xref: /titanic_41/usr/src/uts/common/io/mac/mac_client.c (revision 1e49577a7fcde812700ded04431b49d67cc57d6d)
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 (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24  */
25 
26 /*
27  * - General Introduction:
28  *
29  * This file contains the implementation of the MAC client kernel
30  * API and related code. The MAC client API allows a kernel module
31  * to gain access to a MAC instance (physical NIC, link aggregation, etc).
32  * It allows a MAC client to associate itself with a MAC address,
33  * VLANs, callback functions for data traffic and for promiscuous mode.
34  * The MAC client API is also used to specify the properties associated
35  * with a MAC client, such as bandwidth limits, priority, CPUS, etc.
36  * These properties are further used to determine the hardware resources
37  * to allocate to the various MAC clients.
38  *
39  * - Primary MAC clients:
40  *
41  * The MAC client API refers to "primary MAC clients". A primary MAC
42  * client is a client which "owns" the primary MAC address of
43  * the underlying MAC instance. The primary MAC address is called out
44  * since it is associated with specific semantics: the primary MAC
45  * address is the MAC address which is assigned to the IP interface
46  * when it is plumbed, and the primary MAC address is assigned
47  * to VLAN data-links. The primary address of a MAC instance can
48  * also change dynamically from under the MAC client, for example
49  * as a result of a change of state of a link aggregation. In that
50  * case the MAC layer automatically updates all data-structures which
51  * refer to the current value of the primary MAC address. Typical
52  * primary MAC clients are dls, aggr, and xnb. A typical non-primary
53  * MAC client is the vnic driver.
54  *
55  * - Virtual Switching:
56  *
57  * The MAC layer implements a virtual switch between the MAC clients
58  * (primary and non-primary) defined on top of the same underlying
59  * NIC (physical, link aggregation, etc). The virtual switch is
60  * VLAN-aware, i.e. it allows multiple MAC clients to be member
61  * of one or more VLANs, and the virtual switch will distribute
62  * multicast tagged packets only to the member of the corresponding
63  * VLANs.
64  *
65  * - Upper vs Lower MAC:
66  *
67  * Creating a VNIC on top of a MAC instance effectively causes
68  * two MAC instances to be layered on top of each other, one for
69  * the VNIC(s), one for the underlying MAC instance (physical NIC,
70  * link aggregation, etc). In the code below we refer to the
71  * underlying NIC as the "lower MAC", and we refer to VNICs as
72  * the "upper MAC".
73  *
74  * - Pass-through for VNICs:
75  *
76  * When VNICs are created on top of an underlying MAC, this causes
77  * a layering of two MAC instances. Since the lower MAC already
78  * does the switching and demultiplexing to its MAC clients, the
79  * upper MAC would simply have to pass packets to the layer below
80  * or above it, which would introduce overhead. In order to avoid
81  * this overhead, the MAC layer implements a pass-through mechanism
82  * for VNICs. When a VNIC opens the lower MAC instance, it saves
83  * the MAC client handle it optains from the MAC layer. When a MAC
84  * client opens a VNIC (upper MAC), the MAC layer detects that
85  * the MAC being opened is a VNIC, and gets the MAC client handle
86  * that the VNIC driver obtained from the lower MAC. This exchange
87  * is doing through a private capability between the MAC layer
88  * and the VNIC driver. The upper MAC then returns that handle
89  * directly to its MAC client. Any operation done by the upper
90  * MAC client is now done on the lower MAC client handle, which
91  * allows the VNIC driver to be completely bypassed for the
92  * performance sensitive data-path.
93  *
94  */
95 
96 #include <sys/types.h>
97 #include <sys/conf.h>
98 #include <sys/id_space.h>
99 #include <sys/esunddi.h>
100 #include <sys/stat.h>
101 #include <sys/mkdev.h>
102 #include <sys/stream.h>
103 #include <sys/strsun.h>
104 #include <sys/strsubr.h>
105 #include <sys/dlpi.h>
106 #include <sys/modhash.h>
107 #include <sys/mac_impl.h>
108 #include <sys/mac_client_impl.h>
109 #include <sys/mac_soft_ring.h>
110 #include <sys/mac_stat.h>
111 #include <sys/dls.h>
112 #include <sys/dld.h>
113 #include <sys/modctl.h>
114 #include <sys/fs/dv_node.h>
115 #include <sys/thread.h>
116 #include <sys/proc.h>
117 #include <sys/callb.h>
118 #include <sys/cpuvar.h>
119 #include <sys/atomic.h>
120 #include <sys/sdt.h>
121 #include <sys/mac_flow.h>
122 #include <sys/ddi_intr_impl.h>
123 #include <sys/disp.h>
124 #include <sys/sdt.h>
125 #include <sys/vnic.h>
126 #include <sys/vnic_impl.h>
127 #include <sys/vlan.h>
128 #include <inet/ip.h>
129 #include <inet/ip6.h>
130 #include <sys/exacct.h>
131 #include <sys/exacct_impl.h>
132 #include <inet/nd.h>
133 #include <sys/ethernet.h>
134 
135 kmem_cache_t	*mac_client_impl_cache;
136 kmem_cache_t	*mac_promisc_impl_cache;
137 
138 static boolean_t mac_client_single_rcvr(mac_client_impl_t *);
139 static flow_entry_t *mac_client_swap_mciflent(mac_client_impl_t *);
140 static flow_entry_t *mac_client_get_flow(mac_client_impl_t *,
141     mac_unicast_impl_t *);
142 static void mac_client_remove_flow_from_list(mac_client_impl_t *,
143     flow_entry_t *);
144 static void mac_client_add_to_flow_list(mac_client_impl_t *, flow_entry_t *);
145 static void mac_rename_flow_names(mac_client_impl_t *, const char *);
146 static void mac_virtual_link_update(mac_impl_t *);
147 static int mac_client_datapath_setup(mac_client_impl_t *, uint16_t,
148     uint8_t *, mac_resource_props_t *, boolean_t, mac_unicast_impl_t *);
149 static void mac_client_datapath_teardown(mac_client_handle_t,
150     mac_unicast_impl_t *, flow_entry_t *);
151 
152 /* ARGSUSED */
153 static int
154 i_mac_client_impl_ctor(void *buf, void *arg, int kmflag)
155 {
156 	int	i;
157 	mac_client_impl_t	*mcip = buf;
158 
159 	bzero(buf, MAC_CLIENT_IMPL_SIZE);
160 	mutex_init(&mcip->mci_tx_cb_lock, NULL, MUTEX_DRIVER, NULL);
161 	mcip->mci_tx_notify_cb_info.mcbi_lockp = &mcip->mci_tx_cb_lock;
162 
163 	ASSERT(mac_tx_percpu_cnt >= 0);
164 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
165 		mutex_init(&mcip->mci_tx_pcpu[i].pcpu_tx_lock, NULL,
166 		    MUTEX_DRIVER, NULL);
167 	}
168 	cv_init(&mcip->mci_tx_cv, NULL, CV_DRIVER, NULL);
169 
170 	return (0);
171 }
172 
173 /* ARGSUSED */
174 static void
175 i_mac_client_impl_dtor(void *buf, void *arg)
176 {
177 	int	i;
178 	mac_client_impl_t *mcip = buf;
179 
180 	ASSERT(mcip->mci_promisc_list == NULL);
181 	ASSERT(mcip->mci_unicast_list == NULL);
182 	ASSERT(mcip->mci_state_flags == 0);
183 	ASSERT(mcip->mci_tx_flag == 0);
184 
185 	mutex_destroy(&mcip->mci_tx_cb_lock);
186 
187 	ASSERT(mac_tx_percpu_cnt >= 0);
188 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
189 		ASSERT(mcip->mci_tx_pcpu[i].pcpu_tx_refcnt == 0);
190 		mutex_destroy(&mcip->mci_tx_pcpu[i].pcpu_tx_lock);
191 	}
192 	cv_destroy(&mcip->mci_tx_cv);
193 }
194 
195 /* ARGSUSED */
196 static int
197 i_mac_promisc_impl_ctor(void *buf, void *arg, int kmflag)
198 {
199 	mac_promisc_impl_t	*mpip = buf;
200 
201 	bzero(buf, sizeof (mac_promisc_impl_t));
202 	mpip->mpi_mci_link.mcb_objp = buf;
203 	mpip->mpi_mci_link.mcb_objsize = sizeof (mac_promisc_impl_t);
204 	mpip->mpi_mi_link.mcb_objp = buf;
205 	mpip->mpi_mi_link.mcb_objsize = sizeof (mac_promisc_impl_t);
206 	return (0);
207 }
208 
209 /* ARGSUSED */
210 static void
211 i_mac_promisc_impl_dtor(void *buf, void *arg)
212 {
213 	mac_promisc_impl_t	*mpip = buf;
214 
215 	ASSERT(mpip->mpi_mci_link.mcb_objp != NULL);
216 	ASSERT(mpip->mpi_mci_link.mcb_objsize == sizeof (mac_promisc_impl_t));
217 	ASSERT(mpip->mpi_mi_link.mcb_objp == mpip->mpi_mci_link.mcb_objp);
218 	ASSERT(mpip->mpi_mi_link.mcb_objsize == sizeof (mac_promisc_impl_t));
219 
220 	mpip->mpi_mci_link.mcb_objp = NULL;
221 	mpip->mpi_mci_link.mcb_objsize = 0;
222 	mpip->mpi_mi_link.mcb_objp = NULL;
223 	mpip->mpi_mi_link.mcb_objsize = 0;
224 
225 	ASSERT(mpip->mpi_mci_link.mcb_flags == 0);
226 	mpip->mpi_mci_link.mcb_objsize = 0;
227 }
228 
229 void
230 mac_client_init(void)
231 {
232 	ASSERT(mac_tx_percpu_cnt >= 0);
233 
234 	mac_client_impl_cache = kmem_cache_create("mac_client_impl_cache",
235 	    MAC_CLIENT_IMPL_SIZE, 0, i_mac_client_impl_ctor,
236 	    i_mac_client_impl_dtor, NULL, NULL, NULL, 0);
237 	ASSERT(mac_client_impl_cache != NULL);
238 
239 	mac_promisc_impl_cache = kmem_cache_create("mac_promisc_impl_cache",
240 	    sizeof (mac_promisc_impl_t), 0, i_mac_promisc_impl_ctor,
241 	    i_mac_promisc_impl_dtor, NULL, NULL, NULL, 0);
242 	ASSERT(mac_promisc_impl_cache != NULL);
243 }
244 
245 void
246 mac_client_fini(void)
247 {
248 	kmem_cache_destroy(mac_client_impl_cache);
249 	kmem_cache_destroy(mac_promisc_impl_cache);
250 }
251 
252 /*
253  * Return the lower MAC client handle from the VNIC driver for the
254  * specified VNIC MAC instance.
255  */
256 mac_client_impl_t *
257 mac_vnic_lower(mac_impl_t *mip)
258 {
259 	mac_capab_vnic_t cap;
260 	mac_client_impl_t *mcip;
261 
262 	VERIFY(i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_VNIC, &cap));
263 	mcip = cap.mcv_mac_client_handle(cap.mcv_arg);
264 
265 	return (mcip);
266 }
267 
268 /*
269  * Return the MAC client handle of the primary MAC client for the
270  * specified MAC instance, or NULL otherwise.
271  */
272 mac_client_impl_t *
273 mac_primary_client_handle(mac_impl_t *mip)
274 {
275 	mac_client_impl_t *mcip;
276 
277 	if (mip->mi_state_flags & MIS_IS_VNIC)
278 		return (mac_vnic_lower(mip));
279 
280 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
281 
282 	for (mcip = mip->mi_clients_list; mcip != NULL;
283 	    mcip = mcip->mci_client_next) {
284 		if (MCIP_DATAPATH_SETUP(mcip) && mac_is_primary_client(mcip))
285 			return (mcip);
286 	}
287 	return (NULL);
288 }
289 
290 /*
291  * Open a MAC specified by its MAC name.
292  */
293 int
294 mac_open(const char *macname, mac_handle_t *mhp)
295 {
296 	mac_impl_t	*mip;
297 	int		err;
298 
299 	/*
300 	 * Look up its entry in the global hash table.
301 	 */
302 	if ((err = mac_hold(macname, &mip)) != 0)
303 		return (err);
304 
305 	/*
306 	 * Hold the dip associated to the MAC to prevent it from being
307 	 * detached. For a softmac, its underlying dip is held by the
308 	 * mi_open() callback.
309 	 *
310 	 * This is done to be more tolerant with some defective drivers,
311 	 * which incorrectly handle mac_unregister() failure in their
312 	 * xxx_detach() routine. For example, some drivers ignore the
313 	 * failure of mac_unregister() and free all resources that
314 	 * that are needed for data transmition.
315 	 */
316 	e_ddi_hold_devi(mip->mi_dip);
317 
318 	if (!(mip->mi_callbacks->mc_callbacks & MC_OPEN)) {
319 		*mhp = (mac_handle_t)mip;
320 		return (0);
321 	}
322 
323 	/*
324 	 * The mac perimeter is used in both mac_open and mac_close by the
325 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
326 	 */
327 	i_mac_perim_enter(mip);
328 	mip->mi_oref++;
329 	if (mip->mi_oref != 1 || ((err = mip->mi_open(mip->mi_driver)) == 0)) {
330 		*mhp = (mac_handle_t)mip;
331 		i_mac_perim_exit(mip);
332 		return (0);
333 	}
334 	mip->mi_oref--;
335 	ddi_release_devi(mip->mi_dip);
336 	mac_rele(mip);
337 	i_mac_perim_exit(mip);
338 	return (err);
339 }
340 
341 /*
342  * Open a MAC specified by its linkid.
343  */
344 int
345 mac_open_by_linkid(datalink_id_t linkid, mac_handle_t *mhp)
346 {
347 	dls_dl_handle_t	dlh;
348 	int		err;
349 
350 	if ((err = dls_devnet_hold_tmp(linkid, &dlh)) != 0)
351 		return (err);
352 
353 	dls_devnet_prop_task_wait(dlh);
354 
355 	err = mac_open(dls_devnet_mac(dlh), mhp);
356 
357 	dls_devnet_rele_tmp(dlh);
358 	return (err);
359 }
360 
361 /*
362  * Open a MAC specified by its link name.
363  */
364 int
365 mac_open_by_linkname(const char *link, mac_handle_t *mhp)
366 {
367 	datalink_id_t	linkid;
368 	int		err;
369 
370 	if ((err = dls_mgmt_get_linkid(link, &linkid)) != 0)
371 		return (err);
372 	return (mac_open_by_linkid(linkid, mhp));
373 }
374 
375 /*
376  * Close the specified MAC.
377  */
378 void
379 mac_close(mac_handle_t mh)
380 {
381 	mac_impl_t	*mip = (mac_impl_t *)mh;
382 
383 	i_mac_perim_enter(mip);
384 	/*
385 	 * The mac perimeter is used in both mac_open and mac_close by the
386 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
387 	 */
388 	if (mip->mi_callbacks->mc_callbacks & MC_OPEN) {
389 		ASSERT(mip->mi_oref != 0);
390 		if (--mip->mi_oref == 0) {
391 			if ((mip->mi_callbacks->mc_callbacks & MC_CLOSE))
392 				mip->mi_close(mip->mi_driver);
393 		}
394 	}
395 	i_mac_perim_exit(mip);
396 	ddi_release_devi(mip->mi_dip);
397 	mac_rele(mip);
398 }
399 
400 /*
401  * Misc utility functions to retrieve various information about a MAC
402  * instance or a MAC client.
403  */
404 
405 const mac_info_t *
406 mac_info(mac_handle_t mh)
407 {
408 	return (&((mac_impl_t *)mh)->mi_info);
409 }
410 
411 dev_info_t *
412 mac_devinfo_get(mac_handle_t mh)
413 {
414 	return (((mac_impl_t *)mh)->mi_dip);
415 }
416 
417 void *
418 mac_driver(mac_handle_t mh)
419 {
420 	return (((mac_impl_t *)mh)->mi_driver);
421 }
422 
423 const char *
424 mac_name(mac_handle_t mh)
425 {
426 	return (((mac_impl_t *)mh)->mi_name);
427 }
428 
429 int
430 mac_type(mac_handle_t mh)
431 {
432 	return (((mac_impl_t *)mh)->mi_type->mt_type);
433 }
434 
435 int
436 mac_nativetype(mac_handle_t mh)
437 {
438 	return (((mac_impl_t *)mh)->mi_type->mt_nativetype);
439 }
440 
441 char *
442 mac_client_name(mac_client_handle_t mch)
443 {
444 	return (((mac_client_impl_t *)mch)->mci_name);
445 }
446 
447 minor_t
448 mac_minor(mac_handle_t mh)
449 {
450 	return (((mac_impl_t *)mh)->mi_minor);
451 }
452 
453 /*
454  * Return the VID associated with a MAC client. This function should
455  * be called for clients which are associated with only one VID.
456  */
457 uint16_t
458 mac_client_vid(mac_client_handle_t mch)
459 {
460 	uint16_t		vid = VLAN_ID_NONE;
461 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
462 	flow_desc_t		flow_desc;
463 
464 	if (mcip->mci_nflents == 0)
465 		return (vid);
466 
467 	ASSERT(MCIP_DATAPATH_SETUP(mcip) && mac_client_single_rcvr(mcip));
468 
469 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
470 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
471 		vid = flow_desc.fd_vid;
472 
473 	return (vid);
474 }
475 
476 /*
477  * Return whether the specified MAC client corresponds to a VLAN VNIC.
478  */
479 boolean_t
480 mac_client_is_vlan_vnic(mac_client_handle_t mch)
481 {
482 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
483 
484 	return (((mcip->mci_state_flags & MCIS_IS_VNIC) != 0) &&
485 	    ((mcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) != 0));
486 }
487 
488 /*
489  * Return the link speed associated with the specified MAC client.
490  *
491  * The link speed of a MAC client is equal to the smallest value of
492  * 1) the current link speed of the underlying NIC, or
493  * 2) the bandwidth limit set for the MAC client.
494  *
495  * Note that the bandwidth limit can be higher than the speed
496  * of the underlying NIC. This is allowed to avoid spurious
497  * administration action failures or artifically lowering the
498  * bandwidth limit of a link that may  have temporarily lowered
499  * its link speed due to hardware problem or administrator action.
500  */
501 static uint64_t
502 mac_client_ifspeed(mac_client_impl_t *mcip)
503 {
504 	mac_impl_t *mip = mcip->mci_mip;
505 	uint64_t nic_speed;
506 
507 	nic_speed = mac_stat_get((mac_handle_t)mip, MAC_STAT_IFSPEED);
508 
509 	if (nic_speed == 0) {
510 		return (0);
511 	} else {
512 		uint64_t policy_limit = (uint64_t)-1;
513 
514 		if (MCIP_RESOURCE_PROPS_MASK(mcip) & MRP_MAXBW)
515 			policy_limit = MCIP_RESOURCE_PROPS_MAXBW(mcip);
516 
517 		return (MIN(policy_limit, nic_speed));
518 	}
519 }
520 
521 /*
522  * Return the link state of the specified client. If here are more
523  * than one clients of the underying mac_impl_t, the link state
524  * will always be UP regardless of the link state of the underlying
525  * mac_impl_t. This is needed to allow the MAC clients to continue
526  * to communicate with each other even when the physical link of
527  * their mac_impl_t is down.
528  */
529 static uint64_t
530 mac_client_link_state(mac_client_impl_t *mcip)
531 {
532 	mac_impl_t *mip = mcip->mci_mip;
533 	uint16_t vid;
534 	mac_client_impl_t *mci_list;
535 	mac_unicast_impl_t *mui_list, *oth_mui_list;
536 
537 	/*
538 	 * Returns LINK_STATE_UP if there are other MAC clients defined on
539 	 * mac_impl_t which share same VLAN ID as that of mcip. Note that
540 	 * if 'mcip' has more than one VID's then we match ANY one of the
541 	 * VID's with other MAC client's VID's and return LINK_STATE_UP.
542 	 */
543 	rw_enter(&mcip->mci_rw_lock, RW_READER);
544 	for (mui_list = mcip->mci_unicast_list; mui_list != NULL;
545 	    mui_list = mui_list->mui_next) {
546 		vid = mui_list->mui_vid;
547 		for (mci_list = mip->mi_clients_list; mci_list != NULL;
548 		    mci_list = mci_list->mci_client_next) {
549 			if (mci_list == mcip)
550 				continue;
551 			for (oth_mui_list = mci_list->mci_unicast_list;
552 			    oth_mui_list != NULL; oth_mui_list = oth_mui_list->
553 			    mui_next) {
554 				if (vid == oth_mui_list->mui_vid) {
555 					rw_exit(&mcip->mci_rw_lock);
556 					return (LINK_STATE_UP);
557 				}
558 			}
559 		}
560 	}
561 	rw_exit(&mcip->mci_rw_lock);
562 
563 	return (mac_stat_get((mac_handle_t)mip, MAC_STAT_LINK_STATE));
564 }
565 
566 /*
567  * These statistics are consumed by dladm show-link -s <vnic>,
568  * dladm show-vnic -s and netstat. With the introduction of dlstat,
569  * dladm show-link -s and dladm show-vnic -s witll be EOL'ed while
570  * netstat will consume from kstats introduced for dlstat. This code
571  * will be removed at that time.
572  */
573 
574 /*
575  * Return the statistics of a MAC client. These statistics are different
576  * then the statistics of the underlying MAC which are returned by
577  * mac_stat_get().
578  */
579 uint64_t
580 mac_client_stat_get(mac_client_handle_t mch, uint_t stat)
581 {
582 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
583 	mac_impl_t 		*mip = mcip->mci_mip;
584 	flow_entry_t 		*flent = mcip->mci_flent;
585 	mac_soft_ring_set_t 	*mac_srs;
586 	mac_rx_stats_t		*mac_rx_stat;
587 	mac_tx_stats_t		*mac_tx_stat;
588 	int i;
589 	uint64_t val = 0;
590 
591 	mac_srs = (mac_soft_ring_set_t *)(flent->fe_tx_srs);
592 	mac_tx_stat = &mac_srs->srs_tx.st_stat;
593 
594 	switch (stat) {
595 	case MAC_STAT_LINK_STATE:
596 		val = mac_client_link_state(mcip);
597 		break;
598 	case MAC_STAT_LINK_UP:
599 		val = (mac_client_link_state(mcip) == LINK_STATE_UP);
600 		break;
601 	case MAC_STAT_PROMISC:
602 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_PROMISC);
603 		break;
604 	case MAC_STAT_LOWLINK_STATE:
605 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_LOWLINK_STATE);
606 		break;
607 	case MAC_STAT_IFSPEED:
608 		val = mac_client_ifspeed(mcip);
609 		break;
610 	case MAC_STAT_MULTIRCV:
611 		val = mcip->mci_misc_stat.mms_multircv;
612 		break;
613 	case MAC_STAT_BRDCSTRCV:
614 		val = mcip->mci_misc_stat.mms_brdcstrcv;
615 		break;
616 	case MAC_STAT_MULTIXMT:
617 		val = mcip->mci_misc_stat.mms_multixmt;
618 		break;
619 	case MAC_STAT_BRDCSTXMT:
620 		val = mcip->mci_misc_stat.mms_brdcstxmt;
621 		break;
622 	case MAC_STAT_OBYTES:
623 		val = mac_tx_stat->mts_obytes;
624 		break;
625 	case MAC_STAT_OPACKETS:
626 		val = mac_tx_stat->mts_opackets;
627 		break;
628 	case MAC_STAT_OERRORS:
629 		val = mac_tx_stat->mts_oerrors;
630 		break;
631 	case MAC_STAT_IPACKETS:
632 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
633 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
634 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
635 			val += mac_rx_stat->mrs_intrcnt +
636 			    mac_rx_stat->mrs_pollcnt + mac_rx_stat->mrs_lclcnt;
637 		}
638 		break;
639 	case MAC_STAT_RBYTES:
640 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
641 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
642 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
643 			val += mac_rx_stat->mrs_intrbytes +
644 			    mac_rx_stat->mrs_pollbytes +
645 			    mac_rx_stat->mrs_lclbytes;
646 		}
647 		break;
648 	case MAC_STAT_IERRORS:
649 		for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
650 			mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
651 			mac_rx_stat = &mac_srs->srs_rx.sr_stat;
652 			val += mac_rx_stat->mrs_ierrors;
653 		}
654 		break;
655 	default:
656 		val = mac_driver_stat_default(mip, stat);
657 		break;
658 	}
659 
660 	return (val);
661 }
662 
663 /*
664  * Return the statistics of the specified MAC instance.
665  */
666 uint64_t
667 mac_stat_get(mac_handle_t mh, uint_t stat)
668 {
669 	mac_impl_t	*mip = (mac_impl_t *)mh;
670 	uint64_t	val;
671 	int		ret;
672 
673 	/*
674 	 * The range of stat determines where it is maintained.  Stat
675 	 * values from 0 up to (but not including) MAC_STAT_MIN are
676 	 * mainteined by the mac module itself.  Everything else is
677 	 * maintained by the driver.
678 	 *
679 	 * If the mac_impl_t being queried corresponds to a VNIC,
680 	 * the stats need to be queried from the lower MAC client
681 	 * corresponding to the VNIC. (The mac_link_update()
682 	 * invoked by the driver to the lower MAC causes the *lower
683 	 * MAC* to update its mi_linkstate, and send a notification
684 	 * to its MAC clients. Due to the VNIC passthrough,
685 	 * these notifications are sent to the upper MAC clients
686 	 * of the VNIC directly, and the upper mac_impl_t of the VNIC
687 	 * does not have a valid mi_linkstate.
688 	 */
689 	if (stat < MAC_STAT_MIN && !(mip->mi_state_flags & MIS_IS_VNIC)) {
690 		/* these stats are maintained by the mac module itself */
691 		switch (stat) {
692 		case MAC_STAT_LINK_STATE:
693 			return (mip->mi_linkstate);
694 		case MAC_STAT_LINK_UP:
695 			return (mip->mi_linkstate == LINK_STATE_UP);
696 		case MAC_STAT_PROMISC:
697 			return (mip->mi_devpromisc != 0);
698 		case MAC_STAT_LOWLINK_STATE:
699 			return (mip->mi_lowlinkstate);
700 		default:
701 			ASSERT(B_FALSE);
702 		}
703 	}
704 
705 	/*
706 	 * Call the driver to get the given statistic.
707 	 */
708 	ret = mip->mi_getstat(mip->mi_driver, stat, &val);
709 	if (ret != 0) {
710 		/*
711 		 * The driver doesn't support this statistic.  Get the
712 		 * statistic's default value.
713 		 */
714 		val = mac_driver_stat_default(mip, stat);
715 	}
716 	return (val);
717 }
718 
719 /*
720  * Query hardware rx ring corresponding to the pseudo ring.
721  */
722 uint64_t
723 mac_pseudo_rx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
724 {
725 	return (mac_rx_ring_stat_get(handle, stat));
726 }
727 
728 /*
729  * Query hardware tx ring corresponding to the pseudo ring.
730  */
731 uint64_t
732 mac_pseudo_tx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
733 {
734 	return (mac_tx_ring_stat_get(handle, stat));
735 }
736 
737 /*
738  * Utility function which returns the VID associated with a flow entry.
739  */
740 uint16_t
741 i_mac_flow_vid(flow_entry_t *flent)
742 {
743 	flow_desc_t	flow_desc;
744 
745 	mac_flow_get_desc(flent, &flow_desc);
746 
747 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
748 		return (flow_desc.fd_vid);
749 	return (VLAN_ID_NONE);
750 }
751 
752 /*
753  * Verify the validity of the specified unicast MAC address. Returns B_TRUE
754  * if the address is valid, B_FALSE otherwise (multicast address, or incorrect
755  * length.
756  */
757 boolean_t
758 mac_unicst_verify(mac_handle_t mh, const uint8_t *addr, uint_t len)
759 {
760 	mac_impl_t	*mip = (mac_impl_t *)mh;
761 
762 	/*
763 	 * Verify the address. No lock is needed since mi_type and plugin
764 	 * details don't change after mac_register().
765 	 */
766 	if ((len != mip->mi_type->mt_addr_length) ||
767 	    (mip->mi_type->mt_ops.mtops_unicst_verify(addr,
768 	    mip->mi_pdata)) != 0) {
769 		return (B_FALSE);
770 	} else {
771 		return (B_TRUE);
772 	}
773 }
774 
775 void
776 mac_sdu_get(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu)
777 {
778 	mac_impl_t	*mip = (mac_impl_t *)mh;
779 
780 	if (min_sdu != NULL)
781 		*min_sdu = mip->mi_sdu_min;
782 	if (max_sdu != NULL)
783 		*max_sdu = mip->mi_sdu_max;
784 }
785 
786 /*
787  * Update the MAC unicast address of the specified client's flows. Currently
788  * only one unicast MAC unicast address is allowed per client.
789  */
790 static void
791 mac_unicast_update_client_flow(mac_client_impl_t *mcip)
792 {
793 	mac_impl_t *mip = mcip->mci_mip;
794 	flow_entry_t *flent = mcip->mci_flent;
795 	mac_address_t *map = mcip->mci_unicast;
796 	flow_desc_t flow_desc;
797 
798 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
799 	ASSERT(flent != NULL);
800 
801 	mac_flow_get_desc(flent, &flow_desc);
802 	ASSERT(flow_desc.fd_mask & FLOW_LINK_DST);
803 
804 	bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
805 	mac_flow_set_desc(flent, &flow_desc);
806 
807 	/*
808 	 * The v6 local addr (used by mac protection) needs to be
809 	 * regenerated because our mac address has changed.
810 	 */
811 	mac_protect_update_v6_local_addr(mcip);
812 
813 	/*
814 	 * A MAC client could have one MAC address but multiple
815 	 * VLANs. In that case update the flow entries corresponding
816 	 * to all VLANs of the MAC client.
817 	 */
818 	for (flent = mcip->mci_flent_list; flent != NULL;
819 	    flent = flent->fe_client_next) {
820 		mac_flow_get_desc(flent, &flow_desc);
821 		if (!(flent->fe_type & FLOW_PRIMARY_MAC ||
822 		    flent->fe_type & FLOW_VNIC_MAC))
823 			continue;
824 
825 		bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
826 		mac_flow_set_desc(flent, &flow_desc);
827 	}
828 }
829 
830 /*
831  * Update all clients that share the same unicast address.
832  */
833 void
834 mac_unicast_update_clients(mac_impl_t *mip, mac_address_t *map)
835 {
836 	mac_client_impl_t *mcip;
837 
838 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
839 
840 	/*
841 	 * Find all clients that share the same unicast MAC address and update
842 	 * them appropriately.
843 	 */
844 	for (mcip = mip->mi_clients_list; mcip != NULL;
845 	    mcip = mcip->mci_client_next) {
846 		/*
847 		 * Ignore clients that don't share this MAC address.
848 		 */
849 		if (map != mcip->mci_unicast)
850 			continue;
851 
852 		/*
853 		 * Update those clients with same old unicast MAC address.
854 		 */
855 		mac_unicast_update_client_flow(mcip);
856 	}
857 }
858 
859 /*
860  * Update the unicast MAC address of the specified VNIC MAC client.
861  *
862  * Check whether the operation is valid. Any of following cases should fail:
863  *
864  * 1. It's a VLAN type of VNIC.
865  * 2. The new value is current "primary" MAC address.
866  * 3. The current MAC address is shared with other clients.
867  * 4. The new MAC address has been used. This case will be valid when
868  *    client migration is fully supported.
869  */
870 int
871 mac_vnic_unicast_set(mac_client_handle_t mch, const uint8_t *addr)
872 {
873 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
874 	mac_impl_t *mip = mcip->mci_mip;
875 	mac_address_t *map = mcip->mci_unicast;
876 	int err;
877 
878 	ASSERT(!(mip->mi_state_flags & MIS_IS_VNIC));
879 	ASSERT(mcip->mci_state_flags & MCIS_IS_VNIC);
880 	ASSERT(mcip->mci_flags != MAC_CLIENT_FLAGS_PRIMARY);
881 
882 	i_mac_perim_enter(mip);
883 
884 	/*
885 	 * If this is a VLAN type of VNIC, it's using "primary" MAC address
886 	 * of the underlying interface. Must fail here. Refer to case 1 above.
887 	 */
888 	if (bcmp(map->ma_addr, mip->mi_addr, map->ma_len) == 0) {
889 		i_mac_perim_exit(mip);
890 		return (ENOTSUP);
891 	}
892 
893 	/*
894 	 * If the new address is the "primary" one, must fail. Refer to
895 	 * case 2 above.
896 	 */
897 	if (bcmp(addr, mip->mi_addr, map->ma_len) == 0) {
898 		i_mac_perim_exit(mip);
899 		return (EACCES);
900 	}
901 
902 	/*
903 	 * If the address is shared by multiple clients, must fail. Refer
904 	 * to case 3 above.
905 	 */
906 	if (mac_check_macaddr_shared(map)) {
907 		i_mac_perim_exit(mip);
908 		return (EBUSY);
909 	}
910 
911 	/*
912 	 * If the new address has been used, must fail for now. Refer to
913 	 * case 4 above.
914 	 */
915 	if (mac_find_macaddr(mip, (uint8_t *)addr) != NULL) {
916 		i_mac_perim_exit(mip);
917 		return (ENOTSUP);
918 	}
919 
920 	/*
921 	 * Update the MAC address.
922 	 */
923 	err = mac_update_macaddr(map, (uint8_t *)addr);
924 
925 	if (err != 0) {
926 		i_mac_perim_exit(mip);
927 		return (err);
928 	}
929 
930 	/*
931 	 * Update all flows of this MAC client.
932 	 */
933 	mac_unicast_update_client_flow(mcip);
934 
935 	i_mac_perim_exit(mip);
936 	return (0);
937 }
938 
939 /*
940  * Program the new primary unicast address of the specified MAC.
941  *
942  * Function mac_update_macaddr() takes care different types of underlying
943  * MAC. If the underlying MAC is VNIC, the VNIC driver must have registerd
944  * mi_unicst() entry point, that indirectly calls mac_vnic_unicast_set()
945  * which will take care of updating the MAC address of the corresponding
946  * MAC client.
947  *
948  * This is the only interface that allow the client to update the "primary"
949  * MAC address of the underlying MAC. The new value must have not been
950  * used by other clients.
951  */
952 int
953 mac_unicast_primary_set(mac_handle_t mh, const uint8_t *addr)
954 {
955 	mac_impl_t *mip = (mac_impl_t *)mh;
956 	mac_address_t *map;
957 	int err;
958 
959 	/* verify the address validity */
960 	if (!mac_unicst_verify(mh, addr, mip->mi_type->mt_addr_length))
961 		return (EINVAL);
962 
963 	i_mac_perim_enter(mip);
964 
965 	/*
966 	 * If the new value is the same as the current primary address value,
967 	 * there's nothing to do.
968 	 */
969 	if (bcmp(addr, mip->mi_addr, mip->mi_type->mt_addr_length) == 0) {
970 		i_mac_perim_exit(mip);
971 		return (0);
972 	}
973 
974 	if (mac_find_macaddr(mip, (uint8_t *)addr) != 0) {
975 		i_mac_perim_exit(mip);
976 		return (EBUSY);
977 	}
978 
979 	map = mac_find_macaddr(mip, mip->mi_addr);
980 	ASSERT(map != NULL);
981 
982 	/*
983 	 * Update the MAC address.
984 	 */
985 	if (mip->mi_state_flags & MIS_IS_AGGR) {
986 		mac_capab_aggr_t aggr_cap;
987 
988 		/*
989 		 * If the mac is an aggregation, other than the unicast
990 		 * addresses programming, aggr must be informed about this
991 		 * primary unicst address change to change its mac address
992 		 * policy to be user-specified.
993 		 */
994 		ASSERT(map->ma_type == MAC_ADDRESS_TYPE_UNICAST_CLASSIFIED);
995 		VERIFY(i_mac_capab_get(mh, MAC_CAPAB_AGGR, &aggr_cap));
996 		err = aggr_cap.mca_unicst(mip->mi_driver, addr);
997 		if (err == 0)
998 			bcopy(addr, map->ma_addr, map->ma_len);
999 	} else {
1000 		err = mac_update_macaddr(map, (uint8_t *)addr);
1001 	}
1002 
1003 	if (err != 0) {
1004 		i_mac_perim_exit(mip);
1005 		return (err);
1006 	}
1007 
1008 	mac_unicast_update_clients(mip, map);
1009 
1010 	/*
1011 	 * Save the new primary MAC address in mac_impl_t.
1012 	 */
1013 	bcopy(addr, mip->mi_addr, mip->mi_type->mt_addr_length);
1014 
1015 	i_mac_perim_exit(mip);
1016 
1017 	if (err == 0)
1018 		i_mac_notify(mip, MAC_NOTE_UNICST);
1019 
1020 	return (err);
1021 }
1022 
1023 /*
1024  * Return the current primary MAC address of the specified MAC.
1025  */
1026 void
1027 mac_unicast_primary_get(mac_handle_t mh, uint8_t *addr)
1028 {
1029 	mac_impl_t *mip = (mac_impl_t *)mh;
1030 
1031 	rw_enter(&mip->mi_rw_lock, RW_READER);
1032 	bcopy(mip->mi_addr, addr, mip->mi_type->mt_addr_length);
1033 	rw_exit(&mip->mi_rw_lock);
1034 }
1035 
1036 /*
1037  * Return information about the use of the primary MAC address of the
1038  * specified MAC instance:
1039  *
1040  * - if client_name is non-NULL, it must point to a string of at
1041  *   least MAXNAMELEN bytes, and will be set to the name of the MAC
1042  *   client which uses the primary MAC address.
1043  *
1044  * - if in_use is non-NULL, used to return whether the primary MAC
1045  *   address is currently in use.
1046  */
1047 void
1048 mac_unicast_primary_info(mac_handle_t mh, char *client_name, boolean_t *in_use)
1049 {
1050 	mac_impl_t *mip = (mac_impl_t *)mh;
1051 	mac_client_impl_t *cur_client;
1052 
1053 	if (in_use != NULL)
1054 		*in_use = B_FALSE;
1055 	if (client_name != NULL)
1056 		bzero(client_name, MAXNAMELEN);
1057 
1058 	/*
1059 	 * The mi_rw_lock is used to protect threads that don't hold the
1060 	 * mac perimeter to get a consistent view of the mi_clients_list.
1061 	 * Threads that modify the list must hold both the mac perimeter and
1062 	 * mi_rw_lock(RW_WRITER)
1063 	 */
1064 	rw_enter(&mip->mi_rw_lock, RW_READER);
1065 	for (cur_client = mip->mi_clients_list; cur_client != NULL;
1066 	    cur_client = cur_client->mci_client_next) {
1067 		if (mac_is_primary_client(cur_client) ||
1068 		    (mip->mi_state_flags & MIS_IS_VNIC)) {
1069 			rw_exit(&mip->mi_rw_lock);
1070 			if (in_use != NULL)
1071 				*in_use = B_TRUE;
1072 			if (client_name != NULL) {
1073 				bcopy(cur_client->mci_name, client_name,
1074 				    MAXNAMELEN);
1075 			}
1076 			return;
1077 		}
1078 	}
1079 	rw_exit(&mip->mi_rw_lock);
1080 }
1081 
1082 /*
1083  * Return the current destination MAC address of the specified MAC.
1084  */
1085 boolean_t
1086 mac_dst_get(mac_handle_t mh, uint8_t *addr)
1087 {
1088 	mac_impl_t *mip = (mac_impl_t *)mh;
1089 
1090 	rw_enter(&mip->mi_rw_lock, RW_READER);
1091 	if (mip->mi_dstaddr_set)
1092 		bcopy(mip->mi_dstaddr, addr, mip->mi_type->mt_addr_length);
1093 	rw_exit(&mip->mi_rw_lock);
1094 	return (mip->mi_dstaddr_set);
1095 }
1096 
1097 /*
1098  * Add the specified MAC client to the list of clients which opened
1099  * the specified MAC.
1100  */
1101 static void
1102 mac_client_add(mac_client_impl_t *mcip)
1103 {
1104 	mac_impl_t *mip = mcip->mci_mip;
1105 
1106 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1107 
1108 	/* add VNIC to the front of the list */
1109 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1110 	mcip->mci_client_next = mip->mi_clients_list;
1111 	mip->mi_clients_list = mcip;
1112 	mip->mi_nclients++;
1113 	rw_exit(&mip->mi_rw_lock);
1114 }
1115 
1116 /*
1117  * Remove the specified MAC client from the list of clients which opened
1118  * the specified MAC.
1119  */
1120 static void
1121 mac_client_remove(mac_client_impl_t *mcip)
1122 {
1123 	mac_impl_t *mip = mcip->mci_mip;
1124 	mac_client_impl_t **prev, *cclient;
1125 
1126 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1127 
1128 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1129 	prev = &mip->mi_clients_list;
1130 	cclient = *prev;
1131 	while (cclient != NULL && cclient != mcip) {
1132 		prev = &cclient->mci_client_next;
1133 		cclient = *prev;
1134 	}
1135 	ASSERT(cclient != NULL);
1136 	*prev = cclient->mci_client_next;
1137 	mip->mi_nclients--;
1138 	rw_exit(&mip->mi_rw_lock);
1139 }
1140 
1141 static mac_unicast_impl_t *
1142 mac_client_find_vid(mac_client_impl_t *mcip, uint16_t vid)
1143 {
1144 	mac_unicast_impl_t *muip = mcip->mci_unicast_list;
1145 
1146 	while ((muip != NULL) && (muip->mui_vid != vid))
1147 		muip = muip->mui_next;
1148 
1149 	return (muip);
1150 }
1151 
1152 /*
1153  * Return whether the specified (MAC address, VID) tuple is already used by
1154  * one of the MAC clients associated with the specified MAC.
1155  */
1156 static boolean_t
1157 mac_addr_in_use(mac_impl_t *mip, uint8_t *mac_addr, uint16_t vid)
1158 {
1159 	mac_client_impl_t *client;
1160 	mac_address_t *map;
1161 
1162 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1163 
1164 	for (client = mip->mi_clients_list; client != NULL;
1165 	    client = client->mci_client_next) {
1166 
1167 		/*
1168 		 * Ignore clients that don't have unicast address.
1169 		 */
1170 		if (client->mci_unicast_list == NULL)
1171 			continue;
1172 
1173 		map = client->mci_unicast;
1174 
1175 		if ((bcmp(mac_addr, map->ma_addr, map->ma_len) == 0) &&
1176 		    (mac_client_find_vid(client, vid) != NULL)) {
1177 			return (B_TRUE);
1178 		}
1179 	}
1180 
1181 	return (B_FALSE);
1182 }
1183 
1184 /*
1185  * Generate a random MAC address. The MAC address prefix is
1186  * stored in the array pointed to by mac_addr, and its length, in bytes,
1187  * is specified by prefix_len. The least significant bits
1188  * after prefix_len bytes are generated, and stored after the prefix
1189  * in the mac_addr array.
1190  */
1191 int
1192 mac_addr_random(mac_client_handle_t mch, uint_t prefix_len,
1193     uint8_t *mac_addr, mac_diag_t *diag)
1194 {
1195 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1196 	mac_impl_t *mip = mcip->mci_mip;
1197 	size_t addr_len = mip->mi_type->mt_addr_length;
1198 
1199 	if (prefix_len >= addr_len) {
1200 		*diag = MAC_DIAG_MACPREFIXLEN_INVALID;
1201 		return (EINVAL);
1202 	}
1203 
1204 	/* check the prefix value */
1205 	if (prefix_len > 0) {
1206 		bzero(mac_addr + prefix_len, addr_len - prefix_len);
1207 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr,
1208 		    addr_len)) {
1209 			*diag = MAC_DIAG_MACPREFIX_INVALID;
1210 			return (EINVAL);
1211 		}
1212 	}
1213 
1214 	/* generate the MAC address */
1215 	if (prefix_len < addr_len) {
1216 		(void) random_get_pseudo_bytes(mac_addr +
1217 		    prefix_len, addr_len - prefix_len);
1218 	}
1219 
1220 	*diag = 0;
1221 	return (0);
1222 }
1223 
1224 /*
1225  * Set the priority range for this MAC client. This will be used to
1226  * determine the absolute priority for the threads created for this
1227  * MAC client using the specified "low", "medium" and "high" level.
1228  * This will also be used for any subflows on this MAC client.
1229  */
1230 #define	MAC_CLIENT_SET_PRIORITY_RANGE(mcip, pri) {			\
1231 	(mcip)->mci_min_pri = FLOW_MIN_PRIORITY(MINCLSYSPRI,	\
1232 	    MAXCLSYSPRI, (pri));					\
1233 	(mcip)->mci_max_pri = FLOW_MAX_PRIORITY(MINCLSYSPRI,	\
1234 	    MAXCLSYSPRI, (mcip)->mci_min_pri);				\
1235 	}
1236 
1237 /*
1238  * MAC client open entry point. Return a new MAC client handle. Each
1239  * MAC client is associated with a name, specified through the 'name'
1240  * argument.
1241  */
1242 int
1243 mac_client_open(mac_handle_t mh, mac_client_handle_t *mchp, char *name,
1244     uint16_t flags)
1245 {
1246 	mac_impl_t		*mip = (mac_impl_t *)mh;
1247 	mac_client_impl_t	*mcip;
1248 	int			err = 0;
1249 	boolean_t		share_desired;
1250 	flow_entry_t		*flent = NULL;
1251 
1252 	share_desired = (flags & MAC_OPEN_FLAGS_SHARES_DESIRED) != 0;
1253 	*mchp = NULL;
1254 
1255 	i_mac_perim_enter(mip);
1256 
1257 	if (mip->mi_state_flags & MIS_IS_VNIC) {
1258 		/*
1259 		 * The underlying MAC is a VNIC. Return the MAC client
1260 		 * handle of the lower MAC which was obtained by
1261 		 * the VNIC driver when it did its mac_client_open().
1262 		 */
1263 
1264 		mcip = mac_vnic_lower(mip);
1265 
1266 		/*
1267 		 * Note that multiple mac clients share the same mcip in
1268 		 * this case.
1269 		 */
1270 		if (flags & MAC_OPEN_FLAGS_EXCLUSIVE)
1271 			mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1272 
1273 		if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1274 			mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1275 
1276 		mip->mi_clients_list = mcip;
1277 		i_mac_perim_exit(mip);
1278 		*mchp = (mac_client_handle_t)mcip;
1279 		return (err);
1280 	}
1281 
1282 	mcip = kmem_cache_alloc(mac_client_impl_cache, KM_SLEEP);
1283 
1284 	mcip->mci_mip = mip;
1285 	mcip->mci_upper_mip = NULL;
1286 	mcip->mci_rx_fn = mac_pkt_drop;
1287 	mcip->mci_rx_arg = NULL;
1288 	mcip->mci_rx_p_fn = NULL;
1289 	mcip->mci_rx_p_arg = NULL;
1290 	mcip->mci_p_unicast_list = NULL;
1291 	mcip->mci_direct_rx_fn = NULL;
1292 	mcip->mci_direct_rx_arg = NULL;
1293 
1294 	mcip->mci_unicast_list = NULL;
1295 
1296 	if ((flags & MAC_OPEN_FLAGS_IS_VNIC) != 0)
1297 		mcip->mci_state_flags |= MCIS_IS_VNIC;
1298 
1299 	if ((flags & MAC_OPEN_FLAGS_EXCLUSIVE) != 0)
1300 		mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1301 
1302 	if ((flags & MAC_OPEN_FLAGS_IS_AGGR_PORT) != 0)
1303 		mcip->mci_state_flags |= MCIS_IS_AGGR_PORT;
1304 
1305 	if (mip->mi_state_flags & MIS_IS_AGGR)
1306 		mcip->mci_state_flags |= MCIS_IS_AGGR;
1307 
1308 	if ((flags & MAC_OPEN_FLAGS_USE_DATALINK_NAME) != 0) {
1309 		datalink_id_t	linkid;
1310 
1311 		ASSERT(name == NULL);
1312 		if ((err = dls_devnet_macname2linkid(mip->mi_name,
1313 		    &linkid)) != 0) {
1314 			goto done;
1315 		}
1316 		if ((err = dls_mgmt_get_linkinfo(linkid, mcip->mci_name, NULL,
1317 		    NULL, NULL)) != 0) {
1318 			/*
1319 			 * Use mac name if dlmgmtd is not available.
1320 			 */
1321 			if (err == EBADF) {
1322 				(void) strlcpy(mcip->mci_name, mip->mi_name,
1323 				    sizeof (mcip->mci_name));
1324 				err = 0;
1325 			} else {
1326 				goto done;
1327 			}
1328 		}
1329 		mcip->mci_state_flags |= MCIS_USE_DATALINK_NAME;
1330 	} else {
1331 		ASSERT(name != NULL);
1332 		if (strlen(name) > MAXNAMELEN) {
1333 			err = EINVAL;
1334 			goto done;
1335 		}
1336 		(void) strlcpy(mcip->mci_name, name, sizeof (mcip->mci_name));
1337 	}
1338 
1339 	if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1340 		mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1341 
1342 	if (flags & MAC_OPEN_FLAGS_NO_UNICAST_ADDR)
1343 		mcip->mci_state_flags |= MCIS_NO_UNICAST_ADDR;
1344 
1345 	mac_protect_init(mcip);
1346 
1347 	/* the subflow table will be created dynamically */
1348 	mcip->mci_subflow_tab = NULL;
1349 
1350 	mcip->mci_misc_stat.mms_multircv = 0;
1351 	mcip->mci_misc_stat.mms_brdcstrcv = 0;
1352 	mcip->mci_misc_stat.mms_multixmt = 0;
1353 	mcip->mci_misc_stat.mms_brdcstxmt = 0;
1354 
1355 	/* Create an initial flow */
1356 
1357 	err = mac_flow_create(NULL, NULL, mcip->mci_name, NULL,
1358 	    mcip->mci_state_flags & MCIS_IS_VNIC ? FLOW_VNIC_MAC :
1359 	    FLOW_PRIMARY_MAC, &flent);
1360 	if (err != 0)
1361 		goto done;
1362 	mcip->mci_flent = flent;
1363 	FLOW_MARK(flent, FE_MC_NO_DATAPATH);
1364 	flent->fe_mcip = mcip;
1365 	/*
1366 	 * Place initial creation reference on the flow. This reference
1367 	 * is released in the corresponding delete action viz.
1368 	 * mac_unicast_remove after waiting for all transient refs to
1369 	 * to go away. The wait happens in mac_flow_wait.
1370 	 */
1371 	FLOW_REFHOLD(flent);
1372 
1373 	/*
1374 	 * Do this ahead of the mac_bcast_add() below so that the mi_nclients
1375 	 * will have the right value for mac_rx_srs_setup().
1376 	 */
1377 	mac_client_add(mcip);
1378 
1379 	mcip->mci_share = NULL;
1380 	if (share_desired)
1381 		i_mac_share_alloc(mcip);
1382 
1383 	DTRACE_PROBE2(mac__client__open__allocated, mac_impl_t *,
1384 	    mcip->mci_mip, mac_client_impl_t *, mcip);
1385 	*mchp = (mac_client_handle_t)mcip;
1386 
1387 	/*
1388 	 * We will do mimimal datapath setup to allow a MAC client to
1389 	 * transmit or receive non-unicast packets without waiting
1390 	 * for mac_unicast_add.
1391 	 */
1392 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1393 		if ((err = mac_client_datapath_setup(mcip, VLAN_ID_NONE,
1394 		    NULL, NULL, B_TRUE, NULL)) != 0) {
1395 			goto done;
1396 		}
1397 	}
1398 	i_mac_perim_exit(mip);
1399 	return (0);
1400 
1401 done:
1402 	i_mac_perim_exit(mip);
1403 	mcip->mci_state_flags = 0;
1404 	mcip->mci_tx_flag = 0;
1405 	kmem_cache_free(mac_client_impl_cache, mcip);
1406 	return (err);
1407 }
1408 
1409 /*
1410  * Close the specified MAC client handle.
1411  */
1412 void
1413 mac_client_close(mac_client_handle_t mch, uint16_t flags)
1414 {
1415 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1416 	mac_impl_t		*mip = mcip->mci_mip;
1417 	flow_entry_t		*flent;
1418 
1419 	i_mac_perim_enter(mip);
1420 
1421 	if (flags & MAC_CLOSE_FLAGS_EXCLUSIVE)
1422 		mcip->mci_state_flags &= ~MCIS_EXCLUSIVE;
1423 
1424 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
1425 	    !(flags & MAC_CLOSE_FLAGS_IS_VNIC)) {
1426 		/*
1427 		 * This is an upper VNIC client initiated operation.
1428 		 * The lower MAC client will be closed by the VNIC driver
1429 		 * when the VNIC is deleted.
1430 		 */
1431 
1432 		i_mac_perim_exit(mip);
1433 		return;
1434 	}
1435 
1436 	/* If we have only setup up minimal datapth setup, tear it down */
1437 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1438 		mac_client_datapath_teardown((mac_client_handle_t)mcip, NULL,
1439 		    mcip->mci_flent);
1440 		mcip->mci_state_flags &= ~MCIS_NO_UNICAST_ADDR;
1441 	}
1442 
1443 	/*
1444 	 * Remove the flent associated with the MAC client
1445 	 */
1446 	flent = mcip->mci_flent;
1447 	mcip->mci_flent = NULL;
1448 	FLOW_FINAL_REFRELE(flent);
1449 
1450 	/*
1451 	 * MAC clients must remove the unicast addresses and promisc callbacks
1452 	 * they added before issuing a mac_client_close().
1453 	 */
1454 	ASSERT(mcip->mci_unicast_list == NULL);
1455 	ASSERT(mcip->mci_promisc_list == NULL);
1456 	ASSERT(mcip->mci_tx_notify_cb_list == NULL);
1457 
1458 	i_mac_share_free(mcip);
1459 	mac_protect_fini(mcip);
1460 	mac_client_remove(mcip);
1461 
1462 	i_mac_perim_exit(mip);
1463 	mcip->mci_subflow_tab = NULL;
1464 	mcip->mci_state_flags = 0;
1465 	mcip->mci_tx_flag = 0;
1466 	kmem_cache_free(mac_client_impl_cache, mch);
1467 }
1468 
1469 /*
1470  * Set the rx bypass receive callback.
1471  */
1472 boolean_t
1473 mac_rx_bypass_set(mac_client_handle_t mch, mac_direct_rx_t rx_fn, void *arg1)
1474 {
1475 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1476 	mac_impl_t		*mip = mcip->mci_mip;
1477 
1478 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1479 
1480 	/*
1481 	 * If the mac_client is a VLAN, we should not do DLS bypass and
1482 	 * instead let the packets come up via mac_rx_deliver so the vlan
1483 	 * header can be stripped.
1484 	 */
1485 	if (mcip->mci_nvids > 0)
1486 		return (B_FALSE);
1487 
1488 	/*
1489 	 * These are not accessed directly in the data path, and hence
1490 	 * don't need any protection
1491 	 */
1492 	mcip->mci_direct_rx_fn = rx_fn;
1493 	mcip->mci_direct_rx_arg = arg1;
1494 	return (B_TRUE);
1495 }
1496 
1497 /*
1498  * Enable/Disable rx bypass. By default, bypass is assumed to be enabled.
1499  */
1500 void
1501 mac_rx_bypass_enable(mac_client_handle_t mch)
1502 {
1503 	((mac_client_impl_t *)mch)->mci_state_flags &= ~MCIS_RX_BYPASS_DISABLE;
1504 }
1505 
1506 void
1507 mac_rx_bypass_disable(mac_client_handle_t mch)
1508 {
1509 	((mac_client_impl_t *)mch)->mci_state_flags |= MCIS_RX_BYPASS_DISABLE;
1510 }
1511 
1512 /*
1513  * Set the receive callback for the specified MAC client. There can be
1514  * at most one such callback per MAC client.
1515  */
1516 void
1517 mac_rx_set(mac_client_handle_t mch, mac_rx_t rx_fn, void *arg)
1518 {
1519 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1520 	mac_impl_t	*mip = mcip->mci_mip;
1521 
1522 	/*
1523 	 * Instead of adding an extra set of locks and refcnts in
1524 	 * the datapath at the mac client boundary, we temporarily quiesce
1525 	 * the SRS and related entities. We then change the receive function
1526 	 * without interference from any receive data thread and then reenable
1527 	 * the data flow subsequently.
1528 	 */
1529 	i_mac_perim_enter(mip);
1530 	mac_rx_client_quiesce(mch);
1531 
1532 	mcip->mci_rx_fn = rx_fn;
1533 	mcip->mci_rx_arg = arg;
1534 	mac_rx_client_restart(mch);
1535 	i_mac_perim_exit(mip);
1536 }
1537 
1538 /*
1539  * Reset the receive callback for the specified MAC client.
1540  */
1541 void
1542 mac_rx_clear(mac_client_handle_t mch)
1543 {
1544 	mac_rx_set(mch, mac_pkt_drop, NULL);
1545 }
1546 
1547 /*
1548  * Walk the MAC client subflow table and updates their priority values.
1549  */
1550 static int
1551 mac_update_subflow_priority_cb(flow_entry_t *flent, void *arg)
1552 {
1553 	mac_flow_update_priority(arg, flent);
1554 	return (0);
1555 }
1556 
1557 void
1558 mac_update_subflow_priority(mac_client_impl_t *mcip)
1559 {
1560 	(void) mac_flow_walk(mcip->mci_subflow_tab,
1561 	    mac_update_subflow_priority_cb, mcip);
1562 }
1563 
1564 /*
1565  * Modify the TX or RX ring properties. We could either just move around
1566  * rings, i.e add/remove rings given to a client. Or this might cause the
1567  * client to move from hardware based to software or the other way around.
1568  * If we want to reset this property, then we clear the mask, additionally
1569  * if the client was given a non-default group we remove all rings except
1570  * for 1 and give it back to the default group.
1571  */
1572 int
1573 mac_client_set_rings_prop(mac_client_impl_t *mcip, mac_resource_props_t *mrp,
1574     mac_resource_props_t *tmrp)
1575 {
1576 	mac_impl_t		*mip = mcip->mci_mip;
1577 	flow_entry_t		*flent = mcip->mci_flent;
1578 	uint8_t			*mac_addr;
1579 	int			err = 0;
1580 	mac_group_t		*defgrp;
1581 	mac_group_t		*group;
1582 	mac_group_t		*ngrp;
1583 	mac_resource_props_t	*cmrp = MCIP_RESOURCE_PROPS(mcip);
1584 	uint_t			ringcnt;
1585 	boolean_t		unspec;
1586 
1587 	if (mcip->mci_share != NULL)
1588 		return (EINVAL);
1589 
1590 	if (mrp->mrp_mask & MRP_RX_RINGS) {
1591 		unspec = mrp->mrp_mask & MRP_RXRINGS_UNSPEC;
1592 		group = flent->fe_rx_ring_group;
1593 		defgrp = MAC_DEFAULT_RX_GROUP(mip);
1594 		mac_addr = flent->fe_flow_desc.fd_dst_mac;
1595 
1596 		/*
1597 		 * No resulting change. If we are resetting on a client on
1598 		 * which there was no rx rings property. For dynamic group
1599 		 * if we are setting the same number of rings already set.
1600 		 * For static group if we are requesting a group again.
1601 		 */
1602 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1603 			if (!(tmrp->mrp_mask & MRP_RX_RINGS))
1604 				return (0);
1605 		} else {
1606 			if (unspec) {
1607 				if (tmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1608 					return (0);
1609 			} else if (mip->mi_rx_group_type ==
1610 			    MAC_GROUP_TYPE_DYNAMIC) {
1611 				if ((tmrp->mrp_mask & MRP_RX_RINGS) &&
1612 				    !(tmrp->mrp_mask & MRP_RXRINGS_UNSPEC) &&
1613 				    mrp->mrp_nrxrings == tmrp->mrp_nrxrings) {
1614 					return (0);
1615 				}
1616 			}
1617 		}
1618 		/* Resetting the prop */
1619 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1620 			/*
1621 			 * We will just keep one ring and give others back if
1622 			 * we are not the primary. For the primary we give
1623 			 * all the rings in the default group except the
1624 			 * default ring. If it is a static group, then
1625 			 * we don't do anything, but clear the MRP_RX_RINGS
1626 			 * flag.
1627 			 */
1628 			if (group != defgrp) {
1629 				if (mip->mi_rx_group_type ==
1630 				    MAC_GROUP_TYPE_DYNAMIC) {
1631 					/*
1632 					 * This group has reserved rings
1633 					 * that need to be released now,
1634 					 * so does the group.
1635 					 */
1636 					MAC_RX_RING_RELEASED(mip,
1637 					    group->mrg_cur_count);
1638 					MAC_RX_GRP_RELEASED(mip);
1639 					if ((flent->fe_type &
1640 					    FLOW_PRIMARY_MAC) != 0) {
1641 						if (mip->mi_nactiveclients ==
1642 						    1) {
1643 							(void)
1644 							    mac_rx_switch_group(
1645 							    mcip, group,
1646 							    defgrp);
1647 							return (0);
1648 						} else {
1649 							cmrp->mrp_nrxrings =
1650 							    group->
1651 							    mrg_cur_count +
1652 							    defgrp->
1653 							    mrg_cur_count - 1;
1654 						}
1655 					} else {
1656 						cmrp->mrp_nrxrings = 1;
1657 					}
1658 					(void) mac_group_ring_modify(mcip,
1659 					    group, defgrp);
1660 				} else {
1661 					/*
1662 					 * If this is a static group, we
1663 					 * need to release the group. The
1664 					 * client will remain in the same
1665 					 * group till some other client
1666 					 * needs this group.
1667 					 */
1668 					MAC_RX_GRP_RELEASED(mip);
1669 				}
1670 			/* Let check if we can give this an excl group */
1671 			} else if (group == defgrp) {
1672 				ngrp = 	mac_reserve_rx_group(mcip, mac_addr,
1673 				    B_TRUE);
1674 				/* Couldn't give it a group, that's fine */
1675 				if (ngrp == NULL)
1676 					return (0);
1677 				/* Switch to H/W */
1678 				if (mac_rx_switch_group(mcip, defgrp, ngrp) !=
1679 				    0) {
1680 					mac_stop_group(ngrp);
1681 					return (0);
1682 				}
1683 			}
1684 			/*
1685 			 * If the client is in the default group, we will
1686 			 * just clear the MRP_RX_RINGS and leave it as
1687 			 * it rather than look for an exclusive group
1688 			 * for it.
1689 			 */
1690 			return (0);
1691 		}
1692 
1693 		if (group == defgrp && ((mrp->mrp_nrxrings > 0) || unspec)) {
1694 			ngrp = 	mac_reserve_rx_group(mcip, mac_addr, B_TRUE);
1695 			if (ngrp == NULL)
1696 				return (ENOSPC);
1697 
1698 			/* Switch to H/W */
1699 			if (mac_rx_switch_group(mcip, defgrp, ngrp) != 0) {
1700 				mac_release_rx_group(mcip, ngrp);
1701 				return (ENOSPC);
1702 			}
1703 			MAC_RX_GRP_RESERVED(mip);
1704 			if (mip->mi_rx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1705 				MAC_RX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1706 		} else if (group != defgrp && !unspec &&
1707 		    mrp->mrp_nrxrings == 0) {
1708 			/* Switch to S/W */
1709 			ringcnt = group->mrg_cur_count;
1710 			if (mac_rx_switch_group(mcip, group, defgrp) != 0)
1711 				return (ENOSPC);
1712 			if (tmrp->mrp_mask & MRP_RX_RINGS) {
1713 				MAC_RX_GRP_RELEASED(mip);
1714 				if (mip->mi_rx_group_type ==
1715 				    MAC_GROUP_TYPE_DYNAMIC) {
1716 					MAC_RX_RING_RELEASED(mip, ringcnt);
1717 				}
1718 			}
1719 		} else if (group != defgrp && mip->mi_rx_group_type ==
1720 		    MAC_GROUP_TYPE_DYNAMIC) {
1721 			ringcnt = group->mrg_cur_count;
1722 			err = mac_group_ring_modify(mcip, group, defgrp);
1723 			if (err != 0)
1724 				return (err);
1725 			/*
1726 			 * Update the accounting. If this group
1727 			 * already had explicitly reserved rings,
1728 			 * we need to update the rings based on
1729 			 * the new ring count. If this group
1730 			 * had not explicitly reserved rings,
1731 			 * then we just reserve the rings asked for
1732 			 * and reserve the group.
1733 			 */
1734 			if (tmrp->mrp_mask & MRP_RX_RINGS) {
1735 				if (ringcnt > group->mrg_cur_count) {
1736 					MAC_RX_RING_RELEASED(mip,
1737 					    ringcnt - group->mrg_cur_count);
1738 				} else {
1739 					MAC_RX_RING_RESERVED(mip,
1740 					    group->mrg_cur_count - ringcnt);
1741 				}
1742 			} else {
1743 				MAC_RX_RING_RESERVED(mip, group->mrg_cur_count);
1744 				MAC_RX_GRP_RESERVED(mip);
1745 			}
1746 		}
1747 	}
1748 	if (mrp->mrp_mask & MRP_TX_RINGS) {
1749 		unspec = mrp->mrp_mask & MRP_TXRINGS_UNSPEC;
1750 		group = flent->fe_tx_ring_group;
1751 		defgrp = MAC_DEFAULT_TX_GROUP(mip);
1752 
1753 		/*
1754 		 * For static groups we only allow rings=0 or resetting the
1755 		 * rings property.
1756 		 */
1757 		if (mrp->mrp_ntxrings > 0 &&
1758 		    mip->mi_tx_group_type != MAC_GROUP_TYPE_DYNAMIC) {
1759 			return (ENOTSUP);
1760 		}
1761 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1762 			if (!(tmrp->mrp_mask & MRP_TX_RINGS))
1763 				return (0);
1764 		} else {
1765 			if (unspec) {
1766 				if (tmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1767 					return (0);
1768 			} else if (mip->mi_tx_group_type ==
1769 			    MAC_GROUP_TYPE_DYNAMIC) {
1770 				if ((tmrp->mrp_mask & MRP_TX_RINGS) &&
1771 				    !(tmrp->mrp_mask & MRP_TXRINGS_UNSPEC) &&
1772 				    mrp->mrp_ntxrings == tmrp->mrp_ntxrings) {
1773 					return (0);
1774 				}
1775 			}
1776 		}
1777 		/* Resetting the prop */
1778 		if (mrp->mrp_mask & MRP_RINGS_RESET) {
1779 			if (group != defgrp) {
1780 				if (mip->mi_tx_group_type ==
1781 				    MAC_GROUP_TYPE_DYNAMIC) {
1782 					ringcnt = group->mrg_cur_count;
1783 					if ((flent->fe_type &
1784 					    FLOW_PRIMARY_MAC) != 0) {
1785 						mac_tx_client_quiesce(
1786 						    (mac_client_handle_t)
1787 						    mcip);
1788 						mac_tx_switch_group(mcip,
1789 						    group, defgrp);
1790 						mac_tx_client_restart(
1791 						    (mac_client_handle_t)
1792 						    mcip);
1793 						MAC_TX_GRP_RELEASED(mip);
1794 						MAC_TX_RING_RELEASED(mip,
1795 						    ringcnt);
1796 						return (0);
1797 					}
1798 					cmrp->mrp_ntxrings = 1;
1799 					(void) mac_group_ring_modify(mcip,
1800 					    group, defgrp);
1801 					/*
1802 					 * This group has reserved rings
1803 					 * that need to be released now.
1804 					 */
1805 					MAC_TX_RING_RELEASED(mip, ringcnt);
1806 				}
1807 				/*
1808 				 * If this is a static group, we
1809 				 * need to release the group. The
1810 				 * client will remain in the same
1811 				 * group till some other client
1812 				 * needs this group.
1813 				 */
1814 				MAC_TX_GRP_RELEASED(mip);
1815 			} else if (group == defgrp &&
1816 			    (flent->fe_type & FLOW_PRIMARY_MAC) == 0) {
1817 				ngrp = mac_reserve_tx_group(mcip, B_TRUE);
1818 				if (ngrp == NULL)
1819 					return (0);
1820 				mac_tx_client_quiesce(
1821 				    (mac_client_handle_t)mcip);
1822 				mac_tx_switch_group(mcip, defgrp, ngrp);
1823 				mac_tx_client_restart(
1824 				    (mac_client_handle_t)mcip);
1825 			}
1826 			/*
1827 			 * If the client is in the default group, we will
1828 			 * just clear the MRP_TX_RINGS and leave it as
1829 			 * it rather than look for an exclusive group
1830 			 * for it.
1831 			 */
1832 			return (0);
1833 		}
1834 
1835 		/* Switch to H/W */
1836 		if (group == defgrp && ((mrp->mrp_ntxrings > 0) || unspec)) {
1837 			ngrp = 	mac_reserve_tx_group(mcip, B_TRUE);
1838 			if (ngrp == NULL)
1839 				return (ENOSPC);
1840 			mac_tx_client_quiesce((mac_client_handle_t)mcip);
1841 			mac_tx_switch_group(mcip, defgrp, ngrp);
1842 			mac_tx_client_restart((mac_client_handle_t)mcip);
1843 			MAC_TX_GRP_RESERVED(mip);
1844 			if (mip->mi_tx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1845 				MAC_TX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1846 		/* Switch to S/W */
1847 		} else if (group != defgrp && !unspec &&
1848 		    mrp->mrp_ntxrings == 0) {
1849 			/* Switch to S/W */
1850 			ringcnt = group->mrg_cur_count;
1851 			mac_tx_client_quiesce((mac_client_handle_t)mcip);
1852 			mac_tx_switch_group(mcip, group, defgrp);
1853 			mac_tx_client_restart((mac_client_handle_t)mcip);
1854 			if (tmrp->mrp_mask & MRP_TX_RINGS) {
1855 				MAC_TX_GRP_RELEASED(mip);
1856 				if (mip->mi_tx_group_type ==
1857 				    MAC_GROUP_TYPE_DYNAMIC) {
1858 					MAC_TX_RING_RELEASED(mip, ringcnt);
1859 				}
1860 			}
1861 		} else if (group != defgrp && mip->mi_tx_group_type ==
1862 		    MAC_GROUP_TYPE_DYNAMIC) {
1863 			ringcnt = group->mrg_cur_count;
1864 			err = mac_group_ring_modify(mcip, group, defgrp);
1865 			if (err != 0)
1866 				return (err);
1867 			/*
1868 			 * Update the accounting. If this group
1869 			 * already had explicitly reserved rings,
1870 			 * we need to update the rings based on
1871 			 * the new ring count. If this group
1872 			 * had not explicitly reserved rings,
1873 			 * then we just reserve the rings asked for
1874 			 * and reserve the group.
1875 			 */
1876 			if (tmrp->mrp_mask & MRP_TX_RINGS) {
1877 				if (ringcnt > group->mrg_cur_count) {
1878 					MAC_TX_RING_RELEASED(mip,
1879 					    ringcnt - group->mrg_cur_count);
1880 				} else {
1881 					MAC_TX_RING_RESERVED(mip,
1882 					    group->mrg_cur_count - ringcnt);
1883 				}
1884 			} else {
1885 				MAC_TX_RING_RESERVED(mip, group->mrg_cur_count);
1886 				MAC_TX_GRP_RESERVED(mip);
1887 			}
1888 		}
1889 	}
1890 	return (0);
1891 }
1892 
1893 /*
1894  * When the MAC client is being brought up (i.e. we do a unicast_add) we need
1895  * to initialize the cpu and resource control structure in the
1896  * mac_client_impl_t from the mac_impl_t (i.e if there are any cached
1897  * properties before the flow entry for the unicast address was created).
1898  */
1899 int
1900 mac_resource_ctl_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
1901 {
1902 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
1903 	mac_impl_t		*mip = (mac_impl_t *)mcip->mci_mip;
1904 	int			err = 0;
1905 	flow_entry_t		*flent = mcip->mci_flent;
1906 	mac_resource_props_t	*omrp, *nmrp = MCIP_RESOURCE_PROPS(mcip);
1907 
1908 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1909 
1910 	err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
1911 	    mcip->mci_upper_mip : mip, mrp);
1912 	if (err != 0)
1913 		return (err);
1914 
1915 	/*
1916 	 * Copy over the existing properties since mac_update_resources
1917 	 * will modify the client's mrp. Currently, the saved property
1918 	 * is used to determine the difference between existing and
1919 	 * modified rings property.
1920 	 */
1921 	omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
1922 	bcopy(nmrp, omrp, sizeof (*omrp));
1923 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
1924 	if (MCIP_DATAPATH_SETUP(mcip)) {
1925 		/*
1926 		 * We support rings only for primary client when there are
1927 		 * multiple clients sharing the same MAC address (e.g. VLAN).
1928 		 */
1929 		if (mrp->mrp_mask & MRP_RX_RINGS ||
1930 		    mrp->mrp_mask & MRP_TX_RINGS) {
1931 
1932 			if ((err = mac_client_set_rings_prop(mcip, mrp,
1933 			    omrp)) != 0) {
1934 				if (omrp->mrp_mask & MRP_RX_RINGS) {
1935 					nmrp->mrp_mask |= MRP_RX_RINGS;
1936 					nmrp->mrp_nrxrings = omrp->mrp_nrxrings;
1937 				} else {
1938 					nmrp->mrp_mask &= ~MRP_RX_RINGS;
1939 					nmrp->mrp_nrxrings = 0;
1940 				}
1941 				if (omrp->mrp_mask & MRP_TX_RINGS) {
1942 					nmrp->mrp_mask |= MRP_TX_RINGS;
1943 					nmrp->mrp_ntxrings = omrp->mrp_ntxrings;
1944 				} else {
1945 					nmrp->mrp_mask &= ~MRP_TX_RINGS;
1946 					nmrp->mrp_ntxrings = 0;
1947 				}
1948 				if (omrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1949 					omrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
1950 				else
1951 					omrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
1952 
1953 				if (omrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1954 					omrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
1955 				else
1956 					omrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
1957 				kmem_free(omrp, sizeof (*omrp));
1958 				return (err);
1959 			}
1960 
1961 			/*
1962 			 * If we modified the rings property of the primary
1963 			 * we need to update the property fields of its
1964 			 * VLANs as they inherit the primary's properites.
1965 			 */
1966 			if (mac_is_primary_client(mcip)) {
1967 				mac_set_prim_vlan_rings(mip,
1968 				    MCIP_RESOURCE_PROPS(mcip));
1969 			}
1970 		}
1971 		/*
1972 		 * We have to set this prior to calling mac_flow_modify.
1973 		 */
1974 		if (mrp->mrp_mask & MRP_PRIORITY) {
1975 			if (mrp->mrp_priority == MPL_RESET) {
1976 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1977 				    MPL_LINK_DEFAULT);
1978 			} else {
1979 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1980 				    mrp->mrp_priority);
1981 			}
1982 		}
1983 
1984 		mac_flow_modify(mip->mi_flow_tab, flent, mrp);
1985 		if (mrp->mrp_mask & MRP_PRIORITY)
1986 			mac_update_subflow_priority(mcip);
1987 	}
1988 	kmem_free(omrp, sizeof (*omrp));
1989 	return (0);
1990 }
1991 
1992 void
1993 mac_resource_ctl_get(mac_client_handle_t mch, mac_resource_props_t *mrp)
1994 {
1995 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1996 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
1997 
1998 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
1999 }
2000 
2001 static int
2002 mac_unicast_flow_create(mac_client_impl_t *mcip, uint8_t *mac_addr,
2003     uint16_t vid, boolean_t is_primary, boolean_t first_flow,
2004     flow_entry_t **flent, mac_resource_props_t *mrp)
2005 {
2006 	mac_impl_t	*mip = (mac_impl_t *)mcip->mci_mip;
2007 	flow_desc_t	flow_desc;
2008 	char		flowname[MAXFLOWNAMELEN];
2009 	int		err;
2010 	uint_t		flent_flags;
2011 
2012 	/*
2013 	 * First unicast address being added, create a new flow
2014 	 * for that MAC client.
2015 	 */
2016 	bzero(&flow_desc, sizeof (flow_desc));
2017 
2018 	ASSERT(mac_addr != NULL ||
2019 	    (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR));
2020 	if (mac_addr != NULL) {
2021 		flow_desc.fd_mac_len = mip->mi_type->mt_addr_length;
2022 		bcopy(mac_addr, flow_desc.fd_dst_mac, flow_desc.fd_mac_len);
2023 	}
2024 	flow_desc.fd_mask = FLOW_LINK_DST;
2025 	if (vid != 0) {
2026 		flow_desc.fd_vid = vid;
2027 		flow_desc.fd_mask |= FLOW_LINK_VID;
2028 	}
2029 
2030 	/*
2031 	 * XXX-nicolas. For now I'm keeping the FLOW_PRIMARY_MAC
2032 	 * and FLOW_VNIC. Even though they're a hack inherited
2033 	 * from the SRS code, we'll keep them for now. They're currently
2034 	 * consumed by mac_datapath_setup() to create the SRS.
2035 	 * That code should be eventually moved out of
2036 	 * mac_datapath_setup() and moved to a mac_srs_create()
2037 	 * function of some sort to keep things clean.
2038 	 *
2039 	 * Also, there's no reason why the SRS for the primary MAC
2040 	 * client should be different than any other MAC client. Until
2041 	 * this is cleaned-up, we support only one MAC unicast address
2042 	 * per client.
2043 	 *
2044 	 * We set FLOW_PRIMARY_MAC for the primary MAC address,
2045 	 * FLOW_VNIC for everything else.
2046 	 */
2047 	if (is_primary)
2048 		flent_flags = FLOW_PRIMARY_MAC;
2049 	else
2050 		flent_flags = FLOW_VNIC_MAC;
2051 
2052 	/*
2053 	 * For the first flow we use the mac client's name - mci_name, for
2054 	 * subsequent ones we just create a name with the vid. This is
2055 	 * so that we can add these flows to the same flow table. This is
2056 	 * fine as the flow name (except for the one with the mac client's
2057 	 * name) is not visible. When the first flow is removed, we just replace
2058 	 * its fdesc with another from the list, so we will still retain the
2059 	 * flent with the MAC client's flow name.
2060 	 */
2061 	if (first_flow) {
2062 		bcopy(mcip->mci_name, flowname, MAXFLOWNAMELEN);
2063 	} else {
2064 		(void) sprintf(flowname, "%s%u", mcip->mci_name, vid);
2065 		flent_flags = FLOW_NO_STATS;
2066 	}
2067 
2068 	if ((err = mac_flow_create(&flow_desc, mrp, flowname, NULL,
2069 	    flent_flags, flent)) != 0)
2070 		return (err);
2071 
2072 	mac_misc_stat_create(*flent);
2073 	FLOW_MARK(*flent, FE_INCIPIENT);
2074 	(*flent)->fe_mcip = mcip;
2075 
2076 	/*
2077 	 * Place initial creation reference on the flow. This reference
2078 	 * is released in the corresponding delete action viz.
2079 	 * mac_unicast_remove after waiting for all transient refs to
2080 	 * to go away. The wait happens in mac_flow_wait.
2081 	 * We have already held the reference in mac_client_open().
2082 	 */
2083 	if (!first_flow)
2084 		FLOW_REFHOLD(*flent);
2085 	return (0);
2086 }
2087 
2088 /* Refresh the multicast grouping for this VID. */
2089 int
2090 mac_client_update_mcast(void *arg, boolean_t add, const uint8_t *addrp)
2091 {
2092 	flow_entry_t		*flent = arg;
2093 	mac_client_impl_t	*mcip = flent->fe_mcip;
2094 	uint16_t		vid;
2095 	flow_desc_t		flow_desc;
2096 
2097 	mac_flow_get_desc(flent, &flow_desc);
2098 	vid = (flow_desc.fd_mask & FLOW_LINK_VID) != 0 ?
2099 	    flow_desc.fd_vid : VLAN_ID_NONE;
2100 
2101 	/*
2102 	 * We don't call mac_multicast_add()/mac_multicast_remove() as
2103 	 * we want to add/remove for this specific vid.
2104 	 */
2105 	if (add) {
2106 		return (mac_bcast_add(mcip, addrp, vid,
2107 		    MAC_ADDRTYPE_MULTICAST));
2108 	} else {
2109 		mac_bcast_delete(mcip, addrp, vid);
2110 		return (0);
2111 	}
2112 }
2113 
2114 static void
2115 mac_update_single_active_client(mac_impl_t *mip)
2116 {
2117 	mac_client_impl_t *client = NULL;
2118 
2119 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
2120 
2121 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2122 	if (mip->mi_nactiveclients == 1) {
2123 		/*
2124 		 * Find the one active MAC client from the list of MAC
2125 		 * clients. The active MAC client has at least one
2126 		 * unicast address.
2127 		 */
2128 		for (client = mip->mi_clients_list; client != NULL;
2129 		    client = client->mci_client_next) {
2130 			if (client->mci_unicast_list != NULL)
2131 				break;
2132 		}
2133 		ASSERT(client != NULL);
2134 	}
2135 
2136 	/*
2137 	 * mi_single_active_client is protected by the MAC impl's read/writer
2138 	 * lock, which allows mac_rx() to check the value of that pointer
2139 	 * as a reader.
2140 	 */
2141 	mip->mi_single_active_client = client;
2142 	rw_exit(&mip->mi_rw_lock);
2143 }
2144 
2145 /*
2146  * Set up the data path. Called from i_mac_unicast_add after having
2147  * done all the validations including making sure this is an active
2148  * client (i.e that is ready to process packets.)
2149  */
2150 static int
2151 mac_client_datapath_setup(mac_client_impl_t *mcip, uint16_t vid,
2152     uint8_t *mac_addr, mac_resource_props_t *mrp, boolean_t isprimary,
2153     mac_unicast_impl_t *muip)
2154 {
2155 	mac_impl_t	*mip = mcip->mci_mip;
2156 	boolean_t	mac_started = B_FALSE;
2157 	boolean_t	bcast_added = B_FALSE;
2158 	boolean_t	nactiveclients_added = B_FALSE;
2159 	flow_entry_t	*flent;
2160 	int		err = 0;
2161 	boolean_t	no_unicast;
2162 
2163 	no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2164 
2165 	if ((err = mac_start((mac_handle_t)mip)) != 0)
2166 		goto bail;
2167 
2168 	mac_started = B_TRUE;
2169 
2170 	/* add the MAC client to the broadcast address group by default */
2171 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2172 		err = mac_bcast_add(mcip, mip->mi_type->mt_brdcst_addr, vid,
2173 		    MAC_ADDRTYPE_BROADCAST);
2174 		if (err != 0)
2175 			goto bail;
2176 		bcast_added = B_TRUE;
2177 	}
2178 
2179 	/*
2180 	 * If this is the first unicast address addition for this
2181 	 * client, reuse the pre-allocated larval flow entry associated with
2182 	 * the MAC client.
2183 	 */
2184 	flent = (mcip->mci_nflents == 0) ? mcip->mci_flent : NULL;
2185 
2186 	/* We are configuring the unicast flow now */
2187 	if (!MCIP_DATAPATH_SETUP(mcip)) {
2188 
2189 		if (mrp != NULL) {
2190 			MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
2191 			    (mrp->mrp_mask & MRP_PRIORITY) ? mrp->mrp_priority :
2192 			    MPL_LINK_DEFAULT);
2193 		}
2194 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2195 		    isprimary, B_TRUE, &flent, mrp)) != 0)
2196 			goto bail;
2197 
2198 		mip->mi_nactiveclients++;
2199 		nactiveclients_added = B_TRUE;
2200 
2201 		/*
2202 		 * This will allocate the RX ring group if possible for the
2203 		 * flow and program the software classifier as needed.
2204 		 */
2205 		if ((err = mac_datapath_setup(mcip, flent, SRST_LINK)) != 0)
2206 			goto bail;
2207 
2208 		if (no_unicast)
2209 			goto done_setup;
2210 		/*
2211 		 * The unicast MAC address must have been added successfully.
2212 		 */
2213 		ASSERT(mcip->mci_unicast != NULL);
2214 		/*
2215 		 * Push down the sub-flows that were defined on this link
2216 		 * hitherto. The flows are added to the active flow table
2217 		 * and SRS, softrings etc. are created as needed.
2218 		 */
2219 		mac_link_init_flows((mac_client_handle_t)mcip);
2220 	} else {
2221 		mac_address_t *map = mcip->mci_unicast;
2222 
2223 		ASSERT(!no_unicast);
2224 		/*
2225 		 * A unicast flow already exists for that MAC client,
2226 		 * this flow must be the same mac address but with
2227 		 * different VID. It has been checked by mac_addr_in_use().
2228 		 *
2229 		 * We will use the SRS etc. from the mci_flent. Note that
2230 		 * We don't need to create kstat for this as except for
2231 		 * the fdesc, everything will be used from in the 1st flent.
2232 		 */
2233 
2234 		if (bcmp(mac_addr, map->ma_addr, map->ma_len) != 0) {
2235 			err = EINVAL;
2236 			goto bail;
2237 		}
2238 
2239 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2240 		    isprimary, B_FALSE, &flent, NULL)) != 0) {
2241 			goto bail;
2242 		}
2243 		if ((err = mac_flow_add(mip->mi_flow_tab, flent)) != 0) {
2244 			FLOW_FINAL_REFRELE(flent);
2245 			goto bail;
2246 		}
2247 
2248 		/* update the multicast group for this vid */
2249 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2250 		    (void *)flent, B_TRUE);
2251 
2252 	}
2253 
2254 	/* populate the shared MAC address */
2255 	muip->mui_map = mcip->mci_unicast;
2256 
2257 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2258 	muip->mui_next = mcip->mci_unicast_list;
2259 	mcip->mci_unicast_list = muip;
2260 	rw_exit(&mcip->mci_rw_lock);
2261 
2262 done_setup:
2263 	/*
2264 	 * First add the flent to the flow list of this mcip. Then set
2265 	 * the mip's mi_single_active_client if needed. The Rx path assumes
2266 	 * that mip->mi_single_active_client will always have an associated
2267 	 * flent.
2268 	 */
2269 	mac_client_add_to_flow_list(mcip, flent);
2270 	if (nactiveclients_added)
2271 		mac_update_single_active_client(mip);
2272 	/*
2273 	 * Trigger a renegotiation of the capabilities when the number of
2274 	 * active clients changes from 1 to 2, since some of the capabilities
2275 	 * might have to be disabled. Also send a MAC_NOTE_LINK notification
2276 	 * to all the MAC clients whenever physical link is DOWN.
2277 	 */
2278 	if (mip->mi_nactiveclients == 2) {
2279 		mac_capab_update((mac_handle_t)mip);
2280 		mac_virtual_link_update(mip);
2281 	}
2282 	/*
2283 	 * Now that the setup is complete, clear the INCIPIENT flag.
2284 	 * The flag was set to avoid incoming packets seeing inconsistent
2285 	 * structures while the setup was in progress. Clear the mci_tx_flag
2286 	 * by calling mac_tx_client_block. It is possible that
2287 	 * mac_unicast_remove was called prior to this mac_unicast_add which
2288 	 * could have set the MCI_TX_QUIESCE flag.
2289 	 */
2290 	if (flent->fe_rx_ring_group != NULL)
2291 		mac_rx_group_unmark(flent->fe_rx_ring_group, MR_INCIPIENT);
2292 	FLOW_UNMARK(flent, FE_INCIPIENT);
2293 	FLOW_UNMARK(flent, FE_MC_NO_DATAPATH);
2294 	mac_tx_client_unblock(mcip);
2295 	return (0);
2296 bail:
2297 	if (bcast_added)
2298 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr, vid);
2299 
2300 	if (nactiveclients_added)
2301 		mip->mi_nactiveclients--;
2302 
2303 	if (mac_started)
2304 		mac_stop((mac_handle_t)mip);
2305 
2306 	return (err);
2307 }
2308 
2309 /*
2310  * Return the passive primary MAC client, if present. The passive client is
2311  * a stand-by client that has the same unicast address as another that is
2312  * currenly active. Once the active client goes away, the passive client
2313  * becomes active.
2314  */
2315 static mac_client_impl_t *
2316 mac_get_passive_primary_client(mac_impl_t *mip)
2317 {
2318 	mac_client_impl_t	*mcip;
2319 
2320 	for (mcip = mip->mi_clients_list; mcip != NULL;
2321 	    mcip = mcip->mci_client_next) {
2322 		if (mac_is_primary_client(mcip) &&
2323 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2324 			return (mcip);
2325 		}
2326 	}
2327 	return (NULL);
2328 }
2329 
2330 /*
2331  * Add a new unicast address to the MAC client.
2332  *
2333  * The MAC address can be specified either by value, or the MAC client
2334  * can specify that it wants to use the primary MAC address of the
2335  * underlying MAC. See the introductory comments at the beginning
2336  * of this file for more more information on primary MAC addresses.
2337  *
2338  * Note also the tuple (MAC address, VID) must be unique
2339  * for the MAC clients defined on top of the same underlying MAC
2340  * instance, unless the MAC_UNICAST_NODUPCHECK is specified.
2341  *
2342  * In no case can a client use the PVID for the MAC, if the MAC has one set.
2343  */
2344 int
2345 i_mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2346     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2347 {
2348 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2349 	mac_impl_t		*mip = mcip->mci_mip;
2350 	int			err;
2351 	uint_t			mac_len = mip->mi_type->mt_addr_length;
2352 	boolean_t		check_dups = !(flags & MAC_UNICAST_NODUPCHECK);
2353 	boolean_t		fastpath_disabled = B_FALSE;
2354 	boolean_t		is_primary = (flags & MAC_UNICAST_PRIMARY);
2355 	boolean_t		is_unicast_hw = (flags & MAC_UNICAST_HW);
2356 	mac_resource_props_t	*mrp;
2357 	boolean_t		passive_client = B_FALSE;
2358 	mac_unicast_impl_t	*muip;
2359 	boolean_t		is_vnic_primary =
2360 	    (flags & MAC_UNICAST_VNIC_PRIMARY);
2361 
2362 	/* when VID is non-zero, the underlying MAC can not be VNIC */
2363 	ASSERT(!((mip->mi_state_flags & MIS_IS_VNIC) && (vid != 0)));
2364 
2365 	/*
2366 	 * Can't unicast add if the client asked only for minimal datapath
2367 	 * setup.
2368 	 */
2369 	if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR)
2370 		return (ENOTSUP);
2371 
2372 	/*
2373 	 * Check for an attempted use of the current Port VLAN ID, if enabled.
2374 	 * No client may use it.
2375 	 */
2376 	if (mip->mi_pvid != 0 && vid == mip->mi_pvid)
2377 		return (EBUSY);
2378 
2379 	/*
2380 	 * Check whether it's the primary client and flag it.
2381 	 */
2382 	if (!(mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary && vid == 0)
2383 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PRIMARY;
2384 
2385 	/*
2386 	 * is_vnic_primary is true when we come here as a VLAN VNIC
2387 	 * which uses the primary mac client's address but with a non-zero
2388 	 * VID. In this case the MAC address is not specified by an upper
2389 	 * MAC client.
2390 	 */
2391 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary &&
2392 	    !is_vnic_primary) {
2393 		/*
2394 		 * The address is being set by the upper MAC client
2395 		 * of a VNIC. The MAC address was already set by the
2396 		 * VNIC driver during VNIC creation.
2397 		 *
2398 		 * Note: a VNIC has only one MAC address. We return
2399 		 * the MAC unicast address handle of the lower MAC client
2400 		 * corresponding to the VNIC. We allocate a new entry
2401 		 * which is flagged appropriately, so that mac_unicast_remove()
2402 		 * doesn't attempt to free the original entry that
2403 		 * was allocated by the VNIC driver.
2404 		 */
2405 		ASSERT(mcip->mci_unicast != NULL);
2406 
2407 		/* Check for VLAN flags, if present */
2408 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2409 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2410 
2411 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2412 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2413 
2414 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2415 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2416 
2417 		/*
2418 		 * Ensure that the primary unicast address of the VNIC
2419 		 * is added only once unless we have the
2420 		 * MAC_CLIENT_FLAGS_MULTI_PRIMARY set (and this is not
2421 		 * a passive MAC client).
2422 		 */
2423 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) != 0) {
2424 			if ((mcip->mci_flags &
2425 			    MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2426 			    (mcip->mci_flags &
2427 			    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2428 				return (EBUSY);
2429 			}
2430 			mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2431 			passive_client = B_TRUE;
2432 		}
2433 
2434 		mcip->mci_flags |= MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2435 
2436 		/*
2437 		 * Create a handle for vid 0.
2438 		 */
2439 		ASSERT(vid == 0);
2440 		muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2441 		muip->mui_vid = vid;
2442 		*mah = (mac_unicast_handle_t)muip;
2443 		/*
2444 		 * This will be used by the caller to defer setting the
2445 		 * rx functions.
2446 		 */
2447 		if (passive_client)
2448 			return (EAGAIN);
2449 		return (0);
2450 	}
2451 
2452 	/* primary MAC clients cannot be opened on top of anchor VNICs */
2453 	if ((is_vnic_primary || is_primary) &&
2454 	    i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_ANCHOR_VNIC, NULL)) {
2455 		return (ENXIO);
2456 	}
2457 
2458 	/*
2459 	 * If this is a VNIC/VLAN, disable softmac fast-path.
2460 	 */
2461 	if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2462 		err = mac_fastpath_disable((mac_handle_t)mip);
2463 		if (err != 0)
2464 			return (err);
2465 		fastpath_disabled = B_TRUE;
2466 	}
2467 
2468 	/*
2469 	 * Return EBUSY if:
2470 	 *  - there is an exclusively active mac client exists.
2471 	 *  - this is an exclusive active mac client but
2472 	 *	a. there is already active mac clients exist, or
2473 	 *	b. fastpath streams are already plumbed on this legacy device
2474 	 *  - the mac creator has disallowed active mac clients.
2475 	 */
2476 	if (mip->mi_state_flags & (MIS_EXCLUSIVE|MIS_NO_ACTIVE)) {
2477 		if (fastpath_disabled)
2478 			mac_fastpath_enable((mac_handle_t)mip);
2479 		return (EBUSY);
2480 	}
2481 
2482 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2483 		ASSERT(!fastpath_disabled);
2484 		if (mip->mi_nactiveclients != 0)
2485 			return (EBUSY);
2486 
2487 		if ((mip->mi_state_flags & MIS_LEGACY) &&
2488 		    !(mip->mi_capab_legacy.ml_active_set(mip->mi_driver))) {
2489 			return (EBUSY);
2490 		}
2491 		mip->mi_state_flags |= MIS_EXCLUSIVE;
2492 	}
2493 
2494 	mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
2495 	if (is_primary && !(mcip->mci_state_flags & (MCIS_IS_VNIC |
2496 	    MCIS_IS_AGGR_PORT))) {
2497 		/*
2498 		 * Apply the property cached in the mac_impl_t to the primary
2499 		 * mac client. If the mac client is a VNIC or an aggregation
2500 		 * port, its property should be set in the mcip when the
2501 		 * VNIC/aggr was created.
2502 		 */
2503 		mac_get_resources((mac_handle_t)mip, mrp);
2504 		(void) mac_client_set_resources(mch, mrp);
2505 	} else if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2506 		/*
2507 		 * This is a primary VLAN client, we don't support
2508 		 * specifying rings property for this as it inherits the
2509 		 * rings property from its MAC.
2510 		 */
2511 		if (is_vnic_primary) {
2512 			mac_resource_props_t	*vmrp;
2513 
2514 			vmrp = MCIP_RESOURCE_PROPS(mcip);
2515 			if (vmrp->mrp_mask & MRP_RX_RINGS ||
2516 			    vmrp->mrp_mask & MRP_TX_RINGS) {
2517 				if (fastpath_disabled)
2518 					mac_fastpath_enable((mac_handle_t)mip);
2519 				kmem_free(mrp, sizeof (*mrp));
2520 				return (ENOTSUP);
2521 			}
2522 			/*
2523 			 * Additionally we also need to inherit any
2524 			 * rings property from the MAC.
2525 			 */
2526 			mac_get_resources((mac_handle_t)mip, mrp);
2527 			if (mrp->mrp_mask & MRP_RX_RINGS) {
2528 				vmrp->mrp_mask |= MRP_RX_RINGS;
2529 				vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
2530 			}
2531 			if (mrp->mrp_mask & MRP_TX_RINGS) {
2532 				vmrp->mrp_mask |= MRP_TX_RINGS;
2533 				vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
2534 			}
2535 		}
2536 		bcopy(MCIP_RESOURCE_PROPS(mcip), mrp, sizeof (*mrp));
2537 	}
2538 
2539 	muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2540 	muip->mui_vid = vid;
2541 
2542 	if (is_primary || is_vnic_primary) {
2543 		mac_addr = mip->mi_addr;
2544 	} else {
2545 
2546 		/*
2547 		 * Verify the validity of the specified MAC addresses value.
2548 		 */
2549 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr, mac_len)) {
2550 			*diag = MAC_DIAG_MACADDR_INVALID;
2551 			err = EINVAL;
2552 			goto bail_out;
2553 		}
2554 
2555 		/*
2556 		 * Make sure that the specified MAC address is different
2557 		 * than the unicast MAC address of the underlying NIC.
2558 		 */
2559 		if (check_dups && bcmp(mip->mi_addr, mac_addr, mac_len) == 0) {
2560 			*diag = MAC_DIAG_MACADDR_NIC;
2561 			err = EINVAL;
2562 			goto bail_out;
2563 		}
2564 	}
2565 
2566 	/*
2567 	 * Set the flags here so that if this is a passive client, we
2568 	 * can return  and set it when we call mac_client_datapath_setup
2569 	 * when this becomes the active client. If we defer to using these
2570 	 * flags to mac_client_datapath_setup, then for a passive client,
2571 	 * we'd have to store the flags somewhere (probably fe_flags)
2572 	 * and then use it.
2573 	 */
2574 	if (!MCIP_DATAPATH_SETUP(mcip)) {
2575 		if (is_unicast_hw) {
2576 			/*
2577 			 * The client requires a hardware MAC address slot
2578 			 * for that unicast address. Since we support only
2579 			 * one unicast MAC address per client, flag the
2580 			 * MAC client itself.
2581 			 */
2582 			mcip->mci_state_flags |= MCIS_UNICAST_HW;
2583 		}
2584 
2585 		/* Check for VLAN flags, if present */
2586 		if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2587 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2588 
2589 		if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2590 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2591 
2592 		if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2593 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2594 	} else {
2595 		/*
2596 		 * Assert that the specified flags are consistent with the
2597 		 * flags specified by previous calls to mac_unicast_add().
2598 		 */
2599 		ASSERT(((flags & MAC_UNICAST_TAG_DISABLE) != 0 &&
2600 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) != 0) ||
2601 		    ((flags & MAC_UNICAST_TAG_DISABLE) == 0 &&
2602 		    (mcip->mci_state_flags & MCIS_TAG_DISABLE) == 0));
2603 
2604 		ASSERT(((flags & MAC_UNICAST_STRIP_DISABLE) != 0 &&
2605 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) != 0) ||
2606 		    ((flags & MAC_UNICAST_STRIP_DISABLE) == 0 &&
2607 		    (mcip->mci_state_flags & MCIS_STRIP_DISABLE) == 0));
2608 
2609 		ASSERT(((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0 &&
2610 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) != 0) ||
2611 		    ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) == 0 &&
2612 		    (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) == 0));
2613 
2614 		/*
2615 		 * Make sure the client is consistent about its requests
2616 		 * for MAC addresses. I.e. all requests from the clients
2617 		 * must have the MAC_UNICAST_HW flag set or clear.
2618 		 */
2619 		if ((mcip->mci_state_flags & MCIS_UNICAST_HW) != 0 &&
2620 		    !is_unicast_hw ||
2621 		    (mcip->mci_state_flags & MCIS_UNICAST_HW) == 0 &&
2622 		    is_unicast_hw) {
2623 			err = EINVAL;
2624 			goto bail_out;
2625 		}
2626 	}
2627 	/*
2628 	 * Make sure the MAC address is not already used by
2629 	 * another MAC client defined on top of the same
2630 	 * underlying NIC. Unless we have MAC_CLIENT_FLAGS_MULTI_PRIMARY
2631 	 * set when we allow a passive client to be present which will
2632 	 * be activated when the currently active client goes away - this
2633 	 * works only with primary addresses.
2634 	 */
2635 	if ((check_dups || is_primary || is_vnic_primary) &&
2636 	    mac_addr_in_use(mip, mac_addr, vid)) {
2637 		/*
2638 		 * Must have set the multiple primary address flag when
2639 		 * we did a mac_client_open AND this should be a primary
2640 		 * MAC client AND there should not already be a passive
2641 		 * primary. If all is true then we let this succeed
2642 		 * even if the address is a dup.
2643 		 */
2644 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2645 		    (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) == 0 ||
2646 		    mac_get_passive_primary_client(mip) != NULL) {
2647 			*diag = MAC_DIAG_MACADDR_INUSE;
2648 			err = EEXIST;
2649 			goto bail_out;
2650 		}
2651 		ASSERT((mcip->mci_flags &
2652 		    MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) == 0);
2653 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2654 		kmem_free(mrp, sizeof (*mrp));
2655 
2656 		/*
2657 		 * Stash the unicast address handle, we will use it when
2658 		 * we set up the passive client.
2659 		 */
2660 		mcip->mci_p_unicast_list = muip;
2661 		*mah = (mac_unicast_handle_t)muip;
2662 		return (0);
2663 	}
2664 
2665 	err = mac_client_datapath_setup(mcip, vid, mac_addr, mrp,
2666 	    is_primary || is_vnic_primary, muip);
2667 	if (err != 0)
2668 		goto bail_out;
2669 
2670 	kmem_free(mrp, sizeof (*mrp));
2671 	*mah = (mac_unicast_handle_t)muip;
2672 	return (0);
2673 
2674 bail_out:
2675 	if (fastpath_disabled)
2676 		mac_fastpath_enable((mac_handle_t)mip);
2677 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2678 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2679 		if (mip->mi_state_flags & MIS_LEGACY) {
2680 			mip->mi_capab_legacy.ml_active_clear(
2681 			    mip->mi_driver);
2682 		}
2683 	}
2684 	kmem_free(mrp, sizeof (*mrp));
2685 	kmem_free(muip, sizeof (mac_unicast_impl_t));
2686 	return (err);
2687 }
2688 
2689 /*
2690  * Wrapper function to mac_unicast_add when we want to have the same mac
2691  * client open for two instances, one that is currently active and another
2692  * that will become active when the current one is removed. In this case
2693  * mac_unicast_add will return EGAIN and we will save the rx function and
2694  * arg which will be used when we activate the passive client in
2695  * mac_unicast_remove.
2696  */
2697 int
2698 mac_unicast_add_set_rx(mac_client_handle_t mch, uint8_t *mac_addr,
2699     uint16_t flags, mac_unicast_handle_t *mah,  uint16_t vid, mac_diag_t *diag,
2700     mac_rx_t rx_fn, void *arg)
2701 {
2702 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2703 	uint_t			err;
2704 
2705 	err = mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2706 	if (err != 0 && err != EAGAIN)
2707 		return (err);
2708 	if (err == EAGAIN) {
2709 		if (rx_fn != NULL) {
2710 			mcip->mci_rx_p_fn = rx_fn;
2711 			mcip->mci_rx_p_arg = arg;
2712 		}
2713 		return (0);
2714 	}
2715 	if (rx_fn != NULL)
2716 		mac_rx_set(mch, rx_fn, arg);
2717 	return (err);
2718 }
2719 
2720 int
2721 mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2722     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2723 {
2724 	mac_impl_t *mip = ((mac_client_impl_t *)mch)->mci_mip;
2725 	uint_t err;
2726 
2727 	i_mac_perim_enter(mip);
2728 	err = i_mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2729 	i_mac_perim_exit(mip);
2730 
2731 	return (err);
2732 }
2733 
2734 static void
2735 mac_client_datapath_teardown(mac_client_handle_t mch, mac_unicast_impl_t *muip,
2736     flow_entry_t *flent)
2737 {
2738 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2739 	mac_impl_t		*mip = mcip->mci_mip;
2740 	boolean_t		no_unicast;
2741 
2742 	/*
2743 	 * If we have not added a unicast address for this MAC client, just
2744 	 * teardown the datapath.
2745 	 */
2746 	no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2747 
2748 	if (!no_unicast) {
2749 		/*
2750 		 * We would have initialized subflows etc. only if we brought
2751 		 * up the primary client and set the unicast unicast address
2752 		 * etc. Deactivate the flows. The flow entry will be removed
2753 		 * from the active flow tables, and the associated SRS,
2754 		 * softrings etc will be deleted. But the flow entry itself
2755 		 * won't be destroyed, instead it will continue to be archived
2756 		 * off the  the global flow hash list, for a possible future
2757 		 * activation when say IP is plumbed again.
2758 		 */
2759 		mac_link_release_flows(mch);
2760 	}
2761 	mip->mi_nactiveclients--;
2762 	mac_update_single_active_client(mip);
2763 
2764 	/* Tear down the data path */
2765 	mac_datapath_teardown(mcip, mcip->mci_flent, SRST_LINK);
2766 
2767 	/*
2768 	 * Prevent any future access to the flow entry through the mci_flent
2769 	 * pointer by setting the mci_flent to NULL. Access to mci_flent in
2770 	 * mac_bcast_send is also under mi_rw_lock.
2771 	 */
2772 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2773 	flent = mcip->mci_flent;
2774 	mac_client_remove_flow_from_list(mcip, flent);
2775 
2776 	if (mcip->mci_state_flags & MCIS_DESC_LOGGED)
2777 		mcip->mci_state_flags &= ~MCIS_DESC_LOGGED;
2778 
2779 	/*
2780 	 * This is the last unicast address being removed and there shouldn't
2781 	 * be any outbound data threads at this point coming down from mac
2782 	 * clients. We have waited for the data threads to finish before
2783 	 * starting dld_str_detach. Non-data threads must access TX SRS
2784 	 * under mi_rw_lock.
2785 	 */
2786 	rw_exit(&mip->mi_rw_lock);
2787 
2788 	/*
2789 	 * Don't use FLOW_MARK with FE_MC_NO_DATAPATH, as the flow might
2790 	 * contain other flags, such as FE_CONDEMNED, which we need to
2791 	 * cleared. We don't call mac_flow_cleanup() for this unicast
2792 	 * flow as we have a already cleaned up SRSs etc. (via the teadown
2793 	 * path). We just clear the stats and reset the initial callback
2794 	 * function, the rest will be set when we call mac_flow_create,
2795 	 * if at all.
2796 	 */
2797 	mutex_enter(&flent->fe_lock);
2798 	ASSERT(flent->fe_refcnt == 1 && flent->fe_mbg == NULL &&
2799 	    flent->fe_tx_srs == NULL && flent->fe_rx_srs_cnt == 0);
2800 	flent->fe_flags = FE_MC_NO_DATAPATH;
2801 	flow_stat_destroy(flent);
2802 	mac_misc_stat_delete(flent);
2803 
2804 	/* Initialize the receiver function to a safe routine */
2805 	flent->fe_cb_fn = (flow_fn_t)mac_pkt_drop;
2806 	flent->fe_cb_arg1 = NULL;
2807 	flent->fe_cb_arg2 = NULL;
2808 
2809 	flent->fe_index = -1;
2810 	mutex_exit(&flent->fe_lock);
2811 
2812 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2813 		ASSERT(muip != NULL || no_unicast);
2814 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2815 		    muip != NULL ? muip->mui_vid : VLAN_ID_NONE);
2816 	}
2817 
2818 	if (mip->mi_nactiveclients == 1) {
2819 		mac_capab_update((mac_handle_t)mip);
2820 		mac_virtual_link_update(mip);
2821 	}
2822 
2823 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2824 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2825 
2826 		if (mip->mi_state_flags & MIS_LEGACY)
2827 			mip->mi_capab_legacy.ml_active_clear(mip->mi_driver);
2828 	}
2829 
2830 	mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2831 
2832 	if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2833 		mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2834 
2835 	if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2836 		mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2837 
2838 	if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2839 		mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2840 
2841 	if (muip != NULL)
2842 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2843 	mac_protect_cancel_timer(mcip);
2844 	mac_protect_flush_dhcp(mcip);
2845 
2846 	bzero(&mcip->mci_misc_stat, sizeof (mcip->mci_misc_stat));
2847 	/*
2848 	 * Disable fastpath if this is a VNIC or a VLAN.
2849 	 */
2850 	if (mcip->mci_state_flags & MCIS_IS_VNIC)
2851 		mac_fastpath_enable((mac_handle_t)mip);
2852 	mac_stop((mac_handle_t)mip);
2853 }
2854 
2855 /*
2856  * Remove a MAC address which was previously added by mac_unicast_add().
2857  */
2858 int
2859 mac_unicast_remove(mac_client_handle_t mch, mac_unicast_handle_t mah)
2860 {
2861 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2862 	mac_unicast_impl_t *muip = (mac_unicast_impl_t *)mah;
2863 	mac_unicast_impl_t *pre;
2864 	mac_impl_t *mip = mcip->mci_mip;
2865 	flow_entry_t		*flent;
2866 	uint16_t mui_vid;
2867 
2868 	i_mac_perim_enter(mip);
2869 	if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) {
2870 		/*
2871 		 * Called made by the upper MAC client of a VNIC.
2872 		 * There's nothing much to do, the unicast address will
2873 		 * be removed by the VNIC driver when the VNIC is deleted,
2874 		 * but let's ensure that all our transmit is done before
2875 		 * the client does a mac_client_stop lest it trigger an
2876 		 * assert in the driver.
2877 		 */
2878 		ASSERT(muip->mui_vid == 0);
2879 
2880 		mac_tx_client_flush(mcip);
2881 
2882 		if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2883 			mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2884 			if (mcip->mci_rx_p_fn != NULL) {
2885 				mac_rx_set(mch, mcip->mci_rx_p_fn,
2886 				    mcip->mci_rx_p_arg);
2887 				mcip->mci_rx_p_fn = NULL;
2888 				mcip->mci_rx_p_arg = NULL;
2889 			}
2890 			kmem_free(muip, sizeof (mac_unicast_impl_t));
2891 			i_mac_perim_exit(mip);
2892 			return (0);
2893 		}
2894 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2895 
2896 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2897 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2898 
2899 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2900 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2901 
2902 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2903 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2904 
2905 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2906 		i_mac_perim_exit(mip);
2907 		return (0);
2908 	}
2909 
2910 	ASSERT(muip != NULL);
2911 
2912 	/*
2913 	 * We are removing a passive client, we haven't setup the datapath
2914 	 * for this yet, so nothing much to do.
2915 	 */
2916 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2917 
2918 		ASSERT((mcip->mci_flent->fe_flags & FE_MC_NO_DATAPATH) != 0);
2919 		ASSERT(mcip->mci_p_unicast_list == muip);
2920 
2921 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2922 
2923 		mcip->mci_p_unicast_list = NULL;
2924 		mcip->mci_rx_p_fn = NULL;
2925 		mcip->mci_rx_p_arg = NULL;
2926 
2927 		mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2928 
2929 		if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2930 			mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2931 
2932 		if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2933 			mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2934 
2935 		if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2936 			mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2937 
2938 		kmem_free(muip, sizeof (mac_unicast_impl_t));
2939 		i_mac_perim_exit(mip);
2940 		return (0);
2941 	}
2942 	/*
2943 	 * Remove the VID from the list of client's VIDs.
2944 	 */
2945 	pre = mcip->mci_unicast_list;
2946 	if (muip == pre) {
2947 		mcip->mci_unicast_list = muip->mui_next;
2948 	} else {
2949 		while ((pre->mui_next != NULL) && (pre->mui_next != muip))
2950 			pre = pre->mui_next;
2951 		ASSERT(pre->mui_next == muip);
2952 		rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2953 		pre->mui_next = muip->mui_next;
2954 		rw_exit(&mcip->mci_rw_lock);
2955 	}
2956 
2957 	if (!mac_client_single_rcvr(mcip)) {
2958 		/*
2959 		 * This MAC client is shared by more than one unicast
2960 		 * addresses, so we will just remove the flent
2961 		 * corresponding to the address being removed. We don't invoke
2962 		 * mac_rx_classify_flow_rem() since the additional flow is
2963 		 * not associated with its own separate set of SRS and rings,
2964 		 * and these constructs are still needed for the remaining
2965 		 * flows.
2966 		 */
2967 		flent = mac_client_get_flow(mcip, muip);
2968 		ASSERT(flent != NULL);
2969 
2970 		/*
2971 		 * The first one is disappearing, need to make sure
2972 		 * we replace it with another from the list of
2973 		 * shared clients.
2974 		 */
2975 		if (flent == mcip->mci_flent)
2976 			flent = mac_client_swap_mciflent(mcip);
2977 		mac_client_remove_flow_from_list(mcip, flent);
2978 		mac_flow_remove(mip->mi_flow_tab, flent, B_FALSE);
2979 		mac_flow_wait(flent, FLOW_DRIVER_UPCALL);
2980 
2981 		/*
2982 		 * The multicast groups that were added by the client so
2983 		 * far must be removed from the brodcast domain corresponding
2984 		 * to the VID being removed.
2985 		 */
2986 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2987 		    (void *)flent, B_FALSE);
2988 
2989 		if (mip->mi_type->mt_brdcst_addr != NULL) {
2990 			mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2991 			    muip->mui_vid);
2992 		}
2993 
2994 		FLOW_FINAL_REFRELE(flent);
2995 		ASSERT(!(mcip->mci_state_flags & MCIS_EXCLUSIVE));
2996 		/*
2997 		 * Enable fastpath if this is a VNIC or a VLAN.
2998 		 */
2999 		if (mcip->mci_state_flags & MCIS_IS_VNIC)
3000 			mac_fastpath_enable((mac_handle_t)mip);
3001 		mac_stop((mac_handle_t)mip);
3002 		i_mac_perim_exit(mip);
3003 		return (0);
3004 	}
3005 
3006 	mui_vid = muip->mui_vid;
3007 	mac_client_datapath_teardown(mch, muip, flent);
3008 
3009 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) && mui_vid == 0) {
3010 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PRIMARY;
3011 	} else {
3012 		i_mac_perim_exit(mip);
3013 		return (0);
3014 	}
3015 
3016 	/*
3017 	 * If we are removing the primary, check if we have a passive primary
3018 	 * client that we need to activate now.
3019 	 */
3020 	mcip = mac_get_passive_primary_client(mip);
3021 	if (mcip != NULL) {
3022 		mac_resource_props_t	*mrp;
3023 		mac_unicast_impl_t	*muip;
3024 
3025 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
3026 		mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
3027 
3028 		/*
3029 		 * Apply the property cached in the mac_impl_t to the
3030 		 * primary mac client.
3031 		 */
3032 		mac_get_resources((mac_handle_t)mip, mrp);
3033 		(void) mac_client_set_resources(mch, mrp);
3034 		ASSERT(mcip->mci_p_unicast_list != NULL);
3035 		muip = mcip->mci_p_unicast_list;
3036 		mcip->mci_p_unicast_list = NULL;
3037 		if (mac_client_datapath_setup(mcip, VLAN_ID_NONE,
3038 		    mip->mi_addr, mrp, B_TRUE, muip) == 0) {
3039 			if (mcip->mci_rx_p_fn != NULL) {
3040 				mac_rx_set(mch, mcip->mci_rx_p_fn,
3041 				    mcip->mci_rx_p_arg);
3042 				mcip->mci_rx_p_fn = NULL;
3043 				mcip->mci_rx_p_arg = NULL;
3044 			}
3045 		} else {
3046 			kmem_free(muip, sizeof (mac_unicast_impl_t));
3047 		}
3048 		kmem_free(mrp, sizeof (*mrp));
3049 	}
3050 	i_mac_perim_exit(mip);
3051 	return (0);
3052 }
3053 
3054 /*
3055  * Multicast add function invoked by MAC clients.
3056  */
3057 int
3058 mac_multicast_add(mac_client_handle_t mch, const uint8_t *addr)
3059 {
3060 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3061 	mac_impl_t		*mip = mcip->mci_mip;
3062 	flow_entry_t		*flent = mcip->mci_flent_list;
3063 	flow_entry_t		*prev_fe = NULL;
3064 	uint16_t		vid;
3065 	int			err = 0;
3066 
3067 	/* Verify the address is a valid multicast address */
3068 	if ((err = mip->mi_type->mt_ops.mtops_multicst_verify(addr,
3069 	    mip->mi_pdata)) != 0)
3070 		return (err);
3071 
3072 	i_mac_perim_enter(mip);
3073 	while (flent != NULL) {
3074 		vid = i_mac_flow_vid(flent);
3075 
3076 		err = mac_bcast_add((mac_client_impl_t *)mch, addr, vid,
3077 		    MAC_ADDRTYPE_MULTICAST);
3078 		if (err != 0)
3079 			break;
3080 		prev_fe = flent;
3081 		flent = flent->fe_client_next;
3082 	}
3083 
3084 	/*
3085 	 * If we failed adding, then undo all, rather than partial
3086 	 * success.
3087 	 */
3088 	if (flent != NULL && prev_fe != NULL) {
3089 		flent = mcip->mci_flent_list;
3090 		while (flent != prev_fe->fe_client_next) {
3091 			vid = i_mac_flow_vid(flent);
3092 			mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3093 			flent = flent->fe_client_next;
3094 		}
3095 	}
3096 	i_mac_perim_exit(mip);
3097 	return (err);
3098 }
3099 
3100 /*
3101  * Multicast delete function invoked by MAC clients.
3102  */
3103 void
3104 mac_multicast_remove(mac_client_handle_t mch, const uint8_t *addr)
3105 {
3106 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3107 	mac_impl_t		*mip = mcip->mci_mip;
3108 	flow_entry_t		*flent;
3109 	uint16_t		vid;
3110 
3111 	i_mac_perim_enter(mip);
3112 	for (flent = mcip->mci_flent_list; flent != NULL;
3113 	    flent = flent->fe_client_next) {
3114 		vid = i_mac_flow_vid(flent);
3115 		mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3116 	}
3117 	i_mac_perim_exit(mip);
3118 }
3119 
3120 /*
3121  * When a MAC client desires to capture packets on an interface,
3122  * it registers a promiscuous call back with mac_promisc_add().
3123  * There are three types of promiscuous callbacks:
3124  *
3125  * * MAC_CLIENT_PROMISC_ALL
3126  *   Captures all packets sent and received by the MAC client,
3127  *   the physical interface, as well as all other MAC clients
3128  *   defined on top of the same MAC.
3129  *
3130  * * MAC_CLIENT_PROMISC_FILTERED
3131  *   Captures all packets sent and received by the MAC client,
3132  *   plus all multicast traffic sent and received by the phyisical
3133  *   interface and the other MAC clients.
3134  *
3135  * * MAC_CLIENT_PROMISC_MULTI
3136  *   Captures all broadcast and multicast packets sent and
3137  *   received by the MAC clients as well as the physical interface.
3138  *
3139  * In all cases, the underlying MAC is put in promiscuous mode.
3140  */
3141 int
3142 mac_promisc_add(mac_client_handle_t mch, mac_client_promisc_type_t type,
3143     mac_rx_t fn, void *arg, mac_promisc_handle_t *mphp, uint16_t flags)
3144 {
3145 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3146 	mac_impl_t *mip = mcip->mci_mip;
3147 	mac_promisc_impl_t *mpip;
3148 	mac_cb_info_t	*mcbi;
3149 	int rc;
3150 
3151 	i_mac_perim_enter(mip);
3152 
3153 	if ((rc = mac_start((mac_handle_t)mip)) != 0) {
3154 		i_mac_perim_exit(mip);
3155 		return (rc);
3156 	}
3157 
3158 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
3159 	    type == MAC_CLIENT_PROMISC_ALL) {
3160 		/*
3161 		 * The function is being invoked by the upper MAC client
3162 		 * of a VNIC. The VNIC should only see the traffic
3163 		 * it is entitled to.
3164 		 */
3165 		type = MAC_CLIENT_PROMISC_FILTERED;
3166 	}
3167 
3168 
3169 	/*
3170 	 * Turn on promiscuous mode for the underlying NIC.
3171 	 * This is needed even for filtered callbacks which
3172 	 * expect to receive all multicast traffic on the wire.
3173 	 *
3174 	 * Physical promiscuous mode should not be turned on if
3175 	 * MAC_PROMISC_FLAGS_NO_PHYS is set.
3176 	 */
3177 	if ((flags & MAC_PROMISC_FLAGS_NO_PHYS) == 0) {
3178 		if ((rc = i_mac_promisc_set(mip, B_TRUE)) != 0) {
3179 			mac_stop((mac_handle_t)mip);
3180 			i_mac_perim_exit(mip);
3181 			return (rc);
3182 		}
3183 	}
3184 
3185 	mpip = kmem_cache_alloc(mac_promisc_impl_cache, KM_SLEEP);
3186 
3187 	mpip->mpi_type = type;
3188 	mpip->mpi_fn = fn;
3189 	mpip->mpi_arg = arg;
3190 	mpip->mpi_mcip = mcip;
3191 	mpip->mpi_no_tx_loop = ((flags & MAC_PROMISC_FLAGS_NO_TX_LOOP) != 0);
3192 	mpip->mpi_no_phys = ((flags & MAC_PROMISC_FLAGS_NO_PHYS) != 0);
3193 	mpip->mpi_strip_vlan_tag =
3194 	    ((flags & MAC_PROMISC_FLAGS_VLAN_TAG_STRIP) != 0);
3195 	mpip->mpi_no_copy = ((flags & MAC_PROMISC_FLAGS_NO_COPY) != 0);
3196 
3197 	mcbi = &mip->mi_promisc_cb_info;
3198 	mutex_enter(mcbi->mcbi_lockp);
3199 
3200 	mac_callback_add(&mip->mi_promisc_cb_info, &mcip->mci_promisc_list,
3201 	    &mpip->mpi_mci_link);
3202 	mac_callback_add(&mip->mi_promisc_cb_info, &mip->mi_promisc_list,
3203 	    &mpip->mpi_mi_link);
3204 
3205 	mutex_exit(mcbi->mcbi_lockp);
3206 
3207 	*mphp = (mac_promisc_handle_t)mpip;
3208 	i_mac_perim_exit(mip);
3209 	return (0);
3210 }
3211 
3212 /*
3213  * Remove a multicast address previously aded through mac_promisc_add().
3214  */
3215 void
3216 mac_promisc_remove(mac_promisc_handle_t mph)
3217 {
3218 	mac_promisc_impl_t *mpip = (mac_promisc_impl_t *)mph;
3219 	mac_client_impl_t *mcip = mpip->mpi_mcip;
3220 	mac_impl_t *mip = mcip->mci_mip;
3221 	mac_cb_info_t *mcbi;
3222 	int rv;
3223 
3224 	i_mac_perim_enter(mip);
3225 
3226 	/*
3227 	 * Even if the device can't be reset into normal mode, we still
3228 	 * need to clear the client promisc callbacks. The client may want
3229 	 * to close the mac end point and we can't have stale callbacks.
3230 	 */
3231 	if (!(mpip->mpi_no_phys)) {
3232 		if ((rv = i_mac_promisc_set(mip, B_FALSE)) != 0) {
3233 			cmn_err(CE_WARN, "%s: failed to switch OFF promiscuous"
3234 			    " mode because of error 0x%x", mip->mi_name, rv);
3235 		}
3236 	}
3237 	mcbi = &mip->mi_promisc_cb_info;
3238 	mutex_enter(mcbi->mcbi_lockp);
3239 	if (mac_callback_remove(mcbi, &mip->mi_promisc_list,
3240 	    &mpip->mpi_mi_link)) {
3241 		VERIFY(mac_callback_remove(&mip->mi_promisc_cb_info,
3242 		    &mcip->mci_promisc_list, &mpip->mpi_mci_link));
3243 		kmem_cache_free(mac_promisc_impl_cache, mpip);
3244 	} else {
3245 		mac_callback_remove_wait(&mip->mi_promisc_cb_info);
3246 	}
3247 	mutex_exit(mcbi->mcbi_lockp);
3248 	mac_stop((mac_handle_t)mip);
3249 
3250 	i_mac_perim_exit(mip);
3251 }
3252 
3253 /*
3254  * Reference count the number of active Tx threads. MCI_TX_QUIESCE indicates
3255  * that a control operation wants to quiesce the Tx data flow in which case
3256  * we return an error. Holding any of the per cpu locks ensures that the
3257  * mci_tx_flag won't change.
3258  *
3259  * 'CPU' must be accessed just once and used to compute the index into the
3260  * percpu array, and that index must be used for the entire duration of the
3261  * packet send operation. Note that the thread may be preempted and run on
3262  * another cpu any time and so we can't use 'CPU' more than once for the
3263  * operation.
3264  */
3265 #define	MAC_TX_TRY_HOLD(mcip, mytx, error)				\
3266 {									\
3267 	(error) = 0;							\
3268 	(mytx) = &(mcip)->mci_tx_pcpu[CPU->cpu_seqid & mac_tx_percpu_cnt]; \
3269 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
3270 	if (!((mcip)->mci_tx_flag & MCI_TX_QUIESCE)) {			\
3271 		(mytx)->pcpu_tx_refcnt++;				\
3272 	} else {							\
3273 		(error) = -1;						\
3274 	}								\
3275 	mutex_exit(&(mytx)->pcpu_tx_lock);				\
3276 }
3277 
3278 /*
3279  * Release the reference. If needed, signal any control operation waiting
3280  * for Tx quiescence. The wait and signal are always done using the
3281  * mci_tx_pcpu[0]'s lock
3282  */
3283 #define	MAC_TX_RELE(mcip, mytx) {					\
3284 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
3285 	if (--(mytx)->pcpu_tx_refcnt == 0 &&				\
3286 	    (mcip)->mci_tx_flag & MCI_TX_QUIESCE) {			\
3287 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
3288 		mutex_enter(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
3289 		cv_signal(&(mcip)->mci_tx_cv);				\
3290 		mutex_exit(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
3291 	} else {							\
3292 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
3293 	}								\
3294 }
3295 
3296 /*
3297  * Send function invoked by MAC clients.
3298  */
3299 mac_tx_cookie_t
3300 mac_tx(mac_client_handle_t mch, mblk_t *mp_chain, uintptr_t hint,
3301     uint16_t flag, mblk_t **ret_mp)
3302 {
3303 	mac_tx_cookie_t		cookie = NULL;
3304 	int			error;
3305 	mac_tx_percpu_t		*mytx;
3306 	mac_soft_ring_set_t	*srs;
3307 	flow_entry_t		*flent;
3308 	boolean_t		is_subflow = B_FALSE;
3309 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3310 	mac_impl_t		*mip = mcip->mci_mip;
3311 	mac_srs_tx_t		*srs_tx;
3312 
3313 	/*
3314 	 * Check whether the active Tx threads count is bumped already.
3315 	 */
3316 	if (!(flag & MAC_TX_NO_HOLD)) {
3317 		MAC_TX_TRY_HOLD(mcip, mytx, error);
3318 		if (error != 0) {
3319 			freemsgchain(mp_chain);
3320 			return (NULL);
3321 		}
3322 	}
3323 
3324 	/*
3325 	 * If mac protection is enabled, only the permissible packets will be
3326 	 * returned by mac_protect_check().
3327 	 */
3328 	if ((mcip->mci_flent->
3329 	    fe_resource_props.mrp_mask & MRP_PROTECT) != 0 &&
3330 	    (mp_chain = mac_protect_check(mch, mp_chain)) == NULL)
3331 		goto done;
3332 
3333 	if (mcip->mci_subflow_tab != NULL &&
3334 	    mcip->mci_subflow_tab->ft_flow_count > 0 &&
3335 	    mac_flow_lookup(mcip->mci_subflow_tab, mp_chain,
3336 	    FLOW_OUTBOUND, &flent) == 0) {
3337 		/*
3338 		 * The main assumption here is that if in the event
3339 		 * we get a chain, all the packets will be classified
3340 		 * to the same Flow/SRS. If this changes for any
3341 		 * reason, the following logic should change as well.
3342 		 * I suppose the fanout_hint also assumes this .
3343 		 */
3344 		ASSERT(flent != NULL);
3345 		is_subflow = B_TRUE;
3346 	} else {
3347 		flent = mcip->mci_flent;
3348 	}
3349 
3350 	srs = flent->fe_tx_srs;
3351 	/*
3352 	 * This is to avoid panics with PF_PACKET that can call mac_tx()
3353 	 * against an interface that is not capable of sending. A rewrite
3354 	 * of the mac datapath is required to remove this limitation.
3355 	 */
3356 	if (srs == NULL) {
3357 		freemsgchain(mp_chain);
3358 		goto done;
3359 	}
3360 
3361 	srs_tx = &srs->srs_tx;
3362 	if (srs_tx->st_mode == SRS_TX_DEFAULT &&
3363 	    (srs->srs_state & SRS_ENQUEUED) == 0 &&
3364 	    mip->mi_nactiveclients == 1 && mp_chain->b_next == NULL) {
3365 		uint64_t	obytes;
3366 
3367 		/*
3368 		 * Since dls always opens the underlying MAC, nclients equals
3369 		 * to 1 means that the only active client is dls itself acting
3370 		 * as a primary client of the MAC instance. Since dls will not
3371 		 * send tagged packets in that case, and dls is trusted to send
3372 		 * packets for its allowed VLAN(s), the VLAN tag insertion and
3373 		 * check is required only if nclients is greater than 1.
3374 		 */
3375 		if (mip->mi_nclients > 1) {
3376 			if (MAC_VID_CHECK_NEEDED(mcip)) {
3377 				int	err = 0;
3378 
3379 				MAC_VID_CHECK(mcip, mp_chain, err);
3380 				if (err != 0) {
3381 					freemsg(mp_chain);
3382 					mcip->mci_misc_stat.mms_txerrors++;
3383 					goto done;
3384 				}
3385 			}
3386 			if (MAC_TAG_NEEDED(mcip)) {
3387 				mp_chain = mac_add_vlan_tag(mp_chain, 0,
3388 				    mac_client_vid(mch));
3389 				if (mp_chain == NULL) {
3390 					mcip->mci_misc_stat.mms_txerrors++;
3391 					goto done;
3392 				}
3393 			}
3394 		}
3395 
3396 		obytes = (mp_chain->b_cont == NULL ? MBLKL(mp_chain) :
3397 		    msgdsize(mp_chain));
3398 
3399 		MAC_TX(mip, srs_tx->st_arg2, mp_chain, mcip);
3400 		if (mp_chain == NULL) {
3401 			cookie = NULL;
3402 			SRS_TX_STAT_UPDATE(srs, opackets, 1);
3403 			SRS_TX_STAT_UPDATE(srs, obytes, obytes);
3404 		} else {
3405 			mutex_enter(&srs->srs_lock);
3406 			cookie = mac_tx_srs_no_desc(srs, mp_chain,
3407 			    flag, ret_mp);
3408 			mutex_exit(&srs->srs_lock);
3409 		}
3410 	} else {
3411 		cookie = srs_tx->st_func(srs, mp_chain, hint, flag, ret_mp);
3412 	}
3413 
3414 done:
3415 	if (is_subflow)
3416 		FLOW_REFRELE(flent);
3417 
3418 	if (!(flag & MAC_TX_NO_HOLD))
3419 		MAC_TX_RELE(mcip, mytx);
3420 
3421 	return (cookie);
3422 }
3423 
3424 /*
3425  * mac_tx_is_blocked
3426  *
3427  * Given a cookie, it returns if the ring identified by the cookie is
3428  * flow-controlled or not. If NULL is passed in place of a cookie,
3429  * then it finds out if any of the underlying rings belonging to the
3430  * SRS is flow controlled or not and returns that status.
3431  */
3432 /* ARGSUSED */
3433 boolean_t
3434 mac_tx_is_flow_blocked(mac_client_handle_t mch, mac_tx_cookie_t cookie)
3435 {
3436 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3437 	mac_soft_ring_set_t *mac_srs;
3438 	mac_soft_ring_t *sringp;
3439 	boolean_t blocked = B_FALSE;
3440 	mac_tx_percpu_t *mytx;
3441 	int err;
3442 	int i;
3443 
3444 	/*
3445 	 * Bump the reference count so that mac_srs won't be deleted.
3446 	 * If the client is currently quiesced and we failed to bump
3447 	 * the reference, return B_TRUE so that flow control stays
3448 	 * as enabled.
3449 	 *
3450 	 * Flow control will then be disabled once the client is no
3451 	 * longer quiesced.
3452 	 */
3453 	MAC_TX_TRY_HOLD(mcip, mytx, err);
3454 	if (err != 0)
3455 		return (B_TRUE);
3456 
3457 	if ((mac_srs = MCIP_TX_SRS(mcip)) == NULL) {
3458 		MAC_TX_RELE(mcip, mytx);
3459 		return (B_FALSE);
3460 	}
3461 
3462 	mutex_enter(&mac_srs->srs_lock);
3463 	/*
3464 	 * Only in the case of TX_FANOUT and TX_AGGR, the underlying
3465 	 * softring (s_ring_state) will have the HIWAT set. This is
3466 	 * the multiple Tx ring flow control case. For all other
3467 	 * case, SRS (srs_state) will store the condition.
3468 	 */
3469 	if (mac_srs->srs_tx.st_mode == SRS_TX_FANOUT ||
3470 	    mac_srs->srs_tx.st_mode == SRS_TX_AGGR) {
3471 		if (cookie != NULL) {
3472 			sringp = (mac_soft_ring_t *)cookie;
3473 			mutex_enter(&sringp->s_ring_lock);
3474 			if (sringp->s_ring_state & S_RING_TX_HIWAT)
3475 				blocked = B_TRUE;
3476 			mutex_exit(&sringp->s_ring_lock);
3477 		} else {
3478 			for (i = 0; i < mac_srs->srs_tx_ring_count; i++) {
3479 				sringp = mac_srs->srs_tx_soft_rings[i];
3480 				mutex_enter(&sringp->s_ring_lock);
3481 				if (sringp->s_ring_state & S_RING_TX_HIWAT) {
3482 					blocked = B_TRUE;
3483 					mutex_exit(&sringp->s_ring_lock);
3484 					break;
3485 				}
3486 				mutex_exit(&sringp->s_ring_lock);
3487 			}
3488 		}
3489 	} else {
3490 		blocked = (mac_srs->srs_state & SRS_TX_HIWAT);
3491 	}
3492 	mutex_exit(&mac_srs->srs_lock);
3493 	MAC_TX_RELE(mcip, mytx);
3494 	return (blocked);
3495 }
3496 
3497 /*
3498  * Check if the MAC client is the primary MAC client.
3499  */
3500 boolean_t
3501 mac_is_primary_client(mac_client_impl_t *mcip)
3502 {
3503 	return (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY);
3504 }
3505 
3506 void
3507 mac_ioctl(mac_handle_t mh, queue_t *wq, mblk_t *bp)
3508 {
3509 	mac_impl_t	*mip = (mac_impl_t *)mh;
3510 	int cmd = ((struct iocblk *)bp->b_rptr)->ioc_cmd;
3511 
3512 	if ((cmd == ND_GET && (mip->mi_callbacks->mc_callbacks & MC_GETPROP)) ||
3513 	    (cmd == ND_SET && (mip->mi_callbacks->mc_callbacks & MC_SETPROP))) {
3514 		/*
3515 		 * If ndd props were registered, call them.
3516 		 * Note that ndd ioctls are Obsolete
3517 		 */
3518 		mac_ndd_ioctl(mip, wq, bp);
3519 		return;
3520 	}
3521 
3522 	/*
3523 	 * Call the driver to handle the ioctl.  The driver may not support
3524 	 * any ioctls, in which case we reply with a NAK on its behalf.
3525 	 */
3526 	if (mip->mi_callbacks->mc_callbacks & MC_IOCTL)
3527 		mip->mi_ioctl(mip->mi_driver, wq, bp);
3528 	else
3529 		miocnak(wq, bp, 0, EINVAL);
3530 }
3531 
3532 /*
3533  * Return the link state of the specified MAC instance.
3534  */
3535 link_state_t
3536 mac_link_get(mac_handle_t mh)
3537 {
3538 	return (((mac_impl_t *)mh)->mi_linkstate);
3539 }
3540 
3541 /*
3542  * Add a mac client specified notification callback. Please see the comments
3543  * above mac_callback_add() for general information about mac callback
3544  * addition/deletion in the presence of mac callback list walkers
3545  */
3546 mac_notify_handle_t
3547 mac_notify_add(mac_handle_t mh, mac_notify_t notify_fn, void *arg)
3548 {
3549 	mac_impl_t		*mip = (mac_impl_t *)mh;
3550 	mac_notify_cb_t		*mncb;
3551 	mac_cb_info_t		*mcbi;
3552 
3553 	/*
3554 	 * Allocate a notify callback structure, fill in the details and
3555 	 * use the mac callback list manipulation functions to chain into
3556 	 * the list of callbacks.
3557 	 */
3558 	mncb = kmem_zalloc(sizeof (mac_notify_cb_t), KM_SLEEP);
3559 	mncb->mncb_fn = notify_fn;
3560 	mncb->mncb_arg = arg;
3561 	mncb->mncb_mip = mip;
3562 	mncb->mncb_link.mcb_objp = mncb;
3563 	mncb->mncb_link.mcb_objsize = sizeof (mac_notify_cb_t);
3564 	mncb->mncb_link.mcb_flags = MCB_NOTIFY_CB_T;
3565 
3566 	mcbi = &mip->mi_notify_cb_info;
3567 
3568 	i_mac_perim_enter(mip);
3569 	mutex_enter(mcbi->mcbi_lockp);
3570 
3571 	mac_callback_add(&mip->mi_notify_cb_info, &mip->mi_notify_cb_list,
3572 	    &mncb->mncb_link);
3573 
3574 	mutex_exit(mcbi->mcbi_lockp);
3575 	i_mac_perim_exit(mip);
3576 	return ((mac_notify_handle_t)mncb);
3577 }
3578 
3579 void
3580 mac_notify_remove_wait(mac_handle_t mh)
3581 {
3582 	mac_impl_t	*mip = (mac_impl_t *)mh;
3583 	mac_cb_info_t	*mcbi = &mip->mi_notify_cb_info;
3584 
3585 	mutex_enter(mcbi->mcbi_lockp);
3586 	mac_callback_remove_wait(&mip->mi_notify_cb_info);
3587 	mutex_exit(mcbi->mcbi_lockp);
3588 }
3589 
3590 /*
3591  * Remove a mac client specified notification callback
3592  */
3593 int
3594 mac_notify_remove(mac_notify_handle_t mnh, boolean_t wait)
3595 {
3596 	mac_notify_cb_t	*mncb = (mac_notify_cb_t *)mnh;
3597 	mac_impl_t	*mip = mncb->mncb_mip;
3598 	mac_cb_info_t	*mcbi;
3599 	int		err = 0;
3600 
3601 	mcbi = &mip->mi_notify_cb_info;
3602 
3603 	i_mac_perim_enter(mip);
3604 	mutex_enter(mcbi->mcbi_lockp);
3605 
3606 	ASSERT(mncb->mncb_link.mcb_objp == mncb);
3607 	/*
3608 	 * If there aren't any list walkers, the remove would succeed
3609 	 * inline, else we wait for the deferred remove to complete
3610 	 */
3611 	if (mac_callback_remove(&mip->mi_notify_cb_info,
3612 	    &mip->mi_notify_cb_list, &mncb->mncb_link)) {
3613 		kmem_free(mncb, sizeof (mac_notify_cb_t));
3614 	} else {
3615 		err = EBUSY;
3616 	}
3617 
3618 	mutex_exit(mcbi->mcbi_lockp);
3619 	i_mac_perim_exit(mip);
3620 
3621 	/*
3622 	 * If we failed to remove the notification callback and "wait" is set
3623 	 * to be B_TRUE, wait for the callback to finish after we exit the
3624 	 * mac perimeter.
3625 	 */
3626 	if (err != 0 && wait) {
3627 		mac_notify_remove_wait((mac_handle_t)mip);
3628 		return (0);
3629 	}
3630 
3631 	return (err);
3632 }
3633 
3634 /*
3635  * Associate resource management callbacks with the specified MAC
3636  * clients.
3637  */
3638 
3639 void
3640 mac_resource_set_common(mac_client_handle_t mch, mac_resource_add_t add,
3641     mac_resource_remove_t remove, mac_resource_quiesce_t quiesce,
3642     mac_resource_restart_t restart, mac_resource_bind_t bind,
3643     void *arg)
3644 {
3645 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3646 
3647 	mcip->mci_resource_add = add;
3648 	mcip->mci_resource_remove = remove;
3649 	mcip->mci_resource_quiesce = quiesce;
3650 	mcip->mci_resource_restart = restart;
3651 	mcip->mci_resource_bind = bind;
3652 	mcip->mci_resource_arg = arg;
3653 }
3654 
3655 void
3656 mac_resource_set(mac_client_handle_t mch, mac_resource_add_t add, void *arg)
3657 {
3658 	/* update the 'resource_add' callback */
3659 	mac_resource_set_common(mch, add, NULL, NULL, NULL, NULL, arg);
3660 }
3661 
3662 /*
3663  * Sets up the client resources and enable the polling interface over all the
3664  * SRS's and the soft rings of the client
3665  */
3666 void
3667 mac_client_poll_enable(mac_client_handle_t mch)
3668 {
3669 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3670 	mac_soft_ring_set_t	*mac_srs;
3671 	flow_entry_t		*flent;
3672 	int			i;
3673 
3674 	flent = mcip->mci_flent;
3675 	ASSERT(flent != NULL);
3676 
3677 	mcip->mci_state_flags |= MCIS_CLIENT_POLL_CAPABLE;
3678 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3679 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3680 		ASSERT(mac_srs->srs_mcip == mcip);
3681 		mac_srs_client_poll_enable(mcip, mac_srs);
3682 	}
3683 }
3684 
3685 /*
3686  * Tears down the client resources and disable the polling interface over all
3687  * the SRS's and the soft rings of the client
3688  */
3689 void
3690 mac_client_poll_disable(mac_client_handle_t mch)
3691 {
3692 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3693 	mac_soft_ring_set_t	*mac_srs;
3694 	flow_entry_t		*flent;
3695 	int			i;
3696 
3697 	flent = mcip->mci_flent;
3698 	ASSERT(flent != NULL);
3699 
3700 	mcip->mci_state_flags &= ~MCIS_CLIENT_POLL_CAPABLE;
3701 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3702 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3703 		ASSERT(mac_srs->srs_mcip == mcip);
3704 		mac_srs_client_poll_disable(mcip, mac_srs);
3705 	}
3706 }
3707 
3708 /*
3709  * Associate the CPUs specified by the given property with a MAC client.
3710  */
3711 int
3712 mac_cpu_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
3713 {
3714 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3715 	mac_impl_t *mip = mcip->mci_mip;
3716 	int err = 0;
3717 
3718 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3719 
3720 	if ((err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
3721 	    mcip->mci_upper_mip : mip, mrp)) != 0) {
3722 		return (err);
3723 	}
3724 	if (MCIP_DATAPATH_SETUP(mcip))
3725 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
3726 
3727 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
3728 	return (0);
3729 }
3730 
3731 /*
3732  * Apply the specified properties to the specified MAC client.
3733  */
3734 int
3735 mac_client_set_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3736 {
3737 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3738 	mac_impl_t *mip = mcip->mci_mip;
3739 	int err = 0;
3740 
3741 	i_mac_perim_enter(mip);
3742 
3743 	if ((mrp->mrp_mask & MRP_MAXBW) || (mrp->mrp_mask & MRP_PRIORITY)) {
3744 		err = mac_resource_ctl_set(mch, mrp);
3745 		if (err != 0)
3746 			goto done;
3747 	}
3748 
3749 	if (mrp->mrp_mask & (MRP_CPUS|MRP_POOL)) {
3750 		err = mac_cpu_set(mch, mrp);
3751 		if (err != 0)
3752 			goto done;
3753 	}
3754 
3755 	if (mrp->mrp_mask & MRP_PROTECT) {
3756 		err = mac_protect_set(mch, mrp);
3757 		if (err != 0)
3758 			goto done;
3759 	}
3760 
3761 	if ((mrp->mrp_mask & MRP_RX_RINGS) || (mrp->mrp_mask & MRP_TX_RINGS))
3762 		err = mac_resource_ctl_set(mch, mrp);
3763 
3764 done:
3765 	i_mac_perim_exit(mip);
3766 	return (err);
3767 }
3768 
3769 /*
3770  * Return the properties currently associated with the specified MAC client.
3771  */
3772 void
3773 mac_client_get_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3774 {
3775 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3776 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
3777 
3778 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3779 }
3780 
3781 /*
3782  * Return the effective properties currently associated with the specified
3783  * MAC client.
3784  */
3785 void
3786 mac_client_get_effective_resources(mac_client_handle_t mch,
3787     mac_resource_props_t *mrp)
3788 {
3789 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3790 	mac_resource_props_t	*mcip_mrp = MCIP_EFFECTIVE_PROPS(mcip);
3791 
3792 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3793 }
3794 
3795 /*
3796  * Pass a copy of the specified packet to the promiscuous callbacks
3797  * of the specified MAC.
3798  *
3799  * If sender is NULL, the function is being invoked for a packet chain
3800  * received from the wire. If sender is non-NULL, it points to
3801  * the MAC client from which the packet is being sent.
3802  *
3803  * The packets are distributed to the promiscuous callbacks as follows:
3804  *
3805  * - all packets are sent to the MAC_CLIENT_PROMISC_ALL callbacks
3806  * - all broadcast and multicast packets are sent to the
3807  *   MAC_CLIENT_PROMISC_FILTER and MAC_CLIENT_PROMISC_MULTI.
3808  *
3809  * The unicast packets of MAC_CLIENT_PROMISC_FILTER callbacks are dispatched
3810  * after classification by mac_rx_deliver().
3811  */
3812 
3813 static void
3814 mac_promisc_dispatch_one(mac_promisc_impl_t *mpip, mblk_t *mp,
3815     boolean_t loopback)
3816 {
3817 	mblk_t *mp_copy, *mp_next;
3818 
3819 	if (!mpip->mpi_no_copy || mpip->mpi_strip_vlan_tag) {
3820 		mp_copy = copymsg(mp);
3821 		if (mp_copy == NULL)
3822 			return;
3823 
3824 		if (mpip->mpi_strip_vlan_tag) {
3825 			mp_copy = mac_strip_vlan_tag_chain(mp_copy);
3826 			if (mp_copy == NULL)
3827 				return;
3828 		}
3829 		mp_next = NULL;
3830 	} else {
3831 		mp_copy = mp;
3832 		mp_next = mp->b_next;
3833 	}
3834 	mp_copy->b_next = NULL;
3835 
3836 	mpip->mpi_fn(mpip->mpi_arg, NULL, mp_copy, loopback);
3837 	if (mp_copy == mp)
3838 		mp->b_next = mp_next;
3839 }
3840 
3841 /*
3842  * Return the VID of a packet. Zero if the packet is not tagged.
3843  */
3844 static uint16_t
3845 mac_ether_vid(mblk_t *mp)
3846 {
3847 	struct ether_header *eth = (struct ether_header *)mp->b_rptr;
3848 
3849 	if (ntohs(eth->ether_type) == ETHERTYPE_VLAN) {
3850 		struct ether_vlan_header *t_evhp =
3851 		    (struct ether_vlan_header *)mp->b_rptr;
3852 		return (VLAN_ID(ntohs(t_evhp->ether_tci)));
3853 	}
3854 
3855 	return (0);
3856 }
3857 
3858 /*
3859  * Return whether the specified packet contains a multicast or broadcast
3860  * destination MAC address.
3861  */
3862 static boolean_t
3863 mac_is_mcast(mac_impl_t *mip, mblk_t *mp)
3864 {
3865 	mac_header_info_t hdr_info;
3866 
3867 	if (mac_header_info((mac_handle_t)mip, mp, &hdr_info) != 0)
3868 		return (B_FALSE);
3869 	return ((hdr_info.mhi_dsttype == MAC_ADDRTYPE_BROADCAST) ||
3870 	    (hdr_info.mhi_dsttype == MAC_ADDRTYPE_MULTICAST));
3871 }
3872 
3873 /*
3874  * Send a copy of an mblk chain to the MAC clients of the specified MAC.
3875  * "sender" points to the sender MAC client for outbound packets, and
3876  * is set to NULL for inbound packets.
3877  */
3878 void
3879 mac_promisc_dispatch(mac_impl_t *mip, mblk_t *mp_chain,
3880     mac_client_impl_t *sender)
3881 {
3882 	mac_promisc_impl_t *mpip;
3883 	mac_cb_t *mcb;
3884 	mblk_t *mp;
3885 	boolean_t is_mcast, is_sender;
3886 
3887 	MAC_PROMISC_WALKER_INC(mip);
3888 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3889 		is_mcast = mac_is_mcast(mip, mp);
3890 		/* send packet to interested callbacks */
3891 		for (mcb = mip->mi_promisc_list; mcb != NULL;
3892 		    mcb = mcb->mcb_nextp) {
3893 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3894 			is_sender = (mpip->mpi_mcip == sender);
3895 
3896 			if (is_sender && mpip->mpi_no_tx_loop)
3897 				/*
3898 				 * The sender doesn't want to receive
3899 				 * copies of the packets it sends.
3900 				 */
3901 				continue;
3902 
3903 			/* this client doesn't need any packets (bridge) */
3904 			if (mpip->mpi_fn == NULL)
3905 				continue;
3906 
3907 			/*
3908 			 * For an ethernet MAC, don't displatch a multicast
3909 			 * packet to a non-PROMISC_ALL callbacks unless the VID
3910 			 * of the packet matches the VID of the client.
3911 			 */
3912 			if (is_mcast &&
3913 			    mpip->mpi_type != MAC_CLIENT_PROMISC_ALL &&
3914 			    !mac_client_check_flow_vid(mpip->mpi_mcip,
3915 			    mac_ether_vid(mp)))
3916 				continue;
3917 
3918 			if (is_sender ||
3919 			    mpip->mpi_type == MAC_CLIENT_PROMISC_ALL ||
3920 			    is_mcast)
3921 				mac_promisc_dispatch_one(mpip, mp, is_sender);
3922 		}
3923 	}
3924 	MAC_PROMISC_WALKER_DCR(mip);
3925 }
3926 
3927 void
3928 mac_promisc_client_dispatch(mac_client_impl_t *mcip, mblk_t *mp_chain)
3929 {
3930 	mac_impl_t		*mip = mcip->mci_mip;
3931 	mac_promisc_impl_t	*mpip;
3932 	boolean_t		is_mcast;
3933 	mblk_t			*mp;
3934 	mac_cb_t		*mcb;
3935 
3936 	/*
3937 	 * The unicast packets for the MAC client still
3938 	 * need to be delivered to the MAC_CLIENT_PROMISC_FILTERED
3939 	 * promiscuous callbacks. The broadcast and multicast
3940 	 * packets were delivered from mac_rx().
3941 	 */
3942 	MAC_PROMISC_WALKER_INC(mip);
3943 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3944 		is_mcast = mac_is_mcast(mip, mp);
3945 		for (mcb = mcip->mci_promisc_list; mcb != NULL;
3946 		    mcb = mcb->mcb_nextp) {
3947 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3948 			if (mpip->mpi_type == MAC_CLIENT_PROMISC_FILTERED &&
3949 			    !is_mcast) {
3950 				mac_promisc_dispatch_one(mpip, mp, B_FALSE);
3951 			}
3952 		}
3953 	}
3954 	MAC_PROMISC_WALKER_DCR(mip);
3955 }
3956 
3957 /*
3958  * Return the margin value currently assigned to the specified MAC instance.
3959  */
3960 void
3961 mac_margin_get(mac_handle_t mh, uint32_t *marginp)
3962 {
3963 	mac_impl_t *mip = (mac_impl_t *)mh;
3964 
3965 	rw_enter(&(mip->mi_rw_lock), RW_READER);
3966 	*marginp = mip->mi_margin;
3967 	rw_exit(&(mip->mi_rw_lock));
3968 }
3969 
3970 /*
3971  * mac_info_get() is used for retrieving the mac_info when a DL_INFO_REQ is
3972  * issued before a DL_ATTACH_REQ. we walk the i_mac_impl_hash table and find
3973  * the first mac_impl_t with a matching driver name; then we copy its mac_info_t
3974  * to the caller. we do all this with i_mac_impl_lock held so the mac_impl_t
3975  * cannot disappear while we are accessing it.
3976  */
3977 typedef struct i_mac_info_state_s {
3978 	const char	*mi_name;
3979 	mac_info_t	*mi_infop;
3980 } i_mac_info_state_t;
3981 
3982 /*ARGSUSED*/
3983 static uint_t
3984 i_mac_info_walker(mod_hash_key_t key, mod_hash_val_t *val, void *arg)
3985 {
3986 	i_mac_info_state_t *statep = arg;
3987 	mac_impl_t *mip = (mac_impl_t *)val;
3988 
3989 	if (mip->mi_state_flags & MIS_DISABLED)
3990 		return (MH_WALK_CONTINUE);
3991 
3992 	if (strcmp(statep->mi_name,
3993 	    ddi_driver_name(mip->mi_dip)) != 0)
3994 		return (MH_WALK_CONTINUE);
3995 
3996 	statep->mi_infop = &mip->mi_info;
3997 	return (MH_WALK_TERMINATE);
3998 }
3999 
4000 boolean_t
4001 mac_info_get(const char *name, mac_info_t *minfop)
4002 {
4003 	i_mac_info_state_t state;
4004 
4005 	rw_enter(&i_mac_impl_lock, RW_READER);
4006 	state.mi_name = name;
4007 	state.mi_infop = NULL;
4008 	mod_hash_walk(i_mac_impl_hash, i_mac_info_walker, &state);
4009 	if (state.mi_infop == NULL) {
4010 		rw_exit(&i_mac_impl_lock);
4011 		return (B_FALSE);
4012 	}
4013 	*minfop = *state.mi_infop;
4014 	rw_exit(&i_mac_impl_lock);
4015 	return (B_TRUE);
4016 }
4017 
4018 /*
4019  * To get the capabilities that MAC layer cares about, such as rings, factory
4020  * mac address, vnic or not, it should directly invoke this function.  If the
4021  * link is part of a bridge, then the only "capability" it has is the inability
4022  * to do zero copy.
4023  */
4024 boolean_t
4025 i_mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4026 {
4027 	mac_impl_t *mip = (mac_impl_t *)mh;
4028 
4029 	if (mip->mi_bridge_link != NULL)
4030 		return (cap == MAC_CAPAB_NO_ZCOPY);
4031 	else if (mip->mi_callbacks->mc_callbacks & MC_GETCAPAB)
4032 		return (mip->mi_getcapab(mip->mi_driver, cap, cap_data));
4033 	else
4034 		return (B_FALSE);
4035 }
4036 
4037 /*
4038  * Capability query function. If number of active mac clients is greater than
4039  * 1, only limited capabilities can be advertised to the caller no matter the
4040  * driver has certain capability or not. Else, we query the driver to get the
4041  * capability.
4042  */
4043 boolean_t
4044 mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4045 {
4046 	mac_impl_t *mip = (mac_impl_t *)mh;
4047 
4048 	/*
4049 	 * if mi_nactiveclients > 1, only MAC_CAPAB_LEGACY, MAC_CAPAB_HCKSUM,
4050 	 * MAC_CAPAB_NO_NATIVEVLAN and MAC_CAPAB_NO_ZCOPY can be advertised.
4051 	 */
4052 	if (mip->mi_nactiveclients > 1) {
4053 		switch (cap) {
4054 		case MAC_CAPAB_NO_ZCOPY:
4055 			return (B_TRUE);
4056 		case MAC_CAPAB_LEGACY:
4057 		case MAC_CAPAB_HCKSUM:
4058 		case MAC_CAPAB_NO_NATIVEVLAN:
4059 			break;
4060 		default:
4061 			return (B_FALSE);
4062 		}
4063 	}
4064 
4065 	/* else get capab from driver */
4066 	return (i_mac_capab_get(mh, cap, cap_data));
4067 }
4068 
4069 boolean_t
4070 mac_sap_verify(mac_handle_t mh, uint32_t sap, uint32_t *bind_sap)
4071 {
4072 	mac_impl_t *mip = (mac_impl_t *)mh;
4073 
4074 	return (mip->mi_type->mt_ops.mtops_sap_verify(sap, bind_sap,
4075 	    mip->mi_pdata));
4076 }
4077 
4078 mblk_t *
4079 mac_header(mac_handle_t mh, const uint8_t *daddr, uint32_t sap, mblk_t *payload,
4080     size_t extra_len)
4081 {
4082 	mac_impl_t	*mip = (mac_impl_t *)mh;
4083 	const uint8_t	*hdr_daddr;
4084 
4085 	/*
4086 	 * If the MAC is point-to-point with a fixed destination address, then
4087 	 * we must always use that destination in the MAC header.
4088 	 */
4089 	hdr_daddr = (mip->mi_dstaddr_set ? mip->mi_dstaddr : daddr);
4090 	return (mip->mi_type->mt_ops.mtops_header(mip->mi_addr, hdr_daddr, sap,
4091 	    mip->mi_pdata, payload, extra_len));
4092 }
4093 
4094 int
4095 mac_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4096 {
4097 	mac_impl_t *mip = (mac_impl_t *)mh;
4098 
4099 	return (mip->mi_type->mt_ops.mtops_header_info(mp, mip->mi_pdata,
4100 	    mhip));
4101 }
4102 
4103 int
4104 mac_vlan_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4105 {
4106 	mac_impl_t	*mip = (mac_impl_t *)mh;
4107 	boolean_t	is_ethernet = (mip->mi_info.mi_media == DL_ETHER);
4108 	int		err = 0;
4109 
4110 	/*
4111 	 * Packets should always be at least 16 bit aligned.
4112 	 */
4113 	ASSERT(IS_P2ALIGNED(mp->b_rptr, sizeof (uint16_t)));
4114 
4115 	if ((err = mac_header_info(mh, mp, mhip)) != 0)
4116 		return (err);
4117 
4118 	/*
4119 	 * If this is a VLAN-tagged Ethernet packet, then the SAP in the
4120 	 * mac_header_info_t as returned by mac_header_info() is
4121 	 * ETHERTYPE_VLAN. We need to grab the ethertype from the VLAN header.
4122 	 */
4123 	if (is_ethernet && (mhip->mhi_bindsap == ETHERTYPE_VLAN)) {
4124 		struct ether_vlan_header *evhp;
4125 		uint16_t sap;
4126 		mblk_t *tmp = NULL;
4127 		size_t size;
4128 
4129 		size = sizeof (struct ether_vlan_header);
4130 		if (MBLKL(mp) < size) {
4131 			/*
4132 			 * Pullup the message in order to get the MAC header
4133 			 * infomation. Note that this is a read-only function,
4134 			 * we keep the input packet intact.
4135 			 */
4136 			if ((tmp = msgpullup(mp, size)) == NULL)
4137 				return (EINVAL);
4138 
4139 			mp = tmp;
4140 		}
4141 		evhp = (struct ether_vlan_header *)mp->b_rptr;
4142 		sap = ntohs(evhp->ether_type);
4143 		(void) mac_sap_verify(mh, sap, &mhip->mhi_bindsap);
4144 		mhip->mhi_hdrsize = sizeof (struct ether_vlan_header);
4145 		mhip->mhi_tci = ntohs(evhp->ether_tci);
4146 		mhip->mhi_istagged = B_TRUE;
4147 		freemsg(tmp);
4148 
4149 		if (VLAN_CFI(mhip->mhi_tci) != ETHER_CFI)
4150 			return (EINVAL);
4151 	} else {
4152 		mhip->mhi_istagged = B_FALSE;
4153 		mhip->mhi_tci = 0;
4154 	}
4155 
4156 	return (0);
4157 }
4158 
4159 mblk_t *
4160 mac_header_cook(mac_handle_t mh, mblk_t *mp)
4161 {
4162 	mac_impl_t *mip = (mac_impl_t *)mh;
4163 
4164 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_COOK) {
4165 		if (DB_REF(mp) > 1) {
4166 			mblk_t *newmp = copymsg(mp);
4167 			if (newmp == NULL)
4168 				return (NULL);
4169 			freemsg(mp);
4170 			mp = newmp;
4171 		}
4172 		return (mip->mi_type->mt_ops.mtops_header_cook(mp,
4173 		    mip->mi_pdata));
4174 	}
4175 	return (mp);
4176 }
4177 
4178 mblk_t *
4179 mac_header_uncook(mac_handle_t mh, mblk_t *mp)
4180 {
4181 	mac_impl_t *mip = (mac_impl_t *)mh;
4182 
4183 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_UNCOOK) {
4184 		if (DB_REF(mp) > 1) {
4185 			mblk_t *newmp = copymsg(mp);
4186 			if (newmp == NULL)
4187 				return (NULL);
4188 			freemsg(mp);
4189 			mp = newmp;
4190 		}
4191 		return (mip->mi_type->mt_ops.mtops_header_uncook(mp,
4192 		    mip->mi_pdata));
4193 	}
4194 	return (mp);
4195 }
4196 
4197 uint_t
4198 mac_addr_len(mac_handle_t mh)
4199 {
4200 	mac_impl_t *mip = (mac_impl_t *)mh;
4201 
4202 	return (mip->mi_type->mt_addr_length);
4203 }
4204 
4205 /* True if a MAC is a VNIC */
4206 boolean_t
4207 mac_is_vnic(mac_handle_t mh)
4208 {
4209 	return (((mac_impl_t *)mh)->mi_state_flags & MIS_IS_VNIC);
4210 }
4211 
4212 mac_handle_t
4213 mac_get_lower_mac_handle(mac_handle_t mh)
4214 {
4215 	mac_impl_t *mip = (mac_impl_t *)mh;
4216 
4217 	ASSERT(mac_is_vnic(mh));
4218 	return (((vnic_t *)mip->mi_driver)->vn_lower_mh);
4219 }
4220 
4221 boolean_t
4222 mac_is_vnic_primary(mac_handle_t mh)
4223 {
4224 	mac_impl_t *mip = (mac_impl_t *)mh;
4225 
4226 	ASSERT(mac_is_vnic(mh));
4227 	return (((vnic_t *)mip->mi_driver)->vn_addr_type ==
4228 	    VNIC_MAC_ADDR_TYPE_PRIMARY);
4229 }
4230 
4231 void
4232 mac_update_resources(mac_resource_props_t *nmrp, mac_resource_props_t *cmrp,
4233     boolean_t is_user_flow)
4234 {
4235 	if (nmrp != NULL && cmrp != NULL) {
4236 		if (nmrp->mrp_mask & MRP_PRIORITY) {
4237 			if (nmrp->mrp_priority == MPL_RESET) {
4238 				cmrp->mrp_mask &= ~MRP_PRIORITY;
4239 				if (is_user_flow) {
4240 					cmrp->mrp_priority =
4241 					    MPL_SUBFLOW_DEFAULT;
4242 				} else {
4243 					cmrp->mrp_priority = MPL_LINK_DEFAULT;
4244 				}
4245 			} else {
4246 				cmrp->mrp_mask |= MRP_PRIORITY;
4247 				cmrp->mrp_priority = nmrp->mrp_priority;
4248 			}
4249 		}
4250 		if (nmrp->mrp_mask & MRP_MAXBW) {
4251 			if (nmrp->mrp_maxbw == MRP_MAXBW_RESETVAL) {
4252 				cmrp->mrp_mask &= ~MRP_MAXBW;
4253 				cmrp->mrp_maxbw = 0;
4254 			} else {
4255 				cmrp->mrp_mask |= MRP_MAXBW;
4256 				cmrp->mrp_maxbw = nmrp->mrp_maxbw;
4257 			}
4258 		}
4259 		if (nmrp->mrp_mask & MRP_CPUS)
4260 			MAC_COPY_CPUS(nmrp, cmrp);
4261 
4262 		if (nmrp->mrp_mask & MRP_POOL) {
4263 			if (strlen(nmrp->mrp_pool) == 0) {
4264 				cmrp->mrp_mask &= ~MRP_POOL;
4265 				bzero(cmrp->mrp_pool, sizeof (cmrp->mrp_pool));
4266 			} else {
4267 				cmrp->mrp_mask |= MRP_POOL;
4268 				(void) strncpy(cmrp->mrp_pool, nmrp->mrp_pool,
4269 				    sizeof (cmrp->mrp_pool));
4270 			}
4271 
4272 		}
4273 
4274 		if (nmrp->mrp_mask & MRP_PROTECT)
4275 			mac_protect_update(nmrp, cmrp);
4276 
4277 		/*
4278 		 * Update the rings specified.
4279 		 */
4280 		if (nmrp->mrp_mask & MRP_RX_RINGS) {
4281 			if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4282 				cmrp->mrp_mask &= ~MRP_RX_RINGS;
4283 				if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4284 					cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4285 				cmrp->mrp_nrxrings = 0;
4286 			} else {
4287 				cmrp->mrp_mask |= MRP_RX_RINGS;
4288 				cmrp->mrp_nrxrings = nmrp->mrp_nrxrings;
4289 			}
4290 		}
4291 		if (nmrp->mrp_mask & MRP_TX_RINGS) {
4292 			if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4293 				cmrp->mrp_mask &= ~MRP_TX_RINGS;
4294 				if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4295 					cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4296 				cmrp->mrp_ntxrings = 0;
4297 			} else {
4298 				cmrp->mrp_mask |= MRP_TX_RINGS;
4299 				cmrp->mrp_ntxrings = nmrp->mrp_ntxrings;
4300 			}
4301 		}
4302 		if (nmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4303 			cmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4304 		else if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4305 			cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4306 
4307 		if (nmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4308 			cmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4309 		else if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4310 			cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4311 	}
4312 }
4313 
4314 /*
4315  * i_mac_set_resources:
4316  *
4317  * This routine associates properties with the primary MAC client of
4318  * the specified MAC instance.
4319  * - Cache the properties in mac_impl_t
4320  * - Apply the properties to the primary MAC client if exists
4321  */
4322 int
4323 i_mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4324 {
4325 	mac_impl_t		*mip = (mac_impl_t *)mh;
4326 	mac_client_impl_t	*mcip;
4327 	int			err = 0;
4328 	uint32_t		resmask, newresmask;
4329 	mac_resource_props_t	*tmrp, *umrp;
4330 
4331 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4332 
4333 	err = mac_validate_props(mip, mrp);
4334 	if (err != 0)
4335 		return (err);
4336 
4337 	umrp = kmem_zalloc(sizeof (*umrp), KM_SLEEP);
4338 	bcopy(&mip->mi_resource_props, umrp, sizeof (*umrp));
4339 	resmask = umrp->mrp_mask;
4340 	mac_update_resources(mrp, umrp, B_FALSE);
4341 	newresmask = umrp->mrp_mask;
4342 
4343 	if (resmask == 0 && newresmask != 0) {
4344 		/*
4345 		 * Bandwidth, priority, cpu or pool link properties configured,
4346 		 * must disable fastpath.
4347 		 */
4348 		if ((err = mac_fastpath_disable((mac_handle_t)mip)) != 0) {
4349 			kmem_free(umrp, sizeof (*umrp));
4350 			return (err);
4351 		}
4352 	}
4353 
4354 	/*
4355 	 * Since bind_cpu may be modified by mac_client_set_resources()
4356 	 * we use a copy of bind_cpu and finally cache bind_cpu in mip.
4357 	 * This allows us to cache only user edits in mip.
4358 	 */
4359 	tmrp = kmem_zalloc(sizeof (*tmrp), KM_SLEEP);
4360 	bcopy(mrp, tmrp, sizeof (*tmrp));
4361 	mcip = mac_primary_client_handle(mip);
4362 	if (mcip != NULL && (mcip->mci_state_flags & MCIS_IS_AGGR_PORT) == 0) {
4363 		err = mac_client_set_resources((mac_client_handle_t)mcip, tmrp);
4364 	} else if ((mrp->mrp_mask & MRP_RX_RINGS ||
4365 	    mrp->mrp_mask & MRP_TX_RINGS)) {
4366 		mac_client_impl_t	*vmcip;
4367 
4368 		/*
4369 		 * If the primary is not up, we need to check if there
4370 		 * are any VLANs on this primary. If there are then
4371 		 * we need to set this property on the VLANs since
4372 		 * VLANs follow the primary they are based on. Just
4373 		 * look for the first VLAN and change its properties,
4374 		 * all the other VLANs should be in the same group.
4375 		 */
4376 		for (vmcip = mip->mi_clients_list; vmcip != NULL;
4377 		    vmcip = vmcip->mci_client_next) {
4378 			if ((vmcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) &&
4379 			    mac_client_vid((mac_client_handle_t)vmcip) !=
4380 			    VLAN_ID_NONE) {
4381 				break;
4382 			}
4383 		}
4384 		if (vmcip != NULL) {
4385 			mac_resource_props_t	*omrp;
4386 			mac_resource_props_t	*vmrp;
4387 
4388 			omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
4389 			bcopy(MCIP_RESOURCE_PROPS(vmcip), omrp, sizeof (*omrp));
4390 			/*
4391 			 * We dont' call mac_update_resources since we
4392 			 * want to take only the ring properties and
4393 			 * not all the properties that may have changed.
4394 			 */
4395 			vmrp = MCIP_RESOURCE_PROPS(vmcip);
4396 			if (mrp->mrp_mask & MRP_RX_RINGS) {
4397 				if (mrp->mrp_mask & MRP_RINGS_RESET) {
4398 					vmrp->mrp_mask &= ~MRP_RX_RINGS;
4399 					if (vmrp->mrp_mask &
4400 					    MRP_RXRINGS_UNSPEC) {
4401 						vmrp->mrp_mask &=
4402 						    ~MRP_RXRINGS_UNSPEC;
4403 					}
4404 					vmrp->mrp_nrxrings = 0;
4405 				} else {
4406 					vmrp->mrp_mask |= MRP_RX_RINGS;
4407 					vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
4408 				}
4409 			}
4410 			if (mrp->mrp_mask & MRP_TX_RINGS) {
4411 				if (mrp->mrp_mask & MRP_RINGS_RESET) {
4412 					vmrp->mrp_mask &= ~MRP_TX_RINGS;
4413 					if (vmrp->mrp_mask &
4414 					    MRP_TXRINGS_UNSPEC) {
4415 						vmrp->mrp_mask &=
4416 						    ~MRP_TXRINGS_UNSPEC;
4417 					}
4418 					vmrp->mrp_ntxrings = 0;
4419 				} else {
4420 					vmrp->mrp_mask |= MRP_TX_RINGS;
4421 					vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
4422 				}
4423 			}
4424 			if (mrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4425 				vmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4426 
4427 			if (mrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4428 				vmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4429 
4430 			if ((err = mac_client_set_rings_prop(vmcip, mrp,
4431 			    omrp)) != 0) {
4432 				bcopy(omrp, MCIP_RESOURCE_PROPS(vmcip),
4433 				    sizeof (*omrp));
4434 			} else {
4435 				mac_set_prim_vlan_rings(mip, vmrp);
4436 			}
4437 			kmem_free(omrp, sizeof (*omrp));
4438 		}
4439 	}
4440 
4441 	/* Only update the values if mac_client_set_resources succeeded */
4442 	if (err == 0) {
4443 		bcopy(umrp, &mip->mi_resource_props, sizeof (*umrp));
4444 		/*
4445 		 * If bandwidth, priority or cpu link properties cleared,
4446 		 * renable fastpath.
4447 		 */
4448 		if (resmask != 0 && newresmask == 0)
4449 			mac_fastpath_enable((mac_handle_t)mip);
4450 	} else if (resmask == 0 && newresmask != 0) {
4451 		mac_fastpath_enable((mac_handle_t)mip);
4452 	}
4453 	kmem_free(tmrp, sizeof (*tmrp));
4454 	kmem_free(umrp, sizeof (*umrp));
4455 	return (err);
4456 }
4457 
4458 int
4459 mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4460 {
4461 	int err;
4462 
4463 	i_mac_perim_enter((mac_impl_t *)mh);
4464 	err = i_mac_set_resources(mh, mrp);
4465 	i_mac_perim_exit((mac_impl_t *)mh);
4466 	return (err);
4467 }
4468 
4469 /*
4470  * Get the properties cached for the specified MAC instance.
4471  */
4472 void
4473 mac_get_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4474 {
4475 	mac_impl_t 		*mip = (mac_impl_t *)mh;
4476 	mac_client_impl_t	*mcip;
4477 
4478 	mcip = mac_primary_client_handle(mip);
4479 	if (mcip != NULL) {
4480 		mac_client_get_resources((mac_client_handle_t)mcip, mrp);
4481 		return;
4482 	}
4483 	bcopy(&mip->mi_resource_props, mrp, sizeof (mac_resource_props_t));
4484 }
4485 
4486 /*
4487  * Get the effective properties from the primary client of the
4488  * specified MAC instance.
4489  */
4490 void
4491 mac_get_effective_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4492 {
4493 	mac_impl_t 		*mip = (mac_impl_t *)mh;
4494 	mac_client_impl_t	*mcip;
4495 
4496 	mcip = mac_primary_client_handle(mip);
4497 	if (mcip != NULL) {
4498 		mac_client_get_effective_resources((mac_client_handle_t)mcip,
4499 		    mrp);
4500 		return;
4501 	}
4502 	bzero(mrp, sizeof (mac_resource_props_t));
4503 }
4504 
4505 int
4506 mac_set_pvid(mac_handle_t mh, uint16_t pvid)
4507 {
4508 	mac_impl_t *mip = (mac_impl_t *)mh;
4509 	mac_client_impl_t *mcip;
4510 	mac_unicast_impl_t *muip;
4511 
4512 	i_mac_perim_enter(mip);
4513 	if (pvid != 0) {
4514 		for (mcip = mip->mi_clients_list; mcip != NULL;
4515 		    mcip = mcip->mci_client_next) {
4516 			for (muip = mcip->mci_unicast_list; muip != NULL;
4517 			    muip = muip->mui_next) {
4518 				if (muip->mui_vid == pvid) {
4519 					i_mac_perim_exit(mip);
4520 					return (EBUSY);
4521 				}
4522 			}
4523 		}
4524 	}
4525 	mip->mi_pvid = pvid;
4526 	i_mac_perim_exit(mip);
4527 	return (0);
4528 }
4529 
4530 uint16_t
4531 mac_get_pvid(mac_handle_t mh)
4532 {
4533 	mac_impl_t *mip = (mac_impl_t *)mh;
4534 
4535 	return (mip->mi_pvid);
4536 }
4537 
4538 uint32_t
4539 mac_get_llimit(mac_handle_t mh)
4540 {
4541 	mac_impl_t *mip = (mac_impl_t *)mh;
4542 
4543 	return (mip->mi_llimit);
4544 }
4545 
4546 uint32_t
4547 mac_get_ldecay(mac_handle_t mh)
4548 {
4549 	mac_impl_t *mip = (mac_impl_t *)mh;
4550 
4551 	return (mip->mi_ldecay);
4552 }
4553 
4554 /*
4555  * Rename a mac client, its flow, and the kstat.
4556  */
4557 int
4558 mac_rename_primary(mac_handle_t mh, const char *new_name)
4559 {
4560 	mac_impl_t		*mip = (mac_impl_t *)mh;
4561 	mac_client_impl_t	*cur_clnt = NULL;
4562 	flow_entry_t		*fep;
4563 
4564 	i_mac_perim_enter(mip);
4565 
4566 	/*
4567 	 * VNICs: we need to change the sys flow name and
4568 	 * the associated flow kstat.
4569 	 */
4570 	if (mip->mi_state_flags & MIS_IS_VNIC) {
4571 		mac_client_impl_t *mcip = mac_vnic_lower(mip);
4572 		ASSERT(new_name != NULL);
4573 		mac_rename_flow_names(mcip, new_name);
4574 		mac_stat_rename(mcip);
4575 		goto done;
4576 	}
4577 	/*
4578 	 * This mac may itself be an aggr link, or it may have some client
4579 	 * which is an aggr port. For both cases, we need to change the
4580 	 * aggr port's mac client name, its flow name and the associated flow
4581 	 * kstat.
4582 	 */
4583 	if (mip->mi_state_flags & MIS_IS_AGGR) {
4584 		mac_capab_aggr_t aggr_cap;
4585 		mac_rename_fn_t rename_fn;
4586 		boolean_t ret;
4587 
4588 		ASSERT(new_name != NULL);
4589 		ret = i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_AGGR,
4590 		    (void *)(&aggr_cap));
4591 		ASSERT(ret == B_TRUE);
4592 		rename_fn = aggr_cap.mca_rename_fn;
4593 		rename_fn(new_name, mip->mi_driver);
4594 		/*
4595 		 * The aggr's client name and kstat flow name will be
4596 		 * updated below, i.e. via mac_rename_flow_names.
4597 		 */
4598 	}
4599 
4600 	for (cur_clnt = mip->mi_clients_list; cur_clnt != NULL;
4601 	    cur_clnt = cur_clnt->mci_client_next) {
4602 		if (cur_clnt->mci_state_flags & MCIS_IS_AGGR_PORT) {
4603 			if (new_name != NULL) {
4604 				char *str_st = cur_clnt->mci_name;
4605 				char *str_del = strchr(str_st, '-');
4606 
4607 				ASSERT(str_del != NULL);
4608 				bzero(str_del + 1, MAXNAMELEN -
4609 				    (str_del - str_st + 1));
4610 				bcopy(new_name, str_del + 1,
4611 				    strlen(new_name));
4612 			}
4613 			fep = cur_clnt->mci_flent;
4614 			mac_rename_flow(fep, cur_clnt->mci_name);
4615 			break;
4616 		} else if (new_name != NULL &&
4617 		    cur_clnt->mci_state_flags & MCIS_USE_DATALINK_NAME) {
4618 			mac_rename_flow_names(cur_clnt, new_name);
4619 			break;
4620 		}
4621 	}
4622 
4623 	/* Recreate kstats associated with aggr pseudo rings */
4624 	if (mip->mi_state_flags & MIS_IS_AGGR)
4625 		mac_pseudo_ring_stat_rename(mip);
4626 
4627 done:
4628 	i_mac_perim_exit(mip);
4629 	return (0);
4630 }
4631 
4632 /*
4633  * Rename the MAC client's flow names
4634  */
4635 static void
4636 mac_rename_flow_names(mac_client_impl_t *mcip, const char *new_name)
4637 {
4638 	flow_entry_t	*flent;
4639 	uint16_t	vid;
4640 	char		flowname[MAXFLOWNAMELEN];
4641 	mac_impl_t	*mip = mcip->mci_mip;
4642 
4643 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4644 
4645 	/*
4646 	 * Use mi_rw_lock to ensure that threads not in the mac perimeter
4647 	 * see a self-consistent value for mci_name
4648 	 */
4649 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
4650 	(void) strlcpy(mcip->mci_name, new_name, sizeof (mcip->mci_name));
4651 	rw_exit(&mip->mi_rw_lock);
4652 
4653 	mac_rename_flow(mcip->mci_flent, new_name);
4654 
4655 	if (mcip->mci_nflents == 1)
4656 		return;
4657 
4658 	/*
4659 	 * We have to rename all the others too, no stats to destroy for
4660 	 * these.
4661 	 */
4662 	for (flent = mcip->mci_flent_list; flent != NULL;
4663 	    flent = flent->fe_client_next) {
4664 		if (flent != mcip->mci_flent) {
4665 			vid = i_mac_flow_vid(flent);
4666 			(void) sprintf(flowname, "%s%u", new_name, vid);
4667 			mac_flow_set_name(flent, flowname);
4668 		}
4669 	}
4670 }
4671 
4672 
4673 /*
4674  * Add a flow to the MAC client's flow list - i.e list of MAC/VID tuples
4675  * defined for the specified MAC client.
4676  */
4677 static void
4678 mac_client_add_to_flow_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4679 {
4680 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4681 	/*
4682 	 * The promisc Rx data path walks the mci_flent_list. Protect by
4683 	 * using mi_rw_lock
4684 	 */
4685 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4686 
4687 	/* Add it to the head */
4688 	flent->fe_client_next = mcip->mci_flent_list;
4689 	mcip->mci_flent_list = flent;
4690 	mcip->mci_nflents++;
4691 
4692 	/*
4693 	 * Keep track of the number of non-zero VIDs addresses per MAC
4694 	 * client to avoid figuring it out in the data-path.
4695 	 */
4696 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4697 		mcip->mci_nvids++;
4698 
4699 	rw_exit(&mcip->mci_rw_lock);
4700 }
4701 
4702 /*
4703  * Remove a flow entry from the MAC client's list.
4704  */
4705 static void
4706 mac_client_remove_flow_from_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4707 {
4708 	flow_entry_t	*fe = mcip->mci_flent_list;
4709 	flow_entry_t	*prev_fe = NULL;
4710 
4711 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4712 	/*
4713 	 * The promisc Rx data path walks the mci_flent_list. Protect by
4714 	 * using mci_rw_lock
4715 	 */
4716 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4717 	while ((fe != NULL) && (fe != flent)) {
4718 		prev_fe = fe;
4719 		fe = fe->fe_client_next;
4720 	}
4721 
4722 	ASSERT(fe != NULL);
4723 	if (prev_fe == NULL) {
4724 		/* Deleting the first node */
4725 		mcip->mci_flent_list = fe->fe_client_next;
4726 	} else {
4727 		prev_fe->fe_client_next = fe->fe_client_next;
4728 	}
4729 	mcip->mci_nflents--;
4730 
4731 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4732 		mcip->mci_nvids--;
4733 
4734 	rw_exit(&mcip->mci_rw_lock);
4735 }
4736 
4737 /*
4738  * Check if the given VID belongs to this MAC client.
4739  */
4740 boolean_t
4741 mac_client_check_flow_vid(mac_client_impl_t *mcip, uint16_t vid)
4742 {
4743 	flow_entry_t	*flent;
4744 	uint16_t	mci_vid;
4745 
4746 	/* The mci_flent_list is protected by mci_rw_lock */
4747 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4748 	for (flent = mcip->mci_flent_list; flent != NULL;
4749 	    flent = flent->fe_client_next) {
4750 		mci_vid = i_mac_flow_vid(flent);
4751 		if (vid == mci_vid) {
4752 			rw_exit(&mcip->mci_rw_lock);
4753 			return (B_TRUE);
4754 		}
4755 	}
4756 	rw_exit(&mcip->mci_rw_lock);
4757 	return (B_FALSE);
4758 }
4759 
4760 /*
4761  * Get the flow entry for the specified <MAC addr, VID> tuple.
4762  */
4763 static flow_entry_t *
4764 mac_client_get_flow(mac_client_impl_t *mcip, mac_unicast_impl_t *muip)
4765 {
4766 	mac_address_t *map = mcip->mci_unicast;
4767 	flow_entry_t *flent;
4768 	uint16_t vid;
4769 	flow_desc_t flow_desc;
4770 
4771 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4772 
4773 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
4774 	if (bcmp(flow_desc.fd_dst_mac, map->ma_addr, map->ma_len) != 0)
4775 		return (NULL);
4776 
4777 	for (flent = mcip->mci_flent_list; flent != NULL;
4778 	    flent = flent->fe_client_next) {
4779 		vid = i_mac_flow_vid(flent);
4780 		if (vid == muip->mui_vid) {
4781 			return (flent);
4782 		}
4783 	}
4784 
4785 	return (NULL);
4786 }
4787 
4788 /*
4789  * Since mci_flent has the SRSs, when we want to remove it, we replace
4790  * the flow_desc_t in mci_flent with that of an existing flent and then
4791  * remove that flent instead of mci_flent.
4792  */
4793 static flow_entry_t *
4794 mac_client_swap_mciflent(mac_client_impl_t *mcip)
4795 {
4796 	flow_entry_t	*flent = mcip->mci_flent;
4797 	flow_tab_t	*ft = flent->fe_flow_tab;
4798 	flow_entry_t	*flent1;
4799 	flow_desc_t	fl_desc;
4800 	char		fl_name[MAXFLOWNAMELEN];
4801 	int		err;
4802 
4803 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4804 	ASSERT(mcip->mci_nflents > 1);
4805 
4806 	/* get the next flent following the primary flent  */
4807 	flent1 = mcip->mci_flent_list->fe_client_next;
4808 	ASSERT(flent1 != NULL && flent1->fe_flow_tab == ft);
4809 
4810 	/*
4811 	 * Remove the flent from the flow table before updating the
4812 	 * flow descriptor as the hash depends on the flow descriptor.
4813 	 * This also helps incoming packet classification avoid having
4814 	 * to grab fe_lock. Access to fe_flow_desc of a flent not in the
4815 	 * flow table is done under the fe_lock so that log or stat functions
4816 	 * see a self-consistent fe_flow_desc. The name and desc are specific
4817 	 * to a flow, the rest are shared by all the clients, including
4818 	 * resource control etc.
4819 	 */
4820 	mac_flow_remove(ft, flent, B_TRUE);
4821 	mac_flow_remove(ft, flent1, B_TRUE);
4822 
4823 	bcopy(&flent->fe_flow_desc, &fl_desc, sizeof (flow_desc_t));
4824 	bcopy(flent->fe_flow_name, fl_name, MAXFLOWNAMELEN);
4825 
4826 	/* update the primary flow entry */
4827 	mutex_enter(&flent->fe_lock);
4828 	bcopy(&flent1->fe_flow_desc, &flent->fe_flow_desc,
4829 	    sizeof (flow_desc_t));
4830 	bcopy(&flent1->fe_flow_name, &flent->fe_flow_name, MAXFLOWNAMELEN);
4831 	mutex_exit(&flent->fe_lock);
4832 
4833 	/* update the flow entry that is to be freed */
4834 	mutex_enter(&flent1->fe_lock);
4835 	bcopy(&fl_desc, &flent1->fe_flow_desc, sizeof (flow_desc_t));
4836 	bcopy(fl_name, &flent1->fe_flow_name, MAXFLOWNAMELEN);
4837 	mutex_exit(&flent1->fe_lock);
4838 
4839 	/* now reinsert the flow entries in the table */
4840 	err = mac_flow_add(ft, flent);
4841 	ASSERT(err == 0);
4842 
4843 	err = mac_flow_add(ft, flent1);
4844 	ASSERT(err == 0);
4845 
4846 	return (flent1);
4847 }
4848 
4849 /*
4850  * Return whether there is only one flow entry associated with this
4851  * MAC client.
4852  */
4853 static boolean_t
4854 mac_client_single_rcvr(mac_client_impl_t *mcip)
4855 {
4856 	return (mcip->mci_nflents == 1);
4857 }
4858 
4859 int
4860 mac_validate_props(mac_impl_t *mip, mac_resource_props_t *mrp)
4861 {
4862 	boolean_t		reset;
4863 	uint32_t		rings_needed;
4864 	uint32_t		rings_avail;
4865 	mac_group_type_t	gtype;
4866 	mac_resource_props_t	*mip_mrp;
4867 
4868 	if (mrp == NULL)
4869 		return (0);
4870 
4871 	if (mrp->mrp_mask & MRP_PRIORITY) {
4872 		mac_priority_level_t	pri = mrp->mrp_priority;
4873 
4874 		if (pri < MPL_LOW || pri > MPL_RESET)
4875 			return (EINVAL);
4876 	}
4877 
4878 	if (mrp->mrp_mask & MRP_MAXBW) {
4879 		uint64_t maxbw = mrp->mrp_maxbw;
4880 
4881 		if (maxbw < MRP_MAXBW_MINVAL && maxbw != 0)
4882 			return (EINVAL);
4883 	}
4884 	if (mrp->mrp_mask & MRP_CPUS) {
4885 		int i, j;
4886 		mac_cpu_mode_t	fanout;
4887 
4888 		if (mrp->mrp_ncpus > ncpus)
4889 			return (EINVAL);
4890 
4891 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4892 			for (j = 0; j < mrp->mrp_ncpus; j++) {
4893 				if (i != j &&
4894 				    mrp->mrp_cpu[i] == mrp->mrp_cpu[j]) {
4895 					return (EINVAL);
4896 				}
4897 			}
4898 		}
4899 
4900 		for (i = 0; i < mrp->mrp_ncpus; i++) {
4901 			cpu_t *cp;
4902 			int rv;
4903 
4904 			mutex_enter(&cpu_lock);
4905 			cp = cpu_get(mrp->mrp_cpu[i]);
4906 			if (cp != NULL)
4907 				rv = cpu_is_online(cp);
4908 			else
4909 				rv = 0;
4910 			mutex_exit(&cpu_lock);
4911 			if (rv == 0)
4912 				return (EINVAL);
4913 		}
4914 
4915 		fanout = mrp->mrp_fanout_mode;
4916 		if (fanout < 0 || fanout > MCM_CPUS)
4917 			return (EINVAL);
4918 	}
4919 
4920 	if (mrp->mrp_mask & MRP_PROTECT) {
4921 		int err = mac_protect_validate(mrp);
4922 		if (err != 0)
4923 			return (err);
4924 	}
4925 
4926 	if (!(mrp->mrp_mask & MRP_RX_RINGS) &&
4927 	    !(mrp->mrp_mask & MRP_TX_RINGS)) {
4928 		return (0);
4929 	}
4930 
4931 	/*
4932 	 * mip will be null when we come from mac_flow_create or
4933 	 * mac_link_flow_modify. In the latter case it is a user flow,
4934 	 * for which we don't support rings. In the former we would
4935 	 * have validated the props beforehand (i_mac_unicast_add ->
4936 	 * mac_client_set_resources -> validate for the primary and
4937 	 * vnic_dev_create -> mac_client_set_resources -> validate for
4938 	 * a vnic.
4939 	 */
4940 	if (mip == NULL)
4941 		return (0);
4942 
4943 	/*
4944 	 * We don't support setting rings property for a VNIC that is using a
4945 	 * primary address (VLAN)
4946 	 */
4947 	if ((mip->mi_state_flags & MIS_IS_VNIC) &&
4948 	    mac_is_vnic_primary((mac_handle_t)mip)) {
4949 		return (ENOTSUP);
4950 	}
4951 
4952 	mip_mrp = &mip->mi_resource_props;
4953 	/*
4954 	 * The rings property should be validated against the NICs
4955 	 * resources
4956 	 */
4957 	if (mip->mi_state_flags & MIS_IS_VNIC)
4958 		mip = (mac_impl_t *)mac_get_lower_mac_handle((mac_handle_t)mip);
4959 
4960 	reset = mrp->mrp_mask & MRP_RINGS_RESET;
4961 	/*
4962 	 * If groups are not supported, return error.
4963 	 */
4964 	if (((mrp->mrp_mask & MRP_RX_RINGS) && mip->mi_rx_groups == NULL) ||
4965 	    ((mrp->mrp_mask & MRP_TX_RINGS) && mip->mi_tx_groups == NULL)) {
4966 		return (EINVAL);
4967 	}
4968 	/*
4969 	 * If we are just resetting, there is no validation needed.
4970 	 */
4971 	if (reset)
4972 		return (0);
4973 
4974 	if (mrp->mrp_mask & MRP_RX_RINGS) {
4975 		rings_needed = mrp->mrp_nrxrings;
4976 		/*
4977 		 * We just want to check if the number of additional
4978 		 * rings requested is available.
4979 		 */
4980 		if (mip_mrp->mrp_mask & MRP_RX_RINGS) {
4981 			if (mrp->mrp_nrxrings > mip_mrp->mrp_nrxrings)
4982 				/* Just check for the additional rings */
4983 				rings_needed -= mip_mrp->mrp_nrxrings;
4984 			else
4985 				/* We are not asking for additional rings */
4986 				rings_needed = 0;
4987 		}
4988 		rings_avail = mip->mi_rxrings_avail;
4989 		gtype = mip->mi_rx_group_type;
4990 	} else {
4991 		rings_needed = mrp->mrp_ntxrings;
4992 		/* Similarly for the TX rings */
4993 		if (mip_mrp->mrp_mask & MRP_TX_RINGS) {
4994 			if (mrp->mrp_ntxrings > mip_mrp->mrp_ntxrings)
4995 				/* Just check for the additional rings */
4996 				rings_needed -= mip_mrp->mrp_ntxrings;
4997 			else
4998 				/* We are not asking for additional rings */
4999 				rings_needed = 0;
5000 		}
5001 		rings_avail = mip->mi_txrings_avail;
5002 		gtype = mip->mi_tx_group_type;
5003 	}
5004 
5005 	/* Error if the group is dynamic .. */
5006 	if (gtype == MAC_GROUP_TYPE_DYNAMIC) {
5007 		/*
5008 		 * .. and rings specified are more than available.
5009 		 */
5010 		if (rings_needed > rings_avail)
5011 			return (EINVAL);
5012 	} else {
5013 		/*
5014 		 * OR group is static and we have specified some rings.
5015 		 */
5016 		if (rings_needed > 0)
5017 			return (EINVAL);
5018 	}
5019 	return (0);
5020 }
5021 
5022 /*
5023  * Send a MAC_NOTE_LINK notification to all the MAC clients whenever the
5024  * underlying physical link is down. This is to allow MAC clients to
5025  * communicate with other clients.
5026  */
5027 void
5028 mac_virtual_link_update(mac_impl_t *mip)
5029 {
5030 	if (mip->mi_linkstate != LINK_STATE_UP)
5031 		i_mac_notify(mip, MAC_NOTE_LINK);
5032 }
5033 
5034 /*
5035  * For clients that have a pass-thru MAC, e.g. VNIC, we set the VNIC's
5036  * mac handle in the client.
5037  */
5038 void
5039 mac_set_upper_mac(mac_client_handle_t mch, mac_handle_t mh,
5040     mac_resource_props_t *mrp)
5041 {
5042 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
5043 	mac_impl_t		*mip = (mac_impl_t *)mh;
5044 
5045 	mcip->mci_upper_mip = mip;
5046 	/* If there are any properties, copy it over too */
5047 	if (mrp != NULL) {
5048 		bcopy(mrp, &mip->mi_resource_props,
5049 		    sizeof (mac_resource_props_t));
5050 	}
5051 }
5052 
5053 /*
5054  * Mark the mac as being used exclusively by the single mac client that is
5055  * doing some control operation on this mac. No further opens of this mac
5056  * will be allowed until this client calls mac_unmark_exclusive. The mac
5057  * client calling this function must already be in the mac perimeter
5058  */
5059 int
5060 mac_mark_exclusive(mac_handle_t mh)
5061 {
5062 	mac_impl_t	*mip = (mac_impl_t *)mh;
5063 
5064 	ASSERT(MAC_PERIM_HELD(mh));
5065 	/*
5066 	 * Look up its entry in the global hash table.
5067 	 */
5068 	rw_enter(&i_mac_impl_lock, RW_WRITER);
5069 	if (mip->mi_state_flags & MIS_DISABLED) {
5070 		rw_exit(&i_mac_impl_lock);
5071 		return (ENOENT);
5072 	}
5073 
5074 	/*
5075 	 * A reference to mac is held even if the link is not plumbed.
5076 	 * In i_dls_link_create() we open the MAC interface and hold the
5077 	 * reference. There is an additional reference for the mac_open
5078 	 * done in acquiring the mac perimeter
5079 	 */
5080 	if (mip->mi_ref != 2) {
5081 		rw_exit(&i_mac_impl_lock);
5082 		return (EBUSY);
5083 	}
5084 
5085 	ASSERT(!(mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5086 	mip->mi_state_flags |= MIS_EXCLUSIVE_HELD;
5087 	rw_exit(&i_mac_impl_lock);
5088 	return (0);
5089 }
5090 
5091 void
5092 mac_unmark_exclusive(mac_handle_t mh)
5093 {
5094 	mac_impl_t	*mip = (mac_impl_t *)mh;
5095 
5096 	ASSERT(MAC_PERIM_HELD(mh));
5097 
5098 	rw_enter(&i_mac_impl_lock, RW_WRITER);
5099 	/* 1 for the creation and another for the perimeter */
5100 	ASSERT(mip->mi_ref == 2 && (mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5101 	mip->mi_state_flags &= ~MIS_EXCLUSIVE_HELD;
5102 	rw_exit(&i_mac_impl_lock);
5103 }
5104 
5105 /*
5106  * Set the MTU for the specified MAC.
5107  */
5108 int
5109 mac_set_mtu(mac_handle_t mh, uint_t new_mtu, uint_t *old_mtu_arg)
5110 {
5111 	mac_impl_t *mip = (mac_impl_t *)mh;
5112 	uint_t old_mtu;
5113 	int rv = 0;
5114 
5115 	i_mac_perim_enter(mip);
5116 
5117 	if (!(mip->mi_callbacks->mc_callbacks & (MC_SETPROP|MC_GETPROP))) {
5118 		rv = ENOTSUP;
5119 		goto bail;
5120 	}
5121 
5122 	old_mtu = mip->mi_sdu_max;
5123 
5124 	if (new_mtu == 0 || new_mtu < mip->mi_sdu_min) {
5125 		rv = EINVAL;
5126 		goto bail;
5127 	}
5128 
5129 	if (old_mtu != new_mtu) {
5130 		rv = mip->mi_callbacks->mc_setprop(mip->mi_driver,
5131 		    "mtu", MAC_PROP_MTU, sizeof (uint_t), &new_mtu);
5132 		if (rv != 0)
5133 			goto bail;
5134 		rv = mac_maxsdu_update(mh, new_mtu);
5135 		ASSERT(rv == 0);
5136 	}
5137 
5138 bail:
5139 	i_mac_perim_exit(mip);
5140 
5141 	if (rv == 0 && old_mtu_arg != NULL)
5142 		*old_mtu_arg = old_mtu;
5143 	return (rv);
5144 }
5145 
5146 /*
5147  * Return the RX h/w information for the group indexed by grp_num.
5148  */
5149 void
5150 mac_get_hwrxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5151     uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5152     char *clnts_name)
5153 {
5154 	mac_impl_t *mip = (mac_impl_t *)mh;
5155 	mac_grp_client_t *mcip;
5156 	uint_t i = 0, index = 0;
5157 	mac_ring_t	*ring;
5158 
5159 	/* Revisit when we implement fully dynamic group allocation */
5160 	ASSERT(grp_index >= 0 && grp_index < mip->mi_rx_group_count);
5161 
5162 	rw_enter(&mip->mi_rw_lock, RW_READER);
5163 	*grp_num = mip->mi_rx_groups[grp_index].mrg_index;
5164 	*type = mip->mi_rx_groups[grp_index].mrg_type;
5165 	*n_rings = mip->mi_rx_groups[grp_index].mrg_cur_count;
5166 	ring = mip->mi_rx_groups[grp_index].mrg_rings;
5167 	for (index = 0; index < mip->mi_rx_groups[grp_index].mrg_cur_count;
5168 	    index++) {
5169 		rings[index] = ring->mr_index;
5170 		ring = ring->mr_next;
5171 	}
5172 	/* Assuming the 1st is the default group */
5173 	index = 0;
5174 	if (grp_index == 0) {
5175 		(void) strlcpy(clnts_name, "<default,mcast>,",
5176 		    MAXCLIENTNAMELEN);
5177 		index += strlen("<default,mcast>,");
5178 	}
5179 	for (mcip = mip->mi_rx_groups[grp_index].mrg_clients; mcip != NULL;
5180 	    mcip = mcip->mgc_next) {
5181 		int name_len = strlen(mcip->mgc_client->mci_name);
5182 
5183 		/*
5184 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
5185 		 * names.
5186 		 * XXXX Formating the client name string needs to be moved
5187 		 * to user land when fixing the size of dhi_clnts in
5188 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5189 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
5190 		 */
5191 		if (index + name_len >= MAXCLIENTNAMELEN) {
5192 			index = MAXCLIENTNAMELEN;
5193 			break;
5194 		}
5195 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5196 		    name_len);
5197 		index += name_len;
5198 		clnts_name[index++] = ',';
5199 		i++;
5200 	}
5201 
5202 	/* Get rid of the last , */
5203 	if (index > 0)
5204 		clnts_name[index - 1] = '\0';
5205 	*n_clnts = i;
5206 	rw_exit(&mip->mi_rw_lock);
5207 }
5208 
5209 /*
5210  * Return the TX h/w information for the group indexed by grp_num.
5211  */
5212 void
5213 mac_get_hwtxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5214     uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5215     char *clnts_name)
5216 {
5217 	mac_impl_t *mip = (mac_impl_t *)mh;
5218 	mac_grp_client_t *mcip;
5219 	uint_t i = 0, index = 0;
5220 	mac_ring_t	*ring;
5221 
5222 	/* Revisit when we implement fully dynamic group allocation */
5223 	ASSERT(grp_index >= 0 && grp_index <= mip->mi_tx_group_count);
5224 
5225 	rw_enter(&mip->mi_rw_lock, RW_READER);
5226 	*grp_num = mip->mi_tx_groups[grp_index].mrg_index > 0 ?
5227 	    mip->mi_tx_groups[grp_index].mrg_index : grp_index;
5228 	*type = mip->mi_tx_groups[grp_index].mrg_type;
5229 	*n_rings = mip->mi_tx_groups[grp_index].mrg_cur_count;
5230 	ring = mip->mi_tx_groups[grp_index].mrg_rings;
5231 	for (index = 0; index < mip->mi_tx_groups[grp_index].mrg_cur_count;
5232 	    index++) {
5233 		rings[index] = ring->mr_index;
5234 		ring = ring->mr_next;
5235 	}
5236 	index = 0;
5237 	/* Default group has an index of -1 */
5238 	if (mip->mi_tx_groups[grp_index].mrg_index < 0) {
5239 		(void) strlcpy(clnts_name, "<default>,",
5240 		    MAXCLIENTNAMELEN);
5241 		index += strlen("<default>,");
5242 	}
5243 	for (mcip = mip->mi_tx_groups[grp_index].mrg_clients; mcip != NULL;
5244 	    mcip = mcip->mgc_next) {
5245 		int name_len = strlen(mcip->mgc_client->mci_name);
5246 
5247 		/*
5248 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
5249 		 * names.
5250 		 * XXXX Formating the client name string needs to be moved
5251 		 * to user land when fixing the size of dhi_clnts in
5252 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5253 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
5254 		 */
5255 		if (index + name_len >= MAXCLIENTNAMELEN) {
5256 			index = MAXCLIENTNAMELEN;
5257 			break;
5258 		}
5259 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5260 		    name_len);
5261 		index += name_len;
5262 		clnts_name[index++] = ',';
5263 		i++;
5264 	}
5265 
5266 	/* Get rid of the last , */
5267 	if (index > 0)
5268 		clnts_name[index - 1] = '\0';
5269 	*n_clnts = i;
5270 	rw_exit(&mip->mi_rw_lock);
5271 }
5272 
5273 /*
5274  * Return the group count for RX or TX.
5275  */
5276 uint_t
5277 mac_hwgrp_num(mac_handle_t mh, int type)
5278 {
5279 	mac_impl_t *mip = (mac_impl_t *)mh;
5280 
5281 	/*
5282 	 * Return the Rx and Tx group count; for the Tx we need to
5283 	 * include the default too.
5284 	 */
5285 	return (type == MAC_RING_TYPE_RX ? mip->mi_rx_group_count :
5286 	    mip->mi_tx_groups != NULL ? mip->mi_tx_group_count + 1 : 0);
5287 }
5288 
5289 /*
5290  * The total number of free TX rings for this MAC.
5291  */
5292 uint_t
5293 mac_txavail_get(mac_handle_t mh)
5294 {
5295 	mac_impl_t	*mip = (mac_impl_t *)mh;
5296 
5297 	return (mip->mi_txrings_avail);
5298 }
5299 
5300 /*
5301  * The total number of free RX rings for this MAC.
5302  */
5303 uint_t
5304 mac_rxavail_get(mac_handle_t mh)
5305 {
5306 	mac_impl_t	*mip = (mac_impl_t *)mh;
5307 
5308 	return (mip->mi_rxrings_avail);
5309 }
5310 
5311 /*
5312  * The total number of reserved RX rings on this MAC.
5313  */
5314 uint_t
5315 mac_rxrsvd_get(mac_handle_t mh)
5316 {
5317 	mac_impl_t	*mip = (mac_impl_t *)mh;
5318 
5319 	return (mip->mi_rxrings_rsvd);
5320 }
5321 
5322 /*
5323  * The total number of reserved TX rings on this MAC.
5324  */
5325 uint_t
5326 mac_txrsvd_get(mac_handle_t mh)
5327 {
5328 	mac_impl_t	*mip = (mac_impl_t *)mh;
5329 
5330 	return (mip->mi_txrings_rsvd);
5331 }
5332 
5333 /*
5334  * Total number of free RX groups on this MAC.
5335  */
5336 uint_t
5337 mac_rxhwlnksavail_get(mac_handle_t mh)
5338 {
5339 	mac_impl_t	*mip = (mac_impl_t *)mh;
5340 
5341 	return (mip->mi_rxhwclnt_avail);
5342 }
5343 
5344 /*
5345  * Total number of RX groups reserved on this MAC.
5346  */
5347 uint_t
5348 mac_rxhwlnksrsvd_get(mac_handle_t mh)
5349 {
5350 	mac_impl_t	*mip = (mac_impl_t *)mh;
5351 
5352 	return (mip->mi_rxhwclnt_used);
5353 }
5354 
5355 /*
5356  * Total number of free TX groups on this MAC.
5357  */
5358 uint_t
5359 mac_txhwlnksavail_get(mac_handle_t mh)
5360 {
5361 	mac_impl_t	*mip = (mac_impl_t *)mh;
5362 
5363 	return (mip->mi_txhwclnt_avail);
5364 }
5365 
5366 /*
5367  * Total number of TX groups reserved on this MAC.
5368  */
5369 uint_t
5370 mac_txhwlnksrsvd_get(mac_handle_t mh)
5371 {
5372 	mac_impl_t	*mip = (mac_impl_t *)mh;
5373 
5374 	return (mip->mi_txhwclnt_used);
5375 }
5376 
5377 /*
5378  * Initialize the rings property for a mac client. A non-0 value for
5379  * rxring or txring specifies the number of rings required, a value
5380  * of MAC_RXRINGS_NONE/MAC_TXRINGS_NONE specifies that it doesn't need
5381  * any RX/TX rings and a value of MAC_RXRINGS_DONTCARE/MAC_TXRINGS_DONTCARE
5382  * means the system can decide whether it can give any rings or not.
5383  */
5384 void
5385 mac_client_set_rings(mac_client_handle_t mch, int rxrings, int txrings)
5386 {
5387 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
5388 	mac_resource_props_t	*mrp = MCIP_RESOURCE_PROPS(mcip);
5389 
5390 	if (rxrings != MAC_RXRINGS_DONTCARE) {
5391 		mrp->mrp_mask |= MRP_RX_RINGS;
5392 		mrp->mrp_nrxrings = rxrings;
5393 	}
5394 
5395 	if (txrings != MAC_TXRINGS_DONTCARE) {
5396 		mrp->mrp_mask |= MRP_TX_RINGS;
5397 		mrp->mrp_ntxrings = txrings;
5398 	}
5399 }
5400