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 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
24 */
25
26 #include <sys/types.h>
27 #include <sys/cred.h>
28 #include <sys/sysmacros.h>
29 #include <sys/conf.h>
30 #include <sys/cmn_err.h>
31 #include <sys/list.h>
32 #include <sys/ksynch.h>
33 #include <sys/kmem.h>
34 #include <sys/stream.h>
35 #include <sys/modctl.h>
36 #include <sys/ddi.h>
37 #include <sys/sunddi.h>
38 #include <sys/atomic.h>
39 #include <sys/stat.h>
40 #include <sys/modhash.h>
41 #include <sys/strsubr.h>
42 #include <sys/strsun.h>
43 #include <sys/dlpi.h>
44 #include <sys/mac.h>
45 #include <sys/mac_provider.h>
46 #include <sys/mac_client.h>
47 #include <sys/mac_client_priv.h>
48 #include <sys/mac_ether.h>
49 #include <sys/dls.h>
50 #include <sys/pattr.h>
51 #include <sys/time.h>
52 #include <sys/vlan.h>
53 #include <sys/vnic.h>
54 #include <sys/vnic_impl.h>
55 #include <sys/mac_flow_impl.h>
56 #include <inet/ip_impl.h>
57
58 /*
59 * Note that for best performance, the VNIC is a passthrough design.
60 * For each VNIC corresponds a MAC client of the underlying MAC (lower MAC).
61 * This MAC client is opened by the VNIC driver at VNIC creation,
62 * and closed when the VNIC is deleted.
63 * When a MAC client of the VNIC itself opens a VNIC, the MAC layer
64 * (upper MAC) detects that the MAC being opened is a VNIC. Instead
65 * of allocating a new MAC client, it asks the VNIC driver to return
66 * the lower MAC client handle associated with the VNIC, and that handle
67 * is returned to the upper MAC client directly. This allows access
68 * by upper MAC clients of the VNIC to have direct access to the lower
69 * MAC client for the control path and data path.
70 *
71 * Due to this passthrough, some of the entry points exported by the
72 * VNIC driver are never directly invoked. These entry points include
73 * vnic_m_start, vnic_m_stop, vnic_m_promisc, vnic_m_multicst, etc.
74 */
75
76 static int vnic_m_start(void *);
77 static void vnic_m_stop(void *);
78 static int vnic_m_promisc(void *, boolean_t);
79 static int vnic_m_multicst(void *, boolean_t, const uint8_t *);
80 static int vnic_m_unicst(void *, const uint8_t *);
81 static int vnic_m_stat(void *, uint_t, uint64_t *);
82 static void vnic_m_ioctl(void *, queue_t *, mblk_t *);
83 static int vnic_m_setprop(void *, const char *, mac_prop_id_t, uint_t,
84 const void *);
85 static int vnic_m_getprop(void *, const char *, mac_prop_id_t, uint_t, void *);
86 static void vnic_m_propinfo(void *, const char *, mac_prop_id_t,
87 mac_prop_info_handle_t);
88 static mblk_t *vnic_m_tx(void *, mblk_t *);
89 static boolean_t vnic_m_capab_get(void *, mac_capab_t, void *);
90 static void vnic_notify_cb(void *, mac_notify_type_t);
91
92 static kmem_cache_t *vnic_cache;
93 static krwlock_t vnic_lock;
94 static uint_t vnic_count;
95
96 #define ANCHOR_VNIC_MIN_MTU 576
97 #define ANCHOR_VNIC_MAX_MTU 9000
98
99 /* hash of VNICs (vnic_t's), keyed by VNIC id */
100 static mod_hash_t *vnic_hash;
101 #define VNIC_HASHSZ 64
102 #define VNIC_HASH_KEY(vnic_id) ((mod_hash_key_t)(uintptr_t)vnic_id)
103
104 #define VNIC_M_CALLBACK_FLAGS \
105 (MC_IOCTL | MC_GETCAPAB | MC_SETPROP | MC_GETPROP | MC_PROPINFO)
106
107 static mac_callbacks_t vnic_m_callbacks = {
108 VNIC_M_CALLBACK_FLAGS,
109 vnic_m_stat,
110 vnic_m_start,
111 vnic_m_stop,
112 vnic_m_promisc,
113 vnic_m_multicst,
114 vnic_m_unicst,
115 vnic_m_tx,
116 NULL,
117 vnic_m_ioctl,
118 vnic_m_capab_get,
119 NULL,
120 NULL,
121 vnic_m_setprop,
122 vnic_m_getprop,
123 vnic_m_propinfo
124 };
125
126 void
127 vnic_dev_init(void)
128 {
129 vnic_cache = kmem_cache_create("vnic_cache",
130 sizeof (vnic_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
131
132 vnic_hash = mod_hash_create_idhash("vnic_hash",
133 VNIC_HASHSZ, mod_hash_null_valdtor);
134
135 rw_init(&vnic_lock, NULL, RW_DEFAULT, NULL);
136
137 vnic_count = 0;
138 }
139
140 void
141 vnic_dev_fini(void)
142 {
143 ASSERT(vnic_count == 0);
144
145 rw_destroy(&vnic_lock);
146 mod_hash_destroy_idhash(vnic_hash);
147 kmem_cache_destroy(vnic_cache);
148 }
149
150 uint_t
151 vnic_dev_count(void)
152 {
153 return (vnic_count);
154 }
155
156 static vnic_ioc_diag_t
157 vnic_mac2vnic_diag(mac_diag_t diag)
158 {
159 switch (diag) {
160 case MAC_DIAG_MACADDR_NIC:
161 return (VNIC_IOC_DIAG_MACADDR_NIC);
162 case MAC_DIAG_MACADDR_INUSE:
163 return (VNIC_IOC_DIAG_MACADDR_INUSE);
164 case MAC_DIAG_MACADDR_INVALID:
165 return (VNIC_IOC_DIAG_MACADDR_INVALID);
166 case MAC_DIAG_MACADDRLEN_INVALID:
167 return (VNIC_IOC_DIAG_MACADDRLEN_INVALID);
168 case MAC_DIAG_MACFACTORYSLOTINVALID:
169 return (VNIC_IOC_DIAG_MACFACTORYSLOTINVALID);
170 case MAC_DIAG_MACFACTORYSLOTUSED:
171 return (VNIC_IOC_DIAG_MACFACTORYSLOTUSED);
172 case MAC_DIAG_MACFACTORYSLOTALLUSED:
173 return (VNIC_IOC_DIAG_MACFACTORYSLOTALLUSED);
174 case MAC_DIAG_MACFACTORYNOTSUP:
175 return (VNIC_IOC_DIAG_MACFACTORYNOTSUP);
176 case MAC_DIAG_MACPREFIX_INVALID:
177 return (VNIC_IOC_DIAG_MACPREFIX_INVALID);
178 case MAC_DIAG_MACPREFIXLEN_INVALID:
179 return (VNIC_IOC_DIAG_MACPREFIXLEN_INVALID);
180 case MAC_DIAG_MACNO_HWRINGS:
181 return (VNIC_IOC_DIAG_NO_HWRINGS);
182 default:
183 return (VNIC_IOC_DIAG_NONE);
184 }
185 }
186
187 static int
188 vnic_unicast_add(vnic_t *vnic, vnic_mac_addr_type_t vnic_addr_type,
189 int *addr_slot, uint_t prefix_len, int *addr_len_ptr_arg,
190 uint8_t *mac_addr_arg, uint16_t flags, vnic_ioc_diag_t *diag,
191 uint16_t vid, boolean_t req_hwgrp_flag)
192 {
193 mac_diag_t mac_diag;
194 uint16_t mac_flags = 0;
195 int err;
196 uint_t addr_len;
197
198 if (flags & VNIC_IOC_CREATE_NODUPCHECK)
199 mac_flags |= MAC_UNICAST_NODUPCHECK;
200
201 switch (vnic_addr_type) {
202 case VNIC_MAC_ADDR_TYPE_FIXED:
203 case VNIC_MAC_ADDR_TYPE_VRID:
204 /*
205 * The MAC address value to assign to the VNIC
206 * is already provided in mac_addr_arg. addr_len_ptr_arg
207 * already contains the MAC address length.
208 */
209 break;
210
211 case VNIC_MAC_ADDR_TYPE_RANDOM:
212 /*
213 * Random MAC address. There are two sub-cases:
214 *
215 * 1 - If mac_len == 0, a new MAC address is generated.
216 * The length of the MAC address to generated depends
217 * on the type of MAC used. The prefix to use for the MAC
218 * address is stored in the most significant bytes
219 * of the mac_addr argument, and its length is specified
220 * by the mac_prefix_len argument. This prefix can
221 * correspond to a IEEE OUI in the case of Ethernet,
222 * for example.
223 *
224 * 2 - If mac_len > 0, the address was already picked
225 * randomly, and is now passed back during VNIC
226 * re-creation. The mac_addr argument contains the MAC
227 * address that was generated. We distinguish this
228 * case from the fixed MAC address case, since we
229 * want the user consumers to know, when they query
230 * the list of VNICs, that a VNIC was assigned a
231 * random MAC address vs assigned a fixed address
232 * specified by the user.
233 */
234
235 /*
236 * If it's a pre-generated address, we're done. mac_addr_arg
237 * and addr_len_ptr_arg already contain the MAC address
238 * value and length.
239 */
240 if (*addr_len_ptr_arg > 0)
241 break;
242
243 /* generate a new random MAC address */
244 if ((err = mac_addr_random(vnic->vn_mch,
245 prefix_len, mac_addr_arg, &mac_diag)) != 0) {
246 *diag = vnic_mac2vnic_diag(mac_diag);
247 return (err);
248 }
249 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh);
250 break;
251
252 case VNIC_MAC_ADDR_TYPE_FACTORY:
253 err = mac_addr_factory_reserve(vnic->vn_mch, addr_slot);
254 if (err != 0) {
255 if (err == EINVAL)
256 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTINVALID;
257 if (err == EBUSY)
258 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTUSED;
259 if (err == ENOSPC)
260 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTALLUSED;
261 return (err);
262 }
263
264 mac_addr_factory_value(vnic->vn_lower_mh, *addr_slot,
265 mac_addr_arg, &addr_len, NULL, NULL);
266 *addr_len_ptr_arg = addr_len;
267 break;
268
269 case VNIC_MAC_ADDR_TYPE_AUTO:
270 /* first try to allocate a factory MAC address */
271 err = mac_addr_factory_reserve(vnic->vn_mch, addr_slot);
272 if (err == 0) {
273 mac_addr_factory_value(vnic->vn_lower_mh, *addr_slot,
274 mac_addr_arg, &addr_len, NULL, NULL);
275 vnic_addr_type = VNIC_MAC_ADDR_TYPE_FACTORY;
276 *addr_len_ptr_arg = addr_len;
277 break;
278 }
279
280 /*
281 * Allocating a factory MAC address failed, generate a
282 * random MAC address instead.
283 */
284 if ((err = mac_addr_random(vnic->vn_mch,
285 prefix_len, mac_addr_arg, &mac_diag)) != 0) {
286 *diag = vnic_mac2vnic_diag(mac_diag);
287 return (err);
288 }
289 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh);
290 vnic_addr_type = VNIC_MAC_ADDR_TYPE_RANDOM;
291 break;
292 case VNIC_MAC_ADDR_TYPE_PRIMARY:
293 /*
294 * We get the address here since we copy it in the
295 * vnic's vn_addr.
296 * We can't ask for hardware resources since we
297 * don't currently support hardware classification
298 * for these MAC clients.
299 */
300 if (req_hwgrp_flag) {
301 *diag = VNIC_IOC_DIAG_NO_HWRINGS;
302 return (ENOTSUP);
303 }
304 mac_unicast_primary_get(vnic->vn_lower_mh, mac_addr_arg);
305 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh);
306 mac_flags |= MAC_UNICAST_VNIC_PRIMARY;
307 break;
308 }
309
310 vnic->vn_addr_type = vnic_addr_type;
311
312 err = mac_unicast_add(vnic->vn_mch, mac_addr_arg, mac_flags,
313 &vnic->vn_muh, vid, &mac_diag);
314 if (err != 0) {
315 if (vnic_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) {
316 /* release factory MAC address */
317 mac_addr_factory_release(vnic->vn_mch, *addr_slot);
318 }
319 *diag = vnic_mac2vnic_diag(mac_diag);
320 }
321
322 return (err);
323 }
324
325 /*
326 * Create a new VNIC upon request from administrator.
327 * Returns 0 on success, an errno on failure.
328 */
329 /* ARGSUSED */
330 int
331 vnic_dev_create(datalink_id_t vnic_id, datalink_id_t linkid,
332 vnic_mac_addr_type_t *vnic_addr_type, int *mac_len, uchar_t *mac_addr,
333 int *mac_slot, uint_t mac_prefix_len, uint16_t vid, vrid_t vrid,
334 int af, mac_resource_props_t *mrp, uint32_t flags, vnic_ioc_diag_t *diag,
335 cred_t *credp)
336 {
337 vnic_t *vnic;
338 mac_register_t *mac;
339 int err;
340 boolean_t is_anchor = ((flags & VNIC_IOC_CREATE_ANCHOR) != 0);
341 char vnic_name[MAXNAMELEN];
342 const mac_info_t *minfop;
343 uint32_t req_hwgrp_flag = B_FALSE;
344
345 *diag = VNIC_IOC_DIAG_NONE;
346
347 rw_enter(&vnic_lock, RW_WRITER);
348
349 /* does a VNIC with the same id already exist? */
350 err = mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id),
351 (mod_hash_val_t *)&vnic);
352 if (err == 0) {
353 rw_exit(&vnic_lock);
354 return (EEXIST);
355 }
356
357 vnic = kmem_cache_alloc(vnic_cache, KM_NOSLEEP);
358 if (vnic == NULL) {
359 rw_exit(&vnic_lock);
360 return (ENOMEM);
361 }
362
363 bzero(vnic, sizeof (*vnic));
364
365 vnic->vn_id = vnic_id;
366 vnic->vn_link_id = linkid;
367 vnic->vn_vrid = vrid;
368 vnic->vn_af = af;
369
370 if (!is_anchor) {
371 if (linkid == DATALINK_INVALID_LINKID) {
372 err = EINVAL;
373 goto bail;
374 }
375
376 /*
377 * Open the lower MAC and assign its initial bandwidth and
378 * MAC address. We do this here during VNIC creation and
379 * do not wait until the upper MAC client open so that we
380 * can validate the VNIC creation parameters (bandwidth,
381 * MAC address, etc) and reserve a factory MAC address if
382 * one was requested.
383 */
384 err = mac_open_by_linkid(linkid, &vnic->vn_lower_mh);
385 if (err != 0)
386 goto bail;
387
388 /*
389 * VNIC(vlan) over VNICs(vlans) is not supported.
390 */
391 if (mac_is_vnic(vnic->vn_lower_mh)) {
392 err = EINVAL;
393 goto bail;
394 }
395
396 /* only ethernet support for now */
397 minfop = mac_info(vnic->vn_lower_mh);
398 if (minfop->mi_nativemedia != DL_ETHER) {
399 err = ENOTSUP;
400 goto bail;
401 }
402
403 (void) dls_mgmt_get_linkinfo(vnic_id, vnic_name, NULL, NULL,
404 NULL);
405 err = mac_client_open(vnic->vn_lower_mh, &vnic->vn_mch,
406 vnic_name, MAC_OPEN_FLAGS_IS_VNIC);
407 if (err != 0)
408 goto bail;
409
410 if (mrp != NULL) {
411 if ((mrp->mrp_mask & MRP_RX_RINGS) != 0 ||
412 (mrp->mrp_mask & MRP_TX_RINGS) != 0) {
413 req_hwgrp_flag = B_TRUE;
414 }
415 err = mac_client_set_resources(vnic->vn_mch, mrp);
416 if (err != 0)
417 goto bail;
418 }
419 /* assign a MAC address to the VNIC */
420
421 err = vnic_unicast_add(vnic, *vnic_addr_type, mac_slot,
422 mac_prefix_len, mac_len, mac_addr, flags, diag, vid,
423 req_hwgrp_flag);
424 if (err != 0) {
425 vnic->vn_muh = NULL;
426 if (diag != NULL && req_hwgrp_flag)
427 *diag = VNIC_IOC_DIAG_NO_HWRINGS;
428 goto bail;
429 }
430
431 /* register to receive notification from underlying MAC */
432 vnic->vn_mnh = mac_notify_add(vnic->vn_lower_mh, vnic_notify_cb,
433 vnic);
434
435 *vnic_addr_type = vnic->vn_addr_type;
436 vnic->vn_addr_len = *mac_len;
437 vnic->vn_vid = vid;
438
439 bcopy(mac_addr, vnic->vn_addr, vnic->vn_addr_len);
440
441 if (vnic->vn_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY)
442 vnic->vn_slot_id = *mac_slot;
443
444 /*
445 * Set the initial VNIC capabilities. If the VNIC is created
446 * over MACs which does not support nactive vlan, disable
447 * VNIC's hardware checksum capability if its VID is not 0,
448 * since the underlying MAC would get the hardware checksum
449 * offset wrong in case of VLAN packets.
450 */
451 if (vid == 0 || !mac_capab_get(vnic->vn_lower_mh,
452 MAC_CAPAB_NO_NATIVEVLAN, NULL)) {
453 if (!mac_capab_get(vnic->vn_lower_mh, MAC_CAPAB_HCKSUM,
454 &vnic->vn_hcksum_txflags))
455 vnic->vn_hcksum_txflags = 0;
456 } else {
457 vnic->vn_hcksum_txflags = 0;
458 }
459 }
460
461 /* register with the MAC module */
462 if ((mac = mac_alloc(MAC_VERSION)) == NULL)
463 goto bail;
464
465 mac->m_type_ident = MAC_PLUGIN_IDENT_ETHER;
466 mac->m_driver = vnic;
467 mac->m_dip = vnic_get_dip();
468 mac->m_instance = (uint_t)-1;
469 mac->m_src_addr = vnic->vn_addr;
470 mac->m_callbacks = &vnic_m_callbacks;
471
472 if (!is_anchor) {
473 /*
474 * If this is a VNIC based VLAN, then we check for the
475 * margin unless it has been created with the force
476 * flag. If we are configuring a VLAN over an etherstub,
477 * we don't check the margin even if force is not set.
478 */
479 if (vid == 0 || (flags & VNIC_IOC_CREATE_FORCE) != 0) {
480 if (vid != VLAN_ID_NONE)
481 vnic->vn_force = B_TRUE;
482 /*
483 * As the current margin size of the underlying mac is
484 * used to determine the margin size of the VNIC
485 * itself, request the underlying mac not to change
486 * to a smaller margin size.
487 */
488 err = mac_margin_add(vnic->vn_lower_mh,
489 &vnic->vn_margin, B_TRUE);
490 ASSERT(err == 0);
491 } else {
492 vnic->vn_margin = VLAN_TAGSZ;
493 err = mac_margin_add(vnic->vn_lower_mh,
494 &vnic->vn_margin, B_FALSE);
495 if (err != 0) {
496 mac_free(mac);
497 if (diag != NULL)
498 *diag = VNIC_IOC_DIAG_MACMARGIN_INVALID;
499 goto bail;
500 }
501 }
502
503 mac_sdu_get(vnic->vn_lower_mh, &mac->m_min_sdu,
504 &mac->m_max_sdu);
505 } else {
506 vnic->vn_margin = VLAN_TAGSZ;
507 mac->m_min_sdu = ANCHOR_VNIC_MIN_MTU;
508 mac->m_max_sdu = ANCHOR_VNIC_MAX_MTU;
509 }
510
511 mac->m_margin = vnic->vn_margin;
512
513 err = mac_register(mac, &vnic->vn_mh);
514 mac_free(mac);
515 if (err != 0) {
516 VERIFY(is_anchor || mac_margin_remove(vnic->vn_lower_mh,
517 vnic->vn_margin) == 0);
518 goto bail;
519 }
520
521 /* Set the VNIC's MAC in the client */
522 if (!is_anchor)
523 mac_set_upper_mac(vnic->vn_mch, vnic->vn_mh, mrp);
524
525 err = dls_devnet_create(vnic->vn_mh, vnic->vn_id, crgetzoneid(credp));
526 if (err != 0) {
527 VERIFY(is_anchor || mac_margin_remove(vnic->vn_lower_mh,
528 vnic->vn_margin) == 0);
529 (void) mac_unregister(vnic->vn_mh);
530 goto bail;
531 }
532
533 /* add new VNIC to hash table */
534 err = mod_hash_insert(vnic_hash, VNIC_HASH_KEY(vnic_id),
535 (mod_hash_val_t)vnic);
536 ASSERT(err == 0);
537 vnic_count++;
538
539 vnic->vn_enabled = B_TRUE;
540 rw_exit(&vnic_lock);
541
542 return (0);
543
544 bail:
545 rw_exit(&vnic_lock);
546 if (!is_anchor) {
547 if (vnic->vn_mnh != NULL)
548 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE);
549 if (vnic->vn_muh != NULL)
550 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh);
551 if (vnic->vn_mch != NULL)
552 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC);
553 if (vnic->vn_lower_mh != NULL)
554 mac_close(vnic->vn_lower_mh);
555 }
556
557 kmem_cache_free(vnic_cache, vnic);
558 return (err);
559 }
560
561 /*
562 * Modify the properties of an existing VNIC.
563 */
564 /* ARGSUSED */
565 int
566 vnic_dev_modify(datalink_id_t vnic_id, uint_t modify_mask,
567 vnic_mac_addr_type_t mac_addr_type, uint_t mac_len, uchar_t *mac_addr,
568 uint_t mac_slot, mac_resource_props_t *mrp)
569 {
570 vnic_t *vnic = NULL;
571
572 rw_enter(&vnic_lock, RW_WRITER);
573
574 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id),
575 (mod_hash_val_t *)&vnic) != 0) {
576 rw_exit(&vnic_lock);
577 return (ENOENT);
578 }
579
580 rw_exit(&vnic_lock);
581
582 return (0);
583 }
584
585 /* ARGSUSED */
586 int
587 vnic_dev_delete(datalink_id_t vnic_id, uint32_t flags, cred_t *credp)
588 {
589 vnic_t *vnic = NULL;
590 mod_hash_val_t val;
591 datalink_id_t tmpid;
592 int rc;
593
594 rw_enter(&vnic_lock, RW_WRITER);
595
596 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id),
597 (mod_hash_val_t *)&vnic) != 0) {
598 rw_exit(&vnic_lock);
599 return (ENOENT);
600 }
601
602 if ((rc = dls_devnet_destroy(vnic->vn_mh, &tmpid, B_TRUE)) != 0) {
603 rw_exit(&vnic_lock);
604 return (rc);
605 }
606
607 ASSERT(vnic_id == tmpid);
608
609 /*
610 * We cannot unregister the MAC yet. Unregistering would
611 * free up mac_impl_t which should not happen at this time.
612 * So disable mac_impl_t by calling mac_disable(). This will prevent
613 * any new claims on mac_impl_t.
614 */
615 if ((rc = mac_disable(vnic->vn_mh)) != 0) {
616 (void) dls_devnet_create(vnic->vn_mh, vnic_id,
617 crgetzoneid(credp));
618 rw_exit(&vnic_lock);
619 return (rc);
620 }
621
622 vnic->vn_enabled = B_FALSE;
623 (void) mod_hash_remove(vnic_hash, VNIC_HASH_KEY(vnic_id), &val);
624 ASSERT(vnic == (vnic_t *)val);
625 vnic_count--;
626 rw_exit(&vnic_lock);
627
628 /*
629 * XXX-nicolas shouldn't have a void cast here, if it's
630 * expected that the function will never fail, then we should
631 * have an ASSERT().
632 */
633 (void) mac_unregister(vnic->vn_mh);
634
635 if (vnic->vn_lower_mh != NULL) {
636 /*
637 * Check if MAC address for the vnic was obtained from the
638 * factory MAC addresses. If yes, release it.
639 */
640 if (vnic->vn_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) {
641 (void) mac_addr_factory_release(vnic->vn_mch,
642 vnic->vn_slot_id);
643 }
644 (void) mac_margin_remove(vnic->vn_lower_mh, vnic->vn_margin);
645 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE);
646 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh);
647 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC);
648 mac_close(vnic->vn_lower_mh);
649 }
650
651 kmem_cache_free(vnic_cache, vnic);
652 return (0);
653 }
654
655 /* ARGSUSED */
656 mblk_t *
657 vnic_m_tx(void *arg, mblk_t *mp_chain)
658 {
659 /*
660 * This function could be invoked for an anchor VNIC when sending
661 * broadcast and multicast packets, and unicast packets which did
662 * not match any local known destination.
663 */
664 freemsgchain(mp_chain);
665 return (NULL);
666 }
667
668 /*ARGSUSED*/
669 static void
670 vnic_m_ioctl(void *arg, queue_t *q, mblk_t *mp)
671 {
672 miocnak(q, mp, 0, ENOTSUP);
673 }
674
675 /*
676 * This entry point cannot be passed-through, since it is invoked
677 * for the per-VNIC kstats which must be exported independently
678 * of the existence of VNIC MAC clients.
679 */
680 static int
681 vnic_m_stat(void *arg, uint_t stat, uint64_t *val)
682 {
683 vnic_t *vnic = arg;
684 int rval = 0;
685
686 if (vnic->vn_lower_mh == NULL) {
687 /*
688 * It's an anchor VNIC, which does not have any
689 * statistics in itself.
690 */
691 return (ENOTSUP);
692 }
693
694 /*
695 * ENOTSUP must be reported for unsupported stats, the VNIC
696 * driver reports a subset of the stats that would
697 * be returned by a real piece of hardware.
698 */
699
700 switch (stat) {
701 case MAC_STAT_LINK_STATE:
702 case MAC_STAT_LINK_UP:
703 case MAC_STAT_PROMISC:
704 case MAC_STAT_IFSPEED:
705 case MAC_STAT_MULTIRCV:
706 case MAC_STAT_MULTIXMT:
707 case MAC_STAT_BRDCSTRCV:
708 case MAC_STAT_BRDCSTXMT:
709 case MAC_STAT_OPACKETS:
710 case MAC_STAT_OBYTES:
711 case MAC_STAT_IERRORS:
712 case MAC_STAT_OERRORS:
713 case MAC_STAT_RBYTES:
714 case MAC_STAT_IPACKETS:
715 *val = mac_client_stat_get(vnic->vn_mch, stat);
716 break;
717 default:
718 rval = ENOTSUP;
719 }
720
721 return (rval);
722 }
723
724 /*
725 * Invoked by the upper MAC to retrieve the lower MAC client handle
726 * corresponding to a VNIC. A pointer to this function is obtained
727 * by the upper MAC via capability query.
728 *
729 * XXX-nicolas Note: this currently causes all VNIC MAC clients to
730 * receive the same MAC client handle for the same VNIC. This is ok
731 * as long as we have only one VNIC MAC client which sends and
732 * receives data, but we don't currently enforce this at the MAC layer.
733 */
734 static void *
735 vnic_mac_client_handle(void *vnic_arg)
736 {
737 vnic_t *vnic = vnic_arg;
738
739 return (vnic->vn_mch);
740 }
741
742
743 /*
744 * Return information about the specified capability.
745 */
746 /* ARGSUSED */
747 static boolean_t
748 vnic_m_capab_get(void *arg, mac_capab_t cap, void *cap_data)
749 {
750 vnic_t *vnic = arg;
751
752 switch (cap) {
753 case MAC_CAPAB_HCKSUM: {
754 uint32_t *hcksum_txflags = cap_data;
755
756 *hcksum_txflags = vnic->vn_hcksum_txflags &
757 (HCKSUM_INET_FULL_V4 | HCKSUM_IPHDRCKSUM |
758 HCKSUM_INET_PARTIAL);
759 break;
760 }
761 case MAC_CAPAB_VNIC: {
762 mac_capab_vnic_t *vnic_capab = cap_data;
763
764 if (vnic->vn_lower_mh == NULL) {
765 /*
766 * It's an anchor VNIC, we don't have an underlying
767 * NIC and MAC client handle.
768 */
769 return (B_FALSE);
770 }
771
772 if (vnic_capab != NULL) {
773 vnic_capab->mcv_arg = vnic;
774 vnic_capab->mcv_mac_client_handle =
775 vnic_mac_client_handle;
776 }
777 break;
778 }
779 case MAC_CAPAB_ANCHOR_VNIC: {
780 /* since it's an anchor VNIC we don't have lower mac handle */
781 if (vnic->vn_lower_mh == NULL) {
782 ASSERT(vnic->vn_link_id == 0);
783 return (B_TRUE);
784 }
785 return (B_FALSE);
786 }
787 case MAC_CAPAB_NO_NATIVEVLAN:
788 return (B_FALSE);
789 case MAC_CAPAB_NO_ZCOPY:
790 return (B_TRUE);
791 case MAC_CAPAB_VRRP: {
792 mac_capab_vrrp_t *vrrp_capab = cap_data;
793
794 if (vnic->vn_vrid != 0) {
795 if (vrrp_capab != NULL)
796 vrrp_capab->mcv_af = vnic->vn_af;
797 return (B_TRUE);
798 }
799 return (B_FALSE);
800 }
801 default:
802 return (B_FALSE);
803 }
804 return (B_TRUE);
805 }
806
807 /* ARGSUSED */
808 static int
809 vnic_m_start(void *arg)
810 {
811 return (0);
812 }
813
814 /* ARGSUSED */
815 static void
816 vnic_m_stop(void *arg)
817 {
818 }
819
820 /* ARGSUSED */
821 static int
822 vnic_m_promisc(void *arg, boolean_t on)
823 {
824 return (0);
825 }
826
827 /* ARGSUSED */
828 static int
829 vnic_m_multicst(void *arg, boolean_t add, const uint8_t *addrp)
830 {
831 return (0);
832 }
833
834 static int
835 vnic_m_unicst(void *arg, const uint8_t *macaddr)
836 {
837 vnic_t *vnic = arg;
838
839 return (mac_vnic_unicast_set(vnic->vn_mch, macaddr));
840 }
841
842 /*
843 * Callback functions for set/get of properties
844 */
845 /*ARGSUSED*/
846 static int
847 vnic_m_setprop(void *m_driver, const char *pr_name, mac_prop_id_t pr_num,
848 uint_t pr_valsize, const void *pr_val)
849 {
850 int err = ENOTSUP;
851 vnic_t *vn = m_driver;
852
853 switch (pr_num) {
854 case MAC_PROP_MTU: {
855 uint32_t mtu;
856
857 /* allow setting MTU only on an etherstub */
858 if (vn->vn_link_id != DATALINK_INVALID_LINKID)
859 return (err);
860
861 if (pr_valsize < sizeof (mtu)) {
862 err = EINVAL;
863 break;
864 }
865 bcopy(pr_val, &mtu, sizeof (mtu));
866 if (mtu < ANCHOR_VNIC_MIN_MTU || mtu > ANCHOR_VNIC_MAX_MTU) {
867 err = EINVAL;
868 break;
869 }
870 err = mac_maxsdu_update(vn->vn_mh, mtu);
871 break;
872 }
873 case MAC_PROP_VN_PROMISC_FILTERED: {
874 boolean_t filtered;
875
876 if (pr_valsize < sizeof (filtered)) {
877 err = EINVAL;
878 break;
879 }
880
881 bcopy(pr_val, &filtered, sizeof (filtered));
882 err = mac_set_promisc_filtered(vn->vn_mch, filtered);
883 }
884 default:
885 break;
886 }
887 return (err);
888 }
889
890 static int
891 vnic_m_getprop(void *arg, const char *pr_name, mac_prop_id_t pr_num,
892 uint_t pr_valsize, void *pr_val)
893 {
894 vnic_t *vn = arg;
895 int ret = 0;
896 boolean_t out;
897
898 switch (pr_num) {
899 case MAC_PROP_VN_PROMISC_FILTERED:
900 out = mac_get_promisc_filtered(vn->vn_mch);
901 ASSERT(pr_valsize >= sizeof (boolean_t));
902 bcopy(&out, pr_val, sizeof (boolean_t));
903 break;
904 default:
905 ret = EINVAL;
906 break;
907 }
908
909 return (ret);
910 }
911
912 /* ARGSUSED */
913 static void vnic_m_propinfo(void *m_driver, const char *pr_name,
914 mac_prop_id_t pr_num, mac_prop_info_handle_t prh)
915 {
916 vnic_t *vn = m_driver;
917
918 /* MTU setting allowed only on an etherstub */
919 if (vn->vn_link_id != DATALINK_INVALID_LINKID)
920 return;
921
922 switch (pr_num) {
923 case MAC_PROP_MTU:
924 mac_prop_info_set_range_uint32(prh,
925 ANCHOR_VNIC_MIN_MTU, ANCHOR_VNIC_MAX_MTU);
926 break;
927 }
928 }
929
930
931 int
932 vnic_info(vnic_info_t *info, cred_t *credp)
933 {
934 vnic_t *vnic;
935 int err;
936
937 /* Make sure that the VNIC link is visible from the caller's zone. */
938 if (!dls_devnet_islinkvisible(info->vn_vnic_id, crgetzoneid(credp)))
939 return (ENOENT);
940
941 rw_enter(&vnic_lock, RW_WRITER);
942
943 err = mod_hash_find(vnic_hash, VNIC_HASH_KEY(info->vn_vnic_id),
944 (mod_hash_val_t *)&vnic);
945 if (err != 0) {
946 rw_exit(&vnic_lock);
947 return (ENOENT);
948 }
949
950 info->vn_link_id = vnic->vn_link_id;
951 info->vn_mac_addr_type = vnic->vn_addr_type;
952 info->vn_mac_len = vnic->vn_addr_len;
953 bcopy(vnic->vn_addr, info->vn_mac_addr, MAXMACADDRLEN);
954 info->vn_mac_slot = vnic->vn_slot_id;
955 info->vn_mac_prefix_len = 0;
956 info->vn_vid = vnic->vn_vid;
957 info->vn_force = vnic->vn_force;
958 info->vn_vrid = vnic->vn_vrid;
959 info->vn_af = vnic->vn_af;
960
961 bzero(&info->vn_resource_props, sizeof (mac_resource_props_t));
962 if (vnic->vn_mch != NULL)
963 mac_resource_ctl_get(vnic->vn_mch, &info->vn_resource_props);
964
965 rw_exit(&vnic_lock);
966 return (0);
967 }
968
969 static void
970 vnic_notify_cb(void *arg, mac_notify_type_t type)
971 {
972 vnic_t *vnic = arg;
973
974 /*
975 * Do not deliver notifications if the vnic is not fully initialized
976 * or is in process of being torn down.
977 */
978 if (!vnic->vn_enabled)
979 return;
980
981 switch (type) {
982 case MAC_NOTE_UNICST:
983 /*
984 * Only the VLAN VNIC needs to be notified with primary MAC
985 * address change.
986 */
987 if (vnic->vn_addr_type != VNIC_MAC_ADDR_TYPE_PRIMARY)
988 return;
989
990 /* the unicast MAC address value */
991 mac_unicast_primary_get(vnic->vn_lower_mh, vnic->vn_addr);
992
993 /* notify its upper layer MAC about MAC address change */
994 mac_unicst_update(vnic->vn_mh, (const uint8_t *)vnic->vn_addr);
995 break;
996
997 case MAC_NOTE_LINK:
998 mac_link_update(vnic->vn_mh,
999 mac_client_stat_get(vnic->vn_mch, MAC_STAT_LINK_STATE));
1000 break;
1001
1002 default:
1003 break;
1004 }
1005 }