xref: /titanic_41/usr/src/uts/common/io/mac/mac_client.c (revision 0db3240d392634cfff2f95fb6da34b56b8dc574f)
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 2010 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/mac_stat.h>
112 #include <sys/dls.h>
113 #include <sys/dld.h>
114 #include <sys/modctl.h>
115 #include <sys/fs/dv_node.h>
116 #include <sys/thread.h>
117 #include <sys/proc.h>
118 #include <sys/callb.h>
119 #include <sys/cpuvar.h>
120 #include <sys/atomic.h>
121 #include <sys/sdt.h>
122 #include <sys/mac_flow.h>
123 #include <sys/ddi_intr_impl.h>
124 #include <sys/disp.h>
125 #include <sys/sdt.h>
126 #include <sys/vnic.h>
127 #include <sys/vnic_impl.h>
128 #include <sys/vlan.h>
129 #include <inet/ip.h>
130 #include <inet/ip6.h>
131 #include <sys/exacct.h>
132 #include <sys/exacct_impl.h>
133 #include <inet/nd.h>
134 #include <sys/ethernet.h>
135 
136 kmem_cache_t	*mac_client_impl_cache;
137 kmem_cache_t	*mac_promisc_impl_cache;
138 
139 static boolean_t mac_client_single_rcvr(mac_client_impl_t *);
140 static flow_entry_t *mac_client_swap_mciflent(mac_client_impl_t *);
141 static flow_entry_t *mac_client_get_flow(mac_client_impl_t *,
142     mac_unicast_impl_t *);
143 static void mac_client_remove_flow_from_list(mac_client_impl_t *,
144     flow_entry_t *);
145 static void mac_client_add_to_flow_list(mac_client_impl_t *, flow_entry_t *);
146 static void mac_rename_flow_names(mac_client_impl_t *, const char *);
147 static void mac_virtual_link_update(mac_impl_t *);
148 static int mac_client_datapath_setup(mac_client_impl_t *, uint16_t,
149     uint8_t *, mac_resource_props_t *, boolean_t, mac_unicast_impl_t *);
150 static void mac_client_datapath_teardown(mac_client_handle_t,
151     mac_unicast_impl_t *, flow_entry_t *);
152 
153 /* ARGSUSED */
154 static int
155 i_mac_client_impl_ctor(void *buf, void *arg, int kmflag)
156 {
157 	int	i;
158 	mac_client_impl_t	*mcip = buf;
159 
160 	bzero(buf, MAC_CLIENT_IMPL_SIZE);
161 	mutex_init(&mcip->mci_tx_cb_lock, NULL, MUTEX_DRIVER, NULL);
162 	mcip->mci_tx_notify_cb_info.mcbi_lockp = &mcip->mci_tx_cb_lock;
163 
164 	ASSERT(mac_tx_percpu_cnt >= 0);
165 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
166 		mutex_init(&mcip->mci_tx_pcpu[i].pcpu_tx_lock, NULL,
167 		    MUTEX_DRIVER, NULL);
168 	}
169 	cv_init(&mcip->mci_tx_cv, NULL, CV_DRIVER, NULL);
170 
171 	return (0);
172 }
173 
174 /* ARGSUSED */
175 static void
176 i_mac_client_impl_dtor(void *buf, void *arg)
177 {
178 	int	i;
179 	mac_client_impl_t *mcip = buf;
180 
181 	ASSERT(mcip->mci_promisc_list == NULL);
182 	ASSERT(mcip->mci_unicast_list == NULL);
183 	ASSERT(mcip->mci_state_flags == 0);
184 	ASSERT(mcip->mci_tx_flag == 0);
185 
186 	mutex_destroy(&mcip->mci_tx_cb_lock);
187 
188 	ASSERT(mac_tx_percpu_cnt >= 0);
189 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
190 		ASSERT(mcip->mci_tx_pcpu[i].pcpu_tx_refcnt == 0);
191 		mutex_destroy(&mcip->mci_tx_pcpu[i].pcpu_tx_lock);
192 	}
193 	cv_destroy(&mcip->mci_tx_cv);
194 }
195 
196 /* ARGSUSED */
197 static int
198 i_mac_promisc_impl_ctor(void *buf, void *arg, int kmflag)
199 {
200 	mac_promisc_impl_t	*mpip = buf;
201 
202 	bzero(buf, sizeof (mac_promisc_impl_t));
203 	mpip->mpi_mci_link.mcb_objp = buf;
204 	mpip->mpi_mci_link.mcb_objsize = sizeof (mac_promisc_impl_t);
205 	mpip->mpi_mi_link.mcb_objp = buf;
206 	mpip->mpi_mi_link.mcb_objsize = sizeof (mac_promisc_impl_t);
207 	return (0);
208 }
209 
210 /* ARGSUSED */
211 static void
212 i_mac_promisc_impl_dtor(void *buf, void *arg)
213 {
214 	mac_promisc_impl_t	*mpip = buf;
215 
216 	ASSERT(mpip->mpi_mci_link.mcb_objp != NULL);
217 	ASSERT(mpip->mpi_mci_link.mcb_objsize == sizeof (mac_promisc_impl_t));
218 	ASSERT(mpip->mpi_mi_link.mcb_objp == mpip->mpi_mci_link.mcb_objp);
219 	ASSERT(mpip->mpi_mi_link.mcb_objsize == sizeof (mac_promisc_impl_t));
220 
221 	mpip->mpi_mci_link.mcb_objp = NULL;
222 	mpip->mpi_mci_link.mcb_objsize = 0;
223 	mpip->mpi_mi_link.mcb_objp = NULL;
224 	mpip->mpi_mi_link.mcb_objsize = 0;
225 
226 	ASSERT(mpip->mpi_mci_link.mcb_flags == 0);
227 	mpip->mpi_mci_link.mcb_objsize = 0;
228 }
229 
230 void
231 mac_client_init(void)
232 {
233 	ASSERT(mac_tx_percpu_cnt >= 0);
234 
235 	mac_client_impl_cache = kmem_cache_create("mac_client_impl_cache",
236 	    MAC_CLIENT_IMPL_SIZE, 0, i_mac_client_impl_ctor,
237 	    i_mac_client_impl_dtor, NULL, NULL, NULL, 0);
238 	ASSERT(mac_client_impl_cache != NULL);
239 
240 	mac_promisc_impl_cache = kmem_cache_create("mac_promisc_impl_cache",
241 	    sizeof (mac_promisc_impl_t), 0, i_mac_promisc_impl_ctor,
242 	    i_mac_promisc_impl_dtor, NULL, NULL, NULL, 0);
243 	ASSERT(mac_promisc_impl_cache != NULL);
244 }
245 
246 void
247 mac_client_fini(void)
248 {
249 	kmem_cache_destroy(mac_client_impl_cache);
250 	kmem_cache_destroy(mac_promisc_impl_cache);
251 }
252 
253 /*
254  * Return the lower MAC client handle from the VNIC driver for the
255  * specified VNIC MAC instance.
256  */
257 mac_client_impl_t *
258 mac_vnic_lower(mac_impl_t *mip)
259 {
260 	mac_capab_vnic_t cap;
261 	mac_client_impl_t *mcip;
262 
263 	VERIFY(i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_VNIC, &cap));
264 	mcip = cap.mcv_mac_client_handle(cap.mcv_arg);
265 
266 	return (mcip);
267 }
268 
269 /*
270  * Return the MAC client handle of the primary MAC client for the
271  * specified MAC instance, or NULL otherwise.
272  */
273 mac_client_impl_t *
274 mac_primary_client_handle(mac_impl_t *mip)
275 {
276 	mac_client_impl_t *mcip;
277 
278 	if (mip->mi_state_flags & MIS_IS_VNIC)
279 		return (mac_vnic_lower(mip));
280 
281 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
282 
283 	for (mcip = mip->mi_clients_list; mcip != NULL;
284 	    mcip = mcip->mci_client_next) {
285 		if (MCIP_DATAPATH_SETUP(mcip) && mac_is_primary_client(mcip))
286 			return (mcip);
287 	}
288 	return (NULL);
289 }
290 
291 /*
292  * Open a MAC specified by its MAC name.
293  */
294 int
295 mac_open(const char *macname, mac_handle_t *mhp)
296 {
297 	mac_impl_t	*mip;
298 	int		err;
299 
300 	/*
301 	 * Look up its entry in the global hash table.
302 	 */
303 	if ((err = mac_hold(macname, &mip)) != 0)
304 		return (err);
305 
306 	/*
307 	 * Hold the dip associated to the MAC to prevent it from being
308 	 * detached. For a softmac, its underlying dip is held by the
309 	 * mi_open() callback.
310 	 *
311 	 * This is done to be more tolerant with some defective drivers,
312 	 * which incorrectly handle mac_unregister() failure in their
313 	 * xxx_detach() routine. For example, some drivers ignore the
314 	 * failure of mac_unregister() and free all resources that
315 	 * that are needed for data transmition.
316 	 */
317 	e_ddi_hold_devi(mip->mi_dip);
318 
319 	if (!(mip->mi_callbacks->mc_callbacks & MC_OPEN)) {
320 		*mhp = (mac_handle_t)mip;
321 		return (0);
322 	}
323 
324 	/*
325 	 * The mac perimeter is used in both mac_open and mac_close by the
326 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
327 	 */
328 	i_mac_perim_enter(mip);
329 	mip->mi_oref++;
330 	if (mip->mi_oref != 1 || ((err = mip->mi_open(mip->mi_driver)) == 0)) {
331 		*mhp = (mac_handle_t)mip;
332 		i_mac_perim_exit(mip);
333 		return (0);
334 	}
335 	mip->mi_oref--;
336 	ddi_release_devi(mip->mi_dip);
337 	mac_rele(mip);
338 	i_mac_perim_exit(mip);
339 	return (err);
340 }
341 
342 /*
343  * Open a MAC specified by its linkid.
344  */
345 int
346 mac_open_by_linkid(datalink_id_t linkid, mac_handle_t *mhp)
347 {
348 	dls_dl_handle_t	dlh;
349 	int		err;
350 
351 	if ((err = dls_devnet_hold_tmp(linkid, &dlh)) != 0)
352 		return (err);
353 
354 	dls_devnet_prop_task_wait(dlh);
355 
356 	err = mac_open(dls_devnet_mac(dlh), mhp);
357 
358 	dls_devnet_rele_tmp(dlh);
359 	return (err);
360 }
361 
362 /*
363  * Open a MAC specified by its link name.
364  */
365 int
366 mac_open_by_linkname(const char *link, mac_handle_t *mhp)
367 {
368 	datalink_id_t	linkid;
369 	int		err;
370 
371 	if ((err = dls_mgmt_get_linkid(link, &linkid)) != 0)
372 		return (err);
373 	return (mac_open_by_linkid(linkid, mhp));
374 }
375 
376 /*
377  * Close the specified MAC.
378  */
379 void
380 mac_close(mac_handle_t mh)
381 {
382 	mac_impl_t	*mip = (mac_impl_t *)mh;
383 
384 	i_mac_perim_enter(mip);
385 	/*
386 	 * The mac perimeter is used in both mac_open and mac_close by the
387 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
388 	 */
389 	if (mip->mi_callbacks->mc_callbacks & MC_OPEN) {
390 		ASSERT(mip->mi_oref != 0);
391 		if (--mip->mi_oref == 0) {
392 			if ((mip->mi_callbacks->mc_callbacks & MC_CLOSE))
393 				mip->mi_close(mip->mi_driver);
394 		}
395 	}
396 	i_mac_perim_exit(mip);
397 	ddi_release_devi(mip->mi_dip);
398 	mac_rele(mip);
399 }
400 
401 /*
402  * Misc utility functions to retrieve various information about a MAC
403  * instance or a MAC client.
404  */
405 
406 const mac_info_t *
407 mac_info(mac_handle_t mh)
408 {
409 	return (&((mac_impl_t *)mh)->mi_info);
410 }
411 
412 dev_info_t *
413 mac_devinfo_get(mac_handle_t mh)
414 {
415 	return (((mac_impl_t *)mh)->mi_dip);
416 }
417 
418 void *
419 mac_driver(mac_handle_t mh)
420 {
421 	return (((mac_impl_t *)mh)->mi_driver);
422 }
423 
424 const char *
425 mac_name(mac_handle_t mh)
426 {
427 	return (((mac_impl_t *)mh)->mi_name);
428 }
429 
430 int
431 mac_type(mac_handle_t mh)
432 {
433 	return (((mac_impl_t *)mh)->mi_type->mt_type);
434 }
435 
436 int
437 mac_nativetype(mac_handle_t mh)
438 {
439 	return (((mac_impl_t *)mh)->mi_type->mt_nativetype);
440 }
441 
442 char *
443 mac_client_name(mac_client_handle_t mch)
444 {
445 	return (((mac_client_impl_t *)mch)->mci_name);
446 }
447 
448 minor_t
449 mac_minor(mac_handle_t mh)
450 {
451 	return (((mac_impl_t *)mh)->mi_minor);
452 }
453 
454 /*
455  * Return the VID associated with a MAC client. This function should
456  * be called for clients which are associated with only one VID.
457  */
458 uint16_t
459 mac_client_vid(mac_client_handle_t mch)
460 {
461 	uint16_t		vid = VLAN_ID_NONE;
462 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
463 	flow_desc_t		flow_desc;
464 
465 	if (mcip->mci_nflents == 0)
466 		return (vid);
467 
468 	ASSERT(MCIP_DATAPATH_SETUP(mcip) && mac_client_single_rcvr(mcip));
469 
470 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
471 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
472 		vid = flow_desc.fd_vid;
473 
474 	return (vid);
475 }
476 
477 /*
478  * Return whether the specified MAC client corresponds to a VLAN VNIC.
479  */
480 boolean_t
481 mac_client_is_vlan_vnic(mac_client_handle_t mch)
482 {
483 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
484 
485 	return (((mcip->mci_state_flags & MCIS_IS_VNIC) != 0) &&
486 	    ((mcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) != 0));
487 }
488 
489 /*
490  * Return the link speed associated with the specified MAC client.
491  *
492  * The link speed of a MAC client is equal to the smallest value of
493  * 1) the current link speed of the underlying NIC, or
494  * 2) the bandwidth limit set for the MAC client.
495  *
496  * Note that the bandwidth limit can be higher than the speed
497  * of the underlying NIC. This is allowed to avoid spurious
498  * administration action failures or artifically lowering the
499  * bandwidth limit of a link that may  have temporarily lowered
500  * its link speed due to hardware problem or administrator action.
501  */
502 static uint64_t
503 mac_client_ifspeed(mac_client_impl_t *mcip)
504 {
505 	mac_impl_t *mip = mcip->mci_mip;
506 	uint64_t nic_speed;
507 
508 	nic_speed = mac_stat_get((mac_handle_t)mip, MAC_STAT_IFSPEED);
509 
510 	if (nic_speed == 0) {
511 		return (0);
512 	} else {
513 		uint64_t policy_limit = (uint64_t)-1;
514 
515 		if (MCIP_RESOURCE_PROPS_MASK(mcip) & MRP_MAXBW)
516 			policy_limit = MCIP_RESOURCE_PROPS_MAXBW(mcip);
517 
518 		return (MIN(policy_limit, nic_speed));
519 	}
520 }
521 
522 /*
523  * Return the link state of the specified client. If here are more
524  * than one clients of the underying mac_impl_t, the link state
525  * will always be UP regardless of the link state of the underlying
526  * mac_impl_t. This is needed to allow the MAC clients to continue
527  * to communicate with each other even when the physical link of
528  * their mac_impl_t is down.
529  */
530 static uint64_t
531 mac_client_link_state(mac_client_impl_t *mcip)
532 {
533 	mac_impl_t *mip = mcip->mci_mip;
534 	uint16_t vid;
535 	mac_client_impl_t *mci_list;
536 	mac_unicast_impl_t *mui_list, *oth_mui_list;
537 
538 	/*
539 	 * Returns LINK_STATE_UP if there are other MAC clients defined on
540 	 * mac_impl_t which share same VLAN ID as that of mcip. Note that
541 	 * if 'mcip' has more than one VID's then we match ANY one of the
542 	 * VID's with other MAC client's VID's and return LINK_STATE_UP.
543 	 */
544 	rw_enter(&mcip->mci_rw_lock, RW_READER);
545 	for (mui_list = mcip->mci_unicast_list; mui_list != NULL;
546 	    mui_list = mui_list->mui_next) {
547 		vid = mui_list->mui_vid;
548 		for (mci_list = mip->mi_clients_list; mci_list != NULL;
549 		    mci_list = mci_list->mci_client_next) {
550 			if (mci_list == mcip)
551 				continue;
552 			for (oth_mui_list = mci_list->mci_unicast_list;
553 			    oth_mui_list != NULL; oth_mui_list = oth_mui_list->
554 			    mui_next) {
555 				if (vid == oth_mui_list->mui_vid) {
556 					rw_exit(&mcip->mci_rw_lock);
557 					return (LINK_STATE_UP);
558 				}
559 			}
560 		}
561 	}
562 	rw_exit(&mcip->mci_rw_lock);
563 
564 	return (mac_stat_get((mac_handle_t)mip, MAC_STAT_LINK_STATE));
565 }
566 
567 /*
568  * These statistics are consumed by dladm show-link -s <vnic>,
569  * dladm show-vnic -s and netstat. With the introduction of dlstat,
570  * dladm show-link -s and dladm show-vnic -s witll be EOL'ed while
571  * netstat will consume from kstats introduced for dlstat. This code
572  * will be removed at that time.
573  */
574 
575 /*
576  * Return the statistics of a MAC client. These statistics are different
577  * then the statistics of the underlying MAC which are returned by
578  * mac_stat_get().
579  */
580 uint64_t
581 mac_client_stat_get(mac_client_handle_t mch, uint_t stat)
582 {
583 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
584 	mac_impl_t 		*mip = mcip->mci_mip;
585 	flow_entry_t 		*flent = mcip->mci_flent;
586 	mac_soft_ring_set_t 	*mac_srs;
587 	mac_rx_stats_t		*mac_rx_stat;
588 	mac_tx_stats_t		*mac_tx_stat;
589 	int i;
590 	uint64_t val = 0;
591 
592 	mac_srs = (mac_soft_ring_set_t *)(flent->fe_tx_srs);
593 	mac_tx_stat = &mac_srs->srs_tx.st_stat;
594 
595 	switch (stat) {
596 	case MAC_STAT_LINK_STATE:
597 		val = mac_client_link_state(mcip);
598 		break;
599 	case MAC_STAT_LINK_UP:
600 		val = (mac_client_link_state(mcip) == LINK_STATE_UP);
601 		break;
602 	case MAC_STAT_PROMISC:
603 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_PROMISC);
604 		break;
605 	case MAC_STAT_LOWLINK_STATE:
606 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_LOWLINK_STATE);
607 		break;
608 	case MAC_STAT_IFSPEED:
609 		val = mac_client_ifspeed(mcip);
610 		break;
611 	case MAC_STAT_MULTIRCV:
612 		val = mcip->mci_misc_stat.mms_multircv;
613 		break;
614 	case MAC_STAT_BRDCSTRCV:
615 		val = mcip->mci_misc_stat.mms_brdcstrcv;
616 		break;
617 	case MAC_STAT_MULTIXMT:
618 		val = mcip->mci_misc_stat.mms_multixmt;
619 		break;
620 	case MAC_STAT_BRDCSTXMT:
621 		val = mcip->mci_misc_stat.mms_brdcstxmt;
622 		break;
623 	case MAC_STAT_OBYTES:
624 		val = mac_tx_stat->mts_obytes;
625 		break;
626 	case MAC_STAT_OPACKETS:
627 		val = mac_tx_stat->mts_opackets;
628 		break;
629 	case MAC_STAT_OERRORS:
630 		val = mac_tx_stat->mts_oerrors;
631 		break;
632 	case MAC_STAT_IPACKETS:
633 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
634 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
635 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
636 			val += mac_rx_stat->mrs_intrcnt +
637 			    mac_rx_stat->mrs_pollcnt + mac_rx_stat->mrs_lclcnt;
638 		}
639 		break;
640 	case MAC_STAT_RBYTES:
641 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
642 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
643 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
644 			val += mac_rx_stat->mrs_intrbytes +
645 			    mac_rx_stat->mrs_pollbytes +
646 			    mac_rx_stat->mrs_lclbytes;
647 		}
648 		break;
649 	case MAC_STAT_IERRORS:
650 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
651 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
652 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
653 			val += mac_rx_stat->mrs_ierrors;
654 		}
655 		break;
656 	default:
657 		val = mac_driver_stat_default(mip, stat);
658 		break;
659 	}
660 
661 	return (val);
662 }
663 
664 /*
665  * Return the statistics of the specified MAC instance.
666  */
667 uint64_t
668 mac_stat_get(mac_handle_t mh, uint_t stat)
669 {
670 	mac_impl_t	*mip = (mac_impl_t *)mh;
671 	uint64_t	val;
672 	int		ret;
673 
674 	/*
675 	 * The range of stat determines where it is maintained.  Stat
676 	 * values from 0 up to (but not including) MAC_STAT_MIN are
677 	 * mainteined by the mac module itself.  Everything else is
678 	 * maintained by the driver.
679 	 *
680 	 * If the mac_impl_t being queried corresponds to a VNIC,
681 	 * the stats need to be queried from the lower MAC client
682 	 * corresponding to the VNIC. (The mac_link_update()
683 	 * invoked by the driver to the lower MAC causes the *lower
684 	 * MAC* to update its mi_linkstate, and send a notification
685 	 * to its MAC clients. Due to the VNIC passthrough,
686 	 * these notifications are sent to the upper MAC clients
687 	 * of the VNIC directly, and the upper mac_impl_t of the VNIC
688 	 * does not have a valid mi_linkstate.
689 	 */
690 	if (stat < MAC_STAT_MIN && !(mip->mi_state_flags & MIS_IS_VNIC)) {
691 		/* these stats are maintained by the mac module itself */
692 		switch (stat) {
693 		case MAC_STAT_LINK_STATE:
694 			return (mip->mi_linkstate);
695 		case MAC_STAT_LINK_UP:
696 			return (mip->mi_linkstate == LINK_STATE_UP);
697 		case MAC_STAT_PROMISC:
698 			return (mip->mi_devpromisc != 0);
699 		case MAC_STAT_LOWLINK_STATE:
700 			return (mip->mi_lowlinkstate);
701 		default:
702 			ASSERT(B_FALSE);
703 		}
704 	}
705 
706 	/*
707 	 * Call the driver to get the given statistic.
708 	 */
709 	ret = mip->mi_getstat(mip->mi_driver, stat, &val);
710 	if (ret != 0) {
711 		/*
712 		 * The driver doesn't support this statistic.  Get the
713 		 * statistic's default value.
714 		 */
715 		val = mac_driver_stat_default(mip, stat);
716 	}
717 	return (val);
718 }
719 
720 /*
721  * Query hardware rx ring corresponding to the pseudo ring.
722  */
723 uint64_t
724 mac_pseudo_rx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
725 {
726 	return (mac_rx_ring_stat_get(handle, stat));
727 }
728 
729 /*
730  * Query hardware tx ring corresponding to the pseudo ring.
731  */
732 uint64_t
733 mac_pseudo_tx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
734 {
735 	return (mac_tx_ring_stat_get(handle, stat));
736 }
737 
738 /*
739  * Utility function which returns the VID associated with a flow entry.
740  */
741 uint16_t
742 i_mac_flow_vid(flow_entry_t *flent)
743 {
744 	flow_desc_t	flow_desc;
745 
746 	mac_flow_get_desc(flent, &flow_desc);
747 
748 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
749 		return (flow_desc.fd_vid);
750 	return (VLAN_ID_NONE);
751 }
752 
753 /*
754  * Verify the validity of the specified unicast MAC address. Returns B_TRUE
755  * if the address is valid, B_FALSE otherwise (multicast address, or incorrect
756  * length.
757  */
758 boolean_t
759 mac_unicst_verify(mac_handle_t mh, const uint8_t *addr, uint_t len)
760 {
761 	mac_impl_t	*mip = (mac_impl_t *)mh;
762 
763 	/*
764 	 * Verify the address. No lock is needed since mi_type and plugin
765 	 * details don't change after mac_register().
766 	 */
767 	if ((len != mip->mi_type->mt_addr_length) ||
768 	    (mip->mi_type->mt_ops.mtops_unicst_verify(addr,
769 	    mip->mi_pdata)) != 0) {
770 		return (B_FALSE);
771 	} else {
772 		return (B_TRUE);
773 	}
774 }
775 
776 void
777 mac_sdu_get(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu)
778 {
779 	mac_impl_t	*mip = (mac_impl_t *)mh;
780 
781 	if (min_sdu != NULL)
782 		*min_sdu = mip->mi_sdu_min;
783 	if (max_sdu != NULL)
784 		*max_sdu = mip->mi_sdu_max;
785 }
786 
787 /*
788  * Update the MAC unicast address of the specified client's flows. Currently
789  * only one unicast MAC unicast address is allowed per client.
790  */
791 static void
792 mac_unicast_update_client_flow(mac_client_impl_t *mcip)
793 {
794 	mac_impl_t *mip = mcip->mci_mip;
795 	flow_entry_t *flent = mcip->mci_flent;
796 	mac_address_t *map = mcip->mci_unicast;
797 	flow_desc_t flow_desc;
798 
799 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
800 	ASSERT(flent != NULL);
801 
802 	mac_flow_get_desc(flent, &flow_desc);
803 	ASSERT(flow_desc.fd_mask & FLOW_LINK_DST);
804 
805 	bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
806 	mac_flow_set_desc(flent, &flow_desc);
807 
808 	/*
809 	 * The v6 local addr (used by mac protection) needs to be
810 	 * regenerated because our mac address has changed.
811 	 */
812 	mac_protect_update_v6_local_addr(mcip);
813 
814 	/*
815 	 * A MAC client could have one MAC address but multiple
816 	 * VLANs. In that case update the flow entries corresponding
817 	 * to all VLANs of the MAC client.
818 	 */
819 	for (flent = mcip->mci_flent_list; flent != NULL;
820 	    flent = flent->fe_client_next) {
821 		mac_flow_get_desc(flent, &flow_desc);
822 		if (!(flent->fe_type & FLOW_PRIMARY_MAC ||
823 		    flent->fe_type & FLOW_VNIC_MAC))
824 			continue;
825 
826 		bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
827 		mac_flow_set_desc(flent, &flow_desc);
828 	}
829 }
830 
831 /*
832  * Update all clients that share the same unicast address.
833  */
834 void
835 mac_unicast_update_clients(mac_impl_t *mip, mac_address_t *map)
836 {
837 	mac_client_impl_t *mcip;
838 
839 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
840 
841 	/*
842 	 * Find all clients that share the same unicast MAC address and update
843 	 * them appropriately.
844 	 */
845 	for (mcip = mip->mi_clients_list; mcip != NULL;
846 	    mcip = mcip->mci_client_next) {
847 		/*
848 		 * Ignore clients that don't share this MAC address.
849 		 */
850 		if (map != mcip->mci_unicast)
851 			continue;
852 
853 		/*
854 		 * Update those clients with same old unicast MAC address.
855 		 */
856 		mac_unicast_update_client_flow(mcip);
857 	}
858 }
859 
860 /*
861  * Update the unicast MAC address of the specified VNIC MAC client.
862  *
863  * Check whether the operation is valid. Any of following cases should fail:
864  *
865  * 1. It's a VLAN type of VNIC.
866  * 2. The new value is current "primary" MAC address.
867  * 3. The current MAC address is shared with other clients.
868  * 4. The new MAC address has been used. This case will be valid when
869  *    client migration is fully supported.
870  */
871 int
872 mac_vnic_unicast_set(mac_client_handle_t mch, const uint8_t *addr)
873 {
874 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
875 	mac_impl_t *mip = mcip->mci_mip;
876 	mac_address_t *map = mcip->mci_unicast;
877 	int err;
878 
879 	ASSERT(!(mip->mi_state_flags & MIS_IS_VNIC));
880 	ASSERT(mcip->mci_state_flags & MCIS_IS_VNIC);
881 	ASSERT(mcip->mci_flags != MAC_CLIENT_FLAGS_PRIMARY);
882 
883 	i_mac_perim_enter(mip);
884 
885 	/*
886 	 * If this is a VLAN type of VNIC, it's using "primary" MAC address
887 	 * of the underlying interface. Must fail here. Refer to case 1 above.
888 	 */
889 	if (bcmp(map->ma_addr, mip->mi_addr, map->ma_len) == 0) {
890 		i_mac_perim_exit(mip);
891 		return (ENOTSUP);
892 	}
893 
894 	/*
895 	 * If the new address is the "primary" one, must fail. Refer to
896 	 * case 2 above.
897 	 */
898 	if (bcmp(addr, mip->mi_addr, map->ma_len) == 0) {
899 		i_mac_perim_exit(mip);
900 		return (EACCES);
901 	}
902 
903 	/*
904 	 * If the address is shared by multiple clients, must fail. Refer
905 	 * to case 3 above.
906 	 */
907 	if (mac_check_macaddr_shared(map)) {
908 		i_mac_perim_exit(mip);
909 		return (EBUSY);
910 	}
911 
912 	/*
913 	 * If the new address has been used, must fail for now. Refer to
914 	 * case 4 above.
915 	 */
916 	if (mac_find_macaddr(mip, (uint8_t *)addr) != NULL) {
917 		i_mac_perim_exit(mip);
918 		return (ENOTSUP);
919 	}
920 
921 	/*
922 	 * Update the MAC address.
923 	 */
924 	err = mac_update_macaddr(map, (uint8_t *)addr);
925 
926 	if (err != 0) {
927 		i_mac_perim_exit(mip);
928 		return (err);
929 	}
930 
931 	/*
932 	 * Update all flows of this MAC client.
933 	 */
934 	mac_unicast_update_client_flow(mcip);
935 
936 	i_mac_perim_exit(mip);
937 	return (0);
938 }
939 
940 /*
941  * Program the new primary unicast address of the specified MAC.
942  *
943  * Function mac_update_macaddr() takes care different types of underlying
944  * MAC. If the underlying MAC is VNIC, the VNIC driver must have registerd
945  * mi_unicst() entry point, that indirectly calls mac_vnic_unicast_set()
946  * which will take care of updating the MAC address of the corresponding
947  * MAC client.
948  *
949  * This is the only interface that allow the client to update the "primary"
950  * MAC address of the underlying MAC. The new value must have not been
951  * used by other clients.
952  */
953 int
954 mac_unicast_primary_set(mac_handle_t mh, const uint8_t *addr)
955 {
956 	mac_impl_t *mip = (mac_impl_t *)mh;
957 	mac_address_t *map;
958 	int err;
959 
960 	/* verify the address validity */
961 	if (!mac_unicst_verify(mh, addr, mip->mi_type->mt_addr_length))
962 		return (EINVAL);
963 
964 	i_mac_perim_enter(mip);
965 
966 	/*
967 	 * If the new value is the same as the current primary address value,
968 	 * there's nothing to do.
969 	 */
970 	if (bcmp(addr, mip->mi_addr, mip->mi_type->mt_addr_length) == 0) {
971 		i_mac_perim_exit(mip);
972 		return (0);
973 	}
974 
975 	if (mac_find_macaddr(mip, (uint8_t *)addr) != 0) {
976 		i_mac_perim_exit(mip);
977 		return (EBUSY);
978 	}
979 
980 	map = mac_find_macaddr(mip, mip->mi_addr);
981 	ASSERT(map != NULL);
982 
983 	/*
984 	 * Update the MAC address.
985 	 */
986 	if (mip->mi_state_flags & MIS_IS_AGGR) {
987 		mac_capab_aggr_t aggr_cap;
988 
989 		/*
990 		 * If the mac is an aggregation, other than the unicast
991 		 * addresses programming, aggr must be informed about this
992 		 * primary unicst address change to change its mac address
993 		 * policy to be user-specified.
994 		 */
995 		ASSERT(map->ma_type == MAC_ADDRESS_TYPE_UNICAST_CLASSIFIED);
996 		VERIFY(i_mac_capab_get(mh, MAC_CAPAB_AGGR, &aggr_cap));
997 		err = aggr_cap.mca_unicst(mip->mi_driver, addr);
998 		if (err == 0)
999 			bcopy(addr, map->ma_addr, map->ma_len);
1000 	} else {
1001 		err = mac_update_macaddr(map, (uint8_t *)addr);
1002 	}
1003 
1004 	if (err != 0) {
1005 		i_mac_perim_exit(mip);
1006 		return (err);
1007 	}
1008 
1009 	mac_unicast_update_clients(mip, map);
1010 
1011 	/*
1012 	 * Save the new primary MAC address in mac_impl_t.
1013 	 */
1014 	bcopy(addr, mip->mi_addr, mip->mi_type->mt_addr_length);
1015 
1016 	i_mac_perim_exit(mip);
1017 
1018 	if (err == 0)
1019 		i_mac_notify(mip, MAC_NOTE_UNICST);
1020 
1021 	return (err);
1022 }
1023 
1024 /*
1025  * Return the current primary MAC address of the specified MAC.
1026  */
1027 void
1028 mac_unicast_primary_get(mac_handle_t mh, uint8_t *addr)
1029 {
1030 	mac_impl_t *mip = (mac_impl_t *)mh;
1031 
1032 	rw_enter(&mip->mi_rw_lock, RW_READER);
1033 	bcopy(mip->mi_addr, addr, mip->mi_type->mt_addr_length);
1034 	rw_exit(&mip->mi_rw_lock);
1035 }
1036 
1037 /*
1038  * Return information about the use of the primary MAC address of the
1039  * specified MAC instance:
1040  *
1041  * - if client_name is non-NULL, it must point to a string of at
1042  *   least MAXNAMELEN bytes, and will be set to the name of the MAC
1043  *   client which uses the primary MAC address.
1044  *
1045  * - if in_use is non-NULL, used to return whether the primary MAC
1046  *   address is currently in use.
1047  */
1048 void
1049 mac_unicast_primary_info(mac_handle_t mh, char *client_name, boolean_t *in_use)
1050 {
1051 	mac_impl_t *mip = (mac_impl_t *)mh;
1052 	mac_client_impl_t *cur_client;
1053 
1054 	if (in_use != NULL)
1055 		*in_use = B_FALSE;
1056 	if (client_name != NULL)
1057 		bzero(client_name, MAXNAMELEN);
1058 
1059 	/*
1060 	 * The mi_rw_lock is used to protect threads that don't hold the
1061 	 * mac perimeter to get a consistent view of the mi_clients_list.
1062 	 * Threads that modify the list must hold both the mac perimeter and
1063 	 * mi_rw_lock(RW_WRITER)
1064 	 */
1065 	rw_enter(&mip->mi_rw_lock, RW_READER);
1066 	for (cur_client = mip->mi_clients_list; cur_client != NULL;
1067 	    cur_client = cur_client->mci_client_next) {
1068 		if (mac_is_primary_client(cur_client) ||
1069 		    (mip->mi_state_flags & MIS_IS_VNIC)) {
1070 			rw_exit(&mip->mi_rw_lock);
1071 			if (in_use != NULL)
1072 				*in_use = B_TRUE;
1073 			if (client_name != NULL) {
1074 				bcopy(cur_client->mci_name, client_name,
1075 				    MAXNAMELEN);
1076 			}
1077 			return;
1078 		}
1079 	}
1080 	rw_exit(&mip->mi_rw_lock);
1081 }
1082 
1083 /*
1084  * Return the current destination MAC address of the specified MAC.
1085  */
1086 boolean_t
1087 mac_dst_get(mac_handle_t mh, uint8_t *addr)
1088 {
1089 	mac_impl_t *mip = (mac_impl_t *)mh;
1090 
1091 	rw_enter(&mip->mi_rw_lock, RW_READER);
1092 	if (mip->mi_dstaddr_set)
1093 		bcopy(mip->mi_dstaddr, addr, mip->mi_type->mt_addr_length);
1094 	rw_exit(&mip->mi_rw_lock);
1095 	return (mip->mi_dstaddr_set);
1096 }
1097 
1098 /*
1099  * Add the specified MAC client to the list of clients which opened
1100  * the specified MAC.
1101  */
1102 static void
1103 mac_client_add(mac_client_impl_t *mcip)
1104 {
1105 	mac_impl_t *mip = mcip->mci_mip;
1106 
1107 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1108 
1109 	/* add VNIC to the front of the list */
1110 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1111 	mcip->mci_client_next = mip->mi_clients_list;
1112 	mip->mi_clients_list = mcip;
1113 	mip->mi_nclients++;
1114 	rw_exit(&mip->mi_rw_lock);
1115 }
1116 
1117 /*
1118  * Remove the specified MAC client from the list of clients which opened
1119  * the specified MAC.
1120  */
1121 static void
1122 mac_client_remove(mac_client_impl_t *mcip)
1123 {
1124 	mac_impl_t *mip = mcip->mci_mip;
1125 	mac_client_impl_t **prev, *cclient;
1126 
1127 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1128 
1129 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1130 	prev = &mip->mi_clients_list;
1131 	cclient = *prev;
1132 	while (cclient != NULL && cclient != mcip) {
1133 		prev = &cclient->mci_client_next;
1134 		cclient = *prev;
1135 	}
1136 	ASSERT(cclient != NULL);
1137 	*prev = cclient->mci_client_next;
1138 	mip->mi_nclients--;
1139 	rw_exit(&mip->mi_rw_lock);
1140 }
1141 
1142 static mac_unicast_impl_t *
1143 mac_client_find_vid(mac_client_impl_t *mcip, uint16_t vid)
1144 {
1145 	mac_unicast_impl_t *muip = mcip->mci_unicast_list;
1146 
1147 	while ((muip != NULL) && (muip->mui_vid != vid))
1148 		muip = muip->mui_next;
1149 
1150 	return (muip);
1151 }
1152 
1153 /*
1154  * Return whether the specified (MAC address, VID) tuple is already used by
1155  * one of the MAC clients associated with the specified MAC.
1156  */
1157 static boolean_t
1158 mac_addr_in_use(mac_impl_t *mip, uint8_t *mac_addr, uint16_t vid)
1159 {
1160 	mac_client_impl_t *client;
1161 	mac_address_t *map;
1162 
1163 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1164 
1165 	for (client = mip->mi_clients_list; client != NULL;
1166 	    client = client->mci_client_next) {
1167 
1168 		/*
1169 		 * Ignore clients that don't have unicast address.
1170 		 */
1171 		if (client->mci_unicast_list == NULL)
1172 			continue;
1173 
1174 		map = client->mci_unicast;
1175 
1176 		if ((bcmp(mac_addr, map->ma_addr, map->ma_len) == 0) &&
1177 		    (mac_client_find_vid(client, vid) != NULL)) {
1178 			return (B_TRUE);
1179 		}
1180 	}
1181 
1182 	return (B_FALSE);
1183 }
1184 
1185 /*
1186  * Generate a random MAC address. The MAC address prefix is
1187  * stored in the array pointed to by mac_addr, and its length, in bytes,
1188  * is specified by prefix_len. The least significant bits
1189  * after prefix_len bytes are generated, and stored after the prefix
1190  * in the mac_addr array.
1191  */
1192 int
1193 mac_addr_random(mac_client_handle_t mch, uint_t prefix_len,
1194     uint8_t *mac_addr, mac_diag_t *diag)
1195 {
1196 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1197 	mac_impl_t *mip = mcip->mci_mip;
1198 	size_t addr_len = mip->mi_type->mt_addr_length;
1199 
1200 	if (prefix_len >= addr_len) {
1201 		*diag = MAC_DIAG_MACPREFIXLEN_INVALID;
1202 		return (EINVAL);
1203 	}
1204 
1205 	/* check the prefix value */
1206 	if (prefix_len > 0) {
1207 		bzero(mac_addr + prefix_len, addr_len - prefix_len);
1208 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr,
1209 		    addr_len)) {
1210 			*diag = MAC_DIAG_MACPREFIX_INVALID;
1211 			return (EINVAL);
1212 		}
1213 	}
1214 
1215 	/* generate the MAC address */
1216 	if (prefix_len < addr_len) {
1217 		(void) random_get_pseudo_bytes(mac_addr +
1218 		    prefix_len, addr_len - prefix_len);
1219 	}
1220 
1221 	*diag = 0;
1222 	return (0);
1223 }
1224 
1225 /*
1226  * Set the priority range for this MAC client. This will be used to
1227  * determine the absolute priority for the threads created for this
1228  * MAC client using the specified "low", "medium" and "high" level.
1229  * This will also be used for any subflows on this MAC client.
1230  */
1231 #define	MAC_CLIENT_SET_PRIORITY_RANGE(mcip, pri) {			\
1232 	(mcip)->mci_min_pri = FLOW_MIN_PRIORITY(MINCLSYSPRI,	\
1233 	    MAXCLSYSPRI, (pri));					\
1234 	(mcip)->mci_max_pri = FLOW_MAX_PRIORITY(MINCLSYSPRI,	\
1235 	    MAXCLSYSPRI, (mcip)->mci_min_pri);				\
1236 	}
1237 
1238 /*
1239  * MAC client open entry point. Return a new MAC client handle. Each
1240  * MAC client is associated with a name, specified through the 'name'
1241  * argument.
1242  */
1243 int
1244 mac_client_open(mac_handle_t mh, mac_client_handle_t *mchp, char *name,
1245     uint16_t flags)
1246 {
1247 	mac_impl_t		*mip = (mac_impl_t *)mh;
1248 	mac_client_impl_t	*mcip;
1249 	int			err = 0;
1250 	boolean_t		share_desired;
1251 	flow_entry_t		*flent = NULL;
1252 
1253 	share_desired = (flags & MAC_OPEN_FLAGS_SHARES_DESIRED) != 0;
1254 	*mchp = NULL;
1255 
1256 	i_mac_perim_enter(mip);
1257 
1258 	if (mip->mi_state_flags & MIS_IS_VNIC) {
1259 		/*
1260 		 * The underlying MAC is a VNIC. Return the MAC client
1261 		 * handle of the lower MAC which was obtained by
1262 		 * the VNIC driver when it did its mac_client_open().
1263 		 */
1264 
1265 		mcip = mac_vnic_lower(mip);
1266 
1267 		/*
1268 		 * Note that multiple mac clients share the same mcip in
1269 		 * this case.
1270 		 */
1271 		if (flags & MAC_OPEN_FLAGS_EXCLUSIVE)
1272 			mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1273 
1274 		if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1275 			mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1276 
1277 		mip->mi_clients_list = mcip;
1278 		i_mac_perim_exit(mip);
1279 		*mchp = (mac_client_handle_t)mcip;
1280 		return (err);
1281 	}
1282 
1283 	mcip = kmem_cache_alloc(mac_client_impl_cache, KM_SLEEP);
1284 
1285 	mcip->mci_mip = mip;
1286 	mcip->mci_upper_mip = NULL;
1287 	mcip->mci_rx_fn = mac_pkt_drop;
1288 	mcip->mci_rx_arg = NULL;
1289 	mcip->mci_rx_p_fn = NULL;
1290 	mcip->mci_rx_p_arg = NULL;
1291 	mcip->mci_p_unicast_list = NULL;
1292 	mcip->mci_direct_rx_fn = NULL;
1293 	mcip->mci_direct_rx_arg = NULL;
1294 
1295 	mcip->mci_unicast_list = NULL;
1296 
1297 	if ((flags & MAC_OPEN_FLAGS_IS_VNIC) != 0)
1298 		mcip->mci_state_flags |= MCIS_IS_VNIC;
1299 
1300 	if ((flags & MAC_OPEN_FLAGS_EXCLUSIVE) != 0)
1301 		mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1302 
1303 	if ((flags & MAC_OPEN_FLAGS_IS_AGGR_PORT) != 0)
1304 		mcip->mci_state_flags |= MCIS_IS_AGGR_PORT;
1305 
1306 	if (mip->mi_state_flags & MIS_IS_AGGR)
1307 		mcip->mci_state_flags |= MCIS_IS_AGGR;
1308 
1309 	if ((flags & MAC_OPEN_FLAGS_USE_DATALINK_NAME) != 0) {
1310 		datalink_id_t	linkid;
1311 
1312 		ASSERT(name == NULL);
1313 		if ((err = dls_devnet_macname2linkid(mip->mi_name,
1314 		    &linkid)) != 0) {
1315 			goto done;
1316 		}
1317 		if ((err = dls_mgmt_get_linkinfo(linkid, mcip->mci_name, NULL,
1318 		    NULL, NULL)) != 0) {
1319 			/*
1320 			 * Use mac name if dlmgmtd is not available.
1321 			 */
1322 			if (err == EBADF) {
1323 				(void) strlcpy(mcip->mci_name, mip->mi_name,
1324 				    sizeof (mcip->mci_name));
1325 				err = 0;
1326 			} else {
1327 				goto done;
1328 			}
1329 		}
1330 		mcip->mci_state_flags |= MCIS_USE_DATALINK_NAME;
1331 	} else {
1332 		ASSERT(name != NULL);
1333 		if (strlen(name) > MAXNAMELEN) {
1334 			err = EINVAL;
1335 			goto done;
1336 		}
1337 		(void) strlcpy(mcip->mci_name, name, sizeof (mcip->mci_name));
1338 	}
1339 
1340 	if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1341 		mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1342 
1343 	if (flags & MAC_OPEN_FLAGS_NO_UNICAST_ADDR)
1344 		mcip->mci_state_flags |= MCIS_NO_UNICAST_ADDR;
1345 
1346 	mac_protect_init(mcip);
1347 
1348 	/* the subflow table will be created dynamically */
1349 	mcip->mci_subflow_tab = NULL;
1350 
1351 	mcip->mci_misc_stat.mms_multircv = 0;
1352 	mcip->mci_misc_stat.mms_brdcstrcv = 0;
1353 	mcip->mci_misc_stat.mms_multixmt = 0;
1354 	mcip->mci_misc_stat.mms_brdcstxmt = 0;
1355 
1356 	/* Create an initial flow */
1357 
1358 	err = mac_flow_create(NULL, NULL, mcip->mci_name, NULL,
1359 	    mcip->mci_state_flags & MCIS_IS_VNIC ? FLOW_VNIC_MAC :
1360 	    FLOW_PRIMARY_MAC, &flent);
1361 	if (err != 0)
1362 		goto done;
1363 	mcip->mci_flent = flent;
1364 	FLOW_MARK(flent, FE_MC_NO_DATAPATH);
1365 	flent->fe_mcip = mcip;
1366 	/*
1367 	 * Place initial creation reference on the flow. This reference
1368 	 * is released in the corresponding delete action viz.
1369 	 * mac_unicast_remove after waiting for all transient refs to
1370 	 * to go away. The wait happens in mac_flow_wait.
1371 	 */
1372 	FLOW_REFHOLD(flent);
1373 
1374 	/*
1375 	 * Do this ahead of the mac_bcast_add() below so that the mi_nclients
1376 	 * will have the right value for mac_rx_srs_setup().
1377 	 */
1378 	mac_client_add(mcip);
1379 
1380 	mcip->mci_share = NULL;
1381 	if (share_desired)
1382 		i_mac_share_alloc(mcip);
1383 
1384 	DTRACE_PROBE2(mac__client__open__allocated, mac_impl_t *,
1385 	    mcip->mci_mip, mac_client_impl_t *, mcip);
1386 	*mchp = (mac_client_handle_t)mcip;
1387 
1388 	/*
1389 	 * We will do mimimal datapath setup to allow a MAC client to
1390 	 * transmit or receive non-unicast packets without waiting
1391 	 * for mac_unicast_add.
1392 	 */
1393 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1394 		if ((err = mac_client_datapath_setup(mcip, VLAN_ID_NONE,
1395 		    NULL, NULL, B_TRUE, NULL)) != 0) {
1396 			goto done;
1397 		}
1398 	}
1399 	i_mac_perim_exit(mip);
1400 	return (0);
1401 
1402 done:
1403 	i_mac_perim_exit(mip);
1404 	mcip->mci_state_flags = 0;
1405 	mcip->mci_tx_flag = 0;
1406 	kmem_cache_free(mac_client_impl_cache, mcip);
1407 	return (err);
1408 }
1409 
1410 /*
1411  * Close the specified MAC client handle.
1412  */
1413 void
1414 mac_client_close(mac_client_handle_t mch, uint16_t flags)
1415 {
1416 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1417 	mac_impl_t		*mip = mcip->mci_mip;
1418 	flow_entry_t		*flent;
1419 
1420 	i_mac_perim_enter(mip);
1421 
1422 	if (flags & MAC_CLOSE_FLAGS_EXCLUSIVE)
1423 		mcip->mci_state_flags &= ~MCIS_EXCLUSIVE;
1424 
1425 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
1426 	    !(flags & MAC_CLOSE_FLAGS_IS_VNIC)) {
1427 		/*
1428 		 * This is an upper VNIC client initiated operation.
1429 		 * The lower MAC client will be closed by the VNIC driver
1430 		 * when the VNIC is deleted.
1431 		 */
1432 
1433 		i_mac_perim_exit(mip);
1434 		return;
1435 	}
1436 
1437 	/* If we have only setup up minimal datapth setup, tear it down */
1438 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1439 		mac_client_datapath_teardown((mac_client_handle_t)mcip, NULL,
1440 		    mcip->mci_flent);
1441 		mcip->mci_state_flags &= ~MCIS_NO_UNICAST_ADDR;
1442 	}
1443 
1444 	/*
1445 	 * Remove the flent associated with the MAC client
1446 	 */
1447 	flent = mcip->mci_flent;
1448 	mcip->mci_flent = NULL;
1449 	FLOW_FINAL_REFRELE(flent);
1450 
1451 	/*
1452 	 * MAC clients must remove the unicast addresses and promisc callbacks
1453 	 * they added before issuing a mac_client_close().
1454 	 */
1455 	ASSERT(mcip->mci_unicast_list == NULL);
1456 	ASSERT(mcip->mci_promisc_list == NULL);
1457 	ASSERT(mcip->mci_tx_notify_cb_list == NULL);
1458 
1459 	i_mac_share_free(mcip);
1460 	mac_protect_fini(mcip);
1461 	mac_client_remove(mcip);
1462 
1463 	i_mac_perim_exit(mip);
1464 	mcip->mci_subflow_tab = NULL;
1465 	mcip->mci_state_flags = 0;
1466 	mcip->mci_tx_flag = 0;
1467 	kmem_cache_free(mac_client_impl_cache, mch);
1468 }
1469 
1470 /*
1471  * Set the rx bypass receive callback.
1472  */
1473 boolean_t
1474 mac_rx_bypass_set(mac_client_handle_t mch, mac_direct_rx_t rx_fn, void *arg1)
1475 {
1476 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1477 	mac_impl_t		*mip = mcip->mci_mip;
1478 
1479 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1480 
1481 	/*
1482 	 * If the mac_client is a VLAN, we should not do DLS bypass and
1483 	 * instead let the packets come up via mac_rx_deliver so the vlan
1484 	 * header can be stripped.
1485 	 */
1486 	if (mcip->mci_nvids > 0)
1487 		return (B_FALSE);
1488 
1489 	/*
1490 	 * These are not accessed directly in the data path, and hence
1491 	 * don't need any protection
1492 	 */
1493 	mcip->mci_direct_rx_fn = rx_fn;
1494 	mcip->mci_direct_rx_arg = arg1;
1495 	return (B_TRUE);
1496 }
1497 
1498 /*
1499  * Enable/Disable rx bypass. By default, bypass is assumed to be enabled.
1500  */
1501 void
1502 mac_rx_bypass_enable(mac_client_handle_t mch)
1503 {
1504 	((mac_client_impl_t *)mch)->mci_state_flags &= ~MCIS_RX_BYPASS_DISABLE;
1505 }
1506 
1507 void
1508 mac_rx_bypass_disable(mac_client_handle_t mch)
1509 {
1510 	((mac_client_impl_t *)mch)->mci_state_flags |= MCIS_RX_BYPASS_DISABLE;
1511 }
1512 
1513 /*
1514  * Set the receive callback for the specified MAC client. There can be
1515  * at most one such callback per MAC client.
1516  */
1517 void
1518 mac_rx_set(mac_client_handle_t mch, mac_rx_t rx_fn, void *arg)
1519 {
1520 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1521 	mac_impl_t	*mip = mcip->mci_mip;
1522 
1523 	/*
1524 	 * Instead of adding an extra set of locks and refcnts in
1525 	 * the datapath at the mac client boundary, we temporarily quiesce
1526 	 * the SRS and related entities. We then change the receive function
1527 	 * without interference from any receive data thread and then reenable
1528 	 * the data flow subsequently.
1529 	 */
1530 	i_mac_perim_enter(mip);
1531 	mac_rx_client_quiesce(mch);
1532 
1533 	mcip->mci_rx_fn = rx_fn;
1534 	mcip->mci_rx_arg = arg;
1535 	mac_rx_client_restart(mch);
1536 	i_mac_perim_exit(mip);
1537 }
1538 
1539 /*
1540  * Reset the receive callback for the specified MAC client.
1541  */
1542 void
1543 mac_rx_clear(mac_client_handle_t mch)
1544 {
1545 	mac_rx_set(mch, mac_pkt_drop, NULL);
1546 }
1547 
1548 /*
1549  * Walk the MAC client subflow table and updates their priority values.
1550  */
1551 static int
1552 mac_update_subflow_priority_cb(flow_entry_t *flent, void *arg)
1553 {
1554 	mac_flow_update_priority(arg, flent);
1555 	return (0);
1556 }
1557 
1558 void
1559 mac_update_subflow_priority(mac_client_impl_t *mcip)
1560 {
1561 	(void) mac_flow_walk(mcip->mci_subflow_tab,
1562 	    mac_update_subflow_priority_cb, mcip);
1563 }
1564 
1565 /*
1566  * Modify the TX or RX ring properties. We could either just move around
1567  * rings, i.e add/remove rings given to a client. Or this might cause the
1568  * client to move from hardware based to software or the other way around.
1569  * If we want to reset this property, then we clear the mask, additionally
1570  * if the client was given a non-default group we remove all rings except
1571  * for 1 and give it back to the default group.
1572  */
1573 int
1574 mac_client_set_rings_prop(mac_client_impl_t *mcip, mac_resource_props_t *mrp,
1575     mac_resource_props_t *tmrp)
1576 {
1577 	mac_impl_t		*mip = mcip->mci_mip;
1578 	flow_entry_t		*flent = mcip->mci_flent;
1579 	uint8_t			*mac_addr;
1580 	int			err = 0;
1581 	mac_group_t		*defgrp;
1582 	mac_group_t		*group;
1583 	mac_group_t		*ngrp;
1584 	mac_resource_props_t	*cmrp = MCIP_RESOURCE_PROPS(mcip);
1585 	uint_t			ringcnt;
1586 	boolean_t		unspec;
1587 
1588 	if (mcip->mci_share != NULL)
1589 		return (EINVAL);
1590 
1591 	if (mrp->mrp_mask & MRP_RX_RINGS) {
1592 		unspec = mrp->mrp_mask & MRP_RXRINGS_UNSPEC;
1593 		group = flent->fe_rx_ring_group;
1594 		defgrp = MAC_DEFAULT_RX_GROUP(mip);
1595 		mac_addr = flent->fe_flow_desc.fd_dst_mac;
1596 
1597 		/*
1598 		 * No resulting change. If we are resetting on a client on
1599 		 * which there was no rx rings property. For dynamic group
1600 		 * if we are setting the same number of rings already set.
1601 		 * For static group if we are requesting a group again.
1602 		 */
1603 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1604 			if (!(tmrp->mrp_mask & MRP_RX_RINGS))
1605 				return (0);
1606 		} else {
1607 			if (unspec) {
1608 				if (tmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1609 					return (0);
1610 			} else if (mip->mi_rx_group_type ==
1611 			    MAC_GROUP_TYPE_DYNAMIC) {
1612 				if ((tmrp->mrp_mask & MRP_RX_RINGS) &&
1613 				    !(tmrp->mrp_mask & MRP_RXRINGS_UNSPEC) &&
1614 				    mrp->mrp_nrxrings == tmrp->mrp_nrxrings) {
1615 					return (0);
1616 				}
1617 			}
1618 		}
1619 		/* Resetting the prop */
1620 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1621 			/*
1622 			 * We will just keep one ring and give others back if
1623 			 * we are not the primary. For the primary we give
1624 			 * all the rings in the default group except the
1625 			 * default ring. If it is a static group, then
1626 			 * we don't do anything, but clear the MRP_RX_RINGS
1627 			 * flag.
1628 			 */
1629 			if (group != defgrp) {
1630 				if (mip->mi_rx_group_type ==
1631 				    MAC_GROUP_TYPE_DYNAMIC) {
1632 					/*
1633 					 * This group has reserved rings
1634 					 * that need to be released now,
1635 					 * so does the group.
1636 					 */
1637 					MAC_RX_RING_RELEASED(mip,
1638 					    group->mrg_cur_count);
1639 					MAC_RX_GRP_RELEASED(mip);
1640 					if ((flent->fe_type &
1641 					    FLOW_PRIMARY_MAC) != 0) {
1642 						if (mip->mi_nactiveclients ==
1643 						    1) {
1644 							(void)
1645 							    mac_rx_switch_group(
1646 							    mcip, group,
1647 							    defgrp);
1648 							return (0);
1649 						} else {
1650 							cmrp->mrp_nrxrings =
1651 							    group->
1652 							    mrg_cur_count +
1653 							    defgrp->
1654 							    mrg_cur_count - 1;
1655 						}
1656 					} else {
1657 						cmrp->mrp_nrxrings = 1;
1658 					}
1659 					(void) mac_group_ring_modify(mcip,
1660 					    group, defgrp);
1661 				} else {
1662 					/*
1663 					 * If this is a static group, we
1664 					 * need to release the group. The
1665 					 * client will remain in the same
1666 					 * group till some other client
1667 					 * needs this group.
1668 					 */
1669 					MAC_RX_GRP_RELEASED(mip);
1670 				}
1671 			/* Let check if we can give this an excl group */
1672 			} else if (group == defgrp) {
1673 				ngrp = 	mac_reserve_rx_group(mcip, mac_addr,
1674 				    B_TRUE);
1675 				/* Couldn't give it a group, that's fine */
1676 				if (ngrp == NULL)
1677 					return (0);
1678 				/* Switch to H/W */
1679 				if (mac_rx_switch_group(mcip, defgrp, ngrp) !=
1680 				    0) {
1681 					mac_stop_group(ngrp);
1682 					return (0);
1683 				}
1684 			}
1685 			/*
1686 			 * If the client is in the default group, we will
1687 			 * just clear the MRP_RX_RINGS and leave it as
1688 			 * it rather than look for an exclusive group
1689 			 * for it.
1690 			 */
1691 			return (0);
1692 		}
1693 
1694 		if (group == defgrp && ((mrp->mrp_nrxrings > 0) || unspec)) {
1695 			ngrp = 	mac_reserve_rx_group(mcip, mac_addr, B_TRUE);
1696 			if (ngrp == NULL)
1697 				return (ENOSPC);
1698 
1699 			/* Switch to H/W */
1700 			if (mac_rx_switch_group(mcip, defgrp, ngrp) != 0) {
1701 				mac_release_rx_group(mcip, ngrp);
1702 				return (ENOSPC);
1703 			}
1704 			MAC_RX_GRP_RESERVED(mip);
1705 			if (mip->mi_rx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1706 				MAC_RX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1707 		} else if (group != defgrp && !unspec &&
1708 		    mrp->mrp_nrxrings == 0) {
1709 			/* Switch to S/W */
1710 			ringcnt = group->mrg_cur_count;
1711 			if (mac_rx_switch_group(mcip, group, defgrp) != 0)
1712 				return (ENOSPC);
1713 			if (tmrp->mrp_mask & MRP_RX_RINGS) {
1714 				MAC_RX_GRP_RELEASED(mip);
1715 				if (mip->mi_rx_group_type ==
1716 				    MAC_GROUP_TYPE_DYNAMIC) {
1717 					MAC_RX_RING_RELEASED(mip, ringcnt);
1718 				}
1719 			}
1720 		} else if (group != defgrp && mip->mi_rx_group_type ==
1721 		    MAC_GROUP_TYPE_DYNAMIC) {
1722 			ringcnt = group->mrg_cur_count;
1723 			err = mac_group_ring_modify(mcip, group, defgrp);
1724 			if (err != 0)
1725 				return (err);
1726 			/*
1727 			 * Update the accounting. If this group
1728 			 * already had explicitly reserved rings,
1729 			 * we need to update the rings based on
1730 			 * the new ring count. If this group
1731 			 * had not explicitly reserved rings,
1732 			 * then we just reserve the rings asked for
1733 			 * and reserve the group.
1734 			 */
1735 			if (tmrp->mrp_mask & MRP_RX_RINGS) {
1736 				if (ringcnt > group->mrg_cur_count) {
1737 					MAC_RX_RING_RELEASED(mip,
1738 					    ringcnt - group->mrg_cur_count);
1739 				} else {
1740 					MAC_RX_RING_RESERVED(mip,
1741 					    group->mrg_cur_count - ringcnt);
1742 				}
1743 			} else {
1744 				MAC_RX_RING_RESERVED(mip, group->mrg_cur_count);
1745 				MAC_RX_GRP_RESERVED(mip);
1746 			}
1747 		}
1748 	}
1749 	if (mrp->mrp_mask & MRP_TX_RINGS) {
1750 		unspec = mrp->mrp_mask & MRP_TXRINGS_UNSPEC;
1751 		group = flent->fe_tx_ring_group;
1752 		defgrp = MAC_DEFAULT_TX_GROUP(mip);
1753 
1754 		/*
1755 		 * For static groups we only allow rings=0 or resetting the
1756 		 * rings property.
1757 		 */
1758 		if (mrp->mrp_ntxrings > 0 &&
1759 		    mip->mi_tx_group_type != MAC_GROUP_TYPE_DYNAMIC) {
1760 			return (ENOTSUP);
1761 		}
1762 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1763 			if (!(tmrp->mrp_mask & MRP_TX_RINGS))
1764 				return (0);
1765 		} else {
1766 			if (unspec) {
1767 				if (tmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1768 					return (0);
1769 			} else if (mip->mi_tx_group_type ==
1770 			    MAC_GROUP_TYPE_DYNAMIC) {
1771 				if ((tmrp->mrp_mask & MRP_TX_RINGS) &&
1772 				    !(tmrp->mrp_mask & MRP_TXRINGS_UNSPEC) &&
1773 				    mrp->mrp_ntxrings == tmrp->mrp_ntxrings) {
1774 					return (0);
1775 				}
1776 			}
1777 		}
1778 		/* Resetting the prop */
1779 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1780 			if (group != defgrp) {
1781 				if (mip->mi_tx_group_type ==
1782 				    MAC_GROUP_TYPE_DYNAMIC) {
1783 					ringcnt = group->mrg_cur_count;
1784 					if ((flent->fe_type &
1785 					    FLOW_PRIMARY_MAC) != 0) {
1786 						mac_tx_client_quiesce(
1787 						    (mac_client_handle_t)
1788 						    mcip);
1789 						mac_tx_switch_group(mcip,
1790 						    group, defgrp);
1791 						mac_tx_client_restart(
1792 						    (mac_client_handle_t)
1793 						    mcip);
1794 						MAC_TX_GRP_RELEASED(mip);
1795 						MAC_TX_RING_RELEASED(mip,
1796 						    ringcnt);
1797 						return (0);
1798 					}
1799 					cmrp->mrp_ntxrings = 1;
1800 					(void) mac_group_ring_modify(mcip,
1801 					    group, defgrp);
1802 					/*
1803 					 * This group has reserved rings
1804 					 * that need to be released now.
1805 					 */
1806 					MAC_TX_RING_RELEASED(mip, ringcnt);
1807 				}
1808 				/*
1809 				 * If this is a static group, we
1810 				 * need to release the group. The
1811 				 * client will remain in the same
1812 				 * group till some other client
1813 				 * needs this group.
1814 				 */
1815 				MAC_TX_GRP_RELEASED(mip);
1816 			} else if (group == defgrp &&
1817 			    (flent->fe_type & FLOW_PRIMARY_MAC) == 0) {
1818 				ngrp = mac_reserve_tx_group(mcip, B_TRUE);
1819 				if (ngrp == NULL)
1820 					return (0);
1821 				mac_tx_client_quiesce(
1822 				    (mac_client_handle_t)mcip);
1823 				mac_tx_switch_group(mcip, defgrp, ngrp);
1824 				mac_tx_client_restart(
1825 				    (mac_client_handle_t)mcip);
1826 			}
1827 			/*
1828 			 * If the client is in the default group, we will
1829 			 * just clear the MRP_TX_RINGS and leave it as
1830 			 * it rather than look for an exclusive group
1831 			 * for it.
1832 			 */
1833 			return (0);
1834 		}
1835 
1836 		/* Switch to H/W */
1837 		if (group == defgrp && ((mrp->mrp_ntxrings > 0) || unspec)) {
1838 			ngrp = 	mac_reserve_tx_group(mcip, B_TRUE);
1839 			if (ngrp == NULL)
1840 				return (ENOSPC);
1841 			mac_tx_client_quiesce((mac_client_handle_t)mcip);
1842 			mac_tx_switch_group(mcip, defgrp, ngrp);
1843 			mac_tx_client_restart((mac_client_handle_t)mcip);
1844 			MAC_TX_GRP_RESERVED(mip);
1845 			if (mip->mi_tx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1846 				MAC_TX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1847 		/* Switch to S/W */
1848 		} else if (group != defgrp && !unspec &&
1849 		    mrp->mrp_ntxrings == 0) {
1850 			/* Switch to S/W */
1851 			ringcnt = group->mrg_cur_count;
1852 			mac_tx_client_quiesce((mac_client_handle_t)mcip);
1853 			mac_tx_switch_group(mcip, group, defgrp);
1854 			mac_tx_client_restart((mac_client_handle_t)mcip);
1855 			if (tmrp->mrp_mask & MRP_TX_RINGS) {
1856 				MAC_TX_GRP_RELEASED(mip);
1857 				if (mip->mi_tx_group_type ==
1858 				    MAC_GROUP_TYPE_DYNAMIC) {
1859 					MAC_TX_RING_RELEASED(mip, ringcnt);
1860 				}
1861 			}
1862 		} else if (group != defgrp && mip->mi_tx_group_type ==
1863 		    MAC_GROUP_TYPE_DYNAMIC) {
1864 			ringcnt = group->mrg_cur_count;
1865 			err = mac_group_ring_modify(mcip, group, defgrp);
1866 			if (err != 0)
1867 				return (err);
1868 			/*
1869 			 * Update the accounting. If this group
1870 			 * already had explicitly reserved rings,
1871 			 * we need to update the rings based on
1872 			 * the new ring count. If this group
1873 			 * had not explicitly reserved rings,
1874 			 * then we just reserve the rings asked for
1875 			 * and reserve the group.
1876 			 */
1877 			if (tmrp->mrp_mask & MRP_TX_RINGS) {
1878 				if (ringcnt > group->mrg_cur_count) {
1879 					MAC_TX_RING_RELEASED(mip,
1880 					    ringcnt - group->mrg_cur_count);
1881 				} else {
1882 					MAC_TX_RING_RESERVED(mip,
1883 					    group->mrg_cur_count - ringcnt);
1884 				}
1885 			} else {
1886 				MAC_TX_RING_RESERVED(mip, group->mrg_cur_count);
1887 				MAC_TX_GRP_RESERVED(mip);
1888 			}
1889 		}
1890 	}
1891 	return (0);
1892 }
1893 
1894 /*
1895  * When the MAC client is being brought up (i.e. we do a unicast_add) we need
1896  * to initialize the cpu and resource control structure in the
1897  * mac_client_impl_t from the mac_impl_t (i.e if there are any cached
1898  * properties before the flow entry for the unicast address was created).
1899  */
1900 int
1901 mac_resource_ctl_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
1902 {
1903 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
1904 	mac_impl_t		*mip = (mac_impl_t *)mcip->mci_mip;
1905 	int			err = 0;
1906 	flow_entry_t		*flent = mcip->mci_flent;
1907 	mac_resource_props_t	*omrp, *nmrp = MCIP_RESOURCE_PROPS(mcip);
1908 
1909 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1910 
1911 	err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
1912 	    mcip->mci_upper_mip : mip, mrp);
1913 	if (err != 0)
1914 		return (err);
1915 
1916 	/*
1917 	 * Copy over the existing properties since mac_update_resources
1918 	 * will modify the client's mrp. Currently, the saved property
1919 	 * is used to determine the difference between existing and
1920 	 * modified rings property.
1921 	 */
1922 	omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
1923 	bcopy(nmrp, omrp, sizeof (*omrp));
1924 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
1925 	if (MCIP_DATAPATH_SETUP(mcip)) {
1926 		/*
1927 		 * We support rings only for primary client when there are
1928 		 * multiple clients sharing the same MAC address (e.g. VLAN).
1929 		 */
1930 		if (mrp->mrp_mask & MRP_RX_RINGS ||
1931 		    mrp->mrp_mask & MRP_TX_RINGS) {
1932 
1933 			if ((err = mac_client_set_rings_prop(mcip, mrp,
1934 			    omrp)) != 0) {
1935 				if (omrp->mrp_mask & MRP_RX_RINGS) {
1936 					nmrp->mrp_mask |= MRP_RX_RINGS;
1937 					nmrp->mrp_nrxrings = omrp->mrp_nrxrings;
1938 				} else {
1939 					nmrp->mrp_mask &= ~MRP_RX_RINGS;
1940 					nmrp->mrp_nrxrings = 0;
1941 				}
1942 				if (omrp->mrp_mask & MRP_TX_RINGS) {
1943 					nmrp->mrp_mask |= MRP_TX_RINGS;
1944 					nmrp->mrp_ntxrings = omrp->mrp_ntxrings;
1945 				} else {
1946 					nmrp->mrp_mask &= ~MRP_TX_RINGS;
1947 					nmrp->mrp_ntxrings = 0;
1948 				}
1949 				if (omrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1950 					omrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
1951 				else
1952 					omrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
1953 
1954 				if (omrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1955 					omrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
1956 				else
1957 					omrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
1958 				kmem_free(omrp, sizeof (*omrp));
1959 				return (err);
1960 			}
1961 
1962 			/*
1963 			 * If we modified the rings property of the primary
1964 			 * we need to update the property fields of its
1965 			 * VLANs as they inherit the primary's properites.
1966 			 */
1967 			if (mac_is_primary_client(mcip)) {
1968 				mac_set_prim_vlan_rings(mip,
1969 				    MCIP_RESOURCE_PROPS(mcip));
1970 			}
1971 		}
1972 		/*
1973 		 * We have to set this prior to calling mac_flow_modify.
1974 		 */
1975 		if (mrp->mrp_mask & MRP_PRIORITY) {
1976 			if (mrp->mrp_priority == MPL_RESET) {
1977 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1978 				    MPL_LINK_DEFAULT);
1979 			} else {
1980 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1981 				    mrp->mrp_priority);
1982 			}
1983 		}
1984 
1985 		mac_flow_modify(mip->mi_flow_tab, flent, mrp);
1986 		if (mrp->mrp_mask & MRP_PRIORITY)
1987 			mac_update_subflow_priority(mcip);
1988 	}
1989 	kmem_free(omrp, sizeof (*omrp));
1990 	return (0);
1991 }
1992 
1993 void
1994 mac_resource_ctl_get(mac_client_handle_t mch, mac_resource_props_t *mrp)
1995 {
1996 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1997 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
1998 
1999 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
2000 }
2001 
2002 static int
2003 mac_unicast_flow_create(mac_client_impl_t *mcip, uint8_t *mac_addr,
2004     uint16_t vid, boolean_t is_primary, boolean_t first_flow,
2005     flow_entry_t **flent, mac_resource_props_t *mrp)
2006 {
2007 	mac_impl_t	*mip = (mac_impl_t *)mcip->mci_mip;
2008 	flow_desc_t	flow_desc;
2009 	char		flowname[MAXFLOWNAMELEN];
2010 	int		err;
2011 	uint_t		flent_flags;
2012 
2013 	/*
2014 	 * First unicast address being added, create a new flow
2015 	 * for that MAC client.
2016 	 */
2017 	bzero(&flow_desc, sizeof (flow_desc));
2018 
2019 	ASSERT(mac_addr != NULL ||
2020 	    (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR));
2021 	if (mac_addr != NULL) {
2022 		flow_desc.fd_mac_len = mip->mi_type->mt_addr_length;
2023 		bcopy(mac_addr, flow_desc.fd_dst_mac, flow_desc.fd_mac_len);
2024 	}
2025 	flow_desc.fd_mask = FLOW_LINK_DST;
2026 	if (vid != 0) {
2027 		flow_desc.fd_vid = vid;
2028 		flow_desc.fd_mask |= FLOW_LINK_VID;
2029 	}
2030 
2031 	/*
2032 	 * XXX-nicolas. For now I'm keeping the FLOW_PRIMARY_MAC
2033 	 * and FLOW_VNIC. Even though they're a hack inherited
2034 	 * from the SRS code, we'll keep them for now. They're currently
2035 	 * consumed by mac_datapath_setup() to create the SRS.
2036 	 * That code should be eventually moved out of
2037 	 * mac_datapath_setup() and moved to a mac_srs_create()
2038 	 * function of some sort to keep things clean.
2039 	 *
2040 	 * Also, there's no reason why the SRS for the primary MAC
2041 	 * client should be different than any other MAC client. Until
2042 	 * this is cleaned-up, we support only one MAC unicast address
2043 	 * per client.
2044 	 *
2045 	 * We set FLOW_PRIMARY_MAC for the primary MAC address,
2046 	 * FLOW_VNIC for everything else.
2047 	 */
2048 	if (is_primary)
2049 		flent_flags = FLOW_PRIMARY_MAC;
2050 	else
2051 		flent_flags = FLOW_VNIC_MAC;
2052 
2053 	/*
2054 	 * For the first flow we use the mac client's name - mci_name, for
2055 	 * subsequent ones we just create a name with the vid. This is
2056 	 * so that we can add these flows to the same flow table. This is
2057 	 * fine as the flow name (except for the one with the mac client's
2058 	 * name) is not visible. When the first flow is removed, we just replace
2059 	 * its fdesc with another from the list, so we will still retain the
2060 	 * flent with the MAC client's flow name.
2061 	 */
2062 	if (first_flow) {
2063 		bcopy(mcip->mci_name, flowname, MAXFLOWNAMELEN);
2064 	} else {
2065 		(void) sprintf(flowname, "%s%u", mcip->mci_name, vid);
2066 		flent_flags = FLOW_NO_STATS;
2067 	}
2068 
2069 	if ((err = mac_flow_create(&flow_desc, mrp, flowname, NULL,
2070 	    flent_flags, flent)) != 0)
2071 		return (err);
2072 
2073 	mac_misc_stat_create(*flent);
2074 	FLOW_MARK(*flent, FE_INCIPIENT);
2075 	(*flent)->fe_mcip = mcip;
2076 
2077 	/*
2078 	 * Place initial creation reference on the flow. This reference
2079 	 * is released in the corresponding delete action viz.
2080 	 * mac_unicast_remove after waiting for all transient refs to
2081 	 * to go away. The wait happens in mac_flow_wait.
2082 	 * We have already held the reference in mac_client_open().
2083 	 */
2084 	if (!first_flow)
2085 		FLOW_REFHOLD(*flent);
2086 	return (0);
2087 }
2088 
2089 /* Refresh the multicast grouping for this VID. */
2090 int
2091 mac_client_update_mcast(void *arg, boolean_t add, const uint8_t *addrp)
2092 {
2093 	flow_entry_t		*flent = arg;
2094 	mac_client_impl_t	*mcip = flent->fe_mcip;
2095 	uint16_t		vid;
2096 	flow_desc_t		flow_desc;
2097 
2098 	mac_flow_get_desc(flent, &flow_desc);
2099 	vid = (flow_desc.fd_mask & FLOW_LINK_VID) != 0 ?
2100 	    flow_desc.fd_vid : VLAN_ID_NONE;
2101 
2102 	/*
2103 	 * We don't call mac_multicast_add()/mac_multicast_remove() as
2104 	 * we want to add/remove for this specific vid.
2105 	 */
2106 	if (add) {
2107 		return (mac_bcast_add(mcip, addrp, vid,
2108 		    MAC_ADDRTYPE_MULTICAST));
2109 	} else {
2110 		mac_bcast_delete(mcip, addrp, vid);
2111 		return (0);
2112 	}
2113 }
2114 
2115 static void
2116 mac_update_single_active_client(mac_impl_t *mip)
2117 {
2118 	mac_client_impl_t *client = NULL;
2119 
2120 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
2121 
2122 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2123 	if (mip->mi_nactiveclients == 1) {
2124 		/*
2125 		 * Find the one active MAC client from the list of MAC
2126 		 * clients. The active MAC client has at least one
2127 		 * unicast address.
2128 		 */
2129 		for (client = mip->mi_clients_list; client != NULL;
2130 		    client = client->mci_client_next) {
2131 			if (client->mci_unicast_list != NULL)
2132 				break;
2133 		}
2134 		ASSERT(client != NULL);
2135 	}
2136 
2137 	/*
2138 	 * mi_single_active_client is protected by the MAC impl's read/writer
2139 	 * lock, which allows mac_rx() to check the value of that pointer
2140 	 * as a reader.
2141 	 */
2142 	mip->mi_single_active_client = client;
2143 	rw_exit(&mip->mi_rw_lock);
2144 }
2145 
2146 /*
2147  * Set up the data path. Called from i_mac_unicast_add after having
2148  * done all the validations including making sure this is an active
2149  * client (i.e that is ready to process packets.)
2150  */
2151 static int
2152 mac_client_datapath_setup(mac_client_impl_t *mcip, uint16_t vid,
2153     uint8_t *mac_addr, mac_resource_props_t *mrp, boolean_t isprimary,
2154     mac_unicast_impl_t *muip)
2155 {
2156 	mac_impl_t	*mip = mcip->mci_mip;
2157 	boolean_t	mac_started = B_FALSE;
2158 	boolean_t	bcast_added = B_FALSE;
2159 	boolean_t	nactiveclients_added = B_FALSE;
2160 	flow_entry_t	*flent;
2161 	int		err = 0;
2162 	boolean_t	no_unicast;
2163 
2164 	no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2165 
2166 	if ((err = mac_start((mac_handle_t)mip)) != 0)
2167 		goto bail;
2168 
2169 	mac_started = B_TRUE;
2170 
2171 	/* add the MAC client to the broadcast address group by default */
2172 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2173 		err = mac_bcast_add(mcip, mip->mi_type->mt_brdcst_addr, vid,
2174 		    MAC_ADDRTYPE_BROADCAST);
2175 		if (err != 0)
2176 			goto bail;
2177 		bcast_added = B_TRUE;
2178 	}
2179 
2180 	/*
2181 	 * If this is the first unicast address addition for this
2182 	 * client, reuse the pre-allocated larval flow entry associated with
2183 	 * the MAC client.
2184 	 */
2185 	flent = (mcip->mci_nflents == 0) ? mcip->mci_flent : NULL;
2186 
2187 	/* We are configuring the unicast flow now */
2188 	if (!MCIP_DATAPATH_SETUP(mcip)) {
2189 
2190 		if (mrp != NULL) {
2191 			MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
2192 			    (mrp->mrp_mask & MRP_PRIORITY) ? mrp->mrp_priority :
2193 			    MPL_LINK_DEFAULT);
2194 		}
2195 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2196 		    isprimary, B_TRUE, &flent, mrp)) != 0)
2197 			goto bail;
2198 
2199 		mip->mi_nactiveclients++;
2200 		nactiveclients_added = B_TRUE;
2201 
2202 		/*
2203 		 * This will allocate the RX ring group if possible for the
2204 		 * flow and program the software classifier as needed.
2205 		 */
2206 		if ((err = mac_datapath_setup(mcip, flent, SRST_LINK)) != 0)
2207 			goto bail;
2208 
2209 		if (no_unicast)
2210 			goto done_setup;
2211 		/*
2212 		 * The unicast MAC address must have been added successfully.
2213 		 */
2214 		ASSERT(mcip->mci_unicast != NULL);
2215 		/*
2216 		 * Push down the sub-flows that were defined on this link
2217 		 * hitherto. The flows are added to the active flow table
2218 		 * and SRS, softrings etc. are created as needed.
2219 		 */
2220 		mac_link_init_flows((mac_client_handle_t)mcip);
2221 	} else {
2222 		mac_address_t *map = mcip->mci_unicast;
2223 
2224 		ASSERT(!no_unicast);
2225 		/*
2226 		 * A unicast flow already exists for that MAC client,
2227 		 * this flow must be the same mac address but with
2228 		 * different VID. It has been checked by mac_addr_in_use().
2229 		 *
2230 		 * We will use the SRS etc. from the mci_flent. Note that
2231 		 * We don't need to create kstat for this as except for
2232 		 * the fdesc, everything will be used from in the 1st flent.
2233 		 */
2234 
2235 		if (bcmp(mac_addr, map->ma_addr, map->ma_len) != 0) {
2236 			err = EINVAL;
2237 			goto bail;
2238 		}
2239 
2240 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2241 		    isprimary, B_FALSE, &flent, NULL)) != 0) {
2242 			goto bail;
2243 		}
2244 		if ((err = mac_flow_add(mip->mi_flow_tab, flent)) != 0) {
2245 			FLOW_FINAL_REFRELE(flent);
2246 			goto bail;
2247 		}
2248 
2249 		/* update the multicast group for this vid */
2250 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2251 		    (void *)flent, B_TRUE);
2252 
2253 	}
2254 
2255 	/* populate the shared MAC address */
2256 	muip->mui_map = mcip->mci_unicast;
2257 
2258 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2259 	muip->mui_next = mcip->mci_unicast_list;
2260 	mcip->mci_unicast_list = muip;
2261 	rw_exit(&mcip->mci_rw_lock);
2262 
2263 done_setup:
2264 	/*
2265 	 * First add the flent to the flow list of this mcip. Then set
2266 	 * the mip's mi_single_active_client if needed. The Rx path assumes
2267 	 * that mip->mi_single_active_client will always have an associated
2268 	 * flent.
2269 	 */
2270 	mac_client_add_to_flow_list(mcip, flent);
2271 	if (nactiveclients_added)
2272 		mac_update_single_active_client(mip);
2273 	/*
2274 	 * Trigger a renegotiation of the capabilities when the number of
2275 	 * active clients changes from 1 to 2, since some of the capabilities
2276 	 * might have to be disabled. Also send a MAC_NOTE_LINK notification
2277 	 * to all the MAC clients whenever physical link is DOWN.
2278 	 */
2279 	if (mip->mi_nactiveclients == 2) {
2280 		mac_capab_update((mac_handle_t)mip);
2281 		mac_virtual_link_update(mip);
2282 	}
2283 	/*
2284 	 * Now that the setup is complete, clear the INCIPIENT flag.
2285 	 * The flag was set to avoid incoming packets seeing inconsistent
2286 	 * structures while the setup was in progress. Clear the mci_tx_flag
2287 	 * by calling mac_tx_client_block. It is possible that
2288 	 * mac_unicast_remove was called prior to this mac_unicast_add which
2289 	 * could have set the MCI_TX_QUIESCE flag.
2290 	 */
2291 	if (flent->fe_rx_ring_group != NULL)
2292 		mac_rx_group_unmark(flent->fe_rx_ring_group, MR_INCIPIENT);
2293 	FLOW_UNMARK(flent, FE_INCIPIENT);
2294 	FLOW_UNMARK(flent, FE_MC_NO_DATAPATH);
2295 	mac_tx_client_unblock(mcip);
2296 	return (0);
2297 bail:
2298 	if (bcast_added)
2299 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr, vid);
2300 
2301 	if (nactiveclients_added)
2302 		mip->mi_nactiveclients--;
2303 
2304 	if (mac_started)
2305 		mac_stop((mac_handle_t)mip);
2306 
2307 	return (err);
2308 }
2309 
2310 /*
2311  * Return the passive primary MAC client, if present. The passive client is
2312  * a stand-by client that has the same unicast address as another that is
2313  * currenly active. Once the active client goes away, the passive client
2314  * becomes active.
2315  */
2316 static mac_client_impl_t *
2317 mac_get_passive_primary_client(mac_impl_t *mip)
2318 {
2319 	mac_client_impl_t	*mcip;
2320 
2321 	for (mcip = mip->mi_clients_list; mcip != NULL;
2322 	    mcip = mcip->mci_client_next) {
2323 		if (mac_is_primary_client(mcip) &&
2324 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2325 			return (mcip);
2326 		}
2327 	}
2328 	return (NULL);
2329 }
2330 
2331 /*
2332  * Add a new unicast address to the MAC client.
2333  *
2334  * The MAC address can be specified either by value, or the MAC client
2335  * can specify that it wants to use the primary MAC address of the
2336  * underlying MAC. See the introductory comments at the beginning
2337  * of this file for more more information on primary MAC addresses.
2338  *
2339  * Note also the tuple (MAC address, VID) must be unique
2340  * for the MAC clients defined on top of the same underlying MAC
2341  * instance, unless the MAC_UNICAST_NODUPCHECK is specified.
2342  *
2343  * In no case can a client use the PVID for the MAC, if the MAC has one set.
2344  */
2345 int
2346 i_mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2347     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2348 {
2349 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2350 	mac_impl_t		*mip = mcip->mci_mip;
2351 	int			err;
2352 	uint_t			mac_len = mip->mi_type->mt_addr_length;
2353 	boolean_t		check_dups = !(flags & MAC_UNICAST_NODUPCHECK);
2354 	boolean_t		fastpath_disabled = B_FALSE;
2355 	boolean_t		is_primary = (flags & MAC_UNICAST_PRIMARY);
2356 	boolean_t		is_unicast_hw = (flags & MAC_UNICAST_HW);
2357 	mac_resource_props_t	*mrp;
2358 	boolean_t		passive_client = B_FALSE;
2359 	mac_unicast_impl_t	*muip;
2360 	boolean_t		is_vnic_primary =
2361 	    (flags & MAC_UNICAST_VNIC_PRIMARY);
2362 
2363 	/* when VID is non-zero, the underlying MAC can not be VNIC */
2364 	ASSERT(!((mip->mi_state_flags & MIS_IS_VNIC) && (vid != 0)));
2365 
2366 	/*
2367 	 * Can't unicast add if the client asked only for minimal datapath
2368 	 * setup.
2369 	 */
2370 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR)
2371 		return (ENOTSUP);
2372 
2373 	/*
2374 	 * Check for an attempted use of the current Port VLAN ID, if enabled.
2375 	 * No client may use it.
2376 	 */
2377 	if (mip->mi_pvid != 0 && vid == mip->mi_pvid)
2378 		return (EBUSY);
2379 
2380 	/*
2381 	 * Check whether it's the primary client and flag it.
2382 	 */
2383 	if (!(mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary && vid == 0)
2384 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PRIMARY;
2385 
2386 	/*
2387 	 * is_vnic_primary is true when we come here as a VLAN VNIC
2388 	 * which uses the primary mac client's address but with a non-zero
2389 	 * VID. In this case the MAC address is not specified by an upper
2390 	 * MAC client.
2391 	 */
2392 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary &&
2393 	    !is_vnic_primary) {
2394 		/*
2395 		 * The address is being set by the upper MAC client
2396 		 * of a VNIC. The MAC address was already set by the
2397 		 * VNIC driver during VNIC creation.
2398 		 *
2399 		 * Note: a VNIC has only one MAC address. We return
2400 		 * the MAC unicast address handle of the lower MAC client
2401 		 * corresponding to the VNIC. We allocate a new entry
2402 		 * which is flagged appropriately, so that mac_unicast_remove()
2403 		 * doesn't attempt to free the original entry that
2404 		 * was allocated by the VNIC driver.
2405 		 */
2406 		ASSERT(mcip->mci_unicast != NULL);
2407 
2408 		/* Check for VLAN flags, if present */
2409 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2410 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2411 
2412 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2413 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2414 
2415 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2416 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2417 
2418 		/*
2419 		 * Ensure that the primary unicast address of the VNIC
2420 		 * is added only once unless we have the
2421 		 * MAC_CLIENT_FLAGS_MULTI_PRIMARY set (and this is not
2422 		 * a passive MAC client).
2423 		 */
2424 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) != 0) {
2425 			if ((mcip->mci_flags &
2426 			    MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2427 			    (mcip->mci_flags &
2428 			    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2429 				return (EBUSY);
2430 			}
2431 			mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2432 			passive_client = B_TRUE;
2433 		}
2434 
2435 		mcip->mci_flags |= MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2436 
2437 		/*
2438 		 * Create a handle for vid 0.
2439 		 */
2440 		ASSERT(vid == 0);
2441 		muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2442 		muip->mui_vid = vid;
2443 		*mah = (mac_unicast_handle_t)muip;
2444 		/*
2445 		 * This will be used by the caller to defer setting the
2446 		 * rx functions.
2447 		 */
2448 		if (passive_client)
2449 			return (EAGAIN);
2450 		return (0);
2451 	}
2452 
2453 	/* primary MAC clients cannot be opened on top of anchor VNICs */
2454 	if ((is_vnic_primary || is_primary) &&
2455 	    i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_ANCHOR_VNIC, NULL)) {
2456 		return (ENXIO);
2457 	}
2458 
2459 	/*
2460 	 * If this is a VNIC/VLAN, disable softmac fast-path.
2461 	 */
2462 	if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2463 		err = mac_fastpath_disable((mac_handle_t)mip);
2464 		if (err != 0)
2465 			return (err);
2466 		fastpath_disabled = B_TRUE;
2467 	}
2468 
2469 	/*
2470 	 * Return EBUSY if:
2471 	 *  - there is an exclusively active mac client exists.
2472 	 *  - this is an exclusive active mac client but
2473 	 *	a. there is already active mac clients exist, or
2474 	 *	b. fastpath streams are already plumbed on this legacy device
2475 	 *  - the mac creator has disallowed active mac clients.
2476 	 */
2477 	if (mip->mi_state_flags & (MIS_EXCLUSIVE|MIS_NO_ACTIVE)) {
2478 		if (fastpath_disabled)
2479 			mac_fastpath_enable((mac_handle_t)mip);
2480 		return (EBUSY);
2481 	}
2482 
2483 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2484 		ASSERT(!fastpath_disabled);
2485 		if (mip->mi_nactiveclients != 0)
2486 			return (EBUSY);
2487 
2488 		if ((mip->mi_state_flags & MIS_LEGACY) &&
2489 		    !(mip->mi_capab_legacy.ml_active_set(mip->mi_driver))) {
2490 			return (EBUSY);
2491 		}
2492 		mip->mi_state_flags |= MIS_EXCLUSIVE;
2493 	}
2494 
2495 	mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
2496 	if (is_primary && !(mcip->mci_state_flags & (MCIS_IS_VNIC |
2497 	    MCIS_IS_AGGR_PORT))) {
2498 		/*
2499 		 * Apply the property cached in the mac_impl_t to the primary
2500 		 * mac client. If the mac client is a VNIC or an aggregation
2501 		 * port, its property should be set in the mcip when the
2502 		 * VNIC/aggr was created.
2503 		 */
2504 		mac_get_resources((mac_handle_t)mip, mrp);
2505 		(void) mac_client_set_resources(mch, mrp);
2506 	} else if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2507 		/*
2508 		 * This is a primary VLAN client, we don't support
2509 		 * specifying rings property for this as it inherits the
2510 		 * rings property from its MAC.
2511 		 */
2512 		if (is_vnic_primary) {
2513 			mac_resource_props_t	*vmrp;
2514 
2515 			vmrp = MCIP_RESOURCE_PROPS(mcip);
2516 			if (vmrp->mrp_mask & MRP_RX_RINGS ||
2517 			    vmrp->mrp_mask & MRP_TX_RINGS) {
2518 				if (fastpath_disabled)
2519 					mac_fastpath_enable((mac_handle_t)mip);
2520 				kmem_free(mrp, sizeof (*mrp));
2521 				return (ENOTSUP);
2522 			}
2523 			/*
2524 			 * Additionally we also need to inherit any
2525 			 * rings property from the MAC.
2526 			 */
2527 			mac_get_resources((mac_handle_t)mip, mrp);
2528 			if (mrp->mrp_mask & MRP_RX_RINGS) {
2529 				vmrp->mrp_mask |= MRP_RX_RINGS;
2530 				vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
2531 			}
2532 			if (mrp->mrp_mask & MRP_TX_RINGS) {
2533 				vmrp->mrp_mask |= MRP_TX_RINGS;
2534 				vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
2535 			}
2536 		}
2537 		bcopy(MCIP_RESOURCE_PROPS(mcip), mrp, sizeof (*mrp));
2538 	}
2539 
2540 	muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2541 	muip->mui_vid = vid;
2542 
2543 	if (is_primary || is_vnic_primary) {
2544 		mac_addr = mip->mi_addr;
2545 	} else {
2546 
2547 		/*
2548 		 * Verify the validity of the specified MAC addresses value.
2549 		 */
2550 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr, mac_len)) {
2551 			*diag = MAC_DIAG_MACADDR_INVALID;
2552 			err = EINVAL;
2553 			goto bail_out;
2554 		}
2555 
2556 		/*
2557 		 * Make sure that the specified MAC address is different
2558 		 * than the unicast MAC address of the underlying NIC.
2559 		 */
2560 		if (check_dups && bcmp(mip->mi_addr, mac_addr, mac_len) == 0) {
2561 			*diag = MAC_DIAG_MACADDR_NIC;
2562 			err = EINVAL;
2563 			goto bail_out;
2564 		}
2565 	}
2566 
2567 	/*
2568 	 * Set the flags here so that if this is a passive client, we
2569 	 * can return  and set it when we call mac_client_datapath_setup
2570 	 * when this becomes the active client. If we defer to using these
2571 	 * flags to mac_client_datapath_setup, then for a passive client,
2572 	 * we'd have to store the flags somewhere (probably fe_flags)
2573 	 * and then use it.
2574 	 */
2575 	if (!MCIP_DATAPATH_SETUP(mcip)) {
2576 		if (is_unicast_hw) {
2577 			/*
2578 			 * The client requires a hardware MAC address slot
2579 			 * for that unicast address. Since we support only
2580 			 * one unicast MAC address per client, flag the
2581 			 * MAC client itself.
2582 			 */
2583 			mcip->mci_state_flags |= MCIS_UNICAST_HW;
2584 		}
2585 
2586 		/* Check for VLAN flags, if present */
2587 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2588 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2589 
2590 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2591 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2592 
2593 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2594 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2595 	} else {
2596 		/*
2597 		 * Assert that the specified flags are consistent with the
2598 		 * flags specified by previous calls to mac_unicast_add().
2599 		 */
2600 		ASSERT(((flags & MAC_UNICAST_TAG_DISABLE) != 0 &&
2601 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) != 0) ||
2602 		    ((flags & MAC_UNICAST_TAG_DISABLE) == 0 &&
2603 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) == 0));
2604 
2605 		ASSERT(((flags & MAC_UNICAST_STRIP_DISABLE) != 0 &&
2606 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) != 0) ||
2607 		    ((flags & MAC_UNICAST_STRIP_DISABLE) == 0 &&
2608 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) == 0));
2609 
2610 		ASSERT(((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0 &&
2611 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) != 0) ||
2612 		    ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) == 0 &&
2613 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) == 0));
2614 
2615 		/*
2616 		 * Make sure the client is consistent about its requests
2617 		 * for MAC addresses. I.e. all requests from the clients
2618 		 * must have the MAC_UNICAST_HW flag set or clear.
2619 		 */
2620 		if ((mcip->mci_state_flags & MCIS_UNICAST_HW) != 0 &&
2621 		    !is_unicast_hw ||
2622 		    (mcip->mci_state_flags & MCIS_UNICAST_HW) == 0 &&
2623 		    is_unicast_hw) {
2624 			err = EINVAL;
2625 			goto bail_out;
2626 		}
2627 	}
2628 	/*
2629 	 * Make sure the MAC address is not already used by
2630 	 * another MAC client defined on top of the same
2631 	 * underlying NIC. Unless we have MAC_CLIENT_FLAGS_MULTI_PRIMARY
2632 	 * set when we allow a passive client to be present which will
2633 	 * be activated when the currently active client goes away - this
2634 	 * works only with primary addresses.
2635 	 */
2636 	if ((check_dups || is_primary || is_vnic_primary) &&
2637 	    mac_addr_in_use(mip, mac_addr, vid)) {
2638 		/*
2639 		 * Must have set the multiple primary address flag when
2640 		 * we did a mac_client_open AND this should be a primary
2641 		 * MAC client AND there should not already be a passive
2642 		 * primary. If all is true then we let this succeed
2643 		 * even if the address is a dup.
2644 		 */
2645 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2646 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) == 0 ||
2647 		    mac_get_passive_primary_client(mip) != NULL) {
2648 			*diag = MAC_DIAG_MACADDR_INUSE;
2649 			err = EEXIST;
2650 			goto bail_out;
2651 		}
2652 		ASSERT((mcip->mci_flags &
2653 		    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) == 0);
2654 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2655 		kmem_free(mrp, sizeof (*mrp));
2656 
2657 		/*
2658 		 * Stash the unicast address handle, we will use it when
2659 		 * we set up the passive client.
2660 		 */
2661 		mcip->mci_p_unicast_list = muip;
2662 		*mah = (mac_unicast_handle_t)muip;
2663 		return (0);
2664 	}
2665 
2666 	err = mac_client_datapath_setup(mcip, vid, mac_addr, mrp,
2667 	    is_primary || is_vnic_primary, muip);
2668 	if (err != 0)
2669 		goto bail_out;
2670 
2671 	kmem_free(mrp, sizeof (*mrp));
2672 	*mah = (mac_unicast_handle_t)muip;
2673 	return (0);
2674 
2675 bail_out:
2676 	if (fastpath_disabled)
2677 		mac_fastpath_enable((mac_handle_t)mip);
2678 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2679 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2680 		if (mip->mi_state_flags & MIS_LEGACY) {
2681 			mip->mi_capab_legacy.ml_active_clear(
2682 			    mip->mi_driver);
2683 		}
2684 	}
2685 	kmem_free(mrp, sizeof (*mrp));
2686 	kmem_free(muip, sizeof (mac_unicast_impl_t));
2687 	return (err);
2688 }
2689 
2690 /*
2691  * Wrapper function to mac_unicast_add when we want to have the same mac
2692  * client open for two instances, one that is currently active and another
2693  * that will become active when the current one is removed. In this case
2694  * mac_unicast_add will return EGAIN and we will save the rx function and
2695  * arg which will be used when we activate the passive client in
2696  * mac_unicast_remove.
2697  */
2698 int
2699 mac_unicast_add_set_rx(mac_client_handle_t mch, uint8_t *mac_addr,
2700     uint16_t flags, mac_unicast_handle_t *mah,  uint16_t vid, mac_diag_t *diag,
2701     mac_rx_t rx_fn, void *arg)
2702 {
2703 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2704 	uint_t			err;
2705 
2706 	err = mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2707 	if (err != 0 && err != EAGAIN)
2708 		return (err);
2709 	if (err == EAGAIN) {
2710 		if (rx_fn != NULL) {
2711 			mcip->mci_rx_p_fn = rx_fn;
2712 			mcip->mci_rx_p_arg = arg;
2713 		}
2714 		return (0);
2715 	}
2716 	if (rx_fn != NULL)
2717 		mac_rx_set(mch, rx_fn, arg);
2718 	return (err);
2719 }
2720 
2721 int
2722 mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2723     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2724 {
2725 	mac_impl_t *mip = ((mac_client_impl_t *)mch)->mci_mip;
2726 	uint_t err;
2727 
2728 	i_mac_perim_enter(mip);
2729 	err = i_mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2730 	i_mac_perim_exit(mip);
2731 
2732 	return (err);
2733 }
2734 
2735 static void
2736 mac_client_datapath_teardown(mac_client_handle_t mch, mac_unicast_impl_t *muip,
2737     flow_entry_t *flent)
2738 {
2739 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2740 	mac_impl_t		*mip = mcip->mci_mip;
2741 	boolean_t		no_unicast;
2742 
2743 	/*
2744 	 * If we have not added a unicast address for this MAC client, just
2745 	 * teardown the datapath.
2746 	 */
2747 	no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2748 
2749 	if (!no_unicast) {
2750 		/*
2751 		 * We would have initialized subflows etc. only if we brought
2752 		 * up the primary client and set the unicast unicast address
2753 		 * etc. Deactivate the flows. The flow entry will be removed
2754 		 * from the active flow tables, and the associated SRS,
2755 		 * softrings etc will be deleted. But the flow entry itself
2756 		 * won't be destroyed, instead it will continue to be archived
2757 		 * off the  the global flow hash list, for a possible future
2758 		 * activation when say IP is plumbed again.
2759 		 */
2760 		mac_link_release_flows(mch);
2761 	}
2762 	mip->mi_nactiveclients--;
2763 	mac_update_single_active_client(mip);
2764 
2765 	/* Tear down the data path */
2766 	mac_datapath_teardown(mcip, mcip->mci_flent, SRST_LINK);
2767 
2768 	/*
2769 	 * Prevent any future access to the flow entry through the mci_flent
2770 	 * pointer by setting the mci_flent to NULL. Access to mci_flent in
2771 	 * mac_bcast_send is also under mi_rw_lock.
2772 	 */
2773 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2774 	flent = mcip->mci_flent;
2775 	mac_client_remove_flow_from_list(mcip, flent);
2776 
2777 	if (mcip->mci_state_flags & MCIS_DESC_LOGGED)
2778 		mcip->mci_state_flags &= ~MCIS_DESC_LOGGED;
2779 
2780 	/*
2781 	 * This is the last unicast address being removed and there shouldn't
2782 	 * be any outbound data threads at this point coming down from mac
2783 	 * clients. We have waited for the data threads to finish before
2784 	 * starting dld_str_detach. Non-data threads must access TX SRS
2785 	 * under mi_rw_lock.
2786 	 */
2787 	rw_exit(&mip->mi_rw_lock);
2788 
2789 	/*
2790 	 * Don't use FLOW_MARK with FE_MC_NO_DATAPATH, as the flow might
2791 	 * contain other flags, such as FE_CONDEMNED, which we need to
2792 	 * cleared. We don't call mac_flow_cleanup() for this unicast
2793 	 * flow as we have a already cleaned up SRSs etc. (via the teadown
2794 	 * path). We just clear the stats and reset the initial callback
2795 	 * function, the rest will be set when we call mac_flow_create,
2796 	 * if at all.
2797 	 */
2798 	mutex_enter(&flent->fe_lock);
2799 	ASSERT(flent->fe_refcnt == 1 && flent->fe_mbg == NULL &&
2800 	    flent->fe_tx_srs == NULL && flent->fe_rx_srs_cnt == 0);
2801 	flent->fe_flags = FE_MC_NO_DATAPATH;
2802 	flow_stat_destroy(flent);
2803 	mac_misc_stat_delete(flent);
2804 
2805 	/* Initialize the receiver function to a safe routine */
2806 	flent->fe_cb_fn = (flow_fn_t)mac_pkt_drop;
2807 	flent->fe_cb_arg1 = NULL;
2808 	flent->fe_cb_arg2 = NULL;
2809 
2810 	flent->fe_index = -1;
2811 	mutex_exit(&flent->fe_lock);
2812 
2813 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2814 		ASSERT(muip != NULL || no_unicast);
2815 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2816 		    muip != NULL ? muip->mui_vid : VLAN_ID_NONE);
2817 	}
2818 
2819 	if (mip->mi_nactiveclients == 1) {
2820 		mac_capab_update((mac_handle_t)mip);
2821 		mac_virtual_link_update(mip);
2822 	}
2823 
2824 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2825 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2826 
2827 		if (mip->mi_state_flags & MIS_LEGACY)
2828 			mip->mi_capab_legacy.ml_active_clear(mip->mi_driver);
2829 	}
2830 
2831 	mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2832 
2833 	if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2834 		mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2835 
2836 	if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2837 		mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2838 
2839 	if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2840 		mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2841 
2842 	if (muip != NULL)
2843 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2844 	mac_protect_cancel_timer(mcip);
2845 	mac_protect_flush_dhcp(mcip);
2846 
2847 	bzero(&mcip->mci_misc_stat, sizeof (mcip->mci_misc_stat));
2848 	/*
2849 	 * Disable fastpath if this is a VNIC or a VLAN.
2850 	 */
2851 	if (mcip->mci_state_flags & MCIS_IS_VNIC)
2852 		mac_fastpath_enable((mac_handle_t)mip);
2853 	mac_stop((mac_handle_t)mip);
2854 }
2855 
2856 /*
2857  * Remove a MAC address which was previously added by mac_unicast_add().
2858  */
2859 int
2860 mac_unicast_remove(mac_client_handle_t mch, mac_unicast_handle_t mah)
2861 {
2862 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2863 	mac_unicast_impl_t *muip = (mac_unicast_impl_t *)mah;
2864 	mac_unicast_impl_t *pre;
2865 	mac_impl_t *mip = mcip->mci_mip;
2866 	flow_entry_t		*flent;
2867 	uint16_t mui_vid;
2868 
2869 	i_mac_perim_enter(mip);
2870 	if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) {
2871 		/*
2872 		 * Called made by the upper MAC client of a VNIC.
2873 		 * There's nothing much to do, the unicast address will
2874 		 * be removed by the VNIC driver when the VNIC is deleted,
2875 		 * but let's ensure that all our transmit is done before
2876 		 * the client does a mac_client_stop lest it trigger an
2877 		 * assert in the driver.
2878 		 */
2879 		ASSERT(muip->mui_vid == 0);
2880 
2881 		mac_tx_client_flush(mcip);
2882 
2883 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2884 			mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2885 			if (mcip->mci_rx_p_fn != NULL) {
2886 				mac_rx_set(mch, mcip->mci_rx_p_fn,
2887 				    mcip->mci_rx_p_arg);
2888 				mcip->mci_rx_p_fn = NULL;
2889 				mcip->mci_rx_p_arg = NULL;
2890 			}
2891 			kmem_free(muip, sizeof (mac_unicast_impl_t));
2892 			i_mac_perim_exit(mip);
2893 			return (0);
2894 		}
2895 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2896 
2897 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2898 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2899 
2900 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2901 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2902 
2903 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2904 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2905 
2906 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2907 		i_mac_perim_exit(mip);
2908 		return (0);
2909 	}
2910 
2911 	ASSERT(muip != NULL);
2912 
2913 	/*
2914 	 * We are removing a passive client, we haven't setup the datapath
2915 	 * for this yet, so nothing much to do.
2916 	 */
2917 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2918 
2919 		ASSERT((mcip->mci_flent->fe_flags & FE_MC_NO_DATAPATH) != 0);
2920 		ASSERT(mcip->mci_p_unicast_list == muip);
2921 
2922 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2923 
2924 		mcip->mci_p_unicast_list = NULL;
2925 		mcip->mci_rx_p_fn = NULL;
2926 		mcip->mci_rx_p_arg = NULL;
2927 
2928 		mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2929 
2930 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2931 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2932 
2933 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2934 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2935 
2936 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2937 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2938 
2939 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2940 		i_mac_perim_exit(mip);
2941 		return (0);
2942 	}
2943 	/*
2944 	 * Remove the VID from the list of client's VIDs.
2945 	 */
2946 	pre = mcip->mci_unicast_list;
2947 	if (muip == pre) {
2948 		mcip->mci_unicast_list = muip->mui_next;
2949 	} else {
2950 		while ((pre->mui_next != NULL) && (pre->mui_next != muip))
2951 			pre = pre->mui_next;
2952 		ASSERT(pre->mui_next == muip);
2953 		rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2954 		pre->mui_next = muip->mui_next;
2955 		rw_exit(&mcip->mci_rw_lock);
2956 	}
2957 
2958 	if (!mac_client_single_rcvr(mcip)) {
2959 		/*
2960 		 * This MAC client is shared by more than one unicast
2961 		 * addresses, so we will just remove the flent
2962 		 * corresponding to the address being removed. We don't invoke
2963 		 * mac_rx_classify_flow_rem() since the additional flow is
2964 		 * not associated with its own separate set of SRS and rings,
2965 		 * and these constructs are still needed for the remaining
2966 		 * flows.
2967 		 */
2968 		flent = mac_client_get_flow(mcip, muip);
2969 		ASSERT(flent != NULL);
2970 
2971 		/*
2972 		 * The first one is disappearing, need to make sure
2973 		 * we replace it with another from the list of
2974 		 * shared clients.
2975 		 */
2976 		if (flent == mcip->mci_flent)
2977 			flent = mac_client_swap_mciflent(mcip);
2978 		mac_client_remove_flow_from_list(mcip, flent);
2979 		mac_flow_remove(mip->mi_flow_tab, flent, B_FALSE);
2980 		mac_flow_wait(flent, FLOW_DRIVER_UPCALL);
2981 
2982 		/*
2983 		 * The multicast groups that were added by the client so
2984 		 * far must be removed from the brodcast domain corresponding
2985 		 * to the VID being removed.
2986 		 */
2987 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2988 		    (void *)flent, B_FALSE);
2989 
2990 		if (mip->mi_type->mt_brdcst_addr != NULL) {
2991 			mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2992 			    muip->mui_vid);
2993 		}
2994 
2995 		FLOW_FINAL_REFRELE(flent);
2996 		ASSERT(!(mcip->mci_state_flags & MCIS_EXCLUSIVE));
2997 		/*
2998 		 * Enable fastpath if this is a VNIC or a VLAN.
2999 		 */
3000 		if (mcip->mci_state_flags & MCIS_IS_VNIC)
3001 			mac_fastpath_enable((mac_handle_t)mip);
3002 		mac_stop((mac_handle_t)mip);
3003 		i_mac_perim_exit(mip);
3004 		return (0);
3005 	}
3006 
3007 	mui_vid = muip->mui_vid;
3008 	mac_client_datapath_teardown(mch, muip, flent);
3009 
3010 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) && mui_vid == 0) {
3011 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PRIMARY;
3012 	} else {
3013 		i_mac_perim_exit(mip);
3014 		return (0);
3015 	}
3016 
3017 	/*
3018 	 * If we are removing the primary, check if we have a passive primary
3019 	 * client that we need to activate now.
3020 	 */
3021 	mcip = mac_get_passive_primary_client(mip);
3022 	if (mcip != NULL) {
3023 		mac_resource_props_t	*mrp;
3024 		mac_unicast_impl_t	*muip;
3025 
3026 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
3027 		mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
3028 
3029 		/*
3030 		 * Apply the property cached in the mac_impl_t to the
3031 		 * primary mac client.
3032 		 */
3033 		mac_get_resources((mac_handle_t)mip, mrp);
3034 		(void) mac_client_set_resources(mch, mrp);
3035 		ASSERT(mcip->mci_p_unicast_list != NULL);
3036 		muip = mcip->mci_p_unicast_list;
3037 		mcip->mci_p_unicast_list = NULL;
3038 		if (mac_client_datapath_setup(mcip, VLAN_ID_NONE,
3039 		    mip->mi_addr, mrp, B_TRUE, muip) == 0) {
3040 			if (mcip->mci_rx_p_fn != NULL) {
3041 				mac_rx_set(mch, mcip->mci_rx_p_fn,
3042 				    mcip->mci_rx_p_arg);
3043 				mcip->mci_rx_p_fn = NULL;
3044 				mcip->mci_rx_p_arg = NULL;
3045 			}
3046 		} else {
3047 			kmem_free(muip, sizeof (mac_unicast_impl_t));
3048 		}
3049 		kmem_free(mrp, sizeof (*mrp));
3050 	}
3051 	i_mac_perim_exit(mip);
3052 	return (0);
3053 }
3054 
3055 /*
3056  * Multicast add function invoked by MAC clients.
3057  */
3058 int
3059 mac_multicast_add(mac_client_handle_t mch, const uint8_t *addr)
3060 {
3061 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3062 	mac_impl_t		*mip = mcip->mci_mip;
3063 	flow_entry_t		*flent = mcip->mci_flent_list;
3064 	flow_entry_t		*prev_fe = NULL;
3065 	uint16_t		vid;
3066 	int			err = 0;
3067 
3068 	/* Verify the address is a valid multicast address */
3069 	if ((err = mip->mi_type->mt_ops.mtops_multicst_verify(addr,
3070 	    mip->mi_pdata)) != 0)
3071 		return (err);
3072 
3073 	i_mac_perim_enter(mip);
3074 	while (flent != NULL) {
3075 		vid = i_mac_flow_vid(flent);
3076 
3077 		err = mac_bcast_add((mac_client_impl_t *)mch, addr, vid,
3078 		    MAC_ADDRTYPE_MULTICAST);
3079 		if (err != 0)
3080 			break;
3081 		prev_fe = flent;
3082 		flent = flent->fe_client_next;
3083 	}
3084 
3085 	/*
3086 	 * If we failed adding, then undo all, rather than partial
3087 	 * success.
3088 	 */
3089 	if (flent != NULL && prev_fe != NULL) {
3090 		flent = mcip->mci_flent_list;
3091 		while (flent != prev_fe->fe_client_next) {
3092 			vid = i_mac_flow_vid(flent);
3093 			mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3094 			flent = flent->fe_client_next;
3095 		}
3096 	}
3097 	i_mac_perim_exit(mip);
3098 	return (err);
3099 }
3100 
3101 /*
3102  * Multicast delete function invoked by MAC clients.
3103  */
3104 void
3105 mac_multicast_remove(mac_client_handle_t mch, const uint8_t *addr)
3106 {
3107 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3108 	mac_impl_t		*mip = mcip->mci_mip;
3109 	flow_entry_t		*flent;
3110 	uint16_t		vid;
3111 
3112 	i_mac_perim_enter(mip);
3113 	for (flent = mcip->mci_flent_list; flent != NULL;
3114 	    flent = flent->fe_client_next) {
3115 		vid = i_mac_flow_vid(flent);
3116 		mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3117 	}
3118 	i_mac_perim_exit(mip);
3119 }
3120 
3121 /*
3122  * When a MAC client desires to capture packets on an interface,
3123  * it registers a promiscuous call back with mac_promisc_add().
3124  * There are three types of promiscuous callbacks:
3125  *
3126  * * MAC_CLIENT_PROMISC_ALL
3127  *   Captures all packets sent and received by the MAC client,
3128  *   the physical interface, as well as all other MAC clients
3129  *   defined on top of the same MAC.
3130  *
3131  * * MAC_CLIENT_PROMISC_FILTERED
3132  *   Captures all packets sent and received by the MAC client,
3133  *   plus all multicast traffic sent and received by the phyisical
3134  *   interface and the other MAC clients.
3135  *
3136  * * MAC_CLIENT_PROMISC_MULTI
3137  *   Captures all broadcast and multicast packets sent and
3138  *   received by the MAC clients as well as the physical interface.
3139  *
3140  * In all cases, the underlying MAC is put in promiscuous mode.
3141  */
3142 int
3143 mac_promisc_add(mac_client_handle_t mch, mac_client_promisc_type_t type,
3144     mac_rx_t fn, void *arg, mac_promisc_handle_t *mphp, uint16_t flags)
3145 {
3146 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3147 	mac_impl_t *mip = mcip->mci_mip;
3148 	mac_promisc_impl_t *mpip;
3149 	mac_cb_info_t	*mcbi;
3150 	int rc;
3151 
3152 	i_mac_perim_enter(mip);
3153 
3154 	if ((rc = mac_start((mac_handle_t)mip)) != 0) {
3155 		i_mac_perim_exit(mip);
3156 		return (rc);
3157 	}
3158 
3159 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
3160 	    type == MAC_CLIENT_PROMISC_ALL) {
3161 		/*
3162 		 * The function is being invoked by the upper MAC client
3163 		 * of a VNIC. The VNIC should only see the traffic
3164 		 * it is entitled to.
3165 		 */
3166 		type = MAC_CLIENT_PROMISC_FILTERED;
3167 	}
3168 
3169 
3170 	/*
3171 	 * Turn on promiscuous mode for the underlying NIC.
3172 	 * This is needed even for filtered callbacks which
3173 	 * expect to receive all multicast traffic on the wire.
3174 	 *
3175 	 * Physical promiscuous mode should not be turned on if
3176 	 * MAC_PROMISC_FLAGS_NO_PHYS is set.
3177 	 */
3178 	if ((flags & MAC_PROMISC_FLAGS_NO_PHYS) == 0) {
3179 		if ((rc = i_mac_promisc_set(mip, B_TRUE)) != 0) {
3180 			mac_stop((mac_handle_t)mip);
3181 			i_mac_perim_exit(mip);
3182 			return (rc);
3183 		}
3184 	}
3185 
3186 	mpip = kmem_cache_alloc(mac_promisc_impl_cache, KM_SLEEP);
3187 
3188 	mpip->mpi_type = type;
3189 	mpip->mpi_fn = fn;
3190 	mpip->mpi_arg = arg;
3191 	mpip->mpi_mcip = mcip;
3192 	mpip->mpi_no_tx_loop = ((flags & MAC_PROMISC_FLAGS_NO_TX_LOOP) != 0);
3193 	mpip->mpi_no_phys = ((flags & MAC_PROMISC_FLAGS_NO_PHYS) != 0);
3194 	mpip->mpi_strip_vlan_tag =
3195 	    ((flags & MAC_PROMISC_FLAGS_VLAN_TAG_STRIP) != 0);
3196 	mpip->mpi_no_copy = ((flags & MAC_PROMISC_FLAGS_NO_COPY) != 0);
3197 
3198 	mcbi = &mip->mi_promisc_cb_info;
3199 	mutex_enter(mcbi->mcbi_lockp);
3200 
3201 	mac_callback_add(&mip->mi_promisc_cb_info, &mcip->mci_promisc_list,
3202 	    &mpip->mpi_mci_link);
3203 	mac_callback_add(&mip->mi_promisc_cb_info, &mip->mi_promisc_list,
3204 	    &mpip->mpi_mi_link);
3205 
3206 	mutex_exit(mcbi->mcbi_lockp);
3207 
3208 	*mphp = (mac_promisc_handle_t)mpip;
3209 	i_mac_perim_exit(mip);
3210 	return (0);
3211 }
3212 
3213 /*
3214  * Remove a multicast address previously aded through mac_promisc_add().
3215  */
3216 void
3217 mac_promisc_remove(mac_promisc_handle_t mph)
3218 {
3219 	mac_promisc_impl_t *mpip = (mac_promisc_impl_t *)mph;
3220 	mac_client_impl_t *mcip = mpip->mpi_mcip;
3221 	mac_impl_t *mip = mcip->mci_mip;
3222 	mac_cb_info_t *mcbi;
3223 	int rv;
3224 
3225 	i_mac_perim_enter(mip);
3226 
3227 	/*
3228 	 * Even if the device can't be reset into normal mode, we still
3229 	 * need to clear the client promisc callbacks. The client may want
3230 	 * to close the mac end point and we can't have stale callbacks.
3231 	 */
3232 	if (!(mpip->mpi_no_phys)) {
3233 		if ((rv = i_mac_promisc_set(mip, B_FALSE)) != 0) {
3234 			cmn_err(CE_WARN, "%s: failed to switch OFF promiscuous"
3235 			    " mode because of error 0x%x", mip->mi_name, rv);
3236 		}
3237 	}
3238 	mcbi = &mip->mi_promisc_cb_info;
3239 	mutex_enter(mcbi->mcbi_lockp);
3240 	if (mac_callback_remove(mcbi, &mip->mi_promisc_list,
3241 	    &mpip->mpi_mi_link)) {
3242 		VERIFY(mac_callback_remove(&mip->mi_promisc_cb_info,
3243 		    &mcip->mci_promisc_list, &mpip->mpi_mci_link));
3244 		kmem_cache_free(mac_promisc_impl_cache, mpip);
3245 	} else {
3246 		mac_callback_remove_wait(&mip->mi_promisc_cb_info);
3247 	}
3248 	mutex_exit(mcbi->mcbi_lockp);
3249 	mac_stop((mac_handle_t)mip);
3250 
3251 	i_mac_perim_exit(mip);
3252 }
3253 
3254 /*
3255  * Reference count the number of active Tx threads. MCI_TX_QUIESCE indicates
3256  * that a control operation wants to quiesce the Tx data flow in which case
3257  * we return an error. Holding any of the per cpu locks ensures that the
3258  * mci_tx_flag won't change.
3259  *
3260  * 'CPU' must be accessed just once and used to compute the index into the
3261  * percpu array, and that index must be used for the entire duration of the
3262  * packet send operation. Note that the thread may be preempted and run on
3263  * another cpu any time and so we can't use 'CPU' more than once for the
3264  * operation.
3265  */
3266 #define	MAC_TX_TRY_HOLD(mcip, mytx, error)				\
3267 {									\
3268 	(error) = 0;							\
3269 	(mytx) = &(mcip)->mci_tx_pcpu[CPU->cpu_seqid & mac_tx_percpu_cnt]; \
3270 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
3271 	if (!((mcip)->mci_tx_flag & MCI_TX_QUIESCE)) {			\
3272 		(mytx)->pcpu_tx_refcnt++;				\
3273 	} else {							\
3274 		(error) = -1;						\
3275 	}								\
3276 	mutex_exit(&(mytx)->pcpu_tx_lock);				\
3277 }
3278 
3279 /*
3280  * Release the reference. If needed, signal any control operation waiting
3281  * for Tx quiescence. The wait and signal are always done using the
3282  * mci_tx_pcpu[0]'s lock
3283  */
3284 #define	MAC_TX_RELE(mcip, mytx) {					\
3285 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
3286 	if (--(mytx)->pcpu_tx_refcnt == 0 &&				\
3287 	    (mcip)->mci_tx_flag & MCI_TX_QUIESCE) {			\
3288 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
3289 		mutex_enter(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
3290 		cv_signal(&(mcip)->mci_tx_cv);				\
3291 		mutex_exit(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
3292 	} else {							\
3293 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
3294 	}								\
3295 }
3296 
3297 /*
3298  * Send function invoked by MAC clients.
3299  */
3300 mac_tx_cookie_t
3301 mac_tx(mac_client_handle_t mch, mblk_t *mp_chain, uintptr_t hint,
3302     uint16_t flag, mblk_t **ret_mp)
3303 {
3304 	mac_tx_cookie_t		cookie = NULL;
3305 	int			error;
3306 	mac_tx_percpu_t		*mytx;
3307 	mac_soft_ring_set_t	*srs;
3308 	flow_entry_t		*flent;
3309 	boolean_t		is_subflow = B_FALSE;
3310 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3311 	mac_impl_t		*mip = mcip->mci_mip;
3312 	mac_srs_tx_t		*srs_tx;
3313 
3314 	/*
3315 	 * Check whether the active Tx threads count is bumped already.
3316 	 */
3317 	if (!(flag & MAC_TX_NO_HOLD)) {
3318 		MAC_TX_TRY_HOLD(mcip, mytx, error);
3319 		if (error != 0) {
3320 			freemsgchain(mp_chain);
3321 			return (NULL);
3322 		}
3323 	}
3324 
3325 	/*
3326 	 * If mac protection is enabled, only the permissible packets will be
3327 	 * returned by mac_protect_check().
3328 	 */
3329 	if ((mcip->mci_flent->
3330 	    fe_resource_props.mrp_mask & MRP_PROTECT) != 0 &&
3331 	    (mp_chain = mac_protect_check(mch, mp_chain)) == NULL)
3332 		goto done;
3333 
3334 	if (mcip->mci_subflow_tab != NULL &&
3335 	    mcip->mci_subflow_tab->ft_flow_count > 0 &&
3336 	    mac_flow_lookup(mcip->mci_subflow_tab, mp_chain,
3337 	    FLOW_OUTBOUND, &flent) == 0) {
3338 		/*
3339 		 * The main assumption here is that if in the event
3340 		 * we get a chain, all the packets will be classified
3341 		 * to the same Flow/SRS. If this changes for any
3342 		 * reason, the following logic should change as well.
3343 		 * I suppose the fanout_hint also assumes this .
3344 		 */
3345 		ASSERT(flent != NULL);
3346 		is_subflow = B_TRUE;
3347 	} else {
3348 		flent = mcip->mci_flent;
3349 	}
3350 
3351 	srs = flent->fe_tx_srs;
3352 	/*
3353 	 * This is to avoid panics with PF_PACKET that can call mac_tx()
3354 	 * against an interface that is not capable of sending. A rewrite
3355 	 * of the mac datapath is required to remove this limitation.
3356 	 */
3357 	if (srs == NULL) {
3358 		freemsgchain(mp_chain);
3359 		goto done;
3360 	}
3361 
3362 	srs_tx = &srs->srs_tx;
3363 	if (srs_tx->st_mode == SRS_TX_DEFAULT &&
3364 	    (srs->srs_state & SRS_ENQUEUED) == 0 &&
3365 	    mip->mi_nactiveclients == 1 && mp_chain->b_next == NULL) {
3366 		uint64_t	obytes;
3367 
3368 		/*
3369 		 * Since dls always opens the underlying MAC, nclients equals
3370 		 * to 1 means that the only active client is dls itself acting
3371 		 * as a primary client of the MAC instance. Since dls will not
3372 		 * send tagged packets in that case, and dls is trusted to send
3373 		 * packets for its allowed VLAN(s), the VLAN tag insertion and
3374 		 * check is required only if nclients is greater than 1.
3375 		 */
3376 		if (mip->mi_nclients > 1) {
3377 			if (MAC_VID_CHECK_NEEDED(mcip)) {
3378 				int	err = 0;
3379 
3380 				MAC_VID_CHECK(mcip, mp_chain, err);
3381 				if (err != 0) {
3382 					freemsg(mp_chain);
3383 					mcip->mci_misc_stat.mms_txerrors++;
3384 					goto done;
3385 				}
3386 			}
3387 			if (MAC_TAG_NEEDED(mcip)) {
3388 				mp_chain = mac_add_vlan_tag(mp_chain, 0,
3389 				    mac_client_vid(mch));
3390 				if (mp_chain == NULL) {
3391 					mcip->mci_misc_stat.mms_txerrors++;
3392 					goto done;
3393 				}
3394 			}
3395 		}
3396 
3397 		obytes = (mp_chain->b_cont == NULL ? MBLKL(mp_chain) :
3398 		    msgdsize(mp_chain));
3399 
3400 		MAC_TX(mip, srs_tx->st_arg2, mp_chain, mcip);
3401 		if (mp_chain == NULL) {
3402 			cookie = NULL;
3403 			SRS_TX_STAT_UPDATE(srs, opackets, 1);
3404 			SRS_TX_STAT_UPDATE(srs, obytes, obytes);
3405 		} else {
3406 			mutex_enter(&srs->srs_lock);
3407 			cookie = mac_tx_srs_no_desc(srs, mp_chain,
3408 			    flag, ret_mp);
3409 			mutex_exit(&srs->srs_lock);
3410 		}
3411 	} else {
3412 		cookie = srs_tx->st_func(srs, mp_chain, hint, flag, ret_mp);
3413 	}
3414 
3415 done:
3416 	if (is_subflow)
3417 		FLOW_REFRELE(flent);
3418 
3419 	if (!(flag & MAC_TX_NO_HOLD))
3420 		MAC_TX_RELE(mcip, mytx);
3421 
3422 	return (cookie);
3423 }
3424 
3425 /*
3426  * mac_tx_is_blocked
3427  *
3428  * Given a cookie, it returns if the ring identified by the cookie is
3429  * flow-controlled or not. If NULL is passed in place of a cookie,
3430  * then it finds out if any of the underlying rings belonging to the
3431  * SRS is flow controlled or not and returns that status.
3432  */
3433 /* ARGSUSED */
3434 boolean_t
3435 mac_tx_is_flow_blocked(mac_client_handle_t mch, mac_tx_cookie_t cookie)
3436 {
3437 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3438 	mac_soft_ring_set_t *mac_srs;
3439 	mac_soft_ring_t *sringp;
3440 	boolean_t blocked = B_FALSE;
3441 	mac_tx_percpu_t *mytx;
3442 	int err;
3443 	int i;
3444 
3445 	/*
3446 	 * Bump the reference count so that mac_srs won't be deleted.
3447 	 * If the client is currently quiesced and we failed to bump
3448 	 * the reference, return B_TRUE so that flow control stays
3449 	 * as enabled.
3450 	 *
3451 	 * Flow control will then be disabled once the client is no
3452 	 * longer quiesced.
3453 	 */
3454 	MAC_TX_TRY_HOLD(mcip, mytx, err);
3455 	if (err != 0)
3456 		return (B_TRUE);
3457 
3458 	if ((mac_srs = MCIP_TX_SRS(mcip)) == NULL) {
3459 		MAC_TX_RELE(mcip, mytx);
3460 		return (B_FALSE);
3461 	}
3462 
3463 	mutex_enter(&mac_srs->srs_lock);
3464 	/*
3465 	 * Only in the case of TX_FANOUT and TX_AGGR, the underlying
3466 	 * softring (s_ring_state) will have the HIWAT set. This is
3467 	 * the multiple Tx ring flow control case. For all other
3468 	 * case, SRS (srs_state) will store the condition.
3469 	 */
3470 	if (mac_srs->srs_tx.st_mode == SRS_TX_FANOUT ||
3471 	    mac_srs->srs_tx.st_mode == SRS_TX_AGGR) {
3472 		if (cookie != NULL) {
3473 			sringp = (mac_soft_ring_t *)cookie;
3474 			mutex_enter(&sringp->s_ring_lock);
3475 			if (sringp->s_ring_state & S_RING_TX_HIWAT)
3476 				blocked = B_TRUE;
3477 			mutex_exit(&sringp->s_ring_lock);
3478 		} else {
3479 			for (i = 0; i < mac_srs->srs_tx_ring_count; i++) {
3480 				sringp = mac_srs->srs_tx_soft_rings[i];
3481 				mutex_enter(&sringp->s_ring_lock);
3482 				if (sringp->s_ring_state & S_RING_TX_HIWAT) {
3483 					blocked = B_TRUE;
3484 					mutex_exit(&sringp->s_ring_lock);
3485 					break;
3486 				}
3487 				mutex_exit(&sringp->s_ring_lock);
3488 			}
3489 		}
3490 	} else {
3491 		blocked = (mac_srs->srs_state & SRS_TX_HIWAT);
3492 	}
3493 	mutex_exit(&mac_srs->srs_lock);
3494 	MAC_TX_RELE(mcip, mytx);
3495 	return (blocked);
3496 }
3497 
3498 /*
3499  * Check if the MAC client is the primary MAC client.
3500  */
3501 boolean_t
3502 mac_is_primary_client(mac_client_impl_t *mcip)
3503 {
3504 	return (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY);
3505 }
3506 
3507 void
3508 mac_ioctl(mac_handle_t mh, queue_t *wq, mblk_t *bp)
3509 {
3510 	mac_impl_t	*mip = (mac_impl_t *)mh;
3511 	int cmd = ((struct iocblk *)bp->b_rptr)->ioc_cmd;
3512 
3513 	if ((cmd == ND_GET && (mip->mi_callbacks->mc_callbacks & MC_GETPROP)) ||
3514 	    (cmd == ND_SET && (mip->mi_callbacks->mc_callbacks & MC_SETPROP))) {
3515 		/*
3516 		 * If ndd props were registered, call them.
3517 		 * Note that ndd ioctls are Obsolete
3518 		 */
3519 		mac_ndd_ioctl(mip, wq, bp);
3520 		return;
3521 	}
3522 
3523 	/*
3524 	 * Call the driver to handle the ioctl.  The driver may not support
3525 	 * any ioctls, in which case we reply with a NAK on its behalf.
3526 	 */
3527 	if (mip->mi_callbacks->mc_callbacks & MC_IOCTL)
3528 		mip->mi_ioctl(mip->mi_driver, wq, bp);
3529 	else
3530 		miocnak(wq, bp, 0, EINVAL);
3531 }
3532 
3533 /*
3534  * Return the link state of the specified MAC instance.
3535  */
3536 link_state_t
3537 mac_link_get(mac_handle_t mh)
3538 {
3539 	return (((mac_impl_t *)mh)->mi_linkstate);
3540 }
3541 
3542 /*
3543  * Add a mac client specified notification callback. Please see the comments
3544  * above mac_callback_add() for general information about mac callback
3545  * addition/deletion in the presence of mac callback list walkers
3546  */
3547 mac_notify_handle_t
3548 mac_notify_add(mac_handle_t mh, mac_notify_t notify_fn, void *arg)
3549 {
3550 	mac_impl_t		*mip = (mac_impl_t *)mh;
3551 	mac_notify_cb_t		*mncb;
3552 	mac_cb_info_t		*mcbi;
3553 
3554 	/*
3555 	 * Allocate a notify callback structure, fill in the details and
3556 	 * use the mac callback list manipulation functions to chain into
3557 	 * the list of callbacks.
3558 	 */
3559 	mncb = kmem_zalloc(sizeof (mac_notify_cb_t), KM_SLEEP);
3560 	mncb->mncb_fn = notify_fn;
3561 	mncb->mncb_arg = arg;
3562 	mncb->mncb_mip = mip;
3563 	mncb->mncb_link.mcb_objp = mncb;
3564 	mncb->mncb_link.mcb_objsize = sizeof (mac_notify_cb_t);
3565 	mncb->mncb_link.mcb_flags = MCB_NOTIFY_CB_T;
3566 
3567 	mcbi = &mip->mi_notify_cb_info;
3568 
3569 	i_mac_perim_enter(mip);
3570 	mutex_enter(mcbi->mcbi_lockp);
3571 
3572 	mac_callback_add(&mip->mi_notify_cb_info, &mip->mi_notify_cb_list,
3573 	    &mncb->mncb_link);
3574 
3575 	mutex_exit(mcbi->mcbi_lockp);
3576 	i_mac_perim_exit(mip);
3577 	return ((mac_notify_handle_t)mncb);
3578 }
3579 
3580 void
3581 mac_notify_remove_wait(mac_handle_t mh)
3582 {
3583 	mac_impl_t	*mip = (mac_impl_t *)mh;
3584 	mac_cb_info_t	*mcbi = &mip->mi_notify_cb_info;
3585 
3586 	mutex_enter(mcbi->mcbi_lockp);
3587 	mac_callback_remove_wait(&mip->mi_notify_cb_info);
3588 	mutex_exit(mcbi->mcbi_lockp);
3589 }
3590 
3591 /*
3592  * Remove a mac client specified notification callback
3593  */
3594 int
3595 mac_notify_remove(mac_notify_handle_t mnh, boolean_t wait)
3596 {
3597 	mac_notify_cb_t	*mncb = (mac_notify_cb_t *)mnh;
3598 	mac_impl_t	*mip = mncb->mncb_mip;
3599 	mac_cb_info_t	*mcbi;
3600 	int		err = 0;
3601 
3602 	mcbi = &mip->mi_notify_cb_info;
3603 
3604 	i_mac_perim_enter(mip);
3605 	mutex_enter(mcbi->mcbi_lockp);
3606 
3607 	ASSERT(mncb->mncb_link.mcb_objp == mncb);
3608 	/*
3609 	 * If there aren't any list walkers, the remove would succeed
3610 	 * inline, else we wait for the deferred remove to complete
3611 	 */
3612 	if (mac_callback_remove(&mip->mi_notify_cb_info,
3613 	    &mip->mi_notify_cb_list, &mncb->mncb_link)) {
3614 		kmem_free(mncb, sizeof (mac_notify_cb_t));
3615 	} else {
3616 		err = EBUSY;
3617 	}
3618 
3619 	mutex_exit(mcbi->mcbi_lockp);
3620 	i_mac_perim_exit(mip);
3621 
3622 	/*
3623 	 * If we failed to remove the notification callback and "wait" is set
3624 	 * to be B_TRUE, wait for the callback to finish after we exit the
3625 	 * mac perimeter.
3626 	 */
3627 	if (err != 0 && wait) {
3628 		mac_notify_remove_wait((mac_handle_t)mip);
3629 		return (0);
3630 	}
3631 
3632 	return (err);
3633 }
3634 
3635 /*
3636  * Associate resource management callbacks with the specified MAC
3637  * clients.
3638  */
3639 
3640 void
3641 mac_resource_set_common(mac_client_handle_t mch, mac_resource_add_t add,
3642     mac_resource_remove_t remove, mac_resource_quiesce_t quiesce,
3643     mac_resource_restart_t restart, mac_resource_bind_t bind,
3644     void *arg)
3645 {
3646 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3647 
3648 	mcip->mci_resource_add = add;
3649 	mcip->mci_resource_remove = remove;
3650 	mcip->mci_resource_quiesce = quiesce;
3651 	mcip->mci_resource_restart = restart;
3652 	mcip->mci_resource_bind = bind;
3653 	mcip->mci_resource_arg = arg;
3654 }
3655 
3656 void
3657 mac_resource_set(mac_client_handle_t mch, mac_resource_add_t add, void *arg)
3658 {
3659 	/* update the 'resource_add' callback */
3660 	mac_resource_set_common(mch, add, NULL, NULL, NULL, NULL, arg);
3661 }
3662 
3663 /*
3664  * Sets up the client resources and enable the polling interface over all the
3665  * SRS's and the soft rings of the client
3666  */
3667 void
3668 mac_client_poll_enable(mac_client_handle_t mch)
3669 {
3670 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3671 	mac_soft_ring_set_t	*mac_srs;
3672 	flow_entry_t		*flent;
3673 	int			i;
3674 
3675 	flent = mcip->mci_flent;
3676 	ASSERT(flent != NULL);
3677 
3678 	mcip->mci_state_flags |= MCIS_CLIENT_POLL_CAPABLE;
3679 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3680 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3681 		ASSERT(mac_srs->srs_mcip == mcip);
3682 		mac_srs_client_poll_enable(mcip, mac_srs);
3683 	}
3684 }
3685 
3686 /*
3687  * Tears down the client resources and disable the polling interface over all
3688  * the SRS's and the soft rings of the client
3689  */
3690 void
3691 mac_client_poll_disable(mac_client_handle_t mch)
3692 {
3693 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3694 	mac_soft_ring_set_t	*mac_srs;
3695 	flow_entry_t		*flent;
3696 	int			i;
3697 
3698 	flent = mcip->mci_flent;
3699 	ASSERT(flent != NULL);
3700 
3701 	mcip->mci_state_flags &= ~MCIS_CLIENT_POLL_CAPABLE;
3702 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3703 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3704 		ASSERT(mac_srs->srs_mcip == mcip);
3705 		mac_srs_client_poll_disable(mcip, mac_srs);
3706 	}
3707 }
3708 
3709 /*
3710  * Associate the CPUs specified by the given property with a MAC client.
3711  */
3712 int
3713 mac_cpu_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
3714 {
3715 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3716 	mac_impl_t *mip = mcip->mci_mip;
3717 	int err = 0;
3718 
3719 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3720 
3721 	if ((err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
3722 	    mcip->mci_upper_mip : mip, mrp)) != 0) {
3723 		return (err);
3724 	}
3725 	if (MCIP_DATAPATH_SETUP(mcip))
3726 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
3727 
3728 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
3729 	return (0);
3730 }
3731 
3732 /*
3733  * Apply the specified properties to the specified MAC client.
3734  */
3735 int
3736 mac_client_set_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3737 {
3738 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3739 	mac_impl_t *mip = mcip->mci_mip;
3740 	int err = 0;
3741 
3742 	i_mac_perim_enter(mip);
3743 
3744 	if ((mrp->mrp_mask & MRP_MAXBW) || (mrp->mrp_mask & MRP_PRIORITY)) {
3745 		err = mac_resource_ctl_set(mch, mrp);
3746 		if (err != 0)
3747 			goto done;
3748 	}
3749 
3750 	if (mrp->mrp_mask & (MRP_CPUS|MRP_POOL)) {
3751 		err = mac_cpu_set(mch, mrp);
3752 		if (err != 0)
3753 			goto done;
3754 	}
3755 
3756 	if (mrp->mrp_mask & MRP_PROTECT) {
3757 		err = mac_protect_set(mch, mrp);
3758 		if (err != 0)
3759 			goto done;
3760 	}
3761 
3762 	if ((mrp->mrp_mask & MRP_RX_RINGS) || (mrp->mrp_mask & MRP_TX_RINGS))
3763 		err = mac_resource_ctl_set(mch, mrp);
3764 
3765 done:
3766 	i_mac_perim_exit(mip);
3767 	return (err);
3768 }
3769 
3770 /*
3771  * Return the properties currently associated with the specified MAC client.
3772  */
3773 void
3774 mac_client_get_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3775 {
3776 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3777 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
3778 
3779 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3780 }
3781 
3782 /*
3783  * Return the effective properties currently associated with the specified
3784  * MAC client.
3785  */
3786 void
3787 mac_client_get_effective_resources(mac_client_handle_t mch,
3788     mac_resource_props_t *mrp)
3789 {
3790 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3791 	mac_resource_props_t	*mcip_mrp = MCIP_EFFECTIVE_PROPS(mcip);
3792 
3793 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3794 }
3795 
3796 /*
3797  * Pass a copy of the specified packet to the promiscuous callbacks
3798  * of the specified MAC.
3799  *
3800  * If sender is NULL, the function is being invoked for a packet chain
3801  * received from the wire. If sender is non-NULL, it points to
3802  * the MAC client from which the packet is being sent.
3803  *
3804  * The packets are distributed to the promiscuous callbacks as follows:
3805  *
3806  * - all packets are sent to the MAC_CLIENT_PROMISC_ALL callbacks
3807  * - all broadcast and multicast packets are sent to the
3808  *   MAC_CLIENT_PROMISC_FILTER and MAC_CLIENT_PROMISC_MULTI.
3809  *
3810  * The unicast packets of MAC_CLIENT_PROMISC_FILTER callbacks are dispatched
3811  * after classification by mac_rx_deliver().
3812  */
3813 
3814 static void
3815 mac_promisc_dispatch_one(mac_promisc_impl_t *mpip, mblk_t *mp,
3816     boolean_t loopback)
3817 {
3818 	mblk_t *mp_copy, *mp_next;
3819 
3820 	if (!mpip->mpi_no_copy || mpip->mpi_strip_vlan_tag) {
3821 		mp_copy = copymsg(mp);
3822 		if (mp_copy == NULL)
3823 			return;
3824 
3825 		if (mpip->mpi_strip_vlan_tag) {
3826 			mp_copy = mac_strip_vlan_tag_chain(mp_copy);
3827 			if (mp_copy == NULL)
3828 				return;
3829 		}
3830 		mp_next = NULL;
3831 	} else {
3832 		mp_copy = mp;
3833 		mp_next = mp->b_next;
3834 	}
3835 	mp_copy->b_next = NULL;
3836 
3837 	mpip->mpi_fn(mpip->mpi_arg, NULL, mp_copy, loopback);
3838 	if (mp_copy == mp)
3839 		mp->b_next = mp_next;
3840 }
3841 
3842 /*
3843  * Return the VID of a packet. Zero if the packet is not tagged.
3844  */
3845 static uint16_t
3846 mac_ether_vid(mblk_t *mp)
3847 {
3848 	struct ether_header *eth = (struct ether_header *)mp->b_rptr;
3849 
3850 	if (ntohs(eth->ether_type) == ETHERTYPE_VLAN) {
3851 		struct ether_vlan_header *t_evhp =
3852 		    (struct ether_vlan_header *)mp->b_rptr;
3853 		return (VLAN_ID(ntohs(t_evhp->ether_tci)));
3854 	}
3855 
3856 	return (0);
3857 }
3858 
3859 /*
3860  * Return whether the specified packet contains a multicast or broadcast
3861  * destination MAC address.
3862  */
3863 static boolean_t
3864 mac_is_mcast(mac_impl_t *mip, mblk_t *mp)
3865 {
3866 	mac_header_info_t hdr_info;
3867 
3868 	if (mac_header_info((mac_handle_t)mip, mp, &hdr_info) != 0)
3869 		return (B_FALSE);
3870 	return ((hdr_info.mhi_dsttype == MAC_ADDRTYPE_BROADCAST) ||
3871 	    (hdr_info.mhi_dsttype == MAC_ADDRTYPE_MULTICAST));
3872 }
3873 
3874 /*
3875  * Send a copy of an mblk chain to the MAC clients of the specified MAC.
3876  * "sender" points to the sender MAC client for outbound packets, and
3877  * is set to NULL for inbound packets.
3878  */
3879 void
3880 mac_promisc_dispatch(mac_impl_t *mip, mblk_t *mp_chain,
3881     mac_client_impl_t *sender)
3882 {
3883 	mac_promisc_impl_t *mpip;
3884 	mac_cb_t *mcb;
3885 	mblk_t *mp;
3886 	boolean_t is_mcast, is_sender;
3887 
3888 	MAC_PROMISC_WALKER_INC(mip);
3889 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3890 		is_mcast = mac_is_mcast(mip, mp);
3891 		/* send packet to interested callbacks */
3892 		for (mcb = mip->mi_promisc_list; mcb != NULL;
3893 		    mcb = mcb->mcb_nextp) {
3894 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3895 			is_sender = (mpip->mpi_mcip == sender);
3896 
3897 			if (is_sender && mpip->mpi_no_tx_loop)
3898 				/*
3899 				 * The sender doesn't want to receive
3900 				 * copies of the packets it sends.
3901 				 */
3902 				continue;
3903 
3904 			/* this client doesn't need any packets (bridge) */
3905 			if (mpip->mpi_fn == NULL)
3906 				continue;
3907 
3908 			/*
3909 			 * For an ethernet MAC, don't displatch a multicast
3910 			 * packet to a non-PROMISC_ALL callbacks unless the VID
3911 			 * of the packet matches the VID of the client.
3912 			 */
3913 			if (is_mcast &&
3914 			    mpip->mpi_type != MAC_CLIENT_PROMISC_ALL &&
3915 			    !mac_client_check_flow_vid(mpip->mpi_mcip,
3916 			    mac_ether_vid(mp)))
3917 				continue;
3918 
3919 			if (is_sender ||
3920 			    mpip->mpi_type == MAC_CLIENT_PROMISC_ALL ||
3921 			    is_mcast)
3922 				mac_promisc_dispatch_one(mpip, mp, is_sender);
3923 		}
3924 	}
3925 	MAC_PROMISC_WALKER_DCR(mip);
3926 }
3927 
3928 void
3929 mac_promisc_client_dispatch(mac_client_impl_t *mcip, mblk_t *mp_chain)
3930 {
3931 	mac_impl_t		*mip = mcip->mci_mip;
3932 	mac_promisc_impl_t	*mpip;
3933 	boolean_t		is_mcast;
3934 	mblk_t			*mp;
3935 	mac_cb_t		*mcb;
3936 
3937 	/*
3938 	 * The unicast packets for the MAC client still
3939 	 * need to be delivered to the MAC_CLIENT_PROMISC_FILTERED
3940 	 * promiscuous callbacks. The broadcast and multicast
3941 	 * packets were delivered from mac_rx().
3942 	 */
3943 	MAC_PROMISC_WALKER_INC(mip);
3944 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3945 		is_mcast = mac_is_mcast(mip, mp);
3946 		for (mcb = mcip->mci_promisc_list; mcb != NULL;
3947 		    mcb = mcb->mcb_nextp) {
3948 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3949 			if (mpip->mpi_type == MAC_CLIENT_PROMISC_FILTERED &&
3950 			    !is_mcast) {
3951 				mac_promisc_dispatch_one(mpip, mp, B_FALSE);
3952 			}
3953 		}
3954 	}
3955 	MAC_PROMISC_WALKER_DCR(mip);
3956 }
3957 
3958 /*
3959  * Return the margin value currently assigned to the specified MAC instance.
3960  */
3961 void
3962 mac_margin_get(mac_handle_t mh, uint32_t *marginp)
3963 {
3964 	mac_impl_t *mip = (mac_impl_t *)mh;
3965 
3966 	rw_enter(&(mip->mi_rw_lock), RW_READER);
3967 	*marginp = mip->mi_margin;
3968 	rw_exit(&(mip->mi_rw_lock));
3969 }
3970 
3971 /*
3972  * mac_info_get() is used for retrieving the mac_info when a DL_INFO_REQ is
3973  * issued before a DL_ATTACH_REQ. we walk the i_mac_impl_hash table and find
3974  * the first mac_impl_t with a matching driver name; then we copy its mac_info_t
3975  * to the caller. we do all this with i_mac_impl_lock held so the mac_impl_t
3976  * cannot disappear while we are accessing it.
3977  */
3978 typedef struct i_mac_info_state_s {
3979 	const char	*mi_name;
3980 	mac_info_t	*mi_infop;
3981 } i_mac_info_state_t;
3982 
3983 /*ARGSUSED*/
3984 static uint_t
3985 i_mac_info_walker(mod_hash_key_t key, mod_hash_val_t *val, void *arg)
3986 {
3987 	i_mac_info_state_t *statep = arg;
3988 	mac_impl_t *mip = (mac_impl_t *)val;
3989 
3990 	if (mip->mi_state_flags & MIS_DISABLED)
3991 		return (MH_WALK_CONTINUE);
3992 
3993 	if (strcmp(statep->mi_name,
3994 	    ddi_driver_name(mip->mi_dip)) != 0)
3995 		return (MH_WALK_CONTINUE);
3996 
3997 	statep->mi_infop = &mip->mi_info;
3998 	return (MH_WALK_TERMINATE);
3999 }
4000 
4001 boolean_t
4002 mac_info_get(const char *name, mac_info_t *minfop)
4003 {
4004 	i_mac_info_state_t state;
4005 
4006 	rw_enter(&i_mac_impl_lock, RW_READER);
4007 	state.mi_name = name;
4008 	state.mi_infop = NULL;
4009 	mod_hash_walk(i_mac_impl_hash, i_mac_info_walker, &state);
4010 	if (state.mi_infop == NULL) {
4011 		rw_exit(&i_mac_impl_lock);
4012 		return (B_FALSE);
4013 	}
4014 	*minfop = *state.mi_infop;
4015 	rw_exit(&i_mac_impl_lock);
4016 	return (B_TRUE);
4017 }
4018 
4019 /*
4020  * To get the capabilities that MAC layer cares about, such as rings, factory
4021  * mac address, vnic or not, it should directly invoke this function.  If the
4022  * link is part of a bridge, then the only "capability" it has is the inability
4023  * to do zero copy.
4024  */
4025 boolean_t
4026 i_mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4027 {
4028 	mac_impl_t *mip = (mac_impl_t *)mh;
4029 
4030 	if (mip->mi_bridge_link != NULL)
4031 		return (cap == MAC_CAPAB_NO_ZCOPY);
4032 	else if (mip->mi_callbacks->mc_callbacks & MC_GETCAPAB)
4033 		return (mip->mi_getcapab(mip->mi_driver, cap, cap_data));
4034 	else
4035 		return (B_FALSE);
4036 }
4037 
4038 /*
4039  * Capability query function. If number of active mac clients is greater than
4040  * 1, only limited capabilities can be advertised to the caller no matter the
4041  * driver has certain capability or not. Else, we query the driver to get the
4042  * capability.
4043  */
4044 boolean_t
4045 mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4046 {
4047 	mac_impl_t *mip = (mac_impl_t *)mh;
4048 
4049 	/*
4050 	 * if mi_nactiveclients > 1, only MAC_CAPAB_LEGACY, MAC_CAPAB_HCKSUM,
4051 	 * MAC_CAPAB_NO_NATIVEVLAN and MAC_CAPAB_NO_ZCOPY can be advertised.
4052 	 */
4053 	if (mip->mi_nactiveclients > 1) {
4054 		switch (cap) {
4055 		case MAC_CAPAB_NO_NATIVEVLAN:
4056 		case MAC_CAPAB_NO_ZCOPY:
4057 			return (B_TRUE);
4058 		case MAC_CAPAB_LEGACY:
4059 		case MAC_CAPAB_HCKSUM:
4060 			break;
4061 		default:
4062 			return (B_FALSE);
4063 		}
4064 	}
4065 
4066 	/* else get capab from driver */
4067 	return (i_mac_capab_get(mh, cap, cap_data));
4068 }
4069 
4070 boolean_t
4071 mac_sap_verify(mac_handle_t mh, uint32_t sap, uint32_t *bind_sap)
4072 {
4073 	mac_impl_t *mip = (mac_impl_t *)mh;
4074 
4075 	return (mip->mi_type->mt_ops.mtops_sap_verify(sap, bind_sap,
4076 	    mip->mi_pdata));
4077 }
4078 
4079 mblk_t *
4080 mac_header(mac_handle_t mh, const uint8_t *daddr, uint32_t sap, mblk_t *payload,
4081     size_t extra_len)
4082 {
4083 	mac_impl_t	*mip = (mac_impl_t *)mh;
4084 	const uint8_t	*hdr_daddr;
4085 
4086 	/*
4087 	 * If the MAC is point-to-point with a fixed destination address, then
4088 	 * we must always use that destination in the MAC header.
4089 	 */
4090 	hdr_daddr = (mip->mi_dstaddr_set ? mip->mi_dstaddr : daddr);
4091 	return (mip->mi_type->mt_ops.mtops_header(mip->mi_addr, hdr_daddr, sap,
4092 	    mip->mi_pdata, payload, extra_len));
4093 }
4094 
4095 int
4096 mac_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4097 {
4098 	mac_impl_t *mip = (mac_impl_t *)mh;
4099 
4100 	return (mip->mi_type->mt_ops.mtops_header_info(mp, mip->mi_pdata,
4101 	    mhip));
4102 }
4103 
4104 int
4105 mac_vlan_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4106 {
4107 	mac_impl_t	*mip = (mac_impl_t *)mh;
4108 	boolean_t	is_ethernet = (mip->mi_info.mi_media == DL_ETHER);
4109 	int		err = 0;
4110 
4111 	/*
4112 	 * Packets should always be at least 16 bit aligned.
4113 	 */
4114 	ASSERT(IS_P2ALIGNED(mp->b_rptr, sizeof (uint16_t)));
4115 
4116 	if ((err = mac_header_info(mh, mp, mhip)) != 0)
4117 		return (err);
4118 
4119 	/*
4120 	 * If this is a VLAN-tagged Ethernet packet, then the SAP in the
4121 	 * mac_header_info_t as returned by mac_header_info() is
4122 	 * ETHERTYPE_VLAN. We need to grab the ethertype from the VLAN header.
4123 	 */
4124 	if (is_ethernet && (mhip->mhi_bindsap == ETHERTYPE_VLAN)) {
4125 		struct ether_vlan_header *evhp;
4126 		uint16_t sap;
4127 		mblk_t *tmp = NULL;
4128 		size_t size;
4129 
4130 		size = sizeof (struct ether_vlan_header);
4131 		if (MBLKL(mp) < size) {
4132 			/*
4133 			 * Pullup the message in order to get the MAC header
4134 			 * infomation. Note that this is a read-only function,
4135 			 * we keep the input packet intact.
4136 			 */
4137 			if ((tmp = msgpullup(mp, size)) == NULL)
4138 				return (EINVAL);
4139 
4140 			mp = tmp;
4141 		}
4142 		evhp = (struct ether_vlan_header *)mp->b_rptr;
4143 		sap = ntohs(evhp->ether_type);
4144 		(void) mac_sap_verify(mh, sap, &mhip->mhi_bindsap);
4145 		mhip->mhi_hdrsize = sizeof (struct ether_vlan_header);
4146 		mhip->mhi_tci = ntohs(evhp->ether_tci);
4147 		mhip->mhi_istagged = B_TRUE;
4148 		freemsg(tmp);
4149 
4150 		if (VLAN_CFI(mhip->mhi_tci) != ETHER_CFI)
4151 			return (EINVAL);
4152 	} else {
4153 		mhip->mhi_istagged = B_FALSE;
4154 		mhip->mhi_tci = 0;
4155 	}
4156 
4157 	return (0);
4158 }
4159 
4160 mblk_t *
4161 mac_header_cook(mac_handle_t mh, mblk_t *mp)
4162 {
4163 	mac_impl_t *mip = (mac_impl_t *)mh;
4164 
4165 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_COOK) {
4166 		if (DB_REF(mp) > 1) {
4167 			mblk_t *newmp = copymsg(mp);
4168 			if (newmp == NULL)
4169 				return (NULL);
4170 			freemsg(mp);
4171 			mp = newmp;
4172 		}
4173 		return (mip->mi_type->mt_ops.mtops_header_cook(mp,
4174 		    mip->mi_pdata));
4175 	}
4176 	return (mp);
4177 }
4178 
4179 mblk_t *
4180 mac_header_uncook(mac_handle_t mh, mblk_t *mp)
4181 {
4182 	mac_impl_t *mip = (mac_impl_t *)mh;
4183 
4184 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_UNCOOK) {
4185 		if (DB_REF(mp) > 1) {
4186 			mblk_t *newmp = copymsg(mp);
4187 			if (newmp == NULL)
4188 				return (NULL);
4189 			freemsg(mp);
4190 			mp = newmp;
4191 		}
4192 		return (mip->mi_type->mt_ops.mtops_header_uncook(mp,
4193 		    mip->mi_pdata));
4194 	}
4195 	return (mp);
4196 }
4197 
4198 uint_t
4199 mac_addr_len(mac_handle_t mh)
4200 {
4201 	mac_impl_t *mip = (mac_impl_t *)mh;
4202 
4203 	return (mip->mi_type->mt_addr_length);
4204 }
4205 
4206 /* True if a MAC is a VNIC */
4207 boolean_t
4208 mac_is_vnic(mac_handle_t mh)
4209 {
4210 	return (((mac_impl_t *)mh)->mi_state_flags & MIS_IS_VNIC);
4211 }
4212 
4213 mac_handle_t
4214 mac_get_lower_mac_handle(mac_handle_t mh)
4215 {
4216 	mac_impl_t *mip = (mac_impl_t *)mh;
4217 
4218 	ASSERT(mac_is_vnic(mh));
4219 	return (((vnic_t *)mip->mi_driver)->vn_lower_mh);
4220 }
4221 
4222 boolean_t
4223 mac_is_vnic_primary(mac_handle_t mh)
4224 {
4225 	mac_impl_t *mip = (mac_impl_t *)mh;
4226 
4227 	ASSERT(mac_is_vnic(mh));
4228 	return (((vnic_t *)mip->mi_driver)->vn_addr_type ==
4229 	    VNIC_MAC_ADDR_TYPE_PRIMARY);
4230 }
4231 
4232 void
4233 mac_update_resources(mac_resource_props_t *nmrp, mac_resource_props_t *cmrp,
4234     boolean_t is_user_flow)
4235 {
4236 	if (nmrp != NULL && cmrp != NULL) {
4237 		if (nmrp->mrp_mask & MRP_PRIORITY) {
4238 			if (nmrp->mrp_priority == MPL_RESET) {
4239 				cmrp->mrp_mask &= ~MRP_PRIORITY;
4240 				if (is_user_flow) {
4241 					cmrp->mrp_priority =
4242 					    MPL_SUBFLOW_DEFAULT;
4243 				} else {
4244 					cmrp->mrp_priority = MPL_LINK_DEFAULT;
4245 				}
4246 			} else {
4247 				cmrp->mrp_mask |= MRP_PRIORITY;
4248 				cmrp->mrp_priority = nmrp->mrp_priority;
4249 			}
4250 		}
4251 		if (nmrp->mrp_mask & MRP_MAXBW) {
4252 			if (nmrp->mrp_maxbw == MRP_MAXBW_RESETVAL) {
4253 				cmrp->mrp_mask &= ~MRP_MAXBW;
4254 				cmrp->mrp_maxbw = 0;
4255 			} else {
4256 				cmrp->mrp_mask |= MRP_MAXBW;
4257 				cmrp->mrp_maxbw = nmrp->mrp_maxbw;
4258 			}
4259 		}
4260 		if (nmrp->mrp_mask & MRP_CPUS)
4261 			MAC_COPY_CPUS(nmrp, cmrp);
4262 
4263 		if (nmrp->mrp_mask & MRP_POOL) {
4264 			if (strlen(nmrp->mrp_pool) == 0) {
4265 				cmrp->mrp_mask &= ~MRP_POOL;
4266 				bzero(cmrp->mrp_pool, sizeof (cmrp->mrp_pool));
4267 			} else {
4268 				cmrp->mrp_mask |= MRP_POOL;
4269 				(void) strncpy(cmrp->mrp_pool, nmrp->mrp_pool,
4270 				    sizeof (cmrp->mrp_pool));
4271 			}
4272 
4273 		}
4274 
4275 		if (nmrp->mrp_mask & MRP_PROTECT)
4276 			mac_protect_update(nmrp, cmrp);
4277 
4278 		/*
4279 		 * Update the rings specified.
4280 		 */
4281 		if (nmrp->mrp_mask & MRP_RX_RINGS) {
4282 			if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4283 				cmrp->mrp_mask &= ~MRP_RX_RINGS;
4284 				if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4285 					cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4286 				cmrp->mrp_nrxrings = 0;
4287 			} else {
4288 				cmrp->mrp_mask |= MRP_RX_RINGS;
4289 				cmrp->mrp_nrxrings = nmrp->mrp_nrxrings;
4290 			}
4291 		}
4292 		if (nmrp->mrp_mask & MRP_TX_RINGS) {
4293 			if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4294 				cmrp->mrp_mask &= ~MRP_TX_RINGS;
4295 				if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4296 					cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4297 				cmrp->mrp_ntxrings = 0;
4298 			} else {
4299 				cmrp->mrp_mask |= MRP_TX_RINGS;
4300 				cmrp->mrp_ntxrings = nmrp->mrp_ntxrings;
4301 			}
4302 		}
4303 		if (nmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4304 			cmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4305 		else if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4306 			cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4307 
4308 		if (nmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4309 			cmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4310 		else if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4311 			cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4312 	}
4313 }
4314 
4315 /*
4316  * i_mac_set_resources:
4317  *
4318  * This routine associates properties with the primary MAC client of
4319  * the specified MAC instance.
4320  * - Cache the properties in mac_impl_t
4321  * - Apply the properties to the primary MAC client if exists
4322  */
4323 int
4324 i_mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4325 {
4326 	mac_impl_t		*mip = (mac_impl_t *)mh;
4327 	mac_client_impl_t	*mcip;
4328 	int			err = 0;
4329 	uint32_t		resmask, newresmask;
4330 	mac_resource_props_t	*tmrp, *umrp;
4331 
4332 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4333 
4334 	err = mac_validate_props(mip, mrp);
4335 	if (err != 0)
4336 		return (err);
4337 
4338 	umrp = kmem_zalloc(sizeof (*umrp), KM_SLEEP);
4339 	bcopy(&mip->mi_resource_props, umrp, sizeof (*umrp));
4340 	resmask = umrp->mrp_mask;
4341 	mac_update_resources(mrp, umrp, B_FALSE);
4342 	newresmask = umrp->mrp_mask;
4343 
4344 	if (resmask == 0 && newresmask != 0) {
4345 		/*
4346 		 * Bandwidth, priority, cpu or pool link properties configured,
4347 		 * must disable fastpath.
4348 		 */
4349 		if ((err = mac_fastpath_disable((mac_handle_t)mip)) != 0) {
4350 			kmem_free(umrp, sizeof (*umrp));
4351 			return (err);
4352 		}
4353 	}
4354 
4355 	/*
4356 	 * Since bind_cpu may be modified by mac_client_set_resources()
4357 	 * we use a copy of bind_cpu and finally cache bind_cpu in mip.
4358 	 * This allows us to cache only user edits in mip.
4359 	 */
4360 	tmrp = kmem_zalloc(sizeof (*tmrp), KM_SLEEP);
4361 	bcopy(mrp, tmrp, sizeof (*tmrp));
4362 	mcip = mac_primary_client_handle(mip);
4363 	if (mcip != NULL && (mcip->mci_state_flags & MCIS_IS_AGGR_PORT) == 0) {
4364 		err = mac_client_set_resources((mac_client_handle_t)mcip, tmrp);
4365 	} else if ((mrp->mrp_mask & MRP_RX_RINGS ||
4366 	    mrp->mrp_mask & MRP_TX_RINGS)) {
4367 		mac_client_impl_t	*vmcip;
4368 
4369 		/*
4370 		 * If the primary is not up, we need to check if there
4371 		 * are any VLANs on this primary. If there are then
4372 		 * we need to set this property on the VLANs since
4373 		 * VLANs follow the primary they are based on. Just
4374 		 * look for the first VLAN and change its properties,
4375 		 * all the other VLANs should be in the same group.
4376 		 */
4377 		for (vmcip = mip->mi_clients_list; vmcip != NULL;
4378 		    vmcip = vmcip->mci_client_next) {
4379 			if ((vmcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) &&
4380 			    mac_client_vid((mac_client_handle_t)vmcip) !=
4381 			    VLAN_ID_NONE) {
4382 				break;
4383 			}
4384 		}
4385 		if (vmcip != NULL) {
4386 			mac_resource_props_t	*omrp;
4387 			mac_resource_props_t	*vmrp;
4388 
4389 			omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
4390 			bcopy(MCIP_RESOURCE_PROPS(vmcip), omrp, sizeof (*omrp));
4391 			/*
4392 			 * We dont' call mac_update_resources since we
4393 			 * want to take only the ring properties and
4394 			 * not all the properties that may have changed.
4395 			 */
4396 			vmrp = MCIP_RESOURCE_PROPS(vmcip);
4397 			if (mrp->mrp_mask & MRP_RX_RINGS) {
4398 				if (mrp->mrp_mask & MRP_RINGS_RESET) {
4399 					vmrp->mrp_mask &= ~MRP_RX_RINGS;
4400 					if (vmrp->mrp_mask &
4401 					    MRP_RXRINGS_UNSPEC) {
4402 						vmrp->mrp_mask &=
4403 						    ~MRP_RXRINGS_UNSPEC;
4404 					}
4405 					vmrp->mrp_nrxrings = 0;
4406 				} else {
4407 					vmrp->mrp_mask |= MRP_RX_RINGS;
4408 					vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
4409 				}
4410 			}
4411 			if (mrp->mrp_mask & MRP_TX_RINGS) {
4412 				if (mrp->mrp_mask & MRP_RINGS_RESET) {
4413 					vmrp->mrp_mask &= ~MRP_TX_RINGS;
4414 					if (vmrp->mrp_mask &
4415 					    MRP_TXRINGS_UNSPEC) {
4416 						vmrp->mrp_mask &=
4417 						    ~MRP_TXRINGS_UNSPEC;
4418 					}
4419 					vmrp->mrp_ntxrings = 0;
4420 				} else {
4421 					vmrp->mrp_mask |= MRP_TX_RINGS;
4422 					vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
4423 				}
4424 			}
4425 			if (mrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4426 				vmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4427 
4428 			if (mrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4429 				vmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4430 
4431 			if ((err = mac_client_set_rings_prop(vmcip, mrp,
4432 			    omrp)) != 0) {
4433 				bcopy(omrp, MCIP_RESOURCE_PROPS(vmcip),
4434 				    sizeof (*omrp));
4435 			} else {
4436 				mac_set_prim_vlan_rings(mip, vmrp);
4437 			}
4438 			kmem_free(omrp, sizeof (*omrp));
4439 		}
4440 	}
4441 
4442 	/* Only update the values if mac_client_set_resources succeeded */
4443 	if (err == 0) {
4444 		bcopy(umrp, &mip->mi_resource_props, sizeof (*umrp));
4445 		/*
4446 		 * If bandwidth, priority or cpu link properties cleared,
4447 		 * renable fastpath.
4448 		 */
4449 		if (resmask != 0 && newresmask == 0)
4450 			mac_fastpath_enable((mac_handle_t)mip);
4451 	} else if (resmask == 0 && newresmask != 0) {
4452 		mac_fastpath_enable((mac_handle_t)mip);
4453 	}
4454 	kmem_free(tmrp, sizeof (*tmrp));
4455 	kmem_free(umrp, sizeof (*umrp));
4456 	return (err);
4457 }
4458 
4459 int
4460 mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4461 {
4462 	int err;
4463 
4464 	i_mac_perim_enter((mac_impl_t *)mh);
4465 	err = i_mac_set_resources(mh, mrp);
4466 	i_mac_perim_exit((mac_impl_t *)mh);
4467 	return (err);
4468 }
4469 
4470 /*
4471  * Get the properties cached for the specified MAC instance.
4472  */
4473 void
4474 mac_get_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4475 {
4476 	mac_impl_t 		*mip = (mac_impl_t *)mh;
4477 	mac_client_impl_t	*mcip;
4478 
4479 	mcip = mac_primary_client_handle(mip);
4480 	if (mcip != NULL) {
4481 		mac_client_get_resources((mac_client_handle_t)mcip, mrp);
4482 		return;
4483 	}
4484 	bcopy(&mip->mi_resource_props, mrp, sizeof (mac_resource_props_t));
4485 }
4486 
4487 /*
4488  * Get the effective properties from the primary client of the
4489  * specified MAC instance.
4490  */
4491 void
4492 mac_get_effective_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4493 {
4494 	mac_impl_t 		*mip = (mac_impl_t *)mh;
4495 	mac_client_impl_t	*mcip;
4496 
4497 	mcip = mac_primary_client_handle(mip);
4498 	if (mcip != NULL) {
4499 		mac_client_get_effective_resources((mac_client_handle_t)mcip,
4500 		    mrp);
4501 		return;
4502 	}
4503 	bzero(mrp, sizeof (mac_resource_props_t));
4504 }
4505 
4506 int
4507 mac_set_pvid(mac_handle_t mh, uint16_t pvid)
4508 {
4509 	mac_impl_t *mip = (mac_impl_t *)mh;
4510 	mac_client_impl_t *mcip;
4511 	mac_unicast_impl_t *muip;
4512 
4513 	i_mac_perim_enter(mip);
4514 	if (pvid != 0) {
4515 		for (mcip = mip->mi_clients_list; mcip != NULL;
4516 		    mcip = mcip->mci_client_next) {
4517 			for (muip = mcip->mci_unicast_list; muip != NULL;
4518 			    muip = muip->mui_next) {
4519 				if (muip->mui_vid == pvid) {
4520 					i_mac_perim_exit(mip);
4521 					return (EBUSY);
4522 				}
4523 			}
4524 		}
4525 	}
4526 	mip->mi_pvid = pvid;
4527 	i_mac_perim_exit(mip);
4528 	return (0);
4529 }
4530 
4531 uint16_t
4532 mac_get_pvid(mac_handle_t mh)
4533 {
4534 	mac_impl_t *mip = (mac_impl_t *)mh;
4535 
4536 	return (mip->mi_pvid);
4537 }
4538 
4539 uint32_t
4540 mac_get_llimit(mac_handle_t mh)
4541 {
4542 	mac_impl_t *mip = (mac_impl_t *)mh;
4543 
4544 	return (mip->mi_llimit);
4545 }
4546 
4547 uint32_t
4548 mac_get_ldecay(mac_handle_t mh)
4549 {
4550 	mac_impl_t *mip = (mac_impl_t *)mh;
4551 
4552 	return (mip->mi_ldecay);
4553 }
4554 
4555 /*
4556  * Rename a mac client, its flow, and the kstat.
4557  */
4558 int
4559 mac_rename_primary(mac_handle_t mh, const char *new_name)
4560 {
4561 	mac_impl_t		*mip = (mac_impl_t *)mh;
4562 	mac_client_impl_t	*cur_clnt = NULL;
4563 	flow_entry_t		*fep;
4564 
4565 	i_mac_perim_enter(mip);
4566 
4567 	/*
4568 	 * VNICs: we need to change the sys flow name and
4569 	 * the associated flow kstat.
4570 	 */
4571 	if (mip->mi_state_flags & MIS_IS_VNIC) {
4572 		mac_client_impl_t *mcip = mac_vnic_lower(mip);
4573 		ASSERT(new_name != NULL);
4574 		mac_rename_flow_names(mcip, new_name);
4575 		mac_stat_rename(mcip);
4576 		goto done;
4577 	}
4578 	/*
4579 	 * This mac may itself be an aggr link, or it may have some client
4580 	 * which is an aggr port. For both cases, we need to change the
4581 	 * aggr port's mac client name, its flow name and the associated flow
4582 	 * kstat.
4583 	 */
4584 	if (mip->mi_state_flags & MIS_IS_AGGR) {
4585 		mac_capab_aggr_t aggr_cap;
4586 		mac_rename_fn_t rename_fn;
4587 		boolean_t ret;
4588 
4589 		ASSERT(new_name != NULL);
4590 		ret = i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_AGGR,
4591 		    (void *)(&aggr_cap));
4592 		ASSERT(ret == B_TRUE);
4593 		rename_fn = aggr_cap.mca_rename_fn;
4594 		rename_fn(new_name, mip->mi_driver);
4595 		/*
4596 		 * The aggr's client name and kstat flow name will be
4597 		 * updated below, i.e. via mac_rename_flow_names.
4598 		 */
4599 	}
4600 
4601 	for (cur_clnt = mip->mi_clients_list; cur_clnt != NULL;
4602 	    cur_clnt = cur_clnt->mci_client_next) {
4603 		if (cur_clnt->mci_state_flags & MCIS_IS_AGGR_PORT) {
4604 			if (new_name != NULL) {
4605 				char *str_st = cur_clnt->mci_name;
4606 				char *str_del = strchr(str_st, '-');
4607 
4608 				ASSERT(str_del != NULL);
4609 				bzero(str_del + 1, MAXNAMELEN -
4610 				    (str_del - str_st + 1));
4611 				bcopy(new_name, str_del + 1,
4612 				    strlen(new_name));
4613 			}
4614 			fep = cur_clnt->mci_flent;
4615 			mac_rename_flow(fep, cur_clnt->mci_name);
4616 			break;
4617 		} else if (new_name != NULL &&
4618 		    cur_clnt->mci_state_flags & MCIS_USE_DATALINK_NAME) {
4619 			mac_rename_flow_names(cur_clnt, new_name);
4620 			break;
4621 		}
4622 	}
4623 
4624 	/* Recreate kstats associated with aggr pseudo rings */
4625 	if (mip->mi_state_flags & MIS_IS_AGGR)
4626 		mac_pseudo_ring_stat_rename(mip);
4627 
4628 done:
4629 	i_mac_perim_exit(mip);
4630 	return (0);
4631 }
4632 
4633 /*
4634  * Rename the MAC client's flow names
4635  */
4636 static void
4637 mac_rename_flow_names(mac_client_impl_t *mcip, const char *new_name)
4638 {
4639 	flow_entry_t	*flent;
4640 	uint16_t	vid;
4641 	char		flowname[MAXFLOWNAMELEN];
4642 	mac_impl_t	*mip = mcip->mci_mip;
4643 
4644 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4645 
4646 	/*
4647 	 * Use mi_rw_lock to ensure that threads not in the mac perimeter
4648 	 * see a self-consistent value for mci_name
4649 	 */
4650 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
4651 	(void) strlcpy(mcip->mci_name, new_name, sizeof (mcip->mci_name));
4652 	rw_exit(&mip->mi_rw_lock);
4653 
4654 	mac_rename_flow(mcip->mci_flent, new_name);
4655 
4656 	if (mcip->mci_nflents == 1)
4657 		return;
4658 
4659 	/*
4660 	 * We have to rename all the others too, no stats to destroy for
4661 	 * these.
4662 	 */
4663 	for (flent = mcip->mci_flent_list; flent != NULL;
4664 	    flent = flent->fe_client_next) {
4665 		if (flent != mcip->mci_flent) {
4666 			vid = i_mac_flow_vid(flent);
4667 			(void) sprintf(flowname, "%s%u", new_name, vid);
4668 			mac_flow_set_name(flent, flowname);
4669 		}
4670 	}
4671 }
4672 
4673 
4674 /*
4675  * Add a flow to the MAC client's flow list - i.e list of MAC/VID tuples
4676  * defined for the specified MAC client.
4677  */
4678 static void
4679 mac_client_add_to_flow_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4680 {
4681 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4682 	/*
4683 	 * The promisc Rx data path walks the mci_flent_list. Protect by
4684 	 * using mi_rw_lock
4685 	 */
4686 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4687 
4688 	/* Add it to the head */
4689 	flent->fe_client_next = mcip->mci_flent_list;
4690 	mcip->mci_flent_list = flent;
4691 	mcip->mci_nflents++;
4692 
4693 	/*
4694 	 * Keep track of the number of non-zero VIDs addresses per MAC
4695 	 * client to avoid figuring it out in the data-path.
4696 	 */
4697 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4698 		mcip->mci_nvids++;
4699 
4700 	rw_exit(&mcip->mci_rw_lock);
4701 }
4702 
4703 /*
4704  * Remove a flow entry from the MAC client's list.
4705  */
4706 static void
4707 mac_client_remove_flow_from_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4708 {
4709 	flow_entry_t	*fe = mcip->mci_flent_list;
4710 	flow_entry_t	*prev_fe = NULL;
4711 
4712 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4713 	/*
4714 	 * The promisc Rx data path walks the mci_flent_list. Protect by
4715 	 * using mci_rw_lock
4716 	 */
4717 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4718 	while ((fe != NULL) && (fe != flent)) {
4719 		prev_fe = fe;
4720 		fe = fe->fe_client_next;
4721 	}
4722 
4723 	ASSERT(fe != NULL);
4724 	if (prev_fe == NULL) {
4725 		/* Deleting the first node */
4726 		mcip->mci_flent_list = fe->fe_client_next;
4727 	} else {
4728 		prev_fe->fe_client_next = fe->fe_client_next;
4729 	}
4730 	mcip->mci_nflents--;
4731 
4732 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4733 		mcip->mci_nvids--;
4734 
4735 	rw_exit(&mcip->mci_rw_lock);
4736 }
4737 
4738 /*
4739  * Check if the given VID belongs to this MAC client.
4740  */
4741 boolean_t
4742 mac_client_check_flow_vid(mac_client_impl_t *mcip, uint16_t vid)
4743 {
4744 	flow_entry_t	*flent;
4745 	uint16_t	mci_vid;
4746 
4747 	/* The mci_flent_list is protected by mci_rw_lock */
4748 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4749 	for (flent = mcip->mci_flent_list; flent != NULL;
4750 	    flent = flent->fe_client_next) {
4751 		mci_vid = i_mac_flow_vid(flent);
4752 		if (vid == mci_vid) {
4753 			rw_exit(&mcip->mci_rw_lock);
4754 			return (B_TRUE);
4755 		}
4756 	}
4757 	rw_exit(&mcip->mci_rw_lock);
4758 	return (B_FALSE);
4759 }
4760 
4761 /*
4762  * Get the flow entry for the specified <MAC addr, VID> tuple.
4763  */
4764 static flow_entry_t *
4765 mac_client_get_flow(mac_client_impl_t *mcip, mac_unicast_impl_t *muip)
4766 {
4767 	mac_address_t *map = mcip->mci_unicast;
4768 	flow_entry_t *flent;
4769 	uint16_t vid;
4770 	flow_desc_t flow_desc;
4771 
4772 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4773 
4774 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
4775 	if (bcmp(flow_desc.fd_dst_mac, map->ma_addr, map->ma_len) != 0)
4776 		return (NULL);
4777 
4778 	for (flent = mcip->mci_flent_list; flent != NULL;
4779 	    flent = flent->fe_client_next) {
4780 		vid = i_mac_flow_vid(flent);
4781 		if (vid == muip->mui_vid) {
4782 			return (flent);
4783 		}
4784 	}
4785 
4786 	return (NULL);
4787 }
4788 
4789 /*
4790  * Since mci_flent has the SRSs, when we want to remove it, we replace
4791  * the flow_desc_t in mci_flent with that of an existing flent and then
4792  * remove that flent instead of mci_flent.
4793  */
4794 static flow_entry_t *
4795 mac_client_swap_mciflent(mac_client_impl_t *mcip)
4796 {
4797 	flow_entry_t	*flent = mcip->mci_flent;
4798 	flow_tab_t	*ft = flent->fe_flow_tab;
4799 	flow_entry_t	*flent1;
4800 	flow_desc_t	fl_desc;
4801 	char		fl_name[MAXFLOWNAMELEN];
4802 	int		err;
4803 
4804 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4805 	ASSERT(mcip->mci_nflents > 1);
4806 
4807 	/* get the next flent following the primary flent  */
4808 	flent1 = mcip->mci_flent_list->fe_client_next;
4809 	ASSERT(flent1 != NULL && flent1->fe_flow_tab == ft);
4810 
4811 	/*
4812 	 * Remove the flent from the flow table before updating the
4813 	 * flow descriptor as the hash depends on the flow descriptor.
4814 	 * This also helps incoming packet classification avoid having
4815 	 * to grab fe_lock. Access to fe_flow_desc of a flent not in the
4816 	 * flow table is done under the fe_lock so that log or stat functions
4817 	 * see a self-consistent fe_flow_desc. The name and desc are specific
4818 	 * to a flow, the rest are shared by all the clients, including
4819 	 * resource control etc.
4820 	 */
4821 	mac_flow_remove(ft, flent, B_TRUE);
4822 	mac_flow_remove(ft, flent1, B_TRUE);
4823 
4824 	bcopy(&flent->fe_flow_desc, &fl_desc, sizeof (flow_desc_t));
4825 	bcopy(flent->fe_flow_name, fl_name, MAXFLOWNAMELEN);
4826 
4827 	/* update the primary flow entry */
4828 	mutex_enter(&flent->fe_lock);
4829 	bcopy(&flent1->fe_flow_desc, &flent->fe_flow_desc,
4830 	    sizeof (flow_desc_t));
4831 	bcopy(&flent1->fe_flow_name, &flent->fe_flow_name, MAXFLOWNAMELEN);
4832 	mutex_exit(&flent->fe_lock);
4833 
4834 	/* update the flow entry that is to be freed */
4835 	mutex_enter(&flent1->fe_lock);
4836 	bcopy(&fl_desc, &flent1->fe_flow_desc, sizeof (flow_desc_t));
4837 	bcopy(fl_name, &flent1->fe_flow_name, MAXFLOWNAMELEN);
4838 	mutex_exit(&flent1->fe_lock);
4839 
4840 	/* now reinsert the flow entries in the table */
4841 	err = mac_flow_add(ft, flent);
4842 	ASSERT(err == 0);
4843 
4844 	err = mac_flow_add(ft, flent1);
4845 	ASSERT(err == 0);
4846 
4847 	return (flent1);
4848 }
4849 
4850 /*
4851  * Return whether there is only one flow entry associated with this
4852  * MAC client.
4853  */
4854 static boolean_t
4855 mac_client_single_rcvr(mac_client_impl_t *mcip)
4856 {
4857 	return (mcip->mci_nflents == 1);
4858 }
4859 
4860 int
4861 mac_validate_props(mac_impl_t *mip, mac_resource_props_t *mrp)
4862 {
4863 	boolean_t		reset;
4864 	uint32_t		rings_needed;
4865 	uint32_t		rings_avail;
4866 	mac_group_type_t	gtype;
4867 	mac_resource_props_t	*mip_mrp;
4868 
4869 	if (mrp == NULL)
4870 		return (0);
4871 
4872 	if (mrp->mrp_mask & MRP_PRIORITY) {
4873 		mac_priority_level_t	pri = mrp->mrp_priority;
4874 
4875 		if (pri < MPL_LOW || pri > MPL_RESET)
4876 			return (EINVAL);
4877 	}
4878 
4879 	if (mrp->mrp_mask & MRP_MAXBW) {
4880 		uint64_t maxbw = mrp->mrp_maxbw;
4881 
4882 		if (maxbw < MRP_MAXBW_MINVAL && maxbw != 0)
4883 			return (EINVAL);
4884 	}
4885 	if (mrp->mrp_mask & MRP_CPUS) {
4886 		int i, j;
4887 		mac_cpu_mode_t	fanout;
4888 
4889 		if (mrp->mrp_ncpus > ncpus || mrp->mrp_ncpus > MAX_SR_FANOUT)
4890 			return (EINVAL);
4891 
4892 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4893 			for (j = 0; j < mrp->mrp_ncpus; j++) {
4894 				if (i != j &&
4895 				    mrp->mrp_cpu[i] == mrp->mrp_cpu[j]) {
4896 					return (EINVAL);
4897 				}
4898 			}
4899 		}
4900 
4901 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4902 			cpu_t *cp;
4903 			int rv;
4904 
4905 			mutex_enter(&cpu_lock);
4906 			cp = cpu_get(mrp->mrp_cpu[i]);
4907 			if (cp != NULL)
4908 				rv = cpu_is_online(cp);
4909 			else
4910 				rv = 0;
4911 			mutex_exit(&cpu_lock);
4912 			if (rv == 0)
4913 				return (EINVAL);
4914 		}
4915 
4916 		fanout = mrp->mrp_fanout_mode;
4917 		if (fanout < 0 || fanout > MCM_CPUS)
4918 			return (EINVAL);
4919 	}
4920 
4921 	if (mrp->mrp_mask & MRP_PROTECT) {
4922 		int err = mac_protect_validate(mrp);
4923 		if (err != 0)
4924 			return (err);
4925 	}
4926 
4927 	if (!(mrp->mrp_mask & MRP_RX_RINGS) &&
4928 	    !(mrp->mrp_mask & MRP_TX_RINGS)) {
4929 		return (0);
4930 	}
4931 
4932 	/*
4933 	 * mip will be null when we come from mac_flow_create or
4934 	 * mac_link_flow_modify. In the latter case it is a user flow,
4935 	 * for which we don't support rings. In the former we would
4936 	 * have validated the props beforehand (i_mac_unicast_add ->
4937 	 * mac_client_set_resources -> validate for the primary and
4938 	 * vnic_dev_create -> mac_client_set_resources -> validate for
4939 	 * a vnic.
4940 	 */
4941 	if (mip == NULL)
4942 		return (0);
4943 
4944 	/*
4945 	 * We don't support setting rings property for a VNIC that is using a
4946 	 * primary address (VLAN)
4947 	 */
4948 	if ((mip->mi_state_flags & MIS_IS_VNIC) &&
4949 	    mac_is_vnic_primary((mac_handle_t)mip)) {
4950 		return (ENOTSUP);
4951 	}
4952 
4953 	mip_mrp = &mip->mi_resource_props;
4954 	/*
4955 	 * The rings property should be validated against the NICs
4956 	 * resources
4957 	 */
4958 	if (mip->mi_state_flags & MIS_IS_VNIC)
4959 		mip = (mac_impl_t *)mac_get_lower_mac_handle((mac_handle_t)mip);
4960 
4961 	reset = mrp->mrp_mask & MRP_RINGS_RESET;
4962 	/*
4963 	 * If groups are not supported, return error.
4964 	 */
4965 	if (((mrp->mrp_mask & MRP_RX_RINGS) && mip->mi_rx_groups == NULL) ||
4966 	    ((mrp->mrp_mask & MRP_TX_RINGS) && mip->mi_tx_groups == NULL)) {
4967 		return (EINVAL);
4968 	}
4969 	/*
4970 	 * If we are just resetting, there is no validation needed.
4971 	 */
4972 	if (reset)
4973 		return (0);
4974 
4975 	if (mrp->mrp_mask & MRP_RX_RINGS) {
4976 		rings_needed = mrp->mrp_nrxrings;
4977 		/*
4978 		 * We just want to check if the number of additional
4979 		 * rings requested is available.
4980 		 */
4981 		if (mip_mrp->mrp_mask & MRP_RX_RINGS) {
4982 			if (mrp->mrp_nrxrings > mip_mrp->mrp_nrxrings)
4983 				/* Just check for the additional rings */
4984 				rings_needed -= mip_mrp->mrp_nrxrings;
4985 			else
4986 				/* We are not asking for additional rings */
4987 				rings_needed = 0;
4988 		}
4989 		rings_avail = mip->mi_rxrings_avail;
4990 		gtype = mip->mi_rx_group_type;
4991 	} else {
4992 		rings_needed = mrp->mrp_ntxrings;
4993 		/* Similarly for the TX rings */
4994 		if (mip_mrp->mrp_mask & MRP_TX_RINGS) {
4995 			if (mrp->mrp_ntxrings > mip_mrp->mrp_ntxrings)
4996 				/* Just check for the additional rings */
4997 				rings_needed -= mip_mrp->mrp_ntxrings;
4998 			else
4999 				/* We are not asking for additional rings */
5000 				rings_needed = 0;
5001 		}
5002 		rings_avail = mip->mi_txrings_avail;
5003 		gtype = mip->mi_tx_group_type;
5004 	}
5005 
5006 	/* Error if the group is dynamic .. */
5007 	if (gtype == MAC_GROUP_TYPE_DYNAMIC) {
5008 		/*
5009 		 * .. and rings specified are more than available.
5010 		 */
5011 		if (rings_needed > rings_avail)
5012 			return (EINVAL);
5013 	} else {
5014 		/*
5015 		 * OR group is static and we have specified some rings.
5016 		 */
5017 		if (rings_needed > 0)
5018 			return (EINVAL);
5019 	}
5020 	return (0);
5021 }
5022 
5023 /*
5024  * Send a MAC_NOTE_LINK notification to all the MAC clients whenever the
5025  * underlying physical link is down. This is to allow MAC clients to
5026  * communicate with other clients.
5027  */
5028 void
5029 mac_virtual_link_update(mac_impl_t *mip)
5030 {
5031 	if (mip->mi_linkstate != LINK_STATE_UP)
5032 		i_mac_notify(mip, MAC_NOTE_LINK);
5033 }
5034 
5035 /*
5036  * For clients that have a pass-thru MAC, e.g. VNIC, we set the VNIC's
5037  * mac handle in the client.
5038  */
5039 void
5040 mac_set_upper_mac(mac_client_handle_t mch, mac_handle_t mh,
5041     mac_resource_props_t *mrp)
5042 {
5043 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
5044 	mac_impl_t		*mip = (mac_impl_t *)mh;
5045 
5046 	mcip->mci_upper_mip = mip;
5047 	/* If there are any properties, copy it over too */
5048 	if (mrp != NULL) {
5049 		bcopy(mrp, &mip->mi_resource_props,
5050 		    sizeof (mac_resource_props_t));
5051 	}
5052 }
5053 
5054 /*
5055  * Mark the mac as being used exclusively by the single mac client that is
5056  * doing some control operation on this mac. No further opens of this mac
5057  * will be allowed until this client calls mac_unmark_exclusive. The mac
5058  * client calling this function must already be in the mac perimeter
5059  */
5060 int
5061 mac_mark_exclusive(mac_handle_t mh)
5062 {
5063 	mac_impl_t	*mip = (mac_impl_t *)mh;
5064 
5065 	ASSERT(MAC_PERIM_HELD(mh));
5066 	/*
5067 	 * Look up its entry in the global hash table.
5068 	 */
5069 	rw_enter(&i_mac_impl_lock, RW_WRITER);
5070 	if (mip->mi_state_flags & MIS_DISABLED) {
5071 		rw_exit(&i_mac_impl_lock);
5072 		return (ENOENT);
5073 	}
5074 
5075 	/*
5076 	 * A reference to mac is held even if the link is not plumbed.
5077 	 * In i_dls_link_create() we open the MAC interface and hold the
5078 	 * reference. There is an additional reference for the mac_open
5079 	 * done in acquiring the mac perimeter
5080 	 */
5081 	if (mip->mi_ref != 2) {
5082 		rw_exit(&i_mac_impl_lock);
5083 		return (EBUSY);
5084 	}
5085 
5086 	ASSERT(!(mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5087 	mip->mi_state_flags |= MIS_EXCLUSIVE_HELD;
5088 	rw_exit(&i_mac_impl_lock);
5089 	return (0);
5090 }
5091 
5092 void
5093 mac_unmark_exclusive(mac_handle_t mh)
5094 {
5095 	mac_impl_t	*mip = (mac_impl_t *)mh;
5096 
5097 	ASSERT(MAC_PERIM_HELD(mh));
5098 
5099 	rw_enter(&i_mac_impl_lock, RW_WRITER);
5100 	/* 1 for the creation and another for the perimeter */
5101 	ASSERT(mip->mi_ref == 2 && (mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5102 	mip->mi_state_flags &= ~MIS_EXCLUSIVE_HELD;
5103 	rw_exit(&i_mac_impl_lock);
5104 }
5105 
5106 /*
5107  * Set the MTU for the specified MAC.
5108  */
5109 int
5110 mac_set_mtu(mac_handle_t mh, uint_t new_mtu, uint_t *old_mtu_arg)
5111 {
5112 	mac_impl_t *mip = (mac_impl_t *)mh;
5113 	uint_t old_mtu;
5114 	int rv = 0;
5115 
5116 	i_mac_perim_enter(mip);
5117 
5118 	if (!(mip->mi_callbacks->mc_callbacks & (MC_SETPROP|MC_GETPROP))) {
5119 		rv = ENOTSUP;
5120 		goto bail;
5121 	}
5122 
5123 	old_mtu = mip->mi_sdu_max;
5124 
5125 	if (new_mtu == 0 || new_mtu < mip->mi_sdu_min) {
5126 		rv = EINVAL;
5127 		goto bail;
5128 	}
5129 
5130 	if (old_mtu != new_mtu) {
5131 		rv = mip->mi_callbacks->mc_setprop(mip->mi_driver,
5132 		    "mtu", MAC_PROP_MTU, sizeof (uint_t), &new_mtu);
5133 		if (rv != 0)
5134 			goto bail;
5135 		rv = mac_maxsdu_update(mh, new_mtu);
5136 		ASSERT(rv == 0);
5137 	}
5138 
5139 bail:
5140 	i_mac_perim_exit(mip);
5141 
5142 	if (rv == 0 && old_mtu_arg != NULL)
5143 		*old_mtu_arg = old_mtu;
5144 	return (rv);
5145 }
5146 
5147 /*
5148  * Return the RX h/w information for the group indexed by grp_num.
5149  */
5150 void
5151 mac_get_hwrxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5152     uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5153     char *clnts_name)
5154 {
5155 	mac_impl_t *mip = (mac_impl_t *)mh;
5156 	mac_grp_client_t *mcip;
5157 	uint_t i = 0, index = 0;
5158 	mac_ring_t	*ring;
5159 
5160 	/* Revisit when we implement fully dynamic group allocation */
5161 	ASSERT(grp_index >= 0 && grp_index < mip->mi_rx_group_count);
5162 
5163 	rw_enter(&mip->mi_rw_lock, RW_READER);
5164 	*grp_num = mip->mi_rx_groups[grp_index].mrg_index;
5165 	*type = mip->mi_rx_groups[grp_index].mrg_type;
5166 	*n_rings = mip->mi_rx_groups[grp_index].mrg_cur_count;
5167 	ring = mip->mi_rx_groups[grp_index].mrg_rings;
5168 	for (index = 0; index < mip->mi_rx_groups[grp_index].mrg_cur_count;
5169 	    index++) {
5170 		rings[index] = ring->mr_index;
5171 		ring = ring->mr_next;
5172 	}
5173 	/* Assuming the 1st is the default group */
5174 	index = 0;
5175 	if (grp_index == 0) {
5176 		(void) strlcpy(clnts_name, "<default,mcast>,",
5177 		    MAXCLIENTNAMELEN);
5178 		index += strlen("<default,mcast>,");
5179 	}
5180 	for (mcip = mip->mi_rx_groups[grp_index].mrg_clients; mcip != NULL;
5181 	    mcip = mcip->mgc_next) {
5182 		int name_len = strlen(mcip->mgc_client->mci_name);
5183 
5184 		/*
5185 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
5186 		 * names.
5187 		 * XXXX Formating the client name string needs to be moved
5188 		 * to user land when fixing the size of dhi_clnts in
5189 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5190 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
5191 		 */
5192 		if (index + name_len >= MAXCLIENTNAMELEN) {
5193 			index = MAXCLIENTNAMELEN;
5194 			break;
5195 		}
5196 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5197 		    name_len);
5198 		index += name_len;
5199 		clnts_name[index++] = ',';
5200 		i++;
5201 	}
5202 
5203 	/* Get rid of the last , */
5204 	if (index > 0)
5205 		clnts_name[index - 1] = '\0';
5206 	*n_clnts = i;
5207 	rw_exit(&mip->mi_rw_lock);
5208 }
5209 
5210 /*
5211  * Return the TX h/w information for the group indexed by grp_num.
5212  */
5213 void
5214 mac_get_hwtxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5215     uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5216     char *clnts_name)
5217 {
5218 	mac_impl_t *mip = (mac_impl_t *)mh;
5219 	mac_grp_client_t *mcip;
5220 	uint_t i = 0, index = 0;
5221 	mac_ring_t	*ring;
5222 
5223 	/* Revisit when we implement fully dynamic group allocation */
5224 	ASSERT(grp_index >= 0 && grp_index <= mip->mi_tx_group_count);
5225 
5226 	rw_enter(&mip->mi_rw_lock, RW_READER);
5227 	*grp_num = mip->mi_tx_groups[grp_index].mrg_index > 0 ?
5228 	    mip->mi_tx_groups[grp_index].mrg_index : grp_index;
5229 	*type = mip->mi_tx_groups[grp_index].mrg_type;
5230 	*n_rings = mip->mi_tx_groups[grp_index].mrg_cur_count;
5231 	ring = mip->mi_tx_groups[grp_index].mrg_rings;
5232 	for (index = 0; index < mip->mi_tx_groups[grp_index].mrg_cur_count;
5233 	    index++) {
5234 		rings[index] = ring->mr_index;
5235 		ring = ring->mr_next;
5236 	}
5237 	index = 0;
5238 	/* Default group has an index of -1 */
5239 	if (mip->mi_tx_groups[grp_index].mrg_index < 0) {
5240 		(void) strlcpy(clnts_name, "<default>,",
5241 		    MAXCLIENTNAMELEN);
5242 		index += strlen("<default>,");
5243 	}
5244 	for (mcip = mip->mi_tx_groups[grp_index].mrg_clients; mcip != NULL;
5245 	    mcip = mcip->mgc_next) {
5246 		int name_len = strlen(mcip->mgc_client->mci_name);
5247 
5248 		/*
5249 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
5250 		 * names.
5251 		 * XXXX Formating the client name string needs to be moved
5252 		 * to user land when fixing the size of dhi_clnts in
5253 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5254 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
5255 		 */
5256 		if (index + name_len >= MAXCLIENTNAMELEN) {
5257 			index = MAXCLIENTNAMELEN;
5258 			break;
5259 		}
5260 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5261 		    name_len);
5262 		index += name_len;
5263 		clnts_name[index++] = ',';
5264 		i++;
5265 	}
5266 
5267 	/* Get rid of the last , */
5268 	if (index > 0)
5269 		clnts_name[index - 1] = '\0';
5270 	*n_clnts = i;
5271 	rw_exit(&mip->mi_rw_lock);
5272 }
5273 
5274 /*
5275  * Return the group count for RX or TX.
5276  */
5277 uint_t
5278 mac_hwgrp_num(mac_handle_t mh, int type)
5279 {
5280 	mac_impl_t *mip = (mac_impl_t *)mh;
5281 
5282 	/*
5283 	 * Return the Rx and Tx group count; for the Tx we need to
5284 	 * include the default too.
5285 	 */
5286 	return (type == MAC_RING_TYPE_RX ? mip->mi_rx_group_count :
5287 	    mip->mi_tx_groups != NULL ? mip->mi_tx_group_count + 1 : 0);
5288 }
5289 
5290 /*
5291  * The total number of free TX rings for this MAC.
5292  */
5293 uint_t
5294 mac_txavail_get(mac_handle_t mh)
5295 {
5296 	mac_impl_t	*mip = (mac_impl_t *)mh;
5297 
5298 	return (mip->mi_txrings_avail);
5299 }
5300 
5301 /*
5302  * The total number of free RX rings for this MAC.
5303  */
5304 uint_t
5305 mac_rxavail_get(mac_handle_t mh)
5306 {
5307 	mac_impl_t	*mip = (mac_impl_t *)mh;
5308 
5309 	return (mip->mi_rxrings_avail);
5310 }
5311 
5312 /*
5313  * The total number of reserved RX rings on this MAC.
5314  */
5315 uint_t
5316 mac_rxrsvd_get(mac_handle_t mh)
5317 {
5318 	mac_impl_t	*mip = (mac_impl_t *)mh;
5319 
5320 	return (mip->mi_rxrings_rsvd);
5321 }
5322 
5323 /*
5324  * The total number of reserved TX rings on this MAC.
5325  */
5326 uint_t
5327 mac_txrsvd_get(mac_handle_t mh)
5328 {
5329 	mac_impl_t	*mip = (mac_impl_t *)mh;
5330 
5331 	return (mip->mi_txrings_rsvd);
5332 }
5333 
5334 /*
5335  * Total number of free RX groups on this MAC.
5336  */
5337 uint_t
5338 mac_rxhwlnksavail_get(mac_handle_t mh)
5339 {
5340 	mac_impl_t	*mip = (mac_impl_t *)mh;
5341 
5342 	return (mip->mi_rxhwclnt_avail);
5343 }
5344 
5345 /*
5346  * Total number of RX groups reserved on this MAC.
5347  */
5348 uint_t
5349 mac_rxhwlnksrsvd_get(mac_handle_t mh)
5350 {
5351 	mac_impl_t	*mip = (mac_impl_t *)mh;
5352 
5353 	return (mip->mi_rxhwclnt_used);
5354 }
5355 
5356 /*
5357  * Total number of free TX groups on this MAC.
5358  */
5359 uint_t
5360 mac_txhwlnksavail_get(mac_handle_t mh)
5361 {
5362 	mac_impl_t	*mip = (mac_impl_t *)mh;
5363 
5364 	return (mip->mi_txhwclnt_avail);
5365 }
5366 
5367 /*
5368  * Total number of TX groups reserved on this MAC.
5369  */
5370 uint_t
5371 mac_txhwlnksrsvd_get(mac_handle_t mh)
5372 {
5373 	mac_impl_t	*mip = (mac_impl_t *)mh;
5374 
5375 	return (mip->mi_txhwclnt_used);
5376 }
5377 
5378 /*
5379  * Initialize the rings property for a mac client. A non-0 value for
5380  * rxring or txring specifies the number of rings required, a value
5381  * of MAC_RXRINGS_NONE/MAC_TXRINGS_NONE specifies that it doesn't need
5382  * any RX/TX rings and a value of MAC_RXRINGS_DONTCARE/MAC_TXRINGS_DONTCARE
5383  * means the system can decide whether it can give any rings or not.
5384  */
5385 void
5386 mac_client_set_rings(mac_client_handle_t mch, int rxrings, int txrings)
5387 {
5388 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
5389 	mac_resource_props_t	*mrp = MCIP_RESOURCE_PROPS(mcip);
5390 
5391 	if (rxrings != MAC_RXRINGS_DONTCARE) {
5392 		mrp->mrp_mask |= MRP_RX_RINGS;
5393 		mrp->mrp_nrxrings = rxrings;
5394 	}
5395 
5396 	if (txrings != MAC_TXRINGS_DONTCARE) {
5397 		mrp->mrp_mask |= MRP_TX_RINGS;
5398 		mrp->mrp_ntxrings = txrings;
5399 	}
5400 }
5401