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 #include <trace/events/intel_iommu.h> 35 36 #include "../irq_remapping.h" 37 #include "perf.h" 38 39 typedef int (*dmar_res_handler_t)(struct acpi_dmar_header *, void *); 40 struct dmar_res_callback { 41 dmar_res_handler_t cb[ACPI_DMAR_TYPE_RESERVED]; 42 void *arg[ACPI_DMAR_TYPE_RESERVED]; 43 bool ignore_unhandled; 44 bool print_entry; 45 }; 46 47 /* 48 * Assumptions: 49 * 1) The hotplug framework guarentees that DMAR unit will be hot-added 50 * before IO devices managed by that unit. 51 * 2) The hotplug framework guarantees that DMAR unit will be hot-removed 52 * after IO devices managed by that unit. 53 * 3) Hotplug events are rare. 54 * 55 * Locking rules for DMA and interrupt remapping related global data structures: 56 * 1) Use dmar_global_lock in process context 57 * 2) Use RCU in interrupt context 58 */ 59 DECLARE_RWSEM(dmar_global_lock); 60 LIST_HEAD(dmar_drhd_units); 61 62 struct acpi_table_header * __initdata dmar_tbl; 63 static int dmar_dev_scope_status = 1; 64 static unsigned long dmar_seq_ids[BITS_TO_LONGS(DMAR_UNITS_SUPPORTED)]; 65 66 static int alloc_iommu(struct dmar_drhd_unit *drhd); 67 static void free_iommu(struct intel_iommu *iommu); 68 69 extern const struct iommu_ops intel_iommu_ops; 70 71 static void dmar_register_drhd_unit(struct dmar_drhd_unit *drhd) 72 { 73 /* 74 * add INCLUDE_ALL at the tail, so scan the list will find it at 75 * the very end. 76 */ 77 if (drhd->include_all) 78 list_add_tail_rcu(&drhd->list, &dmar_drhd_units); 79 else 80 list_add_rcu(&drhd->list, &dmar_drhd_units); 81 } 82 83 void *dmar_alloc_dev_scope(void *start, void *end, int *cnt) 84 { 85 struct acpi_dmar_device_scope *scope; 86 87 *cnt = 0; 88 while (start < end) { 89 scope = start; 90 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_NAMESPACE || 91 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT || 92 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE) 93 (*cnt)++; 94 else if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_IOAPIC && 95 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_HPET) { 96 pr_warn("Unsupported device scope\n"); 97 } 98 start += scope->length; 99 } 100 if (*cnt == 0) 101 return NULL; 102 103 return kcalloc(*cnt, sizeof(struct dmar_dev_scope), GFP_KERNEL); 104 } 105 106 void dmar_free_dev_scope(struct dmar_dev_scope **devices, int *cnt) 107 { 108 int i; 109 struct device *tmp_dev; 110 111 if (*devices && *cnt) { 112 for_each_active_dev_scope(*devices, *cnt, i, tmp_dev) 113 put_device(tmp_dev); 114 kfree(*devices); 115 } 116 117 *devices = NULL; 118 *cnt = 0; 119 } 120 121 /* Optimize out kzalloc()/kfree() for normal cases */ 122 static char dmar_pci_notify_info_buf[64]; 123 124 static struct dmar_pci_notify_info * 125 dmar_alloc_pci_notify_info(struct pci_dev *dev, unsigned long event) 126 { 127 int level = 0; 128 size_t size; 129 struct pci_dev *tmp; 130 struct dmar_pci_notify_info *info; 131 132 BUG_ON(dev->is_virtfn); 133 134 /* 135 * Ignore devices that have a domain number higher than what can 136 * be looked up in DMAR, e.g. VMD subdevices with domain 0x10000 137 */ 138 if (pci_domain_nr(dev->bus) > U16_MAX) 139 return NULL; 140 141 /* Only generate path[] for device addition event */ 142 if (event == BUS_NOTIFY_ADD_DEVICE) 143 for (tmp = dev; tmp; tmp = tmp->bus->self) 144 level++; 145 146 size = struct_size(info, path, level); 147 if (size <= sizeof(dmar_pci_notify_info_buf)) { 148 info = (struct dmar_pci_notify_info *)dmar_pci_notify_info_buf; 149 } else { 150 info = kzalloc(size, GFP_KERNEL); 151 if (!info) { 152 pr_warn("Out of memory when allocating notify_info " 153 "for %s.\n", pci_name(dev)); 154 if (dmar_dev_scope_status == 0) 155 dmar_dev_scope_status = -ENOMEM; 156 return NULL; 157 } 158 } 159 160 info->event = event; 161 info->dev = dev; 162 info->seg = pci_domain_nr(dev->bus); 163 info->level = level; 164 if (event == BUS_NOTIFY_ADD_DEVICE) { 165 for (tmp = dev; tmp; tmp = tmp->bus->self) { 166 level--; 167 info->path[level].bus = tmp->bus->number; 168 info->path[level].device = PCI_SLOT(tmp->devfn); 169 info->path[level].function = PCI_FUNC(tmp->devfn); 170 if (pci_is_root_bus(tmp->bus)) 171 info->bus = tmp->bus->number; 172 } 173 } 174 175 return info; 176 } 177 178 static inline void dmar_free_pci_notify_info(struct dmar_pci_notify_info *info) 179 { 180 if ((void *)info != dmar_pci_notify_info_buf) 181 kfree(info); 182 } 183 184 static bool dmar_match_pci_path(struct dmar_pci_notify_info *info, int bus, 185 struct acpi_dmar_pci_path *path, int count) 186 { 187 int i; 188 189 if (info->bus != bus) 190 goto fallback; 191 if (info->level != count) 192 goto fallback; 193 194 for (i = 0; i < count; i++) { 195 if (path[i].device != info->path[i].device || 196 path[i].function != info->path[i].function) 197 goto fallback; 198 } 199 200 return true; 201 202 fallback: 203 204 if (count != 1) 205 return false; 206 207 i = info->level - 1; 208 if (bus == info->path[i].bus && 209 path[0].device == info->path[i].device && 210 path[0].function == info->path[i].function) { 211 pr_info(FW_BUG "RMRR entry for device %02x:%02x.%x is broken - applying workaround\n", 212 bus, path[0].device, path[0].function); 213 return true; 214 } 215 216 return false; 217 } 218 219 /* Return: > 0 if match found, 0 if no match found, < 0 if error happens */ 220 int dmar_insert_dev_scope(struct dmar_pci_notify_info *info, 221 void *start, void*end, u16 segment, 222 struct dmar_dev_scope *devices, 223 int devices_cnt) 224 { 225 int i, level; 226 struct device *tmp, *dev = &info->dev->dev; 227 struct acpi_dmar_device_scope *scope; 228 struct acpi_dmar_pci_path *path; 229 230 if (segment != info->seg) 231 return 0; 232 233 for (; start < end; start += scope->length) { 234 scope = start; 235 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_ENDPOINT && 236 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_BRIDGE) 237 continue; 238 239 path = (struct acpi_dmar_pci_path *)(scope + 1); 240 level = (scope->length - sizeof(*scope)) / sizeof(*path); 241 if (!dmar_match_pci_path(info, scope->bus, path, level)) 242 continue; 243 244 /* 245 * We expect devices with endpoint scope to have normal PCI 246 * headers, and devices with bridge scope to have bridge PCI 247 * headers. However PCI NTB devices may be listed in the 248 * DMAR table with bridge scope, even though they have a 249 * normal PCI header. NTB devices are identified by class 250 * "BRIDGE_OTHER" (0680h) - we don't declare a socpe mismatch 251 * for this special case. 252 */ 253 if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT && 254 info->dev->hdr_type != PCI_HEADER_TYPE_NORMAL) || 255 (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE && 256 (info->dev->hdr_type == PCI_HEADER_TYPE_NORMAL && 257 info->dev->class >> 16 != PCI_BASE_CLASS_BRIDGE))) { 258 pr_warn("Device scope type does not match for %s\n", 259 pci_name(info->dev)); 260 return -EINVAL; 261 } 262 263 for_each_dev_scope(devices, devices_cnt, i, tmp) 264 if (tmp == NULL) { 265 devices[i].bus = info->dev->bus->number; 266 devices[i].devfn = info->dev->devfn; 267 rcu_assign_pointer(devices[i].dev, 268 get_device(dev)); 269 return 1; 270 } 271 BUG_ON(i >= devices_cnt); 272 } 273 274 return 0; 275 } 276 277 int dmar_remove_dev_scope(struct dmar_pci_notify_info *info, u16 segment, 278 struct dmar_dev_scope *devices, int count) 279 { 280 int index; 281 struct device *tmp; 282 283 if (info->seg != segment) 284 return 0; 285 286 for_each_active_dev_scope(devices, count, index, tmp) 287 if (tmp == &info->dev->dev) { 288 RCU_INIT_POINTER(devices[index].dev, NULL); 289 synchronize_rcu(); 290 put_device(tmp); 291 return 1; 292 } 293 294 return 0; 295 } 296 297 static int dmar_pci_bus_add_dev(struct dmar_pci_notify_info *info) 298 { 299 int ret = 0; 300 struct dmar_drhd_unit *dmaru; 301 struct acpi_dmar_hardware_unit *drhd; 302 303 for_each_drhd_unit(dmaru) { 304 if (dmaru->include_all) 305 continue; 306 307 drhd = container_of(dmaru->hdr, 308 struct acpi_dmar_hardware_unit, header); 309 ret = dmar_insert_dev_scope(info, (void *)(drhd + 1), 310 ((void *)drhd) + drhd->header.length, 311 dmaru->segment, 312 dmaru->devices, dmaru->devices_cnt); 313 if (ret) 314 break; 315 } 316 if (ret >= 0) 317 ret = dmar_iommu_notify_scope_dev(info); 318 if (ret < 0 && dmar_dev_scope_status == 0) 319 dmar_dev_scope_status = ret; 320 321 if (ret >= 0) 322 intel_irq_remap_add_device(info); 323 324 return ret; 325 } 326 327 static void dmar_pci_bus_del_dev(struct dmar_pci_notify_info *info) 328 { 329 struct dmar_drhd_unit *dmaru; 330 331 for_each_drhd_unit(dmaru) 332 if (dmar_remove_dev_scope(info, dmaru->segment, 333 dmaru->devices, dmaru->devices_cnt)) 334 break; 335 dmar_iommu_notify_scope_dev(info); 336 } 337 338 static inline void vf_inherit_msi_domain(struct pci_dev *pdev) 339 { 340 struct pci_dev *physfn = pci_physfn(pdev); 341 342 dev_set_msi_domain(&pdev->dev, dev_get_msi_domain(&physfn->dev)); 343 } 344 345 static int dmar_pci_bus_notifier(struct notifier_block *nb, 346 unsigned long action, void *data) 347 { 348 struct pci_dev *pdev = to_pci_dev(data); 349 struct dmar_pci_notify_info *info; 350 351 /* Only care about add/remove events for physical functions. 352 * For VFs we actually do the lookup based on the corresponding 353 * PF in device_to_iommu() anyway. */ 354 if (pdev->is_virtfn) { 355 /* 356 * Ensure that the VF device inherits the irq domain of the 357 * PF device. Ideally the device would inherit the domain 358 * from the bus, but DMAR can have multiple units per bus 359 * which makes this impossible. The VF 'bus' could inherit 360 * from the PF device, but that's yet another x86'sism to 361 * inflict on everybody else. 362 */ 363 if (action == BUS_NOTIFY_ADD_DEVICE) 364 vf_inherit_msi_domain(pdev); 365 return NOTIFY_DONE; 366 } 367 368 if (action != BUS_NOTIFY_ADD_DEVICE && 369 action != BUS_NOTIFY_REMOVED_DEVICE) 370 return NOTIFY_DONE; 371 372 info = dmar_alloc_pci_notify_info(pdev, action); 373 if (!info) 374 return NOTIFY_DONE; 375 376 down_write(&dmar_global_lock); 377 if (action == BUS_NOTIFY_ADD_DEVICE) 378 dmar_pci_bus_add_dev(info); 379 else if (action == BUS_NOTIFY_REMOVED_DEVICE) 380 dmar_pci_bus_del_dev(info); 381 up_write(&dmar_global_lock); 382 383 dmar_free_pci_notify_info(info); 384 385 return NOTIFY_OK; 386 } 387 388 static struct notifier_block dmar_pci_bus_nb = { 389 .notifier_call = dmar_pci_bus_notifier, 390 .priority = INT_MIN, 391 }; 392 393 static struct dmar_drhd_unit * 394 dmar_find_dmaru(struct acpi_dmar_hardware_unit *drhd) 395 { 396 struct dmar_drhd_unit *dmaru; 397 398 list_for_each_entry_rcu(dmaru, &dmar_drhd_units, list, 399 dmar_rcu_check()) 400 if (dmaru->segment == drhd->segment && 401 dmaru->reg_base_addr == drhd->address) 402 return dmaru; 403 404 return NULL; 405 } 406 407 /* 408 * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition 409 * structure which uniquely represent one DMA remapping hardware unit 410 * present in the platform 411 */ 412 static int dmar_parse_one_drhd(struct acpi_dmar_header *header, void *arg) 413 { 414 struct acpi_dmar_hardware_unit *drhd; 415 struct dmar_drhd_unit *dmaru; 416 int ret; 417 418 drhd = (struct acpi_dmar_hardware_unit *)header; 419 dmaru = dmar_find_dmaru(drhd); 420 if (dmaru) 421 goto out; 422 423 dmaru = kzalloc(sizeof(*dmaru) + header->length, GFP_KERNEL); 424 if (!dmaru) 425 return -ENOMEM; 426 427 /* 428 * If header is allocated from slab by ACPI _DSM method, we need to 429 * copy the content because the memory buffer will be freed on return. 430 */ 431 dmaru->hdr = (void *)(dmaru + 1); 432 memcpy(dmaru->hdr, header, header->length); 433 dmaru->reg_base_addr = drhd->address; 434 dmaru->segment = drhd->segment; 435 dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */ 436 dmaru->devices = dmar_alloc_dev_scope((void *)(drhd + 1), 437 ((void *)drhd) + drhd->header.length, 438 &dmaru->devices_cnt); 439 if (dmaru->devices_cnt && dmaru->devices == NULL) { 440 kfree(dmaru); 441 return -ENOMEM; 442 } 443 444 ret = alloc_iommu(dmaru); 445 if (ret) { 446 dmar_free_dev_scope(&dmaru->devices, 447 &dmaru->devices_cnt); 448 kfree(dmaru); 449 return ret; 450 } 451 dmar_register_drhd_unit(dmaru); 452 453 out: 454 if (arg) 455 (*(int *)arg)++; 456 457 return 0; 458 } 459 460 static void dmar_free_drhd(struct dmar_drhd_unit *dmaru) 461 { 462 if (dmaru->devices && dmaru->devices_cnt) 463 dmar_free_dev_scope(&dmaru->devices, &dmaru->devices_cnt); 464 if (dmaru->iommu) 465 free_iommu(dmaru->iommu); 466 kfree(dmaru); 467 } 468 469 static int __init dmar_parse_one_andd(struct acpi_dmar_header *header, 470 void *arg) 471 { 472 struct acpi_dmar_andd *andd = (void *)header; 473 474 /* Check for NUL termination within the designated length */ 475 if (strnlen(andd->device_name, header->length - 8) == header->length - 8) { 476 pr_warn(FW_BUG 477 "Your BIOS is broken; ANDD object name is not NUL-terminated\n" 478 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 479 dmi_get_system_info(DMI_BIOS_VENDOR), 480 dmi_get_system_info(DMI_BIOS_VERSION), 481 dmi_get_system_info(DMI_PRODUCT_VERSION)); 482 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 483 return -EINVAL; 484 } 485 pr_info("ANDD device: %x name: %s\n", andd->device_number, 486 andd->device_name); 487 488 return 0; 489 } 490 491 #ifdef CONFIG_ACPI_NUMA 492 static int dmar_parse_one_rhsa(struct acpi_dmar_header *header, void *arg) 493 { 494 struct acpi_dmar_rhsa *rhsa; 495 struct dmar_drhd_unit *drhd; 496 497 rhsa = (struct acpi_dmar_rhsa *)header; 498 for_each_drhd_unit(drhd) { 499 if (drhd->reg_base_addr == rhsa->base_address) { 500 int node = pxm_to_node(rhsa->proximity_domain); 501 502 if (!node_online(node)) 503 node = NUMA_NO_NODE; 504 drhd->iommu->node = node; 505 return 0; 506 } 507 } 508 pr_warn(FW_BUG 509 "Your BIOS is broken; RHSA refers to non-existent DMAR unit at %llx\n" 510 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 511 rhsa->base_address, 512 dmi_get_system_info(DMI_BIOS_VENDOR), 513 dmi_get_system_info(DMI_BIOS_VERSION), 514 dmi_get_system_info(DMI_PRODUCT_VERSION)); 515 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 516 517 return 0; 518 } 519 #else 520 #define dmar_parse_one_rhsa dmar_res_noop 521 #endif 522 523 static void 524 dmar_table_print_dmar_entry(struct acpi_dmar_header *header) 525 { 526 struct acpi_dmar_hardware_unit *drhd; 527 struct acpi_dmar_reserved_memory *rmrr; 528 struct acpi_dmar_atsr *atsr; 529 struct acpi_dmar_rhsa *rhsa; 530 struct acpi_dmar_satc *satc; 531 532 switch (header->type) { 533 case ACPI_DMAR_TYPE_HARDWARE_UNIT: 534 drhd = container_of(header, struct acpi_dmar_hardware_unit, 535 header); 536 pr_info("DRHD base: %#016Lx flags: %#x\n", 537 (unsigned long long)drhd->address, drhd->flags); 538 break; 539 case ACPI_DMAR_TYPE_RESERVED_MEMORY: 540 rmrr = container_of(header, struct acpi_dmar_reserved_memory, 541 header); 542 pr_info("RMRR base: %#016Lx end: %#016Lx\n", 543 (unsigned long long)rmrr->base_address, 544 (unsigned long long)rmrr->end_address); 545 break; 546 case ACPI_DMAR_TYPE_ROOT_ATS: 547 atsr = container_of(header, struct acpi_dmar_atsr, header); 548 pr_info("ATSR flags: %#x\n", atsr->flags); 549 break; 550 case ACPI_DMAR_TYPE_HARDWARE_AFFINITY: 551 rhsa = container_of(header, struct acpi_dmar_rhsa, header); 552 pr_info("RHSA base: %#016Lx proximity domain: %#x\n", 553 (unsigned long long)rhsa->base_address, 554 rhsa->proximity_domain); 555 break; 556 case ACPI_DMAR_TYPE_NAMESPACE: 557 /* We don't print this here because we need to sanity-check 558 it first. So print it in dmar_parse_one_andd() instead. */ 559 break; 560 case ACPI_DMAR_TYPE_SATC: 561 satc = container_of(header, struct acpi_dmar_satc, header); 562 pr_info("SATC flags: 0x%x\n", satc->flags); 563 break; 564 } 565 } 566 567 /** 568 * dmar_table_detect - checks to see if the platform supports DMAR devices 569 */ 570 static int __init dmar_table_detect(void) 571 { 572 acpi_status status = AE_OK; 573 574 /* if we could find DMAR table, then there are DMAR devices */ 575 status = acpi_get_table(ACPI_SIG_DMAR, 0, &dmar_tbl); 576 577 if (ACPI_SUCCESS(status) && !dmar_tbl) { 578 pr_warn("Unable to map DMAR\n"); 579 status = AE_NOT_FOUND; 580 } 581 582 return ACPI_SUCCESS(status) ? 0 : -ENOENT; 583 } 584 585 static int dmar_walk_remapping_entries(struct acpi_dmar_header *start, 586 size_t len, struct dmar_res_callback *cb) 587 { 588 struct acpi_dmar_header *iter, *next; 589 struct acpi_dmar_header *end = ((void *)start) + len; 590 591 for (iter = start; iter < end; iter = next) { 592 next = (void *)iter + iter->length; 593 if (iter->length == 0) { 594 /* Avoid looping forever on bad ACPI tables */ 595 pr_debug(FW_BUG "Invalid 0-length structure\n"); 596 break; 597 } else if (next > end) { 598 /* Avoid passing table end */ 599 pr_warn(FW_BUG "Record passes table end\n"); 600 return -EINVAL; 601 } 602 603 if (cb->print_entry) 604 dmar_table_print_dmar_entry(iter); 605 606 if (iter->type >= ACPI_DMAR_TYPE_RESERVED) { 607 /* continue for forward compatibility */ 608 pr_debug("Unknown DMAR structure type %d\n", 609 iter->type); 610 } else if (cb->cb[iter->type]) { 611 int ret; 612 613 ret = cb->cb[iter->type](iter, cb->arg[iter->type]); 614 if (ret) 615 return ret; 616 } else if (!cb->ignore_unhandled) { 617 pr_warn("No handler for DMAR structure type %d\n", 618 iter->type); 619 return -EINVAL; 620 } 621 } 622 623 return 0; 624 } 625 626 static inline int dmar_walk_dmar_table(struct acpi_table_dmar *dmar, 627 struct dmar_res_callback *cb) 628 { 629 return dmar_walk_remapping_entries((void *)(dmar + 1), 630 dmar->header.length - sizeof(*dmar), cb); 631 } 632 633 /** 634 * parse_dmar_table - parses the DMA reporting table 635 */ 636 static int __init 637 parse_dmar_table(void) 638 { 639 struct acpi_table_dmar *dmar; 640 int drhd_count = 0; 641 int ret; 642 struct dmar_res_callback cb = { 643 .print_entry = true, 644 .ignore_unhandled = true, 645 .arg[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &drhd_count, 646 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_parse_one_drhd, 647 .cb[ACPI_DMAR_TYPE_RESERVED_MEMORY] = &dmar_parse_one_rmrr, 648 .cb[ACPI_DMAR_TYPE_ROOT_ATS] = &dmar_parse_one_atsr, 649 .cb[ACPI_DMAR_TYPE_HARDWARE_AFFINITY] = &dmar_parse_one_rhsa, 650 .cb[ACPI_DMAR_TYPE_NAMESPACE] = &dmar_parse_one_andd, 651 .cb[ACPI_DMAR_TYPE_SATC] = &dmar_parse_one_satc, 652 }; 653 654 /* 655 * Do it again, earlier dmar_tbl mapping could be mapped with 656 * fixed map. 657 */ 658 dmar_table_detect(); 659 660 /* 661 * ACPI tables may not be DMA protected by tboot, so use DMAR copy 662 * SINIT saved in SinitMleData in TXT heap (which is DMA protected) 663 */ 664 dmar_tbl = tboot_get_dmar_table(dmar_tbl); 665 666 dmar = (struct acpi_table_dmar *)dmar_tbl; 667 if (!dmar) 668 return -ENODEV; 669 670 if (dmar->width < PAGE_SHIFT - 1) { 671 pr_warn("Invalid DMAR haw\n"); 672 return -EINVAL; 673 } 674 675 pr_info("Host address width %d\n", dmar->width + 1); 676 ret = dmar_walk_dmar_table(dmar, &cb); 677 if (ret == 0 && drhd_count == 0) 678 pr_warn(FW_BUG "No DRHD structure found in DMAR table\n"); 679 680 return ret; 681 } 682 683 static int dmar_pci_device_match(struct dmar_dev_scope devices[], 684 int cnt, struct pci_dev *dev) 685 { 686 int index; 687 struct device *tmp; 688 689 while (dev) { 690 for_each_active_dev_scope(devices, cnt, index, tmp) 691 if (dev_is_pci(tmp) && dev == to_pci_dev(tmp)) 692 return 1; 693 694 /* Check our parent */ 695 dev = dev->bus->self; 696 } 697 698 return 0; 699 } 700 701 struct dmar_drhd_unit * 702 dmar_find_matched_drhd_unit(struct pci_dev *dev) 703 { 704 struct dmar_drhd_unit *dmaru; 705 struct acpi_dmar_hardware_unit *drhd; 706 707 dev = pci_physfn(dev); 708 709 rcu_read_lock(); 710 for_each_drhd_unit(dmaru) { 711 drhd = container_of(dmaru->hdr, 712 struct acpi_dmar_hardware_unit, 713 header); 714 715 if (dmaru->include_all && 716 drhd->segment == pci_domain_nr(dev->bus)) 717 goto out; 718 719 if (dmar_pci_device_match(dmaru->devices, 720 dmaru->devices_cnt, dev)) 721 goto out; 722 } 723 dmaru = NULL; 724 out: 725 rcu_read_unlock(); 726 727 return dmaru; 728 } 729 730 static void __init dmar_acpi_insert_dev_scope(u8 device_number, 731 struct acpi_device *adev) 732 { 733 struct dmar_drhd_unit *dmaru; 734 struct acpi_dmar_hardware_unit *drhd; 735 struct acpi_dmar_device_scope *scope; 736 struct device *tmp; 737 int i; 738 struct acpi_dmar_pci_path *path; 739 740 for_each_drhd_unit(dmaru) { 741 drhd = container_of(dmaru->hdr, 742 struct acpi_dmar_hardware_unit, 743 header); 744 745 for (scope = (void *)(drhd + 1); 746 (unsigned long)scope < ((unsigned long)drhd) + drhd->header.length; 747 scope = ((void *)scope) + scope->length) { 748 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_NAMESPACE) 749 continue; 750 if (scope->enumeration_id != device_number) 751 continue; 752 753 path = (void *)(scope + 1); 754 pr_info("ACPI device \"%s\" under DMAR at %llx as %02x:%02x.%d\n", 755 dev_name(&adev->dev), dmaru->reg_base_addr, 756 scope->bus, path->device, path->function); 757 for_each_dev_scope(dmaru->devices, dmaru->devices_cnt, i, tmp) 758 if (tmp == NULL) { 759 dmaru->devices[i].bus = scope->bus; 760 dmaru->devices[i].devfn = PCI_DEVFN(path->device, 761 path->function); 762 rcu_assign_pointer(dmaru->devices[i].dev, 763 get_device(&adev->dev)); 764 return; 765 } 766 BUG_ON(i >= dmaru->devices_cnt); 767 } 768 } 769 pr_warn("No IOMMU scope found for ANDD enumeration ID %d (%s)\n", 770 device_number, dev_name(&adev->dev)); 771 } 772 773 static int __init dmar_acpi_dev_scope_init(void) 774 { 775 struct acpi_dmar_andd *andd; 776 777 if (dmar_tbl == NULL) 778 return -ENODEV; 779 780 for (andd = (void *)dmar_tbl + sizeof(struct acpi_table_dmar); 781 ((unsigned long)andd) < ((unsigned long)dmar_tbl) + dmar_tbl->length; 782 andd = ((void *)andd) + andd->header.length) { 783 if (andd->header.type == ACPI_DMAR_TYPE_NAMESPACE) { 784 acpi_handle h; 785 struct acpi_device *adev; 786 787 if (!ACPI_SUCCESS(acpi_get_handle(ACPI_ROOT_OBJECT, 788 andd->device_name, 789 &h))) { 790 pr_err("Failed to find handle for ACPI object %s\n", 791 andd->device_name); 792 continue; 793 } 794 if (acpi_bus_get_device(h, &adev)) { 795 pr_err("Failed to get device for ACPI object %s\n", 796 andd->device_name); 797 continue; 798 } 799 dmar_acpi_insert_dev_scope(andd->device_number, adev); 800 } 801 } 802 return 0; 803 } 804 805 int __init dmar_dev_scope_init(void) 806 { 807 struct pci_dev *dev = NULL; 808 struct dmar_pci_notify_info *info; 809 810 if (dmar_dev_scope_status != 1) 811 return dmar_dev_scope_status; 812 813 if (list_empty(&dmar_drhd_units)) { 814 dmar_dev_scope_status = -ENODEV; 815 } else { 816 dmar_dev_scope_status = 0; 817 818 dmar_acpi_dev_scope_init(); 819 820 for_each_pci_dev(dev) { 821 if (dev->is_virtfn) 822 continue; 823 824 info = dmar_alloc_pci_notify_info(dev, 825 BUS_NOTIFY_ADD_DEVICE); 826 if (!info) { 827 return dmar_dev_scope_status; 828 } else { 829 dmar_pci_bus_add_dev(info); 830 dmar_free_pci_notify_info(info); 831 } 832 } 833 } 834 835 return dmar_dev_scope_status; 836 } 837 838 void __init dmar_register_bus_notifier(void) 839 { 840 bus_register_notifier(&pci_bus_type, &dmar_pci_bus_nb); 841 } 842 843 844 int __init dmar_table_init(void) 845 { 846 static int dmar_table_initialized; 847 int ret; 848 849 if (dmar_table_initialized == 0) { 850 ret = parse_dmar_table(); 851 if (ret < 0) { 852 if (ret != -ENODEV) 853 pr_info("Parse DMAR table failure.\n"); 854 } else if (list_empty(&dmar_drhd_units)) { 855 pr_info("No DMAR devices found\n"); 856 ret = -ENODEV; 857 } 858 859 if (ret < 0) 860 dmar_table_initialized = ret; 861 else 862 dmar_table_initialized = 1; 863 } 864 865 return dmar_table_initialized < 0 ? dmar_table_initialized : 0; 866 } 867 868 static void warn_invalid_dmar(u64 addr, const char *message) 869 { 870 pr_warn_once(FW_BUG 871 "Your BIOS is broken; DMAR reported at address %llx%s!\n" 872 "BIOS vendor: %s; Ver: %s; Product Version: %s\n", 873 addr, message, 874 dmi_get_system_info(DMI_BIOS_VENDOR), 875 dmi_get_system_info(DMI_BIOS_VERSION), 876 dmi_get_system_info(DMI_PRODUCT_VERSION)); 877 add_taint(TAINT_FIRMWARE_WORKAROUND, LOCKDEP_STILL_OK); 878 } 879 880 static int __ref 881 dmar_validate_one_drhd(struct acpi_dmar_header *entry, void *arg) 882 { 883 struct acpi_dmar_hardware_unit *drhd; 884 void __iomem *addr; 885 u64 cap, ecap; 886 887 drhd = (void *)entry; 888 if (!drhd->address) { 889 warn_invalid_dmar(0, ""); 890 return -EINVAL; 891 } 892 893 if (arg) 894 addr = ioremap(drhd->address, VTD_PAGE_SIZE); 895 else 896 addr = early_ioremap(drhd->address, VTD_PAGE_SIZE); 897 if (!addr) { 898 pr_warn("Can't validate DRHD address: %llx\n", drhd->address); 899 return -EINVAL; 900 } 901 902 cap = dmar_readq(addr + DMAR_CAP_REG); 903 ecap = dmar_readq(addr + DMAR_ECAP_REG); 904 905 if (arg) 906 iounmap(addr); 907 else 908 early_iounmap(addr, VTD_PAGE_SIZE); 909 910 if (cap == (uint64_t)-1 && ecap == (uint64_t)-1) { 911 warn_invalid_dmar(drhd->address, " returns all ones"); 912 return -EINVAL; 913 } 914 915 return 0; 916 } 917 918 int __init detect_intel_iommu(void) 919 { 920 int ret; 921 struct dmar_res_callback validate_drhd_cb = { 922 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_validate_one_drhd, 923 .ignore_unhandled = true, 924 }; 925 926 down_write(&dmar_global_lock); 927 ret = dmar_table_detect(); 928 if (!ret) 929 ret = dmar_walk_dmar_table((struct acpi_table_dmar *)dmar_tbl, 930 &validate_drhd_cb); 931 if (!ret && !no_iommu && !iommu_detected && 932 (!dmar_disabled || dmar_platform_optin())) { 933 iommu_detected = 1; 934 /* Make sure ACS will be enabled */ 935 pci_request_acs(); 936 } 937 938 #ifdef CONFIG_X86 939 if (!ret) { 940 x86_init.iommu.iommu_init = intel_iommu_init; 941 x86_platform.iommu_shutdown = intel_iommu_shutdown; 942 } 943 944 #endif 945 946 if (dmar_tbl) { 947 acpi_put_table(dmar_tbl); 948 dmar_tbl = NULL; 949 } 950 up_write(&dmar_global_lock); 951 952 return ret ? ret : 1; 953 } 954 955 static void unmap_iommu(struct intel_iommu *iommu) 956 { 957 iounmap(iommu->reg); 958 release_mem_region(iommu->reg_phys, iommu->reg_size); 959 } 960 961 /** 962 * map_iommu: map the iommu's registers 963 * @iommu: the iommu to map 964 * @phys_addr: the physical address of the base resgister 965 * 966 * Memory map the iommu's registers. Start w/ a single page, and 967 * possibly expand if that turns out to be insufficent. 968 */ 969 static int map_iommu(struct intel_iommu *iommu, u64 phys_addr) 970 { 971 int map_size, err=0; 972 973 iommu->reg_phys = phys_addr; 974 iommu->reg_size = VTD_PAGE_SIZE; 975 976 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, iommu->name)) { 977 pr_err("Can't reserve memory\n"); 978 err = -EBUSY; 979 goto out; 980 } 981 982 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size); 983 if (!iommu->reg) { 984 pr_err("Can't map the region\n"); 985 err = -ENOMEM; 986 goto release; 987 } 988 989 iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG); 990 iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG); 991 992 if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) { 993 err = -EINVAL; 994 warn_invalid_dmar(phys_addr, " returns all ones"); 995 goto unmap; 996 } 997 if (ecap_vcs(iommu->ecap)) 998 iommu->vccap = dmar_readq(iommu->reg + DMAR_VCCAP_REG); 999 1000 /* the registers might be more than one page */ 1001 map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap), 1002 cap_max_fault_reg_offset(iommu->cap)); 1003 map_size = VTD_PAGE_ALIGN(map_size); 1004 if (map_size > iommu->reg_size) { 1005 iounmap(iommu->reg); 1006 release_mem_region(iommu->reg_phys, iommu->reg_size); 1007 iommu->reg_size = map_size; 1008 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, 1009 iommu->name)) { 1010 pr_err("Can't reserve memory\n"); 1011 err = -EBUSY; 1012 goto out; 1013 } 1014 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size); 1015 if (!iommu->reg) { 1016 pr_err("Can't map the region\n"); 1017 err = -ENOMEM; 1018 goto release; 1019 } 1020 } 1021 err = 0; 1022 goto out; 1023 1024 unmap: 1025 iounmap(iommu->reg); 1026 release: 1027 release_mem_region(iommu->reg_phys, iommu->reg_size); 1028 out: 1029 return err; 1030 } 1031 1032 static int dmar_alloc_seq_id(struct intel_iommu *iommu) 1033 { 1034 iommu->seq_id = find_first_zero_bit(dmar_seq_ids, 1035 DMAR_UNITS_SUPPORTED); 1036 if (iommu->seq_id >= DMAR_UNITS_SUPPORTED) { 1037 iommu->seq_id = -1; 1038 } else { 1039 set_bit(iommu->seq_id, dmar_seq_ids); 1040 sprintf(iommu->name, "dmar%d", iommu->seq_id); 1041 } 1042 1043 return iommu->seq_id; 1044 } 1045 1046 static void dmar_free_seq_id(struct intel_iommu *iommu) 1047 { 1048 if (iommu->seq_id >= 0) { 1049 clear_bit(iommu->seq_id, dmar_seq_ids); 1050 iommu->seq_id = -1; 1051 } 1052 } 1053 1054 static int alloc_iommu(struct dmar_drhd_unit *drhd) 1055 { 1056 struct intel_iommu *iommu; 1057 u32 ver, sts; 1058 int agaw = -1; 1059 int msagaw = -1; 1060 int err; 1061 1062 if (!drhd->reg_base_addr) { 1063 warn_invalid_dmar(0, ""); 1064 return -EINVAL; 1065 } 1066 1067 iommu = kzalloc(sizeof(*iommu), GFP_KERNEL); 1068 if (!iommu) 1069 return -ENOMEM; 1070 1071 if (dmar_alloc_seq_id(iommu) < 0) { 1072 pr_err("Failed to allocate seq_id\n"); 1073 err = -ENOSPC; 1074 goto error; 1075 } 1076 1077 err = map_iommu(iommu, drhd->reg_base_addr); 1078 if (err) { 1079 pr_err("Failed to map %s\n", iommu->name); 1080 goto error_free_seq_id; 1081 } 1082 1083 err = -EINVAL; 1084 if (cap_sagaw(iommu->cap) == 0) { 1085 pr_info("%s: No supported address widths. Not attempting DMA translation.\n", 1086 iommu->name); 1087 drhd->ignored = 1; 1088 } 1089 1090 if (!drhd->ignored) { 1091 agaw = iommu_calculate_agaw(iommu); 1092 if (agaw < 0) { 1093 pr_err("Cannot get a valid agaw for iommu (seq_id = %d)\n", 1094 iommu->seq_id); 1095 drhd->ignored = 1; 1096 } 1097 } 1098 if (!drhd->ignored) { 1099 msagaw = iommu_calculate_max_sagaw(iommu); 1100 if (msagaw < 0) { 1101 pr_err("Cannot get a valid max agaw for iommu (seq_id = %d)\n", 1102 iommu->seq_id); 1103 drhd->ignored = 1; 1104 agaw = -1; 1105 } 1106 } 1107 iommu->agaw = agaw; 1108 iommu->msagaw = msagaw; 1109 iommu->segment = drhd->segment; 1110 1111 iommu->node = NUMA_NO_NODE; 1112 1113 ver = readl(iommu->reg + DMAR_VER_REG); 1114 pr_info("%s: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n", 1115 iommu->name, 1116 (unsigned long long)drhd->reg_base_addr, 1117 DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver), 1118 (unsigned long long)iommu->cap, 1119 (unsigned long long)iommu->ecap); 1120 1121 /* Reflect status in gcmd */ 1122 sts = readl(iommu->reg + DMAR_GSTS_REG); 1123 if (sts & DMA_GSTS_IRES) 1124 iommu->gcmd |= DMA_GCMD_IRE; 1125 if (sts & DMA_GSTS_TES) 1126 iommu->gcmd |= DMA_GCMD_TE; 1127 if (sts & DMA_GSTS_QIES) 1128 iommu->gcmd |= DMA_GCMD_QIE; 1129 1130 raw_spin_lock_init(&iommu->register_lock); 1131 1132 /* 1133 * This is only for hotplug; at boot time intel_iommu_enabled won't 1134 * be set yet. When intel_iommu_init() runs, it registers the units 1135 * present at boot time, then sets intel_iommu_enabled. 1136 */ 1137 if (intel_iommu_enabled && !drhd->ignored) { 1138 err = iommu_device_sysfs_add(&iommu->iommu, NULL, 1139 intel_iommu_groups, 1140 "%s", iommu->name); 1141 if (err) 1142 goto err_unmap; 1143 1144 err = iommu_device_register(&iommu->iommu, &intel_iommu_ops, NULL); 1145 if (err) 1146 goto err_sysfs; 1147 } 1148 1149 drhd->iommu = iommu; 1150 iommu->drhd = drhd; 1151 1152 return 0; 1153 1154 err_sysfs: 1155 iommu_device_sysfs_remove(&iommu->iommu); 1156 err_unmap: 1157 unmap_iommu(iommu); 1158 error_free_seq_id: 1159 dmar_free_seq_id(iommu); 1160 error: 1161 kfree(iommu); 1162 return err; 1163 } 1164 1165 static void free_iommu(struct intel_iommu *iommu) 1166 { 1167 if (intel_iommu_enabled && !iommu->drhd->ignored) { 1168 iommu_device_unregister(&iommu->iommu); 1169 iommu_device_sysfs_remove(&iommu->iommu); 1170 } 1171 1172 if (iommu->irq) { 1173 if (iommu->pr_irq) { 1174 free_irq(iommu->pr_irq, iommu); 1175 dmar_free_hwirq(iommu->pr_irq); 1176 iommu->pr_irq = 0; 1177 } 1178 free_irq(iommu->irq, iommu); 1179 dmar_free_hwirq(iommu->irq); 1180 iommu->irq = 0; 1181 } 1182 1183 if (iommu->qi) { 1184 free_page((unsigned long)iommu->qi->desc); 1185 kfree(iommu->qi->desc_status); 1186 kfree(iommu->qi); 1187 } 1188 1189 if (iommu->reg) 1190 unmap_iommu(iommu); 1191 1192 dmar_free_seq_id(iommu); 1193 kfree(iommu); 1194 } 1195 1196 /* 1197 * Reclaim all the submitted descriptors which have completed its work. 1198 */ 1199 static inline void reclaim_free_desc(struct q_inval *qi) 1200 { 1201 while (qi->desc_status[qi->free_tail] == QI_DONE || 1202 qi->desc_status[qi->free_tail] == QI_ABORT) { 1203 qi->desc_status[qi->free_tail] = QI_FREE; 1204 qi->free_tail = (qi->free_tail + 1) % QI_LENGTH; 1205 qi->free_cnt++; 1206 } 1207 } 1208 1209 static const char *qi_type_string(u8 type) 1210 { 1211 switch (type) { 1212 case QI_CC_TYPE: 1213 return "Context-cache Invalidation"; 1214 case QI_IOTLB_TYPE: 1215 return "IOTLB Invalidation"; 1216 case QI_DIOTLB_TYPE: 1217 return "Device-TLB Invalidation"; 1218 case QI_IEC_TYPE: 1219 return "Interrupt Entry Cache Invalidation"; 1220 case QI_IWD_TYPE: 1221 return "Invalidation Wait"; 1222 case QI_EIOTLB_TYPE: 1223 return "PASID-based IOTLB Invalidation"; 1224 case QI_PC_TYPE: 1225 return "PASID-cache Invalidation"; 1226 case QI_DEIOTLB_TYPE: 1227 return "PASID-based Device-TLB Invalidation"; 1228 case QI_PGRP_RESP_TYPE: 1229 return "Page Group Response"; 1230 default: 1231 return "UNKNOWN"; 1232 } 1233 } 1234 1235 static void qi_dump_fault(struct intel_iommu *iommu, u32 fault) 1236 { 1237 unsigned int head = dmar_readl(iommu->reg + DMAR_IQH_REG); 1238 u64 iqe_err = dmar_readq(iommu->reg + DMAR_IQER_REG); 1239 struct qi_desc *desc = iommu->qi->desc + head; 1240 1241 if (fault & DMA_FSTS_IQE) 1242 pr_err("VT-d detected Invalidation Queue Error: Reason %llx", 1243 DMAR_IQER_REG_IQEI(iqe_err)); 1244 if (fault & DMA_FSTS_ITE) 1245 pr_err("VT-d detected Invalidation Time-out Error: SID %llx", 1246 DMAR_IQER_REG_ITESID(iqe_err)); 1247 if (fault & DMA_FSTS_ICE) 1248 pr_err("VT-d detected Invalidation Completion Error: SID %llx", 1249 DMAR_IQER_REG_ICESID(iqe_err)); 1250 1251 pr_err("QI HEAD: %s qw0 = 0x%llx, qw1 = 0x%llx\n", 1252 qi_type_string(desc->qw0 & 0xf), 1253 (unsigned long long)desc->qw0, 1254 (unsigned long long)desc->qw1); 1255 1256 head = ((head >> qi_shift(iommu)) + QI_LENGTH - 1) % QI_LENGTH; 1257 head <<= qi_shift(iommu); 1258 desc = iommu->qi->desc + head; 1259 1260 pr_err("QI PRIOR: %s qw0 = 0x%llx, qw1 = 0x%llx\n", 1261 qi_type_string(desc->qw0 & 0xf), 1262 (unsigned long long)desc->qw0, 1263 (unsigned long long)desc->qw1); 1264 } 1265 1266 static int qi_check_fault(struct intel_iommu *iommu, int index, int wait_index) 1267 { 1268 u32 fault; 1269 int head, tail; 1270 struct q_inval *qi = iommu->qi; 1271 int shift = qi_shift(iommu); 1272 1273 if (qi->desc_status[wait_index] == QI_ABORT) 1274 return -EAGAIN; 1275 1276 fault = readl(iommu->reg + DMAR_FSTS_REG); 1277 if (fault & (DMA_FSTS_IQE | DMA_FSTS_ITE | DMA_FSTS_ICE)) 1278 qi_dump_fault(iommu, fault); 1279 1280 /* 1281 * If IQE happens, the head points to the descriptor associated 1282 * with the error. No new descriptors are fetched until the IQE 1283 * is cleared. 1284 */ 1285 if (fault & DMA_FSTS_IQE) { 1286 head = readl(iommu->reg + DMAR_IQH_REG); 1287 if ((head >> shift) == index) { 1288 struct qi_desc *desc = qi->desc + head; 1289 1290 /* 1291 * desc->qw2 and desc->qw3 are either reserved or 1292 * used by software as private data. We won't print 1293 * out these two qw's for security consideration. 1294 */ 1295 memcpy(desc, qi->desc + (wait_index << shift), 1296 1 << shift); 1297 writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG); 1298 pr_info("Invalidation Queue Error (IQE) cleared\n"); 1299 return -EINVAL; 1300 } 1301 } 1302 1303 /* 1304 * If ITE happens, all pending wait_desc commands are aborted. 1305 * No new descriptors are fetched until the ITE is cleared. 1306 */ 1307 if (fault & DMA_FSTS_ITE) { 1308 head = readl(iommu->reg + DMAR_IQH_REG); 1309 head = ((head >> shift) - 1 + QI_LENGTH) % QI_LENGTH; 1310 head |= 1; 1311 tail = readl(iommu->reg + DMAR_IQT_REG); 1312 tail = ((tail >> shift) - 1 + QI_LENGTH) % QI_LENGTH; 1313 1314 writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG); 1315 pr_info("Invalidation Time-out Error (ITE) cleared\n"); 1316 1317 do { 1318 if (qi->desc_status[head] == QI_IN_USE) 1319 qi->desc_status[head] = QI_ABORT; 1320 head = (head - 2 + QI_LENGTH) % QI_LENGTH; 1321 } while (head != tail); 1322 1323 if (qi->desc_status[wait_index] == QI_ABORT) 1324 return -EAGAIN; 1325 } 1326 1327 if (fault & DMA_FSTS_ICE) { 1328 writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG); 1329 pr_info("Invalidation Completion Error (ICE) cleared\n"); 1330 } 1331 1332 return 0; 1333 } 1334 1335 /* 1336 * Function to submit invalidation descriptors of all types to the queued 1337 * invalidation interface(QI). Multiple descriptors can be submitted at a 1338 * time, a wait descriptor will be appended to each submission to ensure 1339 * hardware has completed the invalidation before return. Wait descriptors 1340 * can be part of the submission but it will not be polled for completion. 1341 */ 1342 int qi_submit_sync(struct intel_iommu *iommu, struct qi_desc *desc, 1343 unsigned int count, unsigned long options) 1344 { 1345 struct q_inval *qi = iommu->qi; 1346 s64 devtlb_start_ktime = 0; 1347 s64 iotlb_start_ktime = 0; 1348 s64 iec_start_ktime = 0; 1349 struct qi_desc wait_desc; 1350 int wait_index, index; 1351 unsigned long flags; 1352 int offset, shift; 1353 int rc, i; 1354 u64 type; 1355 1356 if (!qi) 1357 return 0; 1358 1359 type = desc->qw0 & GENMASK_ULL(3, 0); 1360 1361 if ((type == QI_IOTLB_TYPE || type == QI_EIOTLB_TYPE) && 1362 dmar_latency_enabled(iommu, DMAR_LATENCY_INV_IOTLB)) 1363 iotlb_start_ktime = ktime_to_ns(ktime_get()); 1364 1365 if ((type == QI_DIOTLB_TYPE || type == QI_DEIOTLB_TYPE) && 1366 dmar_latency_enabled(iommu, DMAR_LATENCY_INV_DEVTLB)) 1367 devtlb_start_ktime = ktime_to_ns(ktime_get()); 1368 1369 if (type == QI_IEC_TYPE && 1370 dmar_latency_enabled(iommu, DMAR_LATENCY_INV_IEC)) 1371 iec_start_ktime = ktime_to_ns(ktime_get()); 1372 1373 restart: 1374 rc = 0; 1375 1376 raw_spin_lock_irqsave(&qi->q_lock, flags); 1377 /* 1378 * Check if we have enough empty slots in the queue to submit, 1379 * the calculation is based on: 1380 * # of desc + 1 wait desc + 1 space between head and tail 1381 */ 1382 while (qi->free_cnt < count + 2) { 1383 raw_spin_unlock_irqrestore(&qi->q_lock, flags); 1384 cpu_relax(); 1385 raw_spin_lock_irqsave(&qi->q_lock, flags); 1386 } 1387 1388 index = qi->free_head; 1389 wait_index = (index + count) % QI_LENGTH; 1390 shift = qi_shift(iommu); 1391 1392 for (i = 0; i < count; i++) { 1393 offset = ((index + i) % QI_LENGTH) << shift; 1394 memcpy(qi->desc + offset, &desc[i], 1 << shift); 1395 qi->desc_status[(index + i) % QI_LENGTH] = QI_IN_USE; 1396 trace_qi_submit(iommu, desc[i].qw0, desc[i].qw1, 1397 desc[i].qw2, desc[i].qw3); 1398 } 1399 qi->desc_status[wait_index] = QI_IN_USE; 1400 1401 wait_desc.qw0 = QI_IWD_STATUS_DATA(QI_DONE) | 1402 QI_IWD_STATUS_WRITE | QI_IWD_TYPE; 1403 if (options & QI_OPT_WAIT_DRAIN) 1404 wait_desc.qw0 |= QI_IWD_PRQ_DRAIN; 1405 wait_desc.qw1 = virt_to_phys(&qi->desc_status[wait_index]); 1406 wait_desc.qw2 = 0; 1407 wait_desc.qw3 = 0; 1408 1409 offset = wait_index << shift; 1410 memcpy(qi->desc + offset, &wait_desc, 1 << shift); 1411 1412 qi->free_head = (qi->free_head + count + 1) % QI_LENGTH; 1413 qi->free_cnt -= count + 1; 1414 1415 /* 1416 * update the HW tail register indicating the presence of 1417 * new descriptors. 1418 */ 1419 writel(qi->free_head << shift, iommu->reg + DMAR_IQT_REG); 1420 1421 while (qi->desc_status[wait_index] != QI_DONE) { 1422 /* 1423 * We will leave the interrupts disabled, to prevent interrupt 1424 * context to queue another cmd while a cmd is already submitted 1425 * and waiting for completion on this cpu. This is to avoid 1426 * a deadlock where the interrupt context can wait indefinitely 1427 * for free slots in the queue. 1428 */ 1429 rc = qi_check_fault(iommu, index, wait_index); 1430 if (rc) 1431 break; 1432 1433 raw_spin_unlock(&qi->q_lock); 1434 cpu_relax(); 1435 raw_spin_lock(&qi->q_lock); 1436 } 1437 1438 for (i = 0; i < count; i++) 1439 qi->desc_status[(index + i) % QI_LENGTH] = QI_DONE; 1440 1441 reclaim_free_desc(qi); 1442 raw_spin_unlock_irqrestore(&qi->q_lock, flags); 1443 1444 if (rc == -EAGAIN) 1445 goto restart; 1446 1447 if (iotlb_start_ktime) 1448 dmar_latency_update(iommu, DMAR_LATENCY_INV_IOTLB, 1449 ktime_to_ns(ktime_get()) - iotlb_start_ktime); 1450 1451 if (devtlb_start_ktime) 1452 dmar_latency_update(iommu, DMAR_LATENCY_INV_DEVTLB, 1453 ktime_to_ns(ktime_get()) - devtlb_start_ktime); 1454 1455 if (iec_start_ktime) 1456 dmar_latency_update(iommu, DMAR_LATENCY_INV_IEC, 1457 ktime_to_ns(ktime_get()) - iec_start_ktime); 1458 1459 return rc; 1460 } 1461 1462 /* 1463 * Flush the global interrupt entry cache. 1464 */ 1465 void qi_global_iec(struct intel_iommu *iommu) 1466 { 1467 struct qi_desc desc; 1468 1469 desc.qw0 = QI_IEC_TYPE; 1470 desc.qw1 = 0; 1471 desc.qw2 = 0; 1472 desc.qw3 = 0; 1473 1474 /* should never fail */ 1475 qi_submit_sync(iommu, &desc, 1, 0); 1476 } 1477 1478 void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm, 1479 u64 type) 1480 { 1481 struct qi_desc desc; 1482 1483 desc.qw0 = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did) 1484 | QI_CC_GRAN(type) | QI_CC_TYPE; 1485 desc.qw1 = 0; 1486 desc.qw2 = 0; 1487 desc.qw3 = 0; 1488 1489 qi_submit_sync(iommu, &desc, 1, 0); 1490 } 1491 1492 void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr, 1493 unsigned int size_order, u64 type) 1494 { 1495 u8 dw = 0, dr = 0; 1496 1497 struct qi_desc desc; 1498 int ih = 0; 1499 1500 if (cap_write_drain(iommu->cap)) 1501 dw = 1; 1502 1503 if (cap_read_drain(iommu->cap)) 1504 dr = 1; 1505 1506 desc.qw0 = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw) 1507 | QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE; 1508 desc.qw1 = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih) 1509 | QI_IOTLB_AM(size_order); 1510 desc.qw2 = 0; 1511 desc.qw3 = 0; 1512 1513 qi_submit_sync(iommu, &desc, 1, 0); 1514 } 1515 1516 void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 pfsid, 1517 u16 qdep, u64 addr, unsigned mask) 1518 { 1519 struct qi_desc desc; 1520 1521 if (mask) { 1522 addr |= (1ULL << (VTD_PAGE_SHIFT + mask - 1)) - 1; 1523 desc.qw1 = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE; 1524 } else 1525 desc.qw1 = QI_DEV_IOTLB_ADDR(addr); 1526 1527 if (qdep >= QI_DEV_IOTLB_MAX_INVS) 1528 qdep = 0; 1529 1530 desc.qw0 = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) | 1531 QI_DIOTLB_TYPE | QI_DEV_IOTLB_PFSID(pfsid); 1532 desc.qw2 = 0; 1533 desc.qw3 = 0; 1534 1535 qi_submit_sync(iommu, &desc, 1, 0); 1536 } 1537 1538 /* PASID-based IOTLB invalidation */ 1539 void qi_flush_piotlb(struct intel_iommu *iommu, u16 did, u32 pasid, u64 addr, 1540 unsigned long npages, bool ih) 1541 { 1542 struct qi_desc desc = {.qw2 = 0, .qw3 = 0}; 1543 1544 /* 1545 * npages == -1 means a PASID-selective invalidation, otherwise, 1546 * a positive value for Page-selective-within-PASID invalidation. 1547 * 0 is not a valid input. 1548 */ 1549 if (WARN_ON(!npages)) { 1550 pr_err("Invalid input npages = %ld\n", npages); 1551 return; 1552 } 1553 1554 if (npages == -1) { 1555 desc.qw0 = QI_EIOTLB_PASID(pasid) | 1556 QI_EIOTLB_DID(did) | 1557 QI_EIOTLB_GRAN(QI_GRAN_NONG_PASID) | 1558 QI_EIOTLB_TYPE; 1559 desc.qw1 = 0; 1560 } else { 1561 int mask = ilog2(__roundup_pow_of_two(npages)); 1562 unsigned long align = (1ULL << (VTD_PAGE_SHIFT + mask)); 1563 1564 if (WARN_ON_ONCE(!IS_ALIGNED(addr, align))) 1565 addr = ALIGN_DOWN(addr, align); 1566 1567 desc.qw0 = QI_EIOTLB_PASID(pasid) | 1568 QI_EIOTLB_DID(did) | 1569 QI_EIOTLB_GRAN(QI_GRAN_PSI_PASID) | 1570 QI_EIOTLB_TYPE; 1571 desc.qw1 = QI_EIOTLB_ADDR(addr) | 1572 QI_EIOTLB_IH(ih) | 1573 QI_EIOTLB_AM(mask); 1574 } 1575 1576 qi_submit_sync(iommu, &desc, 1, 0); 1577 } 1578 1579 /* PASID-based device IOTLB Invalidate */ 1580 void qi_flush_dev_iotlb_pasid(struct intel_iommu *iommu, u16 sid, u16 pfsid, 1581 u32 pasid, u16 qdep, u64 addr, unsigned int size_order) 1582 { 1583 unsigned long mask = 1UL << (VTD_PAGE_SHIFT + size_order - 1); 1584 struct qi_desc desc = {.qw1 = 0, .qw2 = 0, .qw3 = 0}; 1585 1586 desc.qw0 = QI_DEV_EIOTLB_PASID(pasid) | QI_DEV_EIOTLB_SID(sid) | 1587 QI_DEV_EIOTLB_QDEP(qdep) | QI_DEIOTLB_TYPE | 1588 QI_DEV_IOTLB_PFSID(pfsid); 1589 1590 /* 1591 * If S bit is 0, we only flush a single page. If S bit is set, 1592 * The least significant zero bit indicates the invalidation address 1593 * range. VT-d spec 6.5.2.6. 1594 * e.g. address bit 12[0] indicates 8KB, 13[0] indicates 16KB. 1595 * size order = 0 is PAGE_SIZE 4KB 1596 * Max Invs Pending (MIP) is set to 0 for now until we have DIT in 1597 * ECAP. 1598 */ 1599 if (!IS_ALIGNED(addr, VTD_PAGE_SIZE << size_order)) 1600 pr_warn_ratelimited("Invalidate non-aligned address %llx, order %d\n", 1601 addr, size_order); 1602 1603 /* Take page address */ 1604 desc.qw1 = QI_DEV_EIOTLB_ADDR(addr); 1605 1606 if (size_order) { 1607 /* 1608 * Existing 0s in address below size_order may be the least 1609 * significant bit, we must set them to 1s to avoid having 1610 * smaller size than desired. 1611 */ 1612 desc.qw1 |= GENMASK_ULL(size_order + VTD_PAGE_SHIFT - 1, 1613 VTD_PAGE_SHIFT); 1614 /* Clear size_order bit to indicate size */ 1615 desc.qw1 &= ~mask; 1616 /* Set the S bit to indicate flushing more than 1 page */ 1617 desc.qw1 |= QI_DEV_EIOTLB_SIZE; 1618 } 1619 1620 qi_submit_sync(iommu, &desc, 1, 0); 1621 } 1622 1623 void qi_flush_pasid_cache(struct intel_iommu *iommu, u16 did, 1624 u64 granu, u32 pasid) 1625 { 1626 struct qi_desc desc = {.qw1 = 0, .qw2 = 0, .qw3 = 0}; 1627 1628 desc.qw0 = QI_PC_PASID(pasid) | QI_PC_DID(did) | 1629 QI_PC_GRAN(granu) | QI_PC_TYPE; 1630 qi_submit_sync(iommu, &desc, 1, 0); 1631 } 1632 1633 /* 1634 * Disable Queued Invalidation interface. 1635 */ 1636 void dmar_disable_qi(struct intel_iommu *iommu) 1637 { 1638 unsigned long flags; 1639 u32 sts; 1640 cycles_t start_time = get_cycles(); 1641 1642 if (!ecap_qis(iommu->ecap)) 1643 return; 1644 1645 raw_spin_lock_irqsave(&iommu->register_lock, flags); 1646 1647 sts = readl(iommu->reg + DMAR_GSTS_REG); 1648 if (!(sts & DMA_GSTS_QIES)) 1649 goto end; 1650 1651 /* 1652 * Give a chance to HW to complete the pending invalidation requests. 1653 */ 1654 while ((readl(iommu->reg + DMAR_IQT_REG) != 1655 readl(iommu->reg + DMAR_IQH_REG)) && 1656 (DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time))) 1657 cpu_relax(); 1658 1659 iommu->gcmd &= ~DMA_GCMD_QIE; 1660 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG); 1661 1662 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, 1663 !(sts & DMA_GSTS_QIES), sts); 1664 end: 1665 raw_spin_unlock_irqrestore(&iommu->register_lock, flags); 1666 } 1667 1668 /* 1669 * Enable queued invalidation. 1670 */ 1671 static void __dmar_enable_qi(struct intel_iommu *iommu) 1672 { 1673 u32 sts; 1674 unsigned long flags; 1675 struct q_inval *qi = iommu->qi; 1676 u64 val = virt_to_phys(qi->desc); 1677 1678 qi->free_head = qi->free_tail = 0; 1679 qi->free_cnt = QI_LENGTH; 1680 1681 /* 1682 * Set DW=1 and QS=1 in IQA_REG when Scalable Mode capability 1683 * is present. 1684 */ 1685 if (ecap_smts(iommu->ecap)) 1686 val |= (1 << 11) | 1; 1687 1688 raw_spin_lock_irqsave(&iommu->register_lock, flags); 1689 1690 /* write zero to the tail reg */ 1691 writel(0, iommu->reg + DMAR_IQT_REG); 1692 1693 dmar_writeq(iommu->reg + DMAR_IQA_REG, val); 1694 1695 iommu->gcmd |= DMA_GCMD_QIE; 1696 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG); 1697 1698 /* Make sure hardware complete it */ 1699 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts); 1700 1701 raw_spin_unlock_irqrestore(&iommu->register_lock, flags); 1702 } 1703 1704 /* 1705 * Enable Queued Invalidation interface. This is a must to support 1706 * interrupt-remapping. Also used by DMA-remapping, which replaces 1707 * register based IOTLB invalidation. 1708 */ 1709 int dmar_enable_qi(struct intel_iommu *iommu) 1710 { 1711 struct q_inval *qi; 1712 struct page *desc_page; 1713 1714 if (!ecap_qis(iommu->ecap)) 1715 return -ENOENT; 1716 1717 /* 1718 * queued invalidation is already setup and enabled. 1719 */ 1720 if (iommu->qi) 1721 return 0; 1722 1723 iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC); 1724 if (!iommu->qi) 1725 return -ENOMEM; 1726 1727 qi = iommu->qi; 1728 1729 /* 1730 * Need two pages to accommodate 256 descriptors of 256 bits each 1731 * if the remapping hardware supports scalable mode translation. 1732 */ 1733 desc_page = alloc_pages_node(iommu->node, GFP_ATOMIC | __GFP_ZERO, 1734 !!ecap_smts(iommu->ecap)); 1735 if (!desc_page) { 1736 kfree(qi); 1737 iommu->qi = NULL; 1738 return -ENOMEM; 1739 } 1740 1741 qi->desc = page_address(desc_page); 1742 1743 qi->desc_status = kcalloc(QI_LENGTH, sizeof(int), GFP_ATOMIC); 1744 if (!qi->desc_status) { 1745 free_page((unsigned long) qi->desc); 1746 kfree(qi); 1747 iommu->qi = NULL; 1748 return -ENOMEM; 1749 } 1750 1751 raw_spin_lock_init(&qi->q_lock); 1752 1753 __dmar_enable_qi(iommu); 1754 1755 return 0; 1756 } 1757 1758 /* iommu interrupt handling. Most stuff are MSI-like. */ 1759 1760 enum faulttype { 1761 DMA_REMAP, 1762 INTR_REMAP, 1763 UNKNOWN, 1764 }; 1765 1766 static const char *dma_remap_fault_reasons[] = 1767 { 1768 "Software", 1769 "Present bit in root entry is clear", 1770 "Present bit in context entry is clear", 1771 "Invalid context entry", 1772 "Access beyond MGAW", 1773 "PTE Write access is not set", 1774 "PTE Read access is not set", 1775 "Next page table ptr is invalid", 1776 "Root table address invalid", 1777 "Context table ptr is invalid", 1778 "non-zero reserved fields in RTP", 1779 "non-zero reserved fields in CTP", 1780 "non-zero reserved fields in PTE", 1781 "PCE for translation request specifies blocking", 1782 }; 1783 1784 static const char * const dma_remap_sm_fault_reasons[] = { 1785 "SM: Invalid Root Table Address", 1786 "SM: TTM 0 for request with PASID", 1787 "SM: TTM 0 for page group request", 1788 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x33-0x37 */ 1789 "SM: Error attempting to access Root Entry", 1790 "SM: Present bit in Root Entry is clear", 1791 "SM: Non-zero reserved field set in Root Entry", 1792 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x3B-0x3F */ 1793 "SM: Error attempting to access Context Entry", 1794 "SM: Present bit in Context Entry is clear", 1795 "SM: Non-zero reserved field set in the Context Entry", 1796 "SM: Invalid Context Entry", 1797 "SM: DTE field in Context Entry is clear", 1798 "SM: PASID Enable field in Context Entry is clear", 1799 "SM: PASID is larger than the max in Context Entry", 1800 "SM: PRE field in Context-Entry is clear", 1801 "SM: RID_PASID field error in Context-Entry", 1802 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x49-0x4F */ 1803 "SM: Error attempting to access the PASID Directory Entry", 1804 "SM: Present bit in Directory Entry is clear", 1805 "SM: Non-zero reserved field set in PASID Directory Entry", 1806 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x53-0x57 */ 1807 "SM: Error attempting to access PASID Table Entry", 1808 "SM: Present bit in PASID Table Entry is clear", 1809 "SM: Non-zero reserved field set in PASID Table Entry", 1810 "SM: Invalid Scalable-Mode PASID Table Entry", 1811 "SM: ERE field is clear in PASID Table Entry", 1812 "SM: SRE field is clear in PASID Table Entry", 1813 "Unknown", "Unknown",/* 0x5E-0x5F */ 1814 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x60-0x67 */ 1815 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x68-0x6F */ 1816 "SM: Error attempting to access first-level paging entry", 1817 "SM: Present bit in first-level paging entry is clear", 1818 "SM: Non-zero reserved field set in first-level paging entry", 1819 "SM: Error attempting to access FL-PML4 entry", 1820 "SM: First-level entry address beyond MGAW in Nested translation", 1821 "SM: Read permission error in FL-PML4 entry in Nested translation", 1822 "SM: Read permission error in first-level paging entry in Nested translation", 1823 "SM: Write permission error in first-level paging entry in Nested translation", 1824 "SM: Error attempting to access second-level paging entry", 1825 "SM: Read/Write permission error in second-level paging entry", 1826 "SM: Non-zero reserved field set in second-level paging entry", 1827 "SM: Invalid second-level page table pointer", 1828 "SM: A/D bit update needed in second-level entry when set up in no snoop", 1829 "Unknown", "Unknown", "Unknown", /* 0x7D-0x7F */ 1830 "SM: Address in first-level translation is not canonical", 1831 "SM: U/S set 0 for first-level translation with user privilege", 1832 "SM: No execute permission for request with PASID and ER=1", 1833 "SM: Address beyond the DMA hardware max", 1834 "SM: Second-level entry address beyond the max", 1835 "SM: No write permission for Write/AtomicOp request", 1836 "SM: No read permission for Read/AtomicOp request", 1837 "SM: Invalid address-interrupt address", 1838 "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", "Unknown", /* 0x88-0x8F */ 1839 "SM: A/D bit update needed in first-level entry when set up in no snoop", 1840 }; 1841 1842 static const char *irq_remap_fault_reasons[] = 1843 { 1844 "Detected reserved fields in the decoded interrupt-remapped request", 1845 "Interrupt index exceeded the interrupt-remapping table size", 1846 "Present field in the IRTE entry is clear", 1847 "Error accessing interrupt-remapping table pointed by IRTA_REG", 1848 "Detected reserved fields in the IRTE entry", 1849 "Blocked a compatibility format interrupt request", 1850 "Blocked an interrupt request due to source-id verification failure", 1851 }; 1852 1853 static const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type) 1854 { 1855 if (fault_reason >= 0x20 && (fault_reason - 0x20 < 1856 ARRAY_SIZE(irq_remap_fault_reasons))) { 1857 *fault_type = INTR_REMAP; 1858 return irq_remap_fault_reasons[fault_reason - 0x20]; 1859 } else if (fault_reason >= 0x30 && (fault_reason - 0x30 < 1860 ARRAY_SIZE(dma_remap_sm_fault_reasons))) { 1861 *fault_type = DMA_REMAP; 1862 return dma_remap_sm_fault_reasons[fault_reason - 0x30]; 1863 } else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) { 1864 *fault_type = DMA_REMAP; 1865 return dma_remap_fault_reasons[fault_reason]; 1866 } else { 1867 *fault_type = UNKNOWN; 1868 return "Unknown"; 1869 } 1870 } 1871 1872 1873 static inline int dmar_msi_reg(struct intel_iommu *iommu, int irq) 1874 { 1875 if (iommu->irq == irq) 1876 return DMAR_FECTL_REG; 1877 else if (iommu->pr_irq == irq) 1878 return DMAR_PECTL_REG; 1879 else 1880 BUG(); 1881 } 1882 1883 void dmar_msi_unmask(struct irq_data *data) 1884 { 1885 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data); 1886 int reg = dmar_msi_reg(iommu, data->irq); 1887 unsigned long flag; 1888 1889 /* unmask it */ 1890 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1891 writel(0, iommu->reg + reg); 1892 /* Read a reg to force flush the post write */ 1893 readl(iommu->reg + reg); 1894 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1895 } 1896 1897 void dmar_msi_mask(struct irq_data *data) 1898 { 1899 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data); 1900 int reg = dmar_msi_reg(iommu, data->irq); 1901 unsigned long flag; 1902 1903 /* mask it */ 1904 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1905 writel(DMA_FECTL_IM, iommu->reg + reg); 1906 /* Read a reg to force flush the post write */ 1907 readl(iommu->reg + reg); 1908 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1909 } 1910 1911 void dmar_msi_write(int irq, struct msi_msg *msg) 1912 { 1913 struct intel_iommu *iommu = irq_get_handler_data(irq); 1914 int reg = dmar_msi_reg(iommu, irq); 1915 unsigned long flag; 1916 1917 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1918 writel(msg->data, iommu->reg + reg + 4); 1919 writel(msg->address_lo, iommu->reg + reg + 8); 1920 writel(msg->address_hi, iommu->reg + reg + 12); 1921 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1922 } 1923 1924 void dmar_msi_read(int irq, struct msi_msg *msg) 1925 { 1926 struct intel_iommu *iommu = irq_get_handler_data(irq); 1927 int reg = dmar_msi_reg(iommu, irq); 1928 unsigned long flag; 1929 1930 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1931 msg->data = readl(iommu->reg + reg + 4); 1932 msg->address_lo = readl(iommu->reg + reg + 8); 1933 msg->address_hi = readl(iommu->reg + reg + 12); 1934 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 1935 } 1936 1937 static int dmar_fault_do_one(struct intel_iommu *iommu, int type, 1938 u8 fault_reason, u32 pasid, u16 source_id, 1939 unsigned long long addr) 1940 { 1941 const char *reason; 1942 int fault_type; 1943 1944 reason = dmar_get_fault_reason(fault_reason, &fault_type); 1945 1946 if (fault_type == INTR_REMAP) 1947 pr_err("[INTR-REMAP] Request device [0x%02x:0x%02x.%d] fault index 0x%llx [fault reason 0x%02x] %s\n", 1948 source_id >> 8, PCI_SLOT(source_id & 0xFF), 1949 PCI_FUNC(source_id & 0xFF), addr >> 48, 1950 fault_reason, reason); 1951 else if (pasid == INVALID_IOASID) 1952 pr_err("[%s NO_PASID] Request device [0x%02x:0x%02x.%d] fault addr 0x%llx [fault reason 0x%02x] %s\n", 1953 type ? "DMA Read" : "DMA Write", 1954 source_id >> 8, PCI_SLOT(source_id & 0xFF), 1955 PCI_FUNC(source_id & 0xFF), addr, 1956 fault_reason, reason); 1957 else 1958 pr_err("[%s PASID 0x%x] Request device [0x%02x:0x%02x.%d] fault addr 0x%llx [fault reason 0x%02x] %s\n", 1959 type ? "DMA Read" : "DMA Write", pasid, 1960 source_id >> 8, PCI_SLOT(source_id & 0xFF), 1961 PCI_FUNC(source_id & 0xFF), addr, 1962 fault_reason, reason); 1963 1964 return 0; 1965 } 1966 1967 #define PRIMARY_FAULT_REG_LEN (16) 1968 irqreturn_t dmar_fault(int irq, void *dev_id) 1969 { 1970 struct intel_iommu *iommu = dev_id; 1971 int reg, fault_index; 1972 u32 fault_status; 1973 unsigned long flag; 1974 static DEFINE_RATELIMIT_STATE(rs, 1975 DEFAULT_RATELIMIT_INTERVAL, 1976 DEFAULT_RATELIMIT_BURST); 1977 1978 raw_spin_lock_irqsave(&iommu->register_lock, flag); 1979 fault_status = readl(iommu->reg + DMAR_FSTS_REG); 1980 if (fault_status && __ratelimit(&rs)) 1981 pr_err("DRHD: handling fault status reg %x\n", fault_status); 1982 1983 /* TBD: ignore advanced fault log currently */ 1984 if (!(fault_status & DMA_FSTS_PPF)) 1985 goto unlock_exit; 1986 1987 fault_index = dma_fsts_fault_record_index(fault_status); 1988 reg = cap_fault_reg_offset(iommu->cap); 1989 while (1) { 1990 /* Disable printing, simply clear the fault when ratelimited */ 1991 bool ratelimited = !__ratelimit(&rs); 1992 u8 fault_reason; 1993 u16 source_id; 1994 u64 guest_addr; 1995 u32 pasid; 1996 int type; 1997 u32 data; 1998 bool pasid_present; 1999 2000 /* highest 32 bits */ 2001 data = readl(iommu->reg + reg + 2002 fault_index * PRIMARY_FAULT_REG_LEN + 12); 2003 if (!(data & DMA_FRCD_F)) 2004 break; 2005 2006 if (!ratelimited) { 2007 fault_reason = dma_frcd_fault_reason(data); 2008 type = dma_frcd_type(data); 2009 2010 pasid = dma_frcd_pasid_value(data); 2011 data = readl(iommu->reg + reg + 2012 fault_index * PRIMARY_FAULT_REG_LEN + 8); 2013 source_id = dma_frcd_source_id(data); 2014 2015 pasid_present = dma_frcd_pasid_present(data); 2016 guest_addr = dmar_readq(iommu->reg + reg + 2017 fault_index * PRIMARY_FAULT_REG_LEN); 2018 guest_addr = dma_frcd_page_addr(guest_addr); 2019 } 2020 2021 /* clear the fault */ 2022 writel(DMA_FRCD_F, iommu->reg + reg + 2023 fault_index * PRIMARY_FAULT_REG_LEN + 12); 2024 2025 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 2026 2027 if (!ratelimited) 2028 /* Using pasid -1 if pasid is not present */ 2029 dmar_fault_do_one(iommu, type, fault_reason, 2030 pasid_present ? pasid : INVALID_IOASID, 2031 source_id, guest_addr); 2032 2033 fault_index++; 2034 if (fault_index >= cap_num_fault_regs(iommu->cap)) 2035 fault_index = 0; 2036 raw_spin_lock_irqsave(&iommu->register_lock, flag); 2037 } 2038 2039 writel(DMA_FSTS_PFO | DMA_FSTS_PPF | DMA_FSTS_PRO, 2040 iommu->reg + DMAR_FSTS_REG); 2041 2042 unlock_exit: 2043 raw_spin_unlock_irqrestore(&iommu->register_lock, flag); 2044 return IRQ_HANDLED; 2045 } 2046 2047 int dmar_set_interrupt(struct intel_iommu *iommu) 2048 { 2049 int irq, ret; 2050 2051 /* 2052 * Check if the fault interrupt is already initialized. 2053 */ 2054 if (iommu->irq) 2055 return 0; 2056 2057 irq = dmar_alloc_hwirq(iommu->seq_id, iommu->node, iommu); 2058 if (irq > 0) { 2059 iommu->irq = irq; 2060 } else { 2061 pr_err("No free IRQ vectors\n"); 2062 return -EINVAL; 2063 } 2064 2065 ret = request_irq(irq, dmar_fault, IRQF_NO_THREAD, iommu->name, iommu); 2066 if (ret) 2067 pr_err("Can't request irq\n"); 2068 return ret; 2069 } 2070 2071 int __init enable_drhd_fault_handling(void) 2072 { 2073 struct dmar_drhd_unit *drhd; 2074 struct intel_iommu *iommu; 2075 2076 /* 2077 * Enable fault control interrupt. 2078 */ 2079 for_each_iommu(iommu, drhd) { 2080 u32 fault_status; 2081 int ret = dmar_set_interrupt(iommu); 2082 2083 if (ret) { 2084 pr_err("DRHD %Lx: failed to enable fault, interrupt, ret %d\n", 2085 (unsigned long long)drhd->reg_base_addr, ret); 2086 return -1; 2087 } 2088 2089 /* 2090 * Clear any previous faults. 2091 */ 2092 dmar_fault(iommu->irq, iommu); 2093 fault_status = readl(iommu->reg + DMAR_FSTS_REG); 2094 writel(fault_status, iommu->reg + DMAR_FSTS_REG); 2095 } 2096 2097 return 0; 2098 } 2099 2100 /* 2101 * Re-enable Queued Invalidation interface. 2102 */ 2103 int dmar_reenable_qi(struct intel_iommu *iommu) 2104 { 2105 if (!ecap_qis(iommu->ecap)) 2106 return -ENOENT; 2107 2108 if (!iommu->qi) 2109 return -ENOENT; 2110 2111 /* 2112 * First disable queued invalidation. 2113 */ 2114 dmar_disable_qi(iommu); 2115 /* 2116 * Then enable queued invalidation again. Since there is no pending 2117 * invalidation requests now, it's safe to re-enable queued 2118 * invalidation. 2119 */ 2120 __dmar_enable_qi(iommu); 2121 2122 return 0; 2123 } 2124 2125 /* 2126 * Check interrupt remapping support in DMAR table description. 2127 */ 2128 int __init dmar_ir_support(void) 2129 { 2130 struct acpi_table_dmar *dmar; 2131 dmar = (struct acpi_table_dmar *)dmar_tbl; 2132 if (!dmar) 2133 return 0; 2134 return dmar->flags & 0x1; 2135 } 2136 2137 /* Check whether DMAR units are in use */ 2138 static inline bool dmar_in_use(void) 2139 { 2140 return irq_remapping_enabled || intel_iommu_enabled; 2141 } 2142 2143 static int __init dmar_free_unused_resources(void) 2144 { 2145 struct dmar_drhd_unit *dmaru, *dmaru_n; 2146 2147 if (dmar_in_use()) 2148 return 0; 2149 2150 if (dmar_dev_scope_status != 1 && !list_empty(&dmar_drhd_units)) 2151 bus_unregister_notifier(&pci_bus_type, &dmar_pci_bus_nb); 2152 2153 down_write(&dmar_global_lock); 2154 list_for_each_entry_safe(dmaru, dmaru_n, &dmar_drhd_units, list) { 2155 list_del(&dmaru->list); 2156 dmar_free_drhd(dmaru); 2157 } 2158 up_write(&dmar_global_lock); 2159 2160 return 0; 2161 } 2162 2163 late_initcall(dmar_free_unused_resources); 2164 IOMMU_INIT_POST(detect_intel_iommu); 2165 2166 /* 2167 * DMAR Hotplug Support 2168 * For more details, please refer to Intel(R) Virtualization Technology 2169 * for Directed-IO Architecture Specifiction, Rev 2.2, Section 8.8 2170 * "Remapping Hardware Unit Hot Plug". 2171 */ 2172 static guid_t dmar_hp_guid = 2173 GUID_INIT(0xD8C1A3A6, 0xBE9B, 0x4C9B, 2174 0x91, 0xBF, 0xC3, 0xCB, 0x81, 0xFC, 0x5D, 0xAF); 2175 2176 /* 2177 * Currently there's only one revision and BIOS will not check the revision id, 2178 * so use 0 for safety. 2179 */ 2180 #define DMAR_DSM_REV_ID 0 2181 #define DMAR_DSM_FUNC_DRHD 1 2182 #define DMAR_DSM_FUNC_ATSR 2 2183 #define DMAR_DSM_FUNC_RHSA 3 2184 #define DMAR_DSM_FUNC_SATC 4 2185 2186 static inline bool dmar_detect_dsm(acpi_handle handle, int func) 2187 { 2188 return acpi_check_dsm(handle, &dmar_hp_guid, DMAR_DSM_REV_ID, 1 << func); 2189 } 2190 2191 static int dmar_walk_dsm_resource(acpi_handle handle, int func, 2192 dmar_res_handler_t handler, void *arg) 2193 { 2194 int ret = -ENODEV; 2195 union acpi_object *obj; 2196 struct acpi_dmar_header *start; 2197 struct dmar_res_callback callback; 2198 static int res_type[] = { 2199 [DMAR_DSM_FUNC_DRHD] = ACPI_DMAR_TYPE_HARDWARE_UNIT, 2200 [DMAR_DSM_FUNC_ATSR] = ACPI_DMAR_TYPE_ROOT_ATS, 2201 [DMAR_DSM_FUNC_RHSA] = ACPI_DMAR_TYPE_HARDWARE_AFFINITY, 2202 [DMAR_DSM_FUNC_SATC] = ACPI_DMAR_TYPE_SATC, 2203 }; 2204 2205 if (!dmar_detect_dsm(handle, func)) 2206 return 0; 2207 2208 obj = acpi_evaluate_dsm_typed(handle, &dmar_hp_guid, DMAR_DSM_REV_ID, 2209 func, NULL, ACPI_TYPE_BUFFER); 2210 if (!obj) 2211 return -ENODEV; 2212 2213 memset(&callback, 0, sizeof(callback)); 2214 callback.cb[res_type[func]] = handler; 2215 callback.arg[res_type[func]] = arg; 2216 start = (struct acpi_dmar_header *)obj->buffer.pointer; 2217 ret = dmar_walk_remapping_entries(start, obj->buffer.length, &callback); 2218 2219 ACPI_FREE(obj); 2220 2221 return ret; 2222 } 2223 2224 static int dmar_hp_add_drhd(struct acpi_dmar_header *header, void *arg) 2225 { 2226 int ret; 2227 struct dmar_drhd_unit *dmaru; 2228 2229 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2230 if (!dmaru) 2231 return -ENODEV; 2232 2233 ret = dmar_ir_hotplug(dmaru, true); 2234 if (ret == 0) 2235 ret = dmar_iommu_hotplug(dmaru, true); 2236 2237 return ret; 2238 } 2239 2240 static int dmar_hp_remove_drhd(struct acpi_dmar_header *header, void *arg) 2241 { 2242 int i, ret; 2243 struct device *dev; 2244 struct dmar_drhd_unit *dmaru; 2245 2246 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2247 if (!dmaru) 2248 return 0; 2249 2250 /* 2251 * All PCI devices managed by this unit should have been destroyed. 2252 */ 2253 if (!dmaru->include_all && dmaru->devices && dmaru->devices_cnt) { 2254 for_each_active_dev_scope(dmaru->devices, 2255 dmaru->devices_cnt, i, dev) 2256 return -EBUSY; 2257 } 2258 2259 ret = dmar_ir_hotplug(dmaru, false); 2260 if (ret == 0) 2261 ret = dmar_iommu_hotplug(dmaru, false); 2262 2263 return ret; 2264 } 2265 2266 static int dmar_hp_release_drhd(struct acpi_dmar_header *header, void *arg) 2267 { 2268 struct dmar_drhd_unit *dmaru; 2269 2270 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header); 2271 if (dmaru) { 2272 list_del_rcu(&dmaru->list); 2273 synchronize_rcu(); 2274 dmar_free_drhd(dmaru); 2275 } 2276 2277 return 0; 2278 } 2279 2280 static int dmar_hotplug_insert(acpi_handle handle) 2281 { 2282 int ret; 2283 int drhd_count = 0; 2284 2285 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2286 &dmar_validate_one_drhd, (void *)1); 2287 if (ret) 2288 goto out; 2289 2290 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2291 &dmar_parse_one_drhd, (void *)&drhd_count); 2292 if (ret == 0 && drhd_count == 0) { 2293 pr_warn(FW_BUG "No DRHD structures in buffer returned by _DSM method\n"); 2294 goto out; 2295 } else if (ret) { 2296 goto release_drhd; 2297 } 2298 2299 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_RHSA, 2300 &dmar_parse_one_rhsa, NULL); 2301 if (ret) 2302 goto release_drhd; 2303 2304 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2305 &dmar_parse_one_atsr, NULL); 2306 if (ret) 2307 goto release_atsr; 2308 2309 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2310 &dmar_hp_add_drhd, NULL); 2311 if (!ret) 2312 return 0; 2313 2314 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2315 &dmar_hp_remove_drhd, NULL); 2316 release_atsr: 2317 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2318 &dmar_release_one_atsr, NULL); 2319 release_drhd: 2320 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2321 &dmar_hp_release_drhd, NULL); 2322 out: 2323 return ret; 2324 } 2325 2326 static int dmar_hotplug_remove(acpi_handle handle) 2327 { 2328 int ret; 2329 2330 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2331 &dmar_check_one_atsr, NULL); 2332 if (ret) 2333 return ret; 2334 2335 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2336 &dmar_hp_remove_drhd, NULL); 2337 if (ret == 0) { 2338 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR, 2339 &dmar_release_one_atsr, NULL)); 2340 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2341 &dmar_hp_release_drhd, NULL)); 2342 } else { 2343 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD, 2344 &dmar_hp_add_drhd, NULL); 2345 } 2346 2347 return ret; 2348 } 2349 2350 static acpi_status dmar_get_dsm_handle(acpi_handle handle, u32 lvl, 2351 void *context, void **retval) 2352 { 2353 acpi_handle *phdl = retval; 2354 2355 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) { 2356 *phdl = handle; 2357 return AE_CTRL_TERMINATE; 2358 } 2359 2360 return AE_OK; 2361 } 2362 2363 static int dmar_device_hotplug(acpi_handle handle, bool insert) 2364 { 2365 int ret; 2366 acpi_handle tmp = NULL; 2367 acpi_status status; 2368 2369 if (!dmar_in_use()) 2370 return 0; 2371 2372 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) { 2373 tmp = handle; 2374 } else { 2375 status = acpi_walk_namespace(ACPI_TYPE_DEVICE, handle, 2376 ACPI_UINT32_MAX, 2377 dmar_get_dsm_handle, 2378 NULL, NULL, &tmp); 2379 if (ACPI_FAILURE(status)) { 2380 pr_warn("Failed to locate _DSM method.\n"); 2381 return -ENXIO; 2382 } 2383 } 2384 if (tmp == NULL) 2385 return 0; 2386 2387 down_write(&dmar_global_lock); 2388 if (insert) 2389 ret = dmar_hotplug_insert(tmp); 2390 else 2391 ret = dmar_hotplug_remove(tmp); 2392 up_write(&dmar_global_lock); 2393 2394 return ret; 2395 } 2396 2397 int dmar_device_add(acpi_handle handle) 2398 { 2399 return dmar_device_hotplug(handle, true); 2400 } 2401 2402 int dmar_device_remove(acpi_handle handle) 2403 { 2404 return dmar_device_hotplug(handle, false); 2405 } 2406 2407 /* 2408 * dmar_platform_optin - Is %DMA_CTRL_PLATFORM_OPT_IN_FLAG set in DMAR table 2409 * 2410 * Returns true if the platform has %DMA_CTRL_PLATFORM_OPT_IN_FLAG set in 2411 * the ACPI DMAR table. This means that the platform boot firmware has made 2412 * sure no device can issue DMA outside of RMRR regions. 2413 */ 2414 bool dmar_platform_optin(void) 2415 { 2416 struct acpi_table_dmar *dmar; 2417 acpi_status status; 2418 bool ret; 2419 2420 status = acpi_get_table(ACPI_SIG_DMAR, 0, 2421 (struct acpi_table_header **)&dmar); 2422 if (ACPI_FAILURE(status)) 2423 return false; 2424 2425 ret = !!(dmar->flags & DMAR_PLATFORM_OPT_IN); 2426 acpi_put_table((struct acpi_table_header *)dmar); 2427 2428 return ret; 2429 } 2430 EXPORT_SYMBOL_GPL(dmar_platform_optin); 2431