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) 2014, 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 * VNICs support multiple upper mac clients to enable support for 76 * multiple MAC addresses on the VNIC. When the VNIC is created the 77 * initial mac client is the primary upper mac. Any additional mac 78 * clients are secondary macs. 79 */ 80 81 static int vnic_m_start(void *); 82 static void vnic_m_stop(void *); 83 static int vnic_m_promisc(void *, boolean_t); 84 static int vnic_m_multicst(void *, boolean_t, const uint8_t *); 85 static int vnic_m_unicst(void *, const uint8_t *); 86 static int vnic_m_stat(void *, uint_t, uint64_t *); 87 static void vnic_m_ioctl(void *, queue_t *, mblk_t *); 88 static int vnic_m_setprop(void *, const char *, mac_prop_id_t, uint_t, 89 const void *); 90 static int vnic_m_getprop(void *, const char *, mac_prop_id_t, uint_t, void *); 91 static void vnic_m_propinfo(void *, const char *, mac_prop_id_t, 92 mac_prop_info_handle_t); 93 static mblk_t *vnic_m_tx(void *, mblk_t *); 94 static boolean_t vnic_m_capab_get(void *, mac_capab_t, void *); 95 static void vnic_notify_cb(void *, mac_notify_type_t); 96 static void vnic_cleanup_secondary_macs(vnic_t *, int); 97 98 static kmem_cache_t *vnic_cache; 99 static krwlock_t vnic_lock; 100 static uint_t vnic_count; 101 102 #define ANCHOR_VNIC_MIN_MTU 576 103 #define ANCHOR_VNIC_MAX_MTU 9000 104 105 /* hash of VNICs (vnic_t's), keyed by VNIC id */ 106 static mod_hash_t *vnic_hash; 107 #define VNIC_HASHSZ 64 108 #define VNIC_HASH_KEY(vnic_id) ((mod_hash_key_t)(uintptr_t)vnic_id) 109 110 #define VNIC_M_CALLBACK_FLAGS \ 111 (MC_IOCTL | MC_GETCAPAB | MC_SETPROP | MC_GETPROP | MC_PROPINFO) 112 113 static mac_callbacks_t vnic_m_callbacks = { 114 VNIC_M_CALLBACK_FLAGS, 115 vnic_m_stat, 116 vnic_m_start, 117 vnic_m_stop, 118 vnic_m_promisc, 119 vnic_m_multicst, 120 vnic_m_unicst, 121 vnic_m_tx, 122 NULL, 123 vnic_m_ioctl, 124 vnic_m_capab_get, 125 NULL, 126 NULL, 127 vnic_m_setprop, 128 vnic_m_getprop, 129 vnic_m_propinfo 130 }; 131 132 void 133 vnic_dev_init(void) 134 { 135 vnic_cache = kmem_cache_create("vnic_cache", 136 sizeof (vnic_t), 0, NULL, NULL, NULL, NULL, NULL, 0); 137 138 vnic_hash = mod_hash_create_idhash("vnic_hash", 139 VNIC_HASHSZ, mod_hash_null_valdtor); 140 141 rw_init(&vnic_lock, NULL, RW_DEFAULT, NULL); 142 143 vnic_count = 0; 144 } 145 146 void 147 vnic_dev_fini(void) 148 { 149 ASSERT(vnic_count == 0); 150 151 rw_destroy(&vnic_lock); 152 mod_hash_destroy_idhash(vnic_hash); 153 kmem_cache_destroy(vnic_cache); 154 } 155 156 uint_t 157 vnic_dev_count(void) 158 { 159 return (vnic_count); 160 } 161 162 static vnic_ioc_diag_t 163 vnic_mac2vnic_diag(mac_diag_t diag) 164 { 165 switch (diag) { 166 case MAC_DIAG_MACADDR_NIC: 167 return (VNIC_IOC_DIAG_MACADDR_NIC); 168 case MAC_DIAG_MACADDR_INUSE: 169 return (VNIC_IOC_DIAG_MACADDR_INUSE); 170 case MAC_DIAG_MACADDR_INVALID: 171 return (VNIC_IOC_DIAG_MACADDR_INVALID); 172 case MAC_DIAG_MACADDRLEN_INVALID: 173 return (VNIC_IOC_DIAG_MACADDRLEN_INVALID); 174 case MAC_DIAG_MACFACTORYSLOTINVALID: 175 return (VNIC_IOC_DIAG_MACFACTORYSLOTINVALID); 176 case MAC_DIAG_MACFACTORYSLOTUSED: 177 return (VNIC_IOC_DIAG_MACFACTORYSLOTUSED); 178 case MAC_DIAG_MACFACTORYSLOTALLUSED: 179 return (VNIC_IOC_DIAG_MACFACTORYSLOTALLUSED); 180 case MAC_DIAG_MACFACTORYNOTSUP: 181 return (VNIC_IOC_DIAG_MACFACTORYNOTSUP); 182 case MAC_DIAG_MACPREFIX_INVALID: 183 return (VNIC_IOC_DIAG_MACPREFIX_INVALID); 184 case MAC_DIAG_MACPREFIXLEN_INVALID: 185 return (VNIC_IOC_DIAG_MACPREFIXLEN_INVALID); 186 case MAC_DIAG_MACNO_HWRINGS: 187 return (VNIC_IOC_DIAG_NO_HWRINGS); 188 default: 189 return (VNIC_IOC_DIAG_NONE); 190 } 191 } 192 193 static int 194 vnic_unicast_add(vnic_t *vnic, vnic_mac_addr_type_t vnic_addr_type, 195 int *addr_slot, uint_t prefix_len, int *addr_len_ptr_arg, 196 uint8_t *mac_addr_arg, uint16_t flags, vnic_ioc_diag_t *diag, 197 uint16_t vid, boolean_t req_hwgrp_flag) 198 { 199 mac_diag_t mac_diag; 200 uint16_t mac_flags = 0; 201 int err; 202 uint_t addr_len; 203 204 if (flags & VNIC_IOC_CREATE_NODUPCHECK) 205 mac_flags |= MAC_UNICAST_NODUPCHECK; 206 207 switch (vnic_addr_type) { 208 case VNIC_MAC_ADDR_TYPE_FIXED: 209 case VNIC_MAC_ADDR_TYPE_VRID: 210 /* 211 * The MAC address value to assign to the VNIC 212 * is already provided in mac_addr_arg. addr_len_ptr_arg 213 * already contains the MAC address length. 214 */ 215 break; 216 217 case VNIC_MAC_ADDR_TYPE_RANDOM: 218 /* 219 * Random MAC address. There are two sub-cases: 220 * 221 * 1 - If mac_len == 0, a new MAC address is generated. 222 * The length of the MAC address to generated depends 223 * on the type of MAC used. The prefix to use for the MAC 224 * address is stored in the most significant bytes 225 * of the mac_addr argument, and its length is specified 226 * by the mac_prefix_len argument. This prefix can 227 * correspond to a IEEE OUI in the case of Ethernet, 228 * for example. 229 * 230 * 2 - If mac_len > 0, the address was already picked 231 * randomly, and is now passed back during VNIC 232 * re-creation. The mac_addr argument contains the MAC 233 * address that was generated. We distinguish this 234 * case from the fixed MAC address case, since we 235 * want the user consumers to know, when they query 236 * the list of VNICs, that a VNIC was assigned a 237 * random MAC address vs assigned a fixed address 238 * specified by the user. 239 */ 240 241 /* 242 * If it's a pre-generated address, we're done. mac_addr_arg 243 * and addr_len_ptr_arg already contain the MAC address 244 * value and length. 245 */ 246 if (*addr_len_ptr_arg > 0) 247 break; 248 249 /* generate a new random MAC address */ 250 if ((err = mac_addr_random(vnic->vn_mch, 251 prefix_len, mac_addr_arg, &mac_diag)) != 0) { 252 *diag = vnic_mac2vnic_diag(mac_diag); 253 return (err); 254 } 255 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh); 256 break; 257 258 case VNIC_MAC_ADDR_TYPE_FACTORY: 259 err = mac_addr_factory_reserve(vnic->vn_mch, addr_slot); 260 if (err != 0) { 261 if (err == EINVAL) 262 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTINVALID; 263 if (err == EBUSY) 264 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTUSED; 265 if (err == ENOSPC) 266 *diag = VNIC_IOC_DIAG_MACFACTORYSLOTALLUSED; 267 return (err); 268 } 269 270 mac_addr_factory_value(vnic->vn_lower_mh, *addr_slot, 271 mac_addr_arg, &addr_len, NULL, NULL); 272 *addr_len_ptr_arg = addr_len; 273 break; 274 275 case VNIC_MAC_ADDR_TYPE_AUTO: 276 /* first try to allocate a factory MAC address */ 277 err = mac_addr_factory_reserve(vnic->vn_mch, addr_slot); 278 if (err == 0) { 279 mac_addr_factory_value(vnic->vn_lower_mh, *addr_slot, 280 mac_addr_arg, &addr_len, NULL, NULL); 281 vnic_addr_type = VNIC_MAC_ADDR_TYPE_FACTORY; 282 *addr_len_ptr_arg = addr_len; 283 break; 284 } 285 286 /* 287 * Allocating a factory MAC address failed, generate a 288 * random MAC address instead. 289 */ 290 if ((err = mac_addr_random(vnic->vn_mch, 291 prefix_len, mac_addr_arg, &mac_diag)) != 0) { 292 *diag = vnic_mac2vnic_diag(mac_diag); 293 return (err); 294 } 295 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh); 296 vnic_addr_type = VNIC_MAC_ADDR_TYPE_RANDOM; 297 break; 298 case VNIC_MAC_ADDR_TYPE_PRIMARY: 299 /* 300 * We get the address here since we copy it in the 301 * vnic's vn_addr. 302 * We can't ask for hardware resources since we 303 * don't currently support hardware classification 304 * for these MAC clients. 305 */ 306 if (req_hwgrp_flag) { 307 *diag = VNIC_IOC_DIAG_NO_HWRINGS; 308 return (ENOTSUP); 309 } 310 mac_unicast_primary_get(vnic->vn_lower_mh, mac_addr_arg); 311 *addr_len_ptr_arg = mac_addr_len(vnic->vn_lower_mh); 312 mac_flags |= MAC_UNICAST_VNIC_PRIMARY; 313 break; 314 } 315 316 vnic->vn_addr_type = vnic_addr_type; 317 318 err = mac_unicast_add(vnic->vn_mch, mac_addr_arg, mac_flags, 319 &vnic->vn_muh, vid, &mac_diag); 320 if (err != 0) { 321 if (vnic_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) { 322 /* release factory MAC address */ 323 mac_addr_factory_release(vnic->vn_mch, *addr_slot); 324 } 325 *diag = vnic_mac2vnic_diag(mac_diag); 326 } 327 328 return (err); 329 } 330 331 /* 332 * Create a new VNIC upon request from administrator. 333 * Returns 0 on success, an errno on failure. 334 */ 335 /* ARGSUSED */ 336 int 337 vnic_dev_create(datalink_id_t vnic_id, datalink_id_t linkid, 338 vnic_mac_addr_type_t *vnic_addr_type, int *mac_len, uchar_t *mac_addr, 339 int *mac_slot, uint_t mac_prefix_len, uint16_t vid, vrid_t vrid, 340 int af, mac_resource_props_t *mrp, uint32_t flags, vnic_ioc_diag_t *diag, 341 cred_t *credp) 342 { 343 vnic_t *vnic; 344 mac_register_t *mac; 345 int err; 346 boolean_t is_anchor = ((flags & VNIC_IOC_CREATE_ANCHOR) != 0); 347 char vnic_name[MAXNAMELEN]; 348 const mac_info_t *minfop; 349 uint32_t req_hwgrp_flag = B_FALSE; 350 351 *diag = VNIC_IOC_DIAG_NONE; 352 353 rw_enter(&vnic_lock, RW_WRITER); 354 355 /* does a VNIC with the same id already exist? */ 356 err = mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id), 357 (mod_hash_val_t *)&vnic); 358 if (err == 0) { 359 rw_exit(&vnic_lock); 360 return (EEXIST); 361 } 362 363 vnic = kmem_cache_alloc(vnic_cache, KM_NOSLEEP); 364 if (vnic == NULL) { 365 rw_exit(&vnic_lock); 366 return (ENOMEM); 367 } 368 369 bzero(vnic, sizeof (*vnic)); 370 371 vnic->vn_id = vnic_id; 372 vnic->vn_link_id = linkid; 373 vnic->vn_vrid = vrid; 374 vnic->vn_af = af; 375 376 if (!is_anchor) { 377 if (linkid == DATALINK_INVALID_LINKID) { 378 err = EINVAL; 379 goto bail; 380 } 381 382 /* 383 * Open the lower MAC and assign its initial bandwidth and 384 * MAC address. We do this here during VNIC creation and 385 * do not wait until the upper MAC client open so that we 386 * can validate the VNIC creation parameters (bandwidth, 387 * MAC address, etc) and reserve a factory MAC address if 388 * one was requested. 389 */ 390 err = mac_open_by_linkid(linkid, &vnic->vn_lower_mh); 391 if (err != 0) 392 goto bail; 393 394 /* 395 * VNIC(vlan) over VNICs(vlans) is not supported. 396 */ 397 if (mac_is_vnic(vnic->vn_lower_mh)) { 398 err = EINVAL; 399 goto bail; 400 } 401 402 /* only ethernet support for now */ 403 minfop = mac_info(vnic->vn_lower_mh); 404 if (minfop->mi_nativemedia != DL_ETHER) { 405 err = ENOTSUP; 406 goto bail; 407 } 408 409 (void) dls_mgmt_get_linkinfo(vnic_id, vnic_name, NULL, NULL, 410 NULL); 411 err = mac_client_open(vnic->vn_lower_mh, &vnic->vn_mch, 412 vnic_name, MAC_OPEN_FLAGS_IS_VNIC); 413 if (err != 0) 414 goto bail; 415 416 /* assign a MAC address to the VNIC */ 417 418 err = vnic_unicast_add(vnic, *vnic_addr_type, mac_slot, 419 mac_prefix_len, mac_len, mac_addr, flags, diag, vid, 420 req_hwgrp_flag); 421 if (err != 0) { 422 vnic->vn_muh = NULL; 423 if (diag != NULL && req_hwgrp_flag) 424 *diag = VNIC_IOC_DIAG_NO_HWRINGS; 425 goto bail; 426 } 427 428 /* register to receive notification from underlying MAC */ 429 vnic->vn_mnh = mac_notify_add(vnic->vn_lower_mh, vnic_notify_cb, 430 vnic); 431 432 *vnic_addr_type = vnic->vn_addr_type; 433 vnic->vn_addr_len = *mac_len; 434 vnic->vn_vid = vid; 435 436 bcopy(mac_addr, vnic->vn_addr, vnic->vn_addr_len); 437 438 if (vnic->vn_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) 439 vnic->vn_slot_id = *mac_slot; 440 441 /* 442 * Set the initial VNIC capabilities. If the VNIC is created 443 * over MACs which does not support nactive vlan, disable 444 * VNIC's hardware checksum capability if its VID is not 0, 445 * since the underlying MAC would get the hardware checksum 446 * offset wrong in case of VLAN packets. 447 */ 448 if (vid == 0 || !mac_capab_get(vnic->vn_lower_mh, 449 MAC_CAPAB_NO_NATIVEVLAN, NULL)) { 450 if (!mac_capab_get(vnic->vn_lower_mh, MAC_CAPAB_HCKSUM, 451 &vnic->vn_hcksum_txflags)) 452 vnic->vn_hcksum_txflags = 0; 453 } else { 454 vnic->vn_hcksum_txflags = 0; 455 } 456 } 457 458 /* register with the MAC module */ 459 if ((mac = mac_alloc(MAC_VERSION)) == NULL) 460 goto bail; 461 462 mac->m_type_ident = MAC_PLUGIN_IDENT_ETHER; 463 mac->m_driver = vnic; 464 mac->m_dip = vnic_get_dip(); 465 mac->m_instance = (uint_t)-1; 466 mac->m_src_addr = vnic->vn_addr; 467 mac->m_callbacks = &vnic_m_callbacks; 468 469 if (!is_anchor) { 470 /* 471 * If this is a VNIC based VLAN, then we check for the 472 * margin unless it has been created with the force 473 * flag. If we are configuring a VLAN over an etherstub, 474 * we don't check the margin even if force is not set. 475 */ 476 if (vid == 0 || (flags & VNIC_IOC_CREATE_FORCE) != 0) { 477 if (vid != VLAN_ID_NONE) 478 vnic->vn_force = B_TRUE; 479 /* 480 * As the current margin size of the underlying mac is 481 * used to determine the margin size of the VNIC 482 * itself, request the underlying mac not to change 483 * to a smaller margin size. 484 */ 485 err = mac_margin_add(vnic->vn_lower_mh, 486 &vnic->vn_margin, B_TRUE); 487 ASSERT(err == 0); 488 } else { 489 vnic->vn_margin = VLAN_TAGSZ; 490 err = mac_margin_add(vnic->vn_lower_mh, 491 &vnic->vn_margin, B_FALSE); 492 if (err != 0) { 493 mac_free(mac); 494 if (diag != NULL) 495 *diag = VNIC_IOC_DIAG_MACMARGIN_INVALID; 496 goto bail; 497 } 498 } 499 500 mac_sdu_get(vnic->vn_lower_mh, &mac->m_min_sdu, 501 &mac->m_max_sdu); 502 } else { 503 vnic->vn_margin = VLAN_TAGSZ; 504 mac->m_min_sdu = 1; 505 mac->m_max_sdu = ANCHOR_VNIC_MAX_MTU; 506 } 507 508 mac->m_margin = vnic->vn_margin; 509 510 err = mac_register(mac, &vnic->vn_mh); 511 mac_free(mac); 512 if (err != 0) { 513 VERIFY(is_anchor || mac_margin_remove(vnic->vn_lower_mh, 514 vnic->vn_margin) == 0); 515 goto bail; 516 } 517 518 /* Set the VNIC's MAC in the client */ 519 if (!is_anchor) { 520 mac_set_upper_mac(vnic->vn_mch, vnic->vn_mh, mrp); 521 522 if (mrp != NULL) { 523 if ((mrp->mrp_mask & MRP_RX_RINGS) != 0 || 524 (mrp->mrp_mask & MRP_TX_RINGS) != 0) { 525 req_hwgrp_flag = B_TRUE; 526 } 527 err = mac_client_set_resources(vnic->vn_mch, mrp); 528 if (err != 0) { 529 (void) mac_unregister(vnic->vn_mh); 530 goto bail; 531 } 532 } 533 } 534 535 err = dls_devnet_create(vnic->vn_mh, vnic->vn_id, crgetzoneid(credp)); 536 if (err != 0) { 537 VERIFY(is_anchor || mac_margin_remove(vnic->vn_lower_mh, 538 vnic->vn_margin) == 0); 539 (void) mac_unregister(vnic->vn_mh); 540 goto bail; 541 } 542 543 /* add new VNIC to hash table */ 544 err = mod_hash_insert(vnic_hash, VNIC_HASH_KEY(vnic_id), 545 (mod_hash_val_t)vnic); 546 ASSERT(err == 0); 547 vnic_count++; 548 549 vnic->vn_enabled = B_TRUE; 550 rw_exit(&vnic_lock); 551 552 return (0); 553 554 bail: 555 rw_exit(&vnic_lock); 556 if (!is_anchor) { 557 if (vnic->vn_mnh != NULL) 558 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE); 559 if (vnic->vn_muh != NULL) 560 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh); 561 if (vnic->vn_mch != NULL) 562 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC); 563 if (vnic->vn_lower_mh != NULL) 564 mac_close(vnic->vn_lower_mh); 565 } 566 567 kmem_cache_free(vnic_cache, vnic); 568 return (err); 569 } 570 571 /* 572 * Modify the properties of an existing VNIC. 573 */ 574 /* ARGSUSED */ 575 int 576 vnic_dev_modify(datalink_id_t vnic_id, uint_t modify_mask, 577 vnic_mac_addr_type_t mac_addr_type, uint_t mac_len, uchar_t *mac_addr, 578 uint_t mac_slot, mac_resource_props_t *mrp) 579 { 580 vnic_t *vnic = NULL; 581 582 rw_enter(&vnic_lock, RW_WRITER); 583 584 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id), 585 (mod_hash_val_t *)&vnic) != 0) { 586 rw_exit(&vnic_lock); 587 return (ENOENT); 588 } 589 590 rw_exit(&vnic_lock); 591 592 return (0); 593 } 594 595 /* ARGSUSED */ 596 int 597 vnic_dev_delete(datalink_id_t vnic_id, uint32_t flags, cred_t *credp) 598 { 599 vnic_t *vnic = NULL; 600 mod_hash_val_t val; 601 datalink_id_t tmpid; 602 int rc; 603 604 rw_enter(&vnic_lock, RW_WRITER); 605 606 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id), 607 (mod_hash_val_t *)&vnic) != 0) { 608 rw_exit(&vnic_lock); 609 return (ENOENT); 610 } 611 612 if ((rc = dls_devnet_destroy(vnic->vn_mh, &tmpid, B_TRUE)) != 0) { 613 rw_exit(&vnic_lock); 614 return (rc); 615 } 616 617 ASSERT(vnic_id == tmpid); 618 619 /* 620 * We cannot unregister the MAC yet. Unregistering would 621 * free up mac_impl_t which should not happen at this time. 622 * So disable mac_impl_t by calling mac_disable(). This will prevent 623 * any new claims on mac_impl_t. 624 */ 625 if ((rc = mac_disable(vnic->vn_mh)) != 0) { 626 (void) dls_devnet_create(vnic->vn_mh, vnic_id, 627 crgetzoneid(credp)); 628 rw_exit(&vnic_lock); 629 return (rc); 630 } 631 632 vnic_cleanup_secondary_macs(vnic, vnic->vn_nhandles); 633 634 vnic->vn_enabled = B_FALSE; 635 (void) mod_hash_remove(vnic_hash, VNIC_HASH_KEY(vnic_id), &val); 636 ASSERT(vnic == (vnic_t *)val); 637 vnic_count--; 638 rw_exit(&vnic_lock); 639 640 /* 641 * XXX-nicolas shouldn't have a void cast here, if it's 642 * expected that the function will never fail, then we should 643 * have an ASSERT(). 644 */ 645 (void) mac_unregister(vnic->vn_mh); 646 647 if (vnic->vn_lower_mh != NULL) { 648 /* 649 * Check if MAC address for the vnic was obtained from the 650 * factory MAC addresses. If yes, release it. 651 */ 652 if (vnic->vn_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) { 653 (void) mac_addr_factory_release(vnic->vn_mch, 654 vnic->vn_slot_id); 655 } 656 (void) mac_margin_remove(vnic->vn_lower_mh, vnic->vn_margin); 657 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE); 658 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh); 659 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC); 660 mac_close(vnic->vn_lower_mh); 661 } 662 663 kmem_cache_free(vnic_cache, vnic); 664 return (0); 665 } 666 667 /* ARGSUSED */ 668 mblk_t * 669 vnic_m_tx(void *arg, mblk_t *mp_chain) 670 { 671 /* 672 * This function could be invoked for an anchor VNIC when sending 673 * broadcast and multicast packets, and unicast packets which did 674 * not match any local known destination. 675 */ 676 freemsgchain(mp_chain); 677 return (NULL); 678 } 679 680 /*ARGSUSED*/ 681 static void 682 vnic_m_ioctl(void *arg, queue_t *q, mblk_t *mp) 683 { 684 miocnak(q, mp, 0, ENOTSUP); 685 } 686 687 /* 688 * This entry point cannot be passed-through, since it is invoked 689 * for the per-VNIC kstats which must be exported independently 690 * of the existence of VNIC MAC clients. 691 */ 692 static int 693 vnic_m_stat(void *arg, uint_t stat, uint64_t *val) 694 { 695 vnic_t *vnic = arg; 696 int rval = 0; 697 698 if (vnic->vn_lower_mh == NULL) { 699 /* 700 * It's an anchor VNIC, which does not have any 701 * statistics in itself. 702 */ 703 return (ENOTSUP); 704 } 705 706 /* 707 * ENOTSUP must be reported for unsupported stats, the VNIC 708 * driver reports a subset of the stats that would 709 * be returned by a real piece of hardware. 710 */ 711 712 switch (stat) { 713 case MAC_STAT_LINK_STATE: 714 case MAC_STAT_LINK_UP: 715 case MAC_STAT_PROMISC: 716 case MAC_STAT_IFSPEED: 717 case MAC_STAT_MULTIRCV: 718 case MAC_STAT_MULTIXMT: 719 case MAC_STAT_BRDCSTRCV: 720 case MAC_STAT_BRDCSTXMT: 721 case MAC_STAT_OPACKETS: 722 case MAC_STAT_OBYTES: 723 case MAC_STAT_IERRORS: 724 case MAC_STAT_OERRORS: 725 case MAC_STAT_RBYTES: 726 case MAC_STAT_IPACKETS: 727 *val = mac_client_stat_get(vnic->vn_mch, stat); 728 break; 729 default: 730 rval = ENOTSUP; 731 } 732 733 return (rval); 734 } 735 736 /* 737 * Invoked by the upper MAC to retrieve the lower MAC client handle 738 * corresponding to a VNIC. A pointer to this function is obtained 739 * by the upper MAC via capability query. 740 * 741 * XXX-nicolas Note: this currently causes all VNIC MAC clients to 742 * receive the same MAC client handle for the same VNIC. This is ok 743 * as long as we have only one VNIC MAC client which sends and 744 * receives data, but we don't currently enforce this at the MAC layer. 745 */ 746 static void * 747 vnic_mac_client_handle(void *vnic_arg) 748 { 749 vnic_t *vnic = vnic_arg; 750 751 return (vnic->vn_mch); 752 } 753 754 /* 755 * Invoked when updating the primary MAC so that the secondary MACs are 756 * kept in sync. 757 */ 758 static void 759 vnic_mac_secondary_update(void *vnic_arg) 760 { 761 vnic_t *vn = vnic_arg; 762 int i; 763 764 for (i = 1; i <= vn->vn_nhandles; i++) { 765 mac_secondary_dup(vn->vn_mc_handles[0], vn->vn_mc_handles[i]); 766 } 767 } 768 769 /* 770 * Return information about the specified capability. 771 */ 772 /* ARGSUSED */ 773 static boolean_t 774 vnic_m_capab_get(void *arg, mac_capab_t cap, void *cap_data) 775 { 776 vnic_t *vnic = arg; 777 778 switch (cap) { 779 case MAC_CAPAB_HCKSUM: { 780 uint32_t *hcksum_txflags = cap_data; 781 782 *hcksum_txflags = vnic->vn_hcksum_txflags & 783 (HCKSUM_INET_FULL_V4 | HCKSUM_IPHDRCKSUM | 784 HCKSUM_INET_PARTIAL); 785 break; 786 } 787 case MAC_CAPAB_VNIC: { 788 mac_capab_vnic_t *vnic_capab = cap_data; 789 790 if (vnic->vn_lower_mh == NULL) { 791 /* 792 * It's an anchor VNIC, we don't have an underlying 793 * NIC and MAC client handle. 794 */ 795 return (B_FALSE); 796 } 797 798 if (vnic_capab != NULL) { 799 vnic_capab->mcv_arg = vnic; 800 vnic_capab->mcv_mac_client_handle = 801 vnic_mac_client_handle; 802 vnic_capab->mcv_mac_secondary_update = 803 vnic_mac_secondary_update; 804 } 805 break; 806 } 807 case MAC_CAPAB_ANCHOR_VNIC: { 808 /* since it's an anchor VNIC we don't have lower mac handle */ 809 if (vnic->vn_lower_mh == NULL) { 810 ASSERT(vnic->vn_link_id == 0); 811 return (B_TRUE); 812 } 813 return (B_FALSE); 814 } 815 case MAC_CAPAB_NO_NATIVEVLAN: 816 return (B_FALSE); 817 case MAC_CAPAB_NO_ZCOPY: 818 return (B_TRUE); 819 case MAC_CAPAB_VRRP: { 820 mac_capab_vrrp_t *vrrp_capab = cap_data; 821 822 if (vnic->vn_vrid != 0) { 823 if (vrrp_capab != NULL) 824 vrrp_capab->mcv_af = vnic->vn_af; 825 return (B_TRUE); 826 } 827 return (B_FALSE); 828 } 829 default: 830 return (B_FALSE); 831 } 832 return (B_TRUE); 833 } 834 835 /* ARGSUSED */ 836 static int 837 vnic_m_start(void *arg) 838 { 839 return (0); 840 } 841 842 /* ARGSUSED */ 843 static void 844 vnic_m_stop(void *arg) 845 { 846 } 847 848 /* ARGSUSED */ 849 static int 850 vnic_m_promisc(void *arg, boolean_t on) 851 { 852 return (0); 853 } 854 855 /* ARGSUSED */ 856 static int 857 vnic_m_multicst(void *arg, boolean_t add, const uint8_t *addrp) 858 { 859 return (0); 860 } 861 862 static int 863 vnic_m_unicst(void *arg, const uint8_t *macaddr) 864 { 865 vnic_t *vnic = arg; 866 867 return (mac_vnic_unicast_set(vnic->vn_mch, macaddr)); 868 } 869 870 static void 871 vnic_cleanup_secondary_macs(vnic_t *vn, int cnt) 872 { 873 int i; 874 875 /* Remove existing secondaries (primary is at 0) */ 876 for (i = 1; i <= cnt; i++) { 877 mac_rx_clear(vn->vn_mc_handles[i]); 878 879 /* unicast handle might not have been set yet */ 880 if (vn->vn_mu_handles[i] != NULL) 881 (void) mac_unicast_remove(vn->vn_mc_handles[i], 882 vn->vn_mu_handles[i]); 883 884 mac_secondary_cleanup(vn->vn_mc_handles[i]); 885 886 mac_client_close(vn->vn_mc_handles[i], MAC_CLOSE_FLAGS_IS_VNIC); 887 888 vn->vn_mu_handles[i] = NULL; 889 vn->vn_mc_handles[i] = NULL; 890 } 891 892 vn->vn_nhandles = 0; 893 } 894 895 /* 896 * Setup secondary MAC addresses on the vnic. Due to limitations in the mac 897 * code, each mac address must be associated with a mac_client (and the 898 * flow that goes along with the client) so we need to create those clients 899 * here. 900 */ 901 static int 902 vnic_set_secondary_macs(vnic_t *vn, mac_secondary_addr_t *msa) 903 { 904 int i, err; 905 char primary_name[MAXNAMELEN]; 906 907 /* First, remove pre-existing secondaries */ 908 ASSERT(vn->vn_nhandles < MPT_MAXMACADDR); 909 vnic_cleanup_secondary_macs(vn, vn->vn_nhandles); 910 911 if (msa->ms_addrcnt == (uint32_t)-1) 912 msa->ms_addrcnt = 0; 913 914 vn->vn_nhandles = msa->ms_addrcnt; 915 916 (void) dls_mgmt_get_linkinfo(vn->vn_id, primary_name, NULL, NULL, NULL); 917 918 /* 919 * Now add the new secondary MACs 920 * Recall that the primary MAC address is the first element. 921 * The secondary clients are named after the primary with their 922 * index to distinguish them. 923 */ 924 for (i = 1; i <= vn->vn_nhandles; i++) { 925 uint8_t *addr; 926 mac_diag_t mac_diag; 927 char secondary_name[MAXNAMELEN]; 928 929 (void) snprintf(secondary_name, sizeof (secondary_name), 930 "%s%02d", primary_name, i); 931 932 err = mac_client_open(vn->vn_lower_mh, &vn->vn_mc_handles[i], 933 secondary_name, MAC_OPEN_FLAGS_IS_VNIC); 934 if (err != 0) { 935 /* Remove any that we successfully added */ 936 vnic_cleanup_secondary_macs(vn, --i); 937 return (err); 938 } 939 940 /* 941 * Assign a MAC address to the VNIC 942 * 943 * Normally this would be done with vnic_unicast_add but since 944 * we know these are fixed adddresses, and since we need to 945 * save this in the proper array slot, we bypass that function 946 * and go direct. 947 */ 948 addr = msa->ms_addrs[i - 1]; 949 err = mac_unicast_add(vn->vn_mc_handles[i], addr, 0, 950 &vn->vn_mu_handles[i], vn->vn_vid, &mac_diag); 951 if (err != 0) { 952 /* Remove any that we successfully added */ 953 vnic_cleanup_secondary_macs(vn, i); 954 return (err); 955 } 956 957 /* 958 * Setup the secondary the same way as the primary (i.e. 959 * receiver function/argument (e.g. i_dls_link_rx, mac_pkt_drop, 960 * etc.), the promisc list, and the resource controls). 961 */ 962 mac_secondary_dup(vn->vn_mc_handles[0], vn->vn_mc_handles[i]); 963 } 964 965 return (0); 966 } 967 968 static int 969 vnic_get_secondary_macs(vnic_t *vn, uint_t pr_valsize, void *pr_val) 970 { 971 int i; 972 mac_secondary_addr_t msa; 973 974 if (pr_valsize < sizeof (msa)) 975 return (EINVAL); 976 977 /* Get existing addresses (primary is at 0) */ 978 ASSERT(vn->vn_nhandles < MPT_MAXMACADDR); 979 for (i = 1; i <= vn->vn_nhandles; i++) { 980 ASSERT(vn->vn_mc_handles[i] != NULL); 981 mac_unicast_secondary_get(vn->vn_mc_handles[i], 982 msa.ms_addrs[i - 1]); 983 } 984 msa.ms_addrcnt = vn->vn_nhandles; 985 986 bcopy(&msa, pr_val, sizeof (msa)); 987 return (0); 988 } 989 990 /* 991 * Callback functions for set/get of properties 992 */ 993 /*ARGSUSED*/ 994 static int 995 vnic_m_setprop(void *m_driver, const char *pr_name, mac_prop_id_t pr_num, 996 uint_t pr_valsize, const void *pr_val) 997 { 998 int err = 0; 999 vnic_t *vn = m_driver; 1000 1001 switch (pr_num) { 1002 case MAC_PROP_MTU: { 1003 uint32_t mtu; 1004 1005 /* allow setting MTU only on an etherstub */ 1006 if (vn->vn_link_id != DATALINK_INVALID_LINKID) 1007 return (err); 1008 1009 if (pr_valsize < sizeof (mtu)) { 1010 err = EINVAL; 1011 break; 1012 } 1013 bcopy(pr_val, &mtu, sizeof (mtu)); 1014 if (mtu < ANCHOR_VNIC_MIN_MTU || mtu > ANCHOR_VNIC_MAX_MTU) { 1015 err = EINVAL; 1016 break; 1017 } 1018 err = mac_maxsdu_update(vn->vn_mh, mtu); 1019 break; 1020 } 1021 case MAC_PROP_SECONDARY_ADDRS: { 1022 mac_secondary_addr_t msa; 1023 1024 bcopy(pr_val, &msa, sizeof (msa)); 1025 err = vnic_set_secondary_macs(vn, &msa); 1026 break; 1027 } 1028 default: 1029 err = ENOTSUP; 1030 break; 1031 } 1032 return (err); 1033 } 1034 1035 /* ARGSUSED */ 1036 static int 1037 vnic_m_getprop(void *arg, const char *pr_name, mac_prop_id_t pr_num, 1038 uint_t pr_valsize, void *pr_val) 1039 { 1040 vnic_t *vn = arg; 1041 int ret = 0; 1042 1043 switch (pr_num) { 1044 case MAC_PROP_SECONDARY_ADDRS: 1045 ret = vnic_get_secondary_macs(vn, pr_valsize, pr_val); 1046 break; 1047 default: 1048 ret = EINVAL; 1049 break; 1050 } 1051 1052 return (ret); 1053 } 1054 1055 /* ARGSUSED */ 1056 static void vnic_m_propinfo(void *m_driver, const char *pr_name, 1057 mac_prop_id_t pr_num, mac_prop_info_handle_t prh) 1058 { 1059 vnic_t *vn = m_driver; 1060 1061 /* MTU setting allowed only on an etherstub */ 1062 if (vn->vn_link_id != DATALINK_INVALID_LINKID) 1063 return; 1064 1065 switch (pr_num) { 1066 case MAC_PROP_MTU: 1067 mac_prop_info_set_range_uint32(prh, 1068 ANCHOR_VNIC_MIN_MTU, ANCHOR_VNIC_MAX_MTU); 1069 break; 1070 } 1071 } 1072 1073 1074 int 1075 vnic_info(vnic_info_t *info, cred_t *credp) 1076 { 1077 vnic_t *vnic; 1078 int err; 1079 1080 /* Make sure that the VNIC link is visible from the caller's zone. */ 1081 if (!dls_devnet_islinkvisible(info->vn_vnic_id, crgetzoneid(credp))) 1082 return (ENOENT); 1083 1084 rw_enter(&vnic_lock, RW_WRITER); 1085 1086 err = mod_hash_find(vnic_hash, VNIC_HASH_KEY(info->vn_vnic_id), 1087 (mod_hash_val_t *)&vnic); 1088 if (err != 0) { 1089 rw_exit(&vnic_lock); 1090 return (ENOENT); 1091 } 1092 1093 info->vn_link_id = vnic->vn_link_id; 1094 info->vn_mac_addr_type = vnic->vn_addr_type; 1095 info->vn_mac_len = vnic->vn_addr_len; 1096 bcopy(vnic->vn_addr, info->vn_mac_addr, MAXMACADDRLEN); 1097 info->vn_mac_slot = vnic->vn_slot_id; 1098 info->vn_mac_prefix_len = 0; 1099 info->vn_vid = vnic->vn_vid; 1100 info->vn_force = vnic->vn_force; 1101 info->vn_vrid = vnic->vn_vrid; 1102 info->vn_af = vnic->vn_af; 1103 1104 bzero(&info->vn_resource_props, sizeof (mac_resource_props_t)); 1105 if (vnic->vn_mch != NULL) 1106 mac_client_get_resources(vnic->vn_mch, 1107 &info->vn_resource_props); 1108 1109 rw_exit(&vnic_lock); 1110 return (0); 1111 } 1112 1113 static void 1114 vnic_notify_cb(void *arg, mac_notify_type_t type) 1115 { 1116 vnic_t *vnic = arg; 1117 1118 /* 1119 * Do not deliver notifications if the vnic is not fully initialized 1120 * or is in process of being torn down. 1121 */ 1122 if (!vnic->vn_enabled) 1123 return; 1124 1125 switch (type) { 1126 case MAC_NOTE_UNICST: 1127 /* 1128 * Only the VLAN VNIC needs to be notified with primary MAC 1129 * address change. 1130 */ 1131 if (vnic->vn_addr_type != VNIC_MAC_ADDR_TYPE_PRIMARY) 1132 return; 1133 1134 /* the unicast MAC address value */ 1135 mac_unicast_primary_get(vnic->vn_lower_mh, vnic->vn_addr); 1136 1137 /* notify its upper layer MAC about MAC address change */ 1138 mac_unicst_update(vnic->vn_mh, (const uint8_t *)vnic->vn_addr); 1139 break; 1140 1141 case MAC_NOTE_LINK: 1142 mac_link_update(vnic->vn_mh, 1143 mac_client_stat_get(vnic->vn_mch, MAC_STAT_LINK_STATE)); 1144 break; 1145 1146 default: 1147 break; 1148 } 1149 } 1150