 				    MPL_LINK_DEFAULT);
1479 			} else {
1480 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1481 				    mrp->mrp_priority);
1482 			}
1483 		}
1484 
1485 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
1486 		if (mrp->mrp_mask & MRP_PRIORITY)
1487 			mac_update_subflow_priority(mcip);
1488 		return (0);
1489 	}
1490 	return (0);
1491 }
1492 
1493 void
1494 mac_resource_ctl_get(mac_client_handle_t mch, mac_resource_props_t *mrp)
1495 {
1496 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1497 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
1498 
1499 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
1500 }
1501 
1502 static int
1503 mac_unicast_flow_create(mac_client_impl_t *mcip, uint8_t *mac_addr,
1504     uint16_t vid, boolean_t is_primary, boolean_t first_flow,
1505     flow_entry_t **flent, mac_resource_props_t *mrp)
1506 {
1507 	mac_impl_t	*mip = (mac_impl_t *)mcip->mci_mip;
1508 	flow_desc_t	flow_desc;
1509 	char		flowname[MAXFLOWNAMELEN];
1510 	int		err;
1511 	uint_t		flent_flags;
1512 
1513 	/*
1514 	 * First unicast address being added, create a new flow
1515 	 * for that MAC client.
1516 	 */
1517 	bzero(&flow_desc, sizeof (flow_desc));
1518 
1519 	flow_desc.fd_mac_len = mip->mi_type->mt_addr_length;
1520 	bcopy(mac_addr, flow_desc.fd_dst_mac, flow_desc.fd_mac_len);
1521 	flow_desc.fd_mask = FLOW_LINK_DST;
1522 	if (vid != 0) {
1523 		flow_desc.fd_vid = vid;
1524 		flow_desc.fd_mask |= FLOW_LINK_VID;
1525 	}
1526 
1527 	/*
1528 	 * XXX-nicolas. For now I'm keeping the FLOW_PRIMARY_MAC
1529 	 * and FLOW_VNIC. Even though they're a hack inherited
1530 	 * from the SRS code, we'll keep them for now. They're currently
1531 	 * consumed by mac_datapath_setup() to create the SRS.
1532 	 * That code should be eventually moved out of
1533 	 * mac_datapath_setup() and moved to a mac_srs_create()
1534 	 * function of some sort to keep things clean.
1535 	 *
1536 	 * Also, there's no reason why the SRS for the primary MAC
1537 	 * client should be different than any other MAC client. Until
1538 	 * this is cleaned-up, we support only one MAC unicast address
1539 	 * per client.
1540 	 *
1541 	 * We set FLOW_PRIMARY_MAC for the primary MAC address,
1542 	 * FLOW_VNIC for everything else.
1543 	 */
1544 	if (is_primary)
1545 		flent_flags = FLOW_PRIMARY_MAC;
1546 	else
1547 		flent_flags = FLOW_VNIC_MAC;
1548 
1549 	/*
1550 	 * For the first flow we use the mac client's name - mci_name, for
1551 	 * subsequent ones we just create a name with the vid. This is
1552 	 * so that we can add these flows to the same flow table. This is
1553 	 * fine as the flow name (except for the one with the mac client's
1554 	 * name) is not visible. When the first flow is removed, we just replace
1555 	 * its fdesc with another from the list, so we will still retain the
1556 	 * flent with the MAC client's flow name.
1557 	 */
1558 	if (first_flow) {
1559 		bcopy(mcip->mci_name, flowname, MAXFLOWNAMELEN);
1560 	} else {
1561 		(void) sprintf(flowname, "%s%u", mcip->mci_name, vid);
1562 		flent_flags = FLOW_NO_STATS;
1563 	}
1564 
1565 	if ((err = mac_flow_create(&flow_desc, mrp, flowname, NULL,
1566 	    flent_flags, flent)) != 0)
1567 		return (err);
1568 
1569 	FLOW_MARK(*flent, FE_INCIPIENT);
1570 	(*flent)->fe_mcip = mcip;
1571 
1572 	/*
1573 	 * Place initial creation reference on the flow. This reference
1574 	 * is released in the corresponding delete action viz.
1575 	 * mac_unicast_remove after waiting for all transient refs to
1576 	 * to go away. The wait happens in mac_flow_wait.
1577 	 * We have already held the reference in mac_client_open().
1578 	 */
1579 	if (!first_flow)
1580 		FLOW_REFHOLD(*flent);
1581 	return (0);
1582 }
1583 
1584 /* Refresh the multicast grouping for this VID. */
1585 int
1586 mac_client_update_mcast(void *arg, boolean_t add, const uint8_t *addrp)
1587 {
1588 	flow_entry_t		*flent = arg;
1589 	mac_client_impl_t	*mcip = flent->fe_mcip;
1590 	uint16_t		vid;
1591 	flow_desc_t		flow_desc;
1592 
1593 	mac_flow_get_desc(flent, &flow_desc);
1594 	vid = (flow_desc.fd_mask & FLOW_LINK_VID) != 0 ?
1595 	    flow_desc.fd_vid : VLAN_ID_NONE;
1596 
1597 	/*
1598 	 * We don't call mac_multicast_add()/mac_multicast_remove() as
1599 	 * we want to add/remove for this specific vid.
1600 	 */
1601 	if (add) {
1602 		return (mac_bcast_add(mcip, addrp, vid,
1603 		    MAC_ADDRTYPE_MULTICAST));
1604 	} else {
1605 		mac_bcast_delete(mcip, addrp, vid);
1606 		return (0);
1607 	}
1608 }
1609 
1610 static void
1611 mac_update_single_active_client(mac_impl_t *mip)
1612 {
1613 	mac_client_impl_t *client = NULL;
1614 
1615 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1616 
1617 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1618 	if (mip->mi_nactiveclients == 1) {
1619 		/*
1620 		 * Find the one active MAC client from the list of MAC
1621 		 * clients. The active MAC client has at least one
1622 		 * unicast address.
1623 		 */
1624 		for (client = mip->mi_clients_list; client != NULL;
1625 		    client = client->mci_client_next) {
1626 			if (client->mci_unicast_list != NULL)
1627 				break;
1628 		}
1629 		ASSERT(client != NULL);
1630 	}
1631 
1632 	/*
1633 	 * mi_single_active_client is protected by the MAC impl's read/writer
1634 	 * lock, which allows mac_rx() to check the value of that pointer
1635 	 * as a reader.
1636 	 */
1637 	mip->mi_single_active_client = client;
1638 	rw_exit(&mip->mi_rw_lock);
1639 }
1640 
1641 /*
1642  * Set up the data path. Called from i_mac_unicast_add after having
1643  * done all the validations including making sure this is an active
1644  * client (i.e that is ready to process packets.)
1645  */
1646 static int
1647 mac_client_datapath_setup(mac_client_impl_t *mcip, uint16_t vid,
1648     uint8_t *mac_addr, mac_resource_props_t *mrp, boolean_t isprimary,
1649     mac_unicast_impl_t *muip)
1650 {
1651 	mac_impl_t	*mip = mcip->mci_mip;
1652 	boolean_t	mac_started = B_FALSE;
1653 	boolean_t	bcast_added = B_FALSE;
1654 	boolean_t	nactiveclients_added = B_FALSE;
1655 	flow_entry_t	*flent;
1656 	int		err = 0;
1657 
1658 	if ((err = mac_start((mac_handle_t)mip)) != 0)
1659 		goto bail;
1660 
1661 	mac_started = B_TRUE;
1662 
1663 	/* add the MAC client to the broadcast address group by default */
1664 	if (mip->mi_type->mt_brdcst_addr != NULL) {
1665 		err = mac_bcast_add(mcip, mip->mi_type->mt_brdcst_addr, vid,
1666 		    MAC_ADDRTYPE_BROADCAST);
1667 		if (err != 0)
1668 			goto bail;
1669 		bcast_added = B_TRUE;
1670 	}
1671 
1672 	/*
1673 	 * If this is the first unicast address addition for this
1674 	 * client, reuse the pre-allocated larval flow entry associated with
1675 	 * the MAC client.
1676 	 */
1677 	flent = (mcip->mci_nflents == 0) ? mcip->mci_flent : NULL;
1678 
1679 	/* We are configuring the unicast flow now */
1680 	if (!MCIP_DATAPATH_SETUP(mcip)) {
1681 
1682 		MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1683 		    (mrp->mrp_mask & MRP_PRIORITY) ? mrp->mrp_priority :
1684 		    MPL_LINK_DEFAULT);
1685 
1686 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
1687 		    isprimary, B_TRUE, &flent, mrp)) != 0)
1688 			goto bail;
1689 
1690 		mip->mi_nactiveclients++;
1691 		nactiveclients_added = B_TRUE;
1692 
1693 		/*
1694 		 * This will allocate the RX ring group if possible for the
1695 		 * flow and program the software classifier as needed.
1696 		 */
1697 		if ((err = mac_datapath_setup(mcip, flent, SRST_LINK)) != 0)
1698 			goto bail;
1699 
1700 		/*
1701 		 * The unicast MAC address must have been added successfully.
1702 		 */
1703 		ASSERT(mcip->mci_unicast != NULL);
1704 		/*
1705 		 * Push down the sub-flows that were defined on this link
1706 		 * hitherto. The flows are added to the active flow table
1707 		 * and SRS, softrings etc. are created as needed.
1708 		 */
1709 		mac_link_init_flows((mac_client_handle_t)mcip);
1710 	} else {
1711 		mac_address_t *map = mcip->mci_unicast;
1712 
1713 		/*
1714 		 * A unicast flow already exists for that MAC client,
1715 		 * this flow must be the same mac address but with
1716 		 * different VID. It has been checked by mac_addr_in_use().
1717 		 *
1718 		 * We will use the SRS etc. from the mci_flent. Note that
1719 		 * We don't need to create kstat for this as except for
1720 		 * the fdesc, everything will be used from in the 1st flent.
1721 		 */
1722 
1723 		if (bcmp(mac_addr, map->ma_addr, map->ma_len) != 0) {
1724 			err = EINVAL;
1725 			goto bail;
1726 		}
1727 
1728 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
1729 		    isprimary, B_FALSE, &flent, NULL)) != 0) {
1730 			goto bail;
1731 		}
1732 		if ((err = mac_flow_add(mip->mi_flow_tab, flent)) != 0) {
1733 			FLOW_FINAL_REFRELE(flent);
1734 			goto bail;
1735 		}
1736 
1737 		/* update the multicast group for this vid */
1738 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
1739 		    (void *)flent, B_TRUE);
1740 
1741 	}
1742 
1743 	/* populate the shared MAC address */
1744 	muip->mui_map = mcip->mci_unicast;
1745 
1746 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
1747 	muip->mui_next = mcip->mci_unicast_list;
1748 	mcip->mci_unicast_list = muip;
1749 	rw_exit(&mcip->mci_rw_lock);
1750 
1751 
1752 	/*
1753 	 * First add the flent to the flow list of this mcip. Then set
1754 	 * the mip's mi_single_active_client if needed. The Rx path assumes
1755 	 * that mip->mi_single_active_client will always have an associated
1756 	 * flent.
1757 	 */
1758 	mac_client_add_to_flow_list(mcip, flent);
1759 
1760 	if (nactiveclients_added)
1761 		mac_update_single_active_client(mip);
1762 	/*
1763 	 * Trigger a renegotiation of the capabilities when the number of
1764 	 * active clients changes from 1 to 2, since some of the capabilities
1765 	 * might have to be disabled. Also send a MAC_NOTE_LINK notification
1766 	 * to all the MAC clients whenever physical link is DOWN.
1767 	 */
1768 	if (mip->mi_nactiveclients == 2) {
1769 		mac_capab_update((mac_handle_t)mip);
1770 		mac_virtual_link_update(mip);
1771 	}
1772 	/*
1773 	 * Now that the setup is complete, clear the INCIPIENT flag.
1774 	 * The flag was set to avoid incoming packets seeing inconsistent
1775 	 * structures while the setup was in progress. Clear the mci_tx_flag
1776 	 * by calling mac_tx_client_block. It is possible that
1777 	 * mac_unicast_remove was called prior to this mac_unicast_add which
1778 	 * could have set the MCI_TX_QUIESCE flag.
1779 	 */
1780 	if (flent->fe_rx_ring_group != NULL)
1781 		mac_rx_group_unmark(flent->fe_rx_ring_group, MR_INCIPIENT);
1782 	FLOW_UNMARK(flent, FE_INCIPIENT);
1783 	FLOW_UNMARK(flent, FE_MC_NO_DATAPATH);
1784 	mac_tx_client_unblock(mcip);
1785 	return (0);
1786 bail:
1787 	if (bcast_added)
1788 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr, vid);
1789 
1790 	if (nactiveclients_added)
1791 		mip->mi_nactiveclients--;
1792 
1793 	if (mac_started)
1794 		mac_stop((mac_handle_t)mip);
1795 
1796 	return (err);
1797 }
1798 
1799 /*
1800  * Return the passive primary MAC client, if present. The passive client is
1801  * a stand-by client that has the same unicast address as another that is
1802  * currenly active. Once the active client goes away, the passive client
1803  * becomes active.
1804  */
1805 static mac_client_impl_t *
1806 mac_get_passive_primary_client(mac_impl_t *mip)
1807 {
1808 	mac_client_impl_t	*mcip;
1809 
1810 	for (mcip = mip->mi_clients_list; mcip != NULL;
1811 	    mcip = mcip->mci_client_next) {
1812 		if (mac_is_primary_client(mcip) &&
1813 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
1814 			return (mcip);
1815 		}
1816 	}
1817 	return (NULL);
1818 }
1819 
1820 /*
1821  * Add a new unicast address to the MAC client.
1822  *
1823  * The MAC address can be specified either by value, or the MAC client
1824  * can specify that it wants to use the primary MAC address of the
1825  * underlying MAC. See the introductory comments at the beginning
1826  * of this file for more more information on primary MAC addresses.
1827  *
1828  * Note also the tuple (MAC address, VID) must be unique
1829  * for the MAC clients defined on top of the same underlying MAC
1830  * instance, unless the MAC_UNICAST_NODUPCHECK is specified.
1831  */
1832 int
1833 i_mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
1834     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
1835 {
1836 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1837 	mac_impl_t		*mip = mcip->mci_mip;
1838 	int			err;
1839 	uint_t			mac_len = mip->mi_type->mt_addr_length;
1840 	boolean_t		check_dups = !(flags & MAC_UNICAST_NODUPCHECK);
1841 	boolean_t		fastpath_disabled = B_FALSE;
1842 	boolean_t		is_primary = (flags & MAC_UNICAST_PRIMARY);
1843 	boolean_t		is_unicast_hw = (flags & MAC_UNICAST_HW);
1844 	mac_resource_props_t	mrp;
1845 	boolean_t		passive_client = B_FALSE;
1846 	mac_unicast_impl_t	*muip;
1847 	boolean_t		is_vnic_primary =
1848 	    (flags & MAC_UNICAST_VNIC_PRIMARY);
1849 
1850 	/* when VID is non-zero, the underlying MAC can not be VNIC */
1851 	ASSERT(!((mip->mi_state_flags & MIS_IS_VNIC) && (vid != 0)));
1852 
1853 	/*
1854 	 * Check whether it's the primary client and flag it.
1855 	 */
1856 	if (!(mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary && vid == 0)
1857 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PRIMARY;
1858 
1859 	/*
1860 	 * is_vnic_primary is true when we come here as a VLAN VNIC
1861 	 * which uses the primary mac client's address but with a non-zero
1862 	 * VID. In this case the MAC address is not specified by an upper
1863 	 * MAC client.
1864 	 */
1865 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary &&
1866 	    !is_vnic_primary) {
1867 		/*
1868 		 * The address is being set by the upper MAC client
1869 		 * of a VNIC. The MAC address was already set by the
1870 		 * VNIC driver during VNIC creation.
1871 		 *
1872 		 * Note: a VNIC has only one MAC address. We return
1873 		 * the MAC unicast address handle of the lower MAC client
1874 		 * corresponding to the VNIC. We allocate a new entry
1875 		 * which is flagged appropriately, so that mac_unicast_remove()
1876 		 * doesn't attempt to free the original entry that
1877 		 * was allocated by the VNIC driver.
1878 		 */
1879 		ASSERT(mcip->mci_unicast != NULL);
1880 
1881 		/* Check for VLAN flags, if present */
1882 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
1883 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
1884 
1885 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
1886 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
1887 
1888 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
1889 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
1890 
1891 		/*
1892 		 * Ensure that the primary unicast address of the VNIC
1893 		 * is added only once unless we have the
1894 		 * MAC_CLIENT_FLAGS_MULTI_PRIMARY set (and this is not
1895 		 * a passive MAC client).
1896 		 */
1897 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) != 0) {
1898 			if ((mcip->mci_flags &
1899 			    MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
1900 			    (mcip->mci_flags &
1901 			    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
1902 				return (EBUSY);
1903 			}
1904 			mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
1905 			passive_client = B_TRUE;
1906 		}
1907 
1908 		mcip->mci_flags |= MAC_CLIENT_FLAGS_VNIC_PRIMARY;
1909 
1910 		/*
1911 		 * Create a handle for vid 0.
1912 		 */
1913 		ASSERT(vid == 0);
1914 		muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
1915 		muip->mui_vid = vid;
1916 		*mah = (mac_unicast_handle_t)muip;
1917 		/*
1918 		 * This will be used by the caller to defer setting the
1919 		 * rx functions.
1920 		 */
1921 		if (passive_client)
1922 			return (EAGAIN);
1923 		return (0);
1924 	}
1925 
1926 	/* primary MAC clients cannot be opened on top of anchor VNICs */
1927 	if ((is_vnic_primary || is_primary) &&
1928 	    i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_ANCHOR_VNIC, NULL)) {
1929 		return (ENXIO);
1930 	}
1931 
1932 	/*
1933 	 * If this is a VNIC/VLAN, disable softmac fast-path.
1934 	 */
1935 	if (mcip->mci_state_flags & MCIS_IS_VNIC) {
1936 		err = mac_fastpath_disable((mac_handle_t)mip);
1937 		if (err != 0)
1938 			return (err);
1939 		fastpath_disabled = B_TRUE;
1940 	}
1941 
1942 	/*
1943 	 * Return EBUSY if:
1944 	 *  - there is an exclusively active mac client exists.
1945 	 *  - this is an exclusive active mac client but
1946 	 *	a. there is already active mac clients exist, or
1947 	 *	b. fastpath streams are already plumbed on this legacy device
1948 	 */
1949 	if (mip->mi_state_flags & MIS_EXCLUSIVE) {
1950 		if (fastpath_disabled)
1951 			mac_fastpath_enable((mac_handle_t)mip);
1952 		return (EBUSY);
1953 	}
1954 
1955 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
1956 		ASSERT(!fastpath_disabled);
1957 		if (mip->mi_nactiveclients != 0)
1958 			return (EBUSY);
1959 
1960 		if ((mip->mi_state_flags & MIS_LEGACY) &&
1961 		    !(mip->mi_capab_legacy.ml_active_set(mip->mi_driver))) {
1962 			return (EBUSY);
1963 		}
1964 		mip->mi_state_flags |= MIS_EXCLUSIVE;
1965 	}
1966 
1967 	bzero(&mrp, sizeof (mac_resource_props_t));
1968 	if (is_primary && !(mcip->mci_state_flags & (MCIS_IS_VNIC |
1969 	    MCIS_IS_AGGR_PORT))) {
1970 		/*
1971 		 * Apply the property cached in the mac_impl_t to the primary
1972 		 * mac client. If the mac client is a VNIC or an aggregation
1973 		 * port, its property should be set in the mcip when the
1974 		 * VNIC/aggr was created.
1975 		 */
1976 		mac_get_resources((mac_handle_t)mip, &mrp);
1977 		(void) mac_client_set_resources(mch, &mrp);
1978 	} else if (mcip->mci_state_flags & MCIS_IS_VNIC) {
1979 		bcopy(MCIP_RESOURCE_PROPS(mcip), &mrp,
1980 		    sizeof (mac_resource_props_t));
1981 	}
1982 
1983 	muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
1984 	muip->mui_vid = vid;
1985 
1986 	if (is_primary || is_vnic_primary) {
1987 		mac_addr = mip->mi_addr;
1988 	} else {
1989 
1990 		/*
1991 		 * Verify the validity of the specified MAC addresses value.
1992 		 */
1993 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr, mac_len)) {
1994 			*diag = MAC_DIAG_MACADDR_INVALID;
1995 			err = EINVAL;
1996 			goto bail_out;
1997 		}
1998 
1999 		/*
2000 		 * Make sure that the specified MAC address is different
2001 		 * than the unicast MAC address of the underlying NIC.
2002 		 */
2003 		if (check_dups && bcmp(mip->mi_addr, mac_addr, mac_len) == 0) {
2004 			*diag = MAC_DIAG_MACADDR_NIC;
2005 			err = EINVAL;
2006 			goto bail_out;
2007 		}
2008 	}
2009 
2010 	/*
2011 	 * Set the flags here so that if this is a passive client, we
2012 	 * can return  and set it when we call mac_client_datapath_setup
2013 	 * when this becomes the active client. If we defer to using these
2014 	 * flags to mac_client_datapath_setup, then for a passive client,
2015 	 * we'd have to store the flags somewhere (probably fe_flags)
2016 	 * and then use it.
2017 	 */
2018 	if (!MCIP_DATAPATH_SETUP(mcip)) {
2019 		if (is_unicast_hw) {
2020 			/*
2021 			 * The client requires a hardware MAC address slot
2022 			 * for that unicast address. Since we support only
2023 			 * one unicast MAC address per client, flag the
2024 			 * MAC client itself.
2025 			 */
2026 			mcip->mci_state_flags |= MCIS_UNICAST_HW;
2027 		}
2028 
2029 		/* Check for VLAN flags, if present */
2030 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2031 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2032 
2033 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2034 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2035 
2036 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2037 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2038 	} else {
2039 		/*
2040 		 * Assert that the specified flags are consistent with the
2041 		 * flags specified by previous calls to mac_unicast_add().
2042 		 */
2043 		ASSERT(((flags & MAC_UNICAST_TAG_DISABLE) != 0 &&
2044 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) != 0) ||
2045 		    ((flags & MAC_UNICAST_TAG_DISABLE) == 0 &&
2046 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) == 0));
2047 
2048 		ASSERT(((flags & MAC_UNICAST_STRIP_DISABLE) != 0 &&
2049 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) != 0) ||
2050 		    ((flags & MAC_UNICAST_STRIP_DISABLE) == 0 &&
2051 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) == 0));
2052 
2053 		ASSERT(((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0 &&
2054 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) != 0) ||
2055 		    ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) == 0 &&
2056 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) == 0));
2057 
2058 		/*
2059 		 * Make sure the client is consistent about its requests
2060 		 * for MAC addresses. I.e. all requests from the clients
2061 		 * must have the MAC_UNICAST_HW flag set or clear.
2062 		 */
2063 		if ((mcip->mci_state_flags & MCIS_UNICAST_HW) != 0 &&
2064 		    !is_unicast_hw ||
2065 		    (mcip->mci_state_flags & MCIS_UNICAST_HW) == 0 &&
2066 		    is_unicast_hw) {
2067 			err = EINVAL;
2068 			goto bail_out;
2069 		}
2070 	}
2071 	/*
2072 	 * Make sure the MAC address is not already used by
2073 	 * another MAC client defined on top of the same
2074 	 * underlying NIC. Unless we have MAC_CLIENT_FLAGS_MULTI_PRIMARY
2075 	 * set when we allow a passive client to be present which will
2076 	 * be activated when the currently active client goes away - this
2077 	 * works only with primary addresses.
2078 	 */
2079 	if ((check_dups || is_primary || is_vnic_primary) &&
2080 	    mac_addr_in_use(mip, mac_addr, vid)) {
2081 		/*
2082 		 * Must have set the multiple primary address flag when
2083 		 * we did a mac_client_open AND this should be a primary
2084 		 * MAC client AND there should not already be a passive
2085 		 * primary. If all is true then we let this succeed
2086 		 * even if the address is a dup.
2087 		 */
2088 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2089 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) == 0 ||
2090 		    mac_get_passive_primary_client(mip) != NULL) {
2091 			*diag = MAC_DIAG_MACADDR_INUSE;
2092 			err = EEXIST;
2093 			goto bail_out;
2094 		}
2095 		ASSERT((mcip->mci_flags &
2096 		    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) == 0);
2097 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2098 
2099 		/*
2100 		 * Stash the unicast address handle, we will use it when
2101 		 * we set up the passive client.
2102 		 */
2103 		mcip->mci_p_unicast_list = muip;
2104 		*mah = (mac_unicast_handle_t)muip;
2105 		return (0);
2106 	}
2107 
2108 	err = mac_client_datapath_setup(mcip, vid, mac_addr, &mrp,
2109 	    is_primary || is_vnic_primary, muip);
2110 	if (err != 0)
2111 		goto bail_out;
2112 	*mah = (mac_unicast_handle_t)muip;
2113 	return (0);
2114 
2115 bail_out:
2116 	if (fastpath_disabled)
2117 		mac_fastpath_enable((mac_handle_t)mip);
2118 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2119 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2120 		if (mip->mi_state_flags & MIS_LEGACY) {
2121 			mip->mi_capab_legacy.ml_active_clear(
2122 			    mip->mi_driver);
2123 		}
2124 	}
2125 	kmem_free(muip, sizeof (mac_unicast_impl_t));
2126 	return (err);
2127 }
2128 
2129 /*
2130  * Wrapper function to mac_unicast_add when we want to have the same mac
2131  * client open for two instances, one that is currently active and another
2132  * that will become active when the current one is removed. In this case
2133  * mac_unicast_add will return EGAIN and we will save the rx function and
2134  * arg which will be used when we activate the passive client in
2135  * mac_unicast_remove.
2136  */
2137 int
2138 mac_unicast_add_set_rx(mac_client_handle_t mch, uint8_t *mac_addr,
2139     uint16_t flags, mac_unicast_handle_t *mah,  uint16_t vid, mac_diag_t *diag,
2140     mac_rx_t rx_fn, void *arg)
2141 {
2142 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2143 	uint_t			err;
2144 
2145 	err = mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2146 	if (err != 0 && err != EAGAIN)
2147 		return (err);
2148 	if (err == EAGAIN) {
2149 		if (rx_fn != NULL) {
2150 			mcip->mci_rx_p_fn = rx_fn;
2151 			mcip->mci_rx_p_arg = arg;
2152 		}
2153 		return (0);
2154 	}
2155 	if (rx_fn != NULL)
2156 		mac_rx_set(mch, rx_fn, arg);
2157 	return (err);
2158 }
2159 
2160 int
2161 mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2162     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2163 {
2164 	mac_impl_t *mip = ((mac_client_impl_t *)mch)->mci_mip;
2165 	uint_t err;
2166 
2167 	i_mac_perim_enter(mip);
2168 	err = i_mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2169 	i_mac_perim_exit(mip);
2170 
2171 	return (err);
2172 }
2173 
2174 void
2175 mac_client_datapath_teardown(mac_client_handle_t mch, mac_unicast_impl_t *muip,
2176     flow_entry_t *flent)
2177 {
2178 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2179 	mac_impl_t		*mip = mcip->mci_mip;
2180 
2181 	/*
2182 	 * We would have initialized subflows etc. only if we brought up
2183 	 * the primary client and set the unicast unicast address etc.
2184 	 * Deactivate the flows. The flow entry will be removed from the
2185 	 * active flow tables, and the associated SRS, softrings etc will
2186 	 * be deleted. But the flow entry itself won't be destroyed, instead
2187 	 * it will continue to be archived off the  the global flow hash
2188 	 * list, for a possible future activation when say IP is plumbed
2189 	 * again.
2190 	 */
2191 	mac_link_release_flows(mch);
2192 
2193 	mip->mi_nactiveclients--;
2194 	mac_update_single_active_client(mip);
2195 
2196 	/* Tear down the data path */
2197 	mac_datapath_teardown(mcip, mcip->mci_flent, SRST_LINK);
2198 
2199 	/*
2200 	 * Prevent any future access to the flow entry through the mci_flent
2201 	 * pointer by setting the mci_flent to NULL. Access to mci_flent in
2202 	 * mac_bcast_send is also under mi_rw_lock.
2203 	 */
2204 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2205 	flent = mcip->mci_flent;
2206 	mac_client_remove_flow_from_list(mcip, flent);
2207 
2208 	if (mcip->mci_state_flags & MCIS_DESC_LOGGED)
2209 		mcip->mci_state_flags &= ~MCIS_DESC_LOGGED;
2210 
2211 	/*
2212 	 * This is the last unicast address being removed and there shouldn't
2213 	 * be any outbound data threads at this point coming down from mac
2214 	 * clients. We have waited for the data threads to finish before
2215 	 * starting dld_str_detach. Non-data threads must access TX SRS
2216 	 * under mi_rw_lock.
2217 	 */
2218 	rw_exit(&mip->mi_rw_lock);
2219 
2220 	/*
2221 	 * Don't use FLOW_MARK with FE_MC_NO_DATAPATH, as the flow might
2222 	 * contain other flags, such as FE_CONDEMNED, which we need to
2223 	 * cleared. We don't call mac_flow_cleanup() for this unicast
2224 	 * flow as we have a already cleaned up SRSs etc. (via the teadown
2225 	 * path). We just clear the stats and reset the initial callback
2226 	 * function, the rest will be set when we call mac_flow_create,
2227 	 * if at all.
2228 	 */
2229 	mutex_enter(&flent->fe_lock);
2230 	ASSERT(flent->fe_refcnt == 1 && flent->fe_mbg == NULL &&
2231 	    flent->fe_tx_srs == NULL && flent->fe_rx_srs_cnt == 0);
2232 	flent->fe_flags = FE_MC_NO_DATAPATH;
2233 	flow_stat_destroy(flent);
2234 
2235 	/* Initialize the receiver function to a safe routine */
2236 	flent->fe_cb_fn = (flow_fn_t)mac_pkt_drop;
2237 	flent->fe_cb_arg1 = NULL;
2238 	flent->fe_cb_arg2 = NULL;
2239 
2240 	flent->fe_index = -1;
2241 	mutex_exit(&flent->fe_lock);
2242 
2243 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2244 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2245 		    muip->mui_vid);
2246 	}
2247 
2248 	if (mip->mi_nactiveclients == 1) {
2249 		mac_capab_update((mac_handle_t)mip);
2250 		mac_virtual_link_update(mip);
2251 	}
2252 
2253 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2254 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2255 
2256 		if (mip->mi_state_flags & MIS_LEGACY)
2257 			mip->mi_capab_legacy.ml_active_clear(mip->mi_driver);
2258 	}
2259 
2260 	mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2261 
2262 	if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2263 		mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2264 
2265 	if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2266 		mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2267 
2268 	if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2269 		mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2270 
2271 	kmem_free(muip, sizeof (mac_unicast_impl_t));
2272 
2273 	/*
2274 	 * Disable fastpath if this is a VNIC or a VLAN.
2275 	 */
2276 	if (mcip->mci_state_flags & MCIS_IS_VNIC)
2277 		mac_fastpath_enable((mac_handle_t)mip);
2278 	mac_stop((mac_handle_t)mip);
2279 }
2280 
2281 /*
2282  * Remove a MAC address which was previously added by mac_unicast_add().
2283  */
2284 int
2285 mac_unicast_remove(mac_client_handle_t mch, mac_unicast_handle_t mah)
2286 {
2287 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2288 	mac_unicast_impl_t *muip = (mac_unicast_impl_t *)mah;
2289 	mac_unicast_impl_t *pre;
2290 	mac_impl_t *mip = mcip->mci_mip;
2291 	flow_entry_t		*flent;
2292 	boolean_t		isprimary = B_FALSE;
2293 
2294 	i_mac_perim_enter(mip);
2295 	if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) {
2296 		/*
2297 		 * Called made by the upper MAC client of a VNIC.
2298 		 * There's nothing much to do, the unicast address will
2299 		 * be removed by the VNIC driver when the VNIC is deleted,
2300 		 * but let's ensure that all our transmit is done before
2301 		 * the client does a mac_client_stop lest it trigger an
2302 		 * assert in the driver.
2303 		 */
2304 		ASSERT(muip->mui_vid == 0);
2305 
2306 		mac_tx_client_flush(mcip);
2307 
2308 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2309 			mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2310 			if (mcip->mci_rx_p_fn != NULL) {
2311 				mac_rx_set(mch, mcip->mci_rx_p_fn,
2312 				    mcip->mci_rx_p_arg);
2313 				mcip->mci_rx_p_fn = NULL;
2314 				mcip->mci_rx_p_arg = NULL;
2315 			}
2316 			kmem_free(muip, sizeof (mac_unicast_impl_t));
2317 			i_mac_perim_exit(mip);
2318 			return (0);
2319 		}
2320 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2321 
2322 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2323 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2324 
2325 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2326 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2327 
2328 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2329 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2330 
2331 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2332 		i_mac_perim_exit(mip);
2333 		return (0);
2334 	}
2335 
2336 	ASSERT(muip != NULL);
2337 
2338 	/*
2339 	 * We are removing a passive client, we haven't setup the datapath
2340 	 * for this yet, so nothing much to do.
2341 	 */
2342 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2343 
2344 		ASSERT((mcip->mci_flent->fe_flags & FE_MC_NO_DATAPATH) != 0);
2345 		ASSERT(mcip->mci_p_unicast_list == muip);
2346 
2347 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2348 
2349 		mcip->mci_p_unicast_list = NULL;
2350 		mcip->mci_rx_p_fn = NULL;
2351 		mcip->mci_rx_p_arg = NULL;
2352 
2353 		mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2354 
2355 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2356 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2357 
2358 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2359 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2360 
2361 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2362 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2363 
2364 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2365 		i_mac_perim_exit(mip);
2366 		return (0);
2367 	}
2368 	/*
2369 	 * Remove the VID from the list of client's VIDs.
2370 	 */
2371 	pre = mcip->mci_unicast_list;
2372 	if (muip == pre) {
2373 		mcip->mci_unicast_list = muip->mui_next;
2374 	} else {
2375 		while ((pre->mui_next != NULL) && (pre->mui_next != muip))
2376 			pre = pre->mui_next;
2377 		ASSERT(pre->mui_next == muip);
2378 		rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2379 		pre->mui_next = muip->mui_next;
2380 		rw_exit(&mcip->mci_rw_lock);
2381 	}
2382 
2383 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) &&
2384 	    muip->mui_vid == 0) {
2385 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PRIMARY;
2386 		isprimary = B_TRUE;
2387 	}
2388 	if (!mac_client_single_rcvr(mcip)) {
2389 		/*
2390 		 * This MAC client is shared by more than one unicast
2391 		 * addresses, so we will just remove the flent
2392 		 * corresponding to the address being removed. We don't invoke
2393 		 * mac_rx_classify_flow_rem() since the additional flow is
2394 		 * not associated with its own separate set of SRS and rings,
2395 		 * and these constructs are still needed for the remaining
2396 		 * flows.
2397 		 */
2398 		flent = mac_client_get_flow(mcip, muip);
2399 		ASSERT(flent != NULL);
2400 
2401 		/*
2402 		 * The first one is disappearing, need to make sure
2403 		 * we replace it with another from the list of
2404 		 * shared clients.
2405 		 */
2406 		if (flent == mcip->mci_flent)
2407 			flent = mac_client_swap_mciflent(mcip);
2408 		mac_client_remove_flow_from_list(mcip, flent);
2409 		mac_flow_remove(mip->mi_flow_tab, flent, B_FALSE);
2410 		mac_flow_wait(flent, FLOW_DRIVER_UPCALL);
2411 
2412 		/*
2413 		 * The multicast groups that were added by the client so
2414 		 * far must be removed from the brodcast domain corresponding
2415 		 * to the VID being removed.
2416 		 */
2417 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2418 		    (void *)flent, B_FALSE);
2419 
2420 		if (mip->mi_type->mt_brdcst_addr != NULL) {
2421 			mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2422 			    muip->mui_vid);
2423 		}
2424 
2425 		FLOW_FINAL_REFRELE(flent);
2426 		ASSERT(!(mcip->mci_state_flags & MCIS_EXCLUSIVE));
2427 		/*
2428 		 * Enable fastpath if this is a VNIC or a VLAN.
2429 		 */
2430 		if (mcip->mci_state_flags & MCIS_IS_VNIC)
2431 			mac_fastpath_enable((mac_handle_t)mip);
2432 		mac_stop((mac_handle_t)mip);
2433 		i_mac_perim_exit(mip);
2434 		return (0);
2435 	}
2436 
2437 	mac_client_datapath_teardown(mch, muip, flent);
2438 
2439 	/*
2440 	 * If we are removing the primary, check if we have a passive primary
2441 	 * client that we need to activate now.
2442 	 */
2443 	if (!isprimary) {
2444 		i_mac_perim_exit(mip);
2445 		return (0);
2446 	}
2447 	mcip = mac_get_passive_primary_client(mip);
2448 	if (mcip != NULL) {
2449 		mac_resource_props_t	mrp;
2450 		mac_unicast_impl_t	*muip;
2451 
2452 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2453 		bzero(&mrp, sizeof (mac_resource_props_t));
2454 		/*
2455 		 * Apply the property cached in the mac_impl_t to the
2456 		 * primary mac client.
2457 		 */
2458 		mac_get_resources((mac_handle_t)mip, &mrp);
2459 		(void) mac_client_set_resources(mch, &mrp);
2460 		ASSERT(mcip->mci_p_unicast_list != NULL);
2461 		muip = mcip->mci_p_unicast_list;
2462 		mcip->mci_p_unicast_list = NULL;
2463 		if (mac_client_datapath_setup(mcip, VLAN_ID_NONE,
2464 		    mip->mi_addr, &mrp, B_TRUE, muip) == 0) {
2465 			if (mcip->mci_rx_p_fn != NULL) {
2466 				mac_rx_set(mch, mcip->mci_rx_p_fn,
2467 				    mcip->mci_rx_p_arg);
2468 				mcip->mci_rx_p_fn = NULL;
2469 				mcip->mci_rx_p_arg = NULL;
2470 			}
2471 		} else {
2472 			kmem_free(muip, sizeof (mac_unicast_impl_t));
2473 		}
2474 	}
2475 	i_mac_perim_exit(mip);
2476 	return (0);
2477 }
2478 
2479 /*
2480  * Multicast add function invoked by MAC clients.
2481  */
2482 int
2483 mac_multicast_add(mac_client_handle_t mch, const uint8_t *addr)
2484 {
2485 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2486 	mac_impl_t		*mip = mcip->mci_mip;
2487 	flow_entry_t		*flent = mcip->mci_flent_list;
2488 	flow_entry_t		*prev_fe = NULL;
2489 	uint16_t		vid;
2490 	int			err = 0;
2491 
2492 	/* Verify the address is a valid multicast address */
2493 	if ((err = mip->mi_type->mt_ops.mtops_multicst_verify(addr,
2494 	    mip->mi_pdata)) != 0)
2495 		return (err);
2496 
2497 	i_mac_perim_enter(mip);
2498 	while (flent != NULL) {
2499 		vid = i_mac_flow_vid(flent);
2500 
2501 		err = mac_bcast_add((mac_client_impl_t *)mch, addr, vid,
2502 		    MAC_ADDRTYPE_MULTICAST);
2503 		if (err != 0)
2504 			break;
2505 		prev_fe = flent;
2506 		flent = flent->fe_client_next;
2507 	}
2508 
2509 	/*
2510 	 * If we failed adding, then undo all, rather than partial
2511 	 * success.
2512 	 */
2513 	if (flent != NULL && prev_fe != NULL) {
2514 		flent = mcip->mci_flent_list;
2515 		while (flent != prev_fe->fe_client_next) {
2516 			vid = i_mac_flow_vid(flent);
2517 			mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
2518 			flent = flent->fe_client_next;
2519 		}
2520 	}
2521 	i_mac_perim_exit(mip);
2522 	return (err);
2523 }
2524 
2525 /*
2526  * Multicast delete function invoked by MAC clients.
2527  */
2528 void
2529 mac_multicast_remove(mac_client_handle_t mch, const uint8_t *addr)
2530 {
2531 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2532 	mac_impl_t		*mip = mcip->mci_mip;
2533 	flow_entry_t		*flent;
2534 	uint16_t		vid;
2535 
2536 	i_mac_perim_enter(mip);
2537 	for (flent = mcip->mci_flent_list; flent != NULL;
2538 	    flent = flent->fe_client_next) {
2539 		vid = i_mac_flow_vid(flent);
2540 		mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
2541 	}
2542 	i_mac_perim_exit(mip);
2543 }
2544 
2545 /*
2546  * When a MAC client desires to capture packets on an interface,
2547  * it registers a promiscuous call back with mac_promisc_add().
2548  * There are three types of promiscuous callbacks:
2549  *
2550  * * MAC_CLIENT_PROMISC_ALL
2551  *   Captures all packets sent and received by the MAC client,
2552  *   the physical interface, as well as all other MAC clients
2553  *   defined on top of the same MAC.
2554  *
2555  * * MAC_CLIENT_PROMISC_FILTERED
2556  *   Captures all packets sent and received by the MAC client,
2557  *   plus all multicast traffic sent and received by the phyisical
2558  *   interface and the other MAC clients.
2559  *
2560  * * MAC_CLIENT_PROMISC_MULTI
2561  *   Captures all broadcast and multicast packets sent and
2562  *   received by the MAC clients as well as the physical interface.
2563  *
2564  * In all cases, the underlying MAC is put in promiscuous mode.
2565  */
2566 int
2567 mac_promisc_add(mac_client_handle_t mch, mac_client_promisc_type_t type,
2568     mac_rx_t fn, void *arg, mac_promisc_handle_t *mphp, uint16_t flags)
2569 {
2570 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2571 	mac_impl_t *mip = mcip->mci_mip;
2572 	mac_promisc_impl_t *mpip;
2573 	mac_cb_info_t	*mcbi;
2574 	int rc;
2575 
2576 	i_mac_perim_enter(mip);
2577 
2578 	if ((rc = mac_start((mac_handle_t)mip)) != 0) {
2579 		i_mac_perim_exit(mip);
2580 		return (rc);
2581 	}
2582 
2583 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
2584 	    type == MAC_CLIENT_PROMISC_ALL) {
2585 		/*
2586 		 * The function is being invoked by the upper MAC client
2587 		 * of a VNIC. The VNIC should only see the traffic
2588 		 * it is entitled to.
2589 		 */
2590 		type = MAC_CLIENT_PROMISC_FILTERED;
2591 	}
2592 
2593 
2594 	/*
2595 	 * Turn on promiscuous mode for the underlying NIC.
2596 	 * This is needed even for filtered callbacks which
2597 	 * expect to receive all multicast traffic on the wire.
2598 	 *
2599 	 * Physical promiscuous mode should not be turned on if
2600 	 * MAC_PROMISC_FLAGS_NO_PHYS is set.
2601 	 */
2602 	if ((flags & MAC_PROMISC_FLAGS_NO_PHYS) == 0) {
2603 		if ((rc = i_mac_promisc_set(mip, B_TRUE)) != 0) {
2604 			mac_stop((mac_handle_t)mip);
2605 			i_mac_perim_exit(mip);
2606 			return (rc);
2607 		}
2608 	}
2609 
2610 	mpip = kmem_cache_alloc(mac_promisc_impl_cache, KM_SLEEP);
2611 
2612 	mpip->mpi_type = type;
2613 	mpip->mpi_fn = fn;
2614 	mpip->mpi_arg = arg;
2615 	mpip->mpi_mcip = mcip;
2616 	mpip->mpi_no_tx_loop = ((flags & MAC_PROMISC_FLAGS_NO_TX_LOOP) != 0);
2617 	mpip->mpi_no_phys = ((flags & MAC_PROMISC_FLAGS_NO_PHYS) != 0);
2618 	mpip->mpi_strip_vlan_tag =
2619 	    ((flags & MAC_PROMISC_FLAGS_VLAN_TAG_STRIP) != 0);
2620 
2621 	mcbi = &mip->mi_promisc_cb_info;
2622 	mutex_enter(mcbi->mcbi_lockp);
2623 
2624 	mac_callback_add(&mip->mi_promisc_cb_info, &mcip->mci_promisc_list,
2625 	    &mpip->mpi_mci_link);
2626 	mac_callback_add(&mip->mi_promisc_cb_info, &mip->mi_promisc_list,
2627 	    &mpip->mpi_mi_link);
2628 
2629 	mutex_exit(mcbi->mcbi_lockp);
2630 
2631 	*mphp = (mac_promisc_handle_t)mpip;
2632 	i_mac_perim_exit(mip);
2633 	return (0);
2634 }
2635 
2636 /*
2637  * Remove a multicast address previously aded through mac_promisc_add().
2638  */
2639 void
2640 mac_promisc_remove(mac_promisc_handle_t mph)
2641 {
2642 	mac_promisc_impl_t *mpip = (mac_promisc_impl_t *)mph;
2643 	mac_client_impl_t *mcip = mpip->mpi_mcip;
2644 	mac_impl_t *mip = mcip->mci_mip;
2645 	mac_cb_info_t *mcbi;
2646 	int rv;
2647 
2648 	i_mac_perim_enter(mip);
2649 
2650 	/*
2651 	 * Even if the device can't be reset into normal mode, we still
2652 	 * need to clear the client promisc callbacks. The client may want
2653 	 * to close the mac end point and we can't have stale callbacks.
2654 	 */
2655 	if (!(mpip->mpi_no_phys)) {
2656 		if ((rv = i_mac_promisc_set(mip, B_FALSE)) != 0) {
2657 			cmn_err(CE_WARN, "%s: failed to switch OFF promiscuous"
2658 			    " mode because of error 0x%x", mip->mi_name, rv);
2659 		}
2660 	}
2661 	mcbi = &mip->mi_promisc_cb_info;
2662 	mutex_enter(mcbi->mcbi_lockp);
2663 	if (mac_callback_remove(mcbi, &mip->mi_promisc_list,
2664 	    &mpip->mpi_mi_link)) {
2665 		VERIFY(mac_callback_remove(&mip->mi_promisc_cb_info,
2666 		    &mcip->mci_promisc_list, &mpip->mpi_mci_link));
2667 		kmem_cache_free(mac_promisc_impl_cache, mpip);
2668 	} else {
2669 		mac_callback_remove_wait(&mip->mi_promisc_cb_info);
2670 	}
2671 	mutex_exit(mcbi->mcbi_lockp);
2672 	mac_stop((mac_handle_t)mip);
2673 
2674 	i_mac_perim_exit(mip);
2675 }
2676 
2677 /*
2678  * Reference count the number of active Tx threads. MCI_TX_QUIESCE indicates
2679  * that a control operation wants to quiesce the Tx data flow in which case
2680  * we return an error. Holding any of the per cpu locks ensures that the
2681  * mci_tx_flag won't change.
2682  *
2683  * 'CPU' must be accessed just once and used to compute the index into the
2684  * percpu array, and that index must be used for the entire duration of the
2685  * packet send operation. Note that the thread may be preempted and run on
2686  * another cpu any time and so we can't use 'CPU' more than once for the
2687  * operation.
2688  */
2689 #define	MAC_TX_TRY_HOLD(mcip, mytx, error)				\
2690 {									\
2691 	(error) = 0;							\
2692 	(mytx) = &(mcip)->mci_tx_pcpu[CPU->cpu_seqid & mac_tx_percpu_cnt]; \
2693 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
2694 	if (!((mcip)->mci_tx_flag & MCI_TX_QUIESCE)) {			\
2695 		(mytx)->pcpu_tx_refcnt++;				\
2696 	} else {							\
2697 		(error) = -1;						\
2698 	}								\
2699 	mutex_exit(&(mytx)->pcpu_tx_lock);				\
2700 }
2701 
2702 /*
2703  * Release the reference. If needed, signal any control operation waiting
2704  * for Tx quiescence. The wait and signal are always done using the
2705  * mci_tx_pcpu[0]'s lock
2706  */
2707 #define	MAC_TX_RELE(mcip, mytx) {					\
2708 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
2709 	if (--(mytx)->pcpu_tx_refcnt == 0 &&				\
2710 	    (mcip)->mci_tx_flag & MCI_TX_QUIESCE) {			\
2711 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
2712 		mutex_enter(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
2713 		cv_signal(&(mcip)->mci_tx_cv);				\
2714 		mutex_exit(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
2715 	} else {							\
2716 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
2717 	}								\
2718 }
2719 
2720 /*
2721  * Bump the count of the number of active Tx threads. This is maintained as
2722  * a per CPU counter. On (CMT kind of) machines with large number of CPUs,
2723  * a single mci_tx_lock may become contended. However a count of the total
2724  * number of Tx threads per client is needed in order to quiesce the Tx side
2725  * prior to reassigning a Tx ring dynamically to another client. The thread
2726  * that needs to quiesce the Tx traffic grabs all the percpu locks and checks
2727  * the sum of the individual percpu refcnts. Each Tx data thread only grabs
2728  * its own percpu lock and increments its own refcnt.
2729  */
2730 void *
2731 mac_tx_hold(mac_client_handle_t mch)
2732 {
2733 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2734 	mac_tx_percpu_t	*mytx;
2735 	int error;
2736 
2737 	MAC_TX_TRY_HOLD(mcip, mytx, error);
2738 	return (error == 0 ? (void *)mytx : NULL);
2739 }
2740 
2741 void
2742 mac_tx_rele(mac_client_handle_t mch, void *mytx_handle)
2743 {
2744 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2745 	mac_tx_percpu_t	*mytx = mytx_handle;
2746 
2747 	MAC_TX_RELE(mcip, mytx)
2748 }
2749 
2750 /*
2751  * Send function invoked by MAC clients.
2752  */
2753 mac_tx_cookie_t
2754 mac_tx(mac_client_handle_t mch, mblk_t *mp_chain, uintptr_t hint,
2755     uint16_t flag, mblk_t **ret_mp)
2756 {
2757 	mac_tx_cookie_t		cookie;
2758 	int			error;
2759 	mac_tx_percpu_t		*mytx;
2760 	mac_soft_ring_set_t	*srs;
2761 	flow_entry_t		*flent;
2762 	boolean_t		is_subflow = B_FALSE;
2763 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2764 	mac_impl_t		*mip = mcip->mci_mip;
2765 	mac_srs_tx_t		*srs_tx;
2766 
2767 	/*
2768 	 * Check whether the active Tx threads count is bumped already.
2769 	 */
2770 	if (!(flag & MAC_TX_NO_HOLD)) {
2771 		MAC_TX_TRY_HOLD(mcip, mytx, error);
2772 		if (error != 0) {
2773 			freemsgchain(mp_chain);
2774 			return (NULL);
2775 		}
2776 	}
2777 
2778 	if (mcip->mci_subflow_tab != NULL &&
2779 	    mcip->mci_subflow_tab->ft_flow_count > 0 &&
2780 	    mac_flow_lookup(mcip->mci_subflow_tab, mp_chain,
2781 	    FLOW_OUTBOUND, &flent) == 0) {
2782 		/*
2783 		 * The main assumption here is that if in the event
2784 		 * we get a chain, all the packets will be classified
2785 		 * to the same Flow/SRS. If this changes for any
2786 		 * reason, the following logic should change as well.
2787 		 * I suppose the fanout_hint also assumes this .
2788 		 */
2789 		ASSERT(flent != NULL);
2790 		is_subflow = B_TRUE;
2791 	} else {
2792 		flent = mcip->mci_flent;
2793 	}
2794 
2795 	srs = flent->fe_tx_srs;
2796 	srs_tx = &srs->srs_tx;
2797 	if (srs_tx->st_mode == SRS_TX_DEFAULT &&
2798 	    (srs->srs_state & SRS_ENQUEUED) == 0 &&
2799 	    mip->mi_nactiveclients == 1 && mip->mi_promisc_list == NULL &&
2800 	    mp_chain->b_next == NULL) {
2801 		uint64_t	obytes;
2802 
2803 		/*
2804 		 * Since dls always opens the underlying MAC, nclients equals
2805 		 * to 1 means that the only active client is dls itself acting
2806 		 * as a primary client of the MAC instance. Since dls will not
2807 		 * send tagged packets in that case, and dls is trusted to send
2808 		 * packets for its allowed VLAN(s), the VLAN tag insertion and
2809 		 * check is required only if nclients is greater than 1.
2810 		 */
2811 		if (mip->mi_nclients > 1) {
2812 			if (MAC_VID_CHECK_NEEDED(mcip)) {
2813 				int	err = 0;
2814 
2815 				MAC_VID_CHECK(mcip, mp_chain, err);
2816 				if (err != 0) {
2817 					freemsg(mp_chain);
2818 					mcip->mci_stat_oerrors++;
2819 					goto done;
2820 				}
2821 			}
2822 			if (MAC_TAG_NEEDED(mcip)) {
2823 				mp_chain = mac_add_vlan_tag(mp_chain, 0,
2824 				    mac_client_vid(mch));
2825 				if (mp_chain == NULL) {
2826 					mcip->mci_stat_oerrors++;
2827 					goto done;
2828 				}
2829 			}
2830 		}
2831 
2832 		obytes = (mp_chain->b_cont == NULL ? MBLKL(mp_chain) :
2833 		    msgdsize(mp_chain));
2834 
2835 		MAC_TX(mip, srs_tx->st_arg2, mp_chain, mcip);
2836 
2837 		if (mp_chain == NULL) {
2838 			cookie = NULL;
2839 			mcip->mci_stat_obytes += obytes;
2840 			mcip->mci_stat_opackets += 1;
2841 			if ((srs->srs_type & SRST_FLOW) != 0) {
2842 				FLOW_STAT_UPDATE(flent, obytes, obytes);
2843 				FLOW_STAT_UPDATE(flent, opackets, 1);
2844 			}
2845 		} else {
2846 			mutex_enter(&srs->srs_lock);
2847 			cookie = mac_tx_srs_no_desc(srs, mp_chain,
2848 			    flag, ret_mp);
2849 			mutex_exit(&srs->srs_lock);
2850 		}
2851 	} else {
2852 		cookie = srs_tx->st_func(srs, mp_chain, hint, flag, ret_mp);
2853 	}
2854 
2855 done:
2856 	if (is_subflow)
2857 		FLOW_REFRELE(flent);
2858 
2859 	if (!(flag & MAC_TX_NO_HOLD))
2860 		MAC_TX_RELE(mcip, mytx);
2861 
2862 	return (cookie);
2863 }
2864 
2865 /*
2866  * mac_tx_is_blocked
2867  *
2868  * Given a cookie, it returns if the ring identified by the cookie is
2869  * flow-controlled or not. If NULL is passed in place of a cookie,
2870  * then it finds out if any of the underlying rings belonging to the
2871  * SRS is flow controlled or not and returns that status.
2872  */
2873 /* ARGSUSED */
2874 boolean_t
2875 mac_tx_is_flow_blocked(mac_client_handle_t mch, mac_tx_cookie_t cookie)
2876 {
2877 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2878 	mac_soft_ring_set_t *mac_srs;
2879 	mac_soft_ring_t *sringp;
2880 	boolean_t blocked = B_FALSE;
2881 	mac_tx_percpu_t *mytx;
2882 	int err;
2883 	int i;
2884 
2885 	/*
2886 	 * Bump the reference count so that mac_srs won't be deleted.
2887 	 * If the client is currently quiesced and we failed to bump
2888 	 * the reference, return B_TRUE so that flow control stays
2889 	 * as enabled.
2890 	 *
2891 	 * Flow control will then be disabled once the client is no
2892 	 * longer quiesced.
2893 	 */
2894 	MAC_TX_TRY_HOLD(mcip, mytx, err);
2895 	if (err != 0)
2896 		return (B_TRUE);
2897 
2898 	if ((mac_srs = MCIP_TX_SRS(mcip)) == NULL) {
2899 		MAC_TX_RELE(mcip, mytx);
2900 		return (B_FALSE);
2901 	}
2902 
2903 	mutex_enter(&mac_srs->srs_lock);
2904 	if (mac_srs->srs_tx.st_mode == SRS_TX_FANOUT) {
2905 		if (cookie != NULL) {
2906 			sringp = (mac_soft_ring_t *)cookie;
2907 			mutex_enter(&sringp->s_ring_lock);
2908 			if (sringp->s_ring_state & S_RING_TX_HIWAT)
2909 				blocked = B_TRUE;
2910 			mutex_exit(&sringp->s_ring_lock);
2911 		} else {
2912 			for (i = 0; i < mac_srs->srs_oth_ring_count; i++) {
2913 				sringp = mac_srs->srs_oth_soft_rings[i];
2914 				mutex_enter(&sringp->s_ring_lock);
2915 				if (sringp->s_ring_state & S_RING_TX_HIWAT) {
2916 					blocked = B_TRUE;
2917 					mutex_exit(&sringp->s_ring_lock);
2918 					break;
2919 				}
2920 				mutex_exit(&sringp->s_ring_lock);
2921 			}
2922 		}
2923 	} else {
2924 		blocked = (mac_srs->srs_state & SRS_TX_HIWAT);
2925 	}
2926 	mutex_exit(&mac_srs->srs_lock);
2927 	MAC_TX_RELE(mcip, mytx);
2928 	return (blocked);
2929 }
2930 
2931 /*
2932  * Check if the MAC client is the primary MAC client.
2933  */
2934 boolean_t
2935 mac_is_primary_client(mac_client_impl_t *mcip)
2936 {
2937 	return (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY);
2938 }
2939 
2940 void
2941 mac_ioctl(mac_handle_t mh, queue_t *wq, mblk_t *bp)
2942 {
2943 	mac_impl_t	*mip = (mac_impl_t *)mh;
2944 	int cmd = ((struct iocblk *)bp->b_rptr)->ioc_cmd;
2945 
2946 	if ((cmd == ND_GET && (mip->mi_callbacks->mc_callbacks & MC_GETPROP)) ||
2947 	    (cmd == ND_SET && (mip->mi_callbacks->mc_callbacks & MC_SETPROP))) {
2948 		/*
2949 		 * If ndd props were registered, call them.
2950 		 * Note that ndd ioctls are Obsolete
2951 		 */
2952 		mac_ndd_ioctl(mip, wq, bp);
2953 		return;
2954 	}
2955 
2956 	/*
2957 	 * Call the driver to handle the ioctl.  The driver may not support
2958 	 * any ioctls, in which case we reply with a NAK on its behalf.
2959 	 */
2960 	if (mip->mi_callbacks->mc_callbacks & MC_IOCTL)
2961 		mip->mi_ioctl(mip->mi_driver, wq, bp);
2962 	else
2963 		miocnak(wq, bp, 0, EINVAL);
2964 }
2965 
2966 /*
2967  * Return the link state of the specified MAC instance.
2968  */
2969 link_state_t
2970 mac_link_get(mac_handle_t mh)
2971 {
2972 	return (((mac_impl_t *)mh)->mi_linkstate);
2973 }
2974 
2975 /*
2976  * Add a mac client specified notification callback. Please see the comments
2977  * above mac_callback_add() for general information about mac callback
2978  * addition/deletion in the presence of mac callback list walkers
2979  */
2980 mac_notify_handle_t
2981 mac_notify_add(mac_handle_t mh, mac_notify_t notify_fn, void *arg)
2982 {
2983 	mac_impl_t		*mip = (mac_impl_t *)mh;
2984 	mac_notify_cb_t		*mncb;
2985 	mac_cb_info_t		*mcbi;
2986 
2987 	/*
2988 	 * Allocate a notify callback structure, fill in the details and
2989 	 * use the mac callback list manipulation functions to chain into
2990 	 * the list of callbacks.
2991 	 */
2992 	mncb = kmem_zalloc(sizeof (mac_notify_cb_t), KM_SLEEP);
2993 	mncb->mncb_fn = notify_fn;
2994 	mncb->mncb_arg = arg;
2995 	mncb->mncb_mip = mip;
2996 	mncb->mncb_link.mcb_objp = mncb;
2997 	mncb->mncb_link.mcb_objsize = sizeof (mac_notify_cb_t);
2998 	mncb->mncb_link.mcb_flags = MCB_NOTIFY_CB_T;
2999 
3000 	mcbi = &mip->mi_notify_cb_info;
3001 
3002 	i_mac_perim_enter(mip);
3003 	mutex_enter(mcbi->mcbi_lockp);
3004 
3005 	mac_callback_add(&mip->mi_notify_cb_info, &mip->mi_notify_cb_list,
3006 	    &mncb->mncb_link);
3007 
3008 	mutex_exit(mcbi->mcbi_lockp);
3009 	i_mac_perim_exit(mip);
3010 	return ((mac_notify_handle_t)mncb);
3011 }
3012 
3013 void
3014 mac_notify_remove_wait(mac_handle_t mh)
3015 {
3016 	mac_impl_t	*mip = (mac_impl_t *)mh;
3017 	mac_cb_info_t	*mcbi = &mip->mi_notify_cb_info;
3018 
3019 	mutex_enter(mcbi->mcbi_lockp);
3020 	mac_callback_remove_wait(&mip->mi_notify_cb_info);
3021 	mutex_exit(mcbi->mcbi_lockp);
3022 }
3023 
3024 /*
3025  * Remove a mac client specified notification callback
3026  */
3027 int
3028 mac_notify_remove(mac_notify_handle_t mnh, boolean_t wait)
3029 {
3030 	mac_notify_cb_t	*mncb = (mac_notify_cb_t *)mnh;
3031 	mac_impl_t	*mip = mncb->mncb_mip;
3032 	mac_cb_info_t	*mcbi;
3033 	int		err = 0;
3034 
3035 	mcbi = &mip->mi_notify_cb_info;
3036 
3037 	i_mac_perim_enter(mip);
3038 	mutex_enter(mcbi->mcbi_lockp);
3039 
3040 	ASSERT(mncb->mncb_link.mcb_objp == mncb);
3041 	/*
3042 	 * If there aren't any list walkers, the remove would succeed
3043 	 * inline, else we wait for the deferred remove to complete
3044 	 */
3045 	if (mac_callback_remove(&mip->mi_notify_cb_info,
3046 	    &mip->mi_notify_cb_list, &mncb->mncb_link)) {
3047 		kmem_free(mncb, sizeof (mac_notify_cb_t));
3048 	} else {
3049 		err = EBUSY;
3050 	}
3051 
3052 	mutex_exit(mcbi->mcbi_lockp);
3053 	i_mac_perim_exit(mip);
3054 
3055 	/*
3056 	 * If we failed to remove the notification callback and "wait" is set
3057 	 * to be B_TRUE, wait for the callback to finish after we exit the
3058 	 * mac perimeter.
3059 	 */
3060 	if (err != 0 && wait) {
3061 		mac_notify_remove_wait((mac_handle_t)mip);
3062 		return (0);
3063 	}
3064 
3065 	return (err);
3066 }
3067 
3068 /*
3069  * Associate resource management callbacks with the specified MAC
3070  * clients.
3071  */
3072 
3073 void
3074 mac_resource_set_common(mac_client_handle_t mch, mac_resource_add_t add,
3075     mac_resource_remove_t remove, mac_resource_quiesce_t quiesce,
3076     mac_resource_restart_t restart, mac_resource_bind_t bind,
3077     void *arg)
3078 {
3079 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3080 
3081 	mcip->mci_resource_add = add;
3082 	mcip->mci_resource_remove = remove;
3083 	mcip->mci_resource_quiesce = quiesce;
3084 	mcip->mci_resource_restart = restart;
3085 	mcip->mci_resource_bind = bind;
3086 	mcip->mci_resource_arg = arg;
3087 
3088 	if (arg == NULL)
3089 		mcip->mci_state_flags &= ~MCIS_CLIENT_POLL_CAPABLE;
3090 }
3091 
3092 void
3093 mac_resource_set(mac_client_handle_t mch, mac_resource_add_t add, void *arg)
3094 {
3095 	/* update the 'resource_add' callback */
3096 	mac_resource_set_common(mch, add, NULL, NULL, NULL, NULL, arg);
3097 }
3098 
3099 /*
3100  * Sets up the client resources and enable the polling interface over all the
3101  * SRS's and the soft rings of the client
3102  */
3103 void
3104 mac_client_poll_enable(mac_client_handle_t mch)
3105 {
3106 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3107 	mac_soft_ring_set_t	*mac_srs;
3108 	flow_entry_t		*flent;
3109 	int			i;
3110 
3111 	flent = mcip->mci_flent;
3112 	ASSERT(flent != NULL);
3113 
3114 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3115 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3116 		ASSERT(mac_srs->srs_mcip == mcip);
3117 		mac_srs_client_poll_enable(mcip, mac_srs);
3118 	}
3119 }
3120 
3121 /*
3122  * Tears down the client resources and disable the polling interface over all
3123  * the SRS's and the soft rings of the client
3124  */
3125 void
3126 mac_client_poll_disable(mac_client_handle_t mch)
3127 {
3128 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3129 	mac_soft_ring_set_t	*mac_srs;
3130 	flow_entry_t		*flent;
3131 	int			i;
3132 
3133 	flent = mcip->mci_flent;
3134 	ASSERT(flent != NULL);
3135 
3136 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3137 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3138 		ASSERT(mac_srs->srs_mcip == mcip);
3139 		mac_srs_client_poll_disable(mcip, mac_srs);
3140 	}
3141 }
3142 
3143 /*
3144  * Associate the CPUs specified by the given property with a MAC client.
3145  */
3146 int
3147 mac_cpu_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
3148 {
3149 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3150 	mac_impl_t *mip = mcip->mci_mip;
3151 	int err = 0;
3152 
3153 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3154 
3155 	if ((err = mac_validate_props(mrp)) != 0)
3156 		return (err);
3157 
3158 	if (MCIP_DATAPATH_SETUP(mcip))
3159 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
3160 
3161 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
3162 	return (0);
3163 }
3164 
3165 /*
3166  * Apply the specified properties to the specified MAC client.
3167  */
3168 int
3169 mac_client_set_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3170 {
3171 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3172 	mac_impl_t *mip = mcip->mci_mip;
3173 	int err = 0;
3174 
3175 	i_mac_perim_enter(mip);
3176 
3177 	if ((mrp->mrp_mask & MRP_MAXBW) || (mrp->mrp_mask & MRP_PRIORITY)) {
3178 		err = mac_resource_ctl_set(mch, mrp);
3179 		if (err != 0) {
3180 			i_mac_perim_exit(mip);
3181 			return (err);
3182 		}
3183 	}
3184 
3185 	if (mrp->mrp_mask & MRP_CPUS)
3186 		err = mac_cpu_set(mch, mrp);
3187 
3188 	i_mac_perim_exit(mip);
3189 	return (err);
3190 }
3191 
3192 /*
3193  * Return the properties currently associated with the specified MAC client.
3194  */
3195 void
3196 mac_client_get_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3197 {
3198 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3199 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
3200 
3201 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3202 }
3203 
3204 /*
3205  * Pass a copy of the specified packet to the promiscuous callbacks
3206  * of the specified MAC.
3207  *
3208  * If sender is NULL, the function is being invoked for a packet chain
3209  * received from the wire. If sender is non-NULL, it points to
3210  * the MAC client from which the packet is being sent.
3211  *
3212  * The packets are distributed to the promiscuous callbacks as follows:
3213  *
3214  * - all packets are sent to the MAC_CLIENT_PROMISC_ALL callbacks
3215  * - all broadcast and multicast packets are sent to the
3216  *   MAC_CLIENT_PROMISC_FILTER and MAC_CLIENT_PROMISC_MULTI.
3217  *
3218  * The unicast packets of MAC_CLIENT_PROMISC_FILTER callbacks are dispatched
3219  * after classification by mac_rx_deliver().
3220  */
3221 
3222 static void
3223 mac_promisc_dispatch_one(mac_promisc_impl_t *mpip, mblk_t *mp,
3224     boolean_t loopback)
3225 {
3226 	mblk_t *mp_copy;
3227 
3228 	mp_copy = copymsg(mp);
3229 	if (mp_copy == NULL)
3230 		return;
3231 	mp_copy->b_next = NULL;
3232 
3233 	if (mpip->mpi_strip_vlan_tag) {
3234 		if ((mp_copy = mac_strip_vlan_tag_chain(mp_copy)) == NULL)
3235 			return;
3236 	}
3237 	mpip->mpi_fn(mpip->mpi_arg, NULL, mp_copy, loopback);
3238 }
3239 
3240 /*
3241  * Return the VID of a packet. Zero if the packet is not tagged.
3242  */
3243 static uint16_t
3244 mac_ether_vid(mblk_t *mp)
3245 {
3246 	struct ether_header *eth = (struct ether_header *)mp->b_rptr;
3247 
3248 	if (ntohs(eth->ether_type) == ETHERTYPE_VLAN) {
3249 		struct ether_vlan_header *t_evhp =
3250 		    (struct ether_vlan_header *)mp->b_rptr;
3251 		return (VLAN_ID(ntohs(t_evhp->ether_tci)));
3252 	}
3253 
3254 	return (0);
3255 }
3256 
3257 /*
3258  * Return whether the specified packet contains a multicast or broadcast
3259  * destination MAC address.
3260  */
3261 static boolean_t
3262 mac_is_mcast(mac_impl_t *mip, mblk_t *mp)
3263 {
3264 	mac_header_info_t hdr_info;
3265 
3266 	if (mac_header_info((mac_handle_t)mip, mp, &hdr_info) != 0)
3267 		return (B_FALSE);
3268 	return ((hdr_info.mhi_dsttype == MAC_ADDRTYPE_BROADCAST) ||
3269 	    (hdr_info.mhi_dsttype == MAC_ADDRTYPE_MULTICAST));
3270 }
3271 
3272 /*
3273  * Send a copy of an mblk chain to the MAC clients of the specified MAC.
3274  * "sender" points to the sender MAC client for outbound packets, and
3275  * is set to NULL for inbound packets.
3276  */
3277 void
3278 mac_promisc_dispatch(mac_impl_t *mip, mblk_t *mp_chain,
3279     mac_client_impl_t *sender)
3280 {
3281 	mac_promisc_impl_t *mpip;
3282 	mac_cb_t *mcb;
3283 	mblk_t *mp;
3284 	boolean_t is_mcast, is_sender;
3285 
3286 	MAC_PROMISC_WALKER_INC(mip);
3287 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3288 		is_mcast = mac_is_mcast(mip, mp);
3289 		/* send packet to interested callbacks */
3290 		for (mcb = mip->mi_promisc_list; mcb != NULL;
3291 		    mcb = mcb->mcb_nextp) {
3292 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3293 			is_sender = (mpip->mpi_mcip == sender);
3294 
3295 			if (is_sender && mpip->mpi_no_tx_loop)
3296 				/*
3297 				 * The sender doesn't want to receive
3298 				 * copies of the packets it sends.
3299 				 */
3300 				continue;
3301 
3302 			/*
3303 			 * For an ethernet MAC, don't displatch a multicast
3304 			 * packet to a non-PROMISC_ALL callbacks unless the VID
3305 			 * of the packet matches the VID of the client.
3306 			 */
3307 			if (is_mcast &&
3308 			    mpip->mpi_type != MAC_CLIENT_PROMISC_ALL &&
3309 			    !mac_client_check_flow_vid(mpip->mpi_mcip,
3310 			    mac_ether_vid(mp)))
3311 				continue;
3312 
3313 			if (is_sender ||
3314 			    mpip->mpi_type == MAC_CLIENT_PROMISC_ALL ||
3315 			    is_mcast)
3316 				mac_promisc_dispatch_one(mpip, mp, is_sender);
3317 		}
3318 	}
3319 	MAC_PROMISC_WALKER_DCR(mip);
3320 }
3321 
3322 void
3323 mac_promisc_client_dispatch(mac_client_impl_t *mcip, mblk_t *mp_chain)
3324 {
3325 	mac_impl_t		*mip = mcip->mci_mip;
3326 	mac_promisc_impl_t	*mpip;
3327 	boolean_t		is_mcast;
3328 	mblk_t			*mp;
3329 	mac_cb_t		*mcb;
3330 
3331 	/*
3332 	 * The unicast packets for the MAC client still
3333 	 * need to be delivered to the MAC_CLIENT_PROMISC_FILTERED
3334 	 * promiscuous callbacks. The broadcast and multicast
3335 	 * packets were delivered from mac_rx().
3336 	 */
3337 	MAC_PROMISC_WALKER_INC(mip);
3338 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3339 		is_mcast = mac_is_mcast(mip, mp);
3340 		for (mcb = mcip->mci_promisc_list; mcb != NULL;
3341 		    mcb = mcb->mcb_nextp) {
3342 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3343 			if (mpip->mpi_type == MAC_CLIENT_PROMISC_FILTERED &&
3344 			    !is_mcast) {
3345 				mac_promisc_dispatch_one(mpip, mp, B_FALSE);
3346 			}
3347 		}
3348 	}
3349 	MAC_PROMISC_WALKER_DCR(mip);
3350 }
3351 
3352 /*
3353  * Return the margin value currently assigned to the specified MAC instance.
3354  */
3355 void
3356 mac_margin_get(mac_handle_t mh, uint32_t *marginp)
3357 {
3358 	mac_impl_t *mip = (mac_impl_t *)mh;
3359 
3360 	rw_enter(&(mip->mi_rw_lock), RW_READER);
3361 	*marginp = mip->mi_margin;
3362 	rw_exit(&(mip->mi_rw_lock));
3363 }
3364 
3365 /*
3366  * mac_info_get() is used for retrieving the mac_info when a DL_INFO_REQ is
3367  * issued before a DL_ATTACH_REQ. we walk the i_mac_impl_hash table and find
3368  * the first mac_impl_t with a matching driver name; then we copy its mac_info_t
3369  * to the caller. we do all this with i_mac_impl_lock held so the mac_impl_t
3370  * cannot disappear while we are accessing it.
3371  */
3372 typedef struct i_mac_info_state_s {
3373 	const char	*mi_name;
3374 	mac_info_t	*mi_infop;
3375 } i_mac_info_state_t;
3376 
3377 /*ARGSUSED*/
3378 static uint_t
3379 i_mac_info_walker(mod_hash_key_t key, mod_hash_val_t *val, void *arg)
3380 {
3381 	i_mac_info_state_t *statep = arg;
3382 	mac_impl_t *mip = (mac_impl_t *)val;
3383 
3384 	if (mip->mi_state_flags & MIS_DISABLED)
3385 		return (MH_WALK_CONTINUE);
3386 
3387 	if (strcmp(statep->mi_name,
3388 	    ddi_driver_name(mip->mi_dip)) != 0)
3389 		return (MH_WALK_CONTINUE);
3390 
3391 	statep->mi_infop = &mip->mi_info;
3392 	return (MH_WALK_TERMINATE);
3393 }
3394 
3395 boolean_t
3396 mac_info_get(const char *name, mac_info_t *minfop)
3397 {
3398 	i_mac_info_state_t state;
3399 
3400 	rw_enter(&i_mac_impl_lock, RW_READER);
3401 	state.mi_name = name;
3402 	state.mi_infop = NULL;
3403 	mod_hash_walk(i_mac_impl_hash, i_mac_info_walker, &state);
3404 	if (state.mi_infop == NULL) {
3405 		rw_exit(&i_mac_impl_lock);
3406 		return (B_FALSE);
3407 	}
3408 	*minfop = *state.mi_infop;
3409 	rw_exit(&i_mac_impl_lock);
3410 	return (B_TRUE);
3411 }
3412 
3413 /*
3414  * To get the capabilities that MAC layer cares about, such as rings, factory
3415  * mac address, vnic or not, it should directly invoke this function
3416  */
3417 boolean_t
3418 i_mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
3419 {
3420 	mac_impl_t *mip = (mac_impl_t *)mh;
3421 
3422 	if (mip->mi_callbacks->mc_callbacks & MC_GETCAPAB)
3423 		return (mip->mi_getcapab(mip->mi_driver, cap, cap_data));
3424 	else
3425 		return (B_FALSE);
3426 }
3427 
3428 /*
3429  * Capability query function. If number of active mac clients is greater than
3430  * 1, only limited capabilities can be advertised to the caller no matter the
3431  * driver has certain capability or not. Else, we query the driver to get the
3432  * capability.
3433  */
3434 boolean_t
3435 mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
3436 {
3437 	mac_impl_t *mip = (mac_impl_t *)mh;
3438 
3439 	/*
3440 	 * if mi_nactiveclients > 1, only MAC_CAPAB_LEGACY, MAC_CAPAB_HCKSUM,
3441 	 * MAC_CAPAB_NO_NATIVEVLAN and MAC_CAPAB_NO_ZCOPY can be advertised.
3442 	 */
3443 	if (mip->mi_nactiveclients > 1) {
3444 		switch (cap) {
3445 		case MAC_CAPAB_NO_NATIVEVLAN:
3446 		case MAC_CAPAB_NO_ZCOPY:
3447 			return (B_TRUE);
3448 		case MAC_CAPAB_LEGACY:
3449 		case MAC_CAPAB_HCKSUM:
3450 			break;
3451 		default:
3452 			return (B_FALSE);
3453 		}
3454 	}
3455 
3456 	/* else get capab from driver */
3457 	return (i_mac_capab_get(mh, cap, cap_data));
3458 }
3459 
3460 boolean_t
3461 mac_sap_verify(mac_handle_t mh, uint32_t sap, uint32_t *bind_sap)
3462 {
3463 	mac_impl_t *mip = (mac_impl_t *)mh;
3464 
3465 	return (mip->mi_type->mt_ops.mtops_sap_verify(sap, bind_sap,
3466 	    mip->mi_pdata));
3467 }
3468 
3469 mblk_t *
3470 mac_header(mac_handle_t mh, const uint8_t *daddr, uint32_t sap, mblk_t *payload,
3471     size_t extra_len)
3472 {
3473 	mac_impl_t *mip = (mac_impl_t *)mh;
3474 
3475 	return (mip->mi_type->mt_ops.mtops_header(mip->mi_addr, daddr, sap,
3476 	    mip->mi_pdata, payload, extra_len));
3477 }
3478 
3479 int
3480 mac_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
3481 {
3482 	mac_impl_t *mip = (mac_impl_t *)mh;
3483 
3484 	return (mip->mi_type->mt_ops.mtops_header_info(mp, mip->mi_pdata,
3485 	    mhip));
3486 }
3487 
3488 mblk_t *
3489 mac_header_cook(mac_handle_t mh, mblk_t *mp)
3490 {
3491 	mac_impl_t *mip = (mac_impl_t *)mh;
3492 
3493 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_COOK) {
3494 		if (DB_REF(mp) > 1) {
3495 			mblk_t *newmp = copymsg(mp);
3496 			if (newmp == NULL)
3497 				return (NULL);
3498 			freemsg(mp);
3499 			mp = newmp;
3500 		}
3501 		return (mip->mi_type->mt_ops.mtops_header_cook(mp,
3502 		    mip->mi_pdata));
3503 	}
3504 	return (mp);
3505 }
3506 
3507 mblk_t *
3508 mac_header_uncook(mac_handle_t mh, mblk_t *mp)
3509 {
3510 	mac_impl_t *mip = (mac_impl_t *)mh;
3511 
3512 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_UNCOOK) {
3513 		if (DB_REF(mp) > 1) {
3514 			mblk_t *newmp = copymsg(mp);
3515 			if (newmp == NULL)
3516 				return (NULL);
3517 			freemsg(mp);
3518 			mp = newmp;
3519 		}
3520 		return (mip->mi_type->mt_ops.mtops_header_uncook(mp,
3521 		    mip->mi_pdata));
3522 	}
3523 	return (mp);
3524 }
3525 
3526 uint_t
3527 mac_addr_len(mac_handle_t mh)
3528 {
3529 	mac_impl_t *mip = (mac_impl_t *)mh;
3530 
3531 	return (mip->mi_type->mt_addr_length);
3532 }
3533 
3534 /* True if a MAC is a VNIC */
3535 boolean_t
3536 mac_is_vnic(mac_handle_t mh)
3537 {
3538 	return (((mac_impl_t *)mh)->mi_state_flags & MIS_IS_VNIC);
3539 }
3540 
3541 mac_handle_t
3542 mac_get_lower_mac_handle(mac_handle_t mh)
3543 {
3544 	mac_impl_t *mip = (mac_impl_t *)mh;
3545 
3546 	ASSERT(mac_is_vnic(mh));
3547 	return (((vnic_t *)mip->mi_driver)->vn_lower_mh);
3548 }
3549 
3550 void
3551 mac_update_resources(mac_resource_props_t *nmrp, mac_resource_props_t *cmrp,
3552     boolean_t is_user_flow)
3553 {
3554 	if (nmrp != NULL && cmrp != NULL) {
3555 		if (nmrp->mrp_mask & MRP_PRIORITY) {
3556 			if (nmrp->mrp_priority == MPL_RESET) {
3557 				cmrp->mrp_mask &= ~MRP_PRIORITY;
3558 				if (is_user_flow) {
3559 					cmrp->mrp_priority =
3560 					    MPL_SUBFLOW_DEFAULT;
3561 				} else {
3562 					cmrp->mrp_priority = MPL_LINK_DEFAULT;
3563 				}
3564 			} else {
3565 				cmrp->mrp_mask |= MRP_PRIORITY;
3566 				cmrp->mrp_priority = nmrp->mrp_priority;
3567 			}
3568 		}
3569 		if (nmrp->mrp_mask & MRP_MAXBW) {
3570 			cmrp->mrp_maxbw = nmrp->mrp_maxbw;
3571 			if (nmrp->mrp_maxbw == MRP_MAXBW_RESETVAL)
3572 				cmrp->mrp_mask &= ~MRP_MAXBW;
3573 			else
3574 				cmrp->mrp_mask |= MRP_MAXBW;
3575 		}
3576 		if (nmrp->mrp_mask & MRP_CPUS)
3577 			MAC_COPY_CPUS(nmrp, cmrp);
3578 	}
3579 }
3580 
3581 /*
3582  * i_mac_set_resources:
3583  *
3584  * This routine associates properties with the primary MAC client of
3585  * the specified MAC instance.
3586  * - Cache the properties in mac_impl_t
3587  * - Apply the properties to the primary MAC client if exists
3588  */
3589 int
3590 i_mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3591 {
3592 	mac_impl_t		*mip = (mac_impl_t *)mh;
3593 	mac_client_impl_t	*mcip;
3594 	int			err = 0;
3595 	uint32_t		resmask, newresmask;
3596 	mac_resource_props_t	tmrp, umrp;
3597 
3598 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3599 
3600 	err = mac_validate_props(mrp);
3601 	if (err != 0)
3602 		return (err);
3603 
3604 	bcopy(&mip->mi_resource_props, &umrp, sizeof (mac_resource_props_t));
3605 	resmask = umrp.mrp_mask;
3606 	mac_update_resources(mrp, &umrp, B_FALSE);
3607 	newresmask = umrp.mrp_mask;
3608 
3609 	if (resmask == 0 && newresmask != 0) {
3610 		/*
3611 		 * Bandwidth, priority or cpu link properties configured,
3612 		 * must disable fastpath.
3613 		 */
3614 		if ((err = mac_fastpath_disable((mac_handle_t)mip)) != 0)
3615 			return (err);
3616 	}
3617 
3618 	/*
3619 	 * Since bind_cpu may be modified by mac_client_set_resources()
3620 	 * we use a copy of bind_cpu and finally cache bind_cpu in mip.
3621 	 * This allows us to cache only user edits in mip.
3622 	 */
3623 	bcopy(mrp, &tmrp, sizeof (mac_resource_props_t));
3624 	mcip = mac_primary_client_handle(mip);
3625 	if (mcip != NULL && (mcip->mci_state_flags & MCIS_IS_AGGR_PORT) == 0) {
3626 		err =
3627 		    mac_client_set_resources((mac_client_handle_t)mcip, &tmrp);
3628 	}
3629 
3630 	/* Only update the values if mac_client_set_resources succeeded */
3631 	if (err == 0) {
3632 		bcopy(&umrp, &mip->mi_resource_props,
3633 		    sizeof (mac_resource_props_t));
3634 		/*
3635 		 * If bankwidth, priority or cpu link properties cleared,
3636 		 * renable fastpath.
3637 		 */
3638 		if (resmask != 0 && newresmask == 0)
3639 			mac_fastpath_enable((mac_handle_t)mip);
3640 	} else if (resmask == 0 && newresmask != 0) {
3641 		mac_fastpath_enable((mac_handle_t)mip);
3642 	}
3643 	return (err);
3644 }
3645 
3646 int
3647 mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3648 {
3649 	int err;
3650 
3651 	i_mac_perim_enter((mac_impl_t *)mh);
3652 	err = i_mac_set_resources(mh, mrp);
3653 	i_mac_perim_exit((mac_impl_t *)mh);
3654 	return (err);
3655 }
3656 
3657 /*
3658  * Get the properties cached for the specified MAC instance.
3659  */
3660 void
3661 mac_get_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3662 {
3663 	mac_impl_t 		*mip = (mac_impl_t *)mh;
3664 	mac_client_impl_t	*mcip;
3665 
3666 	if (mip->mi_state_flags & MIS_IS_VNIC) {
3667 		mcip = mac_primary_client_handle(mip);
3668 		if (mcip != NULL) {
3669 			mac_client_get_resources((mac_client_handle_t)mcip,
3670 			    mrp);
3671 			return;
3672 		}
3673 	}
3674 	bcopy(&mip->mi_resource_props, mrp, sizeof (mac_resource_props_t));
3675 }
3676 
3677 /*
3678  * Rename a mac client, its flow, and the kstat.
3679  */
3680 int
3681 mac_rename_primary(mac_handle_t mh, const char *new_name)
3682 {
3683 	mac_impl_t		*mip = (mac_impl_t *)mh;
3684 	mac_client_impl_t	*cur_clnt = NULL;
3685 	flow_entry_t		*fep;
3686 
3687 	i_mac_perim_enter(mip);
3688 
3689 	/*
3690 	 * VNICs: we need to change the sys flow name and
3691 	 * the associated flow kstat.
3692 	 */
3693 	if (mip->mi_state_flags & MIS_IS_VNIC) {
3694 		ASSERT(new_name != NULL);
3695 		mac_rename_flow_names(mac_vnic_lower(mip), new_name);
3696 		goto done;
3697 	}
3698 	/*
3699 	 * This mac may itself be an aggr link, or it may have some client
3700 	 * which is an aggr port. For both cases, we need to change the
3701 	 * aggr port's mac client name, its flow name and the associated flow
3702 	 * kstat.
3703 	 */
3704 	if (mip->mi_state_flags & MIS_IS_AGGR) {
3705 		mac_capab_aggr_t aggr_cap;
3706 		mac_rename_fn_t rename_fn;
3707 		boolean_t ret;
3708 
3709 		ASSERT(new_name != NULL);
3710 		ret = i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_AGGR,
3711 		    (void *)(&aggr_cap));
3712 		ASSERT(ret == B_TRUE);
3713 		rename_fn = aggr_cap.mca_rename_fn;
3714 		rename_fn(new_name, mip->mi_driver);
3715 		/*
3716 		 * The aggr's client name and kstat flow name will be
3717 		 * updated below, i.e. via mac_rename_flow_names.
3718 		 */
3719 	}
3720 
3721 	for (cur_clnt = mip->mi_clients_list; cur_clnt != NULL;
3722 	    cur_clnt = cur_clnt->mci_client_next) {
3723 		if (cur_clnt->mci_state_flags & MCIS_IS_AGGR_PORT) {
3724 			if (new_name != NULL) {
3725 				char *str_st = cur_clnt->mci_name;
3726 				char *str_del = strchr(str_st, '-');
3727 
3728 				ASSERT(str_del != NULL);
3729 				bzero(str_del + 1, MAXNAMELEN -
3730 				    (str_del - str_st + 1));
3731 				bcopy(new_name, str_del + 1,
3732 				    strlen(new_name));
3733 			}
3734 			fep = cur_clnt->mci_flent;
3735 			mac_rename_flow(fep, cur_clnt->mci_name);
3736 			break;
3737 		} else if (new_name != NULL &&
3738 		    cur_clnt->mci_state_flags & MCIS_USE_DATALINK_NAME) {
3739 			mac_rename_flow_names(cur_clnt, new_name);
3740 			break;
3741 		}
3742 	}
3743 
3744 done:
3745 	i_mac_perim_exit(mip);
3746 	return (0);
3747 }
3748 
3749 /*
3750  * Rename the MAC client's flow names
3751  */
3752 static void
3753 mac_rename_flow_names(mac_client_impl_t *mcip, const char *new_name)
3754 {
3755 	flow_entry_t	*flent;
3756 	uint16_t	vid;
3757 	char		flowname[MAXFLOWNAMELEN];
3758 	mac_impl_t	*mip = mcip->mci_mip;
3759 
3760 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3761 
3762 	/*
3763 	 * Use mi_rw_lock to ensure that threads not in the mac perimeter
3764 	 * see a self-consistent value for mci_name
3765 	 */
3766 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
3767 	(void) strlcpy(mcip->mci_name, new_name, sizeof (mcip->mci_name));
3768 	rw_exit(&mip->mi_rw_lock);
3769 
3770 	mac_rename_flow(mcip->mci_flent, new_name);
3771 
3772 	if (mcip->mci_nflents == 1)
3773 		return;
3774 
3775 	/*
3776 	 * We have to rename all the others too, no stats to destroy for
3777 	 * these.
3778 	 */
3779 	for (flent = mcip->mci_flent_list; flent != NULL;
3780 	    flent = flent->fe_client_next) {
3781 		if (flent != mcip->mci_flent) {
3782 			vid = i_mac_flow_vid(flent);
3783 			(void) sprintf(flowname, "%s%u", new_name, vid);
3784 			mac_flow_set_name(flent, flowname);
3785 		}
3786 	}
3787 }
3788 
3789 
3790 /*
3791  * Add a flow to the MAC client's flow list - i.e list of MAC/VID tuples
3792  * defined for the specified MAC client.
3793  */
3794 static void
3795 mac_client_add_to_flow_list(mac_client_impl_t *mcip, flow_entry_t *flent)
3796 {
3797 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3798 	/*
3799 	 * The promisc Rx data path walks the mci_flent_list. Protect by
3800 	 * using mi_rw_lock
3801 	 */
3802 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3803 
3804 	/* Add it to the head */
3805 	flent->fe_client_next = mcip->mci_flent_list;
3806 	mcip->mci_flent_list = flent;
3807 	mcip->mci_nflents++;
3808 
3809 	/*
3810 	 * Keep track of the number of non-zero VIDs addresses per MAC
3811 	 * client to avoid figuring it out in the data-path.
3812 	 */
3813 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
3814 		mcip->mci_nvids++;
3815 
3816 	rw_exit(&mcip->mci_rw_lock);
3817 }
3818 
3819 /*
3820  * Remove a flow entry from the MAC client's list.
3821  */
3822 static void
3823 mac_client_remove_flow_from_list(mac_client_impl_t *mcip, flow_entry_t *flent)
3824 {
3825 	flow_entry_t	*fe = mcip->mci_flent_list;
3826 	flow_entry_t	*prev_fe = NULL;
3827 
3828 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3829 	/*
3830 	 * The promisc Rx data path walks the mci_flent_list. Protect by
3831 	 * using mci_rw_lock
3832 	 */
3833 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3834 	while ((fe != NULL) && (fe != flent)) {
3835 		prev_fe = fe;
3836 		fe = fe->fe_client_next;
3837 	}
3838 
3839 	ASSERT(fe != NULL);
3840 	if (prev_fe == NULL) {
3841 		/* Deleting the first node */
3842 		mcip->mci_flent_list = fe->fe_client_next;
3843 	} else {
3844 		prev_fe->fe_client_next = fe->fe_client_next;
3845 	}
3846 	mcip->mci_nflents--;
3847 
3848 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
3849 		mcip->mci_nvids--;
3850 
3851 	rw_exit(&mcip->mci_rw_lock);
3852 }
3853 
3854 /*
3855  * Check if the given VID belongs to this MAC client.
3856  */
3857 boolean_t
3858 mac_client_check_flow_vid(mac_client_impl_t *mcip, uint16_t vid)
3859 {
3860 	flow_entry_t	*flent;
3861 	uint16_t	mci_vid;
3862 
3863 	/* The mci_flent_list is protected by mci_rw_lock */
3864 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3865 	for (flent = mcip->mci_flent_list; flent != NULL;
3866 	    flent = flent->fe_client_next) {
3867 		mci_vid = i_mac_flow_vid(flent);
3868 		if (vid == mci_vid) {
3869 			rw_exit(&mcip->mci_rw_lock);
3870 			return (B_TRUE);
3871 		}
3872 	}
3873 	rw_exit(&mcip->mci_rw_lock);
3874 	return (B_FALSE);
3875 }
3876 
3877 /*
3878  * Get the flow entry for the specified <MAC addr, VID> tuple.
3879  */
3880 static flow_entry_t *
3881 mac_client_get_flow(mac_client_impl_t *mcip, mac_unicast_impl_t *muip)
3882 {
3883 	mac_address_t *map = mcip->mci_unicast;
3884 	flow_entry_t *flent;
3885 	uint16_t vid;
3886 	flow_desc_t flow_desc;
3887 
3888 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3889 
3890 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
3891 	if (bcmp(flow_desc.fd_dst_mac, map->ma_addr, map->ma_len) != 0)
3892 		return (NULL);
3893 
3894 	for (flent = mcip->mci_flent_list; flent != NULL;
3895 	    flent = flent->fe_client_next) {
3896 		vid = i_mac_flow_vid(flent);
3897 		if (vid == muip->mui_vid) {
3898 			return (flent);
3899 		}
3900 	}
3901 
3902 	return (NULL);
3903 }
3904 
3905 /*
3906  * Since mci_flent has the SRSs, when we want to remove it, we replace
3907  * the flow_desc_t in mci_flent with that of an existing flent and then
3908  * remove that flent instead of mci_flent.
3909  */
3910 static flow_entry_t *
3911 mac_client_swap_mciflent(mac_client_impl_t *mcip)
3912 {
3913 	flow_entry_t	*flent = mcip->mci_flent;
3914 	flow_tab_t	*ft = flent->fe_flow_tab;
3915 	flow_entry_t	*flent1;
3916 	flow_desc_t	fl_desc;
3917 	char		fl_name[MAXFLOWNAMELEN];
3918 	int		err;
3919 
3920 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3921 	ASSERT(mcip->mci_nflents > 1);
3922 
3923 	/* get the next flent following the primary flent  */
3924 	flent1 = mcip->mci_flent_list->fe_client_next;
3925 	ASSERT(flent1 != NULL && flent1->fe_flow_tab == ft);
3926 
3927 	/*
3928 	 * Remove the flent from the flow table before updating the
3929 	 * flow descriptor as the hash depends on the flow descriptor.
3930 	 * This also helps incoming packet classification avoid having
3931 	 * to grab fe_lock. Access to fe_flow_desc of a flent not in the
3932 	 * flow table is done under the fe_lock so that log or stat functions
3933 	 * see a self-consistent fe_flow_desc. The name and desc are specific
3934 	 * to a flow, the rest are shared by all the clients, including
3935 	 * resource control etc.
3936 	 */
3937 	mac_flow_remove(ft, flent, B_TRUE);
3938 	mac_flow_remove(ft, flent1, B_TRUE);
3939 
3940 	bcopy(&flent->fe_flow_desc, &fl_desc, sizeof (flow_desc_t));
3941 	bcopy(flent->fe_flow_name, fl_name, MAXFLOWNAMELEN);
3942 
3943 	/* update the primary flow entry */
3944 	mutex_enter(&flent->fe_lock);
3945 	bcopy(&flent1->fe_flow_desc, &flent->fe_flow_desc,
3946 	    sizeof (flow_desc_t));
3947 	bcopy(&flent1->fe_flow_name, &flent->fe_flow_name, MAXFLOWNAMELEN);
3948 	mutex_exit(&flent->fe_lock);
3949 
3950 	/* update the flow entry that is to be freed */
3951 	mutex_enter(&flent1->fe_lock);
3952 	bcopy(&fl_desc, &flent1->fe_flow_desc, sizeof (flow_desc_t));
3953 	bcopy(fl_name, &flent1->fe_flow_name, MAXFLOWNAMELEN);
3954 	mutex_exit(&flent1->fe_lock);
3955 
3956 	/* now reinsert the flow entries in the table */
3957 	err = mac_flow_add(ft, flent);
3958 	ASSERT(err == 0);
3959 
3960 	err = mac_flow_add(ft, flent1);
3961 	ASSERT(err == 0);
3962 
3963 	return (flent1);
3964 }
3965 
3966 /*
3967  * Return whether there is only one flow entry associated with this
3968  * MAC client.
3969  */
3970 static boolean_t
3971 mac_client_single_rcvr(mac_client_impl_t *mcip)
3972 {
3973 	return (mcip->mci_nflents == 1);
3974 }
3975 
3976 int
3977 mac_validate_props(mac_resource_props_t *mrp)
3978 {
3979 	if (mrp == NULL)
3980 		return (0);
3981 
3982 	if (mrp->mrp_mask & MRP_PRIORITY) {
3983 		mac_priority_level_t	pri = mrp->mrp_priority;
3984 
3985 		if (pri < MPL_LOW || pri > MPL_RESET)
3986 			return (EINVAL);
3987 	}
3988 
3989 	if (mrp->mrp_mask & MRP_MAXBW) {
3990 		uint64_t maxbw = mrp->mrp_maxbw;
3991 
3992 		if (maxbw < MRP_MAXBW_MINVAL && maxbw != 0)
3993 			return (EINVAL);
3994 	}
3995 	if (mrp->mrp_mask & MRP_CPUS) {
3996 		int i, j;
3997 		mac_cpu_mode_t	fanout;
3998 
3999 		if (mrp->mrp_ncpus > ncpus || mrp->mrp_ncpus > MAX_SR_FANOUT)
4000 			return (EINVAL);
4001 
4002 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4003 			for (j = 0; j < mrp->mrp_ncpus; j++) {
4004 				if (i != j &&
4005 				    mrp->mrp_cpu[i] == mrp->mrp_cpu[j]) {
4006 					return (EINVAL);
4007 				}
4008 			}
4009 		}
4010 
4011 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4012 			cpu_t *cp;
4013 			int rv;
4014 
4015 			mutex_enter(&cpu_lock);
4016 			cp = cpu_get(mrp->mrp_cpu[i]);
4017 			if (cp != NULL)
4018 				rv = cpu_is_online(cp);
4019 			else
4020 				rv = 0;
4021 			mutex_exit(&cpu_lock);
4022 			if (rv == 0)
4023 				return (EINVAL);
4024 		}
4025 
4026 		fanout = mrp->mrp_fanout_mode;
4027 		if (fanout < 0 || fanout > MCM_CPUS)
4028 			return (EINVAL);
4029 	}
4030 	return (0);
4031 }
4032 
4033 /*
4034  * Send a MAC_NOTE_LINK notification to all the MAC clients whenever the
4035  * underlying physical link is down. This is to allow MAC clients to
4036  * communicate with other clients.
4037  */
4038 void
4039 mac_virtual_link_update(mac_impl_t *mip)
4040 {
4041 	if (mip->mi_linkstate != LINK_STATE_UP)
4042 		i_mac_notify(mip, MAC_NOTE_LINK);
4043 }
4044 
4045 /*
4046  * For clients that have a pass-thru MAC, e.g. VNIC, we set the VNIC's
4047  * mac handle in the client.
4048  */
4049 void
4050 mac_set_upper_mac(mac_client_handle_t mch, mac_handle_t mh)
4051 {
4052 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
4053 
4054 	mcip->mci_upper_mip = (mac_impl_t *)mh;
4055 }
4056 
4057 /*
4058  * Mark the mac as being used exclusively by the single mac client that is
4059  * doing some control operation on this mac. No further opens of this mac
4060  * will be allowed until this client calls mac_unmark_exclusive. The mac
4061  * client calling this function must already be in the mac perimeter
4062  */
4063 int
4064 mac_mark_exclusive(mac_handle_t mh)
4065 {
4066 	mac_impl_t	*mip = (mac_impl_t *)mh;
4067 
4068 	ASSERT(MAC_PERIM_HELD(mh));
4069 	/*
4070 	 * Look up its entry in the global hash table.
4071 	 */
4072 	rw_enter(&i_mac_impl_lock, RW_WRITER);
4073 	if (mip->mi_state_flags & MIS_DISABLED) {
4074 		rw_exit(&i_mac_impl_lock);
4075 		return (ENOENT);
4076 	}
4077 
4078 	/*
4079 	 * A reference to mac is held even if the link is not plumbed.
4080 	 * In i_dls_link_create() we open the MAC interface and hold the
4081 	 * reference. There is an additional reference for the mac_open
4082 	 * done in acquiring the mac perimeter
4083 	 */
4084 	if (mip->mi_ref != 2) {
4085 		rw_exit(&i_mac_impl_lock);
4086 		return (EBUSY);
4087 	}
4088 
4089 	ASSERT(!(mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
4090 	mip->mi_state_flags |= MIS_EXCLUSIVE_HELD;
4091 	rw_exit(&i_mac_impl_lock);
4092 	return (0);
4093 }
4094 
4095 void
4096 mac_unmark_exclusive(mac_handle_t mh)
4097 {
4098 	mac_impl_t	*mip = (mac_impl_t *)mh;
4099 
4100 	ASSERT(MAC_PERIM_HELD(mh));
4101 
4102 	rw_enter(&i_mac_impl_lock, RW_WRITER);
4103 	/* 1 for the creation and another for the perimeter */
4104 	ASSERT(mip->mi_ref == 2 && (mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
4105 	mip->mi_state_flags &= ~MIS_EXCLUSIVE_HELD;
4106 	rw_exit(&i_mac_impl_lock);
4107 }
4108 
4109 /*
4110  * Set the MTU for the specified device. The function returns EBUSY if
4111  * another MAC client prevents the caller to become the exclusive client.
4112  * Returns EAGAIN if the client is started.
4113  */
4114 int
4115 mac_set_mtu(mac_handle_t mh, uint_t new_mtu, uint_t *old_mtu_arg)
4116 {
4117 	mac_impl_t *mip = (mac_impl_t *)mh;
4118 	uint_t old_mtu;
4119 	int rv;
4120 	boolean_t exclusive = B_FALSE;
4121 
4122 	i_mac_perim_enter(mip);
4123 
4124 	if ((mip->mi_callbacks->mc_callbacks & MC_SETPROP) == 0 ||
4125 	    (mip->mi_callbacks->mc_callbacks & MC_GETPROP) == 0) {
4126 		rv = ENOTSUP;
4127 		goto bail;
4128 	}
4129 
4130 	if ((rv = mac_mark_exclusive(mh)) != 0)
4131 		goto bail;
4132 	exclusive = B_TRUE;
4133 
4134 	if (mip->mi_active > 0) {
4135 		/*
4136 		 * The MAC instance is started, for example due to the
4137 		 * presence of a promiscuous clients. Fail the operation
4138 		 * since the MAC's MTU cannot be changed while the NIC
4139 		 * is started.
4140 		 */
4141 		rv = EAGAIN;
4142 		goto bail;
4143 	}
4144 
4145 	mac_sdu_get(mh, NULL, &old_mtu);
4146 
4147 	if (old_mtu != new_mtu) {
4148 		rv = mip->mi_callbacks->mc_setprop(mip->mi_driver,
4149 		    "mtu", MAC_PROP_MTU, sizeof (uint_t), &new_mtu);
4150 	}
4151 
4152 bail:
4153 	if (exclusive)
4154 		mac_unmark_exclusive(mh);
4155 	i_mac_perim_exit(mip);
4156 
4157 	if (rv == 0 && old_mtu_arg != NULL)
4158 		*old_mtu_arg = old_mtu;
4159 	return (rv);
4160 }
4161 
4162 void
4163 mac_get_hwgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
4164     uint_t *n_rings, uint_t *type, uint_t *n_clnts, char *clnts_name)
4165 {
4166 	mac_impl_t *mip = (mac_impl_t *)mh;
4167 	mac_grp_client_t *mcip;
4168 	uint_t i = 0, index = 0;
4169 
4170 	/* Revisit when we implement fully dynamic group allocation */
4171 	ASSERT(grp_index >= 0 && grp_index < mip->mi_rx_group_count);
4172 
4173 	rw_enter(&mip->mi_rw_lock, RW_READER);
4174 	*grp_num = mip->mi_rx_groups[grp_index].mrg_index;
4175 	*type = mip->mi_rx_groups[grp_index].mrg_type;
4176 	*n_rings = mip->mi_rx_groups[grp_index].mrg_cur_count;
4177 	for (mcip = mip->mi_rx_groups[grp_index].mrg_clients; mcip != NULL;
4178 	    mcip = mcip->mgc_next) {
4179 		int name_len = strlen(mcip->mgc_client->mci_name);
4180 
4181 		/*
4182 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
4183 		 * names.
4184 		 * XXXX Formating the client name string needs to be moved
4185 		 * to user land when fixing the size of dhi_clnts in
4186 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
4187 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
4188 		 */
4189 		if (index + name_len >= MAXCLIENTNAMELEN) {
4190 			index = MAXCLIENTNAMELEN;
4191 			break;
4192 		}
4193 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
4194 		    name_len);
4195 		index += name_len;
4196 		clnts_name[index++] = ',';
4197 		i++;
4198 	}
4199 
4200 	/* Get rid of the last , */
4201 	if (index > 0)
4202 		clnts_name[index - 1] = '\0';
4203 	*n_clnts = i;
4204 	rw_exit(&mip->mi_rw_lock);
4205 }
4206 
4207 uint_t
4208 mac_hwgrp_num(mac_handle_t mh)
4209 {
4210 	mac_impl_t *mip = (mac_impl_t *)mh;
4211 
4212 	return (mip->mi_rx_group_count);
4213 }
4214