xref: /titanic_41/usr/src/uts/common/io/mac/mac_client.c (revision 0f1702c5201310f0529cd5abb77652e5e9b241b6)
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 2008 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 /*
28  * - General Introduction:
29  *
30  * This file contains the implementation of the MAC client kernel
31  * API and related code. The MAC client API allows a kernel module
32  * to gain access to a MAC instance (physical NIC, link aggregation, etc).
33  * It allows a MAC client to associate itself with a MAC address,
34  * VLANs, callback functions for data traffic and for promiscuous mode.
35  * The MAC client API is also used to specify the properties associated
36  * with a MAC client, such as bandwidth limits, priority, CPUS, etc.
37  * These properties are further used to determine the hardware resources
38  * to allocate to the various MAC clients.
39  *
40  * - Primary MAC clients:
41  *
42  * The MAC client API refers to "primary MAC clients". A primary MAC
43  * client is a client which "owns" the primary MAC address of
44  * the underlying MAC instance. The primary MAC address is called out
45  * since it is associated with specific semantics: the primary MAC
46  * address is the MAC address which is assigned to the IP interface
47  * when it is plumbed, and the primary MAC address is assigned
48  * to VLAN data-links. The primary address of a MAC instance can
49  * also change dynamically from under the MAC client, for example
50  * as a result of a change of state of a link aggregation. In that
51  * case the MAC layer automatically updates all data-structures which
52  * refer to the current value of the primary MAC address. Typical
53  * primary MAC clients are dls, aggr, and xnb. A typical non-primary
54  * MAC client is the vnic driver.
55  *
56  * - Virtual Switching:
57  *
58  * The MAC layer implements a virtual switch between the MAC clients
59  * (primary and non-primary) defined on top of the same underlying
60  * NIC (physical, link aggregation, etc). The virtual switch is
61  * VLAN-aware, i.e. it allows multiple MAC clients to be member
62  * of one or more VLANs, and the virtual switch will distribute
63  * multicast tagged packets only to the member of the corresponding
64  * VLANs.
65  *
66  * - Upper vs Lower MAC:
67  *
68  * Creating a VNIC on top of a MAC instance effectively causes
69  * two MAC instances to be layered on top of each other, one for
70  * the VNIC(s), one for the underlying MAC instance (physical NIC,
71  * link aggregation, etc). In the code below we refer to the
72  * underlying NIC as the "lower MAC", and we refer to VNICs as
73  * the "upper MAC".
74  *
75  * - Pass-through for VNICs:
76  *
77  * When VNICs are created on top of an underlying MAC, this causes
78  * a layering of two MAC instances. Since the lower MAC already
79  * does the switching and demultiplexing to its MAC clients, the
80  * upper MAC would simply have to pass packets to the layer below
81  * or above it, which would introduce overhead. In order to avoid
82  * this overhead, the MAC layer implements a pass-through mechanism
83  * for VNICs. When a VNIC opens the lower MAC instance, it saves
84  * the MAC client handle it optains from the MAC layer. When a MAC
85  * client opens a VNIC (upper MAC), the MAC layer detects that
86  * the MAC being opened is a VNIC, and gets the MAC client handle
87  * that the VNIC driver obtained from the lower MAC. This exchange
88  * is doing through a private capability between the MAC layer
89  * and the VNIC driver. The upper MAC then returns that handle
90  * directly to its MAC client. Any operation done by the upper
91  * MAC client is now done on the lower MAC client handle, which
92  * allows the VNIC driver to be completely bypassed for the
93  * performance sensitive data-path.
94  *
95  */
96 
97 #include <sys/types.h>
98 #include <sys/conf.h>
99 #include <sys/id_space.h>
100 #include <sys/esunddi.h>
101 #include <sys/stat.h>
102 #include <sys/mkdev.h>
103 #include <sys/stream.h>
104 #include <sys/strsun.h>
105 #include <sys/strsubr.h>
106 #include <sys/dlpi.h>
107 #include <sys/modhash.h>
108 #include <sys/mac_impl.h>
109 #include <sys/mac_client_impl.h>
110 #include <sys/mac_soft_ring.h>
111 #include <sys/dls.h>
112 #include <sys/dld.h>
113 #include <sys/modctl.h>
114 #include <sys/fs/dv_node.h>
115 #include <sys/thread.h>
116 #include <sys/proc.h>
117 #include <sys/callb.h>
118 #include <sys/cpuvar.h>
119 #include <sys/atomic.h>
120 #include <sys/sdt.h>
121 #include <sys/mac_flow.h>
122 #include <sys/ddi_intr_impl.h>
123 #include <sys/disp.h>
124 #include <sys/sdt.h>
125 #include <sys/vnic.h>
126 #include <sys/vnic_impl.h>
127 #include <sys/vlan.h>
128 #include <inet/ip.h>
129 #include <inet/ip6.h>
130 #include <sys/exacct.h>
131 #include <sys/exacct_impl.h>
132 #include <inet/nd.h>
133 #include <sys/ethernet.h>
134 
135 kmem_cache_t	*mac_client_impl_cache;
136 kmem_cache_t	*mac_promisc_impl_cache;
137 
138 static boolean_t mac_client_single_rcvr(mac_client_impl_t *);
139 static flow_entry_t *mac_client_swap_mciflent(mac_client_impl_t *);
140 static flow_entry_t *mac_client_get_flow(mac_client_impl_t *,
141     mac_unicast_impl_t *);
142 static void mac_client_remove_flow_from_list(mac_client_impl_t *,
143     flow_entry_t *);
144 static void mac_client_add_to_flow_list(mac_client_impl_t *, flow_entry_t *);
145 static void mac_rename_flow_names(mac_client_impl_t *, const char *);
146 static void mac_virtual_link_update(mac_impl_t *);
147 
148 /* ARGSUSED */
149 static int
150 i_mac_client_impl_ctor(void *buf, void *arg, int kmflag)
151 {
152 	int	i;
153 	mac_client_impl_t	*mcip = buf;
154 
155 	bzero(buf, MAC_CLIENT_IMPL_SIZE);
156 	mutex_init(&mcip->mci_tx_cb_lock, NULL, MUTEX_DRIVER, NULL);
157 	mcip->mci_tx_notify_cb_info.mcbi_lockp = &mcip->mci_tx_cb_lock;
158 
159 	ASSERT(mac_tx_percpu_cnt >= 0);
160 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
161 		mutex_init(&mcip->mci_tx_pcpu[i].pcpu_tx_lock, NULL,
162 		    MUTEX_DRIVER, NULL);
163 	}
164 	cv_init(&mcip->mci_tx_cv, NULL, CV_DRIVER, NULL);
165 
166 	return (0);
167 }
168 
169 /* ARGSUSED */
170 static void
171 i_mac_client_impl_dtor(void *buf, void *arg)
172 {
173 	int	i;
174 	mac_client_impl_t *mcip = buf;
175 
176 	ASSERT(mcip->mci_promisc_list == NULL);
177 	ASSERT(mcip->mci_unicast_list == NULL);
178 	ASSERT(mcip->mci_state_flags == 0);
179 	ASSERT(mcip->mci_tx_flag == 0);
180 
181 	mutex_destroy(&mcip->mci_tx_cb_lock);
182 
183 	ASSERT(mac_tx_percpu_cnt >= 0);
184 	for (i = 0; i <= mac_tx_percpu_cnt; i++) {
185 		ASSERT(mcip->mci_tx_pcpu[i].pcpu_tx_refcnt == 0);
186 		mutex_destroy(&mcip->mci_tx_pcpu[i].pcpu_tx_lock);
187 	}
188 	cv_destroy(&mcip->mci_tx_cv);
189 }
190 
191 /* ARGSUSED */
192 static int
193 i_mac_promisc_impl_ctor(void *buf, void *arg, int kmflag)
194 {
195 	mac_promisc_impl_t	*mpip = buf;
196 
197 	bzero(buf, sizeof (mac_promisc_impl_t));
198 	mpip->mpi_mci_link.mcb_objp = buf;
199 	mpip->mpi_mci_link.mcb_objsize = sizeof (mac_promisc_impl_t);
200 	mpip->mpi_mi_link.mcb_objp = buf;
201 	mpip->mpi_mi_link.mcb_objsize = sizeof (mac_promisc_impl_t);
202 	return (0);
203 }
204 
205 /* ARGSUSED */
206 static void
207 i_mac_promisc_impl_dtor(void *buf, void *arg)
208 {
209 	mac_promisc_impl_t	*mpip = buf;
210 
211 	ASSERT(mpip->mpi_mci_link.mcb_objp != NULL);
212 	ASSERT(mpip->mpi_mci_link.mcb_objsize == sizeof (mac_promisc_impl_t));
213 	ASSERT(mpip->mpi_mi_link.mcb_objp == mpip->mpi_mci_link.mcb_objp);
214 	ASSERT(mpip->mpi_mi_link.mcb_objsize == sizeof (mac_promisc_impl_t));
215 
216 	mpip->mpi_mci_link.mcb_objp = NULL;
217 	mpip->mpi_mci_link.mcb_objsize = 0;
218 	mpip->mpi_mi_link.mcb_objp = NULL;
219 	mpip->mpi_mi_link.mcb_objsize = 0;
220 
221 	ASSERT(mpip->mpi_mci_link.mcb_flags == 0);
222 	mpip->mpi_mci_link.mcb_objsize = 0;
223 }
224 
225 void
226 mac_client_init(void)
227 {
228 	ASSERT(mac_tx_percpu_cnt >= 0);
229 
230 	mac_client_impl_cache = kmem_cache_create("mac_client_impl_cache",
231 	    MAC_CLIENT_IMPL_SIZE, 0, i_mac_client_impl_ctor,
232 	    i_mac_client_impl_dtor, NULL, NULL, NULL, 0);
233 	ASSERT(mac_client_impl_cache != NULL);
234 
235 	mac_promisc_impl_cache = kmem_cache_create("mac_promisc_impl_cache",
236 	    sizeof (mac_promisc_impl_t), 0, i_mac_promisc_impl_ctor,
237 	    i_mac_promisc_impl_dtor, NULL, NULL, NULL, 0);
238 	ASSERT(mac_promisc_impl_cache != NULL);
239 }
240 
241 void
242 mac_client_fini(void)
243 {
244 	kmem_cache_destroy(mac_client_impl_cache);
245 	kmem_cache_destroy(mac_promisc_impl_cache);
246 }
247 
248 /*
249  * Return the lower MAC client handle from the VNIC driver for the
250  * specified VNIC MAC instance.
251  */
252 mac_client_impl_t *
253 mac_vnic_lower(mac_impl_t *mip)
254 {
255 	mac_capab_vnic_t cap;
256 	mac_client_impl_t *mcip;
257 
258 	VERIFY(i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_VNIC, &cap));
259 	mcip = cap.mcv_mac_client_handle(cap.mcv_arg);
260 
261 	return (mcip);
262 }
263 
264 /*
265  * Return the MAC client handle of the primary MAC client for the
266  * specified MAC instance, or NULL otherwise.
267  */
268 mac_client_impl_t *
269 mac_primary_client_handle(mac_impl_t *mip)
270 {
271 	mac_client_impl_t *mcip;
272 
273 	if (mip->mi_state_flags & MIS_IS_VNIC)
274 		return (mac_vnic_lower(mip));
275 
276 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
277 
278 	for (mcip = mip->mi_clients_list; mcip != NULL;
279 	    mcip = mcip->mci_client_next) {
280 		if (MCIP_DATAPATH_SETUP(mcip) && mac_is_primary_client(mcip))
281 			return (mcip);
282 	}
283 	return (NULL);
284 }
285 
286 /*
287  * Open a MAC specified by its MAC name.
288  */
289 int
290 mac_open(const char *macname, mac_handle_t *mhp)
291 {
292 	mac_impl_t	*mip;
293 	int		err;
294 
295 	/*
296 	 * Look up its entry in the global hash table.
297 	 */
298 	if ((err = mac_hold(macname, &mip)) != 0)
299 		return (err);
300 
301 	/*
302 	 * Hold the dip associated to the MAC to prevent it from being
303 	 * detached. For a softmac, its underlying dip is held by the
304 	 * mi_open() callback.
305 	 *
306 	 * This is done to be more tolerant with some defective drivers,
307 	 * which incorrectly handle mac_unregister() failure in their
308 	 * xxx_detach() routine. For example, some drivers ignore the
309 	 * failure of mac_unregister() and free all resources that
310 	 * that are needed for data transmition.
311 	 */
312 	e_ddi_hold_devi(mip->mi_dip);
313 
314 	if (!(mip->mi_callbacks->mc_callbacks & MC_OPEN)) {
315 		*mhp = (mac_handle_t)mip;
316 		return (0);
317 	}
318 
319 	/*
320 	 * The mac perimeter is used in both mac_open and mac_close by the
321 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
322 	 */
323 	i_mac_perim_enter(mip);
324 	mip->mi_oref++;
325 	if (mip->mi_oref != 1 || ((err = mip->mi_open(mip->mi_driver)) == 0)) {
326 		*mhp = (mac_handle_t)mip;
327 		i_mac_perim_exit(mip);
328 		return (0);
329 	}
330 	mip->mi_oref--;
331 	ddi_release_devi(mip->mi_dip);
332 	mac_rele(mip);
333 	i_mac_perim_exit(mip);
334 	return (err);
335 }
336 
337 /*
338  * Open a MAC specified by its linkid.
339  */
340 int
341 mac_open_by_linkid(datalink_id_t linkid, mac_handle_t *mhp)
342 {
343 	dls_dl_handle_t	dlh;
344 	int		err;
345 
346 	if ((err = dls_devnet_hold_tmp(linkid, &dlh)) != 0)
347 		return (err);
348 
349 	dls_devnet_prop_task_wait(dlh);
350 
351 	err = mac_open(dls_devnet_mac(dlh), mhp);
352 
353 	dls_devnet_rele_tmp(dlh);
354 	return (err);
355 }
356 
357 /*
358  * Open a MAC specified by its link name.
359  */
360 int
361 mac_open_by_linkname(const char *link, mac_handle_t *mhp)
362 {
363 	datalink_id_t	linkid;
364 	int		err;
365 
366 	if ((err = dls_mgmt_get_linkid(link, &linkid)) != 0)
367 		return (err);
368 	return (mac_open_by_linkid(linkid, mhp));
369 }
370 
371 /*
372  * Close the specified MAC.
373  */
374 void
375 mac_close(mac_handle_t mh)
376 {
377 	mac_impl_t	*mip = (mac_impl_t *)mh;
378 
379 	i_mac_perim_enter(mip);
380 	/*
381 	 * The mac perimeter is used in both mac_open and mac_close by the
382 	 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
383 	 */
384 	if (mip->mi_callbacks->mc_callbacks & MC_OPEN) {
385 		ASSERT(mip->mi_oref != 0);
386 		if (--mip->mi_oref == 0) {
387 			if ((mip->mi_callbacks->mc_callbacks & MC_CLOSE))
388 				mip->mi_close(mip->mi_driver);
389 		}
390 	}
391 	i_mac_perim_exit(mip);
392 	ddi_release_devi(mip->mi_dip);
393 	mac_rele(mip);
394 }
395 
396 /*
397  * Misc utility functions to retrieve various information about a MAC
398  * instance or a MAC client.
399  */
400 
401 const mac_info_t *
402 mac_info(mac_handle_t mh)
403 {
404 	return (&((mac_impl_t *)mh)->mi_info);
405 }
406 
407 dev_info_t *
408 mac_devinfo_get(mac_handle_t mh)
409 {
410 	return (((mac_impl_t *)mh)->mi_dip);
411 }
412 
413 const char *
414 mac_name(mac_handle_t mh)
415 {
416 	return (((mac_impl_t *)mh)->mi_name);
417 }
418 
419 char *
420 mac_client_name(mac_client_handle_t mch)
421 {
422 	return (((mac_client_impl_t *)mch)->mci_name);
423 }
424 
425 minor_t
426 mac_minor(mac_handle_t mh)
427 {
428 	return (((mac_impl_t *)mh)->mi_minor);
429 }
430 
431 /*
432  * Return the VID associated with a MAC client. This function should
433  * be called for clients which are associated with only one VID.
434  */
435 uint16_t
436 mac_client_vid(mac_client_handle_t mch)
437 {
438 	uint16_t		vid = VLAN_ID_NONE;
439 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
440 	flow_desc_t		flow_desc;
441 
442 	if (mcip->mci_nflents == 0)
443 		return (vid);
444 
445 	ASSERT(MCIP_DATAPATH_SETUP(mcip) && mac_client_single_rcvr(mcip));
446 
447 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
448 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
449 		vid = flow_desc.fd_vid;
450 
451 	return (vid);
452 }
453 
454 /*
455  * Return the link speed associated with the specified MAC client.
456  *
457  * The link speed of a MAC client is equal to the smallest value of
458  * 1) the current link speed of the underlying NIC, or
459  * 2) the bandwidth limit set for the MAC client.
460  *
461  * Note that the bandwidth limit can be higher than the speed
462  * of the underlying NIC. This is allowed to avoid spurious
463  * administration action failures or artifically lowering the
464  * bandwidth limit of a link that may  have temporarily lowered
465  * its link speed due to hardware problem or administrator action.
466  */
467 static uint64_t
468 mac_client_ifspeed(mac_client_impl_t *mcip)
469 {
470 	mac_impl_t *mip = mcip->mci_mip;
471 	uint64_t nic_speed;
472 
473 	nic_speed = mac_stat_get((mac_handle_t)mip, MAC_STAT_IFSPEED);
474 
475 	if (nic_speed == 0) {
476 		return (0);
477 	} else {
478 		uint64_t policy_limit = (uint64_t)-1;
479 
480 		if (MCIP_RESOURCE_PROPS_MASK(mcip) & MRP_MAXBW)
481 			policy_limit = MCIP_RESOURCE_PROPS_MAXBW(mcip);
482 
483 		return (MIN(policy_limit, nic_speed));
484 	}
485 }
486 
487 /*
488  * Return the link state of the specified client. If here are more
489  * than one clients of the underying mac_impl_t, the link state
490  * will always be UP regardless of the link state of the underlying
491  * mac_impl_t. This is needed to allow the MAC clients to continue
492  * to communicate with each other even when the physical link of
493  * their mac_impl_t is down.
494  */
495 static uint64_t
496 mac_client_link_state(mac_client_impl_t *mcip)
497 {
498 	mac_impl_t *mip = mcip->mci_mip;
499 	uint16_t vid;
500 	mac_client_impl_t *mci_list;
501 	mac_unicast_impl_t *mui_list, *oth_mui_list;
502 
503 	/*
504 	 * Returns LINK_STATE_UP if there are other MAC clients defined on
505 	 * mac_impl_t which share same VLAN ID as that of mcip. Note that
506 	 * if 'mcip' has more than one VID's then we match ANY one of the
507 	 * VID's with other MAC client's VID's and return LINK_STATE_UP.
508 	 */
509 	rw_enter(&mcip->mci_rw_lock, RW_READER);
510 	for (mui_list = mcip->mci_unicast_list; mui_list != NULL;
511 	    mui_list = mui_list->mui_next) {
512 		vid = mui_list->mui_vid;
513 		for (mci_list = mip->mi_clients_list; mci_list != NULL;
514 		    mci_list = mci_list->mci_client_next) {
515 			if (mci_list == mcip)
516 				continue;
517 			for (oth_mui_list = mci_list->mci_unicast_list;
518 			    oth_mui_list != NULL; oth_mui_list = oth_mui_list->
519 			    mui_next) {
520 				if (vid == oth_mui_list->mui_vid) {
521 					rw_exit(&mcip->mci_rw_lock);
522 					return (LINK_STATE_UP);
523 				}
524 			}
525 		}
526 	}
527 	rw_exit(&mcip->mci_rw_lock);
528 
529 	return (mac_stat_get((mac_handle_t)mip, MAC_STAT_LINK_STATE));
530 }
531 
532 /*
533  * Return the statistics of a MAC client. These statistics are different
534  * then the statistics of the underlying MAC which are returned by
535  * mac_stat_get().
536  */
537 uint64_t
538 mac_client_stat_get(mac_client_handle_t mch, uint_t stat)
539 {
540 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
541 	mac_impl_t *mip = mcip->mci_mip;
542 	uint64_t val;
543 
544 	switch (stat) {
545 	case MAC_STAT_LINK_STATE:
546 		val = mac_client_link_state(mcip);
547 		break;
548 	case MAC_STAT_LINK_UP:
549 		val = (mac_client_link_state(mcip) == LINK_STATE_UP);
550 		break;
551 	case MAC_STAT_PROMISC:
552 		val = mac_stat_get((mac_handle_t)mip, MAC_STAT_PROMISC);
553 		break;
554 	case MAC_STAT_IFSPEED:
555 		val = mac_client_ifspeed(mcip);
556 		break;
557 	case MAC_STAT_MULTIRCV:
558 		val = mcip->mci_stat_multircv;
559 		break;
560 	case MAC_STAT_BRDCSTRCV:
561 		val = mcip->mci_stat_brdcstrcv;
562 		break;
563 	case MAC_STAT_MULTIXMT:
564 		val = mcip->mci_stat_multixmt;
565 		break;
566 	case MAC_STAT_BRDCSTXMT:
567 		val = mcip->mci_stat_brdcstxmt;
568 		break;
569 	case MAC_STAT_OBYTES:
570 		val = mcip->mci_stat_obytes;
571 		break;
572 	case MAC_STAT_OPACKETS:
573 		val = mcip->mci_stat_opackets;
574 		break;
575 	case MAC_STAT_OERRORS:
576 		val = mcip->mci_stat_oerrors;
577 		break;
578 	case MAC_STAT_IPACKETS:
579 		val = mcip->mci_stat_ipackets;
580 		break;
581 	case MAC_STAT_RBYTES:
582 		val = mcip->mci_stat_ibytes;
583 		break;
584 	case MAC_STAT_IERRORS:
585 		val = mcip->mci_stat_ierrors;
586 		break;
587 	default:
588 		val = mac_stat_default(mip, stat);
589 		break;
590 	}
591 
592 	return (val);
593 }
594 
595 /*
596  * Return the statistics of the specified MAC instance.
597  */
598 uint64_t
599 mac_stat_get(mac_handle_t mh, uint_t stat)
600 {
601 	mac_impl_t	*mip = (mac_impl_t *)mh;
602 	uint64_t	val;
603 	int		ret;
604 
605 	/*
606 	 * The range of stat determines where it is maintained.  Stat
607 	 * values from 0 up to (but not including) MAC_STAT_MIN are
608 	 * mainteined by the mac module itself.  Everything else is
609 	 * maintained by the driver.
610 	 *
611 	 * If the mac_impl_t being queried corresponds to a VNIC,
612 	 * the stats need to be queried from the lower MAC client
613 	 * corresponding to the VNIC. (The mac_link_update()
614 	 * invoked by the driver to the lower MAC causes the *lower
615 	 * MAC* to update its mi_linkstate, and send a notification
616 	 * to its MAC clients. Due to the VNIC passthrough,
617 	 * these notifications are sent to the upper MAC clients
618 	 * of the VNIC directly, and the upper mac_impl_t of the VNIC
619 	 * does not have a valid mi_linkstate.
620 	 */
621 	if (stat < MAC_STAT_MIN && !(mip->mi_state_flags & MIS_IS_VNIC)) {
622 		/* these stats are maintained by the mac module itself */
623 		switch (stat) {
624 		case MAC_STAT_LINK_STATE:
625 			return (mip->mi_linkstate);
626 		case MAC_STAT_LINK_UP:
627 			return (mip->mi_linkstate == LINK_STATE_UP);
628 		case MAC_STAT_PROMISC:
629 			return (mip->mi_devpromisc != 0);
630 		default:
631 			ASSERT(B_FALSE);
632 		}
633 	}
634 
635 	/*
636 	 * Call the driver to get the given statistic.
637 	 */
638 	ret = mip->mi_getstat(mip->mi_driver, stat, &val);
639 	if (ret != 0) {
640 		/*
641 		 * The driver doesn't support this statistic.  Get the
642 		 * statistic's default value.
643 		 */
644 		val = mac_stat_default(mip, stat);
645 	}
646 	return (val);
647 }
648 
649 /*
650  * Utility function which returns the VID associated with a flow entry.
651  */
652 uint16_t
653 i_mac_flow_vid(flow_entry_t *flent)
654 {
655 	flow_desc_t	flow_desc;
656 
657 	mac_flow_get_desc(flent, &flow_desc);
658 
659 	if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
660 		return (flow_desc.fd_vid);
661 	return (VLAN_ID_NONE);
662 }
663 
664 /*
665  * Verify the validity of the specified unicast MAC address. Returns B_TRUE
666  * if the address is valid, B_FALSE otherwise (multicast address, or incorrect
667  * length.
668  */
669 boolean_t
670 mac_unicst_verify(mac_handle_t mh, const uint8_t *addr, uint_t len)
671 {
672 	mac_impl_t	*mip = (mac_impl_t *)mh;
673 
674 	/*
675 	 * Verify the address. No lock is needed since mi_type and plugin
676 	 * details don't change after mac_register().
677 	 */
678 	if ((len != mip->mi_type->mt_addr_length) ||
679 	    (mip->mi_type->mt_ops.mtops_unicst_verify(addr,
680 	    mip->mi_pdata)) != 0) {
681 		return (B_FALSE);
682 	} else {
683 		return (B_TRUE);
684 	}
685 }
686 
687 void
688 mac_sdu_get(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu)
689 {
690 	mac_impl_t	*mip = (mac_impl_t *)mh;
691 
692 	if (min_sdu != NULL)
693 		*min_sdu = mip->mi_sdu_min;
694 	if (max_sdu != NULL)
695 		*max_sdu = mip->mi_sdu_max;
696 }
697 
698 /*
699  * Update the MAC unicast address of the specified client's flows. Currently
700  * only one unicast MAC unicast address is allowed per client.
701  */
702 static void
703 mac_unicast_update_client_flow(mac_client_impl_t *mcip)
704 {
705 	mac_impl_t *mip = mcip->mci_mip;
706 	flow_entry_t *flent = mcip->mci_flent;
707 	mac_address_t *map = mcip->mci_unicast;
708 	flow_desc_t flow_desc;
709 
710 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
711 	ASSERT(flent != NULL);
712 
713 	mac_flow_get_desc(flent, &flow_desc);
714 	ASSERT(flow_desc.fd_mask & FLOW_LINK_DST);
715 
716 	bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
717 	mac_flow_set_desc(flent, &flow_desc);
718 
719 	/*
720 	 * A MAC client could have one MAC address but multiple
721 	 * VLANs. In that case update the flow entries corresponding
722 	 * to all VLANs of the MAC client.
723 	 */
724 	for (flent = mcip->mci_flent_list; flent != NULL;
725 	    flent = flent->fe_client_next) {
726 		mac_flow_get_desc(flent, &flow_desc);
727 		if (!(flent->fe_type & FLOW_PRIMARY_MAC ||
728 		    flent->fe_type & FLOW_VNIC_MAC))
729 			continue;
730 
731 		bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
732 		mac_flow_set_desc(flent, &flow_desc);
733 	}
734 }
735 
736 /*
737  * Update all clients that share the same unicast address.
738  */
739 void
740 mac_unicast_update_clients(mac_impl_t *mip, mac_address_t *map)
741 {
742 	mac_client_impl_t *mcip;
743 
744 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
745 
746 	/*
747 	 * Find all clients that share the same unicast MAC address and update
748 	 * them appropriately.
749 	 */
750 	for (mcip = mip->mi_clients_list; mcip != NULL;
751 	    mcip = mcip->mci_client_next) {
752 		/*
753 		 * Ignore clients that don't share this MAC address.
754 		 */
755 		if (map != mcip->mci_unicast)
756 			continue;
757 
758 		/*
759 		 * Update those clients with same old unicast MAC address.
760 		 */
761 		mac_unicast_update_client_flow(mcip);
762 	}
763 }
764 
765 /*
766  * Update the unicast MAC address of the specified VNIC MAC client.
767  *
768  * Check whether the operation is valid. Any of following cases should fail:
769  *
770  * 1. It's a VLAN type of VNIC.
771  * 2. The new value is current "primary" MAC address.
772  * 3. The current MAC address is shared with other clients.
773  * 4. The new MAC address has been used. This case will be valid when
774  *    client migration is fully supported.
775  */
776 int
777 mac_vnic_unicast_set(mac_client_handle_t mch, const uint8_t *addr)
778 {
779 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
780 	mac_impl_t *mip = mcip->mci_mip;
781 	mac_address_t *map = mcip->mci_unicast;
782 	int err;
783 
784 	ASSERT(!(mip->mi_state_flags & MIS_IS_VNIC));
785 	ASSERT(mcip->mci_state_flags & MCIS_IS_VNIC);
786 	ASSERT(mcip->mci_flags != MAC_CLIENT_FLAGS_PRIMARY);
787 
788 	i_mac_perim_enter(mip);
789 
790 	/*
791 	 * If this is a VLAN type of VNIC, it's using "primary" MAC address
792 	 * of the underlying interface. Must fail here. Refer to case 1 above.
793 	 */
794 	if (bcmp(map->ma_addr, mip->mi_addr, map->ma_len) == 0) {
795 		i_mac_perim_exit(mip);
796 		return (ENOTSUP);
797 	}
798 
799 	/*
800 	 * If the new address is the "primary" one, must fail. Refer to
801 	 * case 2 above.
802 	 */
803 	if (bcmp(addr, mip->mi_addr, map->ma_len) == 0) {
804 		i_mac_perim_exit(mip);
805 		return (EACCES);
806 	}
807 
808 	/*
809 	 * If the address is shared by multiple clients, must fail. Refer
810 	 * to case 3 above.
811 	 */
812 	if (mac_check_macaddr_shared(map)) {
813 		i_mac_perim_exit(mip);
814 		return (EBUSY);
815 	}
816 
817 	/*
818 	 * If the new address has been used, must fail for now. Refer to
819 	 * case 4 above.
820 	 */
821 	if (mac_find_macaddr(mip, (uint8_t *)addr) != NULL) {
822 		i_mac_perim_exit(mip);
823 		return (ENOTSUP);
824 	}
825 
826 	/*
827 	 * Update the MAC address.
828 	 */
829 	err = mac_update_macaddr(map, (uint8_t *)addr);
830 
831 	if (err != 0) {
832 		i_mac_perim_exit(mip);
833 		return (err);
834 	}
835 
836 	/*
837 	 * Update all flows of this MAC client.
838 	 */
839 	mac_unicast_update_client_flow(mcip);
840 
841 	i_mac_perim_exit(mip);
842 	return (0);
843 }
844 
845 /*
846  * Program the new primary unicast address of the specified MAC.
847  *
848  * Function mac_update_macaddr() takes care different types of underlying
849  * MAC. If the underlying MAC is VNIC, the VNIC driver must have registerd
850  * mi_unicst() entry point, that indirectly calls mac_vnic_unicast_set()
851  * which will take care of updating the MAC address of the corresponding
852  * MAC client.
853  *
854  * This is the only interface that allow the client to update the "primary"
855  * MAC address of the underlying MAC. The new value must have not been
856  * used by other clients.
857  */
858 int
859 mac_unicast_primary_set(mac_handle_t mh, const uint8_t *addr)
860 {
861 	mac_impl_t *mip = (mac_impl_t *)mh;
862 	mac_address_t *map;
863 	int err;
864 
865 	/* verify the address validity */
866 	if (!mac_unicst_verify(mh, addr, mip->mi_type->mt_addr_length))
867 		return (EINVAL);
868 
869 	i_mac_perim_enter(mip);
870 
871 	/*
872 	 * If the new value is the same as the current primary address value,
873 	 * there's nothing to do.
874 	 */
875 	if (bcmp(addr, mip->mi_addr, mip->mi_type->mt_addr_length) == 0) {
876 		i_mac_perim_exit(mip);
877 		return (0);
878 	}
879 
880 	if (mac_find_macaddr(mip, (uint8_t *)addr) != 0) {
881 		i_mac_perim_exit(mip);
882 		return (EBUSY);
883 	}
884 
885 	map = mac_find_macaddr(mip, mip->mi_addr);
886 	ASSERT(map != NULL);
887 
888 	/*
889 	 * Update the MAC address.
890 	 */
891 	if (mip->mi_state_flags & MIS_IS_AGGR) {
892 		mac_capab_aggr_t aggr_cap;
893 
894 		/*
895 		 * If the mac is an aggregation, other than the unicast
896 		 * addresses programming, aggr must be informed about this
897 		 * primary unicst address change to change its mac address
898 		 * policy to be user-specified.
899 		 */
900 		ASSERT(map->ma_type == MAC_ADDRESS_TYPE_UNICAST_CLASSIFIED);
901 		VERIFY(i_mac_capab_get(mh, MAC_CAPAB_AGGR, &aggr_cap));
902 		err = aggr_cap.mca_unicst(mip->mi_driver, addr);
903 		if (err == 0)
904 			bcopy(addr, map->ma_addr, map->ma_len);
905 	} else {
906 		err = mac_update_macaddr(map, (uint8_t *)addr);
907 	}
908 
909 	if (err != 0) {
910 		i_mac_perim_exit(mip);
911 		return (err);
912 	}
913 
914 	mac_unicast_update_clients(mip, map);
915 
916 	/*
917 	 * Save the new primary MAC address in mac_impl_t.
918 	 */
919 	bcopy(addr, mip->mi_addr, mip->mi_type->mt_addr_length);
920 
921 	i_mac_perim_exit(mip);
922 
923 	if (err == 0)
924 		i_mac_notify(mip, MAC_NOTE_UNICST);
925 
926 	return (err);
927 }
928 
929 /*
930  * Return the current primary MAC address of the specified MAC.
931  */
932 void
933 mac_unicast_primary_get(mac_handle_t mh, uint8_t *addr)
934 {
935 	mac_impl_t *mip = (mac_impl_t *)mh;
936 
937 	rw_enter(&mip->mi_rw_lock, RW_READER);
938 	bcopy(mip->mi_addr, addr, mip->mi_type->mt_addr_length);
939 	rw_exit(&mip->mi_rw_lock);
940 }
941 
942 /*
943  * Return information about the use of the primary MAC address of the
944  * specified MAC instance:
945  *
946  * - if client_name is non-NULL, it must point to a string of at
947  *   least MAXNAMELEN bytes, and will be set to the name of the MAC
948  *   client which uses the primary MAC address.
949  *
950  * - if in_use is non-NULL, used to return whether the primary MAC
951  *   address is currently in use.
952  */
953 void
954 mac_unicast_primary_info(mac_handle_t mh, char *client_name, boolean_t *in_use)
955 {
956 	mac_impl_t *mip = (mac_impl_t *)mh;
957 	mac_client_impl_t *cur_client;
958 
959 	if (in_use != NULL)
960 		*in_use = B_FALSE;
961 	if (client_name != NULL)
962 		bzero(client_name, MAXNAMELEN);
963 
964 	/*
965 	 * The mi_rw_lock is used to protect threads that don't hold the
966 	 * mac perimeter to get a consistent view of the mi_clients_list.
967 	 * Threads that modify the list must hold both the mac perimeter and
968 	 * mi_rw_lock(RW_WRITER)
969 	 */
970 	rw_enter(&mip->mi_rw_lock, RW_READER);
971 	for (cur_client = mip->mi_clients_list; cur_client != NULL;
972 	    cur_client = cur_client->mci_client_next) {
973 		if (mac_is_primary_client(cur_client) ||
974 		    (mip->mi_state_flags & MIS_IS_VNIC)) {
975 			rw_exit(&mip->mi_rw_lock);
976 			if (in_use != NULL)
977 				*in_use = B_TRUE;
978 			if (client_name != NULL) {
979 				bcopy(cur_client->mci_name, client_name,
980 				    MAXNAMELEN);
981 			}
982 			return;
983 		}
984 	}
985 	rw_exit(&mip->mi_rw_lock);
986 }
987 
988 /*
989  * Add the specified MAC client to the list of clients which opened
990  * the specified MAC.
991  */
992 static void
993 mac_client_add(mac_client_impl_t *mcip)
994 {
995 	mac_impl_t *mip = mcip->mci_mip;
996 
997 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
998 
999 	/* add VNIC to the front of the list */
1000 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1001 	mcip->mci_client_next = mip->mi_clients_list;
1002 	mip->mi_clients_list = mcip;
1003 	mip->mi_nclients++;
1004 	rw_exit(&mip->mi_rw_lock);
1005 }
1006 
1007 /*
1008  * Remove the specified MAC client from the list of clients which opened
1009  * the specified MAC.
1010  */
1011 static void
1012 mac_client_remove(mac_client_impl_t *mcip)
1013 {
1014 	mac_impl_t *mip = mcip->mci_mip;
1015 	mac_client_impl_t **prev, *cclient;
1016 
1017 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1018 
1019 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
1020 	prev = &mip->mi_clients_list;
1021 	cclient = *prev;
1022 	while (cclient != NULL && cclient != mcip) {
1023 		prev = &cclient->mci_client_next;
1024 		cclient = *prev;
1025 	}
1026 	ASSERT(cclient != NULL);
1027 	*prev = cclient->mci_client_next;
1028 	mip->mi_nclients--;
1029 	rw_exit(&mip->mi_rw_lock);
1030 }
1031 
1032 static mac_unicast_impl_t *
1033 mac_client_find_vid(mac_client_impl_t *mcip, uint16_t vid)
1034 {
1035 	mac_unicast_impl_t *muip = mcip->mci_unicast_list;
1036 
1037 	while ((muip != NULL) && (muip->mui_vid != vid))
1038 		muip = muip->mui_next;
1039 
1040 	return (muip);
1041 }
1042 
1043 /*
1044  * Return whether the specified (MAC address, VID) tuple is already used by
1045  * one of the MAC clients associated with the specified MAC.
1046  */
1047 static boolean_t
1048 mac_addr_in_use(mac_impl_t *mip, uint8_t *mac_addr, uint16_t vid)
1049 {
1050 	mac_client_impl_t *client;
1051 	mac_address_t *map;
1052 
1053 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1054 
1055 	for (client = mip->mi_clients_list; client != NULL;
1056 	    client = client->mci_client_next) {
1057 
1058 		/*
1059 		 * Ignore clients that don't have unicast address.
1060 		 */
1061 		if (client->mci_unicast_list == NULL)
1062 			continue;
1063 
1064 		map = client->mci_unicast;
1065 
1066 		if ((bcmp(mac_addr, map->ma_addr, map->ma_len) == 0) &&
1067 		    (mac_client_find_vid(client, vid) != NULL)) {
1068 			return (B_TRUE);
1069 		}
1070 	}
1071 
1072 	return (B_FALSE);
1073 }
1074 
1075 /*
1076  * Generate a random MAC address. The MAC address prefix is
1077  * stored in the array pointed to by mac_addr, and its length, in bytes,
1078  * is specified by prefix_len. The least significant bits
1079  * after prefix_len bytes are generated, and stored after the prefix
1080  * in the mac_addr array.
1081  */
1082 int
1083 mac_addr_random(mac_client_handle_t mch, uint_t prefix_len,
1084     uint8_t *mac_addr, mac_diag_t *diag)
1085 {
1086 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1087 	mac_impl_t *mip = mcip->mci_mip;
1088 	size_t addr_len = mip->mi_type->mt_addr_length;
1089 
1090 	if (prefix_len >= addr_len) {
1091 		*diag = MAC_DIAG_MACPREFIXLEN_INVALID;
1092 		return (EINVAL);
1093 	}
1094 
1095 	/* check the prefix value */
1096 	if (prefix_len > 0) {
1097 		bzero(mac_addr + prefix_len, addr_len - prefix_len);
1098 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr,
1099 		    addr_len)) {
1100 			*diag = MAC_DIAG_MACPREFIX_INVALID;
1101 			return (EINVAL);
1102 		}
1103 	}
1104 
1105 	/* generate the MAC address */
1106 	if (prefix_len < addr_len) {
1107 		(void) random_get_pseudo_bytes(mac_addr +
1108 		    prefix_len, addr_len - prefix_len);
1109 	}
1110 
1111 	*diag = 0;
1112 	return (0);
1113 }
1114 
1115 /*
1116  * Set the priority range for this MAC client. This will be used to
1117  * determine the absolute priority for the threads created for this
1118  * MAC client using the specified "low", "medium" and "high" level.
1119  * This will also be used for any subflows on this MAC client.
1120  */
1121 #define	MAC_CLIENT_SET_PRIORITY_RANGE(mcip, pri) {			\
1122 	(mcip)->mci_min_pri = FLOW_MIN_PRIORITY(MINCLSYSPRI,	\
1123 	    MAXCLSYSPRI, (pri));					\
1124 	(mcip)->mci_max_pri = FLOW_MAX_PRIORITY(MINCLSYSPRI,	\
1125 	    MAXCLSYSPRI, (mcip)->mci_min_pri);				\
1126 	}
1127 
1128 /*
1129  * MAC client open entry point. Return a new MAC client handle. Each
1130  * MAC client is associated with a name, specified through the 'name'
1131  * argument.
1132  */
1133 int
1134 mac_client_open(mac_handle_t mh, mac_client_handle_t *mchp, char *name,
1135     uint16_t flags)
1136 {
1137 	mac_impl_t *mip = (mac_impl_t *)mh;
1138 	mac_client_impl_t *mcip;
1139 	int err = 0;
1140 	boolean_t share_desired =
1141 	    ((flags & MAC_OPEN_FLAGS_SHARES_DESIRED) != 0);
1142 	boolean_t no_hwrings = ((flags & MAC_OPEN_FLAGS_NO_HWRINGS) != 0);
1143 	boolean_t req_hwrings = ((flags & MAC_OPEN_FLAGS_REQ_HWRINGS) != 0);
1144 	flow_entry_t	*flent = NULL;
1145 
1146 	*mchp = NULL;
1147 	if (share_desired && no_hwrings) {
1148 		/* can't have shares but no hardware rings */
1149 		return (EINVAL);
1150 	}
1151 
1152 	i_mac_perim_enter(mip);
1153 
1154 	if (mip->mi_state_flags & MIS_IS_VNIC) {
1155 		/*
1156 		 * The underlying MAC is a VNIC. Return the MAC client
1157 		 * handle of the lower MAC which was obtained by
1158 		 * the VNIC driver when it did its mac_client_open().
1159 		 */
1160 
1161 		mcip = mac_vnic_lower(mip);
1162 		/*
1163 		 * If there are multiple MAC clients of the VNIC, they
1164 		 * all share the same underlying MAC client handle.
1165 		 */
1166 		if ((flags & MAC_OPEN_FLAGS_TAG_DISABLE) != 0)
1167 			mcip->mci_state_flags |= MCIS_TAG_DISABLE;
1168 
1169 		if ((flags & MAC_OPEN_FLAGS_STRIP_DISABLE) != 0)
1170 			mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
1171 
1172 		if ((flags & MAC_OPEN_FLAGS_DISABLE_TX_VID_CHECK) != 0)
1173 			mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
1174 
1175 		/*
1176 		 * Note that multiple mac clients share the same mcip in
1177 		 * this case.
1178 		 */
1179 		if (flags & MAC_OPEN_FLAGS_EXCLUSIVE)
1180 			mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1181 
1182 		mip->mi_clients_list = mcip;
1183 		i_mac_perim_exit(mip);
1184 		*mchp = (mac_client_handle_t)mcip;
1185 		return (err);
1186 	}
1187 
1188 	mcip = kmem_cache_alloc(mac_client_impl_cache, KM_SLEEP);
1189 
1190 	mcip->mci_mip = mip;
1191 	mcip->mci_upper_mip = NULL;
1192 	mcip->mci_rx_fn = mac_pkt_drop;
1193 	mcip->mci_rx_arg = NULL;
1194 	mcip->mci_direct_rx_fn = NULL;
1195 	mcip->mci_direct_rx_arg = NULL;
1196 
1197 	if ((flags & MAC_OPEN_FLAGS_IS_VNIC) != 0)
1198 		mcip->mci_state_flags |= MCIS_IS_VNIC;
1199 
1200 	if ((flags & MAC_OPEN_FLAGS_EXCLUSIVE) != 0)
1201 		mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1202 
1203 	if ((flags & MAC_OPEN_FLAGS_IS_AGGR_PORT) != 0)
1204 		mcip->mci_state_flags |= MCIS_IS_AGGR_PORT;
1205 
1206 	if ((flags & MAC_OPEN_FLAGS_TAG_DISABLE) != 0)
1207 		mcip->mci_state_flags |= MCIS_TAG_DISABLE;
1208 
1209 	if ((flags & MAC_OPEN_FLAGS_STRIP_DISABLE) != 0)
1210 		mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
1211 
1212 	if ((flags & MAC_OPEN_FLAGS_DISABLE_TX_VID_CHECK) != 0)
1213 		mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
1214 
1215 	if ((flags & MAC_OPEN_FLAGS_USE_DATALINK_NAME) != 0) {
1216 		datalink_id_t	linkid;
1217 
1218 		ASSERT(name == NULL);
1219 		if ((err = dls_devnet_macname2linkid(mip->mi_name,
1220 		    &linkid)) != 0) {
1221 			goto done;
1222 		}
1223 		if ((err = dls_mgmt_get_linkinfo(linkid, mcip->mci_name, NULL,
1224 		    NULL, NULL)) != 0) {
1225 			/*
1226 			 * Use mac name if dlmgmtd is not available.
1227 			 */
1228 			if (err == EBADF) {
1229 				(void) strlcpy(mcip->mci_name, mip->mi_name,
1230 				    sizeof (mcip->mci_name));
1231 				err = 0;
1232 			} else {
1233 				goto done;
1234 			}
1235 		}
1236 		mcip->mci_state_flags |= MCIS_USE_DATALINK_NAME;
1237 	} else {
1238 		ASSERT(name != NULL);
1239 		if (strlen(name) > MAXNAMELEN) {
1240 			err = EINVAL;
1241 			goto done;
1242 		}
1243 		(void) strlcpy(mcip->mci_name, name, sizeof (mcip->mci_name));
1244 	}
1245 	/* the subflow table will be created dynamically */
1246 	mcip->mci_subflow_tab = NULL;
1247 	mcip->mci_stat_multircv = 0;
1248 	mcip->mci_stat_brdcstrcv = 0;
1249 	mcip->mci_stat_multixmt = 0;
1250 	mcip->mci_stat_brdcstxmt = 0;
1251 
1252 	mcip->mci_stat_obytes = 0;
1253 	mcip->mci_stat_opackets = 0;
1254 	mcip->mci_stat_oerrors = 0;
1255 	mcip->mci_stat_ibytes = 0;
1256 	mcip->mci_stat_ipackets = 0;
1257 	mcip->mci_stat_ierrors = 0;
1258 
1259 	/* Create an initial flow */
1260 
1261 	err = mac_flow_create(NULL, NULL, mcip->mci_name, NULL,
1262 	    mcip->mci_state_flags & MCIS_IS_VNIC ? FLOW_VNIC_MAC :
1263 	    FLOW_PRIMARY_MAC, &flent);
1264 	if (err != 0)
1265 		goto done;
1266 	mcip->mci_flent = flent;
1267 	FLOW_MARK(flent, FE_MC_NO_DATAPATH);
1268 	flent->fe_mcip = mcip;
1269 	/*
1270 	 * Place initial creation reference on the flow. This reference
1271 	 * is released in the corresponding delete action viz.
1272 	 * mac_unicast_remove after waiting for all transient refs to
1273 	 * to go away. The wait happens in mac_flow_wait.
1274 	 */
1275 	FLOW_REFHOLD(flent);
1276 
1277 	/*
1278 	 * Do this ahead of the mac_bcast_add() below so that the mi_nclients
1279 	 * will have the right value for mac_rx_srs_setup().
1280 	 */
1281 	mac_client_add(mcip);
1282 
1283 	mcip->mci_no_hwrings = no_hwrings;
1284 	mcip->mci_req_hwrings = req_hwrings;
1285 	mcip->mci_share = NULL;
1286 	if (share_desired) {
1287 		ASSERT(!no_hwrings);
1288 		i_mac_share_alloc(mcip);
1289 	}
1290 
1291 	DTRACE_PROBE2(mac__client__open__allocated, mac_impl_t *,
1292 	    mcip->mci_mip, mac_client_impl_t *, mcip);
1293 	*mchp = (mac_client_handle_t)mcip;
1294 
1295 	i_mac_perim_exit(mip);
1296 	return (0);
1297 
1298 done:
1299 	i_mac_perim_exit(mip);
1300 	mcip->mci_state_flags = 0;
1301 	mcip->mci_tx_flag = 0;
1302 	kmem_cache_free(mac_client_impl_cache, mcip);
1303 	return (err);
1304 }
1305 
1306 /*
1307  * Close the specified MAC client handle.
1308  */
1309 void
1310 mac_client_close(mac_client_handle_t mch, uint16_t flags)
1311 {
1312 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1313 	mac_impl_t		*mip = mcip->mci_mip;
1314 	flow_entry_t		*flent;
1315 
1316 	i_mac_perim_enter(mip);
1317 
1318 	if (flags & MAC_CLOSE_FLAGS_EXCLUSIVE)
1319 		mcip->mci_state_flags &= ~MCIS_EXCLUSIVE;
1320 
1321 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
1322 	    !(flags & MAC_CLOSE_FLAGS_IS_VNIC)) {
1323 		/*
1324 		 * This is an upper VNIC client initiated operation.
1325 		 * The lower MAC client will be closed by the VNIC driver
1326 		 * when the VNIC is deleted.
1327 		 */
1328 
1329 		/*
1330 		 * Clear the flags set when the upper client initiated
1331 		 * open.
1332 		 */
1333 		mcip->mci_state_flags &= ~(MCIS_TAG_DISABLE |
1334 		    MCIS_STRIP_DISABLE | MCIS_DISABLE_TX_VID_CHECK);
1335 
1336 		i_mac_perim_exit(mip);
1337 		return;
1338 	}
1339 
1340 	/*
1341 	 * Remove the flent associated with the MAC client
1342 	 */
1343 	flent = mcip->mci_flent;
1344 	mcip->mci_flent = NULL;
1345 	FLOW_FINAL_REFRELE(flent);
1346 
1347 	/*
1348 	 * MAC clients must remove the unicast addresses and promisc callbacks
1349 	 * they added before issuing a mac_client_close().
1350 	 */
1351 	ASSERT(mcip->mci_unicast_list == NULL);
1352 	ASSERT(mcip->mci_promisc_list == NULL);
1353 	ASSERT(mcip->mci_tx_notify_cb_list == NULL);
1354 
1355 	i_mac_share_free(mcip);
1356 
1357 	mac_client_remove(mcip);
1358 
1359 	i_mac_perim_exit(mip);
1360 	mcip->mci_subflow_tab = NULL;
1361 	mcip->mci_state_flags = 0;
1362 	mcip->mci_tx_flag = 0;
1363 	kmem_cache_free(mac_client_impl_cache, mch);
1364 }
1365 
1366 /*
1367  * Enable bypass for the specified MAC client.
1368  */
1369 boolean_t
1370 mac_rx_bypass_set(mac_client_handle_t mch, mac_direct_rx_t rx_fn, void *arg1)
1371 {
1372 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1373 	mac_impl_t		*mip = mcip->mci_mip;
1374 
1375 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1376 
1377 	/*
1378 	 * If the mac_client is a VLAN or native media is non ethernet, we
1379 	 * should not do DLS bypass and instead let the packets go via the
1380 	 * default mac_rx_deliver route so vlan header can be stripped etc.
1381 	 */
1382 	if (mcip->mci_nvids > 0 ||
1383 	    mip->mi_info.mi_nativemedia != DL_ETHER)
1384 		return (B_FALSE);
1385 
1386 	/*
1387 	 * These are not accessed directly in the data path, and hence
1388 	 * don't need any protection
1389 	 */
1390 	mcip->mci_direct_rx_fn = rx_fn;
1391 	mcip->mci_direct_rx_arg = arg1;
1392 	mcip->mci_state_flags |= MCIS_CLIENT_POLL_CAPABLE;
1393 	return (B_TRUE);
1394 }
1395 
1396 /*
1397  * Set the receive callback for the specified MAC client. There can be
1398  * at most one such callback per MAC client.
1399  */
1400 void
1401 mac_rx_set(mac_client_handle_t mch, mac_rx_t rx_fn, void *arg)
1402 {
1403 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1404 	mac_impl_t	*mip = mcip->mci_mip;
1405 
1406 	/*
1407 	 * Instead of adding an extra set of locks and refcnts in
1408 	 * the datapath at the mac client boundary, we temporarily quiesce
1409 	 * the SRS and related entities. We then change the receive function
1410 	 * without interference from any receive data thread and then reenable
1411 	 * the data flow subsequently.
1412 	 */
1413 	i_mac_perim_enter(mip);
1414 	mac_rx_client_quiesce(mch);
1415 
1416 	mcip->mci_rx_fn = rx_fn;
1417 	mcip->mci_rx_arg = arg;
1418 	mac_rx_client_restart(mch);
1419 	i_mac_perim_exit(mip);
1420 }
1421 
1422 /*
1423  * Reset the receive callback for the specified MAC client.
1424  */
1425 void
1426 mac_rx_clear(mac_client_handle_t mch)
1427 {
1428 	mac_rx_set(mch, mac_pkt_drop, NULL);
1429 }
1430 
1431 /*
1432  * Walk the MAC client subflow table and updates their priority values.
1433  */
1434 static int
1435 mac_update_subflow_priority_cb(flow_entry_t *flent, void *arg)
1436 {
1437 	mac_flow_update_priority(arg, flent);
1438 	return (0);
1439 }
1440 
1441 void
1442 mac_update_subflow_priority(mac_client_impl_t *mcip)
1443 {
1444 	(void) mac_flow_walk(mcip->mci_subflow_tab,
1445 	    mac_update_subflow_priority_cb, mcip);
1446 }
1447 
1448 /*
1449  * When the MAC client is being brought up (i.e. we do a unicast_add) we need
1450  * to initialize the cpu and resource control structure in the
1451  * mac_client_impl_t from the mac_impl_t (i.e if there are any cached
1452  * properties before the flow entry for the unicast address was created).
1453  */
1454 int
1455 mac_resource_ctl_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
1456 {
1457 	mac_client_impl_t 	*mcip = (mac_client_impl_t *)mch;
1458 	mac_impl_t		*mip = (mac_impl_t *)mcip->mci_mip;
1459 	int			err = 0;
1460 
1461 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1462 
1463 	err = mac_validate_props(mrp);
1464 	if (err != 0)
1465 		return (err);
1466 
1467 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
1468 	if (MCIP_DATAPATH_SETUP(mcip)) {
1469 		/*
1470 		 * We have to set this prior to calling mac_flow_modify.
1471 		 */
1472 		if (mrp->mrp_mask & MRP_PRIORITY) {
1473 			if (mrp->mrp_priority == MPL_RESET) {
1474 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1475 				    MPL_LINK_DEFAULT);
1476 			} else {
1477 				MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1478 				    mrp->mrp_priority);
1479 			}
1480 		}
1481 
1482 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
1483 		if (mrp->mrp_mask & MRP_PRIORITY)
1484 			mac_update_subflow_priority(mcip);
1485 		return (0);
1486 	}
1487 	return (0);
1488 }
1489 
1490 void
1491 mac_resource_ctl_get(mac_client_handle_t mch, mac_resource_props_t *mrp)
1492 {
1493 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
1494 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
1495 
1496 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
1497 }
1498 
1499 static int
1500 mac_unicast_flow_create(mac_client_impl_t *mcip, uint8_t *mac_addr,
1501     uint16_t vid, boolean_t is_primary, boolean_t first_flow,
1502     flow_entry_t **flent, mac_resource_props_t *mrp)
1503 {
1504 	mac_impl_t	*mip = (mac_impl_t *)mcip->mci_mip;
1505 	flow_desc_t	flow_desc;
1506 	char		flowname[MAXFLOWNAME];
1507 	int		err;
1508 	uint_t		flent_flags;
1509 
1510 	/*
1511 	 * First unicast address being added, create a new flow
1512 	 * for that MAC client.
1513 	 */
1514 	bzero(&flow_desc, sizeof (flow_desc));
1515 
1516 	flow_desc.fd_mac_len = mip->mi_type->mt_addr_length;
1517 	bcopy(mac_addr, flow_desc.fd_dst_mac, flow_desc.fd_mac_len);
1518 	flow_desc.fd_mask = FLOW_LINK_DST;
1519 	if (vid != 0) {
1520 		flow_desc.fd_vid = vid;
1521 		flow_desc.fd_mask |= FLOW_LINK_VID;
1522 	}
1523 
1524 	/*
1525 	 * XXX-nicolas. For now I'm keeping the FLOW_PRIMARY_MAC
1526 	 * and FLOW_VNIC. Even though they're a hack inherited
1527 	 * from the SRS code, we'll keep them for now. They're currently
1528 	 * consumed by mac_datapath_setup() to create the SRS.
1529 	 * That code should be eventually moved out of
1530 	 * mac_datapath_setup() and moved to a mac_srs_create()
1531 	 * function of some sort to keep things clean.
1532 	 *
1533 	 * Also, there's no reason why the SRS for the primary MAC
1534 	 * client should be different than any other MAC client. Until
1535 	 * this is cleaned-up, we support only one MAC unicast address
1536 	 * per client.
1537 	 *
1538 	 * We set FLOW_PRIMARY_MAC for the primary MAC address,
1539 	 * FLOW_VNIC for everything else.
1540 	 */
1541 	if (is_primary)
1542 		flent_flags = FLOW_PRIMARY_MAC;
1543 	else
1544 		flent_flags = FLOW_VNIC_MAC;
1545 
1546 	/*
1547 	 * For the first flow we use the mac client's name - mci_name, for
1548 	 * subsequent ones we just create a name with the vid. This is
1549 	 * so that we can add these flows to the same flow table. This is
1550 	 * fine as the flow name (except for the one with the mac client's
1551 	 * name) is not visible. When the first flow is removed, we just replace
1552 	 * its fdesc with another from the list, so we will still retain the
1553 	 * flent with the MAC client's flow name.
1554 	 */
1555 	if (first_flow) {
1556 		bcopy(mcip->mci_name, flowname, MAXFLOWNAME);
1557 	} else {
1558 		(void) sprintf(flowname, "%s%u", mcip->mci_name, vid);
1559 		flent_flags = FLOW_NO_STATS;
1560 	}
1561 
1562 	if ((err = mac_flow_create(&flow_desc, mrp, flowname, NULL,
1563 	    flent_flags, flent)) != 0)
1564 		return (err);
1565 
1566 	FLOW_MARK(*flent, FE_INCIPIENT);
1567 	(*flent)->fe_mcip = mcip;
1568 
1569 	/*
1570 	 * Place initial creation reference on the flow. This reference
1571 	 * is released in the corresponding delete action viz.
1572 	 * mac_unicast_remove after waiting for all transient refs to
1573 	 * to go away. The wait happens in mac_flow_wait.
1574 	 * We have already held the reference in mac_client_open().
1575 	 */
1576 	if (!first_flow)
1577 		FLOW_REFHOLD(*flent);
1578 	return (0);
1579 }
1580 
1581 /* Refresh the multicast grouping for this VID. */
1582 int
1583 mac_client_update_mcast(void *arg, boolean_t add, const uint8_t *addrp)
1584 {
1585 	flow_entry_t		*flent = arg;
1586 	mac_client_impl_t	*mcip = flent->fe_mcip;
1587 	uint16_t		vid;
1588 	flow_desc_t		flow_desc;
1589 
1590 	mac_flow_get_desc(flent, &flow_desc);
1591 	vid = (flow_desc.fd_mask & FLOW_LINK_VID) != 0 ?
1592 	    flow_desc.fd_vid : VLAN_ID_NONE;
1593 
1594 	/*
1595 	 * We don't call mac_multicast_add()/mac_multicast_remove() as
1596 	 * we want to add/remove for this specific vid.
1597 	 */
1598 	if (add) {
1599 		return (mac_bcast_add(mcip, addrp, vid,
1600 		    MAC_ADDRTYPE_MULTICAST));
1601 	} else {
1602 		mac_bcast_delete(mcip, addrp, vid);
1603 		return (0);
1604 	}
1605 }
1606 
1607 /*
1608  * Add a new unicast address to the MAC client.
1609  *
1610  * The MAC address can be specified either by value, or the MAC client
1611  * can specify that it wants to use the primary MAC address of the
1612  * underlying MAC. See the introductory comments at the beginning
1613  * of this file for more more information on primary MAC addresses.
1614  *
1615  * Note also the tuple (MAC address, VID) must be unique
1616  * for the MAC clients defined on top of the same underlying MAC
1617  * instance, unless the MAC_UNICAST_NODUPCHECK is specified.
1618  */
1619 
1620 int
1621 i_mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
1622     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
1623 {
1624 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1625 	mac_impl_t *mip = mcip->mci_mip;
1626 	mac_unicast_impl_t *muip;
1627 	flow_entry_t *flent;
1628 	int err;
1629 	uint_t mac_len = mip->mi_type->mt_addr_length;
1630 	boolean_t check_dups = !(flags & MAC_UNICAST_NODUPCHECK);
1631 	boolean_t is_primary = (flags & MAC_UNICAST_PRIMARY);
1632 	boolean_t is_vnic_primary = flags & MAC_UNICAST_VNIC_PRIMARY;
1633 	boolean_t bcast_added = B_FALSE;
1634 	boolean_t nactiveclients_added = B_FALSE;
1635 	boolean_t mac_started = B_FALSE;
1636 	mac_resource_props_t mrp;
1637 
1638 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1639 
1640 	/* when VID is non-zero, the underlying MAC can not be VNIC */
1641 	ASSERT(!((mip->mi_state_flags & MIS_IS_VNIC) && (vid != 0)));
1642 
1643 	/*
1644 	 * Check whether it's the primary client and flag it.
1645 	 */
1646 	if (!(mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary && vid == 0)
1647 		mcip->mci_flags |= MAC_CLIENT_FLAGS_PRIMARY;
1648 
1649 	/*
1650 	 * is_vnic_primary is true when we come here as a VLAN VNIC
1651 	 * which uses the primary mac client's address but with a non-zero
1652 	 * VID. In this case the MAC address is not specified by an upper
1653 	 * MAC client.
1654 	 */
1655 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary &&
1656 	    !is_vnic_primary) {
1657 		/*
1658 		 * The address is being set by the upper MAC client
1659 		 * of a VNIC. The MAC address was already set by the
1660 		 * VNIC driver during VNIC creation.
1661 		 *
1662 		 * Note: a VNIC has only one MAC address. We return
1663 		 * the MAC unicast address handle of the lower MAC client
1664 		 * corresponding to the VNIC. We allocate a new entry
1665 		 * which is flagged appropriately, so that mac_unicast_remove()
1666 		 * doesn't attempt to free the original entry that
1667 		 * was allocated by the VNIC driver.
1668 		 */
1669 		ASSERT(mcip->mci_unicast != NULL);
1670 
1671 		/*
1672 		 * Ensure that the primary unicast address of the VNIC
1673 		 * is added only once.
1674 		 */
1675 		if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY)
1676 			return (EBUSY);
1677 
1678 		mcip->mci_flags |= MAC_CLIENT_FLAGS_VNIC_PRIMARY;
1679 
1680 		/*
1681 		 * Create a handle for vid 0.
1682 		 */
1683 		ASSERT(vid == 0);
1684 		muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
1685 		muip->mui_vid = vid;
1686 		*mah = (mac_unicast_handle_t)muip;
1687 		return (0);
1688 	}
1689 
1690 	/* primary MAC clients cannot be opened on top of anchor VNICs */
1691 	if ((is_vnic_primary || is_primary) &&
1692 	    i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_ANCHOR_VNIC, NULL)) {
1693 		return (ENXIO);
1694 	}
1695 
1696 	/*
1697 	 * Return EBUSY if:
1698 	 *  - this is an exclusive active mac client and there already exist
1699 	 *    active mac clients, or
1700 	 *  - there already exist an exclusively active mac client.
1701 	 */
1702 	if ((mcip->mci_state_flags & MCIS_EXCLUSIVE) &&
1703 	    (mip->mi_nactiveclients != 0) || (mip->mi_state_flags &
1704 	    MIS_EXCLUSIVE)) {
1705 		return (EBUSY);
1706 	}
1707 
1708 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE)
1709 		mip->mi_state_flags |= MIS_EXCLUSIVE;
1710 
1711 	bzero(&mrp, sizeof (mac_resource_props_t));
1712 	if (is_primary && !(mcip->mci_state_flags & MCIS_IS_VNIC)) {
1713 		/*
1714 		 * Apply the property cached in the mac_impl_t to the primary
1715 		 * mac client. If the mac client is a VNIC, its property were
1716 		 * already set in the mcip when the VNIC was created.
1717 		 */
1718 		mac_get_resources((mac_handle_t)mip, &mrp);
1719 		(void) mac_client_set_resources(mch, &mrp);
1720 	} else if (mcip->mci_state_flags & MCIS_IS_VNIC) {
1721 		bcopy(MCIP_RESOURCE_PROPS(mcip), &mrp,
1722 		    sizeof (mac_resource_props_t));
1723 	}
1724 
1725 	muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
1726 	muip->mui_vid = vid;
1727 
1728 	if (is_primary || is_vnic_primary) {
1729 		mac_addr = mip->mi_addr;
1730 		check_dups = B_TRUE;
1731 	} else {
1732 
1733 		/*
1734 		 * Verify the validity of the specified MAC addresses value.
1735 		 */
1736 		if (!mac_unicst_verify((mac_handle_t)mip, mac_addr, mac_len)) {
1737 			*diag = MAC_DIAG_MACADDR_INVALID;
1738 			err = EINVAL;
1739 			goto bail;
1740 		}
1741 
1742 		/*
1743 		 * Make sure that the specified MAC address is different
1744 		 * than the unicast MAC address of the underlying NIC.
1745 		 */
1746 		if (check_dups && bcmp(mip->mi_addr, mac_addr, mac_len) == 0) {
1747 			*diag = MAC_DIAG_MACADDR_NIC;
1748 			err = EINVAL;
1749 			goto bail;
1750 		}
1751 	}
1752 
1753 	/*
1754 	 * Make sure the MAC address is not already used by
1755 	 * another MAC client defined on top of the same
1756 	 * underlying NIC.
1757 	 * xxx-venu mac_unicast_add doesnt' seem to be called
1758 	 * with MAC_UNICAST_NODUPCHECK currently, if it does
1759 	 * get called we need to do mac_addr_in_use() just
1760 	 * to check for addr_in_use till 6697876 is fixed.
1761 	 */
1762 	if (check_dups && mac_addr_in_use(mip, mac_addr, vid)) {
1763 		*diag = MAC_DIAG_MACADDR_INUSE;
1764 		err = EEXIST;
1765 		goto bail;
1766 	}
1767 
1768 	if ((err = mac_start(mip)) != 0)
1769 		goto bail;
1770 
1771 	mac_started = B_TRUE;
1772 
1773 	/* add the MAC client to the broadcast address group by default */
1774 	if (mip->mi_type->mt_brdcst_addr != NULL) {
1775 		err = mac_bcast_add(mcip, mip->mi_type->mt_brdcst_addr, vid,
1776 		    MAC_ADDRTYPE_BROADCAST);
1777 		if (err != 0)
1778 			goto bail;
1779 		bcast_added = B_TRUE;
1780 	}
1781 	flent = mcip->mci_flent;
1782 	ASSERT(flent != NULL);
1783 	/* We are configuring the unicast flow now */
1784 	if (!MCIP_DATAPATH_SETUP(mcip)) {
1785 
1786 		MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1787 		    (mrp.mrp_mask & MRP_PRIORITY) ? mrp.mrp_priority :
1788 		    MPL_LINK_DEFAULT);
1789 
1790 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
1791 		    is_primary || is_vnic_primary, B_TRUE, &flent, &mrp)) != 0)
1792 			goto bail;
1793 
1794 		mip->mi_nactiveclients++;
1795 		nactiveclients_added = B_TRUE;
1796 		/*
1797 		 * This will allocate the RX ring group if possible for the
1798 		 * flow and program the software classifier as needed.
1799 		 */
1800 		if ((err = mac_datapath_setup(mcip, flent, SRST_LINK)) != 0)
1801 			goto bail;
1802 
1803 		/*
1804 		 * The unicast MAC address must have been added successfully.
1805 		 */
1806 		ASSERT(mcip->mci_unicast != NULL);
1807 	} else {
1808 		mac_address_t *map = mcip->mci_unicast;
1809 
1810 		/*
1811 		 * A unicast flow already exists for that MAC client,
1812 		 * this flow must be the same mac address but with
1813 		 * different VID. It has been checked by mac_addr_in_use().
1814 		 *
1815 		 * We will use the SRS etc. from the mci_flent. Note that
1816 		 * We don't need to create kstat for this as except for
1817 		 * the fdesc, everything will be used from in the 1st flent.
1818 		 */
1819 
1820 		if (bcmp(mac_addr, map->ma_addr, map->ma_len) != 0) {
1821 			err = EINVAL;
1822 			goto bail;
1823 		}
1824 
1825 		if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
1826 		    is_primary || is_vnic_primary, B_FALSE, &flent, NULL)) != 0)
1827 			goto bail;
1828 
1829 		if ((err = mac_flow_add(mip->mi_flow_tab, flent)) != 0) {
1830 			FLOW_FINAL_REFRELE(flent);
1831 			goto bail;
1832 		}
1833 
1834 		/* update the multicast group for this vid */
1835 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
1836 		    (void *)flent, B_TRUE);
1837 
1838 	}
1839 
1840 	/* populate the shared MAC address */
1841 	muip->mui_map = mcip->mci_unicast;
1842 
1843 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
1844 	muip->mui_next = mcip->mci_unicast_list;
1845 	mcip->mci_unicast_list = muip;
1846 	rw_exit(&mcip->mci_rw_lock);
1847 
1848 	*mah = (mac_unicast_handle_t)muip;
1849 
1850 	/* add it to the flow list of this mcip */
1851 	mac_client_add_to_flow_list(mcip, flent);
1852 
1853 	/*
1854 	 * Trigger a renegotiation of the capabilities when the number of
1855 	 * active clients changes from 1 to 2, since some of the capabilities
1856 	 * might have to be disabled. Also send a MAC_NOTE_LINK notification
1857 	 * to all the MAC clients whenever physical link is DOWN.
1858 	 */
1859 	if (mip->mi_nactiveclients == 2) {
1860 		mac_capab_update((mac_handle_t)mip);
1861 		mac_virtual_link_update(mip);
1862 	}
1863 	/*
1864 	 * Now that the setup is complete, clear the INCIPIENT flag.
1865 	 * The flag was set to avoid incoming packets seeing inconsistent
1866 	 * structures while the setup was in progress. Clear the mci_tx_flag
1867 	 * by calling mac_tx_client_block. It is possible that
1868 	 * mac_unicast_remove was called prior to this mac_unicast_add which
1869 	 * could have set the MCI_TX_QUIESCE flag.
1870 	 */
1871 	if (flent->fe_rx_ring_group != NULL)
1872 		mac_rx_group_unmark(flent->fe_rx_ring_group, MR_INCIPIENT);
1873 	FLOW_UNMARK(flent, FE_INCIPIENT);
1874 	FLOW_UNMARK(flent, FE_MC_NO_DATAPATH);
1875 	mac_tx_client_unblock(mcip);
1876 	return (0);
1877 bail:
1878 	if (bcast_added)
1879 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr, vid);
1880 	if (mac_started)
1881 		mac_stop(mip);
1882 
1883 	if (nactiveclients_added)
1884 		mip->mi_nactiveclients--;
1885 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE)
1886 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
1887 	kmem_free(muip, sizeof (mac_unicast_impl_t));
1888 	return (err);
1889 }
1890 
1891 int
1892 mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
1893     mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
1894 {
1895 	mac_impl_t *mip = ((mac_client_impl_t *)mch)->mci_mip;
1896 	uint_t err;
1897 
1898 	i_mac_perim_enter(mip);
1899 	err = i_mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
1900 	i_mac_perim_exit(mip);
1901 
1902 	return (err);
1903 }
1904 
1905 /*
1906  * Add the primary MAC address to the MAC client. This is a convenience
1907  * function which can be called by primary MAC clients which do not
1908  * need to specify any other additional flags.
1909  *
1910  * It's called in one of following situations:
1911  *   * dls as the primary MAC client
1912  *   * aggr as an exclusive client
1913  *   * by VNIC's client
1914  */
1915 int
1916 mac_unicast_primary_add(mac_client_handle_t mch, mac_unicast_handle_t *mah,
1917     mac_diag_t *diag)
1918 {
1919 	return (mac_unicast_add(mch, NULL, MAC_UNICAST_PRIMARY, mah, 0, diag));
1920 }
1921 
1922 /*
1923  * Remove a MAC address which was previously added by mac_unicast_add().
1924  */
1925 int
1926 mac_unicast_remove(mac_client_handle_t mch, mac_unicast_handle_t mah)
1927 {
1928 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1929 	mac_unicast_impl_t *muip = (mac_unicast_impl_t *)mah;
1930 	mac_unicast_impl_t *pre;
1931 	mac_impl_t *mip = mcip->mci_mip;
1932 	flow_entry_t *flent;
1933 
1934 	i_mac_perim_enter(mip);
1935 	if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) {
1936 		/*
1937 		 * Called made by the upper MAC client of a VNIC.
1938 		 * There's nothing much to do, the unicast address will
1939 		 * be removed by the VNIC driver when the VNIC is deleted,
1940 		 * but let's ensure that all our transmit is done before
1941 		 * the client does a mac_client_stop lest it trigger an
1942 		 * assert in the driver.
1943 		 */
1944 		ASSERT(muip->mui_vid == 0);
1945 
1946 		mac_tx_client_flush(mcip);
1947 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_VNIC_PRIMARY;
1948 
1949 		kmem_free(muip, sizeof (mac_unicast_impl_t));
1950 		i_mac_perim_exit(mip);
1951 		return (0);
1952 	}
1953 
1954 	ASSERT(muip != NULL);
1955 
1956 	/*
1957 	 * Remove the VID from the list of client's VIDs.
1958 	 */
1959 	pre = mcip->mci_unicast_list;
1960 	if (muip == pre)
1961 		mcip->mci_unicast_list = muip->mui_next;
1962 	else {
1963 		while ((pre->mui_next != NULL) && (pre->mui_next != muip))
1964 			pre = pre->mui_next;
1965 		ASSERT(pre->mui_next == muip);
1966 		rw_enter(&mcip->mci_rw_lock, RW_WRITER);
1967 		pre->mui_next = muip->mui_next;
1968 		rw_exit(&mcip->mci_rw_lock);
1969 	}
1970 
1971 	if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) && muip->mui_vid == 0)
1972 		mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PRIMARY;
1973 
1974 	/*
1975 	 * This MAC client is shared, so we will just remove the flent
1976 	 * corresponding to the address being removed. We don't invoke
1977 	 * mac_rx_classify_flow_rem() since the additional flow is
1978 	 * not associated with its own separate set of SRS and rings,
1979 	 * and these constructs are still needed for the remaining flows.
1980 	 */
1981 	if (!mac_client_single_rcvr(mcip)) {
1982 		flent = mac_client_get_flow(mcip, muip);
1983 		ASSERT(flent != NULL);
1984 
1985 		/*
1986 		 * The first one is disappearing, need to make sure
1987 		 * we replace it with another from the list of
1988 		 * shared clients.
1989 		 */
1990 		if (flent == mcip->mci_flent)
1991 			flent = mac_client_swap_mciflent(mcip);
1992 		mac_client_remove_flow_from_list(mcip, flent);
1993 		mac_flow_remove(mip->mi_flow_tab, flent, B_FALSE);
1994 		mac_flow_wait(flent, FLOW_DRIVER_UPCALL);
1995 
1996 		/*
1997 		 * The multicast groups that were added by the client so
1998 		 * far must be removed from the brodcast domain corresponding
1999 		 * to the VID being removed.
2000 		 */
2001 		mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2002 		    (void *)flent, B_FALSE);
2003 
2004 		if (mip->mi_type->mt_brdcst_addr != NULL) {
2005 			mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2006 			    muip->mui_vid);
2007 		}
2008 		mac_stop(mip);
2009 		FLOW_FINAL_REFRELE(flent);
2010 		i_mac_perim_exit(mip);
2011 		return (0);
2012 	}
2013 
2014 	mip->mi_nactiveclients--;
2015 
2016 	/* Tear down the Data path */
2017 	mac_datapath_teardown(mcip, mcip->mci_flent, SRST_LINK);
2018 
2019 	/*
2020 	 * Prevent any future access to the flow entry through the mci_flent
2021 	 * pointer by setting the mci_flent to NULL. Access to mci_flent in
2022 	 * mac_bcast_send is also under mi_rw_lock.
2023 	 */
2024 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
2025 	flent = mcip->mci_flent;
2026 	mac_client_remove_flow_from_list(mcip, flent);
2027 
2028 	if (mcip->mci_state_flags & MCIS_DESC_LOGGED)
2029 		mcip->mci_state_flags &= ~MCIS_DESC_LOGGED;
2030 
2031 	/*
2032 	 * This is the last unicast address being removed and there shouldn't
2033 	 * be any outbound data threads at this point coming down from mac
2034 	 * clients. We have waited for the data threads to finish before
2035 	 * starting dld_str_detach. Non-data threads must access TX SRS
2036 	 * under mi_rw_lock.
2037 	 */
2038 	rw_exit(&mip->mi_rw_lock);
2039 
2040 	/*
2041 	 * Update the multicast group for this vid.
2042 	 */
2043 	mac_client_bcast_refresh(mcip, mac_client_update_mcast, (void *)flent,
2044 	    B_FALSE);
2045 
2046 	/*
2047 	 * Don't use FLOW_MARK with FE_MC_NO_DATAPATH, as the flow might
2048 	 * contain other flags, such as FE_CONDEMNED, which we need to
2049 	 * cleared. We don't call mac_flow_cleanup() for this unicast
2050 	 * flow as we have a already cleaned up SRSs etc. (via the teadown
2051 	 * path). We just clear the stats and reset the initial callback
2052 	 * function, the rest will be set when we call mac_flow_create,
2053 	 * if at all.
2054 	 */
2055 	mutex_enter(&flent->fe_lock);
2056 	ASSERT(flent->fe_refcnt == 1 && flent->fe_mbg == NULL &&
2057 	    flent->fe_tx_srs == NULL && flent->fe_rx_srs_cnt == 0);
2058 	flent->fe_flags = FE_MC_NO_DATAPATH;
2059 	flow_stat_destroy(flent);
2060 
2061 	/* Initialize the receiver function to a safe routine */
2062 	flent->fe_cb_fn = (flow_fn_t)mac_pkt_drop;
2063 	flent->fe_cb_arg1 = NULL;
2064 	flent->fe_cb_arg2 = NULL;
2065 
2066 	flent->fe_index = -1;
2067 	mutex_exit(&flent->fe_lock);
2068 
2069 	if (mip->mi_type->mt_brdcst_addr != NULL) {
2070 		mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2071 		    muip->mui_vid);
2072 	}
2073 
2074 	if (mip->mi_nactiveclients == 1) {
2075 		mac_capab_update((mac_handle_t)mip);
2076 		mac_virtual_link_update(mip);
2077 	}
2078 	if (mcip->mci_state_flags & MCIS_EXCLUSIVE)
2079 		mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2080 
2081 	mac_stop(mip);
2082 
2083 	i_mac_perim_exit(mip);
2084 	kmem_free(muip, sizeof (mac_unicast_impl_t));
2085 	return (0);
2086 }
2087 
2088 /*
2089  * Multicast add function invoked by MAC clients.
2090  */
2091 int
2092 mac_multicast_add(mac_client_handle_t mch, const uint8_t *addr)
2093 {
2094 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2095 	mac_impl_t		*mip = mcip->mci_mip;
2096 	flow_entry_t		*flent = mcip->mci_flent_list;
2097 	flow_entry_t		*prev_fe = NULL;
2098 	uint16_t		vid;
2099 	int			err = 0;
2100 
2101 	/* Verify the address is a valid multicast address */
2102 	if ((err = mip->mi_type->mt_ops.mtops_multicst_verify(addr,
2103 	    mip->mi_pdata)) != 0)
2104 		return (err);
2105 
2106 	i_mac_perim_enter(mip);
2107 	while (flent != NULL) {
2108 		vid = i_mac_flow_vid(flent);
2109 
2110 		err = mac_bcast_add((mac_client_impl_t *)mch, addr, vid,
2111 		    MAC_ADDRTYPE_MULTICAST);
2112 		if (err != 0)
2113 			break;
2114 		prev_fe = flent;
2115 		flent = flent->fe_client_next;
2116 	}
2117 
2118 	/*
2119 	 * If we failed adding, then undo all, rather than partial
2120 	 * success.
2121 	 */
2122 	if (flent != NULL && prev_fe != NULL) {
2123 		flent = mcip->mci_flent_list;
2124 		while (flent != prev_fe->fe_client_next) {
2125 			vid = i_mac_flow_vid(flent);
2126 			mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
2127 			flent = flent->fe_client_next;
2128 		}
2129 	}
2130 	i_mac_perim_exit(mip);
2131 	return (err);
2132 }
2133 
2134 /*
2135  * Multicast delete function invoked by MAC clients.
2136  */
2137 void
2138 mac_multicast_remove(mac_client_handle_t mch, const uint8_t *addr)
2139 {
2140 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2141 	mac_impl_t		*mip = mcip->mci_mip;
2142 	flow_entry_t		*flent;
2143 	uint16_t		vid;
2144 
2145 	i_mac_perim_enter(mip);
2146 	for (flent = mcip->mci_flent_list; flent != NULL;
2147 	    flent = flent->fe_client_next) {
2148 		vid = i_mac_flow_vid(flent);
2149 		mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
2150 	}
2151 	i_mac_perim_exit(mip);
2152 }
2153 
2154 /*
2155  * When a MAC client desires to capture packets on an interface,
2156  * it registers a promiscuous call back with mac_promisc_add().
2157  * There are three types of promiscuous callbacks:
2158  *
2159  * * MAC_CLIENT_PROMISC_ALL
2160  *   Captures all packets sent and received by the MAC client,
2161  *   the physical interface, as well as all other MAC clients
2162  *   defined on top of the same MAC.
2163  *
2164  * * MAC_CLIENT_PROMISC_FILTERED
2165  *   Captures all packets sent and received by the MAC client,
2166  *   plus all multicast traffic sent and received by the phyisical
2167  *   interface and the other MAC clients.
2168  *
2169  * * MAC_CLIENT_PROMISC_MULTI
2170  *   Captures all broadcast and multicast packets sent and
2171  *   received by the MAC clients as well as the physical interface.
2172  *
2173  * In all cases, the underlying MAC is put in promiscuous mode.
2174  */
2175 int
2176 mac_promisc_add(mac_client_handle_t mch, mac_client_promisc_type_t type,
2177     mac_rx_t fn, void *arg, mac_promisc_handle_t *mphp, uint16_t flags)
2178 {
2179 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2180 	mac_impl_t *mip = mcip->mci_mip;
2181 	mac_promisc_impl_t *mpip;
2182 	mac_cb_info_t	*mcbi;
2183 	int rc;
2184 
2185 	i_mac_perim_enter(mip);
2186 
2187 	if ((rc = mac_start(mip)) != 0) {
2188 		i_mac_perim_exit(mip);
2189 		return (rc);
2190 	}
2191 
2192 	if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
2193 	    type == MAC_CLIENT_PROMISC_ALL) {
2194 		/*
2195 		 * The function is being invoked by the upper MAC client
2196 		 * of a VNIC. The VNIC should only see the traffic
2197 		 * it is entitled to.
2198 		 */
2199 		type = MAC_CLIENT_PROMISC_FILTERED;
2200 	}
2201 
2202 
2203 	/*
2204 	 * Turn on promiscuous mode for the underlying NIC.
2205 	 * This is needed even for filtered callbacks which
2206 	 * expect to receive all multicast traffic on the wire.
2207 	 *
2208 	 * Physical promiscuous mode should not be turned on if
2209 	 * MAC_PROMISC_FLAGS_NO_PHYS is set.
2210 	 */
2211 	if ((flags & MAC_PROMISC_FLAGS_NO_PHYS) == 0) {
2212 		if ((rc = i_mac_promisc_set(mip, B_TRUE, MAC_DEVPROMISC))
2213 		    != 0) {
2214 			mac_stop(mip);
2215 			i_mac_perim_exit(mip);
2216 			return (rc);
2217 		}
2218 	}
2219 
2220 	mpip = kmem_cache_alloc(mac_promisc_impl_cache, KM_SLEEP);
2221 
2222 	mpip->mpi_type = type;
2223 	mpip->mpi_fn = fn;
2224 	mpip->mpi_arg = arg;
2225 	mpip->mpi_mcip = mcip;
2226 	mpip->mpi_no_tx_loop = ((flags & MAC_PROMISC_FLAGS_NO_TX_LOOP) != 0);
2227 	mpip->mpi_no_phys = ((flags & MAC_PROMISC_FLAGS_NO_PHYS) != 0);
2228 
2229 	mcbi = &mip->mi_promisc_cb_info;
2230 	mutex_enter(mcbi->mcbi_lockp);
2231 
2232 	mac_callback_add(&mip->mi_promisc_cb_info, &mcip->mci_promisc_list,
2233 	    &mpip->mpi_mci_link);
2234 	mac_callback_add(&mip->mi_promisc_cb_info, &mip->mi_promisc_list,
2235 	    &mpip->mpi_mi_link);
2236 
2237 	mutex_exit(mcbi->mcbi_lockp);
2238 
2239 	*mphp = (mac_promisc_handle_t)mpip;
2240 	i_mac_perim_exit(mip);
2241 	return (0);
2242 }
2243 
2244 /*
2245  * Remove a multicast address previously aded through mac_promisc_add().
2246  */
2247 int
2248 mac_promisc_remove(mac_promisc_handle_t mph)
2249 {
2250 	mac_promisc_impl_t *mpip = (mac_promisc_impl_t *)mph;
2251 	mac_client_impl_t *mcip = mpip->mpi_mcip;
2252 	mac_impl_t *mip = mcip->mci_mip;
2253 	mac_cb_info_t *mcbi;
2254 	int rc = 0;
2255 
2256 	i_mac_perim_enter(mip);
2257 
2258 	/*
2259 	 * Even if the device can't be reset into normal mode, we still
2260 	 * need to clear the client promisc callbacks. The client may want
2261 	 * to close the mac end point and we can't have stale callbacks.
2262 	 */
2263 	if (!(mpip->mpi_no_phys)) {
2264 		rc = mac_promisc_set((mac_handle_t)mip, B_FALSE,
2265 		    MAC_DEVPROMISC);
2266 		if (rc != 0)
2267 			goto done;
2268 	}
2269 	mcbi = &mip->mi_promisc_cb_info;
2270 	mutex_enter(mcbi->mcbi_lockp);
2271 	if (mac_callback_remove(mcbi, &mip->mi_promisc_list,
2272 	    &mpip->mpi_mi_link)) {
2273 		VERIFY(mac_callback_remove(&mip->mi_promisc_cb_info,
2274 		    &mcip->mci_promisc_list, &mpip->mpi_mci_link));
2275 		kmem_cache_free(mac_promisc_impl_cache, mpip);
2276 	} else {
2277 		mac_callback_remove_wait(&mip->mi_promisc_cb_info);
2278 	}
2279 	mutex_exit(mcbi->mcbi_lockp);
2280 	mac_stop(mip);
2281 
2282 done:
2283 	i_mac_perim_exit(mip);
2284 	return (rc);
2285 }
2286 
2287 /*
2288  * Reference count the number of active Tx threads. MCI_TX_QUIESCE indicates
2289  * that a control operation wants to quiesce the Tx data flow in which case
2290  * we return an error. Holding any of the per cpu locks ensures that the
2291  * mci_tx_flag won't change.
2292  *
2293  * 'CPU' must be accessed just once and used to compute the index into the
2294  * percpu array, and that index must be used for the entire duration of the
2295  * packet send operation. Note that the thread may be preempted and run on
2296  * another cpu any time and so we can't use 'CPU' more than once for the
2297  * operation.
2298  */
2299 #define	MAC_TX_TRY_HOLD(mcip, mytx, error)				\
2300 {									\
2301 	(error) = 0;							\
2302 	(mytx) = &(mcip)->mci_tx_pcpu[CPU->cpu_seqid & mac_tx_percpu_cnt]; \
2303 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
2304 	if (!((mcip)->mci_tx_flag & MCI_TX_QUIESCE)) {			\
2305 		(mytx)->pcpu_tx_refcnt++;				\
2306 	} else {							\
2307 		(error) = -1;						\
2308 	}								\
2309 	mutex_exit(&(mytx)->pcpu_tx_lock);				\
2310 }
2311 
2312 /*
2313  * Release the reference. If needed, signal any control operation waiting
2314  * for Tx quiescence. The wait and signal are always done using the
2315  * mci_tx_pcpu[0]'s lock
2316  */
2317 #define	MAC_TX_RELE(mcip, mytx) {					\
2318 	mutex_enter(&(mytx)->pcpu_tx_lock);				\
2319 	if (--(mytx)->pcpu_tx_refcnt == 0 &&				\
2320 	    (mcip)->mci_tx_flag & MCI_TX_QUIESCE) {			\
2321 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
2322 		mutex_enter(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
2323 		cv_signal(&(mcip)->mci_tx_cv);				\
2324 		mutex_exit(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock);	\
2325 	} else {							\
2326 		mutex_exit(&(mytx)->pcpu_tx_lock);			\
2327 	}								\
2328 }
2329 
2330 /*
2331  * Bump the count of the number of active Tx threads. This is maintained as
2332  * a per CPU counter. On (CMT kind of) machines with large number of CPUs,
2333  * a single mci_tx_lock may become contended. However a count of the total
2334  * number of Tx threads per client is needed in order to quiesce the Tx side
2335  * prior to reassigning a Tx ring dynamically to another client. The thread
2336  * that needs to quiesce the Tx traffic grabs all the percpu locks and checks
2337  * the sum of the individual percpu refcnts. Each Tx data thread only grabs
2338  * its own percpu lock and increments its own refcnt.
2339  */
2340 void *
2341 mac_tx_hold(mac_client_handle_t mch)
2342 {
2343 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2344 	mac_tx_percpu_t	*mytx;
2345 	int error;
2346 
2347 	MAC_TX_TRY_HOLD(mcip, mytx, error);
2348 	return (error == 0 ? (void *)mytx : NULL);
2349 }
2350 
2351 void
2352 mac_tx_rele(mac_client_handle_t mch, void *mytx_handle)
2353 {
2354 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2355 	mac_tx_percpu_t	*mytx = mytx_handle;
2356 
2357 	MAC_TX_RELE(mcip, mytx)
2358 }
2359 
2360 /*
2361  * Send function invoked by MAC clients.
2362  */
2363 mac_tx_cookie_t
2364 mac_tx(mac_client_handle_t mch, mblk_t *mp_chain, uintptr_t hint,
2365     uint16_t flag, mblk_t **ret_mp)
2366 {
2367 	mac_tx_cookie_t		cookie;
2368 	int			error;
2369 	mac_tx_percpu_t		*mytx;
2370 	mac_soft_ring_set_t	*srs;
2371 	flow_entry_t		*flent;
2372 	boolean_t		is_subflow = B_FALSE;
2373 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2374 	mac_impl_t		*mip = mcip->mci_mip;
2375 	mac_srs_tx_t		*srs_tx;
2376 
2377 	/*
2378 	 * Check whether the active Tx threads count is bumped already.
2379 	 */
2380 	if (!(flag & MAC_TX_NO_HOLD)) {
2381 		MAC_TX_TRY_HOLD(mcip, mytx, error);
2382 		if (error != 0) {
2383 			freemsgchain(mp_chain);
2384 			return (NULL);
2385 		}
2386 	}
2387 
2388 	if (mcip->mci_subflow_tab != NULL &&
2389 	    mcip->mci_subflow_tab->ft_flow_count > 0 &&
2390 	    mac_flow_lookup(mcip->mci_subflow_tab, mp_chain,
2391 	    FLOW_OUTBOUND, &flent) == 0) {
2392 		/*
2393 		 * The main assumption here is that if in the event
2394 		 * we get a chain, all the packets will be classified
2395 		 * to the same Flow/SRS. If this changes for any
2396 		 * reason, the following logic should change as well.
2397 		 * I suppose the fanout_hint also assumes this .
2398 		 */
2399 		ASSERT(flent != NULL);
2400 		is_subflow = B_TRUE;
2401 	} else {
2402 		flent = mcip->mci_flent;
2403 	}
2404 
2405 	srs = flent->fe_tx_srs;
2406 	srs_tx = &srs->srs_tx;
2407 	if (srs_tx->st_mode == SRS_TX_DEFAULT &&
2408 	    (srs->srs_state & SRS_ENQUEUED) == 0 &&
2409 	    mip->mi_nactiveclients == 1 && mip->mi_promisc_list == NULL &&
2410 	    mp_chain->b_next == NULL) {
2411 		uint64_t	obytes;
2412 
2413 		/*
2414 		 * Since dls always opens the underlying MAC, nclients equals
2415 		 * to 1 means that the only active client is dls itself acting
2416 		 * as a primary client of the MAC instance. Since dls will not
2417 		 * send tagged packets in that case, and dls is trusted to send
2418 		 * packets for its allowed VLAN(s), the VLAN tag insertion and
2419 		 * check is required only if nclients is greater than 1.
2420 		 */
2421 		if (mip->mi_nclients > 1) {
2422 			if (MAC_VID_CHECK_NEEDED(mcip)) {
2423 				int	err = 0;
2424 
2425 				MAC_VID_CHECK(mcip, mp_chain, err);
2426 				if (err != 0) {
2427 					freemsg(mp_chain);
2428 					mcip->mci_stat_oerrors++;
2429 					goto done;
2430 				}
2431 			}
2432 			if (MAC_TAG_NEEDED(mcip)) {
2433 				mp_chain = mac_add_vlan_tag(mp_chain, 0,
2434 				    mac_client_vid(mch));
2435 				if (mp_chain == NULL) {
2436 					mcip->mci_stat_oerrors++;
2437 					goto done;
2438 				}
2439 			}
2440 		}
2441 
2442 		obytes = (mp_chain->b_cont == NULL ? MBLKL(mp_chain) :
2443 		    msgdsize(mp_chain));
2444 
2445 		MAC_TX(mip, srs_tx->st_arg2, mp_chain, mcip);
2446 
2447 		if (mp_chain == NULL) {
2448 			cookie = NULL;
2449 			mcip->mci_stat_obytes += obytes;
2450 			mcip->mci_stat_opackets += 1;
2451 			if ((srs->srs_type & SRST_FLOW) != 0) {
2452 				FLOW_STAT_UPDATE(flent, obytes, obytes);
2453 				FLOW_STAT_UPDATE(flent, opackets, 1);
2454 			}
2455 		} else {
2456 			mutex_enter(&srs->srs_lock);
2457 			cookie = mac_tx_srs_no_desc(srs, mp_chain,
2458 			    flag, ret_mp);
2459 			mutex_exit(&srs->srs_lock);
2460 		}
2461 	} else {
2462 		cookie = srs_tx->st_func(srs, mp_chain, hint, flag, ret_mp);
2463 	}
2464 
2465 done:
2466 	if (is_subflow)
2467 		FLOW_REFRELE(flent);
2468 
2469 	if (!(flag & MAC_TX_NO_HOLD))
2470 		MAC_TX_RELE(mcip, mytx);
2471 
2472 	return (cookie);
2473 }
2474 
2475 /*
2476  * mac_tx_is_blocked
2477  *
2478  * Given a cookie, it returns if the ring identified by the cookie is
2479  * flow-controlled or not (this is not implemented yet). If NULL is
2480  * passed in place of a cookie, then it finds out if any of the
2481  * underlying rings belonging to the SRS is flow controlled or not
2482  * and returns that status.
2483  */
2484 /* ARGSUSED */
2485 boolean_t
2486 mac_tx_is_flow_blocked(mac_client_handle_t mch, mac_tx_cookie_t cookie)
2487 {
2488 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2489 	mac_soft_ring_set_t *mac_srs = MCIP_TX_SRS(mcip);
2490 	mac_soft_ring_t *sringp;
2491 	boolean_t blocked = B_FALSE;
2492 	int i;
2493 
2494 	/*
2495 	 * On etherstubs, there won't be a Tx SRS or an Rx
2496 	 * SRS. Infact there won't even be a flow_entry.
2497 	 */
2498 	if (mac_srs == NULL)
2499 		return (B_FALSE);
2500 
2501 	mutex_enter(&mac_srs->srs_lock);
2502 	if (mac_srs->srs_tx.st_mode == SRS_TX_FANOUT) {
2503 		for (i = 0; i < mac_srs->srs_oth_ring_count; i++) {
2504 			sringp = mac_srs->srs_oth_soft_rings[i];
2505 			mutex_enter(&sringp->s_ring_lock);
2506 			if (sringp->s_ring_state & S_RING_TX_HIWAT) {
2507 				blocked = B_TRUE;
2508 				mutex_exit(&sringp->s_ring_lock);
2509 				break;
2510 			}
2511 			mutex_exit(&sringp->s_ring_lock);
2512 		}
2513 	} else {
2514 		blocked = (mac_srs->srs_state & SRS_TX_HIWAT);
2515 	}
2516 	mutex_exit(&mac_srs->srs_lock);
2517 	return (blocked);
2518 }
2519 
2520 /*
2521  * Check if the MAC client is the primary MAC client.
2522  */
2523 boolean_t
2524 mac_is_primary_client(mac_client_impl_t *mcip)
2525 {
2526 	return (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY);
2527 }
2528 
2529 void
2530 mac_ioctl(mac_handle_t mh, queue_t *wq, mblk_t *bp)
2531 {
2532 	mac_impl_t	*mip = (mac_impl_t *)mh;
2533 	int cmd = ((struct iocblk *)bp->b_rptr)->ioc_cmd;
2534 
2535 	if ((cmd == ND_GET && (mip->mi_callbacks->mc_callbacks & MC_GETPROP)) ||
2536 	    (cmd == ND_SET && (mip->mi_callbacks->mc_callbacks & MC_SETPROP))) {
2537 		/*
2538 		 * If ndd props were registered, call them.
2539 		 * Note that ndd ioctls are Obsolete
2540 		 */
2541 		mac_ndd_ioctl(mip, wq, bp);
2542 		return;
2543 	}
2544 
2545 	/*
2546 	 * Call the driver to handle the ioctl.  The driver may not support
2547 	 * any ioctls, in which case we reply with a NAK on its behalf.
2548 	 */
2549 	if (mip->mi_callbacks->mc_callbacks & MC_IOCTL)
2550 		mip->mi_ioctl(mip->mi_driver, wq, bp);
2551 	else
2552 		miocnak(wq, bp, 0, EINVAL);
2553 }
2554 
2555 /*
2556  * Return the link state of the specified MAC instance.
2557  */
2558 link_state_t
2559 mac_link_get(mac_handle_t mh)
2560 {
2561 	return (((mac_impl_t *)mh)->mi_linkstate);
2562 }
2563 
2564 /*
2565  * Add a mac client specified notification callback. Please see the comments
2566  * above mac_callback_add() for general information about mac callback
2567  * addition/deletion in the presence of mac callback list walkers
2568  */
2569 mac_notify_handle_t
2570 mac_notify_add(mac_handle_t mh, mac_notify_t notify_fn, void *arg)
2571 {
2572 	mac_impl_t		*mip = (mac_impl_t *)mh;
2573 	mac_notify_cb_t		*mncb;
2574 	mac_cb_info_t		*mcbi;
2575 
2576 	/*
2577 	 * Allocate a notify callback structure, fill in the details and
2578 	 * use the mac callback list manipulation functions to chain into
2579 	 * the list of callbacks.
2580 	 */
2581 	mncb = kmem_zalloc(sizeof (mac_notify_cb_t), KM_SLEEP);
2582 	mncb->mncb_fn = notify_fn;
2583 	mncb->mncb_arg = arg;
2584 	mncb->mncb_mip = mip;
2585 	mncb->mncb_link.mcb_objp = mncb;
2586 	mncb->mncb_link.mcb_objsize = sizeof (mac_notify_cb_t);
2587 	mncb->mncb_link.mcb_flags = MCB_NOTIFY_CB_T;
2588 
2589 	mcbi = &mip->mi_notify_cb_info;
2590 
2591 	i_mac_perim_enter(mip);
2592 	mutex_enter(mcbi->mcbi_lockp);
2593 
2594 	mac_callback_add(&mip->mi_notify_cb_info, &mip->mi_notify_cb_list,
2595 	    &mncb->mncb_link);
2596 
2597 	mutex_exit(mcbi->mcbi_lockp);
2598 	i_mac_perim_exit(mip);
2599 	return ((mac_notify_handle_t)mncb);
2600 }
2601 
2602 void
2603 mac_notify_remove_wait(mac_handle_t mh)
2604 {
2605 	mac_impl_t	*mip = (mac_impl_t *)mh;
2606 	mac_cb_info_t	*mcbi = &mip->mi_notify_cb_info;
2607 
2608 	mutex_enter(mcbi->mcbi_lockp);
2609 	mac_callback_remove_wait(&mip->mi_notify_cb_info);
2610 	mutex_exit(mcbi->mcbi_lockp);
2611 }
2612 
2613 /*
2614  * Remove a mac client specified notification callback
2615  */
2616 int
2617 mac_notify_remove(mac_notify_handle_t mnh, boolean_t wait)
2618 {
2619 	mac_notify_cb_t	*mncb = (mac_notify_cb_t *)mnh;
2620 	mac_impl_t	*mip = mncb->mncb_mip;
2621 	mac_cb_info_t	*mcbi;
2622 	int		err = 0;
2623 
2624 	mcbi = &mip->mi_notify_cb_info;
2625 
2626 	i_mac_perim_enter(mip);
2627 	mutex_enter(mcbi->mcbi_lockp);
2628 
2629 	ASSERT(mncb->mncb_link.mcb_objp == mncb);
2630 	/*
2631 	 * If there aren't any list walkers, the remove would succeed
2632 	 * inline, else we wait for the deferred remove to complete
2633 	 */
2634 	if (mac_callback_remove(&mip->mi_notify_cb_info,
2635 	    &mip->mi_notify_cb_list, &mncb->mncb_link)) {
2636 		kmem_free(mncb, sizeof (mac_notify_cb_t));
2637 	} else {
2638 		err = EBUSY;
2639 	}
2640 
2641 	mutex_exit(mcbi->mcbi_lockp);
2642 	i_mac_perim_exit(mip);
2643 
2644 	/*
2645 	 * If we failed to remove the notification callback and "wait" is set
2646 	 * to be B_TRUE, wait for the callback to finish after we exit the
2647 	 * mac perimeter.
2648 	 */
2649 	if (err != 0 && wait) {
2650 		mac_notify_remove_wait((mac_handle_t)mip);
2651 		return (0);
2652 	}
2653 
2654 	return (err);
2655 }
2656 
2657 /*
2658  * Associate resource management callbacks with the specified MAC
2659  * clients.
2660  */
2661 
2662 void
2663 mac_resource_set_common(mac_client_handle_t mch, mac_resource_add_t add,
2664     mac_resource_remove_t remove, mac_resource_quiesce_t quiesce,
2665     mac_resource_restart_t restart, mac_resource_bind_t bind,
2666     void *arg)
2667 {
2668 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2669 
2670 	mcip->mci_resource_add = add;
2671 	mcip->mci_resource_remove = remove;
2672 	mcip->mci_resource_quiesce = quiesce;
2673 	mcip->mci_resource_restart = restart;
2674 	mcip->mci_resource_bind = bind;
2675 	mcip->mci_resource_arg = arg;
2676 
2677 	if (arg == NULL)
2678 		mcip->mci_state_flags &= ~MCIS_CLIENT_POLL_CAPABLE;
2679 }
2680 
2681 void
2682 mac_resource_set(mac_client_handle_t mch, mac_resource_add_t add, void *arg)
2683 {
2684 	/* update the 'resource_add' callback */
2685 	mac_resource_set_common(mch, add, NULL, NULL, NULL, NULL, arg);
2686 }
2687 
2688 /*
2689  * Sets up the client resources and enable the polling interface over all the
2690  * SRS's and the soft rings of the client
2691  */
2692 void
2693 mac_client_poll_enable(mac_client_handle_t mch)
2694 {
2695 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2696 	mac_soft_ring_set_t	*mac_srs;
2697 	flow_entry_t		*flent;
2698 	int			i;
2699 
2700 	flent = mcip->mci_flent;
2701 	ASSERT(flent != NULL);
2702 
2703 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
2704 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
2705 		ASSERT(mac_srs->srs_mcip == mcip);
2706 		mac_srs_client_poll_enable(mcip, mac_srs);
2707 	}
2708 }
2709 
2710 /*
2711  * Tears down the client resources and disable the polling interface over all
2712  * the SRS's and the soft rings of the client
2713  */
2714 void
2715 mac_client_poll_disable(mac_client_handle_t mch)
2716 {
2717 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2718 	mac_soft_ring_set_t	*mac_srs;
2719 	flow_entry_t		*flent;
2720 	int			i;
2721 
2722 	flent = mcip->mci_flent;
2723 	ASSERT(flent != NULL);
2724 
2725 	for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
2726 		mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
2727 		ASSERT(mac_srs->srs_mcip == mcip);
2728 		mac_srs_client_poll_disable(mcip, mac_srs);
2729 	}
2730 }
2731 
2732 /*
2733  * Associate the CPUs specified by the given property with a MAC client.
2734  */
2735 int
2736 mac_cpu_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
2737 {
2738 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2739 	mac_impl_t *mip = mcip->mci_mip;
2740 	int err = 0;
2741 
2742 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
2743 
2744 	if ((err = mac_validate_props(mrp)) != 0)
2745 		return (err);
2746 
2747 	if (MCIP_DATAPATH_SETUP(mcip))
2748 		mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
2749 
2750 	mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
2751 	return (0);
2752 }
2753 
2754 /*
2755  * Apply the specified properties to the specified MAC client.
2756  */
2757 int
2758 mac_client_set_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
2759 {
2760 	mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2761 	mac_impl_t *mip = mcip->mci_mip;
2762 	int err = 0;
2763 
2764 	i_mac_perim_enter(mip);
2765 
2766 	if ((mrp->mrp_mask & MRP_MAXBW) || (mrp->mrp_mask & MRP_PRIORITY)) {
2767 		err = mac_resource_ctl_set(mch, mrp);
2768 		if (err != 0) {
2769 			i_mac_perim_exit(mip);
2770 			return (err);
2771 		}
2772 	}
2773 
2774 	if (mrp->mrp_mask & MRP_CPUS)
2775 		err = mac_cpu_set(mch, mrp);
2776 
2777 	i_mac_perim_exit(mip);
2778 	return (err);
2779 }
2780 
2781 /*
2782  * Return the properties currently associated with the specified MAC client.
2783  */
2784 void
2785 mac_client_get_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
2786 {
2787 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
2788 	mac_resource_props_t	*mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
2789 
2790 	bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
2791 }
2792 
2793 /*
2794  * Pass a copy of the specified packet to the promiscuous callbacks
2795  * of the specified MAC.
2796  *
2797  * If sender is NULL, the function is being invoked for a packet chain
2798  * received from the wire. If sender is non-NULL, it points to
2799  * the MAC client from which the packet is being sent.
2800  *
2801  * The packets are distributed to the promiscuous callbacks as follows:
2802  *
2803  * - all packets are sent to the MAC_CLIENT_PROMISC_ALL callbacks
2804  * - all broadcast and multicast packets are sent to the
2805  *   MAC_CLIENT_PROMISC_FILTER and MAC_CLIENT_PROMISC_MULTI.
2806  *
2807  * The unicast packets of MAC_CLIENT_PROMISC_FILTER callbacks are dispatched
2808  * after classification by mac_rx_deliver().
2809  */
2810 
2811 static void
2812 mac_promisc_dispatch_one(mac_promisc_impl_t *mpip, mblk_t *mp,
2813     boolean_t loopback)
2814 {
2815 	mblk_t *mp_copy;
2816 
2817 	mp_copy = copymsg(mp);
2818 	if (mp_copy == NULL)
2819 		return;
2820 	mp_copy->b_next = NULL;
2821 
2822 	mpip->mpi_fn(mpip->mpi_arg, NULL, mp_copy, loopback);
2823 }
2824 
2825 /*
2826  * Return the VID of a packet. Zero if the packet is not tagged.
2827  */
2828 static uint16_t
2829 mac_ether_vid(mblk_t *mp)
2830 {
2831 	struct ether_header *eth = (struct ether_header *)mp->b_rptr;
2832 
2833 	if (ntohs(eth->ether_type) == ETHERTYPE_VLAN) {
2834 		struct ether_vlan_header *t_evhp =
2835 		    (struct ether_vlan_header *)mp->b_rptr;
2836 		return (VLAN_ID(ntohs(t_evhp->ether_tci)));
2837 	}
2838 
2839 	return (0);
2840 }
2841 
2842 /*
2843  * Return whether the specified packet contains a multicast or broadcast
2844  * destination MAC address.
2845  */
2846 static boolean_t
2847 mac_is_mcast(mac_impl_t *mip, mblk_t *mp)
2848 {
2849 	mac_header_info_t hdr_info;
2850 
2851 	if (mac_header_info((mac_handle_t)mip, mp, &hdr_info) != 0)
2852 		return (B_FALSE);
2853 	return ((hdr_info.mhi_dsttype == MAC_ADDRTYPE_BROADCAST) ||
2854 	    (hdr_info.mhi_dsttype == MAC_ADDRTYPE_MULTICAST));
2855 }
2856 
2857 /*
2858  * Send a copy of an mblk chain to the MAC clients of the specified MAC.
2859  * "sender" points to the sender MAC client for outbound packets, and
2860  * is set to NULL for inbound packets.
2861  */
2862 void
2863 mac_promisc_dispatch(mac_impl_t *mip, mblk_t *mp_chain,
2864     mac_client_impl_t *sender)
2865 {
2866 	mac_promisc_impl_t *mpip;
2867 	mac_cb_t *mcb;
2868 	mblk_t *mp;
2869 	boolean_t is_mcast, is_sender;
2870 
2871 	MAC_PROMISC_WALKER_INC(mip);
2872 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
2873 		is_mcast = mac_is_mcast(mip, mp);
2874 		/* send packet to interested callbacks */
2875 		for (mcb = mip->mi_promisc_list; mcb != NULL;
2876 		    mcb = mcb->mcb_nextp) {
2877 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
2878 			is_sender = (mpip->mpi_mcip == sender);
2879 
2880 			if (is_sender && mpip->mpi_no_tx_loop)
2881 				/*
2882 				 * The sender doesn't want to receive
2883 				 * copies of the packets it sends.
2884 				 */
2885 				continue;
2886 
2887 			/*
2888 			 * For an ethernet MAC, don't displatch a multicast
2889 			 * packet to a non-PROMISC_ALL callbacks unless the VID
2890 			 * of the packet matches the VID of the client.
2891 			 */
2892 			if (is_mcast &&
2893 			    mpip->mpi_type != MAC_CLIENT_PROMISC_ALL &&
2894 			    !mac_client_check_flow_vid(mpip->mpi_mcip,
2895 			    mac_ether_vid(mp)))
2896 				continue;
2897 
2898 			if (is_sender ||
2899 			    mpip->mpi_type == MAC_CLIENT_PROMISC_ALL ||
2900 			    is_mcast)
2901 				mac_promisc_dispatch_one(mpip, mp, is_sender);
2902 		}
2903 	}
2904 	MAC_PROMISC_WALKER_DCR(mip);
2905 }
2906 
2907 void
2908 mac_promisc_client_dispatch(mac_client_impl_t *mcip, mblk_t *mp_chain)
2909 {
2910 	mac_impl_t		*mip = mcip->mci_mip;
2911 	mac_promisc_impl_t	*mpip;
2912 	boolean_t		is_mcast;
2913 	mblk_t			*mp;
2914 	mac_cb_t		*mcb;
2915 
2916 	/*
2917 	 * The unicast packets for the MAC client still
2918 	 * need to be delivered to the MAC_CLIENT_PROMISC_FILTERED
2919 	 * promiscuous callbacks. The broadcast and multicast
2920 	 * packets were delivered from mac_rx().
2921 	 */
2922 	MAC_PROMISC_WALKER_INC(mip);
2923 	for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
2924 		is_mcast = mac_is_mcast(mip, mp);
2925 		for (mcb = mcip->mci_promisc_list; mcb != NULL;
2926 		    mcb = mcb->mcb_nextp) {
2927 			mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
2928 			if (mpip->mpi_type == MAC_CLIENT_PROMISC_FILTERED &&
2929 			    !is_mcast) {
2930 				mac_promisc_dispatch_one(mpip, mp, B_FALSE);
2931 			}
2932 		}
2933 	}
2934 	MAC_PROMISC_WALKER_DCR(mip);
2935 }
2936 
2937 /*
2938  * Return the margin value currently assigned to the specified MAC instance.
2939  */
2940 void
2941 mac_margin_get(mac_handle_t mh, uint32_t *marginp)
2942 {
2943 	mac_impl_t *mip = (mac_impl_t *)mh;
2944 
2945 	rw_enter(&(mip->mi_rw_lock), RW_READER);
2946 	*marginp = mip->mi_margin;
2947 	rw_exit(&(mip->mi_rw_lock));
2948 }
2949 
2950 /*
2951  * mac_info_get() is used for retrieving the mac_info when a DL_INFO_REQ is
2952  * issued before a DL_ATTACH_REQ. we walk the i_mac_impl_hash table and find
2953  * the first mac_impl_t with a matching driver name; then we copy its mac_info_t
2954  * to the caller. we do all this with i_mac_impl_lock held so the mac_impl_t
2955  * cannot disappear while we are accessing it.
2956  */
2957 typedef struct i_mac_info_state_s {
2958 	const char	*mi_name;
2959 	mac_info_t	*mi_infop;
2960 } i_mac_info_state_t;
2961 
2962 /*ARGSUSED*/
2963 static uint_t
2964 i_mac_info_walker(mod_hash_key_t key, mod_hash_val_t *val, void *arg)
2965 {
2966 	i_mac_info_state_t *statep = arg;
2967 	mac_impl_t *mip = (mac_impl_t *)val;
2968 
2969 	if (mip->mi_state_flags & MIS_DISABLED)
2970 		return (MH_WALK_CONTINUE);
2971 
2972 	if (strcmp(statep->mi_name,
2973 	    ddi_driver_name(mip->mi_dip)) != 0)
2974 		return (MH_WALK_CONTINUE);
2975 
2976 	statep->mi_infop = &mip->mi_info;
2977 	return (MH_WALK_TERMINATE);
2978 }
2979 
2980 boolean_t
2981 mac_info_get(const char *name, mac_info_t *minfop)
2982 {
2983 	i_mac_info_state_t state;
2984 
2985 	rw_enter(&i_mac_impl_lock, RW_READER);
2986 	state.mi_name = name;
2987 	state.mi_infop = NULL;
2988 	mod_hash_walk(i_mac_impl_hash, i_mac_info_walker, &state);
2989 	if (state.mi_infop == NULL) {
2990 		rw_exit(&i_mac_impl_lock);
2991 		return (B_FALSE);
2992 	}
2993 	*minfop = *state.mi_infop;
2994 	rw_exit(&i_mac_impl_lock);
2995 	return (B_TRUE);
2996 }
2997 
2998 /*
2999  * To get the capabilities that MAC layer cares about, such as rings, factory
3000  * mac address, vnic or not, it should directly invoke this function
3001  */
3002 boolean_t
3003 i_mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
3004 {
3005 	mac_impl_t *mip = (mac_impl_t *)mh;
3006 
3007 	if (mip->mi_callbacks->mc_callbacks & MC_GETCAPAB)
3008 		return (mip->mi_getcapab(mip->mi_driver, cap, cap_data));
3009 	else
3010 		return (B_FALSE);
3011 }
3012 
3013 /*
3014  * Capability query function. If number of active mac clients is greater than
3015  * 1, only limited capabilities can be advertised to the caller no matter the
3016  * driver has certain capability or not. Else, we query the driver to get the
3017  * capability.
3018  */
3019 boolean_t
3020 mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
3021 {
3022 	mac_impl_t *mip = (mac_impl_t *)mh;
3023 
3024 	/*
3025 	 * if mi_nactiveclients > 1, only MAC_CAPAB_HCKSUM,
3026 	 * MAC_CAPAB_NO_NATIVEVLAN, MAC_CAPAB_NO_ZCOPY can be advertised.
3027 	 */
3028 	if (mip->mi_nactiveclients > 1) {
3029 		switch (cap) {
3030 		case MAC_CAPAB_HCKSUM:
3031 			return (i_mac_capab_get(mh, cap, cap_data));
3032 		case MAC_CAPAB_NO_NATIVEVLAN:
3033 		case MAC_CAPAB_NO_ZCOPY:
3034 			return (B_TRUE);
3035 		default:
3036 			return (B_FALSE);
3037 		}
3038 	}
3039 
3040 	/* else get capab from driver */
3041 	return (i_mac_capab_get(mh, cap, cap_data));
3042 }
3043 
3044 boolean_t
3045 mac_sap_verify(mac_handle_t mh, uint32_t sap, uint32_t *bind_sap)
3046 {
3047 	mac_impl_t *mip = (mac_impl_t *)mh;
3048 
3049 	return (mip->mi_type->mt_ops.mtops_sap_verify(sap, bind_sap,
3050 	    mip->mi_pdata));
3051 }
3052 
3053 mblk_t *
3054 mac_header(mac_handle_t mh, const uint8_t *daddr, uint32_t sap, mblk_t *payload,
3055     size_t extra_len)
3056 {
3057 	mac_impl_t *mip = (mac_impl_t *)mh;
3058 
3059 	return (mip->mi_type->mt_ops.mtops_header(mip->mi_addr, daddr, sap,
3060 	    mip->mi_pdata, payload, extra_len));
3061 }
3062 
3063 int
3064 mac_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
3065 {
3066 	mac_impl_t *mip = (mac_impl_t *)mh;
3067 
3068 	return (mip->mi_type->mt_ops.mtops_header_info(mp, mip->mi_pdata,
3069 	    mhip));
3070 }
3071 
3072 mblk_t *
3073 mac_header_cook(mac_handle_t mh, mblk_t *mp)
3074 {
3075 	mac_impl_t *mip = (mac_impl_t *)mh;
3076 
3077 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_COOK) {
3078 		if (DB_REF(mp) > 1) {
3079 			mblk_t *newmp = copymsg(mp);
3080 			if (newmp == NULL)
3081 				return (NULL);
3082 			freemsg(mp);
3083 			mp = newmp;
3084 		}
3085 		return (mip->mi_type->mt_ops.mtops_header_cook(mp,
3086 		    mip->mi_pdata));
3087 	}
3088 	return (mp);
3089 }
3090 
3091 mblk_t *
3092 mac_header_uncook(mac_handle_t mh, mblk_t *mp)
3093 {
3094 	mac_impl_t *mip = (mac_impl_t *)mh;
3095 
3096 	if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_UNCOOK) {
3097 		if (DB_REF(mp) > 1) {
3098 			mblk_t *newmp = copymsg(mp);
3099 			if (newmp == NULL)
3100 				return (NULL);
3101 			freemsg(mp);
3102 			mp = newmp;
3103 		}
3104 		return (mip->mi_type->mt_ops.mtops_header_uncook(mp,
3105 		    mip->mi_pdata));
3106 	}
3107 	return (mp);
3108 }
3109 
3110 uint_t
3111 mac_addr_len(mac_handle_t mh)
3112 {
3113 	mac_impl_t *mip = (mac_impl_t *)mh;
3114 
3115 	return (mip->mi_type->mt_addr_length);
3116 }
3117 
3118 /* True if a MAC is a VNIC */
3119 boolean_t
3120 mac_is_vnic(mac_handle_t mh)
3121 {
3122 	return (((mac_impl_t *)mh)->mi_state_flags & MIS_IS_VNIC);
3123 }
3124 
3125 mac_handle_t
3126 mac_get_lower_mac_handle(mac_handle_t mh)
3127 {
3128 	mac_impl_t *mip = (mac_impl_t *)mh;
3129 
3130 	ASSERT(mac_is_vnic(mh));
3131 	return (((vnic_t *)mip->mi_driver)->vn_lower_mh);
3132 }
3133 
3134 void
3135 mac_update_resources(mac_resource_props_t *nmrp, mac_resource_props_t *cmrp,
3136     boolean_t is_user_flow)
3137 {
3138 	if (nmrp != NULL && cmrp != NULL) {
3139 		if (nmrp->mrp_mask & MRP_PRIORITY) {
3140 			if (nmrp->mrp_priority == MPL_RESET) {
3141 				cmrp->mrp_mask &= ~MRP_PRIORITY;
3142 				if (is_user_flow) {
3143 					cmrp->mrp_priority =
3144 					    MPL_SUBFLOW_DEFAULT;
3145 				} else {
3146 					cmrp->mrp_priority = MPL_LINK_DEFAULT;
3147 				}
3148 			} else {
3149 				cmrp->mrp_mask |= MRP_PRIORITY;
3150 				cmrp->mrp_priority = nmrp->mrp_priority;
3151 			}
3152 		}
3153 		if (nmrp->mrp_mask & MRP_MAXBW) {
3154 			cmrp->mrp_maxbw = nmrp->mrp_maxbw;
3155 			if (nmrp->mrp_maxbw == MRP_MAXBW_RESETVAL)
3156 				cmrp->mrp_mask &= ~MRP_MAXBW;
3157 			else
3158 				cmrp->mrp_mask |= MRP_MAXBW;
3159 		}
3160 		if (nmrp->mrp_mask & MRP_CPUS)
3161 			MAC_COPY_CPUS(nmrp, cmrp);
3162 	}
3163 }
3164 
3165 /*
3166  * i_mac_set_resources:
3167  *
3168  * This routine associates properties with the primary MAC client of
3169  * the specified MAC instance.
3170  * - Cache the properties in mac_impl_t
3171  * - Apply the properties to the primary MAC client if exists
3172  */
3173 int
3174 i_mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3175 {
3176 	mac_impl_t		*mip = (mac_impl_t *)mh;
3177 	mac_client_impl_t	*mcip;
3178 	int			err = 0;
3179 	mac_resource_props_t	tmrp;
3180 
3181 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3182 
3183 	err = mac_validate_props(mrp);
3184 	if (err != 0)
3185 		return (err);
3186 
3187 	/*
3188 	 * Since bind_cpu may be modified by mac_client_set_resources()
3189 	 * we use a copy of bind_cpu and finally cache bind_cpu in mip.
3190 	 * This allows us to cache only user edits in mip.
3191 	 */
3192 	bcopy(mrp, &tmrp, sizeof (mac_resource_props_t));
3193 	mcip = mac_primary_client_handle(mip);
3194 	if (mcip != NULL) {
3195 		err =
3196 		    mac_client_set_resources((mac_client_handle_t)mcip, &tmrp);
3197 	}
3198 	/* if mac_client_set_resources failed, do not update the values */
3199 	if (err == 0)
3200 		mac_update_resources(mrp, &mip->mi_resource_props, B_FALSE);
3201 	return (err);
3202 }
3203 
3204 int
3205 mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3206 {
3207 	int err;
3208 
3209 	i_mac_perim_enter((mac_impl_t *)mh);
3210 	err = i_mac_set_resources(mh, mrp);
3211 	i_mac_perim_exit((mac_impl_t *)mh);
3212 	return (err);
3213 }
3214 
3215 /*
3216  * Get the properties cached for the specified MAC instance.
3217  */
3218 void
3219 mac_get_resources(mac_handle_t mh, mac_resource_props_t *mrp)
3220 {
3221 	mac_impl_t 		*mip = (mac_impl_t *)mh;
3222 	mac_client_impl_t	*mcip;
3223 
3224 	if (mip->mi_state_flags & MIS_IS_VNIC) {
3225 		mcip = mac_primary_client_handle(mip);
3226 		if (mcip != NULL) {
3227 			mac_client_get_resources((mac_client_handle_t)mcip,
3228 			    mrp);
3229 			return;
3230 		}
3231 	}
3232 	bcopy(&mip->mi_resource_props, mrp, sizeof (mac_resource_props_t));
3233 }
3234 
3235 /*
3236  * Rename a mac client, its flow, and the kstat.
3237  */
3238 int
3239 mac_rename_primary(mac_handle_t mh, const char *new_name)
3240 {
3241 	mac_impl_t		*mip = (mac_impl_t *)mh;
3242 	mac_client_impl_t	*cur_clnt = NULL;
3243 	flow_entry_t		*fep;
3244 
3245 	i_mac_perim_enter(mip);
3246 
3247 	/*
3248 	 * VNICs: we need to change the sys flow name and
3249 	 * the associated flow kstat.
3250 	 */
3251 	if (mip->mi_state_flags & MIS_IS_VNIC) {
3252 		ASSERT(new_name != NULL);
3253 		mac_rename_flow_names(mac_vnic_lower(mip), new_name);
3254 		goto done;
3255 	}
3256 	/*
3257 	 * This mac may itself be an aggr link, or it may have some client
3258 	 * which is an aggr port. For both cases, we need to change the
3259 	 * aggr port's mac client name, its flow name and the associated flow
3260 	 * kstat.
3261 	 */
3262 	if (mip->mi_state_flags & MIS_IS_AGGR) {
3263 		mac_capab_aggr_t aggr_cap;
3264 		mac_rename_fn_t rename_fn;
3265 		boolean_t ret;
3266 
3267 		ASSERT(new_name != NULL);
3268 		ret = i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_AGGR,
3269 		    (void *)(&aggr_cap));
3270 		ASSERT(ret == B_TRUE);
3271 		rename_fn = aggr_cap.mca_rename_fn;
3272 		rename_fn(new_name, mip->mi_driver);
3273 		/*
3274 		 * The aggr's client name and kstat flow name will be
3275 		 * updated below, i.e. via mac_rename_flow_names.
3276 		 */
3277 	}
3278 
3279 	for (cur_clnt = mip->mi_clients_list; cur_clnt != NULL;
3280 	    cur_clnt = cur_clnt->mci_client_next) {
3281 		if (cur_clnt->mci_state_flags & MCIS_IS_AGGR_PORT) {
3282 			if (new_name != NULL) {
3283 				char *str_st = cur_clnt->mci_name;
3284 				char *str_del = strchr(str_st, '-');
3285 
3286 				ASSERT(str_del != NULL);
3287 				bzero(str_del + 1, MAXNAMELEN -
3288 				    (str_del - str_st + 1));
3289 				bcopy(new_name, str_del + 1,
3290 				    strlen(new_name));
3291 			}
3292 			fep = cur_clnt->mci_flent;
3293 			mac_rename_flow(fep, cur_clnt->mci_name);
3294 			break;
3295 		} else if (new_name != NULL &&
3296 		    cur_clnt->mci_state_flags & MCIS_USE_DATALINK_NAME) {
3297 			mac_rename_flow_names(cur_clnt, new_name);
3298 			break;
3299 		}
3300 	}
3301 
3302 done:
3303 	i_mac_perim_exit(mip);
3304 	return (0);
3305 }
3306 
3307 /*
3308  * Rename the MAC client's flow names
3309  */
3310 static void
3311 mac_rename_flow_names(mac_client_impl_t *mcip, const char *new_name)
3312 {
3313 	flow_entry_t	*flent;
3314 	uint16_t	vid;
3315 	char		flowname[MAXFLOWNAME];
3316 	mac_impl_t	*mip = mcip->mci_mip;
3317 
3318 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3319 
3320 	/*
3321 	 * Use mi_rw_lock to ensure that threads not in the mac perimeter
3322 	 * see a self-consistent value for mci_name
3323 	 */
3324 	rw_enter(&mip->mi_rw_lock, RW_WRITER);
3325 	(void) strlcpy(mcip->mci_name, new_name, sizeof (mcip->mci_name));
3326 	rw_exit(&mip->mi_rw_lock);
3327 
3328 	mac_rename_flow(mcip->mci_flent, new_name);
3329 
3330 	if (mcip->mci_nflents == 1)
3331 		return;
3332 
3333 	/*
3334 	 * We have to rename all the others too, no stats to destroy for
3335 	 * these.
3336 	 */
3337 	for (flent = mcip->mci_flent_list; flent != NULL;
3338 	    flent = flent->fe_client_next) {
3339 		if (flent != mcip->mci_flent) {
3340 			vid = i_mac_flow_vid(flent);
3341 			(void) sprintf(flowname, "%s%u", new_name, vid);
3342 			mac_flow_set_name(flent, flowname);
3343 		}
3344 	}
3345 }
3346 
3347 
3348 /*
3349  * Add a flow to the MAC client's flow list - i.e list of MAC/VID tuples
3350  * defined for the specified MAC client.
3351  */
3352 static void
3353 mac_client_add_to_flow_list(mac_client_impl_t *mcip, flow_entry_t *flent)
3354 {
3355 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3356 	/*
3357 	 * The promisc Rx data path walks the mci_flent_list. Protect by
3358 	 * using mi_rw_lock
3359 	 */
3360 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3361 
3362 	/* Add it to the head */
3363 	flent->fe_client_next = mcip->mci_flent_list;
3364 	mcip->mci_flent_list = flent;
3365 	mcip->mci_nflents++;
3366 
3367 	/*
3368 	 * Keep track of the number of non-zero VIDs addresses per MAC
3369 	 * client to avoid figuring it out in the data-path.
3370 	 */
3371 	if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
3372 		mcip->mci_nvids++;
3373 
3374 	rw_exit(&mcip->mci_rw_lock);
3375 }
3376 
3377 /*
3378  * Remove a flow entry from the MAC client's list.
3379  */
3380 static void
3381 mac_client_remove_flow_from_list(mac_client_impl_t *mcip, flow_entry_t *flent)
3382 {
3383 	flow_entry_t	*fe = mcip->mci_flent_list;
3384 	flow_entry_t	*prev_fe = NULL;
3385 
3386 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3387 	/*
3388 	 * The promisc Rx data path walks the mci_flent_list. Protect by
3389 	 * using mci_rw_lock
3390 	 */
3391 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3392 	while ((fe != NULL) && (fe != flent)) {
3393 		prev_fe = fe;
3394 		fe = fe->fe_client_next;
3395 	}
3396 
3397 	/* XXX should be an ASSERT */
3398 	if (fe != NULL) {
3399 		if (prev_fe == NULL) {
3400 			/* Deleting the first node */
3401 			mcip->mci_flent_list = fe->fe_client_next;
3402 		} else {
3403 			prev_fe->fe_client_next = fe->fe_client_next;
3404 		}
3405 		mcip->mci_nflents--;
3406 
3407 		if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
3408 			mcip->mci_nvids--;
3409 	}
3410 	rw_exit(&mcip->mci_rw_lock);
3411 }
3412 
3413 /*
3414  * Check if the given VID belongs to this MAC client.
3415  */
3416 boolean_t
3417 mac_client_check_flow_vid(mac_client_impl_t *mcip, uint16_t vid)
3418 {
3419 	flow_entry_t	*flent;
3420 	uint16_t	mci_vid;
3421 
3422 	/* The mci_flent_list is protected by mci_rw_lock */
3423 	rw_enter(&mcip->mci_rw_lock, RW_WRITER);
3424 	for (flent = mcip->mci_flent_list; flent != NULL;
3425 	    flent = flent->fe_client_next) {
3426 		mci_vid = i_mac_flow_vid(flent);
3427 		if (vid == mci_vid) {
3428 			rw_exit(&mcip->mci_rw_lock);
3429 			return (B_TRUE);
3430 		}
3431 	}
3432 	rw_exit(&mcip->mci_rw_lock);
3433 	return (B_FALSE);
3434 }
3435 
3436 /*
3437  * Get the flow entry for the specified <MAC addr, VID> tuple.
3438  */
3439 static flow_entry_t *
3440 mac_client_get_flow(mac_client_impl_t *mcip, mac_unicast_impl_t *muip)
3441 {
3442 	mac_address_t *map = mcip->mci_unicast;
3443 	flow_entry_t *flent;
3444 	uint16_t vid;
3445 	flow_desc_t flow_desc;
3446 
3447 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3448 
3449 	mac_flow_get_desc(mcip->mci_flent, &flow_desc);
3450 	if (bcmp(flow_desc.fd_dst_mac, map->ma_addr, map->ma_len) != 0)
3451 		return (NULL);
3452 
3453 	for (flent = mcip->mci_flent_list; flent != NULL;
3454 	    flent = flent->fe_client_next) {
3455 		vid = i_mac_flow_vid(flent);
3456 		if (vid == muip->mui_vid) {
3457 			return (flent);
3458 		}
3459 	}
3460 
3461 	return (NULL);
3462 }
3463 
3464 /*
3465  * Since mci_flent has the SRSs, when we want to remove it, we replace
3466  * the flow_desc_t in mci_flent with that of an existing flent and then
3467  * remove that flent instead of mci_flent.
3468  */
3469 static flow_entry_t *
3470 mac_client_swap_mciflent(mac_client_impl_t *mcip)
3471 {
3472 	flow_entry_t	*flent = mcip->mci_flent;
3473 	flow_tab_t	*ft = flent->fe_flow_tab;
3474 	flow_entry_t	*flent1;
3475 	flow_desc_t	fl_desc;
3476 	char		fl_name[MAXFLOWNAME];
3477 	int		err;
3478 
3479 	ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
3480 	ASSERT(mcip->mci_nflents > 1);
3481 
3482 	/* get the next flent following the primary flent  */
3483 	flent1 = mcip->mci_flent_list->fe_client_next;
3484 	ASSERT(flent1 != NULL && flent1->fe_flow_tab == ft);
3485 
3486 	/*
3487 	 * Remove the flent from the flow table before updating the
3488 	 * flow descriptor as the hash depends on the flow descriptor.
3489 	 * This also helps incoming packet classification avoid having
3490 	 * to grab fe_lock. Access to fe_flow_desc of a flent not in the
3491 	 * flow table is done under the fe_lock so that log or stat functions
3492 	 * see a self-consistent fe_flow_desc. The name and desc are specific
3493 	 * to a flow, the rest are shared by all the clients, including
3494 	 * resource control etc.
3495 	 */
3496 	mac_flow_remove(ft, flent, B_TRUE);
3497 	mac_flow_remove(ft, flent1, B_TRUE);
3498 
3499 	bcopy(&flent->fe_flow_desc, &fl_desc, sizeof (flow_desc_t));
3500 	bcopy(flent->fe_flow_name, fl_name, MAXFLOWNAME);
3501 
3502 	/* update the primary flow entry */
3503 	mutex_enter(&flent->fe_lock);
3504 	bcopy(&flent1->fe_flow_desc, &flent->fe_flow_desc,
3505 	    sizeof (flow_desc_t));
3506 	bcopy(&flent1->fe_flow_name, &flent->fe_flow_name, MAXFLOWNAME);
3507 	mutex_exit(&flent->fe_lock);
3508 
3509 	/* update the flow entry that is to be freed */
3510 	mutex_enter(&flent1->fe_lock);
3511 	bcopy(&fl_desc, &flent1->fe_flow_desc, sizeof (flow_desc_t));
3512 	bcopy(fl_name, &flent1->fe_flow_name, MAXFLOWNAME);
3513 	mutex_exit(&flent1->fe_lock);
3514 
3515 	/* now reinsert the flow entries in the table */
3516 	err = mac_flow_add(ft, flent);
3517 	ASSERT(err == 0);
3518 
3519 	err = mac_flow_add(ft, flent1);
3520 	ASSERT(err == 0);
3521 
3522 	return (flent1);
3523 }
3524 
3525 /*
3526  * Return whether there is only one flow entry associated with this
3527  * MAC client.
3528  */
3529 static boolean_t
3530 mac_client_single_rcvr(mac_client_impl_t *mcip)
3531 {
3532 	return (mcip->mci_nflents == 1);
3533 }
3534 
3535 int
3536 mac_validate_props(mac_resource_props_t *mrp)
3537 {
3538 	if (mrp == NULL)
3539 		return (0);
3540 
3541 	if (mrp->mrp_mask & MRP_PRIORITY) {
3542 		mac_priority_level_t	pri = mrp->mrp_priority;
3543 
3544 		if (pri < MPL_LOW || pri > MPL_RESET)
3545 			return (EINVAL);
3546 	}
3547 
3548 	if (mrp->mrp_mask & MRP_MAXBW) {
3549 		uint64_t maxbw = mrp->mrp_maxbw;
3550 
3551 		if (maxbw < MRP_MAXBW_MINVAL && maxbw != 0)
3552 			return (EINVAL);
3553 	}
3554 	if (mrp->mrp_mask & MRP_CPUS) {
3555 		int i;
3556 		mac_cpu_mode_t	fanout;
3557 
3558 		if (mrp->mrp_ncpus > ncpus || mrp->mrp_ncpus > MAX_SR_FANOUT)
3559 			return (EINVAL);
3560 
3561 		for (i = 0; i < mrp->mrp_ncpus; i++) {
3562 			cpu_t *cp;
3563 			int rv;
3564 
3565 			mutex_enter(&cpu_lock);
3566 			cp = cpu_get(mrp->mrp_cpu[i]);
3567 			if (cp != NULL)
3568 				rv = cpu_is_online(cp);
3569 			else
3570 				rv = 0;
3571 			mutex_exit(&cpu_lock);
3572 			if (rv == 0)
3573 				return (EINVAL);
3574 		}
3575 
3576 		fanout = mrp->mrp_fanout_mode;
3577 		if (fanout < 0 || fanout > MCM_CPUS)
3578 			return (EINVAL);
3579 	}
3580 	return (0);
3581 }
3582 
3583 /*
3584  * Send a MAC_NOTE_LINK notification to all the MAC clients whenever the
3585  * underlying physical link is down. This is to allow MAC clients to
3586  * communicate with other clients.
3587  */
3588 void
3589 mac_virtual_link_update(mac_impl_t *mip)
3590 {
3591 	if (mip->mi_linkstate != LINK_STATE_UP)
3592 		i_mac_notify(mip, MAC_NOTE_LINK);
3593 }
3594 
3595 /*
3596  * For clients that have a pass-thru MAC, e.g. VNIC, we set the VNIC's
3597  * mac handle in the client.
3598  */
3599 void
3600 mac_set_upper_mac(mac_client_handle_t mch, mac_handle_t mh)
3601 {
3602 	mac_client_impl_t	*mcip = (mac_client_impl_t *)mch;
3603 
3604 	mcip->mci_upper_mip = (mac_impl_t *)mh;
3605 }
3606 
3607 /*
3608  * Mark the mac as being used exclusively by the single mac client that is
3609  * doing some control operation on this mac. No further opens of this mac
3610  * will be allowed until this client calls mac_unmark_exclusive. The mac
3611  * client calling this function must already be in the mac perimeter
3612  */
3613 int
3614 mac_mark_exclusive(mac_handle_t mh)
3615 {
3616 	mac_impl_t	*mip = (mac_impl_t *)mh;
3617 
3618 	ASSERT(MAC_PERIM_HELD(mh));
3619 	/*
3620 	 * Look up its entry in the global hash table.
3621 	 */
3622 	rw_enter(&i_mac_impl_lock, RW_WRITER);
3623 	if (mip->mi_state_flags & MIS_DISABLED) {
3624 		rw_exit(&i_mac_impl_lock);
3625 		return (ENOENT);
3626 	}
3627 
3628 	/*
3629 	 * A reference to mac is held even if the link is not plumbed.
3630 	 * In i_dls_link_create() we open the MAC interface and hold the
3631 	 * reference. There is an additional reference for the mac_open
3632 	 * done in acquiring the mac perimeter
3633 	 */
3634 	if (mip->mi_ref != 2) {
3635 		rw_exit(&i_mac_impl_lock);
3636 		return (EBUSY);
3637 	}
3638 
3639 	ASSERT(!(mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
3640 	mip->mi_state_flags |= MIS_EXCLUSIVE_HELD;
3641 	rw_exit(&i_mac_impl_lock);
3642 	return (0);
3643 }
3644 
3645 void
3646 mac_unmark_exclusive(mac_handle_t mh)
3647 {
3648 	mac_impl_t	*mip = (mac_impl_t *)mh;
3649 
3650 	ASSERT(MAC_PERIM_HELD(mh));
3651 
3652 	rw_enter(&i_mac_impl_lock, RW_WRITER);
3653 	/* 1 for the creation and another for the perimeter */
3654 	ASSERT(mip->mi_ref == 2 && (mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
3655 	mip->mi_state_flags &= ~MIS_EXCLUSIVE_HELD;
3656 	rw_exit(&i_mac_impl_lock);
3657 }
3658 
3659 /*
3660  * Set the MTU for the specified device. The function returns EBUSY if
3661  * another MAC client prevents the caller to become the exclusive client.
3662  * Returns EAGAIN if the client is started.
3663  */
3664 int
3665 mac_set_mtu(mac_handle_t mh, uint_t new_mtu, uint_t *old_mtu_arg)
3666 {
3667 	mac_impl_t *mip = (mac_impl_t *)mh;
3668 	uint_t old_mtu;
3669 	int rv;
3670 	boolean_t exclusive = B_FALSE;
3671 
3672 	i_mac_perim_enter(mip);
3673 
3674 	if ((mip->mi_callbacks->mc_callbacks & MC_SETPROP) == 0 ||
3675 	    (mip->mi_callbacks->mc_callbacks & MC_GETPROP) == 0) {
3676 		rv = ENOTSUP;
3677 		goto bail;
3678 	}
3679 
3680 	if ((rv = mac_mark_exclusive(mh)) != 0)
3681 		goto bail;
3682 	exclusive = B_TRUE;
3683 
3684 	if (mip->mi_active > 0) {
3685 		/*
3686 		 * The MAC instance is started, for example due to the
3687 		 * presence of a promiscuous clients. Fail the operation
3688 		 * since the MAC's MTU cannot be changed while the NIC
3689 		 * is started.
3690 		 */
3691 		rv = EAGAIN;
3692 		goto bail;
3693 	}
3694 
3695 	mac_sdu_get(mh, NULL, &old_mtu);
3696 
3697 	if (old_mtu != new_mtu) {
3698 		rv = mip->mi_callbacks->mc_setprop(mip->mi_driver,
3699 		    "mtu", MAC_PROP_MTU, sizeof (uint_t), &new_mtu);
3700 	}
3701 
3702 bail:
3703 	if (exclusive)
3704 		mac_unmark_exclusive(mh);
3705 	i_mac_perim_exit(mip);
3706 
3707 	if (rv == 0 && old_mtu_arg != NULL)
3708 		*old_mtu_arg = old_mtu;
3709 	return (rv);
3710 }
3711 
3712 void
3713 mac_get_hwgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
3714     uint_t *n_rings, uint_t *type, uint_t *n_clnts, char *clnts_name)
3715 {
3716 	mac_impl_t *mip = (mac_impl_t *)mh;
3717 	mac_grp_client_t *mcip;
3718 	uint_t i = 0, index = 0;
3719 
3720 	/* Revisit when we implement fully dynamic group allocation */
3721 	ASSERT(grp_index >= 0 && grp_index < mip->mi_rx_group_count);
3722 
3723 	rw_enter(&mip->mi_rw_lock, RW_READER);
3724 	*grp_num = mip->mi_rx_groups[grp_index].mrg_index;
3725 	*type = mip->mi_rx_groups[grp_index].mrg_type;
3726 	*n_rings = mip->mi_rx_groups[grp_index].mrg_cur_count;
3727 	for (mcip = mip->mi_rx_groups[grp_index].mrg_clients; mcip != NULL;
3728 	    mcip = mcip->mgc_next) {
3729 		int name_len = strlen(mcip->mgc_client->mci_name);
3730 
3731 		/*
3732 		 * MAXCLIENTNAMELEN is the buffer size reserved for client
3733 		 * names.
3734 		 * XXXX Formating the client name string needs to be moved
3735 		 * to user land when fixing the size of dhi_clnts in
3736 		 * dld_hwgrpinfo_t. We should use n_clients * client_name for
3737 		 * dhi_clntsin instead of MAXCLIENTNAMELEN
3738 		 */
3739 		if (index + name_len >= MAXCLIENTNAMELEN) {
3740 			index = MAXCLIENTNAMELEN;
3741 			break;
3742 		}
3743 		bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
3744 		    name_len);
3745 		index += name_len;
3746 		clnts_name[index++] = ',';
3747 		i++;
3748 	}
3749 
3750 	/* Get rid of the last , */
3751 	if (index > 0)
3752 		clnts_name[index - 1] = '\0';
3753 	*n_clnts = i;
3754 	rw_exit(&mip->mi_rw_lock);
3755 }
3756 
3757 uint_t
3758 mac_hwgrp_num(mac_handle_t mh)
3759 {
3760 	mac_impl_t *mip = (mac_impl_t *)mh;
3761 
3762 	return (mip->mi_rx_group_count);
3763 }
3764