1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 */
25
26 /*
27 * - General Introduction:
28 *
29 * This file contains the implementation of the MAC client kernel
30 * API and related code. The MAC client API allows a kernel module
31 * to gain access to a MAC instance (physical NIC, link aggregation, etc).
32 * It allows a MAC client to associate itself with a MAC address,
33 * VLANs, callback functions for data traffic and for promiscuous mode.
34 * The MAC client API is also used to specify the properties associated
35 * with a MAC client, such as bandwidth limits, priority, CPUS, etc.
36 * These properties are further used to determine the hardware resources
37 * to allocate to the various MAC clients.
38 *
39 * - Primary MAC clients:
40 *
41 * The MAC client API refers to "primary MAC clients". A primary MAC
42 * client is a client which "owns" the primary MAC address of
43 * the underlying MAC instance. The primary MAC address is called out
44 * since it is associated with specific semantics: the primary MAC
45 * address is the MAC address which is assigned to the IP interface
46 * when it is plumbed, and the primary MAC address is assigned
47 * to VLAN data-links. The primary address of a MAC instance can
48 * also change dynamically from under the MAC client, for example
49 * as a result of a change of state of a link aggregation. In that
50 * case the MAC layer automatically updates all data-structures which
51 * refer to the current value of the primary MAC address. Typical
52 * primary MAC clients are dls, aggr, and xnb. A typical non-primary
53 * MAC client is the vnic driver.
54 *
55 * - Virtual Switching:
56 *
57 * The MAC layer implements a virtual switch between the MAC clients
58 * (primary and non-primary) defined on top of the same underlying
59 * NIC (physical, link aggregation, etc). The virtual switch is
60 * VLAN-aware, i.e. it allows multiple MAC clients to be member
61 * of one or more VLANs, and the virtual switch will distribute
62 * multicast tagged packets only to the member of the corresponding
63 * VLANs.
64 *
65 * - Upper vs Lower MAC:
66 *
67 * Creating a VNIC on top of a MAC instance effectively causes
68 * two MAC instances to be layered on top of each other, one for
69 * the VNIC(s), one for the underlying MAC instance (physical NIC,
70 * link aggregation, etc). In the code below we refer to the
71 * underlying NIC as the "lower MAC", and we refer to VNICs as
72 * the "upper MAC".
73 *
74 * - Pass-through for VNICs:
75 *
76 * When VNICs are created on top of an underlying MAC, this causes
77 * a layering of two MAC instances. Since the lower MAC already
78 * does the switching and demultiplexing to its MAC clients, the
79 * upper MAC would simply have to pass packets to the layer below
80 * or above it, which would introduce overhead. In order to avoid
81 * this overhead, the MAC layer implements a pass-through mechanism
82 * for VNICs. When a VNIC opens the lower MAC instance, it saves
83 * the MAC client handle it optains from the MAC layer. When a MAC
84 * client opens a VNIC (upper MAC), the MAC layer detects that
85 * the MAC being opened is a VNIC, and gets the MAC client handle
86 * that the VNIC driver obtained from the lower MAC. This exchange
87 * is doing through a private capability between the MAC layer
88 * and the VNIC driver. The upper MAC then returns that handle
89 * directly to its MAC client. Any operation done by the upper
90 * MAC client is now done on the lower MAC client handle, which
91 * allows the VNIC driver to be completely bypassed for the
92 * performance sensitive data-path.
93 *
94 */
95
96 #include <sys/types.h>
97 #include <sys/conf.h>
98 #include <sys/id_space.h>
99 #include <sys/esunddi.h>
100 #include <sys/stat.h>
101 #include <sys/mkdev.h>
102 #include <sys/stream.h>
103 #include <sys/strsun.h>
104 #include <sys/strsubr.h>
105 #include <sys/dlpi.h>
106 #include <sys/modhash.h>
107 #include <sys/mac_impl.h>
108 #include <sys/mac_client_impl.h>
109 #include <sys/mac_soft_ring.h>
110 #include <sys/mac_stat.h>
111 #include <sys/dls.h>
112 #include <sys/dld.h>
113 #include <sys/modctl.h>
114 #include <sys/fs/dv_node.h>
115 #include <sys/thread.h>
116 #include <sys/proc.h>
117 #include <sys/callb.h>
118 #include <sys/cpuvar.h>
119 #include <sys/atomic.h>
120 #include <sys/sdt.h>
121 #include <sys/mac_flow.h>
122 #include <sys/ddi_intr_impl.h>
123 #include <sys/disp.h>
124 #include <sys/sdt.h>
125 #include <sys/vnic.h>
126 #include <sys/vnic_impl.h>
127 #include <sys/vlan.h>
128 #include <inet/ip.h>
129 #include <inet/ip6.h>
130 #include <sys/exacct.h>
131 #include <sys/exacct_impl.h>
132 #include <inet/nd.h>
133 #include <sys/ethernet.h>
134
135 kmem_cache_t *mac_client_impl_cache;
136 kmem_cache_t *mac_promisc_impl_cache;
137
138 static boolean_t mac_client_single_rcvr(mac_client_impl_t *);
139 static flow_entry_t *mac_client_swap_mciflent(mac_client_impl_t *);
140 static flow_entry_t *mac_client_get_flow(mac_client_impl_t *,
141 mac_unicast_impl_t *);
142 static void mac_client_remove_flow_from_list(mac_client_impl_t *,
143 flow_entry_t *);
144 static void mac_client_add_to_flow_list(mac_client_impl_t *, flow_entry_t *);
145 static void mac_rename_flow_names(mac_client_impl_t *, const char *);
146 static void mac_virtual_link_update(mac_impl_t *);
147 static int mac_client_datapath_setup(mac_client_impl_t *, uint16_t,
148 uint8_t *, mac_resource_props_t *, boolean_t, mac_unicast_impl_t *);
149 static void mac_client_datapath_teardown(mac_client_handle_t,
150 mac_unicast_impl_t *, flow_entry_t *);
151
152 /* ARGSUSED */
153 static int
154 i_mac_client_impl_ctor(void *buf, void *arg, int kmflag)
155 {
156 int i;
157 mac_client_impl_t *mcip = buf;
158
159 bzero(buf, MAC_CLIENT_IMPL_SIZE);
160 mutex_init(&mcip->mci_tx_cb_lock, NULL, MUTEX_DRIVER, NULL);
161 mcip->mci_tx_notify_cb_info.mcbi_lockp = &mcip->mci_tx_cb_lock;
162
163 ASSERT(mac_tx_percpu_cnt >= 0);
164 for (i = 0; i <= mac_tx_percpu_cnt; i++) {
165 mutex_init(&mcip->mci_tx_pcpu[i].pcpu_tx_lock, NULL,
166 MUTEX_DRIVER, NULL);
167 }
168 cv_init(&mcip->mci_tx_cv, NULL, CV_DRIVER, NULL);
169
170 return (0);
171 }
172
173 /* ARGSUSED */
174 static void
175 i_mac_client_impl_dtor(void *buf, void *arg)
176 {
177 int i;
178 mac_client_impl_t *mcip = buf;
179
180 ASSERT(mcip->mci_promisc_list == NULL);
181 ASSERT(mcip->mci_unicast_list == NULL);
182 ASSERT(mcip->mci_state_flags == 0);
183 ASSERT(mcip->mci_tx_flag == 0);
184
185 mutex_destroy(&mcip->mci_tx_cb_lock);
186
187 ASSERT(mac_tx_percpu_cnt >= 0);
188 for (i = 0; i <= mac_tx_percpu_cnt; i++) {
189 ASSERT(mcip->mci_tx_pcpu[i].pcpu_tx_refcnt == 0);
190 mutex_destroy(&mcip->mci_tx_pcpu[i].pcpu_tx_lock);
191 }
192 cv_destroy(&mcip->mci_tx_cv);
193 }
194
195 /* ARGSUSED */
196 static int
197 i_mac_promisc_impl_ctor(void *buf, void *arg, int kmflag)
198 {
199 mac_promisc_impl_t *mpip = buf;
200
201 bzero(buf, sizeof (mac_promisc_impl_t));
202 mpip->mpi_mci_link.mcb_objp = buf;
203 mpip->mpi_mci_link.mcb_objsize = sizeof (mac_promisc_impl_t);
204 mpip->mpi_mi_link.mcb_objp = buf;
205 mpip->mpi_mi_link.mcb_objsize = sizeof (mac_promisc_impl_t);
206 return (0);
207 }
208
209 /* ARGSUSED */
210 static void
211 i_mac_promisc_impl_dtor(void *buf, void *arg)
212 {
213 mac_promisc_impl_t *mpip = buf;
214
215 ASSERT(mpip->mpi_mci_link.mcb_objp != NULL);
216 ASSERT(mpip->mpi_mci_link.mcb_objsize == sizeof (mac_promisc_impl_t));
217 ASSERT(mpip->mpi_mi_link.mcb_objp == mpip->mpi_mci_link.mcb_objp);
218 ASSERT(mpip->mpi_mi_link.mcb_objsize == sizeof (mac_promisc_impl_t));
219
220 mpip->mpi_mci_link.mcb_objp = NULL;
221 mpip->mpi_mci_link.mcb_objsize = 0;
222 mpip->mpi_mi_link.mcb_objp = NULL;
223 mpip->mpi_mi_link.mcb_objsize = 0;
224
225 ASSERT(mpip->mpi_mci_link.mcb_flags == 0);
226 mpip->mpi_mci_link.mcb_objsize = 0;
227 }
228
229 void
230 mac_client_init(void)
231 {
232 ASSERT(mac_tx_percpu_cnt >= 0);
233
234 mac_client_impl_cache = kmem_cache_create("mac_client_impl_cache",
235 MAC_CLIENT_IMPL_SIZE, 0, i_mac_client_impl_ctor,
236 i_mac_client_impl_dtor, NULL, NULL, NULL, 0);
237 ASSERT(mac_client_impl_cache != NULL);
238
239 mac_promisc_impl_cache = kmem_cache_create("mac_promisc_impl_cache",
240 sizeof (mac_promisc_impl_t), 0, i_mac_promisc_impl_ctor,
241 i_mac_promisc_impl_dtor, NULL, NULL, NULL, 0);
242 ASSERT(mac_promisc_impl_cache != NULL);
243 }
244
245 void
246 mac_client_fini(void)
247 {
248 kmem_cache_destroy(mac_client_impl_cache);
249 kmem_cache_destroy(mac_promisc_impl_cache);
250 }
251
252 /*
253 * Return the lower MAC client handle from the VNIC driver for the
254 * specified VNIC MAC instance.
255 */
256 mac_client_impl_t *
257 mac_vnic_lower(mac_impl_t *mip)
258 {
259 mac_capab_vnic_t cap;
260 mac_client_impl_t *mcip;
261
262 VERIFY(i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_VNIC, &cap));
263 mcip = cap.mcv_mac_client_handle(cap.mcv_arg);
264
265 return (mcip);
266 }
267
268 /*
269 * Return the MAC client handle of the primary MAC client for the
270 * specified MAC instance, or NULL otherwise.
271 */
272 mac_client_impl_t *
273 mac_primary_client_handle(mac_impl_t *mip)
274 {
275 mac_client_impl_t *mcip;
276
277 if (mip->mi_state_flags & MIS_IS_VNIC)
278 return (mac_vnic_lower(mip));
279
280 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
281
282 for (mcip = mip->mi_clients_list; mcip != NULL;
283 mcip = mcip->mci_client_next) {
284 if (MCIP_DATAPATH_SETUP(mcip) && mac_is_primary_client(mcip))
285 return (mcip);
286 }
287 return (NULL);
288 }
289
290 /*
291 * Open a MAC specified by its MAC name.
292 */
293 int
294 mac_open(const char *macname, mac_handle_t *mhp)
295 {
296 mac_impl_t *mip;
297 int err;
298
299 /*
300 * Look up its entry in the global hash table.
301 */
302 if ((err = mac_hold(macname, &mip)) != 0)
303 return (err);
304
305 /*
306 * Hold the dip associated to the MAC to prevent it from being
307 * detached. For a softmac, its underlying dip is held by the
308 * mi_open() callback.
309 *
310 * This is done to be more tolerant with some defective drivers,
311 * which incorrectly handle mac_unregister() failure in their
312 * xxx_detach() routine. For example, some drivers ignore the
313 * failure of mac_unregister() and free all resources that
314 * that are needed for data transmition.
315 */
316 e_ddi_hold_devi(mip->mi_dip);
317
318 if (!(mip->mi_callbacks->mc_callbacks & MC_OPEN)) {
319 *mhp = (mac_handle_t)mip;
320 return (0);
321 }
322
323 /*
324 * The mac perimeter is used in both mac_open and mac_close by the
325 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
326 */
327 i_mac_perim_enter(mip);
328 mip->mi_oref++;
329 if (mip->mi_oref != 1 || ((err = mip->mi_open(mip->mi_driver)) == 0)) {
330 *mhp = (mac_handle_t)mip;
331 i_mac_perim_exit(mip);
332 return (0);
333 }
334 mip->mi_oref--;
335 ddi_release_devi(mip->mi_dip);
336 mac_rele(mip);
337 i_mac_perim_exit(mip);
338 return (err);
339 }
340
341 /*
342 * Open a MAC specified by its linkid.
343 */
344 int
345 mac_open_by_linkid(datalink_id_t linkid, mac_handle_t *mhp)
346 {
347 dls_dl_handle_t dlh;
348 int err;
349
350 if ((err = dls_devnet_hold_tmp(linkid, &dlh)) != 0)
351 return (err);
352
353 dls_devnet_prop_task_wait(dlh);
354
355 err = mac_open(dls_devnet_mac(dlh), mhp);
356
357 dls_devnet_rele_tmp(dlh);
358 return (err);
359 }
360
361 /*
362 * Open a MAC specified by its link name.
363 */
364 int
365 mac_open_by_linkname(const char *link, mac_handle_t *mhp)
366 {
367 datalink_id_t linkid;
368 int err;
369
370 if ((err = dls_mgmt_get_linkid(link, &linkid)) != 0)
371 return (err);
372 return (mac_open_by_linkid(linkid, mhp));
373 }
374
375 /*
376 * Close the specified MAC.
377 */
378 void
379 mac_close(mac_handle_t mh)
380 {
381 mac_impl_t *mip = (mac_impl_t *)mh;
382
383 i_mac_perim_enter(mip);
384 /*
385 * The mac perimeter is used in both mac_open and mac_close by the
386 * framework to single thread the MC_OPEN/MC_CLOSE of drivers.
387 */
388 if (mip->mi_callbacks->mc_callbacks & MC_OPEN) {
389 ASSERT(mip->mi_oref != 0);
390 if (--mip->mi_oref == 0) {
391 if ((mip->mi_callbacks->mc_callbacks & MC_CLOSE))
392 mip->mi_close(mip->mi_driver);
393 }
394 }
395 i_mac_perim_exit(mip);
396 ddi_release_devi(mip->mi_dip);
397 mac_rele(mip);
398 }
399
400 /*
401 * Misc utility functions to retrieve various information about a MAC
402 * instance or a MAC client.
403 */
404
405 const mac_info_t *
406 mac_info(mac_handle_t mh)
407 {
408 return (&((mac_impl_t *)mh)->mi_info);
409 }
410
411 dev_info_t *
412 mac_devinfo_get(mac_handle_t mh)
413 {
414 return (((mac_impl_t *)mh)->mi_dip);
415 }
416
417 void *
418 mac_driver(mac_handle_t mh)
419 {
420 return (((mac_impl_t *)mh)->mi_driver);
421 }
422
423 const char *
424 mac_name(mac_handle_t mh)
425 {
426 return (((mac_impl_t *)mh)->mi_name);
427 }
428
429 int
430 mac_type(mac_handle_t mh)
431 {
432 return (((mac_impl_t *)mh)->mi_type->mt_type);
433 }
434
435 int
436 mac_nativetype(mac_handle_t mh)
437 {
438 return (((mac_impl_t *)mh)->mi_type->mt_nativetype);
439 }
440
441 char *
442 mac_client_name(mac_client_handle_t mch)
443 {
444 return (((mac_client_impl_t *)mch)->mci_name);
445 }
446
447 minor_t
448 mac_minor(mac_handle_t mh)
449 {
450 return (((mac_impl_t *)mh)->mi_minor);
451 }
452
453 /*
454 * Return the VID associated with a MAC client. This function should
455 * be called for clients which are associated with only one VID.
456 */
457 uint16_t
458 mac_client_vid(mac_client_handle_t mch)
459 {
460 uint16_t vid = VLAN_ID_NONE;
461 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
462 flow_desc_t flow_desc;
463
464 if (mcip->mci_nflents == 0)
465 return (vid);
466
467 ASSERT(MCIP_DATAPATH_SETUP(mcip) && mac_client_single_rcvr(mcip));
468
469 mac_flow_get_desc(mcip->mci_flent, &flow_desc);
470 if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
471 vid = flow_desc.fd_vid;
472
473 return (vid);
474 }
475
476 /*
477 * Return whether the specified MAC client corresponds to a VLAN VNIC.
478 */
479 boolean_t
480 mac_client_is_vlan_vnic(mac_client_handle_t mch)
481 {
482 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
483
484 return (((mcip->mci_state_flags & MCIS_IS_VNIC) != 0) &&
485 ((mcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) != 0));
486 }
487
488 /*
489 * Return the link speed associated with the specified MAC client.
490 *
491 * The link speed of a MAC client is equal to the smallest value of
492 * 1) the current link speed of the underlying NIC, or
493 * 2) the bandwidth limit set for the MAC client.
494 *
495 * Note that the bandwidth limit can be higher than the speed
496 * of the underlying NIC. This is allowed to avoid spurious
497 * administration action failures or artifically lowering the
498 * bandwidth limit of a link that may have temporarily lowered
499 * its link speed due to hardware problem or administrator action.
500 */
501 static uint64_t
502 mac_client_ifspeed(mac_client_impl_t *mcip)
503 {
504 mac_impl_t *mip = mcip->mci_mip;
505 uint64_t nic_speed;
506
507 nic_speed = mac_stat_get((mac_handle_t)mip, MAC_STAT_IFSPEED);
508
509 if (nic_speed == 0) {
510 return (0);
511 } else {
512 uint64_t policy_limit = (uint64_t)-1;
513
514 if (MCIP_RESOURCE_PROPS_MASK(mcip) & MRP_MAXBW)
515 policy_limit = MCIP_RESOURCE_PROPS_MAXBW(mcip);
516
517 return (MIN(policy_limit, nic_speed));
518 }
519 }
520
521 /*
522 * Return the link state of the specified client. If here are more
523 * than one clients of the underying mac_impl_t, the link state
524 * will always be UP regardless of the link state of the underlying
525 * mac_impl_t. This is needed to allow the MAC clients to continue
526 * to communicate with each other even when the physical link of
527 * their mac_impl_t is down.
528 */
529 static uint64_t
530 mac_client_link_state(mac_client_impl_t *mcip)
531 {
532 mac_impl_t *mip = mcip->mci_mip;
533 uint16_t vid;
534 mac_client_impl_t *mci_list;
535 mac_unicast_impl_t *mui_list, *oth_mui_list;
536
537 /*
538 * Returns LINK_STATE_UP if there are other MAC clients defined on
539 * mac_impl_t which share same VLAN ID as that of mcip. Note that
540 * if 'mcip' has more than one VID's then we match ANY one of the
541 * VID's with other MAC client's VID's and return LINK_STATE_UP.
542 */
543 rw_enter(&mcip->mci_rw_lock, RW_READER);
544 for (mui_list = mcip->mci_unicast_list; mui_list != NULL;
545 mui_list = mui_list->mui_next) {
546 vid = mui_list->mui_vid;
547 for (mci_list = mip->mi_clients_list; mci_list != NULL;
548 mci_list = mci_list->mci_client_next) {
549 if (mci_list == mcip)
550 continue;
551 for (oth_mui_list = mci_list->mci_unicast_list;
552 oth_mui_list != NULL; oth_mui_list = oth_mui_list->
553 mui_next) {
554 if (vid == oth_mui_list->mui_vid) {
555 rw_exit(&mcip->mci_rw_lock);
556 return (LINK_STATE_UP);
557 }
558 }
559 }
560 }
561 rw_exit(&mcip->mci_rw_lock);
562
563 return (mac_stat_get((mac_handle_t)mip, MAC_STAT_LINK_STATE));
564 }
565
566 /*
567 * These statistics are consumed by dladm show-link -s <vnic>,
568 * dladm show-vnic -s and netstat. With the introduction of dlstat,
569 * dladm show-link -s and dladm show-vnic -s witll be EOL'ed while
570 * netstat will consume from kstats introduced for dlstat. This code
571 * will be removed at that time.
572 */
573
574 /*
575 * Return the statistics of a MAC client. These statistics are different
576 * then the statistics of the underlying MAC which are returned by
577 * mac_stat_get().
578 */
579 uint64_t
580 mac_client_stat_get(mac_client_handle_t mch, uint_t stat)
581 {
582 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
583 mac_impl_t *mip = mcip->mci_mip;
584 flow_entry_t *flent = mcip->mci_flent;
585 mac_soft_ring_set_t *mac_srs;
586 mac_rx_stats_t *mac_rx_stat;
587 mac_tx_stats_t *mac_tx_stat;
588 int i;
589 uint64_t val = 0;
590
591 mac_srs = (mac_soft_ring_set_t *)(flent->fe_tx_srs);
592 mac_tx_stat = &mac_srs->srs_tx.st_stat;
593
594 switch (stat) {
595 case MAC_STAT_LINK_STATE:
596 val = mac_client_link_state(mcip);
597 break;
598 case MAC_STAT_LINK_UP:
599 val = (mac_client_link_state(mcip) == LINK_STATE_UP);
600 break;
601 case MAC_STAT_PROMISC:
602 val = mac_stat_get((mac_handle_t)mip, MAC_STAT_PROMISC);
603 break;
604 case MAC_STAT_LOWLINK_STATE:
605 val = mac_stat_get((mac_handle_t)mip, MAC_STAT_LOWLINK_STATE);
606 break;
607 case MAC_STAT_IFSPEED:
608 val = mac_client_ifspeed(mcip);
609 break;
610 case MAC_STAT_MULTIRCV:
611 val = mcip->mci_misc_stat.mms_multircv;
612 break;
613 case MAC_STAT_BRDCSTRCV:
614 val = mcip->mci_misc_stat.mms_brdcstrcv;
615 break;
616 case MAC_STAT_MULTIXMT:
617 val = mcip->mci_misc_stat.mms_multixmt;
618 break;
619 case MAC_STAT_BRDCSTXMT:
620 val = mcip->mci_misc_stat.mms_brdcstxmt;
621 break;
622 case MAC_STAT_OBYTES:
623 val = mac_tx_stat->mts_obytes;
624 break;
625 case MAC_STAT_OPACKETS:
626 val = mac_tx_stat->mts_opackets;
627 break;
628 case MAC_STAT_OERRORS:
629 val = mac_tx_stat->mts_oerrors;
630 break;
631 case MAC_STAT_IPACKETS:
632 for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
633 mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
634 mac_rx_stat = &mac_srs->srs_rx.sr_stat;
635 val += mac_rx_stat->mrs_intrcnt +
636 mac_rx_stat->mrs_pollcnt + mac_rx_stat->mrs_lclcnt;
637 }
638 break;
639 case MAC_STAT_RBYTES:
640 for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
641 mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
642 mac_rx_stat = &mac_srs->srs_rx.sr_stat;
643 val += mac_rx_stat->mrs_intrbytes +
644 mac_rx_stat->mrs_pollbytes +
645 mac_rx_stat->mrs_lclbytes;
646 }
647 break;
648 case MAC_STAT_IERRORS:
649 for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
650 mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
651 mac_rx_stat = &mac_srs->srs_rx.sr_stat;
652 val += mac_rx_stat->mrs_ierrors;
653 }
654 break;
655 default:
656 val = mac_driver_stat_default(mip, stat);
657 break;
658 }
659
660 return (val);
661 }
662
663 /*
664 * Return the statistics of the specified MAC instance.
665 */
666 uint64_t
667 mac_stat_get(mac_handle_t mh, uint_t stat)
668 {
669 mac_impl_t *mip = (mac_impl_t *)mh;
670 uint64_t val;
671 int ret;
672
673 /*
674 * The range of stat determines where it is maintained. Stat
675 * values from 0 up to (but not including) MAC_STAT_MIN are
676 * mainteined by the mac module itself. Everything else is
677 * maintained by the driver.
678 *
679 * If the mac_impl_t being queried corresponds to a VNIC,
680 * the stats need to be queried from the lower MAC client
681 * corresponding to the VNIC. (The mac_link_update()
682 * invoked by the driver to the lower MAC causes the *lower
683 * MAC* to update its mi_linkstate, and send a notification
684 * to its MAC clients. Due to the VNIC passthrough,
685 * these notifications are sent to the upper MAC clients
686 * of the VNIC directly, and the upper mac_impl_t of the VNIC
687 * does not have a valid mi_linkstate.
688 */
689 if (stat < MAC_STAT_MIN && !(mip->mi_state_flags & MIS_IS_VNIC)) {
690 /* these stats are maintained by the mac module itself */
691 switch (stat) {
692 case MAC_STAT_LINK_STATE:
693 return (mip->mi_linkstate);
694 case MAC_STAT_LINK_UP:
695 return (mip->mi_linkstate == LINK_STATE_UP);
696 case MAC_STAT_PROMISC:
697 return (mip->mi_devpromisc != 0);
698 case MAC_STAT_LOWLINK_STATE:
699 return (mip->mi_lowlinkstate);
700 default:
701 ASSERT(B_FALSE);
702 }
703 }
704
705 /*
706 * Call the driver to get the given statistic.
707 */
708 ret = mip->mi_getstat(mip->mi_driver, stat, &val);
709 if (ret != 0) {
710 /*
711 * The driver doesn't support this statistic. Get the
712 * statistic's default value.
713 */
714 val = mac_driver_stat_default(mip, stat);
715 }
716 return (val);
717 }
718
719 /*
720 * Query hardware rx ring corresponding to the pseudo ring.
721 */
722 uint64_t
723 mac_pseudo_rx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
724 {
725 return (mac_rx_ring_stat_get(handle, stat));
726 }
727
728 /*
729 * Query hardware tx ring corresponding to the pseudo ring.
730 */
731 uint64_t
732 mac_pseudo_tx_ring_stat_get(mac_ring_handle_t handle, uint_t stat)
733 {
734 return (mac_tx_ring_stat_get(handle, stat));
735 }
736
737 /*
738 * Utility function which returns the VID associated with a flow entry.
739 */
740 uint16_t
741 i_mac_flow_vid(flow_entry_t *flent)
742 {
743 flow_desc_t flow_desc;
744
745 mac_flow_get_desc(flent, &flow_desc);
746
747 if ((flow_desc.fd_mask & FLOW_LINK_VID) != 0)
748 return (flow_desc.fd_vid);
749 return (VLAN_ID_NONE);
750 }
751
752 /*
753 * Verify the validity of the specified unicast MAC address. Returns B_TRUE
754 * if the address is valid, B_FALSE otherwise (multicast address, or incorrect
755 * length.
756 */
757 boolean_t
758 mac_unicst_verify(mac_handle_t mh, const uint8_t *addr, uint_t len)
759 {
760 mac_impl_t *mip = (mac_impl_t *)mh;
761
762 /*
763 * Verify the address. No lock is needed since mi_type and plugin
764 * details don't change after mac_register().
765 */
766 if ((len != mip->mi_type->mt_addr_length) ||
767 (mip->mi_type->mt_ops.mtops_unicst_verify(addr,
768 mip->mi_pdata)) != 0) {
769 return (B_FALSE);
770 } else {
771 return (B_TRUE);
772 }
773 }
774
775 void
776 mac_sdu_get(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu)
777 {
778 mac_impl_t *mip = (mac_impl_t *)mh;
779
780 if (min_sdu != NULL)
781 *min_sdu = mip->mi_sdu_min;
782 if (max_sdu != NULL)
783 *max_sdu = mip->mi_sdu_max;
784 }
785
786 void
787 mac_sdu_get2(mac_handle_t mh, uint_t *min_sdu, uint_t *max_sdu,
788 uint_t *multicast_sdu)
789 {
790 mac_impl_t *mip = (mac_impl_t *)mh;
791
792 if (min_sdu != NULL)
793 *min_sdu = mip->mi_sdu_min;
794 if (max_sdu != NULL)
795 *max_sdu = mip->mi_sdu_max;
796 if (multicast_sdu != NULL)
797 *multicast_sdu = mip->mi_sdu_multicast;
798 }
799
800 /*
801 * Update the MAC unicast address of the specified client's flows. Currently
802 * only one unicast MAC unicast address is allowed per client.
803 */
804 static void
805 mac_unicast_update_client_flow(mac_client_impl_t *mcip)
806 {
807 mac_impl_t *mip = mcip->mci_mip;
808 flow_entry_t *flent = mcip->mci_flent;
809 mac_address_t *map = mcip->mci_unicast;
810 flow_desc_t flow_desc;
811
812 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
813 ASSERT(flent != NULL);
814
815 mac_flow_get_desc(flent, &flow_desc);
816 ASSERT(flow_desc.fd_mask & FLOW_LINK_DST);
817
818 bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
819 mac_flow_set_desc(flent, &flow_desc);
820
821 /*
822 * The v6 local addr (used by mac protection) needs to be
823 * regenerated because our mac address has changed.
824 */
825 mac_protect_update_v6_local_addr(mcip);
826
827 /*
828 * A MAC client could have one MAC address but multiple
829 * VLANs. In that case update the flow entries corresponding
830 * to all VLANs of the MAC client.
831 */
832 for (flent = mcip->mci_flent_list; flent != NULL;
833 flent = flent->fe_client_next) {
834 mac_flow_get_desc(flent, &flow_desc);
835 if (!(flent->fe_type & FLOW_PRIMARY_MAC ||
836 flent->fe_type & FLOW_VNIC_MAC))
837 continue;
838
839 bcopy(map->ma_addr, flow_desc.fd_dst_mac, map->ma_len);
840 mac_flow_set_desc(flent, &flow_desc);
841 }
842 }
843
844 /*
845 * Update all clients that share the same unicast address.
846 */
847 void
848 mac_unicast_update_clients(mac_impl_t *mip, mac_address_t *map)
849 {
850 mac_client_impl_t *mcip;
851
852 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
853
854 /*
855 * Find all clients that share the same unicast MAC address and update
856 * them appropriately.
857 */
858 for (mcip = mip->mi_clients_list; mcip != NULL;
859 mcip = mcip->mci_client_next) {
860 /*
861 * Ignore clients that don't share this MAC address.
862 */
863 if (map != mcip->mci_unicast)
864 continue;
865
866 /*
867 * Update those clients with same old unicast MAC address.
868 */
869 mac_unicast_update_client_flow(mcip);
870 }
871 }
872
873 /*
874 * Update the unicast MAC address of the specified VNIC MAC client.
875 *
876 * Check whether the operation is valid. Any of following cases should fail:
877 *
878 * 1. It's a VLAN type of VNIC.
879 * 2. The new value is current "primary" MAC address.
880 * 3. The current MAC address is shared with other clients.
881 * 4. The new MAC address has been used. This case will be valid when
882 * client migration is fully supported.
883 */
884 int
885 mac_vnic_unicast_set(mac_client_handle_t mch, const uint8_t *addr)
886 {
887 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
888 mac_impl_t *mip = mcip->mci_mip;
889 mac_address_t *map = mcip->mci_unicast;
890 int err;
891
892 ASSERT(!(mip->mi_state_flags & MIS_IS_VNIC));
893 ASSERT(mcip->mci_state_flags & MCIS_IS_VNIC);
894 ASSERT(mcip->mci_flags != MAC_CLIENT_FLAGS_PRIMARY);
895
896 i_mac_perim_enter(mip);
897
898 /*
899 * If this is a VLAN type of VNIC, it's using "primary" MAC address
900 * of the underlying interface. Must fail here. Refer to case 1 above.
901 */
902 if (bcmp(map->ma_addr, mip->mi_addr, map->ma_len) == 0) {
903 i_mac_perim_exit(mip);
904 return (ENOTSUP);
905 }
906
907 /*
908 * If the new address is the "primary" one, must fail. Refer to
909 * case 2 above.
910 */
911 if (bcmp(addr, mip->mi_addr, map->ma_len) == 0) {
912 i_mac_perim_exit(mip);
913 return (EACCES);
914 }
915
916 /*
917 * If the address is shared by multiple clients, must fail. Refer
918 * to case 3 above.
919 */
920 if (mac_check_macaddr_shared(map)) {
921 i_mac_perim_exit(mip);
922 return (EBUSY);
923 }
924
925 /*
926 * If the new address has been used, must fail for now. Refer to
927 * case 4 above.
928 */
929 if (mac_find_macaddr(mip, (uint8_t *)addr) != NULL) {
930 i_mac_perim_exit(mip);
931 return (ENOTSUP);
932 }
933
934 /*
935 * Update the MAC address.
936 */
937 err = mac_update_macaddr(map, (uint8_t *)addr);
938
939 if (err != 0) {
940 i_mac_perim_exit(mip);
941 return (err);
942 }
943
944 /*
945 * Update all flows of this MAC client.
946 */
947 mac_unicast_update_client_flow(mcip);
948
949 i_mac_perim_exit(mip);
950 return (0);
951 }
952
953 /*
954 * Program the new primary unicast address of the specified MAC.
955 *
956 * Function mac_update_macaddr() takes care different types of underlying
957 * MAC. If the underlying MAC is VNIC, the VNIC driver must have registerd
958 * mi_unicst() entry point, that indirectly calls mac_vnic_unicast_set()
959 * which will take care of updating the MAC address of the corresponding
960 * MAC client.
961 *
962 * This is the only interface that allow the client to update the "primary"
963 * MAC address of the underlying MAC. The new value must have not been
964 * used by other clients.
965 */
966 int
967 mac_unicast_primary_set(mac_handle_t mh, const uint8_t *addr)
968 {
969 mac_impl_t *mip = (mac_impl_t *)mh;
970 mac_address_t *map;
971 int err;
972
973 /* verify the address validity */
974 if (!mac_unicst_verify(mh, addr, mip->mi_type->mt_addr_length))
975 return (EINVAL);
976
977 i_mac_perim_enter(mip);
978
979 /*
980 * If the new value is the same as the current primary address value,
981 * there's nothing to do.
982 */
983 if (bcmp(addr, mip->mi_addr, mip->mi_type->mt_addr_length) == 0) {
984 i_mac_perim_exit(mip);
985 return (0);
986 }
987
988 if (mac_find_macaddr(mip, (uint8_t *)addr) != 0) {
989 i_mac_perim_exit(mip);
990 return (EBUSY);
991 }
992
993 map = mac_find_macaddr(mip, mip->mi_addr);
994 ASSERT(map != NULL);
995
996 /*
997 * Update the MAC address.
998 */
999 if (mip->mi_state_flags & MIS_IS_AGGR) {
1000 mac_capab_aggr_t aggr_cap;
1001
1002 /*
1003 * If the mac is an aggregation, other than the unicast
1004 * addresses programming, aggr must be informed about this
1005 * primary unicst address change to change its mac address
1006 * policy to be user-specified.
1007 */
1008 ASSERT(map->ma_type == MAC_ADDRESS_TYPE_UNICAST_CLASSIFIED);
1009 VERIFY(i_mac_capab_get(mh, MAC_CAPAB_AGGR, &aggr_cap));
1010 err = aggr_cap.mca_unicst(mip->mi_driver, addr);
1011 if (err == 0)
1012 bcopy(addr, map->ma_addr, map->ma_len);
1013 } else {
1014 err = mac_update_macaddr(map, (uint8_t *)addr);
1015 }
1016
1017 if (err != 0) {
1018 i_mac_perim_exit(mip);
1019 return (err);
1020 }
1021
1022 mac_unicast_update_clients(mip, map);
1023
1024 /*
1025 * Save the new primary MAC address in mac_impl_t.
1026 */
1027 bcopy(addr, mip->mi_addr, mip->mi_type->mt_addr_length);
1028
1029 i_mac_perim_exit(mip);
1030
1031 if (err == 0)
1032 i_mac_notify(mip, MAC_NOTE_UNICST);
1033
1034 return (err);
1035 }
1036
1037 /*
1038 * Return the current primary MAC address of the specified MAC.
1039 */
1040 void
1041 mac_unicast_primary_get(mac_handle_t mh, uint8_t *addr)
1042 {
1043 mac_impl_t *mip = (mac_impl_t *)mh;
1044
1045 rw_enter(&mip->mi_rw_lock, RW_READER);
1046 bcopy(mip->mi_addr, addr, mip->mi_type->mt_addr_length);
1047 rw_exit(&mip->mi_rw_lock);
1048 }
1049
1050 /*
1051 * Return information about the use of the primary MAC address of the
1052 * specified MAC instance:
1053 *
1054 * - if client_name is non-NULL, it must point to a string of at
1055 * least MAXNAMELEN bytes, and will be set to the name of the MAC
1056 * client which uses the primary MAC address.
1057 *
1058 * - if in_use is non-NULL, used to return whether the primary MAC
1059 * address is currently in use.
1060 */
1061 void
1062 mac_unicast_primary_info(mac_handle_t mh, char *client_name, boolean_t *in_use)
1063 {
1064 mac_impl_t *mip = (mac_impl_t *)mh;
1065 mac_client_impl_t *cur_client;
1066
1067 if (in_use != NULL)
1068 *in_use = B_FALSE;
1069 if (client_name != NULL)
1070 bzero(client_name, MAXNAMELEN);
1071
1072 /*
1073 * The mi_rw_lock is used to protect threads that don't hold the
1074 * mac perimeter to get a consistent view of the mi_clients_list.
1075 * Threads that modify the list must hold both the mac perimeter and
1076 * mi_rw_lock(RW_WRITER)
1077 */
1078 rw_enter(&mip->mi_rw_lock, RW_READER);
1079 for (cur_client = mip->mi_clients_list; cur_client != NULL;
1080 cur_client = cur_client->mci_client_next) {
1081 if (mac_is_primary_client(cur_client) ||
1082 (mip->mi_state_flags & MIS_IS_VNIC)) {
1083 rw_exit(&mip->mi_rw_lock);
1084 if (in_use != NULL)
1085 *in_use = B_TRUE;
1086 if (client_name != NULL) {
1087 bcopy(cur_client->mci_name, client_name,
1088 MAXNAMELEN);
1089 }
1090 return;
1091 }
1092 }
1093 rw_exit(&mip->mi_rw_lock);
1094 }
1095
1096 /*
1097 * Return the current destination MAC address of the specified MAC.
1098 */
1099 boolean_t
1100 mac_dst_get(mac_handle_t mh, uint8_t *addr)
1101 {
1102 mac_impl_t *mip = (mac_impl_t *)mh;
1103
1104 rw_enter(&mip->mi_rw_lock, RW_READER);
1105 if (mip->mi_dstaddr_set)
1106 bcopy(mip->mi_dstaddr, addr, mip->mi_type->mt_addr_length);
1107 rw_exit(&mip->mi_rw_lock);
1108 return (mip->mi_dstaddr_set);
1109 }
1110
1111 /*
1112 * Add the specified MAC client to the list of clients which opened
1113 * the specified MAC.
1114 */
1115 static void
1116 mac_client_add(mac_client_impl_t *mcip)
1117 {
1118 mac_impl_t *mip = mcip->mci_mip;
1119
1120 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1121
1122 /* add VNIC to the front of the list */
1123 rw_enter(&mip->mi_rw_lock, RW_WRITER);
1124 mcip->mci_client_next = mip->mi_clients_list;
1125 mip->mi_clients_list = mcip;
1126 mip->mi_nclients++;
1127 rw_exit(&mip->mi_rw_lock);
1128 }
1129
1130 /*
1131 * Remove the specified MAC client from the list of clients which opened
1132 * the specified MAC.
1133 */
1134 static void
1135 mac_client_remove(mac_client_impl_t *mcip)
1136 {
1137 mac_impl_t *mip = mcip->mci_mip;
1138 mac_client_impl_t **prev, *cclient;
1139
1140 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1141
1142 rw_enter(&mip->mi_rw_lock, RW_WRITER);
1143 prev = &mip->mi_clients_list;
1144 cclient = *prev;
1145 while (cclient != NULL && cclient != mcip) {
1146 prev = &cclient->mci_client_next;
1147 cclient = *prev;
1148 }
1149 ASSERT(cclient != NULL);
1150 *prev = cclient->mci_client_next;
1151 mip->mi_nclients--;
1152 rw_exit(&mip->mi_rw_lock);
1153 }
1154
1155 static mac_unicast_impl_t *
1156 mac_client_find_vid(mac_client_impl_t *mcip, uint16_t vid)
1157 {
1158 mac_unicast_impl_t *muip = mcip->mci_unicast_list;
1159
1160 while ((muip != NULL) && (muip->mui_vid != vid))
1161 muip = muip->mui_next;
1162
1163 return (muip);
1164 }
1165
1166 /*
1167 * Return whether the specified (MAC address, VID) tuple is already used by
1168 * one of the MAC clients associated with the specified MAC.
1169 */
1170 static boolean_t
1171 mac_addr_in_use(mac_impl_t *mip, uint8_t *mac_addr, uint16_t vid)
1172 {
1173 mac_client_impl_t *client;
1174 mac_address_t *map;
1175
1176 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1177
1178 for (client = mip->mi_clients_list; client != NULL;
1179 client = client->mci_client_next) {
1180
1181 /*
1182 * Ignore clients that don't have unicast address.
1183 */
1184 if (client->mci_unicast_list == NULL)
1185 continue;
1186
1187 map = client->mci_unicast;
1188
1189 if ((bcmp(mac_addr, map->ma_addr, map->ma_len) == 0) &&
1190 (mac_client_find_vid(client, vid) != NULL)) {
1191 return (B_TRUE);
1192 }
1193 }
1194
1195 return (B_FALSE);
1196 }
1197
1198 /*
1199 * Generate a random MAC address. The MAC address prefix is
1200 * stored in the array pointed to by mac_addr, and its length, in bytes,
1201 * is specified by prefix_len. The least significant bits
1202 * after prefix_len bytes are generated, and stored after the prefix
1203 * in the mac_addr array.
1204 */
1205 int
1206 mac_addr_random(mac_client_handle_t mch, uint_t prefix_len,
1207 uint8_t *mac_addr, mac_diag_t *diag)
1208 {
1209 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1210 mac_impl_t *mip = mcip->mci_mip;
1211 size_t addr_len = mip->mi_type->mt_addr_length;
1212
1213 if (prefix_len >= addr_len) {
1214 *diag = MAC_DIAG_MACPREFIXLEN_INVALID;
1215 return (EINVAL);
1216 }
1217
1218 /* check the prefix value */
1219 if (prefix_len > 0) {
1220 bzero(mac_addr + prefix_len, addr_len - prefix_len);
1221 if (!mac_unicst_verify((mac_handle_t)mip, mac_addr,
1222 addr_len)) {
1223 *diag = MAC_DIAG_MACPREFIX_INVALID;
1224 return (EINVAL);
1225 }
1226 }
1227
1228 /* generate the MAC address */
1229 if (prefix_len < addr_len) {
1230 (void) random_get_pseudo_bytes(mac_addr +
1231 prefix_len, addr_len - prefix_len);
1232 }
1233
1234 *diag = 0;
1235 return (0);
1236 }
1237
1238 /*
1239 * Set the priority range for this MAC client. This will be used to
1240 * determine the absolute priority for the threads created for this
1241 * MAC client using the specified "low", "medium" and "high" level.
1242 * This will also be used for any subflows on this MAC client.
1243 */
1244 #define MAC_CLIENT_SET_PRIORITY_RANGE(mcip, pri) { \
1245 (mcip)->mci_min_pri = FLOW_MIN_PRIORITY(MINCLSYSPRI, \
1246 MAXCLSYSPRI, (pri)); \
1247 (mcip)->mci_max_pri = FLOW_MAX_PRIORITY(MINCLSYSPRI, \
1248 MAXCLSYSPRI, (mcip)->mci_min_pri); \
1249 }
1250
1251 /*
1252 * MAC client open entry point. Return a new MAC client handle. Each
1253 * MAC client is associated with a name, specified through the 'name'
1254 * argument.
1255 */
1256 int
1257 mac_client_open(mac_handle_t mh, mac_client_handle_t *mchp, char *name,
1258 uint16_t flags)
1259 {
1260 mac_impl_t *mip = (mac_impl_t *)mh;
1261 mac_client_impl_t *mcip;
1262 int err = 0;
1263 boolean_t share_desired;
1264 flow_entry_t *flent = NULL;
1265
1266 share_desired = (flags & MAC_OPEN_FLAGS_SHARES_DESIRED) != 0;
1267 *mchp = NULL;
1268
1269 i_mac_perim_enter(mip);
1270
1271 if (mip->mi_state_flags & MIS_IS_VNIC) {
1272 /*
1273 * The underlying MAC is a VNIC. Return the MAC client
1274 * handle of the lower MAC which was obtained by
1275 * the VNIC driver when it did its mac_client_open().
1276 */
1277
1278 mcip = mac_vnic_lower(mip);
1279
1280 /*
1281 * Note that multiple mac clients share the same mcip in
1282 * this case.
1283 */
1284 if (flags & MAC_OPEN_FLAGS_EXCLUSIVE)
1285 mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1286
1287 if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1288 mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1289
1290 mip->mi_clients_list = mcip;
1291 i_mac_perim_exit(mip);
1292 *mchp = (mac_client_handle_t)mcip;
1293 return (err);
1294 }
1295
1296 mcip = kmem_cache_alloc(mac_client_impl_cache, KM_SLEEP);
1297
1298 mcip->mci_mip = mip;
1299 mcip->mci_upper_mip = NULL;
1300 mcip->mci_rx_fn = mac_pkt_drop;
1301 mcip->mci_rx_arg = NULL;
1302 mcip->mci_rx_p_fn = NULL;
1303 mcip->mci_rx_p_arg = NULL;
1304 mcip->mci_p_unicast_list = NULL;
1305 mcip->mci_direct_rx_fn = NULL;
1306 mcip->mci_direct_rx_arg = NULL;
1307
1308 mcip->mci_unicast_list = NULL;
1309
1310 if ((flags & MAC_OPEN_FLAGS_IS_VNIC) != 0)
1311 mcip->mci_state_flags |= MCIS_IS_VNIC;
1312
1313 if ((flags & MAC_OPEN_FLAGS_EXCLUSIVE) != 0)
1314 mcip->mci_state_flags |= MCIS_EXCLUSIVE;
1315
1316 if ((flags & MAC_OPEN_FLAGS_IS_AGGR_PORT) != 0)
1317 mcip->mci_state_flags |= MCIS_IS_AGGR_PORT;
1318
1319 if (mip->mi_state_flags & MIS_IS_AGGR)
1320 mcip->mci_state_flags |= MCIS_IS_AGGR;
1321
1322 if ((flags & MAC_OPEN_FLAGS_USE_DATALINK_NAME) != 0) {
1323 datalink_id_t linkid;
1324
1325 ASSERT(name == NULL);
1326 if ((err = dls_devnet_macname2linkid(mip->mi_name,
1327 &linkid)) != 0) {
1328 goto done;
1329 }
1330 if ((err = dls_mgmt_get_linkinfo(linkid, mcip->mci_name, NULL,
1331 NULL, NULL)) != 0) {
1332 /*
1333 * Use mac name if dlmgmtd is not available.
1334 */
1335 if (err == EBADF) {
1336 (void) strlcpy(mcip->mci_name, mip->mi_name,
1337 sizeof (mcip->mci_name));
1338 err = 0;
1339 } else {
1340 goto done;
1341 }
1342 }
1343 mcip->mci_state_flags |= MCIS_USE_DATALINK_NAME;
1344 } else {
1345 ASSERT(name != NULL);
1346 if (strlen(name) > MAXNAMELEN) {
1347 err = EINVAL;
1348 goto done;
1349 }
1350 (void) strlcpy(mcip->mci_name, name, sizeof (mcip->mci_name));
1351 }
1352
1353 if (flags & MAC_OPEN_FLAGS_MULTI_PRIMARY)
1354 mcip->mci_flags |= MAC_CLIENT_FLAGS_MULTI_PRIMARY;
1355
1356 if (flags & MAC_OPEN_FLAGS_NO_UNICAST_ADDR)
1357 mcip->mci_state_flags |= MCIS_NO_UNICAST_ADDR;
1358
1359 mac_protect_init(mcip);
1360
1361 /* the subflow table will be created dynamically */
1362 mcip->mci_subflow_tab = NULL;
1363
1364 mcip->mci_misc_stat.mms_multircv = 0;
1365 mcip->mci_misc_stat.mms_brdcstrcv = 0;
1366 mcip->mci_misc_stat.mms_multixmt = 0;
1367 mcip->mci_misc_stat.mms_brdcstxmt = 0;
1368
1369 /* Create an initial flow */
1370
1371 err = mac_flow_create(NULL, NULL, mcip->mci_name, NULL,
1372 mcip->mci_state_flags & MCIS_IS_VNIC ? FLOW_VNIC_MAC :
1373 FLOW_PRIMARY_MAC, &flent);
1374 if (err != 0)
1375 goto done;
1376 mcip->mci_flent = flent;
1377 FLOW_MARK(flent, FE_MC_NO_DATAPATH);
1378 flent->fe_mcip = mcip;
1379 /*
1380 * Place initial creation reference on the flow. This reference
1381 * is released in the corresponding delete action viz.
1382 * mac_unicast_remove after waiting for all transient refs to
1383 * to go away. The wait happens in mac_flow_wait.
1384 */
1385 FLOW_REFHOLD(flent);
1386
1387 /*
1388 * Do this ahead of the mac_bcast_add() below so that the mi_nclients
1389 * will have the right value for mac_rx_srs_setup().
1390 */
1391 mac_client_add(mcip);
1392
1393 mcip->mci_share = NULL;
1394 if (share_desired)
1395 i_mac_share_alloc(mcip);
1396
1397 DTRACE_PROBE2(mac__client__open__allocated, mac_impl_t *,
1398 mcip->mci_mip, mac_client_impl_t *, mcip);
1399 *mchp = (mac_client_handle_t)mcip;
1400
1401 /*
1402 * We will do mimimal datapath setup to allow a MAC client to
1403 * transmit or receive non-unicast packets without waiting
1404 * for mac_unicast_add.
1405 */
1406 if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1407 if ((err = mac_client_datapath_setup(mcip, VLAN_ID_NONE,
1408 NULL, NULL, B_TRUE, NULL)) != 0) {
1409 goto done;
1410 }
1411 }
1412 i_mac_perim_exit(mip);
1413 return (0);
1414
1415 done:
1416 i_mac_perim_exit(mip);
1417 mcip->mci_state_flags = 0;
1418 mcip->mci_tx_flag = 0;
1419 kmem_cache_free(mac_client_impl_cache, mcip);
1420 return (err);
1421 }
1422
1423 /*
1424 * Close the specified MAC client handle.
1425 */
1426 void
1427 mac_client_close(mac_client_handle_t mch, uint16_t flags)
1428 {
1429 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1430 mac_impl_t *mip = mcip->mci_mip;
1431 flow_entry_t *flent;
1432
1433 i_mac_perim_enter(mip);
1434
1435 if (flags & MAC_CLOSE_FLAGS_EXCLUSIVE)
1436 mcip->mci_state_flags &= ~MCIS_EXCLUSIVE;
1437
1438 if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
1439 !(flags & MAC_CLOSE_FLAGS_IS_VNIC)) {
1440 /*
1441 * This is an upper VNIC client initiated operation.
1442 * The lower MAC client will be closed by the VNIC driver
1443 * when the VNIC is deleted.
1444 */
1445
1446 i_mac_perim_exit(mip);
1447 return;
1448 }
1449
1450 /* If we have only setup up minimal datapth setup, tear it down */
1451 if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR) {
1452 mac_client_datapath_teardown((mac_client_handle_t)mcip, NULL,
1453 mcip->mci_flent);
1454 mcip->mci_state_flags &= ~MCIS_NO_UNICAST_ADDR;
1455 }
1456
1457 /*
1458 * Remove the flent associated with the MAC client
1459 */
1460 flent = mcip->mci_flent;
1461 mcip->mci_flent = NULL;
1462 FLOW_FINAL_REFRELE(flent);
1463
1464 /*
1465 * MAC clients must remove the unicast addresses and promisc callbacks
1466 * they added before issuing a mac_client_close().
1467 */
1468 ASSERT(mcip->mci_unicast_list == NULL);
1469 ASSERT(mcip->mci_promisc_list == NULL);
1470 ASSERT(mcip->mci_tx_notify_cb_list == NULL);
1471
1472 i_mac_share_free(mcip);
1473 mac_protect_fini(mcip);
1474 mac_client_remove(mcip);
1475
1476 i_mac_perim_exit(mip);
1477 mcip->mci_subflow_tab = NULL;
1478 mcip->mci_state_flags = 0;
1479 mcip->mci_tx_flag = 0;
1480 kmem_cache_free(mac_client_impl_cache, mch);
1481 }
1482
1483 /*
1484 * Set the rx bypass receive callback.
1485 */
1486 boolean_t
1487 mac_rx_bypass_set(mac_client_handle_t mch, mac_direct_rx_t rx_fn, void *arg1)
1488 {
1489 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1490 mac_impl_t *mip = mcip->mci_mip;
1491
1492 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1493
1494 /*
1495 * If the mac_client is a VLAN, we should not do DLS bypass and
1496 * instead let the packets come up via mac_rx_deliver so the vlan
1497 * header can be stripped.
1498 */
1499 if (mcip->mci_nvids > 0)
1500 return (B_FALSE);
1501
1502 /*
1503 * These are not accessed directly in the data path, and hence
1504 * don't need any protection
1505 */
1506 mcip->mci_direct_rx_fn = rx_fn;
1507 mcip->mci_direct_rx_arg = arg1;
1508 return (B_TRUE);
1509 }
1510
1511 /*
1512 * Enable/Disable rx bypass. By default, bypass is assumed to be enabled.
1513 */
1514 void
1515 mac_rx_bypass_enable(mac_client_handle_t mch)
1516 {
1517 ((mac_client_impl_t *)mch)->mci_state_flags &= ~MCIS_RX_BYPASS_DISABLE;
1518 }
1519
1520 void
1521 mac_rx_bypass_disable(mac_client_handle_t mch)
1522 {
1523 ((mac_client_impl_t *)mch)->mci_state_flags |= MCIS_RX_BYPASS_DISABLE;
1524 }
1525
1526 /*
1527 * Set the receive callback for the specified MAC client. There can be
1528 * at most one such callback per MAC client.
1529 */
1530 void
1531 mac_rx_set(mac_client_handle_t mch, mac_rx_t rx_fn, void *arg)
1532 {
1533 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1534 mac_impl_t *mip = mcip->mci_mip;
1535
1536 /*
1537 * Instead of adding an extra set of locks and refcnts in
1538 * the datapath at the mac client boundary, we temporarily quiesce
1539 * the SRS and related entities. We then change the receive function
1540 * without interference from any receive data thread and then reenable
1541 * the data flow subsequently.
1542 */
1543 i_mac_perim_enter(mip);
1544 mac_rx_client_quiesce(mch);
1545
1546 mcip->mci_rx_fn = rx_fn;
1547 mcip->mci_rx_arg = arg;
1548 mac_rx_client_restart(mch);
1549 i_mac_perim_exit(mip);
1550 }
1551
1552 /*
1553 * Reset the receive callback for the specified MAC client.
1554 */
1555 void
1556 mac_rx_clear(mac_client_handle_t mch)
1557 {
1558 mac_rx_set(mch, mac_pkt_drop, NULL);
1559 }
1560
1561 /*
1562 * Walk the MAC client subflow table and updates their priority values.
1563 */
1564 static int
1565 mac_update_subflow_priority_cb(flow_entry_t *flent, void *arg)
1566 {
1567 mac_flow_update_priority(arg, flent);
1568 return (0);
1569 }
1570
1571 void
1572 mac_update_subflow_priority(mac_client_impl_t *mcip)
1573 {
1574 (void) mac_flow_walk(mcip->mci_subflow_tab,
1575 mac_update_subflow_priority_cb, mcip);
1576 }
1577
1578 /*
1579 * Modify the TX or RX ring properties. We could either just move around
1580 * rings, i.e add/remove rings given to a client. Or this might cause the
1581 * client to move from hardware based to software or the other way around.
1582 * If we want to reset this property, then we clear the mask, additionally
1583 * if the client was given a non-default group we remove all rings except
1584 * for 1 and give it back to the default group.
1585 */
1586 int
1587 mac_client_set_rings_prop(mac_client_impl_t *mcip, mac_resource_props_t *mrp,
1588 mac_resource_props_t *tmrp)
1589 {
1590 mac_impl_t *mip = mcip->mci_mip;
1591 flow_entry_t *flent = mcip->mci_flent;
1592 uint8_t *mac_addr;
1593 int err = 0;
1594 mac_group_t *defgrp;
1595 mac_group_t *group;
1596 mac_group_t *ngrp;
1597 mac_resource_props_t *cmrp = MCIP_RESOURCE_PROPS(mcip);
1598 uint_t ringcnt;
1599 boolean_t unspec;
1600
1601 if (mcip->mci_share != NULL)
1602 return (EINVAL);
1603
1604 if (mrp->mrp_mask & MRP_RX_RINGS) {
1605 unspec = mrp->mrp_mask & MRP_RXRINGS_UNSPEC;
1606 group = flent->fe_rx_ring_group;
1607 defgrp = MAC_DEFAULT_RX_GROUP(mip);
1608 mac_addr = flent->fe_flow_desc.fd_dst_mac;
1609
1610 /*
1611 * No resulting change. If we are resetting on a client on
1612 * which there was no rx rings property. For dynamic group
1613 * if we are setting the same number of rings already set.
1614 * For static group if we are requesting a group again.
1615 */
1616 if (mrp->mrp_mask & MRP_RINGS_RESET) {
1617 if (!(tmrp->mrp_mask & MRP_RX_RINGS))
1618 return (0);
1619 } else {
1620 if (unspec) {
1621 if (tmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1622 return (0);
1623 } else if (mip->mi_rx_group_type ==
1624 MAC_GROUP_TYPE_DYNAMIC) {
1625 if ((tmrp->mrp_mask & MRP_RX_RINGS) &&
1626 !(tmrp->mrp_mask & MRP_RXRINGS_UNSPEC) &&
1627 mrp->mrp_nrxrings == tmrp->mrp_nrxrings) {
1628 return (0);
1629 }
1630 }
1631 }
1632 /* Resetting the prop */
1633 if (mrp->mrp_mask & MRP_RINGS_RESET) {
1634 /*
1635 * We will just keep one ring and give others back if
1636 * we are not the primary. For the primary we give
1637 * all the rings in the default group except the
1638 * default ring. If it is a static group, then
1639 * we don't do anything, but clear the MRP_RX_RINGS
1640 * flag.
1641 */
1642 if (group != defgrp) {
1643 if (mip->mi_rx_group_type ==
1644 MAC_GROUP_TYPE_DYNAMIC) {
1645 /*
1646 * This group has reserved rings
1647 * that need to be released now,
1648 * so does the group.
1649 */
1650 MAC_RX_RING_RELEASED(mip,
1651 group->mrg_cur_count);
1652 MAC_RX_GRP_RELEASED(mip);
1653 if ((flent->fe_type &
1654 FLOW_PRIMARY_MAC) != 0) {
1655 if (mip->mi_nactiveclients ==
1656 1) {
1657 (void)
1658 mac_rx_switch_group(
1659 mcip, group,
1660 defgrp);
1661 return (0);
1662 } else {
1663 cmrp->mrp_nrxrings =
1664 group->
1665 mrg_cur_count +
1666 defgrp->
1667 mrg_cur_count - 1;
1668 }
1669 } else {
1670 cmrp->mrp_nrxrings = 1;
1671 }
1672 (void) mac_group_ring_modify(mcip,
1673 group, defgrp);
1674 } else {
1675 /*
1676 * If this is a static group, we
1677 * need to release the group. The
1678 * client will remain in the same
1679 * group till some other client
1680 * needs this group.
1681 */
1682 MAC_RX_GRP_RELEASED(mip);
1683 }
1684 /* Let check if we can give this an excl group */
1685 } else if (group == defgrp) {
1686 ngrp = mac_reserve_rx_group(mcip, mac_addr,
1687 B_TRUE);
1688 /* Couldn't give it a group, that's fine */
1689 if (ngrp == NULL)
1690 return (0);
1691 /* Switch to H/W */
1692 if (mac_rx_switch_group(mcip, defgrp, ngrp) !=
1693 0) {
1694 mac_stop_group(ngrp);
1695 return (0);
1696 }
1697 }
1698 /*
1699 * If the client is in the default group, we will
1700 * just clear the MRP_RX_RINGS and leave it as
1701 * it rather than look for an exclusive group
1702 * for it.
1703 */
1704 return (0);
1705 }
1706
1707 if (group == defgrp && ((mrp->mrp_nrxrings > 0) || unspec)) {
1708 ngrp = mac_reserve_rx_group(mcip, mac_addr, B_TRUE);
1709 if (ngrp == NULL)
1710 return (ENOSPC);
1711
1712 /* Switch to H/W */
1713 if (mac_rx_switch_group(mcip, defgrp, ngrp) != 0) {
1714 mac_release_rx_group(mcip, ngrp);
1715 return (ENOSPC);
1716 }
1717 MAC_RX_GRP_RESERVED(mip);
1718 if (mip->mi_rx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1719 MAC_RX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1720 } else if (group != defgrp && !unspec &&
1721 mrp->mrp_nrxrings == 0) {
1722 /* Switch to S/W */
1723 ringcnt = group->mrg_cur_count;
1724 if (mac_rx_switch_group(mcip, group, defgrp) != 0)
1725 return (ENOSPC);
1726 if (tmrp->mrp_mask & MRP_RX_RINGS) {
1727 MAC_RX_GRP_RELEASED(mip);
1728 if (mip->mi_rx_group_type ==
1729 MAC_GROUP_TYPE_DYNAMIC) {
1730 MAC_RX_RING_RELEASED(mip, ringcnt);
1731 }
1732 }
1733 } else if (group != defgrp && mip->mi_rx_group_type ==
1734 MAC_GROUP_TYPE_DYNAMIC) {
1735 ringcnt = group->mrg_cur_count;
1736 err = mac_group_ring_modify(mcip, group, defgrp);
1737 if (err != 0)
1738 return (err);
1739 /*
1740 * Update the accounting. If this group
1741 * already had explicitly reserved rings,
1742 * we need to update the rings based on
1743 * the new ring count. If this group
1744 * had not explicitly reserved rings,
1745 * then we just reserve the rings asked for
1746 * and reserve the group.
1747 */
1748 if (tmrp->mrp_mask & MRP_RX_RINGS) {
1749 if (ringcnt > group->mrg_cur_count) {
1750 MAC_RX_RING_RELEASED(mip,
1751 ringcnt - group->mrg_cur_count);
1752 } else {
1753 MAC_RX_RING_RESERVED(mip,
1754 group->mrg_cur_count - ringcnt);
1755 }
1756 } else {
1757 MAC_RX_RING_RESERVED(mip, group->mrg_cur_count);
1758 MAC_RX_GRP_RESERVED(mip);
1759 }
1760 }
1761 }
1762 if (mrp->mrp_mask & MRP_TX_RINGS) {
1763 unspec = mrp->mrp_mask & MRP_TXRINGS_UNSPEC;
1764 group = flent->fe_tx_ring_group;
1765 defgrp = MAC_DEFAULT_TX_GROUP(mip);
1766
1767 /*
1768 * For static groups we only allow rings=0 or resetting the
1769 * rings property.
1770 */
1771 if (mrp->mrp_ntxrings > 0 &&
1772 mip->mi_tx_group_type != MAC_GROUP_TYPE_DYNAMIC) {
1773 return (ENOTSUP);
1774 }
1775 if (mrp->mrp_mask & MRP_RINGS_RESET) {
1776 if (!(tmrp->mrp_mask & MRP_TX_RINGS))
1777 return (0);
1778 } else {
1779 if (unspec) {
1780 if (tmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1781 return (0);
1782 } else if (mip->mi_tx_group_type ==
1783 MAC_GROUP_TYPE_DYNAMIC) {
1784 if ((tmrp->mrp_mask & MRP_TX_RINGS) &&
1785 !(tmrp->mrp_mask & MRP_TXRINGS_UNSPEC) &&
1786 mrp->mrp_ntxrings == tmrp->mrp_ntxrings) {
1787 return (0);
1788 }
1789 }
1790 }
1791 /* Resetting the prop */
1792 if (mrp->mrp_mask & MRP_RINGS_RESET) {
1793 if (group != defgrp) {
1794 if (mip->mi_tx_group_type ==
1795 MAC_GROUP_TYPE_DYNAMIC) {
1796 ringcnt = group->mrg_cur_count;
1797 if ((flent->fe_type &
1798 FLOW_PRIMARY_MAC) != 0) {
1799 mac_tx_client_quiesce(
1800 (mac_client_handle_t)
1801 mcip);
1802 mac_tx_switch_group(mcip,
1803 group, defgrp);
1804 mac_tx_client_restart(
1805 (mac_client_handle_t)
1806 mcip);
1807 MAC_TX_GRP_RELEASED(mip);
1808 MAC_TX_RING_RELEASED(mip,
1809 ringcnt);
1810 return (0);
1811 }
1812 cmrp->mrp_ntxrings = 1;
1813 (void) mac_group_ring_modify(mcip,
1814 group, defgrp);
1815 /*
1816 * This group has reserved rings
1817 * that need to be released now.
1818 */
1819 MAC_TX_RING_RELEASED(mip, ringcnt);
1820 }
1821 /*
1822 * If this is a static group, we
1823 * need to release the group. The
1824 * client will remain in the same
1825 * group till some other client
1826 * needs this group.
1827 */
1828 MAC_TX_GRP_RELEASED(mip);
1829 } else if (group == defgrp &&
1830 (flent->fe_type & FLOW_PRIMARY_MAC) == 0) {
1831 ngrp = mac_reserve_tx_group(mcip, B_TRUE);
1832 if (ngrp == NULL)
1833 return (0);
1834 mac_tx_client_quiesce(
1835 (mac_client_handle_t)mcip);
1836 mac_tx_switch_group(mcip, defgrp, ngrp);
1837 mac_tx_client_restart(
1838 (mac_client_handle_t)mcip);
1839 }
1840 /*
1841 * If the client is in the default group, we will
1842 * just clear the MRP_TX_RINGS and leave it as
1843 * it rather than look for an exclusive group
1844 * for it.
1845 */
1846 return (0);
1847 }
1848
1849 /* Switch to H/W */
1850 if (group == defgrp && ((mrp->mrp_ntxrings > 0) || unspec)) {
1851 ngrp = mac_reserve_tx_group(mcip, B_TRUE);
1852 if (ngrp == NULL)
1853 return (ENOSPC);
1854 mac_tx_client_quiesce((mac_client_handle_t)mcip);
1855 mac_tx_switch_group(mcip, defgrp, ngrp);
1856 mac_tx_client_restart((mac_client_handle_t)mcip);
1857 MAC_TX_GRP_RESERVED(mip);
1858 if (mip->mi_tx_group_type == MAC_GROUP_TYPE_DYNAMIC)
1859 MAC_TX_RING_RESERVED(mip, ngrp->mrg_cur_count);
1860 /* Switch to S/W */
1861 } else if (group != defgrp && !unspec &&
1862 mrp->mrp_ntxrings == 0) {
1863 /* Switch to S/W */
1864 ringcnt = group->mrg_cur_count;
1865 mac_tx_client_quiesce((mac_client_handle_t)mcip);
1866 mac_tx_switch_group(mcip, group, defgrp);
1867 mac_tx_client_restart((mac_client_handle_t)mcip);
1868 if (tmrp->mrp_mask & MRP_TX_RINGS) {
1869 MAC_TX_GRP_RELEASED(mip);
1870 if (mip->mi_tx_group_type ==
1871 MAC_GROUP_TYPE_DYNAMIC) {
1872 MAC_TX_RING_RELEASED(mip, ringcnt);
1873 }
1874 }
1875 } else if (group != defgrp && mip->mi_tx_group_type ==
1876 MAC_GROUP_TYPE_DYNAMIC) {
1877 ringcnt = group->mrg_cur_count;
1878 err = mac_group_ring_modify(mcip, group, defgrp);
1879 if (err != 0)
1880 return (err);
1881 /*
1882 * Update the accounting. If this group
1883 * already had explicitly reserved rings,
1884 * we need to update the rings based on
1885 * the new ring count. If this group
1886 * had not explicitly reserved rings,
1887 * then we just reserve the rings asked for
1888 * and reserve the group.
1889 */
1890 if (tmrp->mrp_mask & MRP_TX_RINGS) {
1891 if (ringcnt > group->mrg_cur_count) {
1892 MAC_TX_RING_RELEASED(mip,
1893 ringcnt - group->mrg_cur_count);
1894 } else {
1895 MAC_TX_RING_RESERVED(mip,
1896 group->mrg_cur_count - ringcnt);
1897 }
1898 } else {
1899 MAC_TX_RING_RESERVED(mip, group->mrg_cur_count);
1900 MAC_TX_GRP_RESERVED(mip);
1901 }
1902 }
1903 }
1904 return (0);
1905 }
1906
1907 /*
1908 * When the MAC client is being brought up (i.e. we do a unicast_add) we need
1909 * to initialize the cpu and resource control structure in the
1910 * mac_client_impl_t from the mac_impl_t (i.e if there are any cached
1911 * properties before the flow entry for the unicast address was created).
1912 */
1913 int
1914 mac_resource_ctl_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
1915 {
1916 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
1917 mac_impl_t *mip = (mac_impl_t *)mcip->mci_mip;
1918 int err = 0;
1919 flow_entry_t *flent = mcip->mci_flent;
1920 mac_resource_props_t *omrp, *nmrp = MCIP_RESOURCE_PROPS(mcip);
1921
1922 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
1923
1924 err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
1925 mcip->mci_upper_mip : mip, mrp);
1926 if (err != 0)
1927 return (err);
1928
1929 /*
1930 * Copy over the existing properties since mac_update_resources
1931 * will modify the client's mrp. Currently, the saved property
1932 * is used to determine the difference between existing and
1933 * modified rings property.
1934 */
1935 omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
1936 bcopy(nmrp, omrp, sizeof (*omrp));
1937 mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
1938 if (MCIP_DATAPATH_SETUP(mcip)) {
1939 /*
1940 * We support rings only for primary client when there are
1941 * multiple clients sharing the same MAC address (e.g. VLAN).
1942 */
1943 if (mrp->mrp_mask & MRP_RX_RINGS ||
1944 mrp->mrp_mask & MRP_TX_RINGS) {
1945
1946 if ((err = mac_client_set_rings_prop(mcip, mrp,
1947 omrp)) != 0) {
1948 if (omrp->mrp_mask & MRP_RX_RINGS) {
1949 nmrp->mrp_mask |= MRP_RX_RINGS;
1950 nmrp->mrp_nrxrings = omrp->mrp_nrxrings;
1951 } else {
1952 nmrp->mrp_mask &= ~MRP_RX_RINGS;
1953 nmrp->mrp_nrxrings = 0;
1954 }
1955 if (omrp->mrp_mask & MRP_TX_RINGS) {
1956 nmrp->mrp_mask |= MRP_TX_RINGS;
1957 nmrp->mrp_ntxrings = omrp->mrp_ntxrings;
1958 } else {
1959 nmrp->mrp_mask &= ~MRP_TX_RINGS;
1960 nmrp->mrp_ntxrings = 0;
1961 }
1962 if (omrp->mrp_mask & MRP_RXRINGS_UNSPEC)
1963 omrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
1964 else
1965 omrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
1966
1967 if (omrp->mrp_mask & MRP_TXRINGS_UNSPEC)
1968 omrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
1969 else
1970 omrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
1971 kmem_free(omrp, sizeof (*omrp));
1972 return (err);
1973 }
1974
1975 /*
1976 * If we modified the rings property of the primary
1977 * we need to update the property fields of its
1978 * VLANs as they inherit the primary's properites.
1979 */
1980 if (mac_is_primary_client(mcip)) {
1981 mac_set_prim_vlan_rings(mip,
1982 MCIP_RESOURCE_PROPS(mcip));
1983 }
1984 }
1985 /*
1986 * We have to set this prior to calling mac_flow_modify.
1987 */
1988 if (mrp->mrp_mask & MRP_PRIORITY) {
1989 if (mrp->mrp_priority == MPL_RESET) {
1990 MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1991 MPL_LINK_DEFAULT);
1992 } else {
1993 MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
1994 mrp->mrp_priority);
1995 }
1996 }
1997
1998 mac_flow_modify(mip->mi_flow_tab, flent, mrp);
1999 if (mrp->mrp_mask & MRP_PRIORITY)
2000 mac_update_subflow_priority(mcip);
2001 }
2002 kmem_free(omrp, sizeof (*omrp));
2003 return (0);
2004 }
2005
2006 void
2007 mac_resource_ctl_get(mac_client_handle_t mch, mac_resource_props_t *mrp)
2008 {
2009 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2010 mac_resource_props_t *mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
2011
2012 bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
2013 }
2014
2015 static int
2016 mac_unicast_flow_create(mac_client_impl_t *mcip, uint8_t *mac_addr,
2017 uint16_t vid, boolean_t is_primary, boolean_t first_flow,
2018 flow_entry_t **flent, mac_resource_props_t *mrp)
2019 {
2020 mac_impl_t *mip = (mac_impl_t *)mcip->mci_mip;
2021 flow_desc_t flow_desc;
2022 char flowname[MAXFLOWNAMELEN];
2023 int err;
2024 uint_t flent_flags;
2025
2026 /*
2027 * First unicast address being added, create a new flow
2028 * for that MAC client.
2029 */
2030 bzero(&flow_desc, sizeof (flow_desc));
2031
2032 ASSERT(mac_addr != NULL ||
2033 (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR));
2034 if (mac_addr != NULL) {
2035 flow_desc.fd_mac_len = mip->mi_type->mt_addr_length;
2036 bcopy(mac_addr, flow_desc.fd_dst_mac, flow_desc.fd_mac_len);
2037 }
2038 flow_desc.fd_mask = FLOW_LINK_DST;
2039 if (vid != 0) {
2040 flow_desc.fd_vid = vid;
2041 flow_desc.fd_mask |= FLOW_LINK_VID;
2042 }
2043
2044 /*
2045 * XXX-nicolas. For now I'm keeping the FLOW_PRIMARY_MAC
2046 * and FLOW_VNIC. Even though they're a hack inherited
2047 * from the SRS code, we'll keep them for now. They're currently
2048 * consumed by mac_datapath_setup() to create the SRS.
2049 * That code should be eventually moved out of
2050 * mac_datapath_setup() and moved to a mac_srs_create()
2051 * function of some sort to keep things clean.
2052 *
2053 * Also, there's no reason why the SRS for the primary MAC
2054 * client should be different than any other MAC client. Until
2055 * this is cleaned-up, we support only one MAC unicast address
2056 * per client.
2057 *
2058 * We set FLOW_PRIMARY_MAC for the primary MAC address,
2059 * FLOW_VNIC for everything else.
2060 */
2061 if (is_primary)
2062 flent_flags = FLOW_PRIMARY_MAC;
2063 else
2064 flent_flags = FLOW_VNIC_MAC;
2065
2066 /*
2067 * For the first flow we use the mac client's name - mci_name, for
2068 * subsequent ones we just create a name with the vid. This is
2069 * so that we can add these flows to the same flow table. This is
2070 * fine as the flow name (except for the one with the mac client's
2071 * name) is not visible. When the first flow is removed, we just replace
2072 * its fdesc with another from the list, so we will still retain the
2073 * flent with the MAC client's flow name.
2074 */
2075 if (first_flow) {
2076 bcopy(mcip->mci_name, flowname, MAXFLOWNAMELEN);
2077 } else {
2078 (void) sprintf(flowname, "%s%u", mcip->mci_name, vid);
2079 flent_flags = FLOW_NO_STATS;
2080 }
2081
2082 if ((err = mac_flow_create(&flow_desc, mrp, flowname, NULL,
2083 flent_flags, flent)) != 0)
2084 return (err);
2085
2086 mac_misc_stat_create(*flent);
2087 FLOW_MARK(*flent, FE_INCIPIENT);
2088 (*flent)->fe_mcip = mcip;
2089
2090 /*
2091 * Place initial creation reference on the flow. This reference
2092 * is released in the corresponding delete action viz.
2093 * mac_unicast_remove after waiting for all transient refs to
2094 * to go away. The wait happens in mac_flow_wait.
2095 * We have already held the reference in mac_client_open().
2096 */
2097 if (!first_flow)
2098 FLOW_REFHOLD(*flent);
2099 return (0);
2100 }
2101
2102 /* Refresh the multicast grouping for this VID. */
2103 int
2104 mac_client_update_mcast(void *arg, boolean_t add, const uint8_t *addrp)
2105 {
2106 flow_entry_t *flent = arg;
2107 mac_client_impl_t *mcip = flent->fe_mcip;
2108 uint16_t vid;
2109 flow_desc_t flow_desc;
2110
2111 mac_flow_get_desc(flent, &flow_desc);
2112 vid = (flow_desc.fd_mask & FLOW_LINK_VID) != 0 ?
2113 flow_desc.fd_vid : VLAN_ID_NONE;
2114
2115 /*
2116 * We don't call mac_multicast_add()/mac_multicast_remove() as
2117 * we want to add/remove for this specific vid.
2118 */
2119 if (add) {
2120 return (mac_bcast_add(mcip, addrp, vid,
2121 MAC_ADDRTYPE_MULTICAST));
2122 } else {
2123 mac_bcast_delete(mcip, addrp, vid);
2124 return (0);
2125 }
2126 }
2127
2128 static void
2129 mac_update_single_active_client(mac_impl_t *mip)
2130 {
2131 mac_client_impl_t *client = NULL;
2132
2133 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
2134
2135 rw_enter(&mip->mi_rw_lock, RW_WRITER);
2136 if (mip->mi_nactiveclients == 1) {
2137 /*
2138 * Find the one active MAC client from the list of MAC
2139 * clients. The active MAC client has at least one
2140 * unicast address.
2141 */
2142 for (client = mip->mi_clients_list; client != NULL;
2143 client = client->mci_client_next) {
2144 if (client->mci_unicast_list != NULL)
2145 break;
2146 }
2147 ASSERT(client != NULL);
2148 }
2149
2150 /*
2151 * mi_single_active_client is protected by the MAC impl's read/writer
2152 * lock, which allows mac_rx() to check the value of that pointer
2153 * as a reader.
2154 */
2155 mip->mi_single_active_client = client;
2156 rw_exit(&mip->mi_rw_lock);
2157 }
2158
2159 /*
2160 * Set up the data path. Called from i_mac_unicast_add after having
2161 * done all the validations including making sure this is an active
2162 * client (i.e that is ready to process packets.)
2163 */
2164 static int
2165 mac_client_datapath_setup(mac_client_impl_t *mcip, uint16_t vid,
2166 uint8_t *mac_addr, mac_resource_props_t *mrp, boolean_t isprimary,
2167 mac_unicast_impl_t *muip)
2168 {
2169 mac_impl_t *mip = mcip->mci_mip;
2170 boolean_t mac_started = B_FALSE;
2171 boolean_t bcast_added = B_FALSE;
2172 boolean_t nactiveclients_added = B_FALSE;
2173 flow_entry_t *flent;
2174 int err = 0;
2175 boolean_t no_unicast;
2176
2177 no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2178
2179 if ((err = mac_start((mac_handle_t)mip)) != 0)
2180 goto bail;
2181
2182 mac_started = B_TRUE;
2183
2184 /* add the MAC client to the broadcast address group by default */
2185 if (mip->mi_type->mt_brdcst_addr != NULL) {
2186 err = mac_bcast_add(mcip, mip->mi_type->mt_brdcst_addr, vid,
2187 MAC_ADDRTYPE_BROADCAST);
2188 if (err != 0)
2189 goto bail;
2190 bcast_added = B_TRUE;
2191 }
2192
2193 /*
2194 * If this is the first unicast address addition for this
2195 * client, reuse the pre-allocated larval flow entry associated with
2196 * the MAC client.
2197 */
2198 flent = (mcip->mci_nflents == 0) ? mcip->mci_flent : NULL;
2199
2200 /* We are configuring the unicast flow now */
2201 if (!MCIP_DATAPATH_SETUP(mcip)) {
2202
2203 if (mrp != NULL) {
2204 MAC_CLIENT_SET_PRIORITY_RANGE(mcip,
2205 (mrp->mrp_mask & MRP_PRIORITY) ? mrp->mrp_priority :
2206 MPL_LINK_DEFAULT);
2207 }
2208 if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2209 isprimary, B_TRUE, &flent, mrp)) != 0)
2210 goto bail;
2211
2212 mip->mi_nactiveclients++;
2213 nactiveclients_added = B_TRUE;
2214
2215 /*
2216 * This will allocate the RX ring group if possible for the
2217 * flow and program the software classifier as needed.
2218 */
2219 if ((err = mac_datapath_setup(mcip, flent, SRST_LINK)) != 0)
2220 goto bail;
2221
2222 if (no_unicast)
2223 goto done_setup;
2224 /*
2225 * The unicast MAC address must have been added successfully.
2226 */
2227 ASSERT(mcip->mci_unicast != NULL);
2228 /*
2229 * Push down the sub-flows that were defined on this link
2230 * hitherto. The flows are added to the active flow table
2231 * and SRS, softrings etc. are created as needed.
2232 */
2233 mac_link_init_flows((mac_client_handle_t)mcip);
2234 } else {
2235 mac_address_t *map = mcip->mci_unicast;
2236
2237 ASSERT(!no_unicast);
2238 /*
2239 * A unicast flow already exists for that MAC client,
2240 * this flow must be the same mac address but with
2241 * different VID. It has been checked by mac_addr_in_use().
2242 *
2243 * We will use the SRS etc. from the mci_flent. Note that
2244 * We don't need to create kstat for this as except for
2245 * the fdesc, everything will be used from in the 1st flent.
2246 */
2247
2248 if (bcmp(mac_addr, map->ma_addr, map->ma_len) != 0) {
2249 err = EINVAL;
2250 goto bail;
2251 }
2252
2253 if ((err = mac_unicast_flow_create(mcip, mac_addr, vid,
2254 isprimary, B_FALSE, &flent, NULL)) != 0) {
2255 goto bail;
2256 }
2257 if ((err = mac_flow_add(mip->mi_flow_tab, flent)) != 0) {
2258 FLOW_FINAL_REFRELE(flent);
2259 goto bail;
2260 }
2261
2262 /* update the multicast group for this vid */
2263 mac_client_bcast_refresh(mcip, mac_client_update_mcast,
2264 (void *)flent, B_TRUE);
2265
2266 }
2267
2268 /* populate the shared MAC address */
2269 muip->mui_map = mcip->mci_unicast;
2270
2271 rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2272 muip->mui_next = mcip->mci_unicast_list;
2273 mcip->mci_unicast_list = muip;
2274 rw_exit(&mcip->mci_rw_lock);
2275
2276 done_setup:
2277 /*
2278 * First add the flent to the flow list of this mcip. Then set
2279 * the mip's mi_single_active_client if needed. The Rx path assumes
2280 * that mip->mi_single_active_client will always have an associated
2281 * flent.
2282 */
2283 mac_client_add_to_flow_list(mcip, flent);
2284 if (nactiveclients_added)
2285 mac_update_single_active_client(mip);
2286 /*
2287 * Trigger a renegotiation of the capabilities when the number of
2288 * active clients changes from 1 to 2, since some of the capabilities
2289 * might have to be disabled. Also send a MAC_NOTE_LINK notification
2290 * to all the MAC clients whenever physical link is DOWN.
2291 */
2292 if (mip->mi_nactiveclients == 2) {
2293 mac_capab_update((mac_handle_t)mip);
2294 mac_virtual_link_update(mip);
2295 }
2296 /*
2297 * Now that the setup is complete, clear the INCIPIENT flag.
2298 * The flag was set to avoid incoming packets seeing inconsistent
2299 * structures while the setup was in progress. Clear the mci_tx_flag
2300 * by calling mac_tx_client_block. It is possible that
2301 * mac_unicast_remove was called prior to this mac_unicast_add which
2302 * could have set the MCI_TX_QUIESCE flag.
2303 */
2304 if (flent->fe_rx_ring_group != NULL)
2305 mac_rx_group_unmark(flent->fe_rx_ring_group, MR_INCIPIENT);
2306 FLOW_UNMARK(flent, FE_INCIPIENT);
2307 FLOW_UNMARK(flent, FE_MC_NO_DATAPATH);
2308 mac_tx_client_unblock(mcip);
2309 return (0);
2310 bail:
2311 if (bcast_added)
2312 mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr, vid);
2313
2314 if (nactiveclients_added)
2315 mip->mi_nactiveclients--;
2316
2317 if (mac_started)
2318 mac_stop((mac_handle_t)mip);
2319
2320 return (err);
2321 }
2322
2323 /*
2324 * Return the passive primary MAC client, if present. The passive client is
2325 * a stand-by client that has the same unicast address as another that is
2326 * currenly active. Once the active client goes away, the passive client
2327 * becomes active.
2328 */
2329 static mac_client_impl_t *
2330 mac_get_passive_primary_client(mac_impl_t *mip)
2331 {
2332 mac_client_impl_t *mcip;
2333
2334 for (mcip = mip->mi_clients_list; mcip != NULL;
2335 mcip = mcip->mci_client_next) {
2336 if (mac_is_primary_client(mcip) &&
2337 (mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2338 return (mcip);
2339 }
2340 }
2341 return (NULL);
2342 }
2343
2344 /*
2345 * Add a new unicast address to the MAC client.
2346 *
2347 * The MAC address can be specified either by value, or the MAC client
2348 * can specify that it wants to use the primary MAC address of the
2349 * underlying MAC. See the introductory comments at the beginning
2350 * of this file for more information on primary MAC addresses.
2351 *
2352 * Note also the tuple (MAC address, VID) must be unique
2353 * for the MAC clients defined on top of the same underlying MAC
2354 * instance, unless the MAC_UNICAST_NODUPCHECK is specified.
2355 *
2356 * In no case can a client use the PVID for the MAC, if the MAC has one set.
2357 */
2358 int
2359 i_mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2360 mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2361 {
2362 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2363 mac_impl_t *mip = mcip->mci_mip;
2364 int err;
2365 uint_t mac_len = mip->mi_type->mt_addr_length;
2366 boolean_t check_dups = !(flags & MAC_UNICAST_NODUPCHECK);
2367 boolean_t fastpath_disabled = B_FALSE;
2368 boolean_t is_primary = (flags & MAC_UNICAST_PRIMARY);
2369 boolean_t is_unicast_hw = (flags & MAC_UNICAST_HW);
2370 mac_resource_props_t *mrp;
2371 boolean_t passive_client = B_FALSE;
2372 mac_unicast_impl_t *muip;
2373 boolean_t is_vnic_primary =
2374 (flags & MAC_UNICAST_VNIC_PRIMARY);
2375
2376 /* when VID is non-zero, the underlying MAC can not be VNIC */
2377 ASSERT(!((mip->mi_state_flags & MIS_IS_VNIC) && (vid != 0)));
2378
2379 /*
2380 * Can't unicast add if the client asked only for minimal datapath
2381 * setup.
2382 */
2383 if (mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR)
2384 return (ENOTSUP);
2385
2386 /*
2387 * Check for an attempted use of the current Port VLAN ID, if enabled.
2388 * No client may use it.
2389 */
2390 if (mip->mi_pvid != 0 && vid == mip->mi_pvid)
2391 return (EBUSY);
2392
2393 /*
2394 * Check whether it's the primary client and flag it.
2395 */
2396 if (!(mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary && vid == 0)
2397 mcip->mci_flags |= MAC_CLIENT_FLAGS_PRIMARY;
2398
2399 /*
2400 * is_vnic_primary is true when we come here as a VLAN VNIC
2401 * which uses the primary mac client's address but with a non-zero
2402 * VID. In this case the MAC address is not specified by an upper
2403 * MAC client.
2404 */
2405 if ((mcip->mci_state_flags & MCIS_IS_VNIC) && is_primary &&
2406 !is_vnic_primary) {
2407 /*
2408 * The address is being set by the upper MAC client
2409 * of a VNIC. The MAC address was already set by the
2410 * VNIC driver during VNIC creation.
2411 *
2412 * Note: a VNIC has only one MAC address. We return
2413 * the MAC unicast address handle of the lower MAC client
2414 * corresponding to the VNIC. We allocate a new entry
2415 * which is flagged appropriately, so that mac_unicast_remove()
2416 * doesn't attempt to free the original entry that
2417 * was allocated by the VNIC driver.
2418 */
2419 ASSERT(mcip->mci_unicast != NULL);
2420
2421 /* Check for VLAN flags, if present */
2422 if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2423 mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2424
2425 if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2426 mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2427
2428 if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2429 mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2430
2431 /*
2432 * Ensure that the primary unicast address of the VNIC
2433 * is added only once unless we have the
2434 * MAC_CLIENT_FLAGS_MULTI_PRIMARY set (and this is not
2435 * a passive MAC client).
2436 */
2437 if ((mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) != 0) {
2438 if ((mcip->mci_flags &
2439 MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2440 (mcip->mci_flags &
2441 MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2442 return (EBUSY);
2443 }
2444 mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2445 passive_client = B_TRUE;
2446 }
2447
2448 mcip->mci_flags |= MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2449
2450 /*
2451 * Create a handle for vid 0.
2452 */
2453 ASSERT(vid == 0);
2454 muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2455 muip->mui_vid = vid;
2456 *mah = (mac_unicast_handle_t)muip;
2457 /*
2458 * This will be used by the caller to defer setting the
2459 * rx functions.
2460 */
2461 if (passive_client)
2462 return (EAGAIN);
2463 return (0);
2464 }
2465
2466 /* primary MAC clients cannot be opened on top of anchor VNICs */
2467 if ((is_vnic_primary || is_primary) &&
2468 i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_ANCHOR_VNIC, NULL)) {
2469 return (ENXIO);
2470 }
2471
2472 /*
2473 * If this is a VNIC/VLAN, disable softmac fast-path.
2474 */
2475 if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2476 err = mac_fastpath_disable((mac_handle_t)mip);
2477 if (err != 0)
2478 return (err);
2479 fastpath_disabled = B_TRUE;
2480 }
2481
2482 /*
2483 * Return EBUSY if:
2484 * - there is an exclusively active mac client exists.
2485 * - this is an exclusive active mac client but
2486 * a. there is already active mac clients exist, or
2487 * b. fastpath streams are already plumbed on this legacy device
2488 * - the mac creator has disallowed active mac clients.
2489 */
2490 if (mip->mi_state_flags & (MIS_EXCLUSIVE|MIS_NO_ACTIVE)) {
2491 if (fastpath_disabled)
2492 mac_fastpath_enable((mac_handle_t)mip);
2493 return (EBUSY);
2494 }
2495
2496 if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2497 ASSERT(!fastpath_disabled);
2498 if (mip->mi_nactiveclients != 0)
2499 return (EBUSY);
2500
2501 if ((mip->mi_state_flags & MIS_LEGACY) &&
2502 !(mip->mi_capab_legacy.ml_active_set(mip->mi_driver))) {
2503 return (EBUSY);
2504 }
2505 mip->mi_state_flags |= MIS_EXCLUSIVE;
2506 }
2507
2508 mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
2509 if (is_primary && !(mcip->mci_state_flags & (MCIS_IS_VNIC |
2510 MCIS_IS_AGGR_PORT))) {
2511 /*
2512 * Apply the property cached in the mac_impl_t to the primary
2513 * mac client. If the mac client is a VNIC or an aggregation
2514 * port, its property should be set in the mcip when the
2515 * VNIC/aggr was created.
2516 */
2517 mac_get_resources((mac_handle_t)mip, mrp);
2518 (void) mac_client_set_resources(mch, mrp);
2519 } else if (mcip->mci_state_flags & MCIS_IS_VNIC) {
2520 /*
2521 * This is a primary VLAN client, we don't support
2522 * specifying rings property for this as it inherits the
2523 * rings property from its MAC.
2524 */
2525 if (is_vnic_primary) {
2526 mac_resource_props_t *vmrp;
2527
2528 vmrp = MCIP_RESOURCE_PROPS(mcip);
2529 if (vmrp->mrp_mask & MRP_RX_RINGS ||
2530 vmrp->mrp_mask & MRP_TX_RINGS) {
2531 if (fastpath_disabled)
2532 mac_fastpath_enable((mac_handle_t)mip);
2533 kmem_free(mrp, sizeof (*mrp));
2534 return (ENOTSUP);
2535 }
2536 /*
2537 * Additionally we also need to inherit any
2538 * rings property from the MAC.
2539 */
2540 mac_get_resources((mac_handle_t)mip, mrp);
2541 if (mrp->mrp_mask & MRP_RX_RINGS) {
2542 vmrp->mrp_mask |= MRP_RX_RINGS;
2543 vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
2544 }
2545 if (mrp->mrp_mask & MRP_TX_RINGS) {
2546 vmrp->mrp_mask |= MRP_TX_RINGS;
2547 vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
2548 }
2549 }
2550 bcopy(MCIP_RESOURCE_PROPS(mcip), mrp, sizeof (*mrp));
2551 }
2552
2553 muip = kmem_zalloc(sizeof (mac_unicast_impl_t), KM_SLEEP);
2554 muip->mui_vid = vid;
2555
2556 if (is_primary || is_vnic_primary) {
2557 mac_addr = mip->mi_addr;
2558 } else {
2559
2560 /*
2561 * Verify the validity of the specified MAC addresses value.
2562 */
2563 if (!mac_unicst_verify((mac_handle_t)mip, mac_addr, mac_len)) {
2564 *diag = MAC_DIAG_MACADDR_INVALID;
2565 err = EINVAL;
2566 goto bail_out;
2567 }
2568
2569 /*
2570 * Make sure that the specified MAC address is different
2571 * than the unicast MAC address of the underlying NIC.
2572 */
2573 if (check_dups && bcmp(mip->mi_addr, mac_addr, mac_len) == 0) {
2574 *diag = MAC_DIAG_MACADDR_NIC;
2575 err = EINVAL;
2576 goto bail_out;
2577 }
2578 }
2579
2580 /*
2581 * Set the flags here so that if this is a passive client, we
2582 * can return and set it when we call mac_client_datapath_setup
2583 * when this becomes the active client. If we defer to using these
2584 * flags to mac_client_datapath_setup, then for a passive client,
2585 * we'd have to store the flags somewhere (probably fe_flags)
2586 * and then use it.
2587 */
2588 if (!MCIP_DATAPATH_SETUP(mcip)) {
2589 if (is_unicast_hw) {
2590 /*
2591 * The client requires a hardware MAC address slot
2592 * for that unicast address. Since we support only
2593 * one unicast MAC address per client, flag the
2594 * MAC client itself.
2595 */
2596 mcip->mci_state_flags |= MCIS_UNICAST_HW;
2597 }
2598
2599 /* Check for VLAN flags, if present */
2600 if ((flags & MAC_UNICAST_TAG_DISABLE) != 0)
2601 mcip->mci_state_flags |= MCIS_TAG_DISABLE;
2602
2603 if ((flags & MAC_UNICAST_STRIP_DISABLE) != 0)
2604 mcip->mci_state_flags |= MCIS_STRIP_DISABLE;
2605
2606 if ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0)
2607 mcip->mci_state_flags |= MCIS_DISABLE_TX_VID_CHECK;
2608 } else {
2609 /*
2610 * Assert that the specified flags are consistent with the
2611 * flags specified by previous calls to mac_unicast_add().
2612 */
2613 ASSERT(((flags & MAC_UNICAST_TAG_DISABLE) != 0 &&
2614 (mcip->mci_state_flags & MCIS_TAG_DISABLE) != 0) ||
2615 ((flags & MAC_UNICAST_TAG_DISABLE) == 0 &&
2616 (mcip->mci_state_flags & MCIS_TAG_DISABLE) == 0));
2617
2618 ASSERT(((flags & MAC_UNICAST_STRIP_DISABLE) != 0 &&
2619 (mcip->mci_state_flags & MCIS_STRIP_DISABLE) != 0) ||
2620 ((flags & MAC_UNICAST_STRIP_DISABLE) == 0 &&
2621 (mcip->mci_state_flags & MCIS_STRIP_DISABLE) == 0));
2622
2623 ASSERT(((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) != 0 &&
2624 (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) != 0) ||
2625 ((flags & MAC_UNICAST_DISABLE_TX_VID_CHECK) == 0 &&
2626 (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK) == 0));
2627
2628 /*
2629 * Make sure the client is consistent about its requests
2630 * for MAC addresses. I.e. all requests from the clients
2631 * must have the MAC_UNICAST_HW flag set or clear.
2632 */
2633 if ((mcip->mci_state_flags & MCIS_UNICAST_HW) != 0 &&
2634 !is_unicast_hw ||
2635 (mcip->mci_state_flags & MCIS_UNICAST_HW) == 0 &&
2636 is_unicast_hw) {
2637 err = EINVAL;
2638 goto bail_out;
2639 }
2640 }
2641 /*
2642 * Make sure the MAC address is not already used by
2643 * another MAC client defined on top of the same
2644 * underlying NIC. Unless we have MAC_CLIENT_FLAGS_MULTI_PRIMARY
2645 * set when we allow a passive client to be present which will
2646 * be activated when the currently active client goes away - this
2647 * works only with primary addresses.
2648 */
2649 if ((check_dups || is_primary || is_vnic_primary) &&
2650 mac_addr_in_use(mip, mac_addr, vid)) {
2651 /*
2652 * Must have set the multiple primary address flag when
2653 * we did a mac_client_open AND this should be a primary
2654 * MAC client AND there should not already be a passive
2655 * primary. If all is true then we let this succeed
2656 * even if the address is a dup.
2657 */
2658 if ((mcip->mci_flags & MAC_CLIENT_FLAGS_MULTI_PRIMARY) == 0 ||
2659 (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) == 0 ||
2660 mac_get_passive_primary_client(mip) != NULL) {
2661 *diag = MAC_DIAG_MACADDR_INUSE;
2662 err = EEXIST;
2663 goto bail_out;
2664 }
2665 ASSERT((mcip->mci_flags &
2666 MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) == 0);
2667 mcip->mci_flags |= MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2668 kmem_free(mrp, sizeof (*mrp));
2669
2670 /*
2671 * Stash the unicast address handle, we will use it when
2672 * we set up the passive client.
2673 */
2674 mcip->mci_p_unicast_list = muip;
2675 *mah = (mac_unicast_handle_t)muip;
2676 return (0);
2677 }
2678
2679 err = mac_client_datapath_setup(mcip, vid, mac_addr, mrp,
2680 is_primary || is_vnic_primary, muip);
2681 if (err != 0)
2682 goto bail_out;
2683
2684 kmem_free(mrp, sizeof (*mrp));
2685 *mah = (mac_unicast_handle_t)muip;
2686 return (0);
2687
2688 bail_out:
2689 if (fastpath_disabled)
2690 mac_fastpath_enable((mac_handle_t)mip);
2691 if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2692 mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2693 if (mip->mi_state_flags & MIS_LEGACY) {
2694 mip->mi_capab_legacy.ml_active_clear(
2695 mip->mi_driver);
2696 }
2697 }
2698 kmem_free(mrp, sizeof (*mrp));
2699 kmem_free(muip, sizeof (mac_unicast_impl_t));
2700 return (err);
2701 }
2702
2703 /*
2704 * Wrapper function to mac_unicast_add when we want to have the same mac
2705 * client open for two instances, one that is currently active and another
2706 * that will become active when the current one is removed. In this case
2707 * mac_unicast_add will return EGAIN and we will save the rx function and
2708 * arg which will be used when we activate the passive client in
2709 * mac_unicast_remove.
2710 */
2711 int
2712 mac_unicast_add_set_rx(mac_client_handle_t mch, uint8_t *mac_addr,
2713 uint16_t flags, mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag,
2714 mac_rx_t rx_fn, void *arg)
2715 {
2716 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2717 uint_t err;
2718
2719 err = mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2720 if (err != 0 && err != EAGAIN)
2721 return (err);
2722 if (err == EAGAIN) {
2723 if (rx_fn != NULL) {
2724 mcip->mci_rx_p_fn = rx_fn;
2725 mcip->mci_rx_p_arg = arg;
2726 }
2727 return (0);
2728 }
2729 if (rx_fn != NULL)
2730 mac_rx_set(mch, rx_fn, arg);
2731 return (err);
2732 }
2733
2734 int
2735 mac_unicast_add(mac_client_handle_t mch, uint8_t *mac_addr, uint16_t flags,
2736 mac_unicast_handle_t *mah, uint16_t vid, mac_diag_t *diag)
2737 {
2738 mac_impl_t *mip = ((mac_client_impl_t *)mch)->mci_mip;
2739 uint_t err;
2740
2741 i_mac_perim_enter(mip);
2742 err = i_mac_unicast_add(mch, mac_addr, flags, mah, vid, diag);
2743 i_mac_perim_exit(mip);
2744
2745 return (err);
2746 }
2747
2748 static void
2749 mac_client_datapath_teardown(mac_client_handle_t mch, mac_unicast_impl_t *muip,
2750 flow_entry_t *flent)
2751 {
2752 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2753 mac_impl_t *mip = mcip->mci_mip;
2754 boolean_t no_unicast;
2755
2756 /*
2757 * If we have not added a unicast address for this MAC client, just
2758 * teardown the datapath.
2759 */
2760 no_unicast = mcip->mci_state_flags & MCIS_NO_UNICAST_ADDR;
2761
2762 if (!no_unicast) {
2763 /*
2764 * We would have initialized subflows etc. only if we brought
2765 * up the primary client and set the unicast unicast address
2766 * etc. Deactivate the flows. The flow entry will be removed
2767 * from the active flow tables, and the associated SRS,
2768 * softrings etc will be deleted. But the flow entry itself
2769 * won't be destroyed, instead it will continue to be archived
2770 * off the the global flow hash list, for a possible future
2771 * activation when say IP is plumbed again.
2772 */
2773 mac_link_release_flows(mch);
2774 }
2775 mip->mi_nactiveclients--;
2776 mac_update_single_active_client(mip);
2777
2778 /* Tear down the data path */
2779 mac_datapath_teardown(mcip, mcip->mci_flent, SRST_LINK);
2780
2781 /*
2782 * Prevent any future access to the flow entry through the mci_flent
2783 * pointer by setting the mci_flent to NULL. Access to mci_flent in
2784 * mac_bcast_send is also under mi_rw_lock.
2785 */
2786 rw_enter(&mip->mi_rw_lock, RW_WRITER);
2787 flent = mcip->mci_flent;
2788 mac_client_remove_flow_from_list(mcip, flent);
2789
2790 if (mcip->mci_state_flags & MCIS_DESC_LOGGED)
2791 mcip->mci_state_flags &= ~MCIS_DESC_LOGGED;
2792
2793 /*
2794 * This is the last unicast address being removed and there shouldn't
2795 * be any outbound data threads at this point coming down from mac
2796 * clients. We have waited for the data threads to finish before
2797 * starting dld_str_detach. Non-data threads must access TX SRS
2798 * under mi_rw_lock.
2799 */
2800 rw_exit(&mip->mi_rw_lock);
2801
2802 /*
2803 * Don't use FLOW_MARK with FE_MC_NO_DATAPATH, as the flow might
2804 * contain other flags, such as FE_CONDEMNED, which we need to
2805 * cleared. We don't call mac_flow_cleanup() for this unicast
2806 * flow as we have a already cleaned up SRSs etc. (via the teadown
2807 * path). We just clear the stats and reset the initial callback
2808 * function, the rest will be set when we call mac_flow_create,
2809 * if at all.
2810 */
2811 mutex_enter(&flent->fe_lock);
2812 ASSERT(flent->fe_refcnt == 1 && flent->fe_mbg == NULL &&
2813 flent->fe_tx_srs == NULL && flent->fe_rx_srs_cnt == 0);
2814 flent->fe_flags = FE_MC_NO_DATAPATH;
2815 flow_stat_destroy(flent);
2816 mac_misc_stat_delete(flent);
2817
2818 /* Initialize the receiver function to a safe routine */
2819 flent->fe_cb_fn = (flow_fn_t)mac_pkt_drop;
2820 flent->fe_cb_arg1 = NULL;
2821 flent->fe_cb_arg2 = NULL;
2822
2823 flent->fe_index = -1;
2824 mutex_exit(&flent->fe_lock);
2825
2826 if (mip->mi_type->mt_brdcst_addr != NULL) {
2827 ASSERT(muip != NULL || no_unicast);
2828 mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
2829 muip != NULL ? muip->mui_vid : VLAN_ID_NONE);
2830 }
2831
2832 if (mip->mi_nactiveclients == 1) {
2833 mac_capab_update((mac_handle_t)mip);
2834 mac_virtual_link_update(mip);
2835 }
2836
2837 if (mcip->mci_state_flags & MCIS_EXCLUSIVE) {
2838 mip->mi_state_flags &= ~MIS_EXCLUSIVE;
2839
2840 if (mip->mi_state_flags & MIS_LEGACY)
2841 mip->mi_capab_legacy.ml_active_clear(mip->mi_driver);
2842 }
2843
2844 mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2845
2846 if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2847 mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2848
2849 if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2850 mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2851
2852 if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2853 mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2854
2855 if (muip != NULL)
2856 kmem_free(muip, sizeof (mac_unicast_impl_t));
2857 mac_protect_cancel_timer(mcip);
2858 mac_protect_flush_dhcp(mcip);
2859
2860 bzero(&mcip->mci_misc_stat, sizeof (mcip->mci_misc_stat));
2861 /*
2862 * Disable fastpath if this is a VNIC or a VLAN.
2863 */
2864 if (mcip->mci_state_flags & MCIS_IS_VNIC)
2865 mac_fastpath_enable((mac_handle_t)mip);
2866 mac_stop((mac_handle_t)mip);
2867 }
2868
2869 /*
2870 * Remove a MAC address which was previously added by mac_unicast_add().
2871 */
2872 int
2873 mac_unicast_remove(mac_client_handle_t mch, mac_unicast_handle_t mah)
2874 {
2875 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
2876 mac_unicast_impl_t *muip = (mac_unicast_impl_t *)mah;
2877 mac_unicast_impl_t *pre;
2878 mac_impl_t *mip = mcip->mci_mip;
2879 flow_entry_t *flent;
2880 uint16_t mui_vid;
2881
2882 i_mac_perim_enter(mip);
2883 if (mcip->mci_flags & MAC_CLIENT_FLAGS_VNIC_PRIMARY) {
2884 /*
2885 * Called made by the upper MAC client of a VNIC.
2886 * There's nothing much to do, the unicast address will
2887 * be removed by the VNIC driver when the VNIC is deleted,
2888 * but let's ensure that all our transmit is done before
2889 * the client does a mac_client_stop lest it trigger an
2890 * assert in the driver.
2891 */
2892 ASSERT(muip->mui_vid == 0);
2893
2894 mac_tx_client_flush(mcip);
2895
2896 if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2897 mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2898 if (mcip->mci_rx_p_fn != NULL) {
2899 mac_rx_set(mch, mcip->mci_rx_p_fn,
2900 mcip->mci_rx_p_arg);
2901 mcip->mci_rx_p_fn = NULL;
2902 mcip->mci_rx_p_arg = NULL;
2903 }
2904 kmem_free(muip, sizeof (mac_unicast_impl_t));
2905 i_mac_perim_exit(mip);
2906 return (0);
2907 }
2908 mcip->mci_flags &= ~MAC_CLIENT_FLAGS_VNIC_PRIMARY;
2909
2910 if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2911 mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2912
2913 if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2914 mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2915
2916 if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2917 mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2918
2919 kmem_free(muip, sizeof (mac_unicast_impl_t));
2920 i_mac_perim_exit(mip);
2921 return (0);
2922 }
2923
2924 ASSERT(muip != NULL);
2925
2926 /*
2927 * We are removing a passive client, we haven't setup the datapath
2928 * for this yet, so nothing much to do.
2929 */
2930 if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PASSIVE_PRIMARY) != 0) {
2931
2932 ASSERT((mcip->mci_flent->fe_flags & FE_MC_NO_DATAPATH) != 0);
2933 ASSERT(mcip->mci_p_unicast_list == muip);
2934
2935 mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
2936
2937 mcip->mci_p_unicast_list = NULL;
2938 mcip->mci_rx_p_fn = NULL;
2939 mcip->mci_rx_p_arg = NULL;
2940
2941 mcip->mci_state_flags &= ~MCIS_UNICAST_HW;
2942
2943 if (mcip->mci_state_flags & MCIS_TAG_DISABLE)
2944 mcip->mci_state_flags &= ~MCIS_TAG_DISABLE;
2945
2946 if (mcip->mci_state_flags & MCIS_STRIP_DISABLE)
2947 mcip->mci_state_flags &= ~MCIS_STRIP_DISABLE;
2948
2949 if (mcip->mci_state_flags & MCIS_DISABLE_TX_VID_CHECK)
2950 mcip->mci_state_flags &= ~MCIS_DISABLE_TX_VID_CHECK;
2951
2952 kmem_free(muip, sizeof (mac_unicast_impl_t));
2953 i_mac_perim_exit(mip);
2954 return (0);
2955 }
2956 /*
2957 * Remove the VID from the list of client's VIDs.
2958 */
2959 pre = mcip->mci_unicast_list;
2960 if (muip == pre) {
2961 mcip->mci_unicast_list = muip->mui_next;
2962 } else {
2963 while ((pre->mui_next != NULL) && (pre->mui_next != muip))
2964 pre = pre->mui_next;
2965 ASSERT(pre->mui_next == muip);
2966 rw_enter(&mcip->mci_rw_lock, RW_WRITER);
2967 pre->mui_next = muip->mui_next;
2968 rw_exit(&mcip->mci_rw_lock);
2969 }
2970
2971 if (!mac_client_single_rcvr(mcip)) {
2972 /*
2973 * This MAC client is shared by more than one unicast
2974 * addresses, so we will just remove the flent
2975 * corresponding to the address being removed. We don't invoke
2976 * mac_rx_classify_flow_rem() since the additional flow is
2977 * not associated with its own separate set of SRS and rings,
2978 * and these constructs are still needed for the remaining
2979 * flows.
2980 */
2981 flent = mac_client_get_flow(mcip, muip);
2982 ASSERT(flent != NULL);
2983
2984 /*
2985 * The first one is disappearing, need to make sure
2986 * we replace it with another from the list of
2987 * shared clients.
2988 */
2989 if (flent == mcip->mci_flent)
2990 flent = mac_client_swap_mciflent(mcip);
2991 mac_client_remove_flow_from_list(mcip, flent);
2992 mac_flow_remove(mip->mi_flow_tab, flent, B_FALSE);
2993 mac_flow_wait(flent, FLOW_DRIVER_UPCALL);
2994
2995 /*
2996 * The multicast groups that were added by the client so
2997 * far must be removed from the brodcast domain corresponding
2998 * to the VID being removed.
2999 */
3000 mac_client_bcast_refresh(mcip, mac_client_update_mcast,
3001 (void *)flent, B_FALSE);
3002
3003 if (mip->mi_type->mt_brdcst_addr != NULL) {
3004 mac_bcast_delete(mcip, mip->mi_type->mt_brdcst_addr,
3005 muip->mui_vid);
3006 }
3007
3008 FLOW_FINAL_REFRELE(flent);
3009 ASSERT(!(mcip->mci_state_flags & MCIS_EXCLUSIVE));
3010 /*
3011 * Enable fastpath if this is a VNIC or a VLAN.
3012 */
3013 if (mcip->mci_state_flags & MCIS_IS_VNIC)
3014 mac_fastpath_enable((mac_handle_t)mip);
3015 mac_stop((mac_handle_t)mip);
3016 i_mac_perim_exit(mip);
3017 return (0);
3018 }
3019
3020 mui_vid = muip->mui_vid;
3021 mac_client_datapath_teardown(mch, muip, flent);
3022
3023 if ((mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY) && mui_vid == 0) {
3024 mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PRIMARY;
3025 } else {
3026 i_mac_perim_exit(mip);
3027 return (0);
3028 }
3029
3030 /*
3031 * If we are removing the primary, check if we have a passive primary
3032 * client that we need to activate now.
3033 */
3034 mcip = mac_get_passive_primary_client(mip);
3035 if (mcip != NULL) {
3036 mac_resource_props_t *mrp;
3037 mac_unicast_impl_t *muip;
3038
3039 mcip->mci_flags &= ~MAC_CLIENT_FLAGS_PASSIVE_PRIMARY;
3040 mrp = kmem_zalloc(sizeof (*mrp), KM_SLEEP);
3041
3042 /*
3043 * Apply the property cached in the mac_impl_t to the
3044 * primary mac client.
3045 */
3046 mac_get_resources((mac_handle_t)mip, mrp);
3047 (void) mac_client_set_resources(mch, mrp);
3048 ASSERT(mcip->mci_p_unicast_list != NULL);
3049 muip = mcip->mci_p_unicast_list;
3050 mcip->mci_p_unicast_list = NULL;
3051 if (mac_client_datapath_setup(mcip, VLAN_ID_NONE,
3052 mip->mi_addr, mrp, B_TRUE, muip) == 0) {
3053 if (mcip->mci_rx_p_fn != NULL) {
3054 mac_rx_set(mch, mcip->mci_rx_p_fn,
3055 mcip->mci_rx_p_arg);
3056 mcip->mci_rx_p_fn = NULL;
3057 mcip->mci_rx_p_arg = NULL;
3058 }
3059 } else {
3060 kmem_free(muip, sizeof (mac_unicast_impl_t));
3061 }
3062 kmem_free(mrp, sizeof (*mrp));
3063 }
3064 i_mac_perim_exit(mip);
3065 return (0);
3066 }
3067
3068 /*
3069 * Multicast add function invoked by MAC clients.
3070 */
3071 int
3072 mac_multicast_add(mac_client_handle_t mch, const uint8_t *addr)
3073 {
3074 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3075 mac_impl_t *mip = mcip->mci_mip;
3076 flow_entry_t *flent = mcip->mci_flent_list;
3077 flow_entry_t *prev_fe = NULL;
3078 uint16_t vid;
3079 int err = 0;
3080
3081 /* Verify the address is a valid multicast address */
3082 if ((err = mip->mi_type->mt_ops.mtops_multicst_verify(addr,
3083 mip->mi_pdata)) != 0)
3084 return (err);
3085
3086 i_mac_perim_enter(mip);
3087 while (flent != NULL) {
3088 vid = i_mac_flow_vid(flent);
3089
3090 err = mac_bcast_add((mac_client_impl_t *)mch, addr, vid,
3091 MAC_ADDRTYPE_MULTICAST);
3092 if (err != 0)
3093 break;
3094 prev_fe = flent;
3095 flent = flent->fe_client_next;
3096 }
3097
3098 /*
3099 * If we failed adding, then undo all, rather than partial
3100 * success.
3101 */
3102 if (flent != NULL && prev_fe != NULL) {
3103 flent = mcip->mci_flent_list;
3104 while (flent != prev_fe->fe_client_next) {
3105 vid = i_mac_flow_vid(flent);
3106 mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3107 flent = flent->fe_client_next;
3108 }
3109 }
3110 i_mac_perim_exit(mip);
3111 return (err);
3112 }
3113
3114 /*
3115 * Multicast delete function invoked by MAC clients.
3116 */
3117 void
3118 mac_multicast_remove(mac_client_handle_t mch, const uint8_t *addr)
3119 {
3120 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3121 mac_impl_t *mip = mcip->mci_mip;
3122 flow_entry_t *flent;
3123 uint16_t vid;
3124
3125 i_mac_perim_enter(mip);
3126 for (flent = mcip->mci_flent_list; flent != NULL;
3127 flent = flent->fe_client_next) {
3128 vid = i_mac_flow_vid(flent);
3129 mac_bcast_delete((mac_client_impl_t *)mch, addr, vid);
3130 }
3131 i_mac_perim_exit(mip);
3132 }
3133
3134 /*
3135 * When a MAC client desires to capture packets on an interface,
3136 * it registers a promiscuous call back with mac_promisc_add().
3137 * There are three types of promiscuous callbacks:
3138 *
3139 * * MAC_CLIENT_PROMISC_ALL
3140 * Captures all packets sent and received by the MAC client,
3141 * the physical interface, as well as all other MAC clients
3142 * defined on top of the same MAC.
3143 *
3144 * * MAC_CLIENT_PROMISC_FILTERED
3145 * Captures all packets sent and received by the MAC client,
3146 * plus all multicast traffic sent and received by the phyisical
3147 * interface and the other MAC clients.
3148 *
3149 * * MAC_CLIENT_PROMISC_MULTI
3150 * Captures all broadcast and multicast packets sent and
3151 * received by the MAC clients as well as the physical interface.
3152 *
3153 * In all cases, the underlying MAC is put in promiscuous mode.
3154 */
3155 int
3156 mac_promisc_add(mac_client_handle_t mch, mac_client_promisc_type_t type,
3157 mac_rx_t fn, void *arg, mac_promisc_handle_t *mphp, uint16_t flags)
3158 {
3159 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3160 mac_impl_t *mip = mcip->mci_mip;
3161 mac_promisc_impl_t *mpip;
3162 mac_cb_info_t *mcbi;
3163 int rc;
3164
3165 i_mac_perim_enter(mip);
3166
3167 if ((rc = mac_start((mac_handle_t)mip)) != 0) {
3168 i_mac_perim_exit(mip);
3169 return (rc);
3170 }
3171
3172 if ((mcip->mci_state_flags & MCIS_IS_VNIC) &&
3173 type == MAC_CLIENT_PROMISC_ALL) {
3174 /*
3175 * The function is being invoked by the upper MAC client
3176 * of a VNIC. The VNIC should only see the traffic
3177 * it is entitled to.
3178 */
3179 type = MAC_CLIENT_PROMISC_FILTERED;
3180 }
3181
3182
3183 /*
3184 * Turn on promiscuous mode for the underlying NIC.
3185 * This is needed even for filtered callbacks which
3186 * expect to receive all multicast traffic on the wire.
3187 *
3188 * Physical promiscuous mode should not be turned on if
3189 * MAC_PROMISC_FLAGS_NO_PHYS is set.
3190 */
3191 if ((flags & MAC_PROMISC_FLAGS_NO_PHYS) == 0) {
3192 if ((rc = i_mac_promisc_set(mip, B_TRUE)) != 0) {
3193 mac_stop((mac_handle_t)mip);
3194 i_mac_perim_exit(mip);
3195 return (rc);
3196 }
3197 }
3198
3199 mpip = kmem_cache_alloc(mac_promisc_impl_cache, KM_SLEEP);
3200
3201 mpip->mpi_type = type;
3202 mpip->mpi_fn = fn;
3203 mpip->mpi_arg = arg;
3204 mpip->mpi_mcip = mcip;
3205 mpip->mpi_no_tx_loop = ((flags & MAC_PROMISC_FLAGS_NO_TX_LOOP) != 0);
3206 mpip->mpi_no_phys = ((flags & MAC_PROMISC_FLAGS_NO_PHYS) != 0);
3207 mpip->mpi_strip_vlan_tag =
3208 ((flags & MAC_PROMISC_FLAGS_VLAN_TAG_STRIP) != 0);
3209 mpip->mpi_no_copy = ((flags & MAC_PROMISC_FLAGS_NO_COPY) != 0);
3210
3211 mcbi = &mip->mi_promisc_cb_info;
3212 mutex_enter(mcbi->mcbi_lockp);
3213
3214 mac_callback_add(&mip->mi_promisc_cb_info, &mcip->mci_promisc_list,
3215 &mpip->mpi_mci_link);
3216 mac_callback_add(&mip->mi_promisc_cb_info, &mip->mi_promisc_list,
3217 &mpip->mpi_mi_link);
3218
3219 mutex_exit(mcbi->mcbi_lockp);
3220
3221 *mphp = (mac_promisc_handle_t)mpip;
3222 i_mac_perim_exit(mip);
3223 return (0);
3224 }
3225
3226 /*
3227 * Remove a multicast address previously aded through mac_promisc_add().
3228 */
3229 void
3230 mac_promisc_remove(mac_promisc_handle_t mph)
3231 {
3232 mac_promisc_impl_t *mpip = (mac_promisc_impl_t *)mph;
3233 mac_client_impl_t *mcip = mpip->mpi_mcip;
3234 mac_impl_t *mip = mcip->mci_mip;
3235 mac_cb_info_t *mcbi;
3236 int rv;
3237
3238 i_mac_perim_enter(mip);
3239
3240 /*
3241 * Even if the device can't be reset into normal mode, we still
3242 * need to clear the client promisc callbacks. The client may want
3243 * to close the mac end point and we can't have stale callbacks.
3244 */
3245 if (!(mpip->mpi_no_phys)) {
3246 if ((rv = i_mac_promisc_set(mip, B_FALSE)) != 0) {
3247 cmn_err(CE_WARN, "%s: failed to switch OFF promiscuous"
3248 " mode because of error 0x%x", mip->mi_name, rv);
3249 }
3250 }
3251 mcbi = &mip->mi_promisc_cb_info;
3252 mutex_enter(mcbi->mcbi_lockp);
3253 if (mac_callback_remove(mcbi, &mip->mi_promisc_list,
3254 &mpip->mpi_mi_link)) {
3255 VERIFY(mac_callback_remove(&mip->mi_promisc_cb_info,
3256 &mcip->mci_promisc_list, &mpip->mpi_mci_link));
3257 kmem_cache_free(mac_promisc_impl_cache, mpip);
3258 } else {
3259 mac_callback_remove_wait(&mip->mi_promisc_cb_info);
3260 }
3261 mutex_exit(mcbi->mcbi_lockp);
3262 mac_stop((mac_handle_t)mip);
3263
3264 i_mac_perim_exit(mip);
3265 }
3266
3267 /*
3268 * Reference count the number of active Tx threads. MCI_TX_QUIESCE indicates
3269 * that a control operation wants to quiesce the Tx data flow in which case
3270 * we return an error. Holding any of the per cpu locks ensures that the
3271 * mci_tx_flag won't change.
3272 *
3273 * 'CPU' must be accessed just once and used to compute the index into the
3274 * percpu array, and that index must be used for the entire duration of the
3275 * packet send operation. Note that the thread may be preempted and run on
3276 * another cpu any time and so we can't use 'CPU' more than once for the
3277 * operation.
3278 */
3279 #define MAC_TX_TRY_HOLD(mcip, mytx, error) \
3280 { \
3281 (error) = 0; \
3282 (mytx) = &(mcip)->mci_tx_pcpu[CPU->cpu_seqid & mac_tx_percpu_cnt]; \
3283 mutex_enter(&(mytx)->pcpu_tx_lock); \
3284 if (!((mcip)->mci_tx_flag & MCI_TX_QUIESCE)) { \
3285 (mytx)->pcpu_tx_refcnt++; \
3286 } else { \
3287 (error) = -1; \
3288 } \
3289 mutex_exit(&(mytx)->pcpu_tx_lock); \
3290 }
3291
3292 /*
3293 * Release the reference. If needed, signal any control operation waiting
3294 * for Tx quiescence. The wait and signal are always done using the
3295 * mci_tx_pcpu[0]'s lock
3296 */
3297 #define MAC_TX_RELE(mcip, mytx) { \
3298 mutex_enter(&(mytx)->pcpu_tx_lock); \
3299 if (--(mytx)->pcpu_tx_refcnt == 0 && \
3300 (mcip)->mci_tx_flag & MCI_TX_QUIESCE) { \
3301 mutex_exit(&(mytx)->pcpu_tx_lock); \
3302 mutex_enter(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock); \
3303 cv_signal(&(mcip)->mci_tx_cv); \
3304 mutex_exit(&(mcip)->mci_tx_pcpu[0].pcpu_tx_lock); \
3305 } else { \
3306 mutex_exit(&(mytx)->pcpu_tx_lock); \
3307 } \
3308 }
3309
3310 /*
3311 * Send function invoked by MAC clients.
3312 */
3313 mac_tx_cookie_t
3314 mac_tx(mac_client_handle_t mch, mblk_t *mp_chain, uintptr_t hint,
3315 uint16_t flag, mblk_t **ret_mp)
3316 {
3317 mac_tx_cookie_t cookie = NULL;
3318 int error;
3319 mac_tx_percpu_t *mytx;
3320 mac_soft_ring_set_t *srs;
3321 flow_entry_t *flent;
3322 boolean_t is_subflow = B_FALSE;
3323 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3324 mac_impl_t *mip = mcip->mci_mip;
3325 mac_srs_tx_t *srs_tx;
3326
3327 /*
3328 * Check whether the active Tx threads count is bumped already.
3329 */
3330 if (!(flag & MAC_TX_NO_HOLD)) {
3331 MAC_TX_TRY_HOLD(mcip, mytx, error);
3332 if (error != 0) {
3333 freemsgchain(mp_chain);
3334 return (NULL);
3335 }
3336 }
3337
3338 /*
3339 * If mac protection is enabled, only the permissible packets will be
3340 * returned by mac_protect_check().
3341 */
3342 if ((mcip->mci_flent->
3343 fe_resource_props.mrp_mask & MRP_PROTECT) != 0 &&
3344 (mp_chain = mac_protect_check(mch, mp_chain)) == NULL)
3345 goto done;
3346
3347 if (mcip->mci_subflow_tab != NULL &&
3348 mcip->mci_subflow_tab->ft_flow_count > 0 &&
3349 mac_flow_lookup(mcip->mci_subflow_tab, mp_chain,
3350 FLOW_OUTBOUND, &flent) == 0) {
3351 /*
3352 * The main assumption here is that if in the event
3353 * we get a chain, all the packets will be classified
3354 * to the same Flow/SRS. If this changes for any
3355 * reason, the following logic should change as well.
3356 * I suppose the fanout_hint also assumes this .
3357 */
3358 ASSERT(flent != NULL);
3359 is_subflow = B_TRUE;
3360 } else {
3361 flent = mcip->mci_flent;
3362 }
3363
3364 srs = flent->fe_tx_srs;
3365 /*
3366 * This is to avoid panics with PF_PACKET that can call mac_tx()
3367 * against an interface that is not capable of sending. A rewrite
3368 * of the mac datapath is required to remove this limitation.
3369 */
3370 if (srs == NULL) {
3371 freemsgchain(mp_chain);
3372 goto done;
3373 }
3374
3375 srs_tx = &srs->srs_tx;
3376 if (srs_tx->st_mode == SRS_TX_DEFAULT &&
3377 (srs->srs_state & SRS_ENQUEUED) == 0 &&
3378 mip->mi_nactiveclients == 1 && mp_chain->b_next == NULL) {
3379 uint64_t obytes;
3380
3381 /*
3382 * Since dls always opens the underlying MAC, nclients equals
3383 * to 1 means that the only active client is dls itself acting
3384 * as a primary client of the MAC instance. Since dls will not
3385 * send tagged packets in that case, and dls is trusted to send
3386 * packets for its allowed VLAN(s), the VLAN tag insertion and
3387 * check is required only if nclients is greater than 1.
3388 */
3389 if (mip->mi_nclients > 1) {
3390 if (MAC_VID_CHECK_NEEDED(mcip)) {
3391 int err = 0;
3392
3393 MAC_VID_CHECK(mcip, mp_chain, err);
3394 if (err != 0) {
3395 freemsg(mp_chain);
3396 mcip->mci_misc_stat.mms_txerrors++;
3397 goto done;
3398 }
3399 }
3400 if (MAC_TAG_NEEDED(mcip)) {
3401 mp_chain = mac_add_vlan_tag(mp_chain, 0,
3402 mac_client_vid(mch));
3403 if (mp_chain == NULL) {
3404 mcip->mci_misc_stat.mms_txerrors++;
3405 goto done;
3406 }
3407 }
3408 }
3409
3410 obytes = (mp_chain->b_cont == NULL ? MBLKL(mp_chain) :
3411 msgdsize(mp_chain));
3412
3413 MAC_TX(mip, srs_tx->st_arg2, mp_chain, mcip);
3414 if (mp_chain == NULL) {
3415 cookie = NULL;
3416 SRS_TX_STAT_UPDATE(srs, opackets, 1);
3417 SRS_TX_STAT_UPDATE(srs, obytes, obytes);
3418 } else {
3419 mutex_enter(&srs->srs_lock);
3420 cookie = mac_tx_srs_no_desc(srs, mp_chain,
3421 flag, ret_mp);
3422 mutex_exit(&srs->srs_lock);
3423 }
3424 } else {
3425 cookie = srs_tx->st_func(srs, mp_chain, hint, flag, ret_mp);
3426 }
3427
3428 done:
3429 if (is_subflow)
3430 FLOW_REFRELE(flent);
3431
3432 if (!(flag & MAC_TX_NO_HOLD))
3433 MAC_TX_RELE(mcip, mytx);
3434
3435 return (cookie);
3436 }
3437
3438 /*
3439 * mac_tx_is_blocked
3440 *
3441 * Given a cookie, it returns if the ring identified by the cookie is
3442 * flow-controlled or not. If NULL is passed in place of a cookie,
3443 * then it finds out if any of the underlying rings belonging to the
3444 * SRS is flow controlled or not and returns that status.
3445 */
3446 /* ARGSUSED */
3447 boolean_t
3448 mac_tx_is_flow_blocked(mac_client_handle_t mch, mac_tx_cookie_t cookie)
3449 {
3450 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3451 mac_soft_ring_set_t *mac_srs;
3452 mac_soft_ring_t *sringp;
3453 boolean_t blocked = B_FALSE;
3454 mac_tx_percpu_t *mytx;
3455 int err;
3456 int i;
3457
3458 /*
3459 * Bump the reference count so that mac_srs won't be deleted.
3460 * If the client is currently quiesced and we failed to bump
3461 * the reference, return B_TRUE so that flow control stays
3462 * as enabled.
3463 *
3464 * Flow control will then be disabled once the client is no
3465 * longer quiesced.
3466 */
3467 MAC_TX_TRY_HOLD(mcip, mytx, err);
3468 if (err != 0)
3469 return (B_TRUE);
3470
3471 if ((mac_srs = MCIP_TX_SRS(mcip)) == NULL) {
3472 MAC_TX_RELE(mcip, mytx);
3473 return (B_FALSE);
3474 }
3475
3476 mutex_enter(&mac_srs->srs_lock);
3477 /*
3478 * Only in the case of TX_FANOUT and TX_AGGR, the underlying
3479 * softring (s_ring_state) will have the HIWAT set. This is
3480 * the multiple Tx ring flow control case. For all other
3481 * case, SRS (srs_state) will store the condition.
3482 */
3483 if (mac_srs->srs_tx.st_mode == SRS_TX_FANOUT ||
3484 mac_srs->srs_tx.st_mode == SRS_TX_AGGR) {
3485 if (cookie != NULL) {
3486 sringp = (mac_soft_ring_t *)cookie;
3487 mutex_enter(&sringp->s_ring_lock);
3488 if (sringp->s_ring_state & S_RING_TX_HIWAT)
3489 blocked = B_TRUE;
3490 mutex_exit(&sringp->s_ring_lock);
3491 } else {
3492 for (i = 0; i < mac_srs->srs_tx_ring_count; i++) {
3493 sringp = mac_srs->srs_tx_soft_rings[i];
3494 mutex_enter(&sringp->s_ring_lock);
3495 if (sringp->s_ring_state & S_RING_TX_HIWAT) {
3496 blocked = B_TRUE;
3497 mutex_exit(&sringp->s_ring_lock);
3498 break;
3499 }
3500 mutex_exit(&sringp->s_ring_lock);
3501 }
3502 }
3503 } else {
3504 blocked = (mac_srs->srs_state & SRS_TX_HIWAT);
3505 }
3506 mutex_exit(&mac_srs->srs_lock);
3507 MAC_TX_RELE(mcip, mytx);
3508 return (blocked);
3509 }
3510
3511 /*
3512 * Check if the MAC client is the primary MAC client.
3513 */
3514 boolean_t
3515 mac_is_primary_client(mac_client_impl_t *mcip)
3516 {
3517 return (mcip->mci_flags & MAC_CLIENT_FLAGS_PRIMARY);
3518 }
3519
3520 void
3521 mac_ioctl(mac_handle_t mh, queue_t *wq, mblk_t *bp)
3522 {
3523 mac_impl_t *mip = (mac_impl_t *)mh;
3524 int cmd = ((struct iocblk *)bp->b_rptr)->ioc_cmd;
3525
3526 if ((cmd == ND_GET && (mip->mi_callbacks->mc_callbacks & MC_GETPROP)) ||
3527 (cmd == ND_SET && (mip->mi_callbacks->mc_callbacks & MC_SETPROP))) {
3528 /*
3529 * If ndd props were registered, call them.
3530 * Note that ndd ioctls are Obsolete
3531 */
3532 mac_ndd_ioctl(mip, wq, bp);
3533 return;
3534 }
3535
3536 /*
3537 * Call the driver to handle the ioctl. The driver may not support
3538 * any ioctls, in which case we reply with a NAK on its behalf.
3539 */
3540 if (mip->mi_callbacks->mc_callbacks & MC_IOCTL)
3541 mip->mi_ioctl(mip->mi_driver, wq, bp);
3542 else
3543 miocnak(wq, bp, 0, EINVAL);
3544 }
3545
3546 /*
3547 * Return the link state of the specified MAC instance.
3548 */
3549 link_state_t
3550 mac_link_get(mac_handle_t mh)
3551 {
3552 return (((mac_impl_t *)mh)->mi_linkstate);
3553 }
3554
3555 /*
3556 * Add a mac client specified notification callback. Please see the comments
3557 * above mac_callback_add() for general information about mac callback
3558 * addition/deletion in the presence of mac callback list walkers
3559 */
3560 mac_notify_handle_t
3561 mac_notify_add(mac_handle_t mh, mac_notify_t notify_fn, void *arg)
3562 {
3563 mac_impl_t *mip = (mac_impl_t *)mh;
3564 mac_notify_cb_t *mncb;
3565 mac_cb_info_t *mcbi;
3566
3567 /*
3568 * Allocate a notify callback structure, fill in the details and
3569 * use the mac callback list manipulation functions to chain into
3570 * the list of callbacks.
3571 */
3572 mncb = kmem_zalloc(sizeof (mac_notify_cb_t), KM_SLEEP);
3573 mncb->mncb_fn = notify_fn;
3574 mncb->mncb_arg = arg;
3575 mncb->mncb_mip = mip;
3576 mncb->mncb_link.mcb_objp = mncb;
3577 mncb->mncb_link.mcb_objsize = sizeof (mac_notify_cb_t);
3578 mncb->mncb_link.mcb_flags = MCB_NOTIFY_CB_T;
3579
3580 mcbi = &mip->mi_notify_cb_info;
3581
3582 i_mac_perim_enter(mip);
3583 mutex_enter(mcbi->mcbi_lockp);
3584
3585 mac_callback_add(&mip->mi_notify_cb_info, &mip->mi_notify_cb_list,
3586 &mncb->mncb_link);
3587
3588 mutex_exit(mcbi->mcbi_lockp);
3589 i_mac_perim_exit(mip);
3590 return ((mac_notify_handle_t)mncb);
3591 }
3592
3593 void
3594 mac_notify_remove_wait(mac_handle_t mh)
3595 {
3596 mac_impl_t *mip = (mac_impl_t *)mh;
3597 mac_cb_info_t *mcbi = &mip->mi_notify_cb_info;
3598
3599 mutex_enter(mcbi->mcbi_lockp);
3600 mac_callback_remove_wait(&mip->mi_notify_cb_info);
3601 mutex_exit(mcbi->mcbi_lockp);
3602 }
3603
3604 /*
3605 * Remove a mac client specified notification callback
3606 */
3607 int
3608 mac_notify_remove(mac_notify_handle_t mnh, boolean_t wait)
3609 {
3610 mac_notify_cb_t *mncb = (mac_notify_cb_t *)mnh;
3611 mac_impl_t *mip = mncb->mncb_mip;
3612 mac_cb_info_t *mcbi;
3613 int err = 0;
3614
3615 mcbi = &mip->mi_notify_cb_info;
3616
3617 i_mac_perim_enter(mip);
3618 mutex_enter(mcbi->mcbi_lockp);
3619
3620 ASSERT(mncb->mncb_link.mcb_objp == mncb);
3621 /*
3622 * If there aren't any list walkers, the remove would succeed
3623 * inline, else we wait for the deferred remove to complete
3624 */
3625 if (mac_callback_remove(&mip->mi_notify_cb_info,
3626 &mip->mi_notify_cb_list, &mncb->mncb_link)) {
3627 kmem_free(mncb, sizeof (mac_notify_cb_t));
3628 } else {
3629 err = EBUSY;
3630 }
3631
3632 mutex_exit(mcbi->mcbi_lockp);
3633 i_mac_perim_exit(mip);
3634
3635 /*
3636 * If we failed to remove the notification callback and "wait" is set
3637 * to be B_TRUE, wait for the callback to finish after we exit the
3638 * mac perimeter.
3639 */
3640 if (err != 0 && wait) {
3641 mac_notify_remove_wait((mac_handle_t)mip);
3642 return (0);
3643 }
3644
3645 return (err);
3646 }
3647
3648 /*
3649 * Associate resource management callbacks with the specified MAC
3650 * clients.
3651 */
3652
3653 void
3654 mac_resource_set_common(mac_client_handle_t mch, mac_resource_add_t add,
3655 mac_resource_remove_t remove, mac_resource_quiesce_t quiesce,
3656 mac_resource_restart_t restart, mac_resource_bind_t bind,
3657 void *arg)
3658 {
3659 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3660
3661 mcip->mci_resource_add = add;
3662 mcip->mci_resource_remove = remove;
3663 mcip->mci_resource_quiesce = quiesce;
3664 mcip->mci_resource_restart = restart;
3665 mcip->mci_resource_bind = bind;
3666 mcip->mci_resource_arg = arg;
3667 }
3668
3669 void
3670 mac_resource_set(mac_client_handle_t mch, mac_resource_add_t add, void *arg)
3671 {
3672 /* update the 'resource_add' callback */
3673 mac_resource_set_common(mch, add, NULL, NULL, NULL, NULL, arg);
3674 }
3675
3676 /*
3677 * Sets up the client resources and enable the polling interface over all the
3678 * SRS's and the soft rings of the client
3679 */
3680 void
3681 mac_client_poll_enable(mac_client_handle_t mch)
3682 {
3683 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3684 mac_soft_ring_set_t *mac_srs;
3685 flow_entry_t *flent;
3686 int i;
3687
3688 flent = mcip->mci_flent;
3689 ASSERT(flent != NULL);
3690
3691 mcip->mci_state_flags |= MCIS_CLIENT_POLL_CAPABLE;
3692 for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3693 mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3694 ASSERT(mac_srs->srs_mcip == mcip);
3695 mac_srs_client_poll_enable(mcip, mac_srs);
3696 }
3697 }
3698
3699 /*
3700 * Tears down the client resources and disable the polling interface over all
3701 * the SRS's and the soft rings of the client
3702 */
3703 void
3704 mac_client_poll_disable(mac_client_handle_t mch)
3705 {
3706 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3707 mac_soft_ring_set_t *mac_srs;
3708 flow_entry_t *flent;
3709 int i;
3710
3711 flent = mcip->mci_flent;
3712 ASSERT(flent != NULL);
3713
3714 mcip->mci_state_flags &= ~MCIS_CLIENT_POLL_CAPABLE;
3715 for (i = 0; i < flent->fe_rx_srs_cnt; i++) {
3716 mac_srs = (mac_soft_ring_set_t *)flent->fe_rx_srs[i];
3717 ASSERT(mac_srs->srs_mcip == mcip);
3718 mac_srs_client_poll_disable(mcip, mac_srs);
3719 }
3720 }
3721
3722 /*
3723 * Associate the CPUs specified by the given property with a MAC client.
3724 */
3725 int
3726 mac_cpu_set(mac_client_handle_t mch, mac_resource_props_t *mrp)
3727 {
3728 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3729 mac_impl_t *mip = mcip->mci_mip;
3730 int err = 0;
3731
3732 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
3733
3734 if ((err = mac_validate_props(mcip->mci_state_flags & MCIS_IS_VNIC ?
3735 mcip->mci_upper_mip : mip, mrp)) != 0) {
3736 return (err);
3737 }
3738 if (MCIP_DATAPATH_SETUP(mcip))
3739 mac_flow_modify(mip->mi_flow_tab, mcip->mci_flent, mrp);
3740
3741 mac_update_resources(mrp, MCIP_RESOURCE_PROPS(mcip), B_FALSE);
3742 return (0);
3743 }
3744
3745 /*
3746 * Apply the specified properties to the specified MAC client.
3747 */
3748 int
3749 mac_client_set_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3750 {
3751 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3752 mac_impl_t *mip = mcip->mci_mip;
3753 int err = 0;
3754
3755 i_mac_perim_enter(mip);
3756
3757 if ((mrp->mrp_mask & MRP_MAXBW) || (mrp->mrp_mask & MRP_PRIORITY)) {
3758 err = mac_resource_ctl_set(mch, mrp);
3759 if (err != 0)
3760 goto done;
3761 }
3762
3763 if (mrp->mrp_mask & (MRP_CPUS|MRP_POOL)) {
3764 err = mac_cpu_set(mch, mrp);
3765 if (err != 0)
3766 goto done;
3767 }
3768
3769 if (mrp->mrp_mask & MRP_PROTECT) {
3770 err = mac_protect_set(mch, mrp);
3771 if (err != 0)
3772 goto done;
3773 }
3774
3775 if ((mrp->mrp_mask & MRP_RX_RINGS) || (mrp->mrp_mask & MRP_TX_RINGS))
3776 err = mac_resource_ctl_set(mch, mrp);
3777
3778 done:
3779 i_mac_perim_exit(mip);
3780 return (err);
3781 }
3782
3783 /*
3784 * Return the properties currently associated with the specified MAC client.
3785 */
3786 void
3787 mac_client_get_resources(mac_client_handle_t mch, mac_resource_props_t *mrp)
3788 {
3789 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3790 mac_resource_props_t *mcip_mrp = MCIP_RESOURCE_PROPS(mcip);
3791
3792 bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3793 }
3794
3795 /*
3796 * Return the effective properties currently associated with the specified
3797 * MAC client.
3798 */
3799 void
3800 mac_client_get_effective_resources(mac_client_handle_t mch,
3801 mac_resource_props_t *mrp)
3802 {
3803 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
3804 mac_resource_props_t *mcip_mrp = MCIP_EFFECTIVE_PROPS(mcip);
3805
3806 bcopy(mcip_mrp, mrp, sizeof (mac_resource_props_t));
3807 }
3808
3809 /*
3810 * Pass a copy of the specified packet to the promiscuous callbacks
3811 * of the specified MAC.
3812 *
3813 * If sender is NULL, the function is being invoked for a packet chain
3814 * received from the wire. If sender is non-NULL, it points to
3815 * the MAC client from which the packet is being sent.
3816 *
3817 * The packets are distributed to the promiscuous callbacks as follows:
3818 *
3819 * - all packets are sent to the MAC_CLIENT_PROMISC_ALL callbacks
3820 * - all broadcast and multicast packets are sent to the
3821 * MAC_CLIENT_PROMISC_FILTER and MAC_CLIENT_PROMISC_MULTI.
3822 *
3823 * The unicast packets of MAC_CLIENT_PROMISC_FILTER callbacks are dispatched
3824 * after classification by mac_rx_deliver().
3825 */
3826
3827 static void
3828 mac_promisc_dispatch_one(mac_promisc_impl_t *mpip, mblk_t *mp,
3829 boolean_t loopback)
3830 {
3831 mblk_t *mp_copy, *mp_next;
3832
3833 if (!mpip->mpi_no_copy || mpip->mpi_strip_vlan_tag) {
3834 mp_copy = copymsg(mp);
3835 if (mp_copy == NULL)
3836 return;
3837
3838 if (mpip->mpi_strip_vlan_tag) {
3839 mp_copy = mac_strip_vlan_tag_chain(mp_copy);
3840 if (mp_copy == NULL)
3841 return;
3842 }
3843 mp_next = NULL;
3844 } else {
3845 mp_copy = mp;
3846 mp_next = mp->b_next;
3847 }
3848 mp_copy->b_next = NULL;
3849
3850 mpip->mpi_fn(mpip->mpi_arg, NULL, mp_copy, loopback);
3851 if (mp_copy == mp)
3852 mp->b_next = mp_next;
3853 }
3854
3855 /*
3856 * Return the VID of a packet. Zero if the packet is not tagged.
3857 */
3858 static uint16_t
3859 mac_ether_vid(mblk_t *mp)
3860 {
3861 struct ether_header *eth = (struct ether_header *)mp->b_rptr;
3862
3863 if (ntohs(eth->ether_type) == ETHERTYPE_VLAN) {
3864 struct ether_vlan_header *t_evhp =
3865 (struct ether_vlan_header *)mp->b_rptr;
3866 return (VLAN_ID(ntohs(t_evhp->ether_tci)));
3867 }
3868
3869 return (0);
3870 }
3871
3872 /*
3873 * Return whether the specified packet contains a multicast or broadcast
3874 * destination MAC address.
3875 */
3876 static boolean_t
3877 mac_is_mcast(mac_impl_t *mip, mblk_t *mp)
3878 {
3879 mac_header_info_t hdr_info;
3880
3881 if (mac_header_info((mac_handle_t)mip, mp, &hdr_info) != 0)
3882 return (B_FALSE);
3883 return ((hdr_info.mhi_dsttype == MAC_ADDRTYPE_BROADCAST) ||
3884 (hdr_info.mhi_dsttype == MAC_ADDRTYPE_MULTICAST));
3885 }
3886
3887 /*
3888 * Send a copy of an mblk chain to the MAC clients of the specified MAC.
3889 * "sender" points to the sender MAC client for outbound packets, and
3890 * is set to NULL for inbound packets.
3891 */
3892 void
3893 mac_promisc_dispatch(mac_impl_t *mip, mblk_t *mp_chain,
3894 mac_client_impl_t *sender)
3895 {
3896 mac_promisc_impl_t *mpip;
3897 mac_cb_t *mcb;
3898 mblk_t *mp;
3899 boolean_t is_mcast, is_sender;
3900
3901 MAC_PROMISC_WALKER_INC(mip);
3902 for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3903 is_mcast = mac_is_mcast(mip, mp);
3904 /* send packet to interested callbacks */
3905 for (mcb = mip->mi_promisc_list; mcb != NULL;
3906 mcb = mcb->mcb_nextp) {
3907 mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3908 is_sender = (mpip->mpi_mcip == sender);
3909
3910 if (is_sender && mpip->mpi_no_tx_loop)
3911 /*
3912 * The sender doesn't want to receive
3913 * copies of the packets it sends.
3914 */
3915 continue;
3916
3917 /* this client doesn't need any packets (bridge) */
3918 if (mpip->mpi_fn == NULL)
3919 continue;
3920
3921 /*
3922 * For an ethernet MAC, don't displatch a multicast
3923 * packet to a non-PROMISC_ALL callbacks unless the VID
3924 * of the packet matches the VID of the client.
3925 */
3926 if (is_mcast &&
3927 mpip->mpi_type != MAC_CLIENT_PROMISC_ALL &&
3928 !mac_client_check_flow_vid(mpip->mpi_mcip,
3929 mac_ether_vid(mp)))
3930 continue;
3931
3932 if (is_sender ||
3933 mpip->mpi_type == MAC_CLIENT_PROMISC_ALL ||
3934 is_mcast)
3935 mac_promisc_dispatch_one(mpip, mp, is_sender);
3936 }
3937 }
3938 MAC_PROMISC_WALKER_DCR(mip);
3939 }
3940
3941 void
3942 mac_promisc_client_dispatch(mac_client_impl_t *mcip, mblk_t *mp_chain)
3943 {
3944 mac_impl_t *mip = mcip->mci_mip;
3945 mac_promisc_impl_t *mpip;
3946 boolean_t is_mcast;
3947 mblk_t *mp;
3948 mac_cb_t *mcb;
3949
3950 /*
3951 * The unicast packets for the MAC client still
3952 * need to be delivered to the MAC_CLIENT_PROMISC_FILTERED
3953 * promiscuous callbacks. The broadcast and multicast
3954 * packets were delivered from mac_rx().
3955 */
3956 MAC_PROMISC_WALKER_INC(mip);
3957 for (mp = mp_chain; mp != NULL; mp = mp->b_next) {
3958 is_mcast = mac_is_mcast(mip, mp);
3959 for (mcb = mcip->mci_promisc_list; mcb != NULL;
3960 mcb = mcb->mcb_nextp) {
3961 mpip = (mac_promisc_impl_t *)mcb->mcb_objp;
3962 if (mpip->mpi_type == MAC_CLIENT_PROMISC_FILTERED &&
3963 !is_mcast) {
3964 mac_promisc_dispatch_one(mpip, mp, B_FALSE);
3965 }
3966 }
3967 }
3968 MAC_PROMISC_WALKER_DCR(mip);
3969 }
3970
3971 /*
3972 * Return the margin value currently assigned to the specified MAC instance.
3973 */
3974 void
3975 mac_margin_get(mac_handle_t mh, uint32_t *marginp)
3976 {
3977 mac_impl_t *mip = (mac_impl_t *)mh;
3978
3979 rw_enter(&(mip->mi_rw_lock), RW_READER);
3980 *marginp = mip->mi_margin;
3981 rw_exit(&(mip->mi_rw_lock));
3982 }
3983
3984 /*
3985 * mac_info_get() is used for retrieving the mac_info when a DL_INFO_REQ is
3986 * issued before a DL_ATTACH_REQ. we walk the i_mac_impl_hash table and find
3987 * the first mac_impl_t with a matching driver name; then we copy its mac_info_t
3988 * to the caller. we do all this with i_mac_impl_lock held so the mac_impl_t
3989 * cannot disappear while we are accessing it.
3990 */
3991 typedef struct i_mac_info_state_s {
3992 const char *mi_name;
3993 mac_info_t *mi_infop;
3994 } i_mac_info_state_t;
3995
3996 /*ARGSUSED*/
3997 static uint_t
3998 i_mac_info_walker(mod_hash_key_t key, mod_hash_val_t *val, void *arg)
3999 {
4000 i_mac_info_state_t *statep = arg;
4001 mac_impl_t *mip = (mac_impl_t *)val;
4002
4003 if (mip->mi_state_flags & MIS_DISABLED)
4004 return (MH_WALK_CONTINUE);
4005
4006 if (strcmp(statep->mi_name,
4007 ddi_driver_name(mip->mi_dip)) != 0)
4008 return (MH_WALK_CONTINUE);
4009
4010 statep->mi_infop = &mip->mi_info;
4011 return (MH_WALK_TERMINATE);
4012 }
4013
4014 boolean_t
4015 mac_info_get(const char *name, mac_info_t *minfop)
4016 {
4017 i_mac_info_state_t state;
4018
4019 rw_enter(&i_mac_impl_lock, RW_READER);
4020 state.mi_name = name;
4021 state.mi_infop = NULL;
4022 mod_hash_walk(i_mac_impl_hash, i_mac_info_walker, &state);
4023 if (state.mi_infop == NULL) {
4024 rw_exit(&i_mac_impl_lock);
4025 return (B_FALSE);
4026 }
4027 *minfop = *state.mi_infop;
4028 rw_exit(&i_mac_impl_lock);
4029 return (B_TRUE);
4030 }
4031
4032 /*
4033 * To get the capabilities that MAC layer cares about, such as rings, factory
4034 * mac address, vnic or not, it should directly invoke this function. If the
4035 * link is part of a bridge, then the only "capability" it has is the inability
4036 * to do zero copy.
4037 */
4038 boolean_t
4039 i_mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4040 {
4041 mac_impl_t *mip = (mac_impl_t *)mh;
4042
4043 if (mip->mi_bridge_link != NULL)
4044 return (cap == MAC_CAPAB_NO_ZCOPY);
4045 else if (mip->mi_callbacks->mc_callbacks & MC_GETCAPAB)
4046 return (mip->mi_getcapab(mip->mi_driver, cap, cap_data));
4047 else
4048 return (B_FALSE);
4049 }
4050
4051 /*
4052 * Capability query function. If number of active mac clients is greater than
4053 * 1, only limited capabilities can be advertised to the caller no matter the
4054 * driver has certain capability or not. Else, we query the driver to get the
4055 * capability.
4056 */
4057 boolean_t
4058 mac_capab_get(mac_handle_t mh, mac_capab_t cap, void *cap_data)
4059 {
4060 mac_impl_t *mip = (mac_impl_t *)mh;
4061
4062 /*
4063 * if mi_nactiveclients > 1, only MAC_CAPAB_LEGACY, MAC_CAPAB_HCKSUM,
4064 * MAC_CAPAB_NO_NATIVEVLAN and MAC_CAPAB_NO_ZCOPY can be advertised.
4065 */
4066 if (mip->mi_nactiveclients > 1) {
4067 switch (cap) {
4068 case MAC_CAPAB_NO_ZCOPY:
4069 return (B_TRUE);
4070 case MAC_CAPAB_LEGACY:
4071 case MAC_CAPAB_HCKSUM:
4072 case MAC_CAPAB_NO_NATIVEVLAN:
4073 break;
4074 default:
4075 return (B_FALSE);
4076 }
4077 }
4078
4079 /* else get capab from driver */
4080 return (i_mac_capab_get(mh, cap, cap_data));
4081 }
4082
4083 boolean_t
4084 mac_sap_verify(mac_handle_t mh, uint32_t sap, uint32_t *bind_sap)
4085 {
4086 mac_impl_t *mip = (mac_impl_t *)mh;
4087
4088 return (mip->mi_type->mt_ops.mtops_sap_verify(sap, bind_sap,
4089 mip->mi_pdata));
4090 }
4091
4092 mblk_t *
4093 mac_header(mac_handle_t mh, const uint8_t *daddr, uint32_t sap, mblk_t *payload,
4094 size_t extra_len)
4095 {
4096 mac_impl_t *mip = (mac_impl_t *)mh;
4097 const uint8_t *hdr_daddr;
4098
4099 /*
4100 * If the MAC is point-to-point with a fixed destination address, then
4101 * we must always use that destination in the MAC header.
4102 */
4103 hdr_daddr = (mip->mi_dstaddr_set ? mip->mi_dstaddr : daddr);
4104 return (mip->mi_type->mt_ops.mtops_header(mip->mi_addr, hdr_daddr, sap,
4105 mip->mi_pdata, payload, extra_len));
4106 }
4107
4108 int
4109 mac_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4110 {
4111 mac_impl_t *mip = (mac_impl_t *)mh;
4112
4113 return (mip->mi_type->mt_ops.mtops_header_info(mp, mip->mi_pdata,
4114 mhip));
4115 }
4116
4117 int
4118 mac_vlan_header_info(mac_handle_t mh, mblk_t *mp, mac_header_info_t *mhip)
4119 {
4120 mac_impl_t *mip = (mac_impl_t *)mh;
4121 boolean_t is_ethernet = (mip->mi_info.mi_media == DL_ETHER);
4122 int err = 0;
4123
4124 /*
4125 * Packets should always be at least 16 bit aligned.
4126 */
4127 ASSERT(IS_P2ALIGNED(mp->b_rptr, sizeof (uint16_t)));
4128
4129 if ((err = mac_header_info(mh, mp, mhip)) != 0)
4130 return (err);
4131
4132 /*
4133 * If this is a VLAN-tagged Ethernet packet, then the SAP in the
4134 * mac_header_info_t as returned by mac_header_info() is
4135 * ETHERTYPE_VLAN. We need to grab the ethertype from the VLAN header.
4136 */
4137 if (is_ethernet && (mhip->mhi_bindsap == ETHERTYPE_VLAN)) {
4138 struct ether_vlan_header *evhp;
4139 uint16_t sap;
4140 mblk_t *tmp = NULL;
4141 size_t size;
4142
4143 size = sizeof (struct ether_vlan_header);
4144 if (MBLKL(mp) < size) {
4145 /*
4146 * Pullup the message in order to get the MAC header
4147 * infomation. Note that this is a read-only function,
4148 * we keep the input packet intact.
4149 */
4150 if ((tmp = msgpullup(mp, size)) == NULL)
4151 return (EINVAL);
4152
4153 mp = tmp;
4154 }
4155 evhp = (struct ether_vlan_header *)mp->b_rptr;
4156 sap = ntohs(evhp->ether_type);
4157 (void) mac_sap_verify(mh, sap, &mhip->mhi_bindsap);
4158 mhip->mhi_hdrsize = sizeof (struct ether_vlan_header);
4159 mhip->mhi_tci = ntohs(evhp->ether_tci);
4160 mhip->mhi_istagged = B_TRUE;
4161 freemsg(tmp);
4162
4163 if (VLAN_CFI(mhip->mhi_tci) != ETHER_CFI)
4164 return (EINVAL);
4165 } else {
4166 mhip->mhi_istagged = B_FALSE;
4167 mhip->mhi_tci = 0;
4168 }
4169
4170 return (0);
4171 }
4172
4173 mblk_t *
4174 mac_header_cook(mac_handle_t mh, mblk_t *mp)
4175 {
4176 mac_impl_t *mip = (mac_impl_t *)mh;
4177
4178 if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_COOK) {
4179 if (DB_REF(mp) > 1) {
4180 mblk_t *newmp = copymsg(mp);
4181 if (newmp == NULL)
4182 return (NULL);
4183 freemsg(mp);
4184 mp = newmp;
4185 }
4186 return (mip->mi_type->mt_ops.mtops_header_cook(mp,
4187 mip->mi_pdata));
4188 }
4189 return (mp);
4190 }
4191
4192 mblk_t *
4193 mac_header_uncook(mac_handle_t mh, mblk_t *mp)
4194 {
4195 mac_impl_t *mip = (mac_impl_t *)mh;
4196
4197 if (mip->mi_type->mt_ops.mtops_ops & MTOPS_HEADER_UNCOOK) {
4198 if (DB_REF(mp) > 1) {
4199 mblk_t *newmp = copymsg(mp);
4200 if (newmp == NULL)
4201 return (NULL);
4202 freemsg(mp);
4203 mp = newmp;
4204 }
4205 return (mip->mi_type->mt_ops.mtops_header_uncook(mp,
4206 mip->mi_pdata));
4207 }
4208 return (mp);
4209 }
4210
4211 uint_t
4212 mac_addr_len(mac_handle_t mh)
4213 {
4214 mac_impl_t *mip = (mac_impl_t *)mh;
4215
4216 return (mip->mi_type->mt_addr_length);
4217 }
4218
4219 /* True if a MAC is a VNIC */
4220 boolean_t
4221 mac_is_vnic(mac_handle_t mh)
4222 {
4223 return (((mac_impl_t *)mh)->mi_state_flags & MIS_IS_VNIC);
4224 }
4225
4226 mac_handle_t
4227 mac_get_lower_mac_handle(mac_handle_t mh)
4228 {
4229 mac_impl_t *mip = (mac_impl_t *)mh;
4230
4231 ASSERT(mac_is_vnic(mh));
4232 return (((vnic_t *)mip->mi_driver)->vn_lower_mh);
4233 }
4234
4235 boolean_t
4236 mac_is_vnic_primary(mac_handle_t mh)
4237 {
4238 mac_impl_t *mip = (mac_impl_t *)mh;
4239
4240 ASSERT(mac_is_vnic(mh));
4241 return (((vnic_t *)mip->mi_driver)->vn_addr_type ==
4242 VNIC_MAC_ADDR_TYPE_PRIMARY);
4243 }
4244
4245 void
4246 mac_update_resources(mac_resource_props_t *nmrp, mac_resource_props_t *cmrp,
4247 boolean_t is_user_flow)
4248 {
4249 if (nmrp != NULL && cmrp != NULL) {
4250 if (nmrp->mrp_mask & MRP_PRIORITY) {
4251 if (nmrp->mrp_priority == MPL_RESET) {
4252 cmrp->mrp_mask &= ~MRP_PRIORITY;
4253 if (is_user_flow) {
4254 cmrp->mrp_priority =
4255 MPL_SUBFLOW_DEFAULT;
4256 } else {
4257 cmrp->mrp_priority = MPL_LINK_DEFAULT;
4258 }
4259 } else {
4260 cmrp->mrp_mask |= MRP_PRIORITY;
4261 cmrp->mrp_priority = nmrp->mrp_priority;
4262 }
4263 }
4264 if (nmrp->mrp_mask & MRP_MAXBW) {
4265 if (nmrp->mrp_maxbw == MRP_MAXBW_RESETVAL) {
4266 cmrp->mrp_mask &= ~MRP_MAXBW;
4267 cmrp->mrp_maxbw = 0;
4268 } else {
4269 cmrp->mrp_mask |= MRP_MAXBW;
4270 cmrp->mrp_maxbw = nmrp->mrp_maxbw;
4271 }
4272 }
4273 if (nmrp->mrp_mask & MRP_CPUS)
4274 MAC_COPY_CPUS(nmrp, cmrp);
4275
4276 if (nmrp->mrp_mask & MRP_POOL) {
4277 if (strlen(nmrp->mrp_pool) == 0) {
4278 cmrp->mrp_mask &= ~MRP_POOL;
4279 bzero(cmrp->mrp_pool, sizeof (cmrp->mrp_pool));
4280 } else {
4281 cmrp->mrp_mask |= MRP_POOL;
4282 (void) strncpy(cmrp->mrp_pool, nmrp->mrp_pool,
4283 sizeof (cmrp->mrp_pool));
4284 }
4285
4286 }
4287
4288 if (nmrp->mrp_mask & MRP_PROTECT)
4289 mac_protect_update(nmrp, cmrp);
4290
4291 /*
4292 * Update the rings specified.
4293 */
4294 if (nmrp->mrp_mask & MRP_RX_RINGS) {
4295 if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4296 cmrp->mrp_mask &= ~MRP_RX_RINGS;
4297 if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4298 cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4299 cmrp->mrp_nrxrings = 0;
4300 } else {
4301 cmrp->mrp_mask |= MRP_RX_RINGS;
4302 cmrp->mrp_nrxrings = nmrp->mrp_nrxrings;
4303 }
4304 }
4305 if (nmrp->mrp_mask & MRP_TX_RINGS) {
4306 if (nmrp->mrp_mask & MRP_RINGS_RESET) {
4307 cmrp->mrp_mask &= ~MRP_TX_RINGS;
4308 if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4309 cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4310 cmrp->mrp_ntxrings = 0;
4311 } else {
4312 cmrp->mrp_mask |= MRP_TX_RINGS;
4313 cmrp->mrp_ntxrings = nmrp->mrp_ntxrings;
4314 }
4315 }
4316 if (nmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4317 cmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4318 else if (cmrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4319 cmrp->mrp_mask &= ~MRP_RXRINGS_UNSPEC;
4320
4321 if (nmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4322 cmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4323 else if (cmrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4324 cmrp->mrp_mask &= ~MRP_TXRINGS_UNSPEC;
4325 }
4326 }
4327
4328 /*
4329 * i_mac_set_resources:
4330 *
4331 * This routine associates properties with the primary MAC client of
4332 * the specified MAC instance.
4333 * - Cache the properties in mac_impl_t
4334 * - Apply the properties to the primary MAC client if exists
4335 */
4336 int
4337 i_mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4338 {
4339 mac_impl_t *mip = (mac_impl_t *)mh;
4340 mac_client_impl_t *mcip;
4341 int err = 0;
4342 uint32_t resmask, newresmask;
4343 mac_resource_props_t *tmrp, *umrp;
4344
4345 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4346
4347 err = mac_validate_props(mip, mrp);
4348 if (err != 0)
4349 return (err);
4350
4351 umrp = kmem_zalloc(sizeof (*umrp), KM_SLEEP);
4352 bcopy(&mip->mi_resource_props, umrp, sizeof (*umrp));
4353 resmask = umrp->mrp_mask;
4354 mac_update_resources(mrp, umrp, B_FALSE);
4355 newresmask = umrp->mrp_mask;
4356
4357 if (resmask == 0 && newresmask != 0) {
4358 /*
4359 * Bandwidth, priority, cpu or pool link properties configured,
4360 * must disable fastpath.
4361 */
4362 if ((err = mac_fastpath_disable((mac_handle_t)mip)) != 0) {
4363 kmem_free(umrp, sizeof (*umrp));
4364 return (err);
4365 }
4366 }
4367
4368 /*
4369 * Since bind_cpu may be modified by mac_client_set_resources()
4370 * we use a copy of bind_cpu and finally cache bind_cpu in mip.
4371 * This allows us to cache only user edits in mip.
4372 */
4373 tmrp = kmem_zalloc(sizeof (*tmrp), KM_SLEEP);
4374 bcopy(mrp, tmrp, sizeof (*tmrp));
4375 mcip = mac_primary_client_handle(mip);
4376 if (mcip != NULL && (mcip->mci_state_flags & MCIS_IS_AGGR_PORT) == 0) {
4377 err = mac_client_set_resources((mac_client_handle_t)mcip, tmrp);
4378 } else if ((mrp->mrp_mask & MRP_RX_RINGS ||
4379 mrp->mrp_mask & MRP_TX_RINGS)) {
4380 mac_client_impl_t *vmcip;
4381
4382 /*
4383 * If the primary is not up, we need to check if there
4384 * are any VLANs on this primary. If there are then
4385 * we need to set this property on the VLANs since
4386 * VLANs follow the primary they are based on. Just
4387 * look for the first VLAN and change its properties,
4388 * all the other VLANs should be in the same group.
4389 */
4390 for (vmcip = mip->mi_clients_list; vmcip != NULL;
4391 vmcip = vmcip->mci_client_next) {
4392 if ((vmcip->mci_flent->fe_type & FLOW_PRIMARY_MAC) &&
4393 mac_client_vid((mac_client_handle_t)vmcip) !=
4394 VLAN_ID_NONE) {
4395 break;
4396 }
4397 }
4398 if (vmcip != NULL) {
4399 mac_resource_props_t *omrp;
4400 mac_resource_props_t *vmrp;
4401
4402 omrp = kmem_zalloc(sizeof (*omrp), KM_SLEEP);
4403 bcopy(MCIP_RESOURCE_PROPS(vmcip), omrp, sizeof (*omrp));
4404 /*
4405 * We dont' call mac_update_resources since we
4406 * want to take only the ring properties and
4407 * not all the properties that may have changed.
4408 */
4409 vmrp = MCIP_RESOURCE_PROPS(vmcip);
4410 if (mrp->mrp_mask & MRP_RX_RINGS) {
4411 if (mrp->mrp_mask & MRP_RINGS_RESET) {
4412 vmrp->mrp_mask &= ~MRP_RX_RINGS;
4413 if (vmrp->mrp_mask &
4414 MRP_RXRINGS_UNSPEC) {
4415 vmrp->mrp_mask &=
4416 ~MRP_RXRINGS_UNSPEC;
4417 }
4418 vmrp->mrp_nrxrings = 0;
4419 } else {
4420 vmrp->mrp_mask |= MRP_RX_RINGS;
4421 vmrp->mrp_nrxrings = mrp->mrp_nrxrings;
4422 }
4423 }
4424 if (mrp->mrp_mask & MRP_TX_RINGS) {
4425 if (mrp->mrp_mask & MRP_RINGS_RESET) {
4426 vmrp->mrp_mask &= ~MRP_TX_RINGS;
4427 if (vmrp->mrp_mask &
4428 MRP_TXRINGS_UNSPEC) {
4429 vmrp->mrp_mask &=
4430 ~MRP_TXRINGS_UNSPEC;
4431 }
4432 vmrp->mrp_ntxrings = 0;
4433 } else {
4434 vmrp->mrp_mask |= MRP_TX_RINGS;
4435 vmrp->mrp_ntxrings = mrp->mrp_ntxrings;
4436 }
4437 }
4438 if (mrp->mrp_mask & MRP_RXRINGS_UNSPEC)
4439 vmrp->mrp_mask |= MRP_RXRINGS_UNSPEC;
4440
4441 if (mrp->mrp_mask & MRP_TXRINGS_UNSPEC)
4442 vmrp->mrp_mask |= MRP_TXRINGS_UNSPEC;
4443
4444 if ((err = mac_client_set_rings_prop(vmcip, mrp,
4445 omrp)) != 0) {
4446 bcopy(omrp, MCIP_RESOURCE_PROPS(vmcip),
4447 sizeof (*omrp));
4448 } else {
4449 mac_set_prim_vlan_rings(mip, vmrp);
4450 }
4451 kmem_free(omrp, sizeof (*omrp));
4452 }
4453 }
4454
4455 /* Only update the values if mac_client_set_resources succeeded */
4456 if (err == 0) {
4457 bcopy(umrp, &mip->mi_resource_props, sizeof (*umrp));
4458 /*
4459 * If bandwidth, priority or cpu link properties cleared,
4460 * renable fastpath.
4461 */
4462 if (resmask != 0 && newresmask == 0)
4463 mac_fastpath_enable((mac_handle_t)mip);
4464 } else if (resmask == 0 && newresmask != 0) {
4465 mac_fastpath_enable((mac_handle_t)mip);
4466 }
4467 kmem_free(tmrp, sizeof (*tmrp));
4468 kmem_free(umrp, sizeof (*umrp));
4469 return (err);
4470 }
4471
4472 int
4473 mac_set_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4474 {
4475 int err;
4476
4477 i_mac_perim_enter((mac_impl_t *)mh);
4478 err = i_mac_set_resources(mh, mrp);
4479 i_mac_perim_exit((mac_impl_t *)mh);
4480 return (err);
4481 }
4482
4483 /*
4484 * Get the properties cached for the specified MAC instance.
4485 */
4486 void
4487 mac_get_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4488 {
4489 mac_impl_t *mip = (mac_impl_t *)mh;
4490 mac_client_impl_t *mcip;
4491
4492 mcip = mac_primary_client_handle(mip);
4493 if (mcip != NULL) {
4494 mac_client_get_resources((mac_client_handle_t)mcip, mrp);
4495 return;
4496 }
4497 bcopy(&mip->mi_resource_props, mrp, sizeof (mac_resource_props_t));
4498 }
4499
4500 /*
4501 * Get the effective properties from the primary client of the
4502 * specified MAC instance.
4503 */
4504 void
4505 mac_get_effective_resources(mac_handle_t mh, mac_resource_props_t *mrp)
4506 {
4507 mac_impl_t *mip = (mac_impl_t *)mh;
4508 mac_client_impl_t *mcip;
4509
4510 mcip = mac_primary_client_handle(mip);
4511 if (mcip != NULL) {
4512 mac_client_get_effective_resources((mac_client_handle_t)mcip,
4513 mrp);
4514 return;
4515 }
4516 bzero(mrp, sizeof (mac_resource_props_t));
4517 }
4518
4519 int
4520 mac_set_pvid(mac_handle_t mh, uint16_t pvid)
4521 {
4522 mac_impl_t *mip = (mac_impl_t *)mh;
4523 mac_client_impl_t *mcip;
4524 mac_unicast_impl_t *muip;
4525
4526 i_mac_perim_enter(mip);
4527 if (pvid != 0) {
4528 for (mcip = mip->mi_clients_list; mcip != NULL;
4529 mcip = mcip->mci_client_next) {
4530 for (muip = mcip->mci_unicast_list; muip != NULL;
4531 muip = muip->mui_next) {
4532 if (muip->mui_vid == pvid) {
4533 i_mac_perim_exit(mip);
4534 return (EBUSY);
4535 }
4536 }
4537 }
4538 }
4539 mip->mi_pvid = pvid;
4540 i_mac_perim_exit(mip);
4541 return (0);
4542 }
4543
4544 uint16_t
4545 mac_get_pvid(mac_handle_t mh)
4546 {
4547 mac_impl_t *mip = (mac_impl_t *)mh;
4548
4549 return (mip->mi_pvid);
4550 }
4551
4552 uint32_t
4553 mac_get_llimit(mac_handle_t mh)
4554 {
4555 mac_impl_t *mip = (mac_impl_t *)mh;
4556
4557 return (mip->mi_llimit);
4558 }
4559
4560 uint32_t
4561 mac_get_ldecay(mac_handle_t mh)
4562 {
4563 mac_impl_t *mip = (mac_impl_t *)mh;
4564
4565 return (mip->mi_ldecay);
4566 }
4567
4568 /*
4569 * Rename a mac client, its flow, and the kstat.
4570 */
4571 int
4572 mac_rename_primary(mac_handle_t mh, const char *new_name)
4573 {
4574 mac_impl_t *mip = (mac_impl_t *)mh;
4575 mac_client_impl_t *cur_clnt = NULL;
4576 flow_entry_t *fep;
4577
4578 i_mac_perim_enter(mip);
4579
4580 /*
4581 * VNICs: we need to change the sys flow name and
4582 * the associated flow kstat.
4583 */
4584 if (mip->mi_state_flags & MIS_IS_VNIC) {
4585 mac_client_impl_t *mcip = mac_vnic_lower(mip);
4586 ASSERT(new_name != NULL);
4587 mac_rename_flow_names(mcip, new_name);
4588 mac_stat_rename(mcip);
4589 goto done;
4590 }
4591 /*
4592 * This mac may itself be an aggr link, or it may have some client
4593 * which is an aggr port. For both cases, we need to change the
4594 * aggr port's mac client name, its flow name and the associated flow
4595 * kstat.
4596 */
4597 if (mip->mi_state_flags & MIS_IS_AGGR) {
4598 mac_capab_aggr_t aggr_cap;
4599 mac_rename_fn_t rename_fn;
4600 boolean_t ret;
4601
4602 ASSERT(new_name != NULL);
4603 ret = i_mac_capab_get((mac_handle_t)mip, MAC_CAPAB_AGGR,
4604 (void *)(&aggr_cap));
4605 ASSERT(ret == B_TRUE);
4606 rename_fn = aggr_cap.mca_rename_fn;
4607 rename_fn(new_name, mip->mi_driver);
4608 /*
4609 * The aggr's client name and kstat flow name will be
4610 * updated below, i.e. via mac_rename_flow_names.
4611 */
4612 }
4613
4614 for (cur_clnt = mip->mi_clients_list; cur_clnt != NULL;
4615 cur_clnt = cur_clnt->mci_client_next) {
4616 if (cur_clnt->mci_state_flags & MCIS_IS_AGGR_PORT) {
4617 if (new_name != NULL) {
4618 char *str_st = cur_clnt->mci_name;
4619 char *str_del = strchr(str_st, '-');
4620
4621 ASSERT(str_del != NULL);
4622 bzero(str_del + 1, MAXNAMELEN -
4623 (str_del - str_st + 1));
4624 bcopy(new_name, str_del + 1,
4625 strlen(new_name));
4626 }
4627 fep = cur_clnt->mci_flent;
4628 mac_rename_flow(fep, cur_clnt->mci_name);
4629 break;
4630 } else if (new_name != NULL &&
4631 cur_clnt->mci_state_flags & MCIS_USE_DATALINK_NAME) {
4632 mac_rename_flow_names(cur_clnt, new_name);
4633 break;
4634 }
4635 }
4636
4637 /* Recreate kstats associated with aggr pseudo rings */
4638 if (mip->mi_state_flags & MIS_IS_AGGR)
4639 mac_pseudo_ring_stat_rename(mip);
4640
4641 done:
4642 i_mac_perim_exit(mip);
4643 return (0);
4644 }
4645
4646 /*
4647 * Rename the MAC client's flow names
4648 */
4649 static void
4650 mac_rename_flow_names(mac_client_impl_t *mcip, const char *new_name)
4651 {
4652 flow_entry_t *flent;
4653 uint16_t vid;
4654 char flowname[MAXFLOWNAMELEN];
4655 mac_impl_t *mip = mcip->mci_mip;
4656
4657 ASSERT(MAC_PERIM_HELD((mac_handle_t)mip));
4658
4659 /*
4660 * Use mi_rw_lock to ensure that threads not in the mac perimeter
4661 * see a self-consistent value for mci_name
4662 */
4663 rw_enter(&mip->mi_rw_lock, RW_WRITER);
4664 (void) strlcpy(mcip->mci_name, new_name, sizeof (mcip->mci_name));
4665 rw_exit(&mip->mi_rw_lock);
4666
4667 mac_rename_flow(mcip->mci_flent, new_name);
4668
4669 if (mcip->mci_nflents == 1)
4670 return;
4671
4672 /*
4673 * We have to rename all the others too, no stats to destroy for
4674 * these.
4675 */
4676 for (flent = mcip->mci_flent_list; flent != NULL;
4677 flent = flent->fe_client_next) {
4678 if (flent != mcip->mci_flent) {
4679 vid = i_mac_flow_vid(flent);
4680 (void) sprintf(flowname, "%s%u", new_name, vid);
4681 mac_flow_set_name(flent, flowname);
4682 }
4683 }
4684 }
4685
4686
4687 /*
4688 * Add a flow to the MAC client's flow list - i.e list of MAC/VID tuples
4689 * defined for the specified MAC client.
4690 */
4691 static void
4692 mac_client_add_to_flow_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4693 {
4694 ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4695 /*
4696 * The promisc Rx data path walks the mci_flent_list. Protect by
4697 * using mi_rw_lock
4698 */
4699 rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4700
4701 /* Add it to the head */
4702 flent->fe_client_next = mcip->mci_flent_list;
4703 mcip->mci_flent_list = flent;
4704 mcip->mci_nflents++;
4705
4706 /*
4707 * Keep track of the number of non-zero VIDs addresses per MAC
4708 * client to avoid figuring it out in the data-path.
4709 */
4710 if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4711 mcip->mci_nvids++;
4712
4713 rw_exit(&mcip->mci_rw_lock);
4714 }
4715
4716 /*
4717 * Remove a flow entry from the MAC client's list.
4718 */
4719 static void
4720 mac_client_remove_flow_from_list(mac_client_impl_t *mcip, flow_entry_t *flent)
4721 {
4722 flow_entry_t *fe = mcip->mci_flent_list;
4723 flow_entry_t *prev_fe = NULL;
4724
4725 ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4726 /*
4727 * The promisc Rx data path walks the mci_flent_list. Protect by
4728 * using mci_rw_lock
4729 */
4730 rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4731 while ((fe != NULL) && (fe != flent)) {
4732 prev_fe = fe;
4733 fe = fe->fe_client_next;
4734 }
4735
4736 ASSERT(fe != NULL);
4737 if (prev_fe == NULL) {
4738 /* Deleting the first node */
4739 mcip->mci_flent_list = fe->fe_client_next;
4740 } else {
4741 prev_fe->fe_client_next = fe->fe_client_next;
4742 }
4743 mcip->mci_nflents--;
4744
4745 if (i_mac_flow_vid(flent) != VLAN_ID_NONE)
4746 mcip->mci_nvids--;
4747
4748 rw_exit(&mcip->mci_rw_lock);
4749 }
4750
4751 /*
4752 * Check if the given VID belongs to this MAC client.
4753 */
4754 boolean_t
4755 mac_client_check_flow_vid(mac_client_impl_t *mcip, uint16_t vid)
4756 {
4757 flow_entry_t *flent;
4758 uint16_t mci_vid;
4759
4760 /* The mci_flent_list is protected by mci_rw_lock */
4761 rw_enter(&mcip->mci_rw_lock, RW_WRITER);
4762 for (flent = mcip->mci_flent_list; flent != NULL;
4763 flent = flent->fe_client_next) {
4764 mci_vid = i_mac_flow_vid(flent);
4765 if (vid == mci_vid) {
4766 rw_exit(&mcip->mci_rw_lock);
4767 return (B_TRUE);
4768 }
4769 }
4770 rw_exit(&mcip->mci_rw_lock);
4771 return (B_FALSE);
4772 }
4773
4774 /*
4775 * Get the flow entry for the specified <MAC addr, VID> tuple.
4776 */
4777 static flow_entry_t *
4778 mac_client_get_flow(mac_client_impl_t *mcip, mac_unicast_impl_t *muip)
4779 {
4780 mac_address_t *map = mcip->mci_unicast;
4781 flow_entry_t *flent;
4782 uint16_t vid;
4783 flow_desc_t flow_desc;
4784
4785 ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4786
4787 mac_flow_get_desc(mcip->mci_flent, &flow_desc);
4788 if (bcmp(flow_desc.fd_dst_mac, map->ma_addr, map->ma_len) != 0)
4789 return (NULL);
4790
4791 for (flent = mcip->mci_flent_list; flent != NULL;
4792 flent = flent->fe_client_next) {
4793 vid = i_mac_flow_vid(flent);
4794 if (vid == muip->mui_vid) {
4795 return (flent);
4796 }
4797 }
4798
4799 return (NULL);
4800 }
4801
4802 /*
4803 * Since mci_flent has the SRSs, when we want to remove it, we replace
4804 * the flow_desc_t in mci_flent with that of an existing flent and then
4805 * remove that flent instead of mci_flent.
4806 */
4807 static flow_entry_t *
4808 mac_client_swap_mciflent(mac_client_impl_t *mcip)
4809 {
4810 flow_entry_t *flent = mcip->mci_flent;
4811 flow_tab_t *ft = flent->fe_flow_tab;
4812 flow_entry_t *flent1;
4813 flow_desc_t fl_desc;
4814 char fl_name[MAXFLOWNAMELEN];
4815 int err;
4816
4817 ASSERT(MAC_PERIM_HELD((mac_handle_t)mcip->mci_mip));
4818 ASSERT(mcip->mci_nflents > 1);
4819
4820 /* get the next flent following the primary flent */
4821 flent1 = mcip->mci_flent_list->fe_client_next;
4822 ASSERT(flent1 != NULL && flent1->fe_flow_tab == ft);
4823
4824 /*
4825 * Remove the flent from the flow table before updating the
4826 * flow descriptor as the hash depends on the flow descriptor.
4827 * This also helps incoming packet classification avoid having
4828 * to grab fe_lock. Access to fe_flow_desc of a flent not in the
4829 * flow table is done under the fe_lock so that log or stat functions
4830 * see a self-consistent fe_flow_desc. The name and desc are specific
4831 * to a flow, the rest are shared by all the clients, including
4832 * resource control etc.
4833 */
4834 mac_flow_remove(ft, flent, B_TRUE);
4835 mac_flow_remove(ft, flent1, B_TRUE);
4836
4837 bcopy(&flent->fe_flow_desc, &fl_desc, sizeof (flow_desc_t));
4838 bcopy(flent->fe_flow_name, fl_name, MAXFLOWNAMELEN);
4839
4840 /* update the primary flow entry */
4841 mutex_enter(&flent->fe_lock);
4842 bcopy(&flent1->fe_flow_desc, &flent->fe_flow_desc,
4843 sizeof (flow_desc_t));
4844 bcopy(&flent1->fe_flow_name, &flent->fe_flow_name, MAXFLOWNAMELEN);
4845 mutex_exit(&flent->fe_lock);
4846
4847 /* update the flow entry that is to be freed */
4848 mutex_enter(&flent1->fe_lock);
4849 bcopy(&fl_desc, &flent1->fe_flow_desc, sizeof (flow_desc_t));
4850 bcopy(fl_name, &flent1->fe_flow_name, MAXFLOWNAMELEN);
4851 mutex_exit(&flent1->fe_lock);
4852
4853 /* now reinsert the flow entries in the table */
4854 err = mac_flow_add(ft, flent);
4855 ASSERT(err == 0);
4856
4857 err = mac_flow_add(ft, flent1);
4858 ASSERT(err == 0);
4859
4860 return (flent1);
4861 }
4862
4863 /*
4864 * Return whether there is only one flow entry associated with this
4865 * MAC client.
4866 */
4867 static boolean_t
4868 mac_client_single_rcvr(mac_client_impl_t *mcip)
4869 {
4870 return (mcip->mci_nflents == 1);
4871 }
4872
4873 int
4874 mac_validate_props(mac_impl_t *mip, mac_resource_props_t *mrp)
4875 {
4876 boolean_t reset;
4877 uint32_t rings_needed;
4878 uint32_t rings_avail;
4879 mac_group_type_t gtype;
4880 mac_resource_props_t *mip_mrp;
4881
4882 if (mrp == NULL)
4883 return (0);
4884
4885 if (mrp->mrp_mask & MRP_PRIORITY) {
4886 mac_priority_level_t pri = mrp->mrp_priority;
4887
4888 if (pri < MPL_LOW || pri > MPL_RESET)
4889 return (EINVAL);
4890 }
4891
4892 if (mrp->mrp_mask & MRP_MAXBW) {
4893 uint64_t maxbw = mrp->mrp_maxbw;
4894
4895 if (maxbw < MRP_MAXBW_MINVAL && maxbw != 0)
4896 return (EINVAL);
4897 }
4898 if (mrp->mrp_mask & MRP_CPUS) {
4899 int i, j;
4900 mac_cpu_mode_t fanout;
4901
4902 if (mrp->mrp_ncpus > ncpus)
4903 return (EINVAL);
4904
4905 for (i = 0; i < mrp->mrp_ncpus; i++) {
4906 for (j = 0; j < mrp->mrp_ncpus; j++) {
4907 if (i != j &&
4908 mrp->mrp_cpu[i] == mrp->mrp_cpu[j]) {
4909 return (EINVAL);
4910 }
4911 }
4912 }
4913
4914 for (i = 0; i < mrp->mrp_ncpus; i++) {
4915 cpu_t *cp;
4916 int rv;
4917
4918 mutex_enter(&cpu_lock);
4919 cp = cpu_get(mrp->mrp_cpu[i]);
4920 if (cp != NULL)
4921 rv = cpu_is_online(cp);
4922 else
4923 rv = 0;
4924 mutex_exit(&cpu_lock);
4925 if (rv == 0)
4926 return (EINVAL);
4927 }
4928
4929 fanout = mrp->mrp_fanout_mode;
4930 if (fanout < 0 || fanout > MCM_CPUS)
4931 return (EINVAL);
4932 }
4933
4934 if (mrp->mrp_mask & MRP_PROTECT) {
4935 int err = mac_protect_validate(mrp);
4936 if (err != 0)
4937 return (err);
4938 }
4939
4940 if (!(mrp->mrp_mask & MRP_RX_RINGS) &&
4941 !(mrp->mrp_mask & MRP_TX_RINGS)) {
4942 return (0);
4943 }
4944
4945 /*
4946 * mip will be null when we come from mac_flow_create or
4947 * mac_link_flow_modify. In the latter case it is a user flow,
4948 * for which we don't support rings. In the former we would
4949 * have validated the props beforehand (i_mac_unicast_add ->
4950 * mac_client_set_resources -> validate for the primary and
4951 * vnic_dev_create -> mac_client_set_resources -> validate for
4952 * a vnic.
4953 */
4954 if (mip == NULL)
4955 return (0);
4956
4957 /*
4958 * We don't support setting rings property for a VNIC that is using a
4959 * primary address (VLAN)
4960 */
4961 if ((mip->mi_state_flags & MIS_IS_VNIC) &&
4962 mac_is_vnic_primary((mac_handle_t)mip)) {
4963 return (ENOTSUP);
4964 }
4965
4966 mip_mrp = &mip->mi_resource_props;
4967 /*
4968 * The rings property should be validated against the NICs
4969 * resources
4970 */
4971 if (mip->mi_state_flags & MIS_IS_VNIC)
4972 mip = (mac_impl_t *)mac_get_lower_mac_handle((mac_handle_t)mip);
4973
4974 reset = mrp->mrp_mask & MRP_RINGS_RESET;
4975 /*
4976 * If groups are not supported, return error.
4977 */
4978 if (((mrp->mrp_mask & MRP_RX_RINGS) && mip->mi_rx_groups == NULL) ||
4979 ((mrp->mrp_mask & MRP_TX_RINGS) && mip->mi_tx_groups == NULL)) {
4980 return (EINVAL);
4981 }
4982 /*
4983 * If we are just resetting, there is no validation needed.
4984 */
4985 if (reset)
4986 return (0);
4987
4988 if (mrp->mrp_mask & MRP_RX_RINGS) {
4989 rings_needed = mrp->mrp_nrxrings;
4990 /*
4991 * We just want to check if the number of additional
4992 * rings requested is available.
4993 */
4994 if (mip_mrp->mrp_mask & MRP_RX_RINGS) {
4995 if (mrp->mrp_nrxrings > mip_mrp->mrp_nrxrings)
4996 /* Just check for the additional rings */
4997 rings_needed -= mip_mrp->mrp_nrxrings;
4998 else
4999 /* We are not asking for additional rings */
5000 rings_needed = 0;
5001 }
5002 rings_avail = mip->mi_rxrings_avail;
5003 gtype = mip->mi_rx_group_type;
5004 } else {
5005 rings_needed = mrp->mrp_ntxrings;
5006 /* Similarly for the TX rings */
5007 if (mip_mrp->mrp_mask & MRP_TX_RINGS) {
5008 if (mrp->mrp_ntxrings > mip_mrp->mrp_ntxrings)
5009 /* Just check for the additional rings */
5010 rings_needed -= mip_mrp->mrp_ntxrings;
5011 else
5012 /* We are not asking for additional rings */
5013 rings_needed = 0;
5014 }
5015 rings_avail = mip->mi_txrings_avail;
5016 gtype = mip->mi_tx_group_type;
5017 }
5018
5019 /* Error if the group is dynamic .. */
5020 if (gtype == MAC_GROUP_TYPE_DYNAMIC) {
5021 /*
5022 * .. and rings specified are more than available.
5023 */
5024 if (rings_needed > rings_avail)
5025 return (EINVAL);
5026 } else {
5027 /*
5028 * OR group is static and we have specified some rings.
5029 */
5030 if (rings_needed > 0)
5031 return (EINVAL);
5032 }
5033 return (0);
5034 }
5035
5036 /*
5037 * Send a MAC_NOTE_LINK notification to all the MAC clients whenever the
5038 * underlying physical link is down. This is to allow MAC clients to
5039 * communicate with other clients.
5040 */
5041 void
5042 mac_virtual_link_update(mac_impl_t *mip)
5043 {
5044 if (mip->mi_linkstate != LINK_STATE_UP)
5045 i_mac_notify(mip, MAC_NOTE_LINK);
5046 }
5047
5048 /*
5049 * For clients that have a pass-thru MAC, e.g. VNIC, we set the VNIC's
5050 * mac handle in the client.
5051 */
5052 void
5053 mac_set_upper_mac(mac_client_handle_t mch, mac_handle_t mh,
5054 mac_resource_props_t *mrp)
5055 {
5056 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
5057 mac_impl_t *mip = (mac_impl_t *)mh;
5058
5059 mcip->mci_upper_mip = mip;
5060 /* If there are any properties, copy it over too */
5061 if (mrp != NULL) {
5062 bcopy(mrp, &mip->mi_resource_props,
5063 sizeof (mac_resource_props_t));
5064 }
5065 }
5066
5067 /*
5068 * Mark the mac as being used exclusively by the single mac client that is
5069 * doing some control operation on this mac. No further opens of this mac
5070 * will be allowed until this client calls mac_unmark_exclusive. The mac
5071 * client calling this function must already be in the mac perimeter
5072 */
5073 int
5074 mac_mark_exclusive(mac_handle_t mh)
5075 {
5076 mac_impl_t *mip = (mac_impl_t *)mh;
5077
5078 ASSERT(MAC_PERIM_HELD(mh));
5079 /*
5080 * Look up its entry in the global hash table.
5081 */
5082 rw_enter(&i_mac_impl_lock, RW_WRITER);
5083 if (mip->mi_state_flags & MIS_DISABLED) {
5084 rw_exit(&i_mac_impl_lock);
5085 return (ENOENT);
5086 }
5087
5088 /*
5089 * A reference to mac is held even if the link is not plumbed.
5090 * In i_dls_link_create() we open the MAC interface and hold the
5091 * reference. There is an additional reference for the mac_open
5092 * done in acquiring the mac perimeter
5093 */
5094 if (mip->mi_ref != 2) {
5095 rw_exit(&i_mac_impl_lock);
5096 return (EBUSY);
5097 }
5098
5099 ASSERT(!(mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5100 mip->mi_state_flags |= MIS_EXCLUSIVE_HELD;
5101 rw_exit(&i_mac_impl_lock);
5102 return (0);
5103 }
5104
5105 void
5106 mac_unmark_exclusive(mac_handle_t mh)
5107 {
5108 mac_impl_t *mip = (mac_impl_t *)mh;
5109
5110 ASSERT(MAC_PERIM_HELD(mh));
5111
5112 rw_enter(&i_mac_impl_lock, RW_WRITER);
5113 /* 1 for the creation and another for the perimeter */
5114 ASSERT(mip->mi_ref == 2 && (mip->mi_state_flags & MIS_EXCLUSIVE_HELD));
5115 mip->mi_state_flags &= ~MIS_EXCLUSIVE_HELD;
5116 rw_exit(&i_mac_impl_lock);
5117 }
5118
5119 /*
5120 * Set the MTU for the specified MAC.
5121 */
5122 int
5123 mac_set_mtu(mac_handle_t mh, uint_t new_mtu, uint_t *old_mtu_arg)
5124 {
5125 mac_impl_t *mip = (mac_impl_t *)mh;
5126 uint_t old_mtu;
5127 int rv = 0;
5128
5129 i_mac_perim_enter(mip);
5130
5131 if (!(mip->mi_callbacks->mc_callbacks & (MC_SETPROP|MC_GETPROP))) {
5132 rv = ENOTSUP;
5133 goto bail;
5134 }
5135
5136 old_mtu = mip->mi_sdu_max;
5137
5138 if (new_mtu == 0 || new_mtu < mip->mi_sdu_min) {
5139 rv = EINVAL;
5140 goto bail;
5141 }
5142
5143 if (old_mtu != new_mtu) {
5144 rv = mip->mi_callbacks->mc_setprop(mip->mi_driver,
5145 "mtu", MAC_PROP_MTU, sizeof (uint_t), &new_mtu);
5146 if (rv != 0)
5147 goto bail;
5148 rv = mac_maxsdu_update(mh, new_mtu);
5149 ASSERT(rv == 0);
5150 }
5151
5152 bail:
5153 i_mac_perim_exit(mip);
5154
5155 if (rv == 0 && old_mtu_arg != NULL)
5156 *old_mtu_arg = old_mtu;
5157 return (rv);
5158 }
5159
5160 /*
5161 * Return the RX h/w information for the group indexed by grp_num.
5162 */
5163 void
5164 mac_get_hwrxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5165 uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5166 char *clnts_name)
5167 {
5168 mac_impl_t *mip = (mac_impl_t *)mh;
5169 mac_grp_client_t *mcip;
5170 uint_t i = 0, index = 0;
5171 mac_ring_t *ring;
5172
5173 /* Revisit when we implement fully dynamic group allocation */
5174 ASSERT(grp_index >= 0 && grp_index < mip->mi_rx_group_count);
5175
5176 rw_enter(&mip->mi_rw_lock, RW_READER);
5177 *grp_num = mip->mi_rx_groups[grp_index].mrg_index;
5178 *type = mip->mi_rx_groups[grp_index].mrg_type;
5179 *n_rings = mip->mi_rx_groups[grp_index].mrg_cur_count;
5180 ring = mip->mi_rx_groups[grp_index].mrg_rings;
5181 for (index = 0; index < mip->mi_rx_groups[grp_index].mrg_cur_count;
5182 index++) {
5183 rings[index] = ring->mr_index;
5184 ring = ring->mr_next;
5185 }
5186 /* Assuming the 1st is the default group */
5187 index = 0;
5188 if (grp_index == 0) {
5189 (void) strlcpy(clnts_name, "<default,mcast>,",
5190 MAXCLIENTNAMELEN);
5191 index += strlen("<default,mcast>,");
5192 }
5193 for (mcip = mip->mi_rx_groups[grp_index].mrg_clients; mcip != NULL;
5194 mcip = mcip->mgc_next) {
5195 int name_len = strlen(mcip->mgc_client->mci_name);
5196
5197 /*
5198 * MAXCLIENTNAMELEN is the buffer size reserved for client
5199 * names.
5200 * XXXX Formating the client name string needs to be moved
5201 * to user land when fixing the size of dhi_clnts in
5202 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5203 * dhi_clntsin instead of MAXCLIENTNAMELEN
5204 */
5205 if (index + name_len >= MAXCLIENTNAMELEN) {
5206 index = MAXCLIENTNAMELEN;
5207 break;
5208 }
5209 bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5210 name_len);
5211 index += name_len;
5212 clnts_name[index++] = ',';
5213 i++;
5214 }
5215
5216 /* Get rid of the last , */
5217 if (index > 0)
5218 clnts_name[index - 1] = '\0';
5219 *n_clnts = i;
5220 rw_exit(&mip->mi_rw_lock);
5221 }
5222
5223 /*
5224 * Return the TX h/w information for the group indexed by grp_num.
5225 */
5226 void
5227 mac_get_hwtxgrp_info(mac_handle_t mh, int grp_index, uint_t *grp_num,
5228 uint_t *n_rings, uint_t *rings, uint_t *type, uint_t *n_clnts,
5229 char *clnts_name)
5230 {
5231 mac_impl_t *mip = (mac_impl_t *)mh;
5232 mac_grp_client_t *mcip;
5233 uint_t i = 0, index = 0;
5234 mac_ring_t *ring;
5235
5236 /* Revisit when we implement fully dynamic group allocation */
5237 ASSERT(grp_index >= 0 && grp_index <= mip->mi_tx_group_count);
5238
5239 rw_enter(&mip->mi_rw_lock, RW_READER);
5240 *grp_num = mip->mi_tx_groups[grp_index].mrg_index > 0 ?
5241 mip->mi_tx_groups[grp_index].mrg_index : grp_index;
5242 *type = mip->mi_tx_groups[grp_index].mrg_type;
5243 *n_rings = mip->mi_tx_groups[grp_index].mrg_cur_count;
5244 ring = mip->mi_tx_groups[grp_index].mrg_rings;
5245 for (index = 0; index < mip->mi_tx_groups[grp_index].mrg_cur_count;
5246 index++) {
5247 rings[index] = ring->mr_index;
5248 ring = ring->mr_next;
5249 }
5250 index = 0;
5251 /* Default group has an index of -1 */
5252 if (mip->mi_tx_groups[grp_index].mrg_index < 0) {
5253 (void) strlcpy(clnts_name, "<default>,",
5254 MAXCLIENTNAMELEN);
5255 index += strlen("<default>,");
5256 }
5257 for (mcip = mip->mi_tx_groups[grp_index].mrg_clients; mcip != NULL;
5258 mcip = mcip->mgc_next) {
5259 int name_len = strlen(mcip->mgc_client->mci_name);
5260
5261 /*
5262 * MAXCLIENTNAMELEN is the buffer size reserved for client
5263 * names.
5264 * XXXX Formating the client name string needs to be moved
5265 * to user land when fixing the size of dhi_clnts in
5266 * dld_hwgrpinfo_t. We should use n_clients * client_name for
5267 * dhi_clntsin instead of MAXCLIENTNAMELEN
5268 */
5269 if (index + name_len >= MAXCLIENTNAMELEN) {
5270 index = MAXCLIENTNAMELEN;
5271 break;
5272 }
5273 bcopy(mcip->mgc_client->mci_name, &(clnts_name[index]),
5274 name_len);
5275 index += name_len;
5276 clnts_name[index++] = ',';
5277 i++;
5278 }
5279
5280 /* Get rid of the last , */
5281 if (index > 0)
5282 clnts_name[index - 1] = '\0';
5283 *n_clnts = i;
5284 rw_exit(&mip->mi_rw_lock);
5285 }
5286
5287 /*
5288 * Return the group count for RX or TX.
5289 */
5290 uint_t
5291 mac_hwgrp_num(mac_handle_t mh, int type)
5292 {
5293 mac_impl_t *mip = (mac_impl_t *)mh;
5294
5295 /*
5296 * Return the Rx and Tx group count; for the Tx we need to
5297 * include the default too.
5298 */
5299 return (type == MAC_RING_TYPE_RX ? mip->mi_rx_group_count :
5300 mip->mi_tx_groups != NULL ? mip->mi_tx_group_count + 1 : 0);
5301 }
5302
5303 /*
5304 * The total number of free TX rings for this MAC.
5305 */
5306 uint_t
5307 mac_txavail_get(mac_handle_t mh)
5308 {
5309 mac_impl_t *mip = (mac_impl_t *)mh;
5310
5311 return (mip->mi_txrings_avail);
5312 }
5313
5314 /*
5315 * The total number of free RX rings for this MAC.
5316 */
5317 uint_t
5318 mac_rxavail_get(mac_handle_t mh)
5319 {
5320 mac_impl_t *mip = (mac_impl_t *)mh;
5321
5322 return (mip->mi_rxrings_avail);
5323 }
5324
5325 /*
5326 * The total number of reserved RX rings on this MAC.
5327 */
5328 uint_t
5329 mac_rxrsvd_get(mac_handle_t mh)
5330 {
5331 mac_impl_t *mip = (mac_impl_t *)mh;
5332
5333 return (mip->mi_rxrings_rsvd);
5334 }
5335
5336 /*
5337 * The total number of reserved TX rings on this MAC.
5338 */
5339 uint_t
5340 mac_txrsvd_get(mac_handle_t mh)
5341 {
5342 mac_impl_t *mip = (mac_impl_t *)mh;
5343
5344 return (mip->mi_txrings_rsvd);
5345 }
5346
5347 /*
5348 * Total number of free RX groups on this MAC.
5349 */
5350 uint_t
5351 mac_rxhwlnksavail_get(mac_handle_t mh)
5352 {
5353 mac_impl_t *mip = (mac_impl_t *)mh;
5354
5355 return (mip->mi_rxhwclnt_avail);
5356 }
5357
5358 /*
5359 * Total number of RX groups reserved on this MAC.
5360 */
5361 uint_t
5362 mac_rxhwlnksrsvd_get(mac_handle_t mh)
5363 {
5364 mac_impl_t *mip = (mac_impl_t *)mh;
5365
5366 return (mip->mi_rxhwclnt_used);
5367 }
5368
5369 /*
5370 * Total number of free TX groups on this MAC.
5371 */
5372 uint_t
5373 mac_txhwlnksavail_get(mac_handle_t mh)
5374 {
5375 mac_impl_t *mip = (mac_impl_t *)mh;
5376
5377 return (mip->mi_txhwclnt_avail);
5378 }
5379
5380 /*
5381 * Total number of TX groups reserved on this MAC.
5382 */
5383 uint_t
5384 mac_txhwlnksrsvd_get(mac_handle_t mh)
5385 {
5386 mac_impl_t *mip = (mac_impl_t *)mh;
5387
5388 return (mip->mi_txhwclnt_used);
5389 }
5390
5391 /*
5392 * Initialize the rings property for a mac client. A non-0 value for
5393 * rxring or txring specifies the number of rings required, a value
5394 * of MAC_RXRINGS_NONE/MAC_TXRINGS_NONE specifies that it doesn't need
5395 * any RX/TX rings and a value of MAC_RXRINGS_DONTCARE/MAC_TXRINGS_DONTCARE
5396 * means the system can decide whether it can give any rings or not.
5397 */
5398 void
5399 mac_client_set_rings(mac_client_handle_t mch, int rxrings, int txrings)
5400 {
5401 mac_client_impl_t *mcip = (mac_client_impl_t *)mch;
5402 mac_resource_props_t *mrp = MCIP_RESOURCE_PROPS(mcip);
5403
5404 if (rxrings != MAC_RXRINGS_DONTCARE) {
5405 mrp->mrp_mask |= MRP_RX_RINGS;
5406 mrp->mrp_nrxrings = rxrings;
5407 }
5408
5409 if (txrings != MAC_TXRINGS_DONTCARE) {
5410 mrp->mrp_mask |= MRP_TX_RINGS;
5411 mrp->mrp_ntxrings = txrings;
5412 }
5413 }