1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Copyright (c) 2006, Intel Corporation. 4 * 5 * Copyright (C) 2006-2008 Intel Corporation 6 * Author: Ashok Raj <ashok.raj@intel.com> 7 * Author: Shaohua Li <shaohua.li@intel.com> 8 * Author: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> 9 * 10 * This file implements early detection/parsing of Remapping Devices 11 * reported to OS through BIOS via DMA remapping reporting (DMAR) ACPI 12 * tables. 13 * 14 * These routines are used by both DMA-remapping and Interrupt-remapping 15 */ 16 17 #define pr_fmt(fmt) "DMAR: " fmt 18 19 #include <linux/pci.h> 20 #include <linux/dmar.h> 21 #include <linux/iova.h> 22 #include <linux/intel-iommu.h> 23 #include <linux/timer.h> 24 #include <linux/irq.h> 25 #include <linux/interrupt.h> 26 #include <linux/tboot.h> 27 #include <linux/dmi.h> 28 #include <linux/slab.h> 29 #include <linux/iommu.h> 30 #include <linux/numa.h> 31 #include <linux/limits.h> 32 #include <asm/irq_remapping.h> 33 #include <asm/iommu_table.h> 34 35 #include "../irq_remapping.h" 36 37 typedef int (*dmar_res_handler_t)(struct acpi_dmar_header *, void *); 38 struct dmar_res_callback { 39 dmar_res_handler_t cb[ACPI_DMAR_TYPE_RESERVED]; 40 void *arg[ACPI_DMAR_TYPE_RESERVED]; 41 bool ignore_unhandled; 42 bool print_entry; 43 }; 44 45 /* 46 * Assumptions: 47 * 1) The hotplug framework guarentees that DMAR unit will be hot-added 48 * before IO devices managed by that unit. 49 * 2) The hotplug framework guarantees that DMAR unit will be hot-removed 50 * after IO devices managed by that unit. 51 * 3) Hotplug events are rare. 52 * 53 * Locking rules for DMA and interrupt remapping related global data structures: 54 * 1) Use dmar_global_lock in process context 55 * 2) Use RCU in interrupt context 56 */ 57 DECLARE_RWSEM(dmar_global_lock); 58 LIST_HEAD(dmar_drhd_units); 59 60 struct acpi_table_header * __initdata dmar_tbl; 61 static int dmar_dev_scope_status = 1; 62 static unsigned long dmar_seq_ids[BITS_TO_LONGS(DMAR_UNITS_SUPPORTED)]; 63 64 static int alloc_iommu(struct dmar_drhd_unit *drhd); 65 static void free_iommu(struct intel_iommu *iommu); 66 67 extern const struct iommu_ops intel_iommu_ops; 68 69 static void dmar_register_drhd_unit(struct dmar_drhd_unit *drhd) 70 { 71 /* 72 * add INCLUDE_ALL at the tail, so scan the list will find it at 73 * the very end. 74 */ 75 if (drhd->include_all) 76 list_add_tail_rcu(&drhd->list, &dmar_drhd_units); 77 else 78 list_add_rcu(&drhd->list, &dmar_drhd_units); 79 } 80 81 void *dmar_alloc_dev_scope(void *start, void *end, int *cnt) 82 { 83 struct acpi_dmar_device_scope *scope; 84 85 *cnt = 0; 86 while (start < end) { 87 scope = start; 88 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_NAMESPACE || 89 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT || 90 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE) 91 (*cnt)++; 92 else if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_IOAPIC && 93 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_HPET) { 94 pr_warn("Unsupported device scope\n"); 95 } 96 start += scope->length; 97 } 98 if (*cnt == 0) 99 return NULL; 100 101 return kcalloc(*cnt, sizeof(struct dmar_dev_scope), GFP_KERNEL); 102 } 103 104 void dmar_free_dev_scope(struct dmar_dev_scope **devices, int *cnt) 105 { 106 int i; 107 struct device *tmp_dev; 108 109 if (*devices && *cnt) { 110 for_each_active_dev_scope(*devices, *cnt, i, tmp_dev) 111 put_device(tmp_dev); 112 kfree(*devices); 113 } 114 115 *devices = NULL; 116 *cnt = 0; 117 } 118 119 /* Optimize out kzalloc()/kfree() for normal cases */ 120 static char dmar_pci_notify_info_buf[64]; 121 122 static struct dmar_pci_notify_info * 123 dmar_alloc_pci_notify_info(struct pci_dev *dev, unsigned long event) 124 { 125 int level = 0; 126 size_t size; 127 struct pci_dev *tmp; 128 struct dmar_pci_notify_info *info; 129 130 BUG_ON(dev->is_virtfn); 131 132 /* 133 * Ignore devices that have a domain number higher than what can 134 * be looked up in DMAR, e.g. VMD subdevices with domain 0x10000 135 */ 136 if (pci_domain_nr(dev->bus) > U16_MAX) 137 return NULL; 138 139 /* Only generate path[] for device addition event */ 140 if (event == BUS_NOTIFY_ADD_DEVICE) 141 for (tmp = dev; tmp; tmp = tmp->bus->self) 142 level++; 143 144 size = struct_size(info, path, level); 145 if (size <= sizeof(dmar_pci_notify_info_buf)) { 146 info = (struct dmar_pci_notify_info *)dmar_pci_notify_info_buf; 147 } else { 148 info = kzalloc(size, GFP_KERNEL); 149 if (!info) { 150 pr_warn("Out of memory when allocating notify_info " 151 "for %s.\n", pci_name(dev)); 152 if (dmar_dev_scope_status == 0) 153 dmar_dev_scope_status = -ENOMEM; 154 return NULL; 155 } 156 } 157 158 info->event = event; 159 info->dev = dev; 160 info->seg = pci_domain_nr(dev->bus); 161 info->level = level; 162 if (event == BUS_NOTIFY_ADD_DEVICE) { 163 for (tmp = dev; tmp; tmp = tmp->bus->self) { 164 level--; 165 info->path[level].bus = tmp->bus->number; 166 info->path[level].device = PCI_SLOT(tmp->devfn); 167 info->path[level].function = PCI_FUNC(tmp->devfn); 168 if (pci_is_root_bus(tmp->bus)) 169 info->bus = tmp->bus->number; 170 } 171 } 172 173 return info; 174 } 175 176 static inline void dmar_free_pci_notify_info(struct dmar_pci_notify_info *info) 177 { 178 if ((void *)info != dmar_pci_notify_info_buf) 179 kfree(info); 180 } 181 182 static bool dmar_match_pci_path(struct dmar_pci_notify_info *info, int bus, 183 struct acpi_dmar_pci_path *path, int count) 184 { 185 int i; 186 187 if (info->bus != bus) 188 goto fallback; 189 if (info->level != count) 190 goto fallback; 191 192 for (i = 0; i < count; i++) { 193 if (path[i].device != info->path[i].device || 194 path[i].function != info->path[i].function) 195 goto fallback; 196 } 197 198 return true; 199 200 fallback: 201 202 if (count != 1) 203 return false; 204 205 i = info->level - 1; 206 if (bus == info->path[i].bus && 207 path[0].device == info->path[i].device && 208 path[0].function == info->path[i].function) { 209 pr_info(FW_BUG "RMRR entry for device %02x:%02x.%x is broken - applying workaround\n", 210 bus, path[0].device, path[0].function); 211 return true; 212 } 213 214 return false; 215 } 216 217 /* Return: > 0 if match found, 0 if no match found, < 0 if error happens */ 218 int dmar_insert_dev_scope(struct dmar_pci_notify_info *info, 219 void *start, void*end, u16 segment, 220 struct dmar_dev_scope *devices, 221 int devices_cnt) 222 { 223 int i, level; 224 struct device *tmp, *dev = &info->dev->dev; 225 struct acpi_dmar_device_scope *scope; 226 struct acpi_dmar_pci_path *path; 227 228 if (segment != info->seg) 229 return 0; 230 231 for (; start < end; start += scope->length) { 232 scope = start; 233 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_ENDPOINT && 234 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_BRIDGE) 235 continue; 236 237 path = (struct acpi_dmar_pci_path *)(scope + 1); 238 level = (scope->length - sizeof(*scope)) / sizeof(*path); 239 if (!dmar_match_pci_path(info, scope->bus, path, level)) 240 continue; 241 242 /* 243 * We expect devices with endpoint scope to have normal PCI 244 * headers, and devices with bridge scope to have bridge PCI 245 * headers. However PCI NTB devices may be listed in the 246 * DMAR table with bridge scope, even though they have a 247 * normal PCI header. NTB devices are identified by class 248 * "BRIDGE_OTHER" (0680h) - we don't declare a socpe mismatch 249 * for this special case. 250 */ 251 if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT && 252 info->dev->hdr_type != PCI_HEADER_TYPE_NORMAL) || 253 (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE && 254 (info->dev->hdr_type == PCI_HEADER_TYPE_NORMAL && 255 info->dev->class >> 16 != PCI_BASE_CLASS_BRIDGE))) { 256 pr_warn("Device scope type does not match for %s\n", 257 pci_name(info->dev)); 258 return -EINVAL; 259 } 260 261 for_each_dev_scope(devices, devices_cnt, i, tmp) 262 if (tmp == NULL) { 263 devices[i].bus = info->dev->bus->number; 264 devices[i].devfn = info->dev->devfn; 265 rcu_assign_pointer(devices[i].dev, 266 get_device(dev)); 267 return 1; 268 } 269 BUG_ON(i >= devices_cnt); 270 } 271 272 return 0; 273 } 274 275 int dmar_remove_dev_scope(struct dmar_pci_notify_info *info, u16 segment, 276 struct dmar_dev_scope *devices, int count) 277 { 278 int index; 279 struct device *tmp; 280 281 if (info->seg != segment) 282 return 0; 283 284 for_each_active_dev_scope(devices, count, index, tmp) 285 if (tmp == &info->dev->dev) { 286 RCU_INIT_POINTER(devices[index].dev, NULL); 287 synchronize_rcu(); 288 put_device(tmp); 289 return 1; 290 } 291 292 return 0; 293 } 294 295 static int dmar_pci_bus_add_dev(struct dmar_pci_notify_info *info) 296 { 297 int ret = 0; 298 struct dmar_drhd_unit *dmaru; 299 struct acpi_dmar_hardware_unit *drhd; 300 301 for_each_drhd_unit(dmaru) { 302 if (dmaru->include_all) 303 continue; 304 305 drhd = container_of(dmaru->hdr, 306 struct acpi_dmar_hardware_unit, header); 307 ret = dmar_insert_dev_scope(info, (void *)(drhd + 1), 308 ((void *)drhd) + drhd->header.length, 309 dmaru->segment, 310 dmaru->devices, dmaru->devices_cnt); 311 if (ret) 312 break; 313 } 314 if (ret >= 0) 315 ret = dmar_iommu_notify_scope_dev(info); 316 if (ret < 0 && dmar_dev_scope_status == 0) 317 dmar_dev_scope_status = ret; 318 319 return ret; 320 } 321 322 static void dmar_pci_bus_del_dev(struct dmar_pci_notify_info *info) 323 { 324 struct dmar_drhd_unit *dmaru; 325 326 for_each_drhd_unit(dmaru) 327 if (dmar_remove_dev_scope(info, dmaru->segment, 328 dmaru->devices, dmaru->devices_cnt)) 329 break; 330 dmar_iommu_notify_scope_dev(info); 331 } 332 333 static int dmar_pci_bus_notifier(struct notifier_block *nb, 334 unsigned long action, void *data) 335 { 336 struct pci_dev *pdev = to_pci_dev(data); 337 struct dmar_pci_notify_info *info; 338 339 /* Only care about add/remove events for physical functions. 340 * For VFs we actually do the lookup based on the corresponding 341 * PF in device_to_iommu() anyway. */ 342 if (pdev->is_virtfn) 343 return NOTIFY_DONE; 344 if (action != BUS_NOTIFY_ADD_DEVICE && 345 action != BUS_NOTIFY_REMOVED_DEVICE) 346 return NOTIFY_DONE; 347 348 info = dmar_alloc_pci_notify_info(pdev, action); 349 if (!info) 350 return NOTIFY_DONE; 351 352 down_write(&dmar_global_lock); 353 if (action == BUS_NOTIFY_ADD_DEVICE) 354 dmar_pci_bus_add_dev(info); 355 else if (action == BUS_NOTIFY_REMOVED_DEVICE) 356 dmar_pci_bus_del_dev(info); 357 up_write(&dmar_global_lock); 358 359 dmar_free_pci_notify_info(info); 360 361 return NOTIFY_OK; 362 } 363 364 static struct notifier_block dmar_pci_bus_nb = { 365 .notifier_call = dmar_pci_bus_notifier, 366 .priority = INT_MIN, 367 }; 368 369 static struct dmar_drhd_unit * 370 dmar_find_dmaru(struct acpi_dmar_hardware_unit *drhd) 371 { 372 struct dmar_drhd_unit *dmaru; 373 374 list_for_each_entry_rcu(dmaru, &dmar_drhd_units, list, 375 dmar_rcu_check()) 376 if (dmaru->segment == drhd->segment && 377 dmaru->reg_base_addr == drhd->address) 378 return dmaru; 379 380 return NULL; 381 } 382 383 /** 384 * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition 385 * structure which uniquely represent one DMA remapping hardware unit 386 * present in the platform 387 */ 388 static int dmar_parse_one_drhd(struct acpi_dmar_header *header, void *arg) 389 { 390 struct acpi_dmar_hardware_unit *drhd; 391 struct dmar_drhd_unit *dmaru; 392 int ret; 393 394 drhd = (struct acpi_dmar_hardware_unit *)header; 395 dmaru = dmar_find_dmaru(drhd); 396 if (dmaru) 397 goto out; 398 399 dmaru = kzalloc(sizeof(*dmaru) + header->length, GFP_KERNEL); 400 if (!dmaru) 401 return -ENOMEM; 402 403 /* 404 * If header is allocated from slab by ACPI _DSM method, we need to 405 * copy the content because the memory buffer will be freed on return. 406 */ 407 dmaru->hdr = (void *)(dmaru + 1); 408 memcpy(dmaru->hdr, header, header->length); 409 dmaru->reg_base_addr = drhd->address; 410 dmaru->segment = drhd->segment; 411 dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */ 412 dmaru->devices = dmar_alloc_dev_scope((void *)(drhd + 1), 413 ((void *)drhd) + drhd->header.length, 414 &dmaru->devices_cnt); 415 if (dmaru->devices_cnt && dmaru->devices == NULL) { 416 kfree(dmaru); 417 return -ENOMEM; 418 } 419 420 ret = alloc_iommu(dmaru); 421 if (ret) { 422 dmar_free_dev_scope(&dmaru->devices, 423 &dmaru->devices_cnt); 424 kfree(dmaru); 425 return ret; 426 } 427 dmar_register_drhd_unit(dmaru); 428 429 out: 430 if (arg) 431 (*(int *)arg)++; 432 433 return 0; 434 } 435 436 static void dmar_free_drhd(struct dmar_drhd_unit *dmaru) 437 { 438 if (dmaru->devices && dmaru->devices_cnt) 439 dmar_free_dev_scope(&dmaru->devices, &dmaru->devices_cnt); 440 if (dmaru->iommu) 441 free_iommu(dmaru->iommu); 442 kfree(dmaru); 443 } 444 445 static int __init dmar_parse_one_andd(struct acpi_dmar_header *header, 446 void *arg) 447 { 448 struct acpi_dmar_andd *andd = (void *)header; 449 450 /* Check for NUL termination within the designated length */ 451 if (strnlen(andd->device_name, header->length - 8) == header->length - 8) { 452 pr_warn(FW_BUG 453 "Your BIOS is broken; ANDD object name is not NUL-terminated\n" 454 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 455 dmi_get_system_info(DMI_BIOS_VENDOR), 456 dmi_get_system_info(DMI_BIOS_VERSION), 457 dmi_get_system_info(DMI_PRODUCT_VERSION)); 458 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 459 return -EINVAL; 460 } 461 pr_info("ANDD device: %x name: %s\n", andd->device_number, 462 andd->device_name); 463 464 return 0; 465 } 466 467 #ifdef CONFIG_ACPI_NUMA 468 static int dmar_parse_one_rhsa(struct acpi_dmar_header *header, void *arg) 469 { 470 struct acpi_dmar_rhsa *rhsa; 471 struct dmar_drhd_unit *drhd; 472 473 rhsa = (struct acpi_dmar_rhsa *)header; 474 for_each_drhd_unit(drhd) { 475 if (drhd->reg_base_addr == rhsa->base_address) { 476 int node = acpi_map_pxm_to_node(rhsa->proximity_domain); 477 478 if (!node_online(node)) 479 node = NUMA_NO_NODE; 480 drhd->iommu->node = node; 481 return 0; 482 } 483 } 484 pr_warn(FW_BUG 485 "Your BIOS is broken; RHSA refers to non-existent DMAR unit at %llx\n" 486 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 487 rhsa->base_address, 488 dmi_get_system_info(DMI_BIOS_VENDOR), 489 dmi_get_system_info(DMI_BIOS_VERSION), 490 dmi_get_system_info(DMI_PRODUCT_VERSION)); 491 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 492 493 return 0; 494 } 495 #else 496 #define dmar_parse_one_rhsa dmar_res_noop 497 #endif 498 499 static void 500 dmar_table_print_dmar_entry(struct acpi_dmar_header *header) 501 { 502 struct acpi_dmar_hardware_unit *drhd; 503 struct acpi_dmar_reserved_memory *rmrr; 504 struct acpi_dmar_atsr *atsr; 505 struct acpi_dmar_rhsa *rhsa; 506 507 switch (header->type) { 508 case ACPI_DMAR_TYPE_HARDWARE_UNIT: 509 drhd = container_of(header, struct acpi_dmar_hardware_unit, 510 header); 511 pr_info("DRHD base: %#016Lx flags: %#x\n", 512 (unsigned long long)drhd->address, drhd->flags); 513 break; 514 case ACPI_DMAR_TYPE_RESERVED_MEMORY: 515 rmrr = container_of(header, struct acpi_dmar_reserved_memory, 516 header); 517 pr_info("RMRR base: %#016Lx end: %#016Lx\n", 518 (unsigned long long)rmrr->base_address, 519 (unsigned long long)rmrr->end_address); 520 break; 521 case ACPI_DMAR_TYPE_ROOT_ATS: 522 atsr = container_of(header, struct acpi_dmar_atsr, header); 523 pr_info("ATSR flags: %#x\n", atsr->flags); 524 break; 525 case ACPI_DMAR_TYPE_HARDWARE_AFFINITY: 526 rhsa = container_of(header, struct acpi_dmar_rhsa, header); 527 pr_info("RHSA base: %#016Lx proximity domain: %#x\n", 528 (unsigned long long)rhsa->base_address, 529 rhsa->proximity_domain); 530 break; 531 case ACPI_DMAR_TYPE_NAMESPACE: 532 /* We don't print this here because we need to sanity-check 533 it first. So print it in dmar_parse_one_andd() instead. */ 534 break; 535 } 536 } 537 538 /** 539 * dmar_table_detect - checks to see if the platform supports DMAR devices 540 */ 541 static int __init dmar_table_detect(void) 542 { 543 acpi_status status = AE_OK; 544 545 /* if we could find DMAR table, then there are DMAR devices */ 546 status = acpi_get_table(ACPI_SIG_DMAR, 0, &dmar_tbl); 547 548 if (ACPI_SUCCESS(status) && !dmar_tbl) { 549 pr_warn("Unable to map DMAR\n"); 550 status = AE_NOT_FOUND; 551 } 552 553 return ACPI_SUCCESS(status) ? 0 : -ENOENT; 554 } 555 556 static int dmar_walk_remapping_entries(struct acpi_dmar_header *start, 557 size_t len, struct dmar_res_callback *cb) 558 { 559 struct acpi_dmar_header *iter, *next; 560 struct acpi_dmar_header *end = ((void *)start) + len; 561 562 for (iter = start; iter < end; iter = next) { 563 next = (void *)iter + iter->length; 564 if (iter->length == 0) { 565 /* Avoid looping forever on bad ACPI tables */ 566 pr_debug(FW_BUG "Invalid 0-length structure\n"); 567 break; 568 } else if (next > end) { 569 /* Avoid passing table end */ 570 pr_warn(FW_BUG "Record passes table end\n"); 571 return -EINVAL; 572 } 573 574 if (cb->print_entry) 575 dmar_table_print_dmar_entry(iter); 576 577 if (iter->type >= ACPI_DMAR_TYPE_RESERVED) { 578 /* continue for forward compatibility */ 579 pr_debug("Unknown DMAR structure type %d\n", 580 iter->type); 581 } else if (cb->cb[iter->type]) { 582 int ret; 583 584 ret = cb->cb[iter->type](iter, cb->arg[iter->type]); 585 if (ret) 586 return ret; 587 } else if (!cb->ignore_unhandled) { 588 pr_warn("No handler for DMAR structure type %d\n", 589 iter->type); 590 return -EINVAL; 591 } 592 } 593 594 return 0; 595 } 596 597 static inline int dmar_walk_dmar_table(struct acpi_table_dmar *dmar, 598 struct dmar_res_callback *cb) 599 { 600 return dmar_walk_remapping_entries((void *)(dmar + 1), 601 dmar->header.length - sizeof(*dmar), cb); 602 } 603 604 /** 605 * parse_dmar_table - parses the DMA reporting table 606 */ 607 static int __init 608 parse_dmar_table(void) 609 { 610 struct acpi_table_dmar *dmar; 611 int drhd_count = 0; 612 int ret; 613 struct dmar_res_callback cb = { 614 .print_entry = true, 615 .ignore_unhandled = true, 616 .arg[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &drhd_count, 617 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_parse_one_drhd, 618 .cb[ACPI_DMAR_TYPE_RESERVED_MEMORY] = &dmar_parse_one_rmrr, 619 .cb[ACPI_DMAR_TYPE_ROOT_ATS] = &dmar_parse_one_atsr, 620 .cb[ACPI_DMAR_TYPE_HARDWARE_AFFINITY] = &dmar_parse_one_rhsa, 621 .cb[ACPI_DMAR_TYPE_NAMESPACE] = &dmar_parse_one_andd, 622 }; 623 624 /* 625 * Do it again, earlier dmar_tbl mapping could be mapped with 626 * fixed map. 627 */ 628 dmar_table_detect(); 629 630 /* 631 * ACPI tables may not be DMA protected by tboot, so use DMAR copy 632 * SINIT saved in SinitMleData in TXT heap (which is DMA protected) 633 */ 634 dmar_tbl = tboot_get_dmar_table(dmar_tbl); 635 636 dmar = (struct acpi_table_dmar *)dmar_tbl; 637 if (!dmar) 638 return -ENODEV; 639 640 if (dmar->width < PAGE_SHIFT - 1) { 641 pr_warn("Invalid DMAR haw\n"); 642 return -EINVAL; 643 } 644 645 pr_info("Host address width %d\n", dmar->width + 1); 646 ret = dmar_walk_dmar_table(dmar, &cb); 647 if (ret == 0 && drhd_count == 0) 648 pr_warn(FW_BUG "No DRHD structure found in DMAR table\n"); 649 650 return ret; 651 } 652 653 static int dmar_pci_device_match(struct dmar_dev_scope devices[], 654 int cnt, struct pci_dev *dev) 655 { 656 int index; 657 struct device *tmp; 658 659 while (dev) { 660 for_each_active_dev_scope(devices, cnt, index, tmp) 661 if (dev_is_pci(tmp) && dev == to_pci_dev(tmp)) 662 return 1; 663 664 /* Check our parent */ 665 dev = dev->bus->self; 666 } 667 668 return 0; 669 } 670 671 struct dmar_drhd_unit * 672 dmar_find_matched_drhd_unit(struct pci_dev *dev) 673 { 674 struct dmar_drhd_unit *dmaru; 675 struct acpi_dmar_hardware_unit *drhd; 676 677 dev = pci_physfn(dev); 678 679 rcu_read_lock(); 680 for_each_drhd_unit(dmaru) { 681 drhd = container_of(dmaru->hdr, 682 struct acpi_dmar_hardware_unit, 683 header); 684 685 if (dmaru->include_all && 686 drhd->segment == pci_domain_nr(dev->bus)) 687 goto out; 688 689 if (dmar_pci_device_match(dmaru->devices, 690 dmaru->devices_cnt, dev)) 691 goto out; 692 } 693 dmaru = NULL; 694 out: 695 rcu_read_unlock(); 696 697 return dmaru; 698 } 699 700 static void __init dmar_acpi_insert_dev_scope(u8 device_number, 701 struct acpi_device *adev) 702 { 703 struct dmar_drhd_unit *dmaru; 704 struct acpi_dmar_hardware_unit *drhd; 705 struct acpi_dmar_device_scope *scope; 706 struct device *tmp; 707 int i; 708 struct acpi_dmar_pci_path *path; 709 710 for_each_drhd_unit(dmaru) { 711 drhd = container_of(dmaru->hdr, 712 struct acpi_dmar_hardware_unit, 713 header); 714 715 for (scope = (void *)(drhd + 1); 716 (unsigned long)scope < ((unsigned long)drhd) + drhd->header.length; 717 scope = ((void *)scope) + scope->length) { 718 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_NAMESPACE) 719 continue; 720 if (scope->enumeration_id != device_number) 721 continue; 722 723 path = (void *)(scope + 1); 724 pr_info("ACPI device \"%s\" under DMAR at %llx as %02x:%02x.%d\n", 725 dev_name(&adev->dev), dmaru->reg_base_addr, 726 scope->bus, path->device, path->function); 727 for_each_dev_scope(dmaru->devices, dmaru->devices_cnt, i, tmp) 728 if (tmp == NULL) { 729 dmaru->devices[i].bus = scope->bus; 730 dmaru->devices[i].devfn = PCI_DEVFN(path->device, 731 path->function); 732 rcu_assign_pointer(dmaru->devices[i].dev, 733 get_device(&adev->dev)); 734 return; 735 } 736 BUG_ON(i >= dmaru->devices_cnt); 737 } 738 } 739 pr_warn("No IOMMU scope found for ANDD enumeration ID %d (%s)\n", 740 device_number, dev_name(&adev->dev)); 741 } 742 743 static int __init dmar_acpi_dev_scope_init(void) 744 { 745 struct acpi_dmar_andd *andd; 746 747 if (dmar_tbl == NULL) 748 return -ENODEV; 749 750 for (andd = (void *)dmar_tbl + sizeof(struct acpi_table_dmar); 751 ((unsigned long)andd) < ((unsigned long)dmar_tbl) + dmar_tbl->length; 752 andd = ((void *)andd) + andd->header.length) { 753 if (andd->header.type == ACPI_DMAR_TYPE_NAMESPACE) { 754 acpi_handle h; 755 struct acpi_device *adev; 756 757 if (!ACPI_SUCCESS(acpi_get_handle(ACPI_ROOT_OBJECT, 758 andd->device_name, 759 &h))) { 760 pr_err("Failed to find handle for ACPI object %s\n", 761 andd->device_name); 762 continue; 763 } 764 if (acpi_bus_get_device(h, &adev)) { 765 pr_err("Failed to get device for ACPI object %s\n", 766 andd->device_name); 767 continue; 768 } 769 dmar_acpi_insert_dev_scope(andd->device_number, adev); 770 } 771 } 772 return 0; 773 } 774 775 int __init dmar_dev_scope_init(void) 776 { 777 struct pci_dev *dev = NULL; 778 struct dmar_pci_notify_info *info; 779 780 if (dmar_dev_scope_status != 1) 781 return dmar_dev_scope_status; 782 783 if (list_empty(&dmar_drhd_units)) { 784 dmar_dev_scope_status = -ENODEV; 785 } else { 786 dmar_dev_scope_status = 0; 787 788 dmar_acpi_dev_scope_init(); 789 790 for_each_pci_dev(dev) { 791 if (dev->is_virtfn) 792 continue; 793 794 info = dmar_alloc_pci_notify_info(dev, 795 BUS_NOTIFY_ADD_DEVICE); 796 if (!info) { 797 return dmar_dev_scope_status; 798 } else { 799 dmar_pci_bus_add_dev(info); 800 dmar_free_pci_notify_info(info); 801 } 802 } 803 } 804 805 return dmar_dev_scope_status; 806 } 807 808 void __init dmar_register_bus_notifier(void) 809 { 810 bus_register_notifier(&pci_bus_type, &dmar_pci_bus_nb); 811 } 812 813 814 int __init dmar_table_init(void) 815 { 816 static int dmar_table_initialized; 817 int ret; 818 819 if (dmar_table_initialized == 0) { 820 ret = parse_dmar_table(); 821 if (ret < 0) { 822 if (ret != -ENODEV) 823 pr_info("Parse DMAR table failure.\n"); 824 } else if (list_empty(&dmar_drhd_units)) { 825 pr_info("No DMAR devices found\n"); 826 ret = -ENODEV; 827 } 828 829 if (ret < 0) 830 dmar_table_initialized = ret; 831 else 832 dmar_table_initialized = 1; 833 } 834 835 return dmar_table_initialized < 0 ? dmar_table_initialized : 0; 836 } 837 838 static void warn_invalid_dmar(u64 addr, const char *message) 839 { 840 pr_warn_once(FW_BUG 841 "Your BIOS is broken; DMAR reported at address %llx%s!\n" 842 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 843 addr, message, 844 dmi_get_system_info(DMI_BIOS_VENDOR), 845 dmi_get_system_info(DMI_BIOS_VERSION), 846 dmi_get_system_info(DMI_PRODUCT_VERSION)); 847 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 848 } 849 850 static int __ref 851 dmar_validate_one_drhd(struct acpi_dmar_header *entry, void *arg) 852 { 853 struct acpi_dmar_hardware_unit *drhd; 854 void __iomem *addr; 855 u64 cap, ecap; 856 857 drhd = (void *)entry; 858 if (!drhd->address) { 859 warn_invalid_dmar(0, ""); 860 return -EINVAL; 861 } 862 863 if (arg) 864 addr = ioremap(drhd->address, VTD_PAGE_SIZE); 865 else 866 addr = early_ioremap(drhd->address, VTD_PAGE_SIZE); 867 if (!addr) { 868 pr_warn("Can't validate DRHD address: %llx\n", drhd->address); 869 return -EINVAL; 870 } 871 872 cap = dmar_readq(addr + DMAR_CAP_REG); 873 ecap = dmar_readq(addr + DMAR_ECAP_REG); 874 875 if (arg) 876 iounmap(addr); 877 else 878 early_iounmap(addr, VTD_PAGE_SIZE); 879 880 if (cap == (uint64_t)-1 && ecap == (uint64_t)-1) { 881 warn_invalid_dmar(drhd->address, " returns all ones"); 882 return -EINVAL; 883 } 884 885 return 0; 886 } 887 888 int __init detect_intel_iommu(void) 889 { 890 int ret; 891 struct dmar_res_callback validate_drhd_cb = { 892 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_validate_one_drhd, 893 .ignore_unhandled = true, 894 }; 895 896 down_write(&dmar_global_lock); 897 ret = dmar_table_detect(); 898 if (!ret) 899 ret = dmar_walk_dmar_table((struct acpi_table_dmar *)dmar_tbl, 900 &validate_drhd_cb); 901 if (!ret && !no_iommu && !iommu_detected && 902 (!dmar_disabled || dmar_platform_optin())) { 903 iommu_detected = 1; 904 /* Make sure ACS will be enabled */ 905 pci_request_acs(); 906 } 907 908 #ifdef CONFIG_X86 909 if (!ret) { 910 x86_init.iommu.iommu_init = intel_iommu_init; 911 x86_platform.iommu_shutdown = intel_iommu_shutdown; 912 } 913 914 #endif 915 916 if (dmar_tbl) { 917 acpi_put_table(dmar_tbl); 918 dmar_tbl = NULL; 919 } 920 up_write(&dmar_global_lock); 921 922 return ret ? ret : 1; 923 } 924 925 static void unmap_iommu(struct intel_iommu *iommu) 926 { 927 iounmap(iommu->reg); 928 release_mem_region(iommu->reg_phys, iommu->reg_size); 929 } 930 931 /** 932 * map_iommu: map the iommu's registers 933 * @iommu: the iommu to map 934 * @phys_addr: the physical address of the base resgister 935 * 936 * Memory map the iommu's registers. Start w/ a single page, and 937 * possibly expand if that turns out to be insufficent. 938 */ 939 static int map_iommu(struct intel_iommu *iommu, u64 phys_addr) 940 { 941 int map_size, err=0; 942 943 iommu->reg_phys = phys_addr; 944 iommu->reg_size = VTD_PAGE_SIZE; 945 946 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, iommu->name)) { 947 pr_err("Can't reserve memory\n"); 948 err = -EBUSY; 949 goto out; 950 } 951 952 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size); 953 if (!iommu->reg) { 954 pr_err("Can't map the region\n"); 955 err = -ENOMEM; 956 goto release; 957 } 958 959 iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG); 960 iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG); 961 962 if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) { 963 err = -EINVAL; 964 warn_invalid_dmar(phys_addr, " returns all ones"); 965 goto unmap; 966 } 967 iommu->vccap = dmar_readq(iommu->reg + DMAR_VCCAP_REG); 968 969 /* the registers might be more than one page */ 970 map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap), 971 cap_max_fault_reg_offset(iommu->cap)); 972 map_size = VTD_PAGE_ALIGN(map_size); 973 if (map_size > iommu->reg_size) { 974 iounmap(iommu->reg); 975 release_mem_region(iommu->reg_phys, iommu->reg_size); 976 iommu->reg_size = map_size; 977 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, 978 iommu->name)) { 979 pr_err("Can't reserve memory\n"); 980 err = -EBUSY; 981 goto out; 982 } 983 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size); 984 if (!iommu->reg) { 985 pr_err("Can't map the region\n"); 986 err = -ENOMEM; 987 goto release; 988 } 989 } 990 err = 0; 991 goto out; 992 993 unmap: 994 iounmap(iommu->reg); 995 release: 996 release_mem_region(iommu->reg_phys, iommu->reg_size); 997 out: 998 return err; 999 } 1000 1001 static int dmar_alloc_seq_id(struct intel_iommu *iommu) 1002 { 1003 iommu->seq_id = find_first_zero_bit(dmar_seq_ids, 1004 DMAR_UNITS_SUPPORTED); 1005 if (iommu->seq_id >= DMAR_UNITS_SUPPORTED) { 1006 iommu->seq_id = -1; 1007 } else { 1008 set_bit(iommu->seq_id, dmar_seq_ids); 1009 sprintf(iommu->name, "dmar%d", iommu->seq_id); 1010 } 1011 1012 return iommu->seq_id; 1013 } 1014 1015 static void dmar_free_seq_id(struct intel_iommu *iommu) 1016 { 1017 if (iommu->seq_id >= 0) { 1018 clear_bit(iommu->seq_id, dmar_seq_ids); 1019 iommu->seq_id = -1; 1020 } 1021 } 1022 1023 static int alloc_iommu(struct dmar_drhd_unit *drhd) 1024 { 1025 struct intel_iommu *iommu; 1026 u32 ver, sts; 1027 int agaw = 0; 1028 int msagaw = 0; 1029 int err; 1030 1031 if (!drhd->reg_base_addr) { 1032 warn_invalid_dmar(0, ""); 1033 return -EINVAL; 1034 } 1035 1036 iommu = kzalloc(sizeof(*iommu), GFP_KERNEL); 1037 if (!iommu) 1038 return -ENOMEM; 1039 1040 if (dmar_alloc_seq_id(iommu) < 0) { 1041 pr_err("Failed to allocate seq_id\n"); 1042 err = -ENOSPC; 1043 goto error; 1044 } 1045 1046 err = map_iommu(iommu, drhd->reg_base_addr); 1047 if (err) { 1048 pr_err("Failed to map %s\n", iommu->name); 1049 goto error_free_seq_id; 1050 } 1051 1052 err = -EINVAL; 1053 agaw = iommu_calculate_agaw(iommu); 1054 if (agaw < 0) { 1055 pr_err("Cannot get a valid agaw for iommu (seq_id = %d)\n", 1056 iommu->seq_id); 1057 goto err_unmap; 1058 } 1059 msagaw = iommu_calculate_max_sagaw(iommu); 1060 if (msagaw < 0) { 1061 pr_err("Cannot get a valid max agaw for iommu (seq_id = %d)\n", 1062 iommu->seq_id); 1063 goto err_unmap; 1064 } 1065 iommu->agaw = agaw; 1066 iommu->msagaw = msagaw; 1067 iommu->segment = drhd->segment; 1068 1069 iommu->node = NUMA_NO_NODE; 1070 1071 ver = readl(iommu->reg + DMAR_VER_REG); 1072 pr_info("%s: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n", 1073 iommu->name, 1074 (unsigned long long)drhd->reg_base_addr, 1075 DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver), 1076 (unsigned long long)iommu->cap, 1077 (unsigned long long)iommu->ecap); 1078 1079 /* Reflect status in gcmd */ 1080 sts = readl(iommu->reg + DMAR_GSTS_REG); 1081 if (sts & DMA_GSTS_IRES) 1082 iommu->gcmd |= DMA_GCMD_IRE; 1083 if (sts & DMA_GSTS_TES) 1084 iommu->gcmd |= DMA_GCMD_TE; 1085 if (sts & DMA_GSTS_QIES) 1086 iommu->gcmd |= DMA_GCMD_QIE; 1087 1088 raw_spin_lock_init(&iommu->register_lock); 1089 1090 if (intel_iommu_enabled) { 1091 err = iommu_device_sysfs_add(&iommu->iommu, NULL, 1092 intel_iommu_groups, 1093 "%s", iommu->name); 1094 if (err) 1095 goto err_unmap; 1096 1097 iommu_device_set_ops(&iommu->iommu, &intel_iommu_ops); 1098 1099 err = iommu_device_register(&iommu->iommu); 1100 if (err) 1101 goto err_unmap; 1102 } 1103 1104 drhd->iommu = iommu; 1105 iommu->drhd = drhd; 1106 1107 return 0; 1108 1109 err_unmap: 1110 unmap_iommu(iommu); 1111 error_free_seq_id: 1112 dmar_free_seq_id(iommu); 1113 error: 1114 kfree(iommu); 1115 return err; 1116 } 1117 1118 static void free_iommu(struct intel_iommu *iommu) 1119 { 1120 if (intel_iommu_enabled) { 1121 iommu_device_unregister(&iommu->iommu); 1122 iommu_device_sysfs_remove(&iommu->iommu); 1123 } 1124 1125 if (iommu->irq) { 1126 if (iommu->pr_irq) { 1127 free_irq(iommu->pr_irq, iommu); 1128 dmar_free_hwirq(iommu->pr_irq); 1129 iommu->pr_irq = 0; 1130 } 1131 free_irq(iommu->irq, iommu); 1132 dmar_free_hwirq(iommu->irq); 1133 iommu->irq = 0; 1134 } 1135 1136 if (iommu->qi) { 1137 free_page((unsigned long)iommu->qi->desc); 1138 kfree(iommu->qi->desc_status); 1139 kfree(iommu->qi); 1140 } 1141 1142 if (iommu->reg) 1143 unmap_iommu(iommu); 1144 1145 dmar_free_seq_id(iommu); 1146 kfree(iommu); 1147 } 1148 1149 /* 1150 * Reclaim all the submitted descriptors which have completed its work. 1151 */ 1152 static inline void reclaim_free_desc(struct q_inval *qi) 1153 { 1154 while (qi->desc_status[qi->free_tail] == QI_DONE || 1155 qi->desc_status[qi->free_tail] == QI_ABORT) { 1156 qi->desc_status[qi->free_tail] = QI_FREE; 1157 qi->free_tail = (qi->free_tail + 1) % QI_LENGTH; 1158 qi->free_cnt++; 1159 } 1160 } 1161 1162 static int qi_check_fault(struct intel_iommu *iommu, int index, int wait_index) 1163 { 1164 u32 fault; 1165 int head, tail; 1166 struct q_inval *qi = iommu->qi; 1167 int shift = qi_shift(iommu); 1168 1169 if (qi->desc_status[wait_index] == QI_ABORT) 1170 return -EAGAIN; 1171 1172 fault = readl(iommu->reg + DMAR_FSTS_REG); 1173 1174 /* 1175 * If IQE happens, the head points to the descriptor associated 1176 * with the error. No new descriptors are fetched until the IQE 1177 * is cleared. 1178 */ 1179 if (fault & DMA_FSTS_IQE) { 1180 head = readl(iommu->reg + DMAR_IQH_REG); 1181 if ((head >> shift) == index) { 1182 struct qi_desc *desc = qi->desc + head; 1183 1184 /* 1185 * desc->qw2 and desc->qw3 are either reserved or 1186 * used by software as private data. We won't print 1187 * out these two qw's for security consideration. 1188 */ 1189 pr_err("VT-d detected invalid descriptor: qw0 = %llx, qw1 = %llx\n", 1190 (unsigned long long)desc->qw0, 1191 (unsigned long long)desc->qw1); 1192 memcpy(desc, qi->desc + (wait_index << shift), 1193 1 << shift); 1194 writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG); 1195 return -EINVAL; 1196 } 1197 } 1198 1199 /* 1200 * If ITE happens, all pending wait_desc commands are aborted. 1201 * No new descriptors are fetched until the ITE is cleared. 1202 */ 1203 if (fault & DMA_FSTS_ITE) { 1204 head = readl(iommu->reg + DMAR_IQH_REG); 1205 head = ((head >> shift) - 1 + QI_LENGTH) % QI_LENGTH; 1206 head |= 1; 1207 tail = readl(iommu->reg + DMAR_IQT_REG); 1208 tail = ((tail >> shift) - 1 + QI_LENGTH) % QI_LENGTH; 1209 1210 writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG); 1211 1212 do { 1213 if (qi->desc_status[head] == QI_IN_USE) 1214 qi->desc_status[head] = QI_ABORT; 1215 head = (head - 2 + QI_LENGTH) % QI_LENGTH; 1216 } while (head != tail); 1217 1218 if (qi->desc_status[wait_index] == QI_ABORT) 1219 return -EAGAIN; 1220 } 1221 1222 if (fault & DMA_FSTS_ICE) 1223 writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG); 1224 1225 return 0; 1226 } 1227 1228 /* 1229 * Function to submit invalidation descriptors of all types to the queued 1230 * invalidation interface(QI). Multiple descriptors can be submitted at a 1231 * time, a wait descriptor will be appended to each submission to ensure 1232 * hardware has completed the invalidation before return. Wait descriptors 1233 * can be part of the submission but it will not be polled for completion. 1234 */ 1235 int qi_submit_sync(struct intel_iommu *iommu, struct qi_desc *desc, 1236 unsigned int count, unsigned long options) 1237 { 1238 struct q_inval *qi = iommu->qi; 1239 struct qi_desc wait_desc; 1240 int wait_index, index; 1241 unsigned long flags; 1242 int offset, shift; 1243 int rc, i; 1244 1245 if (!qi) 1246 return 0; 1247 1248 restart: 1249 rc = 0; 1250 1251 raw_spin_lock_irqsave(&qi->q_lock, flags); 1252 /* 1253 * Check if we have enough empty slots in the queue to submit, 1254 * the calculation is based on: 1255 * # of desc + 1 wait desc + 1 space between head and tail 1256 */ 1257 while (qi->free_cnt < count + 2) { 1258 raw_spin_unlock_irqrestore(&qi->q_lock, flags); 1259 cpu_relax(); 1260 raw_spin_lock_irqsave(&qi->q_lock, flags); 1261 } 1262 1263 index = qi->free_head; 1264 wait_index = (index + count) % QI_LENGTH; 1265 shift = qi_shift(iommu); 1266 1267 for (i = 0; i < count; i++) { 1268 offset = ((index + i) % QI_LENGTH) << shift; 1269 memcpy(qi->desc + offset, &desc[i], 1 << shift); 1270 qi->desc_status[(index + i) % QI_LENGTH] = QI_IN_USE; 1271 } 1272 qi->desc_status[wait_index] = QI_IN_USE; 1273 1274 wait_desc.qw0 = QI_IWD_STATUS_DATA(QI_DONE) | 1275 QI_IWD_STATUS_WRITE | QI_IWD_TYPE; 1276 if (options & QI_OPT_WAIT_DRAIN) 1277 wait_desc.qw0 |= QI_IWD_PRQ_DRAIN; 1278 wait_desc.qw1 = virt_to_phys(&qi->desc_status[wait_index]); 1279 wait_desc.qw2 = 0; 1280 wait_desc.qw3 = 0; 1281 1282 offset = wait_index << shift; 1283 memcpy(qi->desc + offset, &wait_desc, 1 << shift); 1284 1285 qi->free_head = (qi->free_head + count + 1) % QI_LENGTH; 1286 qi->free_cnt -= count + 1; 1287 1288 /* 1289 * update the HW tail register indicating the presence of 1290 * new descriptors. 1291 */ 1292 writel(qi->free_head << shift, iommu->reg + DMAR_IQT_REG); 1293 1294 while (qi->desc_status[wait_index] != QI_DONE) { 1295 /* 1296 * We will leave the interrupts disabled, to prevent interrupt 1297 * context to queue another cmd while a cmd is already submitted 1298 * and waiting for completion on this cpu. This is to avoid 1299 * a deadlock where the interrupt context can wait indefinitely 1300 * for free slots in the queue. 1301 */ 1302 rc = qi_check_fault(iommu, index, wait_index); 1303 if (rc) 1304 break; 1305 1306 raw_spin_unlock(&qi->q_lock); 1307 cpu_relax(); 1308 raw_spin_lock(&qi->q_lock); 1309 } 1310 1311 for (i = 0; i < count; i++) 1312 qi->desc_status[(index + i) % QI_LENGTH] = QI_DONE; 1313 1314 reclaim_free_desc(qi); 1315 raw_spin_unlock_irqrestore(&qi->q_lock, flags); 1316 1317 if (rc == -EAGAIN) 1318 goto restart; 1319 1320 return rc; 1321 } 1322 1323 /* 1324 * Flush the global interrupt entry cache. 1325 */ 1326 void qi_global_iec(struct intel_iommu *iommu) 1327 { 1328 struct qi_desc desc; 1329 1330 desc.qw0 = QI_IEC_TYPE; 1331 desc.qw1 = 0; 1332 desc.qw2 = 0; 1333 desc.qw3 = 0; 1334 1335 /* should never fail */ 1336 qi_submit_sync(iommu, &desc, 1, 0); 1337 } 1338 1339 void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm, 1340 u64 type) 1341 { 1342 struct qi_desc desc; 1343 1344 desc.qw0 = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did) 1345 | QI_CC_GRAN(type) | QI_CC_TYPE; 1346 desc.qw1 = 0; 1347 desc.qw2 = 0; 1348 desc.qw3 = 0; 1349 1350 qi_submit_sync(iommu, &desc, 1, 0); 1351 } 1352 1353 void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr, 1354 unsigned int size_order, u64 type) 1355 { 1356 u8 dw = 0, dr = 0; 1357 1358 struct qi_desc desc; 1359 int ih = 0; 1360 1361 if (cap_write_drain(iommu->cap)) 1362 dw = 1; 1363 1364 if (cap_read_drain(iommu->cap)) 1365 dr = 1; 1366 1367 desc.qw0 = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw) 1368 | QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE; 1369 desc.qw1 = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih) 1370 | QI_IOTLB_AM(size_order); 1371 desc.qw2 = 0; 1372 desc.qw3 = 0; 1373 1374 qi_submit_sync(iommu, &desc, 1, 0); 1375 } 1376 1377 void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 pfsid, 1378 u16 qdep, u64 addr, unsigned mask) 1379 { 1380 struct qi_desc desc; 1381 1382 if (mask) { 1383 addr |= (1ULL << (VTD_PAGE_SHIFT + mask - 1)) - 1; 1384 desc.qw1 = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE; 1385 } else 1386 desc.qw1 = QI_DEV_IOTLB_ADDR(addr); 1387 1388 if (qdep >= QI_DEV_IOTLB_MAX_INVS) 1389 qdep = 0; 1390 1391 desc.qw0 = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) | 1392 QI_DIOTLB_TYPE | QI_DEV_IOTLB_PFSID(pfsid); 1393 desc.qw2 = 0; 1394 desc.qw3 = 0; 1395 1396 qi_submit_sync(iommu, &desc, 1, 0); 1397 } 1398 1399 /* PASID-based IOTLB invalidation */ 1400 void qi_flush_piotlb(struct intel_iommu *iommu, u16 did, u32 pasid, u64 addr, 1401 unsigned long npages, bool ih) 1402 { 1403 struct qi_desc desc = {.qw2 = 0, .qw3 = 0}; 1404 1405 /* 1406 * npages == -1 means a PASID-selective invalidation, otherwise, 1407 * a positive value for Page-selective-within-PASID invalidation. 1408 * 0 is not a valid input. 1409 */ 1410 if (WARN_ON(!npages)) { 1411 pr_err("Invalid input npages = %ld\n", npages); 1412 return; 1413 } 1414 1415 if (npages == -1) { 1416 desc.qw0 = QI_EIOTLB_PASID(pasid) | 1417 QI_EIOTLB_DID(did) | 1418 QI_EIOTLB_GRAN(QI_GRAN_NONG_PASID) | 1419 QI_EIOTLB_TYPE; 1420 desc.qw1 = 0; 1421 } else { 1422 int mask = ilog2(__roundup_pow_of_two(npages)); 1423 unsigned long align = (1ULL << (VTD_PAGE_SHIFT + mask)); 1424 1425 if (WARN_ON_ONCE(!ALIGN(addr, align))) 1426 addr &= ~(align - 1); 1427 1428 desc.qw0 = QI_EIOTLB_PASID(pasid) | 1429 QI_EIOTLB_DID(did) | 1430 QI_EIOTLB_GRAN(QI_GRAN_PSI_PASID) | 1431 QI_EIOTLB_TYPE; 1432 desc.qw1 = QI_EIOTLB_ADDR(addr) | 1433 QI_EIOTLB_IH(ih) | 1434 QI_EIOTLB_AM(mask); 1435 } 1436 1437 qi_submit_sync(iommu, &desc, 1, 0); 1438 } 1439 1440 /* PASID-based device IOTLB Invalidate */ 1441 void qi_flush_dev_iotlb_pasid(struct intel_iommu *iommu, u16 sid, u16 pfsid, 1442 u32 pasid, u16 qdep, u64 addr, unsigned int size_order) 1443 { 1444 unsigned long mask = 1UL << (VTD_PAGE_SHIFT + size_order - 1); 1445 struct qi_desc desc = {.qw1 = 0, .qw2 = 0, .qw3 = 0}; 1446 1447 desc.qw0 = QI_DEV_EIOTLB_PASID(pasid) | QI_DEV_EIOTLB_SID(sid) | 1448 QI_DEV_EIOTLB_QDEP(qdep) | QI_DEIOTLB_TYPE | 1449 QI_DEV_IOTLB_PFSID(pfsid); 1450 1451 /* 1452 * If S bit is 0, we only flush a single page. If S bit is set, 1453 * The least significant zero bit indicates the invalidation address 1454 * range. VT-d spec 6.5.2.6. 1455 * e.g. address bit 12[0] indicates 8KB, 13[0] indicates 16KB. 1456 * size order = 0 is PAGE_SIZE 4KB 1457 * Max Invs Pending (MIP) is set to 0 for now until we have DIT in 1458 * ECAP. 1459 */ 1460 if (addr & GENMASK_ULL(size_order + VTD_PAGE_SHIFT, 0)) 1461 pr_warn_ratelimited("Invalidate non-aligned address %llx, order %d\n", 1462 addr, size_order); 1463 1464 /* Take page address */ 1465 desc.qw1 = QI_DEV_EIOTLB_ADDR(addr); 1466 1467 if (size_order) { 1468 /* 1469 * Existing 0s in address below size_order may be the least 1470 * significant bit, we must set them to 1s to avoid having 1471 * smaller size than desired. 1472 */ 1473 desc.qw1 |= GENMASK_ULL(size_order + VTD_PAGE_SHIFT - 1, 1474 VTD_PAGE_SHIFT); 1475 /* Clear size_order bit to indicate size */ 1476 desc.qw1 &= ~mask; 1477 /* Set the S bit to indicate flushing more than 1 page */ 1478 desc.qw1 |= QI_DEV_EIOTLB_SIZE; 1479 } 1480 1481 qi_submit_sync(iommu, &desc, 1, 0); 1482 } 1483 1484 void qi_flush_pasid_cache(struct intel_iommu *iommu, u16 did, 1485 u64 granu, int pasid) 1486 { 1487 struct qi_desc desc = {.qw1 = 0, .qw2 = 0, .qw3 = 0}; 1488 1489 desc.qw0 = QI_PC_PASID(pasid) | QI_PC_DID(did) | 1490 QI_PC_GRAN(granu) | QI_PC_TYPE; 1491 qi_submit_sync(iommu, &desc, 1, 0); 1492 } 1493 1494 /* 1495 * Disable Queued Invalidation interface. 1496 */ 1497 void dmar_disable_qi(struct intel_iommu *iommu) 1498 { 1499 unsigned long flags; 1500 u32 sts; 1501 cycles_t start_time = get_cycles(); 1502 1503 if (!ecap_qis(iommu->ecap)) 1504 return; 1505 1506 raw_spin_lock_irqsave(&iommu->register_lock, flags); 1507 1508 sts = readl(iommu->reg + DMAR_GSTS_REG); 1509 if (!(sts & DMA_GSTS_QIES)) 1510 goto end; 1511 1512 /* 1513 * Give a chance to HW to complete the pending invalidation requests. 1514 */ 1515 while ((readl(iommu->reg + DMAR_IQT_REG) != 1516 readl(iommu->reg + DMAR_IQH_REG)) && 1517 (DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time))) 1518 cpu_relax(); 1519 1520 iommu->gcmd &= ~DMA_GCMD_QIE; 1521 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG); 1522 1523 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, 1524 !(sts & DMA_GSTS_QIES), sts); 1525 end: 1526 raw_spin_unlock_irqrestore(&iommu->register_lock, flags); 1527 } 1528 1529 /* 1530 * Enable queued invalidation. 1531 */ 1532 static void __dmar_enable_qi(struct intel_iommu *iommu) 1533 { 1534 u32 sts; 1535 unsigned long flags; 1536 struct q_inval *qi = iommu->qi; 1537 u64 val = virt_to_phys(qi->desc); 1538 1539 qi->free_head = qi->free_tail = 0; 1540 qi->free_cnt = QI_LENGTH; 1541 1542 /* 1543 * Set DW=1 and QS=1 in IQA_REG when Scalable Mode capability 1544 * is present. 1545 */ 1546 if (ecap_smts(iommu->ecap)) 1547 val |= (1 << 11) | 1; 1548 1549 raw_spin_lock_irqsave(&iommu->register_lock, flags); 1550 1551 /* write zero to the tail reg */ 1552 writel(0, iommu->reg + DMAR_IQT_REG); 1553 1554 dmar_writeq(iommu->reg + DMAR_IQA_REG, val); 1555 1556 iommu->gcmd |= DMA_GCMD_QIE; 1557 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG); 1558 1559 /* Make sure hardware complete it */ 1560 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts); 1561 1562 raw_spin_unlock_irqrestore(&iommu->register_lock, flags); 1563 } 1564 1565 /* 1566 * Enable Queued Invalidation interface. This is a must to support 1567 * interrupt-remapping. Also used by DMA-remapping, which replaces 1568 * register based IOTLB invalidation. 1569 */ 1570 int dmar_enable_qi(struct intel_iommu *iommu) 1571 { 1572 struct q_inval *qi; 1573 struct page *desc_page; 1574 1575 if (!ecap_qis(iommu->ecap)) 1576 return -ENOENT; 1577 1578 /* 1579 * queued invalidation is already setup and enabled. 1580 */ 1581 if (iommu->qi) 1582 return 0; 1583 1584 iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC); 1585 if (!iommu->qi) 1586 return -ENOMEM; 1587 1588 qi = iommu->qi; 1589 1590 /* 1591 * Need two pages to accommodate 256 descriptors of 256 bits each 1592 * if the remapping hardware supports scalable mode translation. 1593 */ 1594 desc_page = alloc_pages_node(iommu->node, GFP_ATOMIC | __GFP_ZERO, 1595 !!ecap_smts(iommu->ecap)); 1596 if (!desc_page) { 1597 kfree(qi); 1598 iommu->qi = NULL; 1599 return -ENOMEM; 1600 } 1601 1602 qi->desc = page_address(desc_page); 1603 1604 qi->desc_status = kcalloc(QI_LENGTH, sizeof(int), GFP_ATOMIC); 1605 if (!qi->desc_status) { 1606 free_page((unsigned long) qi->desc); 1607 kfree(qi); 1608 iommu->qi = NULL; 1609 return -ENOMEM; 1610 } 1611 1612 raw_spin_lock_init(&qi->q_lock); 1613 1614 __dmar_enable_qi(iommu); 1615 1616 return 0; 1617 } 1618 1619 /* iommu interrupt handling. Most stuff are MSI-like. */ 1620 1621 enum faulttype { 1622 DMA_REMAP, 1623 INTR_REMAP, 1624 UNKNOWN, 1625 }; 1626 1627 static const char *dma_remap_fault_reasons[] = 1628 { 1629 "Software", 1630 "Present bit in root entry is clear", 1631 "Present bit in context entry is clear", 1632 "Invalid context entry", 1633 "Access beyond MGAW", 1634 "PTE Write access is not set", 1635 "PTE Read access is not set", 1636 "Next page table ptr is invalid", 1637 "Root table address invalid", 1638 "Context table ptr is invalid", 1639 "non-zero reserved fields in RTP", 1640 "non-zero reserved fields in CTP", 1641 "non-zero reserved fields in PTE", 1642 "PCE for translation request specifies blocking", 1643 }; 1644 1645 static const char * const dma_remap_sm_fault_reasons[] = { 1646 "SM: Invalid Root Table Address", 1647 "SM: TTM 0 for request with PASID", 1648 "SM: TTM 0 for page group request", 1649 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x33-0x37 */ 1650 "SM: Error attempting to access Root Entry", 1651 "SM: Present bit in Root Entry is clear", 1652 "SM: Non-zero reserved field set in Root Entry", 1653 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x3B-0x3F */ 1654 "SM: Error attempting to access Context Entry", 1655 "SM: Present bit in Context Entry is clear", 1656 "SM: Non-zero reserved field set in the Context Entry", 1657 "SM: Invalid Context Entry", 1658 "SM: DTE field in Context Entry is clear", 1659 "SM: PASID Enable field in Context Entry is clear", 1660 "SM: PASID is larger than the max in Context Entry", 1661 "SM: PRE field in Context-Entry is clear", 1662 "SM: RID_PASID field error in Context-Entry", 1663 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x49-0x4F */ 1664 "SM: Error attempting to access the PASID Directory Entry", 1665 "SM: Present bit in Directory Entry is clear", 1666 "SM: Non-zero reserved field set in PASID Directory Entry", 1667 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x53-0x57 */ 1668 "SM: Error attempting to access PASID Table Entry", 1669 "SM: Present bit in PASID Table Entry is clear", 1670 "SM: Non-zero reserved field set in PASID Table Entry", 1671 "SM: Invalid Scalable-Mode PASID Table Entry", 1672 "SM: ERE field is clear in PASID Table Entry", 1673 "SM: SRE field is clear in PASID Table Entry", 1674 "Unknown", "Unknown",/* 0x5E-0x5F */ 1675 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x60-0x67 */ 1676 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x68-0x6F */ 1677 "SM: Error attempting to access first-level paging entry", 1678 "SM: Present bit in first-level paging entry is clear", 1679 "SM: Non-zero reserved field set in first-level paging entry", 1680 "SM: Error attempting to access FL-PML4 entry", 1681 "SM: First-level entry address beyond MGAW in Nested translation", 1682 "SM: Read permission error in FL-PML4 entry in Nested translation", 1683 "SM: Read permission error in first-level paging entry in Nested translation", 1684 "SM: Write permission error in first-level paging entry in Nested translation", 1685 "SM: Error attempting to access second-level paging entry", 1686 "SM: Read/Write permission error in second-level paging entry", 1687 "SM: Non-zero reserved field set in second-level paging entry", 1688 "SM: Invalid second-level page table pointer", 1689 "SM: A/D bit update needed in second-level entry when set up in no snoop", 1690 "Unknown", "Unknown", "Unknown", /* 0x7D-0x7F */ 1691 "SM: Address in first-level translation is not canonical", 1692 "SM: U/S set 0 for first-level translation with user privilege", 1693 "SM: No execute permission for request with PASID and ER=1", 1694 "SM: Address beyond the DMA hardware max", 1695 "SM: Second-level entry address beyond the max", 1696 "SM: No write permission for Write/AtomicOp request", 1697 "SM: No read permission for Read/AtomicOp request", 1698 "SM: Invalid address-interrupt address", 1699 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x88-0x8F */ 1700 "SM: A/D bit update needed in first-level entry when set up in no snoop", 1701 }; 1702 1703 static const char *irq_remap_fault_reasons[] = 1704 { 1705 "Detected reserved fields in the decoded interrupt-remapped request", 1706 "Interrupt index exceeded the interrupt-remapping table size", 1707 "Present field in the IRTE entry is clear", 1708 "Error accessing interrupt-remapping table pointed by IRTA_REG", 1709 "Detected reserved fields in the IRTE entry", 1710 "Blocked a compatibility format interrupt request", 1711 "Blocked an interrupt request due to source-id verification failure", 1712 }; 1713 1714 static const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type) 1715 { 1716 if (fault_reason >= 0x20 && (fault_reason - 0x20 < 1717 ARRAY_SIZE(irq_remap_fault_reasons))) { 1718 *fault_type = INTR_REMAP; 1719 return irq_remap_fault_reasons[fault_reason - 0x20]; 1720 } else if (fault_reason >= 0x30 && (fault_reason - 0x30 < 1721 ARRAY_SIZE(dma_remap_sm_fault_reasons))) { 1722 *fault_type = DMA_REMAP; 1723 return dma_remap_sm_fault_reasons[fault_reason - 0x30]; 1724 } else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) { 1725 *fault_type = DMA_REMAP; 1726 return dma_remap_fault_reasons[fault_reason]; 1727 } else { 1728 *fault_type = UNKNOWN; 1729 return "Unknown"; 1730 } 1731 } 1732 1733 1734 static inline int dmar_msi_reg(struct intel_iommu *iommu, int irq) 1735 { 1736 if (iommu->irq == irq) 1737 return DMAR_FECTL_REG; 1738 else if (iommu->pr_irq == irq) 1739 return DMAR_PECTL_REG; 1740 else 1741 BUG(); 1742 } 1743 1744 void dmar_msi_unmask(struct irq_data *data) 1745 { 1746 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data); 1747 int reg = dmar_msi_reg(iommu, data->irq); 1748 unsigned long flag; 1749 1750 /* unmask it */ 1751 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1752 writel(0, iommu->reg + reg); 1753 /* Read a reg to force flush the post write */ 1754 readl(iommu->reg + reg); 1755 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1756 } 1757 1758 void dmar_msi_mask(struct irq_data *data) 1759 { 1760 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data); 1761 int reg = dmar_msi_reg(iommu, data->irq); 1762 unsigned long flag; 1763 1764 /* mask it */ 1765 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1766 writel(DMA_FECTL_IM, iommu->reg + reg); 1767 /* Read a reg to force flush the post write */ 1768 readl(iommu->reg + reg); 1769 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1770 } 1771 1772 void dmar_msi_write(int irq, struct msi_msg *msg) 1773 { 1774 struct intel_iommu *iommu = irq_get_handler_data(irq); 1775 int reg = dmar_msi_reg(iommu, irq); 1776 unsigned long flag; 1777 1778 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1779 writel(msg->data, iommu->reg + reg + 4); 1780 writel(msg->address_lo, iommu->reg + reg + 8); 1781 writel(msg->address_hi, iommu->reg + reg + 12); 1782 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1783 } 1784 1785 void dmar_msi_read(int irq, struct msi_msg *msg) 1786 { 1787 struct intel_iommu *iommu = irq_get_handler_data(irq); 1788 int reg = dmar_msi_reg(iommu, irq); 1789 unsigned long flag; 1790 1791 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1792 msg->data = readl(iommu->reg + reg + 4); 1793 msg->address_lo = readl(iommu->reg + reg + 8); 1794 msg->address_hi = readl(iommu->reg + reg + 12); 1795 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1796 } 1797 1798 static int dmar_fault_do_one(struct intel_iommu *iommu, int type, 1799 u8 fault_reason, int pasid, u16 source_id, 1800 unsigned long long addr) 1801 { 1802 const char *reason; 1803 int fault_type; 1804 1805 reason = dmar_get_fault_reason(fault_reason, &fault_type); 1806 1807 if (fault_type == INTR_REMAP) 1808 pr_err("[INTR-REMAP] Request device [%02x:%02x.%d] fault index %llx [fault reason %02d] %s\n", 1809 source_id >> 8, PCI_SLOT(source_id & 0xFF), 1810 PCI_FUNC(source_id & 0xFF), addr >> 48, 1811 fault_reason, reason); 1812 else 1813 pr_err("[%s] Request device [%02x:%02x.%d] PASID %x fault addr %llx [fault reason %02d] %s\n", 1814 type ? "DMA Read" : "DMA Write", 1815 source_id >> 8, PCI_SLOT(source_id & 0xFF), 1816 PCI_FUNC(source_id & 0xFF), pasid, addr, 1817 fault_reason, reason); 1818 return 0; 1819 } 1820 1821 #define PRIMARY_FAULT_REG_LEN (16) 1822 irqreturn_t dmar_fault(int irq, void *dev_id) 1823 { 1824 struct intel_iommu *iommu = dev_id; 1825 int reg, fault_index; 1826 u32 fault_status; 1827 unsigned long flag; 1828 static DEFINE_RATELIMIT_STATE(rs, 1829 DEFAULT_RATELIMIT_INTERVAL, 1830 DEFAULT_RATELIMIT_BURST); 1831 1832 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1833 fault_status = readl(iommu->reg + DMAR_FSTS_REG); 1834 if (fault_status && __ratelimit(&rs)) 1835 pr_err("DRHD: handling fault status reg %x\n", fault_status); 1836 1837 /* TBD: ignore advanced fault log currently */ 1838 if (!(fault_status & DMA_FSTS_PPF)) 1839 goto unlock_exit; 1840 1841 fault_index = dma_fsts_fault_record_index(fault_status); 1842 reg = cap_fault_reg_offset(iommu->cap); 1843 while (1) { 1844 /* Disable printing, simply clear the fault when ratelimited */ 1845 bool ratelimited = !__ratelimit(&rs); 1846 u8 fault_reason; 1847 u16 source_id; 1848 u64 guest_addr; 1849 int type, pasid; 1850 u32 data; 1851 bool pasid_present; 1852 1853 /* highest 32 bits */ 1854 data = readl(iommu->reg + reg + 1855 fault_index * PRIMARY_FAULT_REG_LEN + 12); 1856 if (!(data & DMA_FRCD_F)) 1857 break; 1858 1859 if (!ratelimited) { 1860 fault_reason = dma_frcd_fault_reason(data); 1861 type = dma_frcd_type(data); 1862 1863 pasid = dma_frcd_pasid_value(data); 1864 data = readl(iommu->reg + reg + 1865 fault_index * PRIMARY_FAULT_REG_LEN + 8); 1866 source_id = dma_frcd_source_id(data); 1867 1868 pasid_present = dma_frcd_pasid_present(data); 1869 guest_addr = dmar_readq(iommu->reg + reg + 1870 fault_index * PRIMARY_FAULT_REG_LEN); 1871 guest_addr = dma_frcd_page_addr(guest_addr); 1872 } 1873 1874 /* clear the fault */ 1875 writel(DMA_FRCD_F, iommu->reg + reg + 1876 fault_index * PRIMARY_FAULT_REG_LEN + 12); 1877 1878 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1879 1880 if (!ratelimited) 1881 /* Using pasid -1 if pasid is not present */ 1882 dmar_fault_do_one(iommu, type, fault_reason, 1883 pasid_present ? pasid : -1, 1884 source_id, guest_addr); 1885 1886 fault_index++; 1887 if (fault_index >= cap_num_fault_regs(iommu->cap)) 1888 fault_index = 0; 1889 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1890 } 1891 1892 writel(DMA_FSTS_PFO | DMA_FSTS_PPF | DMA_FSTS_PRO, 1893 iommu->reg + DMAR_FSTS_REG); 1894 1895 unlock_exit: 1896 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1897 return IRQ_HANDLED; 1898 } 1899 1900 int dmar_set_interrupt(struct intel_iommu *iommu) 1901 { 1902 int irq, ret; 1903 1904 /* 1905 * Check if the fault interrupt is already initialized. 1906 */ 1907 if (iommu->irq) 1908 return 0; 1909 1910 irq = dmar_alloc_hwirq(iommu->seq_id, iommu->node, iommu); 1911 if (irq > 0) { 1912 iommu->irq = irq; 1913 } else { 1914 pr_err("No free IRQ vectors\n"); 1915 return -EINVAL; 1916 } 1917 1918 ret = request_irq(irq, dmar_fault, IRQF_NO_THREAD, iommu->name, iommu); 1919 if (ret) 1920 pr_err("Can't request irq\n"); 1921 return ret; 1922 } 1923 1924 int __init enable_drhd_fault_handling(void) 1925 { 1926 struct dmar_drhd_unit *drhd; 1927 struct intel_iommu *iommu; 1928 1929 /* 1930 * Enable fault control interrupt. 1931 */ 1932 for_each_iommu(iommu, drhd) { 1933 u32 fault_status; 1934 int ret = dmar_set_interrupt(iommu); 1935 1936 if (ret) { 1937 pr_err("DRHD %Lx: failed to enable fault, interrupt, ret %d\n", 1938 (unsigned long long)drhd->reg_base_addr, ret); 1939 return -1; 1940 } 1941 1942 /* 1943 * Clear any previous faults. 1944 */ 1945 dmar_fault(iommu->irq, iommu); 1946 fault_status = readl(iommu->reg + DMAR_FSTS_REG); 1947 writel(fault_status, iommu->reg + DMAR_FSTS_REG); 1948 } 1949 1950 return 0; 1951 } 1952 1953 /* 1954 * Re-enable Queued Invalidation interface. 1955 */ 1956 int dmar_reenable_qi(struct intel_iommu *iommu) 1957 { 1958 if (!ecap_qis(iommu->ecap)) 1959 return -ENOENT; 1960 1961 if (!iommu->qi) 1962 return -ENOENT; 1963 1964 /* 1965 * First disable queued invalidation. 1966 */ 1967 dmar_disable_qi(iommu); 1968 /* 1969 * Then enable queued invalidation again. Since there is no pending 1970 * invalidation requests now, it's safe to re-enable queued 1971 * invalidation. 1972 */ 1973 __dmar_enable_qi(iommu); 1974 1975 return 0; 1976 } 1977 1978 /* 1979 * Check interrupt remapping support in DMAR table description. 1980 */ 1981 int __init dmar_ir_support(void) 1982 { 1983 struct acpi_table_dmar *dmar; 1984 dmar = (struct acpi_table_dmar *)dmar_tbl; 1985 if (!dmar) 1986 return 0; 1987 return dmar->flags & 0x1; 1988 } 1989 1990 /* Check whether DMAR units are in use */ 1991 static inline bool dmar_in_use(void) 1992 { 1993 return irq_remapping_enabled || intel_iommu_enabled; 1994 } 1995 1996 static int __init dmar_free_unused_resources(void) 1997 { 1998 struct dmar_drhd_unit *dmaru, *dmaru_n; 1999 2000 if (dmar_in_use()) 2001 return 0; 2002 2003 if (dmar_dev_scope_status != 1 && !list_empty(&dmar_drhd_units)) 2004 bus_unregister_notifier(&pci_bus_type, &dmar_pci_bus_nb); 2005 2006 down_write(&dmar_global_lock); 2007 list_for_each_entry_safe(dmaru, dmaru_n, &dmar_drhd_units, list) { 2008 list_del(&dmaru->list); 2009 dmar_free_drhd(dmaru); 2010 } 2011 up_write(&dmar_global_lock); 2012 2013 return 0; 2014 } 2015 2016 late_initcall(dmar_free_unused_resources); 2017 IOMMU_INIT_POST(detect_intel_iommu); 2018 2019 /* 2020 * DMAR Hotplug Support 2021 * For more details, please refer to Intel(R) Virtualization Technology 2022 * for Directed-IO Architecture Specifiction, Rev 2.2, Section 8.8 2023 * "Remapping Hardware Unit Hot Plug". 2024 */ 2025 static guid_t dmar_hp_guid = 2026 GUID_INIT(0xD8C1A3A6, 0xBE9B, 0x4C9B, 2027 0x91, 0xBF, 0xC3, 0xCB, 0x81, 0xFC, 0x5D, 0xAF); 2028 2029 /* 2030 * Currently there's only one revision and BIOS will not check the revision id, 2031 * so use 0 for safety. 2032 */ 2033 #define DMAR_DSM_REV_ID 0 2034 #define DMAR_DSM_FUNC_DRHD 1 2035 #define DMAR_DSM_FUNC_ATSR 2 2036 #define DMAR_DSM_FUNC_RHSA 3 2037 2038 static inline bool dmar_detect_dsm(acpi_handle handle, int func) 2039 { 2040 return acpi_check_dsm(handle, &dmar_hp_guid, DMAR_DSM_REV_ID, 1 << func); 2041 } 2042 2043 static int dmar_walk_dsm_resource(acpi_handle handle, int func, 2044 dmar_res_handler_t handler, void *arg) 2045 { 2046 int ret = -ENODEV; 2047 union acpi_object *obj; 2048 struct acpi_dmar_header *start; 2049 struct dmar_res_callback callback; 2050 static int res_type[] = { 2051 [DMAR_DSM_FUNC_DRHD] = ACPI_DMAR_TYPE_HARDWARE_UNIT, 2052 [DMAR_DSM_FUNC_ATSR] = ACPI_DMAR_TYPE_ROOT_ATS, 2053 [DMAR_DSM_FUNC_RHSA] = ACPI_DMAR_TYPE_HARDWARE_AFFINITY, 2054 }; 2055 2056 if (!dmar_detect_dsm(handle, func)) 2057 return 0; 2058 2059 obj = acpi_evaluate_dsm_typed(handle, &dmar_hp_guid, DMAR_DSM_REV_ID, 2060 func, NULL, ACPI_TYPE_BUFFER); 2061 if (!obj) 2062 return -ENODEV; 2063 2064 memset(&callback, 0, sizeof(callback)); 2065 callback.cb[res_type[func]] = handler; 2066 callback.arg[res_type[func]] = arg; 2067 start = (struct acpi_dmar_header *)obj->buffer.pointer; 2068 ret = dmar_walk_remapping_entries(start, obj->buffer.length, &callback); 2069 2070 ACPI_FREE(obj); 2071 2072 return ret; 2073 } 2074 2075 static int dmar_hp_add_drhd(struct acpi_dmar_header *header, void *arg) 2076 { 2077 int ret; 2078 struct dmar_drhd_unit *dmaru; 2079 2080 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2081 if (!dmaru) 2082 return -ENODEV; 2083 2084 ret = dmar_ir_hotplug(dmaru, true); 2085 if (ret == 0) 2086 ret = dmar_iommu_hotplug(dmaru, true); 2087 2088 return ret; 2089 } 2090 2091 static int dmar_hp_remove_drhd(struct acpi_dmar_header *header, void *arg) 2092 { 2093 int i, ret; 2094 struct device *dev; 2095 struct dmar_drhd_unit *dmaru; 2096 2097 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2098 if (!dmaru) 2099 return 0; 2100 2101 /* 2102 * All PCI devices managed by this unit should have been destroyed. 2103 */ 2104 if (!dmaru->include_all && dmaru->devices && dmaru->devices_cnt) { 2105 for_each_active_dev_scope(dmaru->devices, 2106 dmaru->devices_cnt, i, dev) 2107 return -EBUSY; 2108 } 2109 2110 ret = dmar_ir_hotplug(dmaru, false); 2111 if (ret == 0) 2112 ret = dmar_iommu_hotplug(dmaru, false); 2113 2114 return ret; 2115 } 2116 2117 static int dmar_hp_release_drhd(struct acpi_dmar_header *header, void *arg) 2118 { 2119 struct dmar_drhd_unit *dmaru; 2120 2121 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2122 if (dmaru) { 2123 list_del_rcu(&dmaru->list); 2124 synchronize_rcu(); 2125 dmar_free_drhd(dmaru); 2126 } 2127 2128 return 0; 2129 } 2130 2131 static int dmar_hotplug_insert(acpi_handle handle) 2132 { 2133 int ret; 2134 int drhd_count = 0; 2135 2136 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2137 &dmar_validate_one_drhd, (void *)1); 2138 if (ret) 2139 goto out; 2140 2141 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2142 &dmar_parse_one_drhd, (void *)&drhd_count); 2143 if (ret == 0 && drhd_count == 0) { 2144 pr_warn(FW_BUG "No DRHD structures in buffer returned by _DSM method\n"); 2145 goto out; 2146 } else if (ret) { 2147 goto release_drhd; 2148 } 2149 2150 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_RHSA, 2151 &dmar_parse_one_rhsa, NULL); 2152 if (ret) 2153 goto release_drhd; 2154 2155 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2156 &dmar_parse_one_atsr, NULL); 2157 if (ret) 2158 goto release_atsr; 2159 2160 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2161 &dmar_hp_add_drhd, NULL); 2162 if (!ret) 2163 return 0; 2164 2165 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2166 &dmar_hp_remove_drhd, NULL); 2167 release_atsr: 2168 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2169 &dmar_release_one_atsr, NULL); 2170 release_drhd: 2171 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2172 &dmar_hp_release_drhd, NULL); 2173 out: 2174 return ret; 2175 } 2176 2177 static int dmar_hotplug_remove(acpi_handle handle) 2178 { 2179 int ret; 2180 2181 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2182 &dmar_check_one_atsr, NULL); 2183 if (ret) 2184 return ret; 2185 2186 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2187 &dmar_hp_remove_drhd, NULL); 2188 if (ret == 0) { 2189 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2190 &dmar_release_one_atsr, NULL)); 2191 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2192 &dmar_hp_release_drhd, NULL)); 2193 } else { 2194 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2195 &dmar_hp_add_drhd, NULL); 2196 } 2197 2198 return ret; 2199 } 2200 2201 static acpi_status dmar_get_dsm_handle(acpi_handle handle, u32 lvl, 2202 void *context, void **retval) 2203 { 2204 acpi_handle *phdl = retval; 2205 2206 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) { 2207 *phdl = handle; 2208 return AE_CTRL_TERMINATE; 2209 } 2210 2211 return AE_OK; 2212 } 2213 2214 static int dmar_device_hotplug(acpi_handle handle, bool insert) 2215 { 2216 int ret; 2217 acpi_handle tmp = NULL; 2218 acpi_status status; 2219 2220 if (!dmar_in_use()) 2221 return 0; 2222 2223 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) { 2224 tmp = handle; 2225 } else { 2226 status = acpi_walk_namespace(ACPI_TYPE_DEVICE, handle, 2227 ACPI_UINT32_MAX, 2228 dmar_get_dsm_handle, 2229 NULL, NULL, &tmp); 2230 if (ACPI_FAILURE(status)) { 2231 pr_warn("Failed to locate _DSM method.\n"); 2232 return -ENXIO; 2233 } 2234 } 2235 if (tmp == NULL) 2236 return 0; 2237 2238 down_write(&dmar_global_lock); 2239 if (insert) 2240 ret = dmar_hotplug_insert(tmp); 2241 else 2242 ret = dmar_hotplug_remove(tmp); 2243 up_write(&dmar_global_lock); 2244 2245 return ret; 2246 } 2247 2248 int dmar_device_add(acpi_handle handle) 2249 { 2250 return dmar_device_hotplug(handle, true); 2251 } 2252 2253 int dmar_device_remove(acpi_handle handle) 2254 { 2255 return dmar_device_hotplug(handle, false); 2256 } 2257 2258 /* 2259 * dmar_platform_optin - Is %DMA_CTRL_PLATFORM_OPT_IN_FLAG set in DMAR table 2260 * 2261 * Returns true if the platform has %DMA_CTRL_PLATFORM_OPT_IN_FLAG set in 2262 * the ACPI DMAR table. This means that the platform boot firmware has made 2263 * sure no device can issue DMA outside of RMRR regions. 2264 */ 2265 bool dmar_platform_optin(void) 2266 { 2267 struct acpi_table_dmar *dmar; 2268 acpi_status status; 2269 bool ret; 2270 2271 status = acpi_get_table(ACPI_SIG_DMAR, 0, 2272 (struct acpi_table_header **)&dmar); 2273 if (ACPI_FAILURE(status)) 2274 return false; 2275 2276 ret = !!(dmar->flags & DMAR_PLATFORM_OPT_IN); 2277 acpi_put_table((struct acpi_table_header *)dmar); 2278 2279 return ret; 2280 } 2281 EXPORT_SYMBOL_GPL(dmar_platform_optin); 2282