drivers/pci/dmar.c

   1 /*
   2  * Copyright (c) 2006, Intel Corporation.
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms and conditions of the GNU General Public License,
   6  * version 2, as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope it will be useful, but WITHOUT
   9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  10  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  11  * more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along with
  14  * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
  15  * Place - Suite 330, Boston, MA 02111-1307 USA.
  16  *
  17  * Copyright (C) 2006-2008 Intel Corporation
  18  * Author: Ashok Raj <ashok.raj@intel.com>
  19  * Author: Shaohua Li <shaohua.li@intel.com>
  20  * Author: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
  21  *
  22  * This file implements early detection/parsing of Remapping Devices
  23  * reported to OS through BIOS via DMA remapping reporting (DMAR) ACPI
  24  * tables.
  25  *
  26  * These routines are used by both DMA-remapping and Interrupt-remapping
  27  */
  28
  29 #include <linux/pci.h>
  30 #include <linux/dmar.h>
  31 #include <linux/iova.h>
  32 #include <linux/intel-iommu.h>
  33 #include <linux/timer.h>
  34 #include <linux/irq.h>
  35 #include <linux/interrupt.h>
  36 #include <linux/tboot.h>
  37 #include <linux/dmi.h>
  38
  39 #define PREFIX "DMAR: "
  40
  41 /* No locks are needed as DMA remapping hardware unit
  42  * list is constructed at boot time and hotplug of
  43  * these units are not supported by the architecture.
  44  */
  45 LIST_HEAD(dmar_drhd_units);
  46
  47 static struct acpi_table_header * __initdata dmar_tbl;
  48 static acpi_size dmar_tbl_size;
  49
  50 static void __init dmar_register_drhd_unit(struct dmar_drhd_unit *drhd)
  51 {
  52         /*
  53          * add INCLUDE_ALL at the tail, so scan the list will find it at
  54          * the very end.
  55          */
  56         if (drhd->include_all)
  57                 list_add_tail(&drhd->list, &dmar_drhd_units);
  58         else
  59                 list_add(&drhd->list, &dmar_drhd_units);
  60 }
  61
  62 static int __init dmar_parse_one_dev_scope(struct acpi_dmar_device_scope *scope,
  63                                            struct pci_dev **dev, u16 segment)
  64 {
  65         struct pci_bus *bus;
  66         struct pci_dev *pdev = NULL;
  67         struct acpi_dmar_pci_path *path;
  68         int count;
  69
  70         bus = pci_find_bus(segment, scope->bus);
  71         path = (struct acpi_dmar_pci_path *)(scope + 1);
  72         count = (scope->length - sizeof(struct acpi_dmar_device_scope))
  73                 / sizeof(struct acpi_dmar_pci_path);
  74
  75         while (count) {
  76                 if (pdev)
  77                         pci_dev_put(pdev);
  78                 /*
  79                  * Some BIOSes list non-exist devices in DMAR table, just
  80                  * ignore it
  81                  */
  82                 if (!bus) {
  83                         printk(KERN_WARNING
  84                         PREFIX "Device scope bus [%d] not found\n",
  85                         scope->bus);
  86                         break;
  87                 }
  88                 pdev = pci_get_slot(bus, PCI_DEVFN(path->dev, path->fn));
  89                 if (!pdev) {
  90                         printk(KERN_WARNING PREFIX
  91                         "Device scope device [%04x:%02x:%02x.%02x] not found\n",
  92                                 segment, bus->number, path->dev, path->fn);
  93                         break;
  94                 }
  95                 path ++;
  96                 count --;
  97                 bus = pdev->subordinate;
  98         }
  99         if (!pdev) {
 100                 printk(KERN_WARNING PREFIX
 101                 "Device scope device [%04x:%02x:%02x.%02x] not found\n",
 102                 segment, scope->bus, path->dev, path->fn);
 103                 *dev = NULL;
 104                 return 0;
 105         }
 106         if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT && \
 107                         pdev->subordinate) || (scope->entry_type == \
 108                         ACPI_DMAR_SCOPE_TYPE_BRIDGE && !pdev->subordinate)) {
 109                 pci_dev_put(pdev);
 110                 printk(KERN_WARNING PREFIX
 111                         "Device scope type does not match for %s\n",
 112                          pci_name(pdev));
 113                 return -EINVAL;
 114         }
 115         *dev = pdev;
 116         return 0;
 117 }
 118
 119 static int __init dmar_parse_dev_scope(void *start, void *end, int *cnt,
 120                                        struct pci_dev ***devices, u16 segment)
 121 {
 122         struct acpi_dmar_device_scope *scope;
 123         void * tmp = start;
 124         int index;
 125         int ret;
 126
 127         *cnt = 0;
 128         while (start < end) {
 129                 scope = start;
 130                 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
 131                     scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE)
 132                         (*cnt)++;
 133                 else
 134                         printk(KERN_WARNING PREFIX
 135                                 "Unsupported device scope\n");
 136                 start += scope->length;
 137         }
 138         if (*cnt == 0)
 139                 return 0;
 140
 141         *devices = kcalloc(*cnt, sizeof(struct pci_dev *), GFP_KERNEL);
 142         if (!*devices)
 143                 return -ENOMEM;
 144
 145         start = tmp;
 146         index = 0;
 147         while (start < end) {
 148                 scope = start;
 149                 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
 150                     scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE) {
 151                         ret = dmar_parse_one_dev_scope(scope,
 152                                 &(*devices)[index], segment);
 153                         if (ret) {
 154                                 kfree(*devices);
 155                                 return ret;
 156                         }
 157                         index ++;
 158                 }
 159                 start += scope->length;
 160         }
 161
 162         return 0;
 163 }
 164
 165 /**
 166  * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition
 167  * structure which uniquely represent one DMA remapping hardware unit
 168  * present in the platform
 169  */
 170 static int __init
 171 dmar_parse_one_drhd(struct acpi_dmar_header *header)
 172 {
 173         struct acpi_dmar_hardware_unit *drhd;
 174         struct dmar_drhd_unit *dmaru;
 175         int ret = 0;
 176
 177         drhd = (struct acpi_dmar_hardware_unit *)header;
 178         dmaru = kzalloc(sizeof(*dmaru), GFP_KERNEL);
 179         if (!dmaru)
 180                 return -ENOMEM;
 181
 182         dmaru->hdr = header;
 183         dmaru->reg_base_addr = drhd->address;
 184         dmaru->segment = drhd->segment;
 185         dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */
 186
 187         ret = alloc_iommu(dmaru);
 188         if (ret) {
 189                 kfree(dmaru);
 190                 return ret;
 191         }
 192         dmar_register_drhd_unit(dmaru);
 193         return 0;
 194 }
 195
 196 static int __init dmar_parse_dev(struct dmar_drhd_unit *dmaru)
 197 {
 198         struct acpi_dmar_hardware_unit *drhd;
 199         int ret = 0;
 200
 201         drhd = (struct acpi_dmar_hardware_unit *) dmaru->hdr;
 202
 203         if (dmaru->include_all)
 204                 return 0;
 205
 206         ret = dmar_parse_dev_scope((void *)(drhd + 1),
 207                                 ((void *)drhd) + drhd->header.length,
 208                                 &dmaru->devices_cnt, &dmaru->devices,
 209                                 drhd->segment);
 210         if (ret) {
 211                 list_del(&dmaru->list);
 212                 kfree(dmaru);
 213         }
 214         return ret;
 215 }
 216
 217 #ifdef CONFIG_DMAR
 218 LIST_HEAD(dmar_rmrr_units);
 219
 220 static void __init dmar_register_rmrr_unit(struct dmar_rmrr_unit *rmrr)
 221 {
 222         list_add(&rmrr->list, &dmar_rmrr_units);
 223 }
 224
 225
 226 static int __init
 227 dmar_parse_one_rmrr(struct acpi_dmar_header *header)
 228 {
 229         struct acpi_dmar_reserved_memory *rmrr;
 230         struct dmar_rmrr_unit *rmrru;
 231
 232         rmrru = kzalloc(sizeof(*rmrru), GFP_KERNEL);
 233         if (!rmrru)
 234                 return -ENOMEM;
 235
 236         rmrru->hdr = header;
 237         rmrr = (struct acpi_dmar_reserved_memory *)header;
 238         rmrru->base_address = rmrr->base_address;
 239         rmrru->end_address = rmrr->end_address;
 240
 241         dmar_register_rmrr_unit(rmrru);
 242         return 0;
 243 }
 244
 245 static int __init
 246 rmrr_parse_dev(struct dmar_rmrr_unit *rmrru)
 247 {
 248         struct acpi_dmar_reserved_memory *rmrr;
 249         int ret;
 250
 251         rmrr = (struct acpi_dmar_reserved_memory *) rmrru->hdr;
 252         ret = dmar_parse_dev_scope((void *)(rmrr + 1),
 253                 ((void *)rmrr) + rmrr->header.length,
 254                 &rmrru->devices_cnt, &rmrru->devices, rmrr->segment);
 255
 256         if (ret || (rmrru->devices_cnt == 0)) {
 257                 list_del(&rmrru->list);
 258                 kfree(rmrru);
 259         }
 260         return ret;
 261 }
 262
 263 static LIST_HEAD(dmar_atsr_units);
 264
 265 static int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
 266 {
 267         struct acpi_dmar_atsr *atsr;
 268         struct dmar_atsr_unit *atsru;
 269
 270         atsr = container_of(hdr, struct acpi_dmar_atsr, header);
 271         atsru = kzalloc(sizeof(*atsru), GFP_KERNEL);
 272         if (!atsru)
 273                 return -ENOMEM;
 274
 275         atsru->hdr = hdr;
 276         atsru->include_all = atsr->flags & 0x1;
 277
 278         list_add(&atsru->list, &dmar_atsr_units);
 279
 280         return 0;
 281 }
 282
 283 static int __init atsr_parse_dev(struct dmar_atsr_unit *atsru)
 284 {
 285         int rc;
 286         struct acpi_dmar_atsr *atsr;
 287
 288         if (atsru->include_all)
 289                 return 0;
 290
 291         atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
 292         rc = dmar_parse_dev_scope((void *)(atsr + 1),
 293                                 (void *)atsr + atsr->header.length,
 294                                 &atsru->devices_cnt, &atsru->devices,
 295                                 atsr->segment);
 296         if (rc || !atsru->devices_cnt) {
 297                 list_del(&atsru->list);
 298                 kfree(atsru);
 299         }
 300
 301         return rc;
 302 }
 303
 304 int dmar_find_matched_atsr_unit(struct pci_dev *dev)
 305 {
 306         int i;
 307         struct pci_bus *bus;
 308         struct acpi_dmar_atsr *atsr;
 309         struct dmar_atsr_unit *atsru;
 310
 311         list_for_each_entry(atsru, &dmar_atsr_units, list) {
 312                 atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
 313                 if (atsr->segment == pci_domain_nr(dev->bus))
 314                         goto found;
 315         }
 316
 317         return 0;
 318
 319 found:
 320         for (bus = dev->bus; bus; bus = bus->parent) {
 321                 struct pci_dev *bridge = bus->self;
 322
 323                 if (!bridge || !bridge->is_pcie ||
 324                     bridge->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
 325                         return 0;
 326
 327                 if (bridge->pcie_type == PCI_EXP_TYPE_ROOT_PORT) {
 328                         for (i = 0; i < atsru->devices_cnt; i++)
 329                                 if (atsru->devices[i] == bridge)
 330                                         return 1;
 331                         break;
 332                 }
 333         }
 334
 335         if (atsru->include_all)
 336                 return 1;
 337
 338         return 0;
 339 }
 340 #endif
 341
 342 static void __init
 343 dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
 344 {
 345         struct acpi_dmar_hardware_unit *drhd;
 346         struct acpi_dmar_reserved_memory *rmrr;
 347         struct acpi_dmar_atsr *atsr;
 348         struct acpi_dmar_rhsa *rhsa;
 349
 350         switch (header->type) {
 351         case ACPI_DMAR_TYPE_HARDWARE_UNIT:
 352                 drhd = container_of(header, struct acpi_dmar_hardware_unit,
 353                                     header);
 354                 printk (KERN_INFO PREFIX
 355                         "DRHD base: %#016Lx flags: %#x\n",
 356                         (unsigned long long)drhd->address, drhd->flags);
 357                 break;
 358         case ACPI_DMAR_TYPE_RESERVED_MEMORY:
 359                 rmrr = container_of(header, struct acpi_dmar_reserved_memory,
 360                                     header);
 361                 printk (KERN_INFO PREFIX
 362                         "RMRR base: %#016Lx end: %#016Lx\n",
 363                         (unsigned long long)rmrr->base_address,
 364                         (unsigned long long)rmrr->end_address);
 365                 break;
 366         case ACPI_DMAR_TYPE_ATSR:
 367                 atsr = container_of(header, struct acpi_dmar_atsr, header);
 368                 printk(KERN_INFO PREFIX "ATSR flags: %#x\n", atsr->flags);
 369                 break;
 370         case ACPI_DMAR_HARDWARE_AFFINITY:
 371                 rhsa = container_of(header, struct acpi_dmar_rhsa, header);
 372                 printk(KERN_INFO PREFIX "RHSA base: %#016Lx proximity domain: %#x\n",
 373                        (unsigned long long)rhsa->base_address,
 374                        rhsa->proximity_domain);
 375                 break;
 376         }
 377 }
 378
 379 /**
 380  * dmar_table_detect - checks to see if the platform supports DMAR devices
 381  */
 382 static int __init dmar_table_detect(void)
 383 {
 384         acpi_status status = AE_OK;
 385
 386         /* if we could find DMAR table, then there are DMAR devices */
 387         status = acpi_get_table_with_size(ACPI_SIG_DMAR, 0,
 388                                 (struct acpi_table_header **)&dmar_tbl,
 389                                 &dmar_tbl_size);
 390
 391         if (ACPI_SUCCESS(status) && !dmar_tbl) {
 392                 printk (KERN_WARNING PREFIX "Unable to map DMAR\n");
 393                 status = AE_NOT_FOUND;
 394         }
 395
 396         return (ACPI_SUCCESS(status) ? 1 : 0);
 397 }
 398
 399 /**
 400  * parse_dmar_table - parses the DMA reporting table
 401  */
 402 static int __init
 403 parse_dmar_table(void)
 404 {
 405         struct acpi_table_dmar *dmar;
 406         struct acpi_dmar_header *entry_header;
 407         int ret = 0;
 408
 409         /*
 410          * Do it again, earlier dmar_tbl mapping could be mapped with
 411          * fixed map.
 412          */
 413         dmar_table_detect();
 414
 415         /*
 416          * ACPI tables may not be DMA protected by tboot, so use DMAR copy
 417          * SINIT saved in SinitMleData in TXT heap (which is DMA protected)
 418          */
 419         dmar_tbl = tboot_get_dmar_table(dmar_tbl);
 420
 421         dmar = (struct acpi_table_dmar *)dmar_tbl;
 422         if (!dmar)
 423                 return -ENODEV;
 424
 425         if (dmar->width < PAGE_SHIFT - 1) {
 426                 printk(KERN_WARNING PREFIX "Invalid DMAR haw\n");
 427                 return -EINVAL;
 428         }
 429
 430         printk (KERN_INFO PREFIX "Host address width %d\n",
 431                 dmar->width + 1);
 432
 433         entry_header = (struct acpi_dmar_header *)(dmar + 1);
 434         while (((unsigned long)entry_header) <
 435                         (((unsigned long)dmar) + dmar_tbl->length)) {
 436                 /* Avoid looping forever on bad ACPI tables */
 437                 if (entry_header->length == 0) {
 438                         printk(KERN_WARNING PREFIX
 439                                 "Invalid 0-length structure\n");
 440                         ret = -EINVAL;
 441                         break;
 442                 }
 443
 444                 dmar_table_print_dmar_entry(entry_header);
 445
 446                 switch (entry_header->type) {
 447                 case ACPI_DMAR_TYPE_HARDWARE_UNIT:
 448                         ret = dmar_parse_one_drhd(entry_header);
 449                         break;
 450                 case ACPI_DMAR_TYPE_RESERVED_MEMORY:
 451 #ifdef CONFIG_DMAR
 452                         ret = dmar_parse_one_rmrr(entry_header);
 453 #endif
 454                         break;
 455                 case ACPI_DMAR_TYPE_ATSR:
 456 #ifdef CONFIG_DMAR
 457                         ret = dmar_parse_one_atsr(entry_header);
 458 #endif
 459                         break;
 460                 case ACPI_DMAR_HARDWARE_AFFINITY:
 461                         /* We don't do anything with RHSA (yet?) */
 462                         break;
 463                 default:
 464                         printk(KERN_WARNING PREFIX
 465                                 "Unknown DMAR structure type %d\n",
 466                                 entry_header->type);
 467                         ret = 0; /* for forward compatibility */
 468                         break;
 469                 }
 470                 if (ret)
 471                         break;
 472
 473                 entry_header = ((void *)entry_header + entry_header->length);
 474         }
 475         return ret;
 476 }
 477
 478 int dmar_pci_device_match(struct pci_dev *devices[], int cnt,
 479                           struct pci_dev *dev)
 480 {
 481         int index;
 482
 483         while (dev) {
 484                 for (index = 0; index < cnt; index++)
 485                         if (dev == devices[index])
 486                                 return 1;
 487
 488                 /* Check our parent */
 489                 dev = dev->bus->self;
 490         }
 491
 492         return 0;
 493 }
 494
 495 struct dmar_drhd_unit *
 496 dmar_find_matched_drhd_unit(struct pci_dev *dev)
 497 {
 498         struct dmar_drhd_unit *dmaru = NULL;
 499         struct acpi_dmar_hardware_unit *drhd;
 500
 501         list_for_each_entry(dmaru, &dmar_drhd_units, list) {
 502                 drhd = container_of(dmaru->hdr,
 503                                     struct acpi_dmar_hardware_unit,
 504                                     header);
 505
 506                 if (dmaru->include_all &&
 507                     drhd->segment == pci_domain_nr(dev->bus))
 508                         return dmaru;
 509
 510                 if (dmar_pci_device_match(dmaru->devices,
 511                                           dmaru->devices_cnt, dev))
 512                         return dmaru;
 513         }
 514
 515         return NULL;
 516 }
 517
 518 int __init dmar_dev_scope_init(void)
 519 {
 520         struct dmar_drhd_unit *drhd, *drhd_n;
 521         int ret = -ENODEV;
 522
 523         list_for_each_entry_safe(drhd, drhd_n, &dmar_drhd_units, list) {
 524                 ret = dmar_parse_dev(drhd);
 525                 if (ret)
 526                         return ret;
 527         }
 528
 529 #ifdef CONFIG_DMAR
 530         {
 531                 struct dmar_rmrr_unit *rmrr, *rmrr_n;
 532                 struct dmar_atsr_unit *atsr, *atsr_n;
 533
 534                 list_for_each_entry_safe(rmrr, rmrr_n, &dmar_rmrr_units, list) {
 535                         ret = rmrr_parse_dev(rmrr);
 536                         if (ret)
 537                                 return ret;
 538                 }
 539
 540                 list_for_each_entry_safe(atsr, atsr_n, &dmar_atsr_units, list) {
 541                         ret = atsr_parse_dev(atsr);
 542                         if (ret)
 543                                 return ret;
 544                 }
 545         }
 546 #endif
 547
 548         return ret;
 549 }
 550
 551
 552 int __init dmar_table_init(void)
 553 {
 554         static int dmar_table_initialized;
 555         int ret;
 556
 557         if (dmar_table_initialized)
 558                 return 0;
 559
 560         dmar_table_initialized = 1;
 561
 562         ret = parse_dmar_table();
 563         if (ret) {
 564                 if (ret != -ENODEV)
 565                         printk(KERN_INFO PREFIX "parse DMAR table failure.\n");
 566                 return ret;
 567         }
 568
 569         if (list_empty(&dmar_drhd_units)) {
 570                 printk(KERN_INFO PREFIX "No DMAR devices found\n");
 571                 return -ENODEV;
 572         }
 573
 574 #ifdef CONFIG_DMAR
 575         if (list_empty(&dmar_rmrr_units))
 576                 printk(KERN_INFO PREFIX "No RMRR found\n");
 577
 578         if (list_empty(&dmar_atsr_units))
 579                 printk(KERN_INFO PREFIX "No ATSR found\n");
 580 #endif
 581
 582         return 0;
 583 }
 584
 585 int __init check_zero_address(void)
 586 {
 587         struct acpi_table_dmar *dmar;
 588         struct acpi_dmar_header *entry_header;
 589         struct acpi_dmar_hardware_unit *drhd;
 590
 591         dmar = (struct acpi_table_dmar *)dmar_tbl;
 592         entry_header = (struct acpi_dmar_header *)(dmar + 1);
 593
 594         while (((unsigned long)entry_header) <
 595                         (((unsigned long)dmar) + dmar_tbl->length)) {
 596                 /* Avoid looping forever on bad ACPI tables */
 597                 if (entry_header->length == 0) {
 598                         printk(KERN_WARNING PREFIX
 599                                 "Invalid 0-length structure\n");
 600                         return 0;
 601                 }
 602
 603                 if (entry_header->type == ACPI_DMAR_TYPE_HARDWARE_UNIT) {
 604                         drhd = (void *)entry_header;
 605                         if (!drhd->address) {
 606                                 /* Promote an attitude of violence to a BIOS engineer today */
 607                                 WARN(1, "Your BIOS is broken; DMAR reported at address zero!\n"
 608                                      "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
 609                                      dmi_get_system_info(DMI_BIOS_VENDOR),
 610                                      dmi_get_system_info(DMI_BIOS_VERSION),
 611                                      dmi_get_system_info(DMI_PRODUCT_VERSION));
 612 #ifdef CONFIG_DMAR
 613                                 dmar_disabled = 1;
 614 #endif
 615                                 return 0;
 616                         }
 617                         break;
 618                 }
 619
 620                 entry_header = ((void *)entry_header + entry_header->length);
 621         }
 622         return 1;
 623 }
 624
 625 void __init detect_intel_iommu(void)
 626 {
 627         int ret;
 628
 629         ret = dmar_table_detect();
 630         if (ret)
 631                 ret = check_zero_address();
 632         {
 633 #ifdef CONFIG_INTR_REMAP
 634                 struct acpi_table_dmar *dmar;
 635                 /*
 636                  * for now we will disable dma-remapping when interrupt
 637                  * remapping is enabled.
 638                  * When support for queued invalidation for IOTLB invalidation
 639                  * is added, we will not need this any more.
 640                  */
 641                 dmar = (struct acpi_table_dmar *) dmar_tbl;
 642                 if (ret && cpu_has_x2apic && dmar->flags & 0x1)
 643                         printk(KERN_INFO
 644                                "Queued invalidation will be enabled to support "
 645                                "x2apic and Intr-remapping.\n");
 646 #endif
 647 #ifdef CONFIG_DMAR
 648                 if (ret && !no_iommu && !iommu_detected && !dmar_disabled)
 649                         iommu_detected = 1;
 650 #endif
 651 #ifdef CONFIG_X86
 652                 if (ret)
 653                         x86_init.iommu.iommu_init = intel_iommu_init;
 654 #endif
 655         }
 656         early_acpi_os_unmap_memory(dmar_tbl, dmar_tbl_size);
 657         dmar_tbl = NULL;
 658 }
 659
 660
 661 int alloc_iommu(struct dmar_drhd_unit *drhd)
 662 {
 663         struct intel_iommu *iommu;
 664         int map_size;
 665         u32 ver;
 666         static int iommu_allocated = 0;
 667         int agaw = 0;
 668         int msagaw = 0;
 669
 670         iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
 671         if (!iommu)
 672                 return -ENOMEM;
 673
 674         iommu->seq_id = iommu_allocated++;
 675         sprintf (iommu->name, "dmar%d", iommu->seq_id);
 676
 677         iommu->reg = ioremap(drhd->reg_base_addr, VTD_PAGE_SIZE);
 678         if (!iommu->reg) {
 679                 printk(KERN_ERR "IOMMU: can't map the region\n");
 680                 goto error;
 681         }
 682         iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
 683         iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
 684
 685         if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) {
 686                 /* Promote an attitude of violence to a BIOS engineer today */
 687                 WARN(1, "Your BIOS is broken; DMAR reported at address %llx returns all ones!\n"
 688                      "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
 689                      drhd->reg_base_addr,
 690                      dmi_get_system_info(DMI_BIOS_VENDOR),
 691                      dmi_get_system_info(DMI_BIOS_VERSION),
 692                      dmi_get_system_info(DMI_PRODUCT_VERSION));
 693                 goto err_unmap;
 694         }
 695
 696 #ifdef CONFIG_DMAR
 697         agaw = iommu_calculate_agaw(iommu);
 698         if (agaw < 0) {
 699                 printk(KERN_ERR
 700                        "Cannot get a valid agaw for iommu (seq_id = %d)\n",
 701                        iommu->seq_id);
 702                 goto err_unmap;
 703         }
 704         msagaw = iommu_calculate_max_sagaw(iommu);
 705         if (msagaw < 0) {
 706                 printk(KERN_ERR
 707                         "Cannot get a valid max agaw for iommu (seq_id = %d)\n",
 708                         iommu->seq_id);
 709                 goto err_unmap;
 710         }
 711 #endif
 712         iommu->agaw = agaw;
 713         iommu->msagaw = msagaw;
 714
 715         /* the registers might be more than one page */
 716         map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
 717                 cap_max_fault_reg_offset(iommu->cap));
 718         map_size = VTD_PAGE_ALIGN(map_size);
 719         if (map_size > VTD_PAGE_SIZE) {
 720                 iounmap(iommu->reg);
 721                 iommu->reg = ioremap(drhd->reg_base_addr, map_size);
 722                 if (!iommu->reg) {
 723                         printk(KERN_ERR "IOMMU: can't map the region\n");
 724                         goto error;
 725                 }
 726         }
 727
 728         ver = readl(iommu->reg + DMAR_VER_REG);
 729         pr_info("IOMMU %llx: ver %d:%d cap %llx ecap %llx\n",
 730                 (unsigned long long)drhd->reg_base_addr,
 731                 DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
 732                 (unsigned long long)iommu->cap,
 733                 (unsigned long long)iommu->ecap);
 734
 735         spin_lock_init(&iommu->register_lock);
 736
 737         drhd->iommu = iommu;
 738         return 0;
 739
 740  err_unmap:
 741         iounmap(iommu->reg);
 742  error:
 743         kfree(iommu);
 744         return -1;
 745 }
 746
 747 void free_iommu(struct intel_iommu *iommu)
 748 {
 749         if (!iommu)
 750                 return;
 751
 752 #ifdef CONFIG_DMAR
 753         free_dmar_iommu(iommu);
 754 #endif
 755
 756         if (iommu->reg)
 757                 iounmap(iommu->reg);
 758         kfree(iommu);
 759 }
 760
 761 /*
 762  * Reclaim all the submitted descriptors which have completed its work.
 763  */
 764 static inline void reclaim_free_desc(struct q_inval *qi)
 765 {
 766         while (qi->desc_status[qi->free_tail] == QI_DONE ||
 767                qi->desc_status[qi->free_tail] == QI_ABORT) {
 768                 qi->desc_status[qi->free_tail] = QI_FREE;
 769                 qi->free_tail = (qi->free_tail + 1) % QI_LENGTH;
 770                 qi->free_cnt++;
 771         }
 772 }
 773
 774 static int qi_check_fault(struct intel_iommu *iommu, int index)
 775 {
 776         u32 fault;
 777         int head, tail;
 778         struct q_inval *qi = iommu->qi;
 779         int wait_index = (index + 1) % QI_LENGTH;
 780
 781         if (qi->desc_status[wait_index] == QI_ABORT)
 782                 return -EAGAIN;
 783
 784         fault = readl(iommu->reg + DMAR_FSTS_REG);
 785
 786         /*
 787          * If IQE happens, the head points to the descriptor associated
 788          * with the error. No new descriptors are fetched until the IQE
 789          * is cleared.
 790          */
 791         if (fault & DMA_FSTS_IQE) {
 792                 head = readl(iommu->reg + DMAR_IQH_REG);
 793                 if ((head >> DMAR_IQ_SHIFT) == index) {
 794                         printk(KERN_ERR "VT-d detected invalid descriptor: "
 795                                 "low=%llx, high=%llx\n",
 796                                 (unsigned long long)qi->desc[index].low,
 797                                 (unsigned long long)qi->desc[index].high);
 798                         memcpy(&qi->desc[index], &qi->desc[wait_index],
 799                                         sizeof(struct qi_desc));
 800                         __iommu_flush_cache(iommu, &qi->desc[index],
 801                                         sizeof(struct qi_desc));
 802                         writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG);
 803                         return -EINVAL;
 804                 }
 805         }
 806
 807         /*
 808          * If ITE happens, all pending wait_desc commands are aborted.
 809          * No new descriptors are fetched until the ITE is cleared.
 810          */
 811         if (fault & DMA_FSTS_ITE) {
 812                 head = readl(iommu->reg + DMAR_IQH_REG);
 813                 head = ((head >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
 814                 head |= 1;
 815                 tail = readl(iommu->reg + DMAR_IQT_REG);
 816                 tail = ((tail >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
 817
 818                 writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG);
 819
 820                 do {
 821                         if (qi->desc_status[head] == QI_IN_USE)
 822                                 qi->desc_status[head] = QI_ABORT;
 823                         head = (head - 2 + QI_LENGTH) % QI_LENGTH;
 824                 } while (head != tail);
 825
 826                 if (qi->desc_status[wait_index] == QI_ABORT)
 827                         return -EAGAIN;
 828         }
 829
 830         if (fault & DMA_FSTS_ICE)
 831                 writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG);
 832
 833         return 0;
 834 }
 835
 836 /*
 837  * Submit the queued invalidation descriptor to the remapping
 838  * hardware unit and wait for its completion.
 839  */
 840 int qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
 841 {
 842         int rc;
 843         struct q_inval *qi = iommu->qi;
 844         struct qi_desc *hw, wait_desc;
 845         int wait_index, index;
 846         unsigned long flags;
 847
 848         if (!qi)
 849                 return 0;
 850
 851         hw = qi->desc;
 852
 853 restart:
 854         rc = 0;
 855
 856         spin_lock_irqsave(&qi->q_lock, flags);
 857         while (qi->free_cnt < 3) {
 858                 spin_unlock_irqrestore(&qi->q_lock, flags);
 859                 cpu_relax();
 860                 spin_lock_irqsave(&qi->q_lock, flags);
 861         }
 862
 863         index = qi->free_head;
 864         wait_index = (index + 1) % QI_LENGTH;
 865
 866         qi->desc_status[index] = qi->desc_status[wait_index] = QI_IN_USE;
 867
 868         hw[index] = *desc;
 869
 870         wait_desc.low = QI_IWD_STATUS_DATA(QI_DONE) |
 871                         QI_IWD_STATUS_WRITE | QI_IWD_TYPE;
 872         wait_desc.high = virt_to_phys(&qi->desc_status[wait_index]);
 873
 874         hw[wait_index] = wait_desc;
 875
 876         __iommu_flush_cache(iommu, &hw[index], sizeof(struct qi_desc));
 877         __iommu_flush_cache(iommu, &hw[wait_index], sizeof(struct qi_desc));
 878
 879         qi->free_head = (qi->free_head + 2) % QI_LENGTH;
 880         qi->free_cnt -= 2;
 881
 882         /*
 883          * update the HW tail register indicating the presence of
 884          * new descriptors.
 885          */
 886         writel(qi->free_head << DMAR_IQ_SHIFT, iommu->reg + DMAR_IQT_REG);
 887
 888         while (qi->desc_status[wait_index] != QI_DONE) {
 889                 /*
 890                  * We will leave the interrupts disabled, to prevent interrupt
 891                  * context to queue another cmd while a cmd is already submitted
 892                  * and waiting for completion on this cpu. This is to avoid
 893                  * a deadlock where the interrupt context can wait indefinitely
 894                  * for free slots in the queue.
 895                  */
 896                 rc = qi_check_fault(iommu, index);
 897                 if (rc)
 898                         break;
 899
 900                 spin_unlock(&qi->q_lock);
 901                 cpu_relax();
 902                 spin_lock(&qi->q_lock);
 903         }
 904
 905         qi->desc_status[index] = QI_DONE;
 906
 907         reclaim_free_desc(qi);
 908         spin_unlock_irqrestore(&qi->q_lock, flags);
 909
 910         if (rc == -EAGAIN)
 911                 goto restart;
 912
 913         return rc;
 914 }
 915
 916 /*
 917  * Flush the global interrupt entry cache.
 918  */
 919 void qi_global_iec(struct intel_iommu *iommu)
 920 {
 921         struct qi_desc desc;
 922
 923         desc.low = QI_IEC_TYPE;
 924         desc.high = 0;
 925
 926         /* should never fail */
 927         qi_submit_sync(&desc, iommu);
 928 }
 929
 930 void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
 931                       u64 type)
 932 {
 933         struct qi_desc desc;
 934
 935         desc.low = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did)
 936                         | QI_CC_GRAN(type) | QI_CC_TYPE;
 937         desc.high = 0;
 938
 939         qi_submit_sync(&desc, iommu);
 940 }
 941
 942 void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
 943                     unsigned int size_order, u64 type)
 944 {
 945         u8 dw = 0, dr = 0;
 946
 947         struct qi_desc desc;
 948         int ih = 0;
 949
 950         if (cap_write_drain(iommu->cap))
 951                 dw = 1;
 952
 953         if (cap_read_drain(iommu->cap))
 954                 dr = 1;
 955
 956         desc.low = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw)
 957                 | QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE;
 958         desc.high = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih)
 959                 | QI_IOTLB_AM(size_order);
 960
 961         qi_submit_sync(&desc, iommu);
 962 }
 963
 964 void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 qdep,
 965                         u64 addr, unsigned mask)
 966 {
 967         struct qi_desc desc;
 968
 969         if (mask) {
 970                 BUG_ON(addr & ((1 << (VTD_PAGE_SHIFT + mask)) - 1));
 971                 addr |= (1 << (VTD_PAGE_SHIFT + mask - 1)) - 1;
 972                 desc.high = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE;
 973         } else
 974                 desc.high = QI_DEV_IOTLB_ADDR(addr);
 975
 976         if (qdep >= QI_DEV_IOTLB_MAX_INVS)
 977                 qdep = 0;
 978
 979         desc.low = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) |
 980                    QI_DIOTLB_TYPE;
 981
 982         qi_submit_sync(&desc, iommu);
 983 }
 984
 985 /*
 986  * Disable Queued Invalidation interface.
 987  */
 988 void dmar_disable_qi(struct intel_iommu *iommu)
 989 {
 990         unsigned long flags;
 991         u32 sts;
 992         cycles_t start_time = get_cycles();
 993
 994         if (!ecap_qis(iommu->ecap))
 995                 return;
 996
 997         spin_lock_irqsave(&iommu->register_lock, flags);
 998
 999         sts =  dmar_readq(iommu->reg + DMAR_GSTS_REG);
1000         if (!(sts & DMA_GSTS_QIES))
1001                 goto end;
1002
1003         /*
1004          * Give a chance to HW to complete the pending invalidation requests.
1005          */
1006         while ((readl(iommu->reg + DMAR_IQT_REG) !=
1007                 readl(iommu->reg + DMAR_IQH_REG)) &&
1008                 (DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time)))
1009                 cpu_relax();
1010
1011         iommu->gcmd &= ~DMA_GCMD_QIE;
1012         writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1013
1014         IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl,
1015                       !(sts & DMA_GSTS_QIES), sts);
1016 end:
1017         spin_unlock_irqrestore(&iommu->register_lock, flags);
1018 }
1019
1020 /*
1021  * Enable queued invalidation.
1022  */
1023 static void __dmar_enable_qi(struct intel_iommu *iommu)
1024 {
1025         u32 sts;
1026         unsigned long flags;
1027         struct q_inval *qi = iommu->qi;
1028
1029         qi->free_head = qi->free_tail = 0;
1030         qi->free_cnt = QI_LENGTH;
1031
1032         spin_lock_irqsave(&iommu->register_lock, flags);
1033
1034         /* write zero to the tail reg */
1035         writel(0, iommu->reg + DMAR_IQT_REG);
1036
1037         dmar_writeq(iommu->reg + DMAR_IQA_REG, virt_to_phys(qi->desc));
1038
1039         iommu->gcmd |= DMA_GCMD_QIE;
1040         writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1041
1042         /* Make sure hardware complete it */
1043         IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts);
1044
1045         spin_unlock_irqrestore(&iommu->register_lock, flags);
1046 }
1047
1048 /*
1049  * Enable Queued Invalidation interface. This is a must to support
1050  * interrupt-remapping. Also used by DMA-remapping, which replaces
1051  * register based IOTLB invalidation.
1052  */
1053 int dmar_enable_qi(struct intel_iommu *iommu)
1054 {
1055         struct q_inval *qi;
1056
1057         if (!ecap_qis(iommu->ecap))
1058                 return -ENOENT;
1059
1060         /*
1061          * queued invalidation is already setup and enabled.
1062          */
1063         if (iommu->qi)
1064                 return 0;
1065
1066         iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC);
1067         if (!iommu->qi)
1068                 return -ENOMEM;
1069
1070         qi = iommu->qi;
1071
1072         qi->desc = (void *)(get_zeroed_page(GFP_ATOMIC));
1073         if (!qi->desc) {
1074                 kfree(qi);
1075                 iommu->qi = 0;
1076                 return -ENOMEM;
1077         }
1078
1079         qi->desc_status = kmalloc(QI_LENGTH * sizeof(int), GFP_ATOMIC);
1080         if (!qi->desc_status) {
1081                 free_page((unsigned long) qi->desc);
1082                 kfree(qi);
1083                 iommu->qi = 0;
1084                 return -ENOMEM;
1085         }
1086
1087         qi->free_head = qi->free_tail = 0;
1088         qi->free_cnt = QI_LENGTH;
1089
1090         spin_lock_init(&qi->q_lock);
1091
1092         __dmar_enable_qi(iommu);
1093
1094         return 0;
1095 }
1096
1097 /* iommu interrupt handling. Most stuff are MSI-like. */
1098
1099 enum faulttype {
1100         DMA_REMAP,
1101         INTR_REMAP,
1102         UNKNOWN,
1103 };
1104
1105 static const char *dma_remap_fault_reasons[] =
1106 {
1107         "Software",
1108         "Present bit in root entry is clear",
1109         "Present bit in context entry is clear",
1110         "Invalid context entry",
1111         "Access beyond MGAW",
1112         "PTE Write access is not set",
1113         "PTE Read access is not set",
1114         "Next page table ptr is invalid",
1115         "Root table address invalid",
1116         "Context table ptr is invalid",
1117         "non-zero reserved fields in RTP",
1118         "non-zero reserved fields in CTP",
1119         "non-zero reserved fields in PTE",
1120 };
1121
1122 static const char *intr_remap_fault_reasons[] =
1123 {
1124         "Detected reserved fields in the decoded interrupt-remapped request",
1125         "Interrupt index exceeded the interrupt-remapping table size",
1126         "Present field in the IRTE entry is clear",
1127         "Error accessing interrupt-remapping table pointed by IRTA_REG",
1128         "Detected reserved fields in the IRTE entry",
1129         "Blocked a compatibility format interrupt request",
1130         "Blocked an interrupt request due to source-id verification failure",
1131 };
1132
1133 #define MAX_FAULT_REASON_IDX    (ARRAY_SIZE(fault_reason_strings) - 1)
1134
1135 const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type)
1136 {
1137         if (fault_reason >= 0x20 && (fault_reason <= 0x20 +
1138                                      ARRAY_SIZE(intr_remap_fault_reasons))) {
1139                 *fault_type = INTR_REMAP;
1140                 return intr_remap_fault_reasons[fault_reason - 0x20];
1141         } else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) {
1142                 *fault_type = DMA_REMAP;
1143                 return dma_remap_fault_reasons[fault_reason];
1144         } else {
1145                 *fault_type = UNKNOWN;
1146                 return "Unknown";
1147         }
1148 }
1149
1150 void dmar_msi_unmask(unsigned int irq)
1151 {
1152         struct intel_iommu *iommu = get_irq_data(irq);
1153         unsigned long flag;
1154
1155         /* unmask it */
1156         spin_lock_irqsave(&iommu->register_lock, flag);
1157         writel(0, iommu->reg + DMAR_FECTL_REG);
1158         /* Read a reg to force flush the post write */
1159         readl(iommu->reg + DMAR_FECTL_REG);
1160         spin_unlock_irqrestore(&iommu->register_lock, flag);
1161 }
1162
1163 void dmar_msi_mask(unsigned int irq)
1164 {
1165         unsigned long flag;
1166         struct intel_iommu *iommu = get_irq_data(irq);
1167
1168         /* mask it */
1169         spin_lock_irqsave(&iommu->register_lock, flag);
1170         writel(DMA_FECTL_IM, iommu->reg + DMAR_FECTL_REG);
1171         /* Read a reg to force flush the post write */
1172         readl(iommu->reg + DMAR_FECTL_REG);
1173         spin_unlock_irqrestore(&iommu->register_lock, flag);
1174 }
1175
1176 void dmar_msi_write(int irq, struct msi_msg *msg)
1177 {
1178         struct intel_iommu *iommu = get_irq_data(irq);
1179         unsigned long flag;
1180
1181         spin_lock_irqsave(&iommu->register_lock, flag);
1182         writel(msg->data, iommu->reg + DMAR_FEDATA_REG);
1183         writel(msg->address_lo, iommu->reg + DMAR_FEADDR_REG);
1184         writel(msg->address_hi, iommu->reg + DMAR_FEUADDR_REG);
1185         spin_unlock_irqrestore(&iommu->register_lock, flag);
1186 }
1187
1188 void dmar_msi_read(int irq, struct msi_msg *msg)
1189 {
1190         struct intel_iommu *iommu = get_irq_data(irq);
1191         unsigned long flag;
1192
1193         spin_lock_irqsave(&iommu->register_lock, flag);
1194         msg->data = readl(iommu->reg + DMAR_FEDATA_REG);
1195         msg->address_lo = readl(iommu->reg + DMAR_FEADDR_REG);
1196         msg->address_hi = readl(iommu->reg + DMAR_FEUADDR_REG);
1197         spin_unlock_irqrestore(&iommu->register_lock, flag);
1198 }
1199
1200 static int dmar_fault_do_one(struct intel_iommu *iommu, int type,
1201                 u8 fault_reason, u16 source_id, unsigned long long addr)
1202 {
1203         const char *reason;
1204         int fault_type;
1205
1206         reason = dmar_get_fault_reason(fault_reason, &fault_type);
1207
1208         if (fault_type == INTR_REMAP)
1209                 printk(KERN_ERR "INTR-REMAP: Request device [[%02x:%02x.%d] "
1210                        "fault index %llx\n"
1211                         "INTR-REMAP:[fault reason %02d] %s\n",
1212                         (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1213                         PCI_FUNC(source_id & 0xFF), addr >> 48,
1214                         fault_reason, reason);
1215         else
1216                 printk(KERN_ERR
1217                        "DMAR:[%s] Request device [%02x:%02x.%d] "
1218                        "fault addr %llx \n"
1219                        "DMAR:[fault reason %02d] %s\n",
1220                        (type ? "DMA Read" : "DMA Write"),
1221                        (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1222                        PCI_FUNC(source_id & 0xFF), addr, fault_reason, reason);
1223         return 0;
1224 }
1225
1226 #define PRIMARY_FAULT_REG_LEN (16)
1227 irqreturn_t dmar_fault(int irq, void *dev_id)
1228 {
1229         struct intel_iommu *iommu = dev_id;
1230         int reg, fault_index;
1231         u32 fault_status;
1232         unsigned long flag;
1233
1234         spin_lock_irqsave(&iommu->register_lock, flag);
1235         fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1236         if (fault_status)
1237                 printk(KERN_ERR "DRHD: handling fault status reg %x\n",
1238                        fault_status);
1239
1240         /* TBD: ignore advanced fault log currently */
1241         if (!(fault_status & DMA_FSTS_PPF))
1242                 goto clear_rest;
1243
1244         fault_index = dma_fsts_fault_record_index(fault_status);
1245         reg = cap_fault_reg_offset(iommu->cap);
1246         while (1) {
1247                 u8 fault_reason;
1248                 u16 source_id;
1249                 u64 guest_addr;
1250                 int type;
1251                 u32 data;
1252
1253                 /* highest 32 bits */
1254                 data = readl(iommu->reg + reg +
1255                                 fault_index * PRIMARY_FAULT_REG_LEN + 12);
1256                 if (!(data & DMA_FRCD_F))
1257                         break;
1258
1259                 fault_reason = dma_frcd_fault_reason(data);
1260                 type = dma_frcd_type(data);
1261
1262                 data = readl(iommu->reg + reg +
1263                                 fault_index * PRIMARY_FAULT_REG_LEN + 8);
1264                 source_id = dma_frcd_source_id(data);
1265
1266                 guest_addr = dmar_readq(iommu->reg + reg +
1267                                 fault_index * PRIMARY_FAULT_REG_LEN);
1268                 guest_addr = dma_frcd_page_addr(guest_addr);
1269                 /* clear the fault */
1270                 writel(DMA_FRCD_F, iommu->reg + reg +
1271                         fault_index * PRIMARY_FAULT_REG_LEN + 12);
1272
1273                 spin_unlock_irqrestore(&iommu->register_lock, flag);
1274
1275                 dmar_fault_do_one(iommu, type, fault_reason,
1276                                 source_id, guest_addr);
1277
1278                 fault_index++;
1279                 if (fault_index >= cap_num_fault_regs(iommu->cap))
1280                         fault_index = 0;
1281                 spin_lock_irqsave(&iommu->register_lock, flag);
1282         }
1283 clear_rest:
1284         /* clear all the other faults */
1285         fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1286         writel(fault_status, iommu->reg + DMAR_FSTS_REG);
1287
1288         spin_unlock_irqrestore(&iommu->register_lock, flag);
1289         return IRQ_HANDLED;
1290 }
1291
1292 int dmar_set_interrupt(struct intel_iommu *iommu)
1293 {
1294         int irq, ret;
1295
1296         /*
1297          * Check if the fault interrupt is already initialized.
1298          */
1299         if (iommu->irq)
1300                 return 0;
1301
1302         irq = create_irq();
1303         if (!irq) {
1304                 printk(KERN_ERR "IOMMU: no free vectors\n");
1305                 return -EINVAL;
1306         }
1307
1308         set_irq_data(irq, iommu);
1309         iommu->irq = irq;
1310
1311         ret = arch_setup_dmar_msi(irq);
1312         if (ret) {
1313                 set_irq_data(irq, NULL);
1314                 iommu->irq = 0;
1315                 destroy_irq(irq);
1316                 return ret;
1317         }
1318
1319         ret = request_irq(irq, dmar_fault, 0, iommu->name, iommu);
1320         if (ret)
1321                 printk(KERN_ERR "IOMMU: can't request irq\n");
1322         return ret;
1323 }
1324
1325 int __init enable_drhd_fault_handling(void)
1326 {
1327         struct dmar_drhd_unit *drhd;
1328
1329         /*
1330          * Enable fault control interrupt.
1331          */
1332         for_each_drhd_unit(drhd) {
1333                 int ret;
1334                 struct intel_iommu *iommu = drhd->iommu;
1335                 ret = dmar_set_interrupt(iommu);
1336
1337                 if (ret) {
1338                         printk(KERN_ERR "DRHD %Lx: failed to enable fault, "
1339                                " interrupt, ret %d\n",
1340                                (unsigned long long)drhd->reg_base_addr, ret);
1341                         return -1;
1342                 }
1343         }
1344
1345         return 0;
1346 }
1347
1348 /*
1349  * Re-enable Queued Invalidation interface.
1350  */
1351 int dmar_reenable_qi(struct intel_iommu *iommu)
1352 {
1353         if (!ecap_qis(iommu->ecap))
1354                 return -ENOENT;
1355
1356         if (!iommu->qi)
1357                 return -ENOENT;
1358
1359         /*
1360          * First disable queued invalidation.
1361          */
1362         dmar_disable_qi(iommu);
1363         /*
1364          * Then enable queued invalidation again. Since there is no pending
1365          * invalidation requests now, it's safe to re-enable queued
1366          * invalidation.
1367          */
1368         __dmar_enable_qi(iommu);
1369
1370         return 0;
1371 }
1372
1373 /*
1374  * Check interrupt remapping support in DMAR table description.
1375  */
1376 int dmar_ir_support(void)
1377 {
1378         struct acpi_table_dmar *dmar;
1379         dmar = (struct acpi_table_dmar *)dmar_tbl;
1380         return dmar->flags & 0x1;
1381 }