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 err = mac_mtu_add(vnic->vn_lower_mh, &mac->m_max_sdu, B_FALSE); 503 if (err != 0) { 504 VERIFY(mac_margin_remove(vnic->vn_lower_mh, 505 vnic->vn_margin) == 0); 506 mac_free(mac); 507 if (diag != NULL) 508 *diag = VNIC_IOC_DIAG_MACMTU_INVALID; 509 goto bail; 510 } 511 vnic->vn_mtu = mac->m_max_sdu; 512 } else { 513 vnic->vn_margin = VLAN_TAGSZ; 514 mac->m_min_sdu = 1; 515 mac->m_max_sdu = ANCHOR_VNIC_MAX_MTU; 516 vnic->vn_mtu = ANCHOR_VNIC_MAX_MTU; 517 } 518 519 mac->m_margin = vnic->vn_margin; 520 521 err = mac_register(mac, &vnic->vn_mh); 522 mac_free(mac); 523 if (err != 0) { 524 if (!is_anchor) { 525 VERIFY(mac_mtu_remove(vnic->vn_lower_mh, 526 vnic->vn_mtu) == 0); 527 VERIFY(mac_margin_remove(vnic->vn_lower_mh, 528 vnic->vn_margin) == 0); 529 } 530 goto bail; 531 } 532 533 /* Set the VNIC's MAC in the client */ 534 if (!is_anchor) { 535 mac_set_upper_mac(vnic->vn_mch, vnic->vn_mh, mrp); 536 537 if (mrp != NULL) { 538 if ((mrp->mrp_mask & MRP_RX_RINGS) != 0 || 539 (mrp->mrp_mask & MRP_TX_RINGS) != 0) { 540 req_hwgrp_flag = B_TRUE; 541 } 542 err = mac_client_set_resources(vnic->vn_mch, mrp); 543 if (err != 0) { 544 VERIFY(mac_mtu_remove(vnic->vn_lower_mh, 545 vnic->vn_mtu) == 0); 546 VERIFY(mac_margin_remove(vnic->vn_lower_mh, 547 vnic->vn_margin) == 0); 548 (void) mac_unregister(vnic->vn_mh); 549 goto bail; 550 } 551 } 552 } 553 554 err = dls_devnet_create(vnic->vn_mh, vnic->vn_id, crgetzoneid(credp)); 555 if (err != 0) { 556 VERIFY(is_anchor || mac_margin_remove(vnic->vn_lower_mh, 557 vnic->vn_margin) == 0); 558 if (!is_anchor) { 559 VERIFY(mac_mtu_remove(vnic->vn_lower_mh, 560 vnic->vn_mtu) == 0); 561 VERIFY(mac_margin_remove(vnic->vn_lower_mh, 562 vnic->vn_margin) == 0); 563 } 564 (void) mac_unregister(vnic->vn_mh); 565 goto bail; 566 } 567 568 /* add new VNIC to hash table */ 569 err = mod_hash_insert(vnic_hash, VNIC_HASH_KEY(vnic_id), 570 (mod_hash_val_t)vnic); 571 ASSERT(err == 0); 572 vnic_count++; 573 574 vnic->vn_enabled = B_TRUE; 575 rw_exit(&vnic_lock); 576 577 return (0); 578 579 bail: 580 rw_exit(&vnic_lock); 581 if (!is_anchor) { 582 if (vnic->vn_mnh != NULL) 583 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE); 584 if (vnic->vn_muh != NULL) 585 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh); 586 if (vnic->vn_mch != NULL) 587 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC); 588 if (vnic->vn_lower_mh != NULL) 589 mac_close(vnic->vn_lower_mh); 590 } 591 592 kmem_cache_free(vnic_cache, vnic); 593 return (err); 594 } 595 596 /* 597 * Modify the properties of an existing VNIC. 598 */ 599 /* ARGSUSED */ 600 int 601 vnic_dev_modify(datalink_id_t vnic_id, uint_t modify_mask, 602 vnic_mac_addr_type_t mac_addr_type, uint_t mac_len, uchar_t *mac_addr, 603 uint_t mac_slot, mac_resource_props_t *mrp) 604 { 605 vnic_t *vnic = NULL; 606 607 rw_enter(&vnic_lock, RW_WRITER); 608 609 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id), 610 (mod_hash_val_t *)&vnic) != 0) { 611 rw_exit(&vnic_lock); 612 return (ENOENT); 613 } 614 615 rw_exit(&vnic_lock); 616 617 return (0); 618 } 619 620 /* ARGSUSED */ 621 int 622 vnic_dev_delete(datalink_id_t vnic_id, uint32_t flags, cred_t *credp) 623 { 624 vnic_t *vnic = NULL; 625 mod_hash_val_t val; 626 datalink_id_t tmpid; 627 int rc; 628 629 rw_enter(&vnic_lock, RW_WRITER); 630 631 if (mod_hash_find(vnic_hash, VNIC_HASH_KEY(vnic_id), 632 (mod_hash_val_t *)&vnic) != 0) { 633 rw_exit(&vnic_lock); 634 return (ENOENT); 635 } 636 637 if ((rc = dls_devnet_destroy(vnic->vn_mh, &tmpid, B_TRUE)) != 0) { 638 rw_exit(&vnic_lock); 639 return (rc); 640 } 641 642 ASSERT(vnic_id == tmpid); 643 644 /* 645 * We cannot unregister the MAC yet. Unregistering would 646 * free up mac_impl_t which should not happen at this time. 647 * So disable mac_impl_t by calling mac_disable(). This will prevent 648 * any new claims on mac_impl_t. 649 */ 650 if ((rc = mac_disable(vnic->vn_mh)) != 0) { 651 (void) dls_devnet_create(vnic->vn_mh, vnic_id, 652 crgetzoneid(credp)); 653 rw_exit(&vnic_lock); 654 return (rc); 655 } 656 657 vnic_cleanup_secondary_macs(vnic, vnic->vn_nhandles); 658 659 vnic->vn_enabled = B_FALSE; 660 (void) mod_hash_remove(vnic_hash, VNIC_HASH_KEY(vnic_id), &val); 661 ASSERT(vnic == (vnic_t *)val); 662 vnic_count--; 663 rw_exit(&vnic_lock); 664 665 /* 666 * XXX-nicolas shouldn't have a void cast here, if it's 667 * expected that the function will never fail, then we should 668 * have an ASSERT(). 669 */ 670 (void) mac_unregister(vnic->vn_mh); 671 672 if (vnic->vn_lower_mh != NULL) { 673 /* 674 * Check if MAC address for the vnic was obtained from the 675 * factory MAC addresses. If yes, release it. 676 */ 677 if (vnic->vn_addr_type == VNIC_MAC_ADDR_TYPE_FACTORY) { 678 (void) mac_addr_factory_release(vnic->vn_mch, 679 vnic->vn_slot_id); 680 } 681 (void) mac_margin_remove(vnic->vn_lower_mh, vnic->vn_margin); 682 (void) mac_mtu_remove(vnic->vn_lower_mh, vnic->vn_mtu); 683 (void) mac_notify_remove(vnic->vn_mnh, B_TRUE); 684 (void) mac_unicast_remove(vnic->vn_mch, vnic->vn_muh); 685 mac_client_close(vnic->vn_mch, MAC_CLOSE_FLAGS_IS_VNIC); 686 mac_close(vnic->vn_lower_mh); 687 } 688 689 kmem_cache_free(vnic_cache, vnic); 690 return (0); 691 } 692 693 /* ARGSUSED */ 694 mblk_t * 695 vnic_m_tx(void *arg, mblk_t *mp_chain) 696 { 697 /* 698 * This function could be invoked for an anchor VNIC when sending 699 * broadcast and multicast packets, and unicast packets which did 700 * not match any local known destination. 701 */ 702 freemsgchain(mp_chain); 703 return (NULL); 704 } 705 706 /*ARGSUSED*/ 707 static void 708 vnic_m_ioctl(void *arg, queue_t *q, mblk_t *mp) 709 { 710 miocnak(q, mp, 0, ENOTSUP); 711 } 712 713 /* 714 * This entry point cannot be passed-through, since it is invoked 715 * for the per-VNIC kstats which must be exported independently 716 * of the existence of VNIC MAC clients. 717 */ 718 static int 719 vnic_m_stat(void *arg, uint_t stat, uint64_t *val) 720 { 721 vnic_t *vnic = arg; 722 int rval = 0; 723 724 if (vnic->vn_lower_mh == NULL) { 725 /* 726 * It's an anchor VNIC, which does not have any 727 * statistics in itself. 728 */ 729 return (ENOTSUP); 730 } 731 732 /* 733 * ENOTSUP must be reported for unsupported stats, the VNIC 734 * driver reports a subset of the stats that would 735 * be returned by a real piece of hardware. 736 */ 737 738 switch (stat) { 739 case MAC_STAT_LINK_STATE: 740 case MAC_STAT_LINK_UP: 741 case MAC_STAT_PROMISC: 742 case MAC_STAT_IFSPEED: 743 case MAC_STAT_MULTIRCV: 744 case MAC_STAT_MULTIXMT: 745 case MAC_STAT_BRDCSTRCV: 746 case MAC_STAT_BRDCSTXMT: 747 case MAC_STAT_OPACKETS: 748 case MAC_STAT_OBYTES: 749 case MAC_STAT_IERRORS: 750 case MAC_STAT_OERRORS: 751 case MAC_STAT_RBYTES: 752 case MAC_STAT_IPACKETS: 753 *val = mac_client_stat_get(vnic->vn_mch, stat); 754 break; 755 default: 756 rval = ENOTSUP; 757 } 758 759 return (rval); 760 } 761 762 /* 763 * Invoked by the upper MAC to retrieve the lower MAC client handle 764 * corresponding to a VNIC. A pointer to this function is obtained 765 * by the upper MAC via capability query. 766 * 767 * XXX-nicolas Note: this currently causes all VNIC MAC clients to 768 * receive the same MAC client handle for the same VNIC. This is ok 769 * as long as we have only one VNIC MAC client which sends and 770 * receives data, but we don't currently enforce this at the MAC layer. 771 */ 772 static void * 773 vnic_mac_client_handle(void *vnic_arg) 774 { 775 vnic_t *vnic = vnic_arg; 776 777 return (vnic->vn_mch); 778 } 779 780 /* 781 * Invoked when updating the primary MAC so that the secondary MACs are 782 * kept in sync. 783 */ 784 static void 785 vnic_mac_secondary_update(void *vnic_arg) 786 { 787 vnic_t *vn = vnic_arg; 788 int i; 789 790 for (i = 1; i <= vn->vn_nhandles; i++) { 791 mac_secondary_dup(vn->vn_mc_handles[0], vn->vn_mc_handles[i]); 792 } 793 } 794 795 /* 796 * Return information about the specified capability. 797 */ 798 /* ARGSUSED */ 799 static boolean_t 800 vnic_m_capab_get(void *arg, mac_capab_t cap, void *cap_data) 801 { 802 vnic_t *vnic = arg; 803 804 switch (cap) { 805 case MAC_CAPAB_HCKSUM: { 806 uint32_t *hcksum_txflags = cap_data; 807 808 *hcksum_txflags = vnic->vn_hcksum_txflags & 809 (HCKSUM_INET_FULL_V4 | HCKSUM_IPHDRCKSUM | 810 HCKSUM_INET_PARTIAL); 811 break; 812 } 813 case MAC_CAPAB_VNIC: { 814 mac_capab_vnic_t *vnic_capab = cap_data; 815 816 if (vnic->vn_lower_mh == NULL) { 817 /* 818 * It's an anchor VNIC, we don't have an underlying 819 * NIC and MAC client handle. 820 */ 821 return (B_FALSE); 822 } 823 824 if (vnic_capab != NULL) { 825 vnic_capab->mcv_arg = vnic; 826 vnic_capab->mcv_mac_client_handle = 827 vnic_mac_client_handle; 828 vnic_capab->mcv_mac_secondary_update = 829 vnic_mac_secondary_update; 830 } 831 break; 832 } 833 case MAC_CAPAB_ANCHOR_VNIC: { 834 /* since it's an anchor VNIC we don't have lower mac handle */ 835 if (vnic->vn_lower_mh == NULL) { 836 ASSERT(vnic->vn_link_id == 0); 837 return (B_TRUE); 838 } 839 return (B_FALSE); 840 } 841 case MAC_CAPAB_NO_NATIVEVLAN: 842 return (B_FALSE); 843 case MAC_CAPAB_NO_ZCOPY: 844 return (B_TRUE); 845 case MAC_CAPAB_VRRP: { 846 mac_capab_vrrp_t *vrrp_capab = cap_data; 847 848 if (vnic->vn_vrid != 0) { 849 if (vrrp_capab != NULL) 850 vrrp_capab->mcv_af = vnic->vn_af; 851 return (B_TRUE); 852 } 853 return (B_FALSE); 854 } 855 default: 856 return (B_FALSE); 857 } 858 return (B_TRUE); 859 } 860 861 /* ARGSUSED */ 862 static int 863 vnic_m_start(void *arg) 864 { 865 return (0); 866 } 867 868 /* ARGSUSED */ 869 static void 870 vnic_m_stop(void *arg) 871 { 872 } 873 874 /* ARGSUSED */ 875 static int 876 vnic_m_promisc(void *arg, boolean_t on) 877 { 878 return (0); 879 } 880 881 /* ARGSUSED */ 882 static int 883 vnic_m_multicst(void *arg, boolean_t add, const uint8_t *addrp) 884 { 885 return (0); 886 } 887 888 static int 889 vnic_m_unicst(void *arg, const uint8_t *macaddr) 890 { 891 vnic_t *vnic = arg; 892 893 return (mac_vnic_unicast_set(vnic->vn_mch, macaddr)); 894 } 895 896 static void 897 vnic_cleanup_secondary_macs(vnic_t *vn, int cnt) 898 { 899 int i; 900 901 /* Remove existing secondaries (primary is at 0) */ 902 for (i = 1; i <= cnt; i++) { 903 mac_rx_clear(vn->vn_mc_handles[i]); 904 905 /* unicast handle might not have been set yet */ 906 if (vn->vn_mu_handles[i] != NULL) 907 (void) mac_unicast_remove(vn->vn_mc_handles[i], 908 vn->vn_mu_handles[i]); 909 910 mac_secondary_cleanup(vn->vn_mc_handles[i]); 911 912 mac_client_close(vn->vn_mc_handles[i], MAC_CLOSE_FLAGS_IS_VNIC); 913 914 vn->vn_mu_handles[i] = NULL; 915 vn->vn_mc_handles[i] = NULL; 916 } 917 918 vn->vn_nhandles = 0; 919 } 920 921 /* 922 * Setup secondary MAC addresses on the vnic. Due to limitations in the mac 923 * code, each mac address must be associated with a mac_client (and the 924 * flow that goes along with the client) so we need to create those clients 925 * here. 926 */ 927 static int 928 vnic_set_secondary_macs(vnic_t *vn, mac_secondary_addr_t *msa) 929 { 930 int i, err; 931 char primary_name[MAXNAMELEN]; 932 933 /* First, remove pre-existing secondaries */ 934 ASSERT(vn->vn_nhandles < MPT_MAXMACADDR); 935 vnic_cleanup_secondary_macs(vn, vn->vn_nhandles); 936 937 if (msa->ms_addrcnt == (uint32_t)-1) 938 msa->ms_addrcnt = 0; 939 940 vn->vn_nhandles = msa->ms_addrcnt; 941 942 (void) dls_mgmt_get_linkinfo(vn->vn_id, primary_name, NULL, NULL, NULL); 943 944 /* 945 * Now add the new secondary MACs 946 * Recall that the primary MAC address is the first element. 947 * The secondary clients are named after the primary with their 948 * index to distinguish them. 949 */ 950 for (i = 1; i <= vn->vn_nhandles; i++) { 951 uint8_t *addr; 952 mac_diag_t mac_diag; 953 char secondary_name[MAXNAMELEN]; 954 955 (void) snprintf(secondary_name, sizeof (secondary_name), 956 "%s%02d", primary_name, i); 957 958 err = mac_client_open(vn->vn_lower_mh, &vn->vn_mc_handles[i], 959 secondary_name, MAC_OPEN_FLAGS_IS_VNIC); 960 if (err != 0) { 961 /* Remove any that we successfully added */ 962 vnic_cleanup_secondary_macs(vn, --i); 963 return (err); 964 } 965 966 /* 967 * Assign a MAC address to the VNIC 968 * 969 * Normally this would be done with vnic_unicast_add but since 970 * we know these are fixed adddresses, and since we need to 971 * save this in the proper array slot, we bypass that function 972 * and go direct. 973 */ 974 addr = msa->ms_addrs[i - 1]; 975 err = mac_unicast_add(vn->vn_mc_handles[i], addr, 0, 976 &vn->vn_mu_handles[i], vn->vn_vid, &mac_diag); 977 if (err != 0) { 978 /* Remove any that we successfully added */ 979 vnic_cleanup_secondary_macs(vn, i); 980 return (err); 981 } 982 983 /* 984 * Setup the secondary the same way as the primary (i.e. 985 * receiver function/argument (e.g. i_dls_link_rx, mac_pkt_drop, 986 * etc.), the promisc list, and the resource controls). 987 */ 988 mac_secondary_dup(vn->vn_mc_handles[0], vn->vn_mc_handles[i]); 989 } 990 991 return (0); 992 } 993 994 static int 995 vnic_get_secondary_macs(vnic_t *vn, uint_t pr_valsize, void *pr_val) 996 { 997 int i; 998 mac_secondary_addr_t msa; 999 1000 if (pr_valsize < sizeof (msa)) 1001 return (EINVAL); 1002 1003 /* Get existing addresses (primary is at 0) */ 1004 ASSERT(vn->vn_nhandles < MPT_MAXMACADDR); 1005 for (i = 1; i <= vn->vn_nhandles; i++) { 1006 ASSERT(vn->vn_mc_handles[i] != NULL); 1007 mac_unicast_secondary_get(vn->vn_mc_handles[i], 1008 msa.ms_addrs[i - 1]); 1009 } 1010 msa.ms_addrcnt = vn->vn_nhandles; 1011 1012 bcopy(&msa, pr_val, sizeof (msa)); 1013 return (0); 1014 } 1015 1016 /* 1017 * Callback functions for set/get of properties 1018 */ 1019 /*ARGSUSED*/ 1020 static int 1021 vnic_m_setprop(void *m_driver, const char *pr_name, mac_prop_id_t pr_num, 1022 uint_t pr_valsize, const void *pr_val) 1023 { 1024 int err = 0; 1025 vnic_t *vn = m_driver; 1026 1027 switch (pr_num) { 1028 case MAC_PROP_MTU: { 1029 uint32_t mtu; 1030 1031 if (pr_valsize < sizeof (mtu)) { 1032 err = EINVAL; 1033 break; 1034 } 1035 bcopy(pr_val, &mtu, sizeof (mtu)); 1036 1037 if (vn->vn_link_id == DATALINK_INVALID_LINKID) { 1038 if (mtu < ANCHOR_VNIC_MIN_MTU || 1039 mtu > ANCHOR_VNIC_MAX_MTU) { 1040 err = EINVAL; 1041 break; 1042 } 1043 } else { 1044 err = mac_mtu_add(vn->vn_lower_mh, &mtu, B_FALSE); 1045 /* 1046 * If it's not supported to set a value here, translate 1047 * that to EINVAL, so user land gets a better idea of 1048 * what went wrong. This realistically means that they 1049 * violated the output of prop info. 1050 */ 1051 if (err == ENOTSUP) 1052 err = EINVAL; 1053 if (err != 0) 1054 break; 1055 VERIFY(mac_mtu_remove(vn->vn_lower_mh, 1056 vn->vn_mtu) == 0); 1057 } 1058 vn->vn_mtu = mtu; 1059 err = mac_maxsdu_update(vn->vn_mh, mtu); 1060 break; 1061 } 1062 case MAC_PROP_SECONDARY_ADDRS: { 1063 mac_secondary_addr_t msa; 1064 1065 bcopy(pr_val, &msa, sizeof (msa)); 1066 err = vnic_set_secondary_macs(vn, &msa); 1067 break; 1068 } 1069 default: 1070 err = ENOTSUP; 1071 break; 1072 } 1073 return (err); 1074 } 1075 1076 /* ARGSUSED */ 1077 static int 1078 vnic_m_getprop(void *arg, const char *pr_name, mac_prop_id_t pr_num, 1079 uint_t pr_valsize, void *pr_val) 1080 { 1081 vnic_t *vn = arg; 1082 int ret = 0; 1083 1084 switch (pr_num) { 1085 case MAC_PROP_SECONDARY_ADDRS: 1086 ret = vnic_get_secondary_macs(vn, pr_valsize, pr_val); 1087 break; 1088 default: 1089 ret = EINVAL; 1090 break; 1091 } 1092 1093 return (ret); 1094 } 1095 1096 /* ARGSUSED */ 1097 static void vnic_m_propinfo(void *m_driver, const char *pr_name, 1098 mac_prop_id_t pr_num, mac_prop_info_handle_t prh) 1099 { 1100 vnic_t *vn = m_driver; 1101 1102 switch (pr_num) { 1103 case MAC_PROP_MTU: 1104 if (vn->vn_link_id == DATALINK_INVALID_LINKID) { 1105 mac_prop_info_set_range_uint32(prh, 1106 ANCHOR_VNIC_MIN_MTU, ANCHOR_VNIC_MAX_MTU); 1107 } else { 1108 uint32_t max; 1109 mac_perim_handle_t mph; 1110 mac_propval_range_t range; 1111 1112 /* 1113 * The valid range for a VNIC's MTU is the minimum that 1114 * the device supports and the current value of the 1115 * device. A VNIC cannot increase the current MTU of the 1116 * device. Therefore we need to get the range from the 1117 * propinfo endpoint and current mtu from the 1118 * traditional property endpoint. 1119 */ 1120 mac_perim_enter_by_mh(vn->vn_lower_mh, &mph); 1121 if (mac_get_prop(vn->vn_lower_mh, MAC_PROP_MTU, "mtu", 1122 &max, sizeof (uint32_t)) != 0) { 1123 mac_perim_exit(mph); 1124 return; 1125 } 1126 1127 range.mpr_count = 1; 1128 if (mac_prop_info(vn->vn_lower_mh, MAC_PROP_MTU, "mtu", 1129 NULL, 0, &range, NULL) != 0) { 1130 mac_perim_exit(mph); 1131 return; 1132 } 1133 1134 mac_prop_info_set_default_uint32(prh, max); 1135 mac_prop_info_set_range_uint32(prh, 1136 range.mpr_range_uint32[0].mpur_min, max); 1137 mac_perim_exit(mph); 1138 } 1139 break; 1140 } 1141 } 1142 1143 1144 int 1145 vnic_info(vnic_info_t *info, cred_t *credp) 1146 { 1147 vnic_t *vnic; 1148 int err; 1149 1150 /* Make sure that the VNIC link is visible from the caller's zone. */ 1151 if (!dls_devnet_islinkvisible(info->vn_vnic_id, crgetzoneid(credp))) 1152 return (ENOENT); 1153 1154 rw_enter(&vnic_lock, RW_WRITER); 1155 1156 err = mod_hash_find(vnic_hash, VNIC_HASH_KEY(info->vn_vnic_id), 1157 (mod_hash_val_t *)&vnic); 1158 if (err != 0) { 1159 rw_exit(&vnic_lock); 1160 return (ENOENT); 1161 } 1162 1163 info->vn_link_id = vnic->vn_link_id; 1164 info->vn_mac_addr_type = vnic->vn_addr_type; 1165 info->vn_mac_len = vnic->vn_addr_len; 1166 bcopy(vnic->vn_addr, info->vn_mac_addr, MAXMACADDRLEN); 1167 info->vn_mac_slot = vnic->vn_slot_id; 1168 info->vn_mac_prefix_len = 0; 1169 info->vn_vid = vnic->vn_vid; 1170 info->vn_force = vnic->vn_force; 1171 info->vn_vrid = vnic->vn_vrid; 1172 info->vn_af = vnic->vn_af; 1173 1174 bzero(&info->vn_resource_props, sizeof (mac_resource_props_t)); 1175 if (vnic->vn_mch != NULL) 1176 mac_client_get_resources(vnic->vn_mch, 1177 &info->vn_resource_props); 1178 1179 rw_exit(&vnic_lock); 1180 return (0); 1181 } 1182 1183 static void 1184 vnic_notify_cb(void *arg, mac_notify_type_t type) 1185 { 1186 vnic_t *vnic = arg; 1187 1188 /* 1189 * Do not deliver notifications if the vnic is not fully initialized 1190 * or is in process of being torn down. 1191 */ 1192 if (!vnic->vn_enabled) 1193 return; 1194 1195 switch (type) { 1196 case MAC_NOTE_UNICST: 1197 /* 1198 * Only the VLAN VNIC needs to be notified with primary MAC 1199 * address change. 1200 */ 1201 if (vnic->vn_addr_type != VNIC_MAC_ADDR_TYPE_PRIMARY) 1202 return; 1203 1204 /* the unicast MAC address value */ 1205 mac_unicast_primary_get(vnic->vn_lower_mh, vnic->vn_addr); 1206 1207 /* notify its upper layer MAC about MAC address change */ 1208 mac_unicst_update(vnic->vn_mh, (const uint8_t *)vnic->vn_addr); 1209 break; 1210 1211 case MAC_NOTE_LINK: 1212 mac_link_update(vnic->vn_mh, 1213 mac_client_stat_get(vnic->vn_mch, MAC_STAT_LINK_STATE)); 1214 break; 1215 1216 default: 1217 break; 1218 } 1219 } 1220