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 2012, Nexenta Systems, Inc. All rights reserved. 24 */ 25 26 /* 27 * Data-Link Driver 28 */ 29 #include <sys/sysmacros.h> 30 #include <sys/strsubr.h> 31 #include <sys/strsun.h> 32 #include <sys/vlan.h> 33 #include <sys/dld_impl.h> 34 #include <sys/mac_client.h> 35 #include <sys/mac_client_impl.h> 36 #include <sys/mac_client_priv.h> 37 38 typedef void proto_reqfunc_t(dld_str_t *, mblk_t *); 39 40 static proto_reqfunc_t proto_info_req, proto_attach_req, proto_detach_req, 41 proto_bind_req, proto_unbind_req, proto_promiscon_req, proto_promiscoff_req, 42 proto_enabmulti_req, proto_disabmulti_req, proto_physaddr_req, 43 proto_setphysaddr_req, proto_udqos_req, proto_req, proto_capability_req, 44 proto_notify_req, proto_passive_req; 45 46 static void proto_capability_advertise(dld_str_t *, mblk_t *); 47 static int dld_capab_poll_disable(dld_str_t *, dld_capab_poll_t *); 48 static boolean_t check_mod_above(queue_t *, const char *); 49 50 #define DL_ACK_PENDING(state) \ 51 ((state) == DL_ATTACH_PENDING || \ 52 (state) == DL_DETACH_PENDING || \ 53 (state) == DL_BIND_PENDING || \ 54 (state) == DL_UNBIND_PENDING) 55 56 /* 57 * Process a DLPI protocol message. 58 * The primitives DL_BIND_REQ, DL_ENABMULTI_REQ, DL_PROMISCON_REQ, 59 * DL_SET_PHYS_ADDR_REQ put the data link below our dld_str_t into an 60 * 'active' state. The primitive DL_PASSIVE_REQ marks our dld_str_t 61 * as 'passive' and forbids it from being subsequently made 'active' 62 * by the above primitives. 63 */ 64 void 65 dld_proto(dld_str_t *dsp, mblk_t *mp) 66 { 67 t_uscalar_t prim; 68 69 if (MBLKL(mp) < sizeof (t_uscalar_t)) { 70 freemsg(mp); 71 return; 72 } 73 prim = ((union DL_primitives *)mp->b_rptr)->dl_primitive; 74 75 switch (prim) { 76 case DL_INFO_REQ: 77 proto_info_req(dsp, mp); 78 break; 79 case DL_BIND_REQ: 80 proto_bind_req(dsp, mp); 81 break; 82 case DL_UNBIND_REQ: 83 proto_unbind_req(dsp, mp); 84 break; 85 case DL_UNITDATA_REQ: 86 proto_unitdata_req(dsp, mp); 87 break; 88 case DL_UDQOS_REQ: 89 proto_udqos_req(dsp, mp); 90 break; 91 case DL_ATTACH_REQ: 92 proto_attach_req(dsp, mp); 93 break; 94 case DL_DETACH_REQ: 95 proto_detach_req(dsp, mp); 96 break; 97 case DL_ENABMULTI_REQ: 98 proto_enabmulti_req(dsp, mp); 99 break; 100 case DL_DISABMULTI_REQ: 101 proto_disabmulti_req(dsp, mp); 102 break; 103 case DL_PROMISCON_REQ: 104 proto_promiscon_req(dsp, mp); 105 break; 106 case DL_PROMISCOFF_REQ: 107 proto_promiscoff_req(dsp, mp); 108 break; 109 case DL_PHYS_ADDR_REQ: 110 proto_physaddr_req(dsp, mp); 111 break; 112 case DL_SET_PHYS_ADDR_REQ: 113 proto_setphysaddr_req(dsp, mp); 114 break; 115 case DL_NOTIFY_REQ: 116 proto_notify_req(dsp, mp); 117 break; 118 case DL_CAPABILITY_REQ: 119 proto_capability_req(dsp, mp); 120 break; 121 case DL_PASSIVE_REQ: 122 proto_passive_req(dsp, mp); 123 break; 124 default: 125 proto_req(dsp, mp); 126 break; 127 } 128 } 129 130 #define NEG(x) -(x) 131 typedef struct dl_info_ack_wrapper { 132 dl_info_ack_t dl_info; 133 uint8_t dl_addr[MAXMACADDRLEN + sizeof (uint16_t)]; 134 uint8_t dl_brdcst_addr[MAXMACADDRLEN]; 135 dl_qos_cl_range1_t dl_qos_range1; 136 dl_qos_cl_sel1_t dl_qos_sel1; 137 } dl_info_ack_wrapper_t; 138 139 /* 140 * DL_INFO_REQ 141 */ 142 static void 143 proto_info_req(dld_str_t *dsp, mblk_t *mp) 144 { 145 dl_info_ack_wrapper_t *dlwp; 146 dl_info_ack_t *dlp; 147 dl_qos_cl_sel1_t *selp; 148 dl_qos_cl_range1_t *rangep; 149 uint8_t *addr; 150 uint8_t *brdcst_addr; 151 uint_t addr_length; 152 uint_t sap_length; 153 mac_info_t minfo; 154 mac_info_t *minfop; 155 queue_t *q = dsp->ds_wq; 156 157 /* 158 * Swap the request message for one large enough to contain the 159 * wrapper structure defined above. 160 */ 161 if ((mp = mexchange(q, mp, sizeof (dl_info_ack_wrapper_t), 162 M_PCPROTO, 0)) == NULL) 163 return; 164 165 bzero(mp->b_rptr, sizeof (dl_info_ack_wrapper_t)); 166 dlwp = (dl_info_ack_wrapper_t *)mp->b_rptr; 167 168 dlp = &(dlwp->dl_info); 169 ASSERT(dlp == (dl_info_ack_t *)mp->b_rptr); 170 171 dlp->dl_primitive = DL_INFO_ACK; 172 173 /* 174 * Set up the sub-structure pointers. 175 */ 176 addr = dlwp->dl_addr; 177 brdcst_addr = dlwp->dl_brdcst_addr; 178 rangep = &(dlwp->dl_qos_range1); 179 selp = &(dlwp->dl_qos_sel1); 180 181 /* 182 * This driver supports only version 2 connectionless DLPI provider 183 * nodes. 184 */ 185 dlp->dl_service_mode = DL_CLDLS; 186 dlp->dl_version = DL_VERSION_2; 187 188 /* 189 * Set the style of the provider 190 */ 191 dlp->dl_provider_style = dsp->ds_style; 192 ASSERT(dlp->dl_provider_style == DL_STYLE1 || 193 dlp->dl_provider_style == DL_STYLE2); 194 195 /* 196 * Set the current DLPI state. 197 */ 198 dlp->dl_current_state = dsp->ds_dlstate; 199 200 /* 201 * Gratuitously set the media type. This is to deal with modules 202 * that assume the media type is known prior to DL_ATTACH_REQ 203 * being completed. 204 */ 205 dlp->dl_mac_type = DL_ETHER; 206 207 /* 208 * If the stream is not at least attached we try to retrieve the 209 * mac_info using mac_info_get() 210 */ 211 if (dsp->ds_dlstate == DL_UNATTACHED || 212 dsp->ds_dlstate == DL_ATTACH_PENDING || 213 dsp->ds_dlstate == DL_DETACH_PENDING) { 214 if (!mac_info_get(ddi_major_to_name(dsp->ds_major), &minfo)) { 215 /* 216 * Cannot find mac_info. giving up. 217 */ 218 goto done; 219 } 220 minfop = &minfo; 221 } else { 222 minfop = (mac_info_t *)dsp->ds_mip; 223 /* We can only get the sdu if we're attached. */ 224 mac_sdu_get(dsp->ds_mh, &dlp->dl_min_sdu, &dlp->dl_max_sdu); 225 } 226 227 /* 228 * Set the media type (properly this time). 229 */ 230 if (dsp->ds_native) 231 dlp->dl_mac_type = minfop->mi_nativemedia; 232 else 233 dlp->dl_mac_type = minfop->mi_media; 234 235 /* 236 * Set the DLSAP length. We only support 16 bit values and they 237 * appear after the MAC address portion of DLSAP addresses. 238 */ 239 sap_length = sizeof (uint16_t); 240 dlp->dl_sap_length = NEG(sap_length); 241 242 addr_length = minfop->mi_addr_length; 243 244 /* 245 * Copy in the media broadcast address. 246 */ 247 if (minfop->mi_brdcst_addr != NULL) { 248 dlp->dl_brdcst_addr_offset = 249 (uintptr_t)brdcst_addr - (uintptr_t)dlp; 250 bcopy(minfop->mi_brdcst_addr, brdcst_addr, addr_length); 251 dlp->dl_brdcst_addr_length = addr_length; 252 } 253 254 /* Only VLAN links and links that have a normal tag mode support QOS. */ 255 if ((dsp->ds_mch != NULL && 256 mac_client_vid(dsp->ds_mch) != VLAN_ID_NONE) || 257 (dsp->ds_dlp != NULL && 258 dsp->ds_dlp->dl_tagmode == LINK_TAGMODE_NORMAL)) { 259 dlp->dl_qos_range_offset = (uintptr_t)rangep - (uintptr_t)dlp; 260 dlp->dl_qos_range_length = sizeof (dl_qos_cl_range1_t); 261 262 rangep->dl_qos_type = DL_QOS_CL_RANGE1; 263 rangep->dl_trans_delay.dl_target_value = DL_UNKNOWN; 264 rangep->dl_trans_delay.dl_accept_value = DL_UNKNOWN; 265 rangep->dl_protection.dl_min = DL_UNKNOWN; 266 rangep->dl_protection.dl_max = DL_UNKNOWN; 267 rangep->dl_residual_error = DL_UNKNOWN; 268 269 /* 270 * Specify the supported range of priorities. 271 */ 272 rangep->dl_priority.dl_min = 0; 273 rangep->dl_priority.dl_max = (1 << VLAN_PRI_SIZE) - 1; 274 275 dlp->dl_qos_offset = (uintptr_t)selp - (uintptr_t)dlp; 276 dlp->dl_qos_length = sizeof (dl_qos_cl_sel1_t); 277 278 selp->dl_qos_type = DL_QOS_CL_SEL1; 279 selp->dl_trans_delay = DL_UNKNOWN; 280 selp->dl_protection = DL_UNKNOWN; 281 selp->dl_residual_error = DL_UNKNOWN; 282 283 /* 284 * Specify the current priority (which can be changed by 285 * the DL_UDQOS_REQ primitive). 286 */ 287 selp->dl_priority = dsp->ds_pri; 288 } 289 290 dlp->dl_addr_length = addr_length + sizeof (uint16_t); 291 if (dsp->ds_dlstate == DL_IDLE) { 292 /* 293 * The stream is bound. Therefore we can formulate a valid 294 * DLSAP address. 295 */ 296 dlp->dl_addr_offset = (uintptr_t)addr - (uintptr_t)dlp; 297 if (addr_length > 0) 298 mac_unicast_primary_get(dsp->ds_mh, addr); 299 300 *(uint16_t *)(addr + addr_length) = dsp->ds_sap; 301 } 302 303 done: 304 IMPLY(dlp->dl_qos_offset != 0, dlp->dl_qos_length != 0); 305 IMPLY(dlp->dl_qos_range_offset != 0, 306 dlp->dl_qos_range_length != 0); 307 IMPLY(dlp->dl_addr_offset != 0, dlp->dl_addr_length != 0); 308 IMPLY(dlp->dl_brdcst_addr_offset != 0, 309 dlp->dl_brdcst_addr_length != 0); 310 311 qreply(q, mp); 312 } 313 314 /* 315 * DL_ATTACH_REQ 316 */ 317 static void 318 proto_attach_req(dld_str_t *dsp, mblk_t *mp) 319 { 320 dl_attach_req_t *dlp = (dl_attach_req_t *)mp->b_rptr; 321 int err = 0; 322 t_uscalar_t dl_err; 323 queue_t *q = dsp->ds_wq; 324 325 if (MBLKL(mp) < sizeof (dl_attach_req_t) || 326 dlp->dl_ppa < 0 || dsp->ds_style == DL_STYLE1) { 327 dl_err = DL_BADPRIM; 328 goto failed; 329 } 330 331 if (dsp->ds_dlstate != DL_UNATTACHED) { 332 dl_err = DL_OUTSTATE; 333 goto failed; 334 } 335 336 dsp->ds_dlstate = DL_ATTACH_PENDING; 337 338 err = dld_str_attach(dsp, dlp->dl_ppa); 339 if (err != 0) { 340 switch (err) { 341 case ENOENT: 342 dl_err = DL_BADPPA; 343 err = 0; 344 break; 345 default: 346 dl_err = DL_SYSERR; 347 break; 348 } 349 dsp->ds_dlstate = DL_UNATTACHED; 350 goto failed; 351 } 352 ASSERT(dsp->ds_dlstate == DL_UNBOUND); 353 dlokack(q, mp, DL_ATTACH_REQ); 354 return; 355 356 failed: 357 dlerrorack(q, mp, DL_ATTACH_REQ, dl_err, (t_uscalar_t)err); 358 } 359 360 /* 361 * DL_DETACH_REQ 362 */ 363 static void 364 proto_detach_req(dld_str_t *dsp, mblk_t *mp) 365 { 366 queue_t *q = dsp->ds_wq; 367 t_uscalar_t dl_err; 368 369 if (MBLKL(mp) < sizeof (dl_detach_req_t)) { 370 dl_err = DL_BADPRIM; 371 goto failed; 372 } 373 374 if (dsp->ds_dlstate != DL_UNBOUND) { 375 dl_err = DL_OUTSTATE; 376 goto failed; 377 } 378 379 if (dsp->ds_style == DL_STYLE1) { 380 dl_err = DL_BADPRIM; 381 goto failed; 382 } 383 384 ASSERT(dsp->ds_datathr_cnt == 0); 385 dsp->ds_dlstate = DL_DETACH_PENDING; 386 387 dld_str_detach(dsp); 388 dlokack(dsp->ds_wq, mp, DL_DETACH_REQ); 389 return; 390 391 failed: 392 dlerrorack(q, mp, DL_DETACH_REQ, dl_err, 0); 393 } 394 395 /* 396 * DL_BIND_REQ 397 */ 398 static void 399 proto_bind_req(dld_str_t *dsp, mblk_t *mp) 400 { 401 dl_bind_req_t *dlp = (dl_bind_req_t *)mp->b_rptr; 402 int err = 0; 403 uint8_t dlsap_addr[MAXMACADDRLEN + sizeof (uint16_t)]; 404 uint_t dlsap_addr_length; 405 t_uscalar_t dl_err; 406 t_scalar_t sap; 407 queue_t *q = dsp->ds_wq; 408 mac_perim_handle_t mph; 409 void *mdip; 410 int32_t intr_cpu; 411 412 if (MBLKL(mp) < sizeof (dl_bind_req_t)) { 413 dl_err = DL_BADPRIM; 414 goto failed; 415 } 416 417 if (dlp->dl_xidtest_flg != 0) { 418 dl_err = DL_NOAUTO; 419 goto failed; 420 } 421 422 if (dlp->dl_service_mode != DL_CLDLS) { 423 dl_err = DL_UNSUPPORTED; 424 goto failed; 425 } 426 427 if (dsp->ds_dlstate != DL_UNBOUND) { 428 dl_err = DL_OUTSTATE; 429 goto failed; 430 } 431 432 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 433 434 if ((err = dls_active_set(dsp)) != 0) { 435 dl_err = DL_SYSERR; 436 goto failed2; 437 } 438 439 dsp->ds_dlstate = DL_BIND_PENDING; 440 /* 441 * Set the receive callback. 442 */ 443 dls_rx_set(dsp, (dsp->ds_mode == DLD_RAW) ? 444 dld_str_rx_raw : dld_str_rx_unitdata, dsp); 445 446 /* 447 * Bind the channel such that it can receive packets. 448 */ 449 sap = dlp->dl_sap; 450 dsp->ds_nonip = !check_mod_above(dsp->ds_rq, "ip") && 451 !check_mod_above(dsp->ds_rq, "arp"); 452 453 err = dls_bind(dsp, sap); 454 if (err != 0) { 455 switch (err) { 456 case EINVAL: 457 dl_err = DL_BADADDR; 458 err = 0; 459 break; 460 default: 461 dl_err = DL_SYSERR; 462 break; 463 } 464 465 dsp->ds_dlstate = DL_UNBOUND; 466 dls_active_clear(dsp, B_FALSE); 467 goto failed2; 468 } 469 470 intr_cpu = mac_client_intr_cpu(dsp->ds_mch); 471 mdip = mac_get_devinfo(dsp->ds_mh); 472 mac_perim_exit(mph); 473 474 /* 475 * We do this after we get out of the perim to avoid deadlocks 476 * etc. since part of mac_client_retarget_intr is to walk the 477 * device tree in order to find and retarget the interrupts. 478 */ 479 if (intr_cpu != -1) 480 mac_client_set_intr_cpu(mdip, dsp->ds_mch, intr_cpu); 481 482 /* 483 * Copy in MAC address. 484 */ 485 dlsap_addr_length = dsp->ds_mip->mi_addr_length; 486 mac_unicast_primary_get(dsp->ds_mh, dlsap_addr); 487 488 /* 489 * Copy in the SAP. 490 */ 491 *(uint16_t *)(dlsap_addr + dlsap_addr_length) = sap; 492 dlsap_addr_length += sizeof (uint16_t); 493 494 dsp->ds_dlstate = DL_IDLE; 495 dlbindack(q, mp, sap, dlsap_addr, dlsap_addr_length, 0, 0); 496 return; 497 498 failed2: 499 mac_perim_exit(mph); 500 failed: 501 dlerrorack(q, mp, DL_BIND_REQ, dl_err, (t_uscalar_t)err); 502 } 503 504 /* 505 * DL_UNBIND_REQ 506 */ 507 static void 508 proto_unbind_req(dld_str_t *dsp, mblk_t *mp) 509 { 510 queue_t *q = dsp->ds_wq; 511 t_uscalar_t dl_err; 512 mac_perim_handle_t mph; 513 514 if (MBLKL(mp) < sizeof (dl_unbind_req_t)) { 515 dl_err = DL_BADPRIM; 516 goto failed; 517 } 518 519 if (dsp->ds_dlstate != DL_IDLE) { 520 dl_err = DL_OUTSTATE; 521 goto failed; 522 } 523 524 mutex_enter(&dsp->ds_lock); 525 while (dsp->ds_datathr_cnt != 0) 526 cv_wait(&dsp->ds_datathr_cv, &dsp->ds_lock); 527 528 dsp->ds_dlstate = DL_UNBIND_PENDING; 529 mutex_exit(&dsp->ds_lock); 530 531 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 532 /* 533 * Unbind the channel to stop packets being received. 534 */ 535 dls_unbind(dsp); 536 537 /* 538 * Disable polling mode, if it is enabled. 539 */ 540 (void) dld_capab_poll_disable(dsp, NULL); 541 542 /* 543 * Clear LSO flags. 544 */ 545 dsp->ds_lso = B_FALSE; 546 dsp->ds_lso_max = 0; 547 548 /* 549 * Clear the receive callback. 550 */ 551 dls_rx_set(dsp, NULL, NULL); 552 dsp->ds_direct = B_FALSE; 553 554 /* 555 * Set the mode back to the default (unitdata). 556 */ 557 dsp->ds_mode = DLD_UNITDATA; 558 dsp->ds_dlstate = DL_UNBOUND; 559 560 dls_active_clear(dsp, B_FALSE); 561 mac_perim_exit(mph); 562 dlokack(dsp->ds_wq, mp, DL_UNBIND_REQ); 563 return; 564 failed: 565 dlerrorack(q, mp, DL_UNBIND_REQ, dl_err, 0); 566 } 567 568 /* 569 * DL_PROMISCON_REQ 570 */ 571 static void 572 proto_promiscon_req(dld_str_t *dsp, mblk_t *mp) 573 { 574 dl_promiscon_req_t *dlp = (dl_promiscon_req_t *)mp->b_rptr; 575 int err = 0; 576 t_uscalar_t dl_err; 577 uint32_t new_flags, promisc_saved; 578 queue_t *q = dsp->ds_wq; 579 mac_perim_handle_t mph; 580 581 if (MBLKL(mp) < sizeof (dl_promiscon_req_t)) { 582 dl_err = DL_BADPRIM; 583 goto failed; 584 } 585 586 if (dsp->ds_dlstate == DL_UNATTACHED || 587 DL_ACK_PENDING(dsp->ds_dlstate)) { 588 dl_err = DL_OUTSTATE; 589 goto failed; 590 } 591 592 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 593 594 new_flags = promisc_saved = dsp->ds_promisc; 595 switch (dlp->dl_level) { 596 case DL_PROMISC_SAP: 597 new_flags |= DLS_PROMISC_SAP; 598 break; 599 600 case DL_PROMISC_MULTI: 601 new_flags |= DLS_PROMISC_MULTI; 602 break; 603 604 case DL_PROMISC_PHYS: 605 new_flags |= DLS_PROMISC_PHYS; 606 break; 607 608 default: 609 dl_err = DL_NOTSUPPORTED; 610 goto failed2; 611 } 612 613 if ((promisc_saved == 0) && (err = dls_active_set(dsp)) != 0) { 614 ASSERT(dsp->ds_promisc == promisc_saved); 615 dl_err = DL_SYSERR; 616 goto failed2; 617 } 618 619 /* 620 * Adjust channel promiscuity. 621 */ 622 err = dls_promisc(dsp, new_flags); 623 624 if (err != 0) { 625 dl_err = DL_SYSERR; 626 dsp->ds_promisc = promisc_saved; 627 if (promisc_saved == 0) 628 dls_active_clear(dsp, B_FALSE); 629 goto failed2; 630 } 631 632 mac_perim_exit(mph); 633 634 dlokack(q, mp, DL_PROMISCON_REQ); 635 return; 636 637 failed2: 638 mac_perim_exit(mph); 639 failed: 640 dlerrorack(q, mp, DL_PROMISCON_REQ, dl_err, (t_uscalar_t)err); 641 } 642 643 /* 644 * DL_PROMISCOFF_REQ 645 */ 646 static void 647 proto_promiscoff_req(dld_str_t *dsp, mblk_t *mp) 648 { 649 dl_promiscoff_req_t *dlp = (dl_promiscoff_req_t *)mp->b_rptr; 650 int err = 0; 651 t_uscalar_t dl_err; 652 uint32_t new_flags; 653 queue_t *q = dsp->ds_wq; 654 mac_perim_handle_t mph; 655 656 if (MBLKL(mp) < sizeof (dl_promiscoff_req_t)) { 657 dl_err = DL_BADPRIM; 658 goto failed; 659 } 660 661 if (dsp->ds_dlstate == DL_UNATTACHED || 662 DL_ACK_PENDING(dsp->ds_dlstate)) { 663 dl_err = DL_OUTSTATE; 664 goto failed; 665 } 666 667 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 668 669 new_flags = dsp->ds_promisc; 670 switch (dlp->dl_level) { 671 case DL_PROMISC_SAP: 672 if (!(dsp->ds_promisc & DLS_PROMISC_SAP)) { 673 dl_err = DL_NOTENAB; 674 goto failed; 675 } 676 new_flags &= ~DLS_PROMISC_SAP; 677 break; 678 679 case DL_PROMISC_MULTI: 680 if (!(dsp->ds_promisc & DLS_PROMISC_MULTI)) { 681 dl_err = DL_NOTENAB; 682 goto failed; 683 } 684 new_flags &= ~DLS_PROMISC_MULTI; 685 break; 686 687 case DL_PROMISC_PHYS: 688 if (!(dsp->ds_promisc & DLS_PROMISC_PHYS)) { 689 dl_err = DL_NOTENAB; 690 goto failed; 691 } 692 new_flags &= ~DLS_PROMISC_PHYS; 693 break; 694 695 default: 696 dl_err = DL_NOTSUPPORTED; 697 mac_perim_exit(mph); 698 goto failed; 699 } 700 701 /* 702 * Adjust channel promiscuity. 703 */ 704 err = dls_promisc(dsp, new_flags); 705 706 if (err != 0) { 707 mac_perim_exit(mph); 708 dl_err = DL_SYSERR; 709 goto failed; 710 } 711 712 ASSERT(dsp->ds_promisc == new_flags); 713 if (dsp->ds_promisc == 0) 714 dls_active_clear(dsp, B_FALSE); 715 716 mac_perim_exit(mph); 717 718 dlokack(q, mp, DL_PROMISCOFF_REQ); 719 return; 720 failed: 721 dlerrorack(q, mp, DL_PROMISCOFF_REQ, dl_err, (t_uscalar_t)err); 722 } 723 724 /* 725 * DL_ENABMULTI_REQ 726 */ 727 static void 728 proto_enabmulti_req(dld_str_t *dsp, mblk_t *mp) 729 { 730 dl_enabmulti_req_t *dlp = (dl_enabmulti_req_t *)mp->b_rptr; 731 int err = 0; 732 t_uscalar_t dl_err; 733 queue_t *q = dsp->ds_wq; 734 mac_perim_handle_t mph; 735 736 if (dsp->ds_dlstate == DL_UNATTACHED || 737 DL_ACK_PENDING(dsp->ds_dlstate)) { 738 dl_err = DL_OUTSTATE; 739 goto failed; 740 } 741 742 if (MBLKL(mp) < sizeof (dl_enabmulti_req_t) || 743 !MBLKIN(mp, dlp->dl_addr_offset, dlp->dl_addr_length) || 744 dlp->dl_addr_length != dsp->ds_mip->mi_addr_length) { 745 dl_err = DL_BADPRIM; 746 goto failed; 747 } 748 749 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 750 751 if ((dsp->ds_dmap == NULL) && (err = dls_active_set(dsp)) != 0) { 752 dl_err = DL_SYSERR; 753 goto failed2; 754 } 755 756 err = dls_multicst_add(dsp, mp->b_rptr + dlp->dl_addr_offset); 757 if (err != 0) { 758 switch (err) { 759 case EINVAL: 760 dl_err = DL_BADADDR; 761 err = 0; 762 break; 763 case ENOSPC: 764 dl_err = DL_TOOMANY; 765 err = 0; 766 break; 767 default: 768 dl_err = DL_SYSERR; 769 break; 770 } 771 if (dsp->ds_dmap == NULL) 772 dls_active_clear(dsp, B_FALSE); 773 goto failed2; 774 } 775 776 mac_perim_exit(mph); 777 778 dlokack(q, mp, DL_ENABMULTI_REQ); 779 return; 780 781 failed2: 782 mac_perim_exit(mph); 783 failed: 784 dlerrorack(q, mp, DL_ENABMULTI_REQ, dl_err, (t_uscalar_t)err); 785 } 786 787 /* 788 * DL_DISABMULTI_REQ 789 */ 790 static void 791 proto_disabmulti_req(dld_str_t *dsp, mblk_t *mp) 792 { 793 dl_disabmulti_req_t *dlp = (dl_disabmulti_req_t *)mp->b_rptr; 794 int err = 0; 795 t_uscalar_t dl_err; 796 queue_t *q = dsp->ds_wq; 797 mac_perim_handle_t mph; 798 799 if (dsp->ds_dlstate == DL_UNATTACHED || 800 DL_ACK_PENDING(dsp->ds_dlstate)) { 801 dl_err = DL_OUTSTATE; 802 goto failed; 803 } 804 805 if (MBLKL(mp) < sizeof (dl_disabmulti_req_t) || 806 !MBLKIN(mp, dlp->dl_addr_offset, dlp->dl_addr_length) || 807 dlp->dl_addr_length != dsp->ds_mip->mi_addr_length) { 808 dl_err = DL_BADPRIM; 809 goto failed; 810 } 811 812 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 813 err = dls_multicst_remove(dsp, mp->b_rptr + dlp->dl_addr_offset); 814 if ((err == 0) && (dsp->ds_dmap == NULL)) 815 dls_active_clear(dsp, B_FALSE); 816 mac_perim_exit(mph); 817 818 if (err != 0) { 819 switch (err) { 820 case EINVAL: 821 dl_err = DL_BADADDR; 822 err = 0; 823 break; 824 825 case ENOENT: 826 dl_err = DL_NOTENAB; 827 err = 0; 828 break; 829 830 default: 831 dl_err = DL_SYSERR; 832 break; 833 } 834 goto failed; 835 } 836 dlokack(q, mp, DL_DISABMULTI_REQ); 837 return; 838 failed: 839 dlerrorack(q, mp, DL_DISABMULTI_REQ, dl_err, (t_uscalar_t)err); 840 } 841 842 /* 843 * DL_PHYS_ADDR_REQ 844 */ 845 static void 846 proto_physaddr_req(dld_str_t *dsp, mblk_t *mp) 847 { 848 dl_phys_addr_req_t *dlp = (dl_phys_addr_req_t *)mp->b_rptr; 849 queue_t *q = dsp->ds_wq; 850 t_uscalar_t dl_err = 0; 851 char *addr = NULL; 852 uint_t addr_length; 853 854 if (MBLKL(mp) < sizeof (dl_phys_addr_req_t)) { 855 dl_err = DL_BADPRIM; 856 goto done; 857 } 858 859 if (dsp->ds_dlstate == DL_UNATTACHED || 860 DL_ACK_PENDING(dsp->ds_dlstate)) { 861 dl_err = DL_OUTSTATE; 862 goto done; 863 } 864 865 addr_length = dsp->ds_mip->mi_addr_length; 866 if (addr_length > 0) { 867 addr = kmem_alloc(addr_length, KM_SLEEP); 868 switch (dlp->dl_addr_type) { 869 case DL_CURR_PHYS_ADDR: 870 mac_unicast_primary_get(dsp->ds_mh, (uint8_t *)addr); 871 break; 872 case DL_FACT_PHYS_ADDR: 873 bcopy(dsp->ds_mip->mi_unicst_addr, addr, addr_length); 874 break; 875 case DL_CURR_DEST_ADDR: 876 if (!mac_dst_get(dsp->ds_mh, (uint8_t *)addr)) 877 dl_err = DL_NOTSUPPORTED; 878 break; 879 default: 880 dl_err = DL_UNSUPPORTED; 881 } 882 } 883 done: 884 if (dl_err == 0) 885 dlphysaddrack(q, mp, addr, (t_uscalar_t)addr_length); 886 else 887 dlerrorack(q, mp, DL_PHYS_ADDR_REQ, dl_err, 0); 888 if (addr != NULL) 889 kmem_free(addr, addr_length); 890 } 891 892 /* 893 * DL_SET_PHYS_ADDR_REQ 894 */ 895 static void 896 proto_setphysaddr_req(dld_str_t *dsp, mblk_t *mp) 897 { 898 dl_set_phys_addr_req_t *dlp = (dl_set_phys_addr_req_t *)mp->b_rptr; 899 int err = 0; 900 t_uscalar_t dl_err; 901 queue_t *q = dsp->ds_wq; 902 mac_perim_handle_t mph; 903 904 if (dsp->ds_dlstate == DL_UNATTACHED || 905 DL_ACK_PENDING(dsp->ds_dlstate)) { 906 dl_err = DL_OUTSTATE; 907 goto failed; 908 } 909 910 if (MBLKL(mp) < sizeof (dl_set_phys_addr_req_t) || 911 !MBLKIN(mp, dlp->dl_addr_offset, dlp->dl_addr_length) || 912 dlp->dl_addr_length != dsp->ds_mip->mi_addr_length) { 913 dl_err = DL_BADPRIM; 914 goto failed; 915 } 916 917 mac_perim_enter_by_mh(dsp->ds_mh, &mph); 918 919 if ((err = dls_active_set(dsp)) != 0) { 920 dl_err = DL_SYSERR; 921 goto failed2; 922 } 923 924 /* 925 * If mac-nospoof is enabled and the link is owned by a 926 * non-global zone, changing the mac address is not allowed. 927 */ 928 if (dsp->ds_dlp->dl_zid != GLOBAL_ZONEID && 929 mac_protect_enabled(dsp->ds_mch, MPT_MACNOSPOOF)) { 930 dls_active_clear(dsp, B_FALSE); 931 err = EACCES; 932 goto failed2; 933 } 934 935 err = mac_unicast_primary_set(dsp->ds_mh, 936 mp->b_rptr + dlp->dl_addr_offset); 937 if (err != 0) { 938 switch (err) { 939 case EINVAL: 940 dl_err = DL_BADADDR; 941 err = 0; 942 break; 943 944 default: 945 dl_err = DL_SYSERR; 946 break; 947 } 948 dls_active_clear(dsp, B_FALSE); 949 goto failed2; 950 951 } 952 953 mac_perim_exit(mph); 954 955 dlokack(q, mp, DL_SET_PHYS_ADDR_REQ); 956 return; 957 958 failed2: 959 mac_perim_exit(mph); 960 failed: 961 dlerrorack(q, mp, DL_SET_PHYS_ADDR_REQ, dl_err, (t_uscalar_t)err); 962 } 963 964 /* 965 * DL_UDQOS_REQ 966 */ 967 static void 968 proto_udqos_req(dld_str_t *dsp, mblk_t *mp) 969 { 970 dl_udqos_req_t *dlp = (dl_udqos_req_t *)mp->b_rptr; 971 dl_qos_cl_sel1_t *selp; 972 int off, len; 973 t_uscalar_t dl_err; 974 queue_t *q = dsp->ds_wq; 975 976 off = dlp->dl_qos_offset; 977 len = dlp->dl_qos_length; 978 979 if (MBLKL(mp) < sizeof (dl_udqos_req_t) || !MBLKIN(mp, off, len)) { 980 dl_err = DL_BADPRIM; 981 goto failed; 982 } 983 984 selp = (dl_qos_cl_sel1_t *)(mp->b_rptr + off); 985 if (selp->dl_qos_type != DL_QOS_CL_SEL1) { 986 dl_err = DL_BADQOSTYPE; 987 goto failed; 988 } 989 990 if (selp->dl_priority > (1 << VLAN_PRI_SIZE) - 1 || 991 selp->dl_priority < 0) { 992 dl_err = DL_BADQOSPARAM; 993 goto failed; 994 } 995 996 dsp->ds_pri = selp->dl_priority; 997 dlokack(q, mp, DL_UDQOS_REQ); 998 return; 999 failed: 1000 dlerrorack(q, mp, DL_UDQOS_REQ, dl_err, 0); 1001 } 1002 1003 static boolean_t 1004 check_mod_above(queue_t *q, const char *mod) 1005 { 1006 queue_t *next_q; 1007 boolean_t ret = B_TRUE; 1008 1009 claimstr(q); 1010 next_q = q->q_next; 1011 if (strcmp(next_q->q_qinfo->qi_minfo->mi_idname, mod) != 0) 1012 ret = B_FALSE; 1013 releasestr(q); 1014 return (ret); 1015 } 1016 1017 /* 1018 * DL_CAPABILITY_REQ 1019 */ 1020 static void 1021 proto_capability_req(dld_str_t *dsp, mblk_t *mp) 1022 { 1023 dl_capability_req_t *dlp = (dl_capability_req_t *)mp->b_rptr; 1024 dl_capability_sub_t *sp; 1025 size_t size, len; 1026 offset_t off, end; 1027 t_uscalar_t dl_err; 1028 queue_t *q = dsp->ds_wq; 1029 1030 if (MBLKL(mp) < sizeof (dl_capability_req_t)) { 1031 dl_err = DL_BADPRIM; 1032 goto failed; 1033 } 1034 1035 if (dsp->ds_dlstate == DL_UNATTACHED || 1036 DL_ACK_PENDING(dsp->ds_dlstate)) { 1037 dl_err = DL_OUTSTATE; 1038 goto failed; 1039 } 1040 1041 /* 1042 * This request is overloaded. If there are no requested capabilities 1043 * then we just want to acknowledge with all the capabilities we 1044 * support. Otherwise we enable the set of capabilities requested. 1045 */ 1046 if (dlp->dl_sub_length == 0) { 1047 proto_capability_advertise(dsp, mp); 1048 return; 1049 } 1050 1051 if (!MBLKIN(mp, dlp->dl_sub_offset, dlp->dl_sub_length)) { 1052 dl_err = DL_BADPRIM; 1053 goto failed; 1054 } 1055 1056 dlp->dl_primitive = DL_CAPABILITY_ACK; 1057 1058 off = dlp->dl_sub_offset; 1059 len = dlp->dl_sub_length; 1060 1061 /* 1062 * Walk the list of capabilities to be enabled. 1063 */ 1064 for (end = off + len; off < end; ) { 1065 sp = (dl_capability_sub_t *)(mp->b_rptr + off); 1066 size = sizeof (dl_capability_sub_t) + sp->dl_length; 1067 1068 if (off + size > end || 1069 !IS_P2ALIGNED(off, sizeof (uint32_t))) { 1070 dl_err = DL_BADPRIM; 1071 goto failed; 1072 } 1073 1074 switch (sp->dl_cap) { 1075 /* 1076 * TCP/IP checksum offload to hardware. 1077 */ 1078 case DL_CAPAB_HCKSUM: { 1079 dl_capab_hcksum_t *hcksump; 1080 dl_capab_hcksum_t hcksum; 1081 1082 hcksump = (dl_capab_hcksum_t *)&sp[1]; 1083 /* 1084 * Copy for alignment. 1085 */ 1086 bcopy(hcksump, &hcksum, sizeof (dl_capab_hcksum_t)); 1087 dlcapabsetqid(&(hcksum.hcksum_mid), dsp->ds_rq); 1088 bcopy(&hcksum, hcksump, sizeof (dl_capab_hcksum_t)); 1089 break; 1090 } 1091 1092 case DL_CAPAB_DLD: { 1093 dl_capab_dld_t *dldp; 1094 dl_capab_dld_t dld; 1095 1096 dldp = (dl_capab_dld_t *)&sp[1]; 1097 /* 1098 * Copy for alignment. 1099 */ 1100 bcopy(dldp, &dld, sizeof (dl_capab_dld_t)); 1101 dlcapabsetqid(&(dld.dld_mid), dsp->ds_rq); 1102 bcopy(&dld, dldp, sizeof (dl_capab_dld_t)); 1103 break; 1104 } 1105 default: 1106 break; 1107 } 1108 off += size; 1109 } 1110 qreply(q, mp); 1111 return; 1112 failed: 1113 dlerrorack(q, mp, DL_CAPABILITY_REQ, dl_err, 0); 1114 } 1115 1116 /* 1117 * DL_NOTIFY_REQ 1118 */ 1119 static void 1120 proto_notify_req(dld_str_t *dsp, mblk_t *mp) 1121 { 1122 dl_notify_req_t *dlp = (dl_notify_req_t *)mp->b_rptr; 1123 t_uscalar_t dl_err; 1124 queue_t *q = dsp->ds_wq; 1125 uint_t note = 1126 DL_NOTE_PROMISC_ON_PHYS | 1127 DL_NOTE_PROMISC_OFF_PHYS | 1128 DL_NOTE_PHYS_ADDR | 1129 DL_NOTE_LINK_UP | 1130 DL_NOTE_LINK_DOWN | 1131 DL_NOTE_CAPAB_RENEG | 1132 DL_NOTE_FASTPATH_FLUSH | 1133 DL_NOTE_SPEED | 1134 DL_NOTE_SDU_SIZE| 1135 DL_NOTE_SDU_SIZE2| 1136 DL_NOTE_ALLOWED_IPS; 1137 1138 if (MBLKL(mp) < sizeof (dl_notify_req_t)) { 1139 dl_err = DL_BADPRIM; 1140 goto failed; 1141 } 1142 1143 if (dsp->ds_dlstate == DL_UNATTACHED || 1144 DL_ACK_PENDING(dsp->ds_dlstate)) { 1145 dl_err = DL_OUTSTATE; 1146 goto failed; 1147 } 1148 1149 note &= ~(mac_no_notification(dsp->ds_mh)); 1150 1151 /* 1152 * Cache the notifications that are being enabled. 1153 */ 1154 dsp->ds_notifications = dlp->dl_notifications & note; 1155 /* 1156 * The ACK carries all notifications regardless of which set is 1157 * being enabled. 1158 */ 1159 dlnotifyack(q, mp, note); 1160 1161 /* 1162 * Generate DL_NOTIFY_IND messages for each enabled notification. 1163 */ 1164 if (dsp->ds_notifications != 0) { 1165 dld_str_notify_ind(dsp); 1166 } 1167 return; 1168 failed: 1169 dlerrorack(q, mp, DL_NOTIFY_REQ, dl_err, 0); 1170 } 1171 1172 /* 1173 * DL_UINTDATA_REQ 1174 */ 1175 void 1176 proto_unitdata_req(dld_str_t *dsp, mblk_t *mp) 1177 { 1178 queue_t *q = dsp->ds_wq; 1179 dl_unitdata_req_t *dlp = (dl_unitdata_req_t *)mp->b_rptr; 1180 off_t off; 1181 size_t len, size; 1182 const uint8_t *addr; 1183 uint16_t sap; 1184 uint_t addr_length; 1185 mblk_t *bp, *payload; 1186 uint32_t start, stuff, end, value, flags; 1187 t_uscalar_t dl_err; 1188 uint_t max_sdu; 1189 1190 if (MBLKL(mp) < sizeof (dl_unitdata_req_t) || mp->b_cont == NULL) { 1191 dlerrorack(q, mp, DL_UNITDATA_REQ, DL_BADPRIM, 0); 1192 return; 1193 } 1194 1195 mutex_enter(&dsp->ds_lock); 1196 if (dsp->ds_dlstate != DL_IDLE) { 1197 mutex_exit(&dsp->ds_lock); 1198 dlerrorack(q, mp, DL_UNITDATA_REQ, DL_OUTSTATE, 0); 1199 return; 1200 } 1201 DLD_DATATHR_INC(dsp); 1202 mutex_exit(&dsp->ds_lock); 1203 1204 addr_length = dsp->ds_mip->mi_addr_length; 1205 1206 off = dlp->dl_dest_addr_offset; 1207 len = dlp->dl_dest_addr_length; 1208 1209 if (!MBLKIN(mp, off, len) || !IS_P2ALIGNED(off, sizeof (uint16_t))) { 1210 dl_err = DL_BADPRIM; 1211 goto failed; 1212 } 1213 1214 if (len != addr_length + sizeof (uint16_t)) { 1215 dl_err = DL_BADADDR; 1216 goto failed; 1217 } 1218 1219 addr = mp->b_rptr + off; 1220 sap = *(uint16_t *)(mp->b_rptr + off + addr_length); 1221 1222 /* 1223 * Check the length of the packet and the block types. 1224 */ 1225 size = 0; 1226 payload = mp->b_cont; 1227 for (bp = payload; bp != NULL; bp = bp->b_cont) { 1228 if (DB_TYPE(bp) != M_DATA) 1229 goto baddata; 1230 1231 size += MBLKL(bp); 1232 } 1233 1234 mac_sdu_get(dsp->ds_mh, NULL, &max_sdu); 1235 if (size > max_sdu) 1236 goto baddata; 1237 1238 /* 1239 * Build a packet header. 1240 */ 1241 if ((bp = dls_header(dsp, addr, sap, dlp->dl_priority.dl_max, 1242 &payload)) == NULL) { 1243 dl_err = DL_BADADDR; 1244 goto failed; 1245 } 1246 1247 /* 1248 * We no longer need the M_PROTO header, so free it. 1249 */ 1250 freeb(mp); 1251 1252 /* 1253 * Transfer the checksum offload information if it is present. 1254 */ 1255 hcksum_retrieve(payload, NULL, NULL, &start, &stuff, &end, &value, 1256 &flags); 1257 (void) hcksum_assoc(bp, NULL, NULL, start, stuff, end, value, flags, 0); 1258 1259 /* 1260 * Link the payload onto the new header. 1261 */ 1262 ASSERT(bp->b_cont == NULL); 1263 bp->b_cont = payload; 1264 1265 /* 1266 * No lock can be held across modules and putnext()'s, 1267 * which can happen here with the call from DLD_TX(). 1268 */ 1269 if (DLD_TX(dsp, bp, 0, 0) != NULL) { 1270 /* flow-controlled */ 1271 DLD_SETQFULL(dsp); 1272 } 1273 DLD_DATATHR_DCR(dsp); 1274 return; 1275 1276 failed: 1277 dlerrorack(q, mp, DL_UNITDATA_REQ, dl_err, 0); 1278 DLD_DATATHR_DCR(dsp); 1279 return; 1280 1281 baddata: 1282 dluderrorind(q, mp, (void *)addr, len, DL_BADDATA, 0); 1283 DLD_DATATHR_DCR(dsp); 1284 } 1285 1286 /* 1287 * DL_PASSIVE_REQ 1288 */ 1289 static void 1290 proto_passive_req(dld_str_t *dsp, mblk_t *mp) 1291 { 1292 t_uscalar_t dl_err; 1293 1294 /* 1295 * If we've already become active by issuing an active primitive, 1296 * then it's too late to try to become passive. 1297 */ 1298 if (dsp->ds_passivestate == DLD_ACTIVE) { 1299 dl_err = DL_OUTSTATE; 1300 goto failed; 1301 } 1302 1303 if (MBLKL(mp) < sizeof (dl_passive_req_t)) { 1304 dl_err = DL_BADPRIM; 1305 goto failed; 1306 } 1307 1308 dsp->ds_passivestate = DLD_PASSIVE; 1309 dlokack(dsp->ds_wq, mp, DL_PASSIVE_REQ); 1310 return; 1311 failed: 1312 dlerrorack(dsp->ds_wq, mp, DL_PASSIVE_REQ, dl_err, 0); 1313 } 1314 1315 1316 /* 1317 * Catch-all handler. 1318 */ 1319 static void 1320 proto_req(dld_str_t *dsp, mblk_t *mp) 1321 { 1322 union DL_primitives *dlp = (union DL_primitives *)mp->b_rptr; 1323 1324 dlerrorack(dsp->ds_wq, mp, dlp->dl_primitive, DL_UNSUPPORTED, 0); 1325 } 1326 1327 static int 1328 dld_capab_perim(dld_str_t *dsp, void *data, uint_t flags) 1329 { 1330 switch (flags) { 1331 case DLD_ENABLE: 1332 mac_perim_enter_by_mh(dsp->ds_mh, (mac_perim_handle_t *)data); 1333 return (0); 1334 1335 case DLD_DISABLE: 1336 mac_perim_exit((mac_perim_handle_t)data); 1337 return (0); 1338 1339 case DLD_QUERY: 1340 return (mac_perim_held(dsp->ds_mh)); 1341 } 1342 return (0); 1343 } 1344 1345 static int 1346 dld_capab_direct(dld_str_t *dsp, void *data, uint_t flags) 1347 { 1348 dld_capab_direct_t *direct = data; 1349 1350 ASSERT(MAC_PERIM_HELD(dsp->ds_mh)); 1351 1352 switch (flags) { 1353 case DLD_ENABLE: 1354 dls_rx_set(dsp, (dls_rx_t)direct->di_rx_cf, 1355 direct->di_rx_ch); 1356 1357 direct->di_tx_df = (uintptr_t)str_mdata_fastpath_put; 1358 direct->di_tx_dh = dsp; 1359 direct->di_tx_cb_df = (uintptr_t)mac_client_tx_notify; 1360 direct->di_tx_cb_dh = dsp->ds_mch; 1361 direct->di_tx_fctl_df = (uintptr_t)mac_tx_is_flow_blocked; 1362 direct->di_tx_fctl_dh = dsp->ds_mch; 1363 1364 dsp->ds_direct = B_TRUE; 1365 1366 return (0); 1367 1368 case DLD_DISABLE: 1369 dls_rx_set(dsp, (dsp->ds_mode == DLD_FASTPATH) ? 1370 dld_str_rx_fastpath : dld_str_rx_unitdata, (void *)dsp); 1371 dsp->ds_direct = B_FALSE; 1372 1373 return (0); 1374 } 1375 return (ENOTSUP); 1376 } 1377 1378 /* 1379 * dld_capab_poll_enable() 1380 * 1381 * This function is misnamed. All polling and fanouts are run out of the 1382 * lower mac (in case of VNIC and the only mac in case of NICs). The 1383 * availability of Rx ring and promiscous mode is all taken care between 1384 * the soft ring set (mac_srs), the Rx ring, and S/W classifier. Any 1385 * fanout necessary is done by the soft rings that are part of the 1386 * mac_srs (by default mac_srs sends the packets up via a TCP and 1387 * non TCP soft ring). 1388 * 1389 * The mac_srs (or its associated soft rings) always store the ill_rx_ring 1390 * (the cookie returned when they registered with IP during plumb) as their 1391 * 2nd argument which is passed up as mac_resource_handle_t. The upcall 1392 * function and 1st argument is what the caller registered when they 1393 * called mac_rx_classify_flow_add() to register the flow. For VNIC, 1394 * the function is vnic_rx and argument is vnic_t. For regular NIC 1395 * case, it mac_rx_default and mac_handle_t. As explained above, the 1396 * mac_srs (or its soft ring) will add the ill_rx_ring (mac_resource_handle_t) 1397 * from its stored 2nd argument. 1398 */ 1399 static int 1400 dld_capab_poll_enable(dld_str_t *dsp, dld_capab_poll_t *poll) 1401 { 1402 if (dsp->ds_polling) 1403 return (EINVAL); 1404 1405 if ((dld_opt & DLD_OPT_NO_POLL) != 0 || dsp->ds_mode == DLD_RAW) 1406 return (ENOTSUP); 1407 1408 /* 1409 * Enable client polling if and only if DLS bypass is possible. 1410 * Special cases like VLANs need DLS processing in the Rx data path. 1411 * In such a case we can neither allow the client (IP) to directly 1412 * poll the softring (since DLS processing hasn't been done) nor can 1413 * we allow DLS bypass. 1414 */ 1415 if (!mac_rx_bypass_set(dsp->ds_mch, dsp->ds_rx, dsp->ds_rx_arg)) 1416 return (ENOTSUP); 1417 1418 /* 1419 * Register soft ring resources. This will come in handy later if 1420 * the user decides to modify CPU bindings to use more CPUs for the 1421 * device in which case we will switch to fanout using soft rings. 1422 */ 1423 mac_resource_set_common(dsp->ds_mch, 1424 (mac_resource_add_t)poll->poll_ring_add_cf, 1425 (mac_resource_remove_t)poll->poll_ring_remove_cf, 1426 (mac_resource_quiesce_t)poll->poll_ring_quiesce_cf, 1427 (mac_resource_restart_t)poll->poll_ring_restart_cf, 1428 (mac_resource_bind_t)poll->poll_ring_bind_cf, 1429 poll->poll_ring_ch); 1430 1431 mac_client_poll_enable(dsp->ds_mch); 1432 1433 dsp->ds_polling = B_TRUE; 1434 return (0); 1435 } 1436 1437 /* ARGSUSED */ 1438 static int 1439 dld_capab_poll_disable(dld_str_t *dsp, dld_capab_poll_t *poll) 1440 { 1441 if (!dsp->ds_polling) 1442 return (EINVAL); 1443 1444 mac_client_poll_disable(dsp->ds_mch); 1445 mac_resource_set(dsp->ds_mch, NULL, NULL); 1446 1447 dsp->ds_polling = B_FALSE; 1448 return (0); 1449 } 1450 1451 static int 1452 dld_capab_poll(dld_str_t *dsp, void *data, uint_t flags) 1453 { 1454 dld_capab_poll_t *poll = data; 1455 1456 ASSERT(MAC_PERIM_HELD(dsp->ds_mh)); 1457 1458 switch (flags) { 1459 case DLD_ENABLE: 1460 return (dld_capab_poll_enable(dsp, poll)); 1461 case DLD_DISABLE: 1462 return (dld_capab_poll_disable(dsp, poll)); 1463 } 1464 return (ENOTSUP); 1465 } 1466 1467 static int 1468 dld_capab_lso(dld_str_t *dsp, void *data, uint_t flags) 1469 { 1470 dld_capab_lso_t *lso = data; 1471 1472 ASSERT(MAC_PERIM_HELD(dsp->ds_mh)); 1473 1474 switch (flags) { 1475 case DLD_ENABLE: { 1476 mac_capab_lso_t mac_lso; 1477 1478 /* 1479 * Check if LSO is supported on this MAC & enable LSO 1480 * accordingly. 1481 */ 1482 if (mac_capab_get(dsp->ds_mh, MAC_CAPAB_LSO, &mac_lso)) { 1483 lso->lso_max = mac_lso.lso_basic_tcp_ipv4.lso_max; 1484 lso->lso_flags = 0; 1485 /* translate the flag for mac clients */ 1486 if ((mac_lso.lso_flags & LSO_TX_BASIC_TCP_IPV4) != 0) 1487 lso->lso_flags |= DLD_LSO_BASIC_TCP_IPV4; 1488 dsp->ds_lso = B_TRUE; 1489 dsp->ds_lso_max = lso->lso_max; 1490 } else { 1491 dsp->ds_lso = B_FALSE; 1492 dsp->ds_lso_max = 0; 1493 return (ENOTSUP); 1494 } 1495 return (0); 1496 } 1497 case DLD_DISABLE: { 1498 dsp->ds_lso = B_FALSE; 1499 dsp->ds_lso_max = 0; 1500 return (0); 1501 } 1502 } 1503 return (ENOTSUP); 1504 } 1505 1506 static int 1507 dld_capab(dld_str_t *dsp, uint_t type, void *data, uint_t flags) 1508 { 1509 int err; 1510 1511 /* 1512 * Don't enable direct callback capabilities unless the caller is 1513 * the IP client. When a module is inserted in a stream (_I_INSERT) 1514 * the stack initiates capability disable, but due to races, the 1515 * module insertion may complete before the capability disable 1516 * completes. So we limit the check to DLD_ENABLE case. 1517 */ 1518 if ((flags == DLD_ENABLE && type != DLD_CAPAB_PERIM) && 1519 (dsp->ds_sap != ETHERTYPE_IP || 1520 !check_mod_above(dsp->ds_rq, "ip"))) { 1521 return (ENOTSUP); 1522 } 1523 1524 switch (type) { 1525 case DLD_CAPAB_DIRECT: 1526 err = dld_capab_direct(dsp, data, flags); 1527 break; 1528 1529 case DLD_CAPAB_POLL: 1530 err = dld_capab_poll(dsp, data, flags); 1531 break; 1532 1533 case DLD_CAPAB_PERIM: 1534 err = dld_capab_perim(dsp, data, flags); 1535 break; 1536 1537 case DLD_CAPAB_LSO: 1538 err = dld_capab_lso(dsp, data, flags); 1539 break; 1540 1541 default: 1542 err = ENOTSUP; 1543 break; 1544 } 1545 1546 return (err); 1547 } 1548 1549 /* 1550 * DL_CAPABILITY_ACK/DL_ERROR_ACK 1551 */ 1552 static void 1553 proto_capability_advertise(dld_str_t *dsp, mblk_t *mp) 1554 { 1555 dl_capability_ack_t *dlap; 1556 dl_capability_sub_t *dlsp; 1557 size_t subsize; 1558 dl_capab_dld_t dld; 1559 dl_capab_hcksum_t hcksum; 1560 dl_capab_zerocopy_t zcopy; 1561 dl_capab_vrrp_t vrrp; 1562 mac_capab_vrrp_t vrrp_capab; 1563 uint8_t *ptr; 1564 queue_t *q = dsp->ds_wq; 1565 mblk_t *mp1; 1566 boolean_t hcksum_capable = B_FALSE; 1567 boolean_t zcopy_capable = B_FALSE; 1568 boolean_t dld_capable = B_FALSE; 1569 boolean_t vrrp_capable = B_FALSE; 1570 1571 /* 1572 * Initially assume no capabilities. 1573 */ 1574 subsize = 0; 1575 1576 /* 1577 * Check if checksum offload is supported on this MAC. 1578 */ 1579 bzero(&hcksum, sizeof (dl_capab_hcksum_t)); 1580 if (mac_capab_get(dsp->ds_mh, MAC_CAPAB_HCKSUM, 1581 &hcksum.hcksum_txflags)) { 1582 if (hcksum.hcksum_txflags != 0) { 1583 hcksum_capable = B_TRUE; 1584 subsize += sizeof (dl_capability_sub_t) + 1585 sizeof (dl_capab_hcksum_t); 1586 } 1587 } 1588 1589 /* 1590 * Check if zerocopy is supported on this interface. 1591 * If advertising DL_CAPAB_ZEROCOPY has not been explicitly disabled 1592 * then reserve space for that capability. 1593 */ 1594 if (!mac_capab_get(dsp->ds_mh, MAC_CAPAB_NO_ZCOPY, NULL) && 1595 !(dld_opt & DLD_OPT_NO_ZEROCOPY)) { 1596 zcopy_capable = B_TRUE; 1597 subsize += sizeof (dl_capability_sub_t) + 1598 sizeof (dl_capab_zerocopy_t); 1599 } 1600 1601 /* 1602 * Direct capability negotiation interface between IP and DLD 1603 */ 1604 if (dsp->ds_sap == ETHERTYPE_IP && check_mod_above(dsp->ds_rq, "ip")) { 1605 dld_capable = B_TRUE; 1606 subsize += sizeof (dl_capability_sub_t) + 1607 sizeof (dl_capab_dld_t); 1608 } 1609 1610 /* 1611 * Check if vrrp is supported on this interface. If so, reserve 1612 * space for that capability. 1613 */ 1614 if (mac_capab_get(dsp->ds_mh, MAC_CAPAB_VRRP, &vrrp_capab)) { 1615 vrrp_capable = B_TRUE; 1616 subsize += sizeof (dl_capability_sub_t) + 1617 sizeof (dl_capab_vrrp_t); 1618 } 1619 1620 /* 1621 * If there are no capabilities to advertise or if we 1622 * can't allocate a response, send a DL_ERROR_ACK. 1623 */ 1624 if ((mp1 = reallocb(mp, 1625 sizeof (dl_capability_ack_t) + subsize, 0)) == NULL) { 1626 dlerrorack(q, mp, DL_CAPABILITY_REQ, DL_NOTSUPPORTED, 0); 1627 return; 1628 } 1629 1630 mp = mp1; 1631 DB_TYPE(mp) = M_PROTO; 1632 mp->b_wptr = mp->b_rptr + sizeof (dl_capability_ack_t) + subsize; 1633 bzero(mp->b_rptr, MBLKL(mp)); 1634 dlap = (dl_capability_ack_t *)mp->b_rptr; 1635 dlap->dl_primitive = DL_CAPABILITY_ACK; 1636 dlap->dl_sub_offset = sizeof (dl_capability_ack_t); 1637 dlap->dl_sub_length = subsize; 1638 ptr = (uint8_t *)&dlap[1]; 1639 1640 /* 1641 * TCP/IP checksum offload. 1642 */ 1643 if (hcksum_capable) { 1644 dlsp = (dl_capability_sub_t *)ptr; 1645 1646 dlsp->dl_cap = DL_CAPAB_HCKSUM; 1647 dlsp->dl_length = sizeof (dl_capab_hcksum_t); 1648 ptr += sizeof (dl_capability_sub_t); 1649 1650 hcksum.hcksum_version = HCKSUM_VERSION_1; 1651 dlcapabsetqid(&(hcksum.hcksum_mid), dsp->ds_rq); 1652 bcopy(&hcksum, ptr, sizeof (dl_capab_hcksum_t)); 1653 ptr += sizeof (dl_capab_hcksum_t); 1654 } 1655 1656 /* 1657 * Zero copy 1658 */ 1659 if (zcopy_capable) { 1660 dlsp = (dl_capability_sub_t *)ptr; 1661 1662 dlsp->dl_cap = DL_CAPAB_ZEROCOPY; 1663 dlsp->dl_length = sizeof (dl_capab_zerocopy_t); 1664 ptr += sizeof (dl_capability_sub_t); 1665 1666 bzero(&zcopy, sizeof (dl_capab_zerocopy_t)); 1667 zcopy.zerocopy_version = ZEROCOPY_VERSION_1; 1668 zcopy.zerocopy_flags = DL_CAPAB_VMSAFE_MEM; 1669 1670 dlcapabsetqid(&(zcopy.zerocopy_mid), dsp->ds_rq); 1671 bcopy(&zcopy, ptr, sizeof (dl_capab_zerocopy_t)); 1672 ptr += sizeof (dl_capab_zerocopy_t); 1673 } 1674 1675 /* 1676 * VRRP capability negotiation 1677 */ 1678 if (vrrp_capable) { 1679 dlsp = (dl_capability_sub_t *)ptr; 1680 dlsp->dl_cap = DL_CAPAB_VRRP; 1681 dlsp->dl_length = sizeof (dl_capab_vrrp_t); 1682 ptr += sizeof (dl_capability_sub_t); 1683 1684 bzero(&vrrp, sizeof (dl_capab_vrrp_t)); 1685 vrrp.vrrp_af = vrrp_capab.mcv_af; 1686 bcopy(&vrrp, ptr, sizeof (dl_capab_vrrp_t)); 1687 ptr += sizeof (dl_capab_vrrp_t); 1688 } 1689 1690 /* 1691 * Direct capability negotiation interface between IP and DLD. 1692 * Refer to dld.h for details. 1693 */ 1694 if (dld_capable) { 1695 dlsp = (dl_capability_sub_t *)ptr; 1696 dlsp->dl_cap = DL_CAPAB_DLD; 1697 dlsp->dl_length = sizeof (dl_capab_dld_t); 1698 ptr += sizeof (dl_capability_sub_t); 1699 1700 bzero(&dld, sizeof (dl_capab_dld_t)); 1701 dld.dld_version = DLD_CURRENT_VERSION; 1702 dld.dld_capab = (uintptr_t)dld_capab; 1703 dld.dld_capab_handle = (uintptr_t)dsp; 1704 1705 dlcapabsetqid(&(dld.dld_mid), dsp->ds_rq); 1706 bcopy(&dld, ptr, sizeof (dl_capab_dld_t)); 1707 ptr += sizeof (dl_capab_dld_t); 1708 } 1709 1710 ASSERT(ptr == mp->b_rptr + sizeof (dl_capability_ack_t) + subsize); 1711 qreply(q, mp); 1712 } 1713 1714 /* 1715 * Disable any enabled capabilities. 1716 */ 1717 void 1718 dld_capabilities_disable(dld_str_t *dsp) 1719 { 1720 if (dsp->ds_polling) 1721 (void) dld_capab_poll_disable(dsp, NULL); 1722 } 1723