?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sata_sil24: Identify which card suffered IRQ status error
[linux-2.6/btrfs-unstable.git] / drivers / iommu / dmar.c
blob9a4f05e5b23fa480e308d0b44373e6599ad82982
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt /* has to precede printk.h */
30
31 #include <linux/pci.h>
32 #include <linux/dmar.h>
33 #include <linux/iova.h>
34 #include <linux/intel-iommu.h>
35 #include <linux/timer.h>
36 #include <linux/irq.h>
37 #include <linux/interrupt.h>
38 #include <linux/tboot.h>
39 #include <linux/dmi.h>
40 #include <linux/slab.h>
41 #include <asm/irq_remapping.h>
42 #include <asm/iommu_table.h>
43
44 #include "irq_remapping.h"
45
46 /*
47 * Assumptions:
48 * 1) The hotplug framework guarentees that DMAR unit will be hot-added
49 * before IO devices managed by that unit.
50 * 2) The hotplug framework guarantees that DMAR unit will be hot-removed
51 * after IO devices managed by that unit.
52 * 3) Hotplug events are rare.
53 *
54 * Locking rules for DMA and interrupt remapping related global data structures:
55 * 1) Use dmar_global_lock in process context
56 * 2) Use RCU in interrupt context
57 */
58 DECLARE_RWSEM(dmar_global_lock);
59 LIST_HEAD(dmar_drhd_units);
60
61 struct acpi_table_header * __initdata dmar_tbl;
62 static acpi_size dmar_tbl_size;
63 static int dmar_dev_scope_status = 1;
64
65 static int alloc_iommu(struct dmar_drhd_unit *drhd);
66 static void free_iommu(struct intel_iommu *iommu);
67
68 static void __init dmar_register_drhd_unit(struct dmar_drhd_unit *drhd)
69 {
70 /*
71 * add INCLUDE_ALL at the tail, so scan the list will find it at
72 * the very end.
73 */
74 if (drhd->include_all)
75 list_add_tail_rcu(&drhd->list, &dmar_drhd_units);
76 else
77 list_add_rcu(&drhd->list, &dmar_drhd_units);
78 }
79
80 void *dmar_alloc_dev_scope(void *start, void *end, int *cnt)
81 {
82 struct acpi_dmar_device_scope *scope;
83
84 *cnt = 0;
85 while (start < end) {
86 scope = start;
87 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ACPI ||
88 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
89 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE)
90 (*cnt)++;
91 else if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_IOAPIC &&
92 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_HPET) {
93 pr_warn("Unsupported device scope\n");
94 }
95 start += scope->length;
96 }
97 if (*cnt == 0)
98 return NULL;
99
100 return kcalloc(*cnt, sizeof(struct dmar_dev_scope), GFP_KERNEL);
101 }
102
103 void dmar_free_dev_scope(struct dmar_dev_scope **devices, int *cnt)
104 {
105 int i;
106 struct device *tmp_dev;
107
108 if (*devices && *cnt) {
109 for_each_active_dev_scope(*devices, *cnt, i, tmp_dev)
110 put_device(tmp_dev);
111 kfree(*devices);
112 }
113
114 *devices = NULL;
115 *cnt = 0;
116 }
117
118 /* Optimize out kzalloc()/kfree() for normal cases */
119 static char dmar_pci_notify_info_buf[64];
120
121 static struct dmar_pci_notify_info *
122 dmar_alloc_pci_notify_info(struct pci_dev *dev, unsigned long event)
123 {
124 int level = 0;
125 size_t size;
126 struct pci_dev *tmp;
127 struct dmar_pci_notify_info *info;
128
129 BUG_ON(dev->is_virtfn);
130
131 /* Only generate path[] for device addition event */
132 if (event == BUS_NOTIFY_ADD_DEVICE)
133 for (tmp = dev; tmp; tmp = tmp->bus->self)
134 level++;
135
136 size = sizeof(*info) + level * sizeof(struct acpi_dmar_pci_path);
137 if (size <= sizeof(dmar_pci_notify_info_buf)) {
138 info = (struct dmar_pci_notify_info *)dmar_pci_notify_info_buf;
139 } else {
140 info = kzalloc(size, GFP_KERNEL);
141 if (!info) {
142 pr_warn("Out of memory when allocating notify_info "
143 "for %s.\n", pci_name(dev));
144 if (dmar_dev_scope_status == 0)
145 dmar_dev_scope_status = -ENOMEM;
146 return NULL;
147 }
148 }
149
150 info->event = event;
151 info->dev = dev;
152 info->seg = pci_domain_nr(dev->bus);
153 info->level = level;
154 if (event == BUS_NOTIFY_ADD_DEVICE) {
155 for (tmp = dev; tmp; tmp = tmp->bus->self) {
156 level--;
157 info->path[level].device = PCI_SLOT(tmp->devfn);
158 info->path[level].function = PCI_FUNC(tmp->devfn);
159 if (pci_is_root_bus(tmp->bus))
160 info->bus = tmp->bus->number;
161 }
162 }
163
164 return info;
165 }
166
167 static inline void dmar_free_pci_notify_info(struct dmar_pci_notify_info *info)
168 {
169 if ((void *)info != dmar_pci_notify_info_buf)
170 kfree(info);
171 }
172
173 static bool dmar_match_pci_path(struct dmar_pci_notify_info *info, int bus,
174 struct acpi_dmar_pci_path *path, int count)
175 {
176 int i;
177
178 if (info->bus != bus)
179 return false;
180 if (info->level != count)
181 return false;
182
183 for (i = 0; i < count; i++) {
184 if (path[i].device != info->path[i].device ||
185 path[i].function != info->path[i].function)
186 return false;
187 }
188
189 return true;
190 }
191
192 /* Return: > 0 if match found, 0 if no match found, < 0 if error happens */
193 int dmar_insert_dev_scope(struct dmar_pci_notify_info *info,
194 void *start, void*end, u16 segment,
195 struct dmar_dev_scope *devices,
196 int devices_cnt)
197 {
198 int i, level;
199 struct device *tmp, *dev = &info->dev->dev;
200 struct acpi_dmar_device_scope *scope;
201 struct acpi_dmar_pci_path *path;
202
203 if (segment != info->seg)
204 return 0;
205
206 for (; start < end; start += scope->length) {
207 scope = start;
208 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_ENDPOINT &&
209 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_BRIDGE)
210 continue;
211
212 path = (struct acpi_dmar_pci_path *)(scope + 1);
213 level = (scope->length - sizeof(*scope)) / sizeof(*path);
214 if (!dmar_match_pci_path(info, scope->bus, path, level))
215 continue;
216
217 if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT) ^
218 (info->dev->hdr_type == PCI_HEADER_TYPE_NORMAL)) {
219 pr_warn("Device scope type does not match for %s\n",
220 pci_name(info->dev));
221 return -EINVAL;
222 }
223
224 for_each_dev_scope(devices, devices_cnt, i, tmp)
225 if (tmp == NULL) {
226 devices[i].bus = info->dev->bus->number;
227 devices[i].devfn = info->dev->devfn;
228 rcu_assign_pointer(devices[i].dev,
229 get_device(dev));
230 return 1;
231 }
232 BUG_ON(i >= devices_cnt);
233 }
234
235 return 0;
236 }
237
238 int dmar_remove_dev_scope(struct dmar_pci_notify_info *info, u16 segment,
239 struct dmar_dev_scope *devices, int count)
240 {
241 int index;
242 struct device *tmp;
243
244 if (info->seg != segment)
245 return 0;
246
247 for_each_active_dev_scope(devices, count, index, tmp)
248 if (tmp == &info->dev->dev) {
249 rcu_assign_pointer(devices[index].dev, NULL);
250 synchronize_rcu();
251 put_device(tmp);
252 return 1;
253 }
254
255 return 0;
256 }
257
258 static int dmar_pci_bus_add_dev(struct dmar_pci_notify_info *info)
259 {
260 int ret = 0;
261 struct dmar_drhd_unit *dmaru;
262 struct acpi_dmar_hardware_unit *drhd;
263
264 for_each_drhd_unit(dmaru) {
265 if (dmaru->include_all)
266 continue;
267
268 drhd = container_of(dmaru->hdr,
269 struct acpi_dmar_hardware_unit, header);
270 ret = dmar_insert_dev_scope(info, (void *)(drhd + 1),
271 ((void *)drhd) + drhd->header.length,
272 dmaru->segment,
273 dmaru->devices, dmaru->devices_cnt);
274 if (ret != 0)
275 break;
276 }
277 if (ret >= 0)
278 ret = dmar_iommu_notify_scope_dev(info);
279 if (ret < 0 && dmar_dev_scope_status == 0)
280 dmar_dev_scope_status = ret;
281
282 return ret;
283 }
284
285 static void dmar_pci_bus_del_dev(struct dmar_pci_notify_info *info)
286 {
287 struct dmar_drhd_unit *dmaru;
288
289 for_each_drhd_unit(dmaru)
290 if (dmar_remove_dev_scope(info, dmaru->segment,
291 dmaru->devices, dmaru->devices_cnt))
292 break;
293 dmar_iommu_notify_scope_dev(info);
294 }
295
296 static int dmar_pci_bus_notifier(struct notifier_block *nb,
297 unsigned long action, void *data)
298 {
299 struct pci_dev *pdev = to_pci_dev(data);
300 struct dmar_pci_notify_info *info;
301
302 /* Only care about add/remove events for physical functions */
303 if (pdev->is_virtfn)
304 return NOTIFY_DONE;
305 if (action != BUS_NOTIFY_ADD_DEVICE && action != BUS_NOTIFY_DEL_DEVICE)
306 return NOTIFY_DONE;
307
308 info = dmar_alloc_pci_notify_info(pdev, action);
309 if (!info)
310 return NOTIFY_DONE;
311
312 down_write(&dmar_global_lock);
313 if (action == BUS_NOTIFY_ADD_DEVICE)
314 dmar_pci_bus_add_dev(info);
315 else if (action == BUS_NOTIFY_DEL_DEVICE)
316 dmar_pci_bus_del_dev(info);
317 up_write(&dmar_global_lock);
318
319 dmar_free_pci_notify_info(info);
320
321 return NOTIFY_OK;
322 }
323
324 static struct notifier_block dmar_pci_bus_nb = {
325 .notifier_call = dmar_pci_bus_notifier,
326 .priority = INT_MIN,
327 };
328
329 /**
330 * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition
331 * structure which uniquely represent one DMA remapping hardware unit
332 * present in the platform
333 */
334 static int __init
335 dmar_parse_one_drhd(struct acpi_dmar_header *header)
336 {
337 struct acpi_dmar_hardware_unit *drhd;
338 struct dmar_drhd_unit *dmaru;
339 int ret = 0;
340
341 drhd = (struct acpi_dmar_hardware_unit *)header;
342 dmaru = kzalloc(sizeof(*dmaru), GFP_KERNEL);
343 if (!dmaru)
344 return -ENOMEM;
345
346 dmaru->hdr = header;
347 dmaru->reg_base_addr = drhd->address;
348 dmaru->segment = drhd->segment;
349 dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */
350 dmaru->devices = dmar_alloc_dev_scope((void *)(drhd + 1),
351 ((void *)drhd) + drhd->header.length,
352 &dmaru->devices_cnt);
353 if (dmaru->devices_cnt && dmaru->devices == NULL) {
354 kfree(dmaru);
355 return -ENOMEM;
356 }
357
358 ret = alloc_iommu(dmaru);
359 if (ret) {
360 dmar_free_dev_scope(&dmaru->devices,
361 &dmaru->devices_cnt);
362 kfree(dmaru);
363 return ret;
364 }
365 dmar_register_drhd_unit(dmaru);
366 return 0;
367 }
368
369 static void dmar_free_drhd(struct dmar_drhd_unit *dmaru)
370 {
371 if (dmaru->devices && dmaru->devices_cnt)
372 dmar_free_dev_scope(&dmaru->devices, &dmaru->devices_cnt);
373 if (dmaru->iommu)
374 free_iommu(dmaru->iommu);
375 kfree(dmaru);
376 }
377
378 static int __init dmar_parse_one_andd(struct acpi_dmar_header *header)
379 {
380 struct acpi_dmar_andd *andd = (void *)header;
381
382 /* Check for NUL termination within the designated length */
383 if (strnlen(andd->object_name, header->length - 8) == header->length - 8) {
384 WARN_TAINT(1, TAINT_FIRMWARE_WORKAROUND,
385 "Your BIOS is broken; ANDD object name is not NUL-terminated\n"
386 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
387 dmi_get_system_info(DMI_BIOS_VENDOR),
388 dmi_get_system_info(DMI_BIOS_VERSION),
389 dmi_get_system_info(DMI_PRODUCT_VERSION));
390 return -EINVAL;
391 }
392 pr_info("ANDD device: %x name: %s\n", andd->device_number,
393 andd->object_name);
394
395 return 0;
396 }
397
398 #ifdef CONFIG_ACPI_NUMA
399 static int __init
400 dmar_parse_one_rhsa(struct acpi_dmar_header *header)
401 {
402 struct acpi_dmar_rhsa *rhsa;
403 struct dmar_drhd_unit *drhd;
404
405 rhsa = (struct acpi_dmar_rhsa *)header;
406 for_each_drhd_unit(drhd) {
407 if (drhd->reg_base_addr == rhsa->base_address) {
408 int node = acpi_map_pxm_to_node(rhsa->proximity_domain);
409
410 if (!node_online(node))
411 node = -1;
412 drhd->iommu->node = node;
413 return 0;
414 }
415 }
416 WARN_TAINT(
417 1, TAINT_FIRMWARE_WORKAROUND,
418 "Your BIOS is broken; RHSA refers to non-existent DMAR unit at %llx\n"
419 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
420 drhd->reg_base_addr,
421 dmi_get_system_info(DMI_BIOS_VENDOR),
422 dmi_get_system_info(DMI_BIOS_VERSION),
423 dmi_get_system_info(DMI_PRODUCT_VERSION));
424
425 return 0;
426 }
427 #endif
428
429 static void __init
430 dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
431 {
432 struct acpi_dmar_hardware_unit *drhd;
433 struct acpi_dmar_reserved_memory *rmrr;
434 struct acpi_dmar_atsr *atsr;
435 struct acpi_dmar_rhsa *rhsa;
436
437 switch (header->type) {
438 case ACPI_DMAR_TYPE_HARDWARE_UNIT:
439 drhd = container_of(header, struct acpi_dmar_hardware_unit,
440 header);
441 pr_info("DRHD base: %#016Lx flags: %#x\n",
442 (unsigned long long)drhd->address, drhd->flags);
443 break;
444 case ACPI_DMAR_TYPE_RESERVED_MEMORY:
445 rmrr = container_of(header, struct acpi_dmar_reserved_memory,
446 header);
447 pr_info("RMRR base: %#016Lx end: %#016Lx\n",
448 (unsigned long long)rmrr->base_address,
449 (unsigned long long)rmrr->end_address);
450 break;
451 case ACPI_DMAR_TYPE_ATSR:
452 atsr = container_of(header, struct acpi_dmar_atsr, header);
453 pr_info("ATSR flags: %#x\n", atsr->flags);
454 break;
455 case ACPI_DMAR_HARDWARE_AFFINITY:
456 rhsa = container_of(header, struct acpi_dmar_rhsa, header);
457 pr_info("RHSA base: %#016Lx proximity domain: %#x\n",
458 (unsigned long long)rhsa->base_address,
459 rhsa->proximity_domain);
460 break;
461 case ACPI_DMAR_TYPE_ANDD:
462 /* We don't print this here because we need to sanity-check
463 it first. So print it in dmar_parse_one_andd() instead. */
464 break;
465 }
466 }
467
468 /**
469 * dmar_table_detect - checks to see if the platform supports DMAR devices
470 */
471 static int __init dmar_table_detect(void)
472 {
473 acpi_status status = AE_OK;
474
475 /* if we could find DMAR table, then there are DMAR devices */
476 status = acpi_get_table_with_size(ACPI_SIG_DMAR, 0,
477 (struct acpi_table_header **)&dmar_tbl,
478 &dmar_tbl_size);
479
480 if (ACPI_SUCCESS(status) && !dmar_tbl) {
481 pr_warn("Unable to map DMAR\n");
482 status = AE_NOT_FOUND;
483 }
484
485 return (ACPI_SUCCESS(status) ? 1 : 0);
486 }
487
488 /**
489 * parse_dmar_table - parses the DMA reporting table
490 */
491 static int __init
492 parse_dmar_table(void)
493 {
494 struct acpi_table_dmar *dmar;
495 struct acpi_dmar_header *entry_header;
496 int ret = 0;
497 int drhd_count = 0;
498
499 /*
500 * Do it again, earlier dmar_tbl mapping could be mapped with
501 * fixed map.
502 */
503 dmar_table_detect();
504
505 /*
506 * ACPI tables may not be DMA protected by tboot, so use DMAR copy
507 * SINIT saved in SinitMleData in TXT heap (which is DMA protected)
508 */
509 dmar_tbl = tboot_get_dmar_table(dmar_tbl);
510
511 dmar = (struct acpi_table_dmar *)dmar_tbl;
512 if (!dmar)
513 return -ENODEV;
514
515 if (dmar->width < PAGE_SHIFT - 1) {
516 pr_warn("Invalid DMAR haw\n");
517 return -EINVAL;
518 }
519
520 pr_info("Host address width %d\n", dmar->width + 1);
521
522 entry_header = (struct acpi_dmar_header *)(dmar + 1);
523 while (((unsigned long)entry_header) <
524 (((unsigned long)dmar) + dmar_tbl->length)) {
525 /* Avoid looping forever on bad ACPI tables */
526 if (entry_header->length == 0) {
527 pr_warn("Invalid 0-length structure\n");
528 ret = -EINVAL;
529 break;
530 }
531
532 dmar_table_print_dmar_entry(entry_header);
533
534 switch (entry_header->type) {
535 case ACPI_DMAR_TYPE_HARDWARE_UNIT:
536 drhd_count++;
537 ret = dmar_parse_one_drhd(entry_header);
538 break;
539 case ACPI_DMAR_TYPE_RESERVED_MEMORY:
540 ret = dmar_parse_one_rmrr(entry_header);
541 break;
542 case ACPI_DMAR_TYPE_ATSR:
543 ret = dmar_parse_one_atsr(entry_header);
544 break;
545 case ACPI_DMAR_HARDWARE_AFFINITY:
546 #ifdef CONFIG_ACPI_NUMA
547 ret = dmar_parse_one_rhsa(entry_header);
548 #endif
549 break;
550 case ACPI_DMAR_TYPE_ANDD:
551 ret = dmar_parse_one_andd(entry_header);
552 break;
553 default:
554 pr_warn("Unknown DMAR structure type %d\n",
555 entry_header->type);
556 ret = 0; /* for forward compatibility */
557 break;
558 }
559 if (ret)
560 break;
561
562 entry_header = ((void *)entry_header + entry_header->length);
563 }
564 if (drhd_count == 0)
565 pr_warn(FW_BUG "No DRHD structure found in DMAR table\n");
566 return ret;
567 }
568
569 static int dmar_pci_device_match(struct dmar_dev_scope devices[],
570 int cnt, struct pci_dev *dev)
571 {
572 int index;
573 struct device *tmp;
574
575 while (dev) {
576 for_each_active_dev_scope(devices, cnt, index, tmp)
577 if (dev_is_pci(tmp) && dev == to_pci_dev(tmp))
578 return 1;
579
580 /* Check our parent */
581 dev = dev->bus->self;
582 }
583
584 return 0;
585 }
586
587 struct dmar_drhd_unit *
588 dmar_find_matched_drhd_unit(struct pci_dev *dev)
589 {
590 struct dmar_drhd_unit *dmaru;
591 struct acpi_dmar_hardware_unit *drhd;
592
593 dev = pci_physfn(dev);
594
595 rcu_read_lock();
596 for_each_drhd_unit(dmaru) {
597 drhd = container_of(dmaru->hdr,
598 struct acpi_dmar_hardware_unit,
599 header);
600
601 if (dmaru->include_all &&
602 drhd->segment == pci_domain_nr(dev->bus))
603 goto out;
604
605 if (dmar_pci_device_match(dmaru->devices,
606 dmaru->devices_cnt, dev))
607 goto out;
608 }
609 dmaru = NULL;
610 out:
611 rcu_read_unlock();
612
613 return dmaru;
614 }
615
616 static void __init dmar_acpi_insert_dev_scope(u8 device_number,
617 struct acpi_device *adev)
618 {
619 struct dmar_drhd_unit *dmaru;
620 struct acpi_dmar_hardware_unit *drhd;
621 struct acpi_dmar_device_scope *scope;
622 struct device *tmp;
623 int i;
624 struct acpi_dmar_pci_path *path;
625
626 for_each_drhd_unit(dmaru) {
627 drhd = container_of(dmaru->hdr,
628 struct acpi_dmar_hardware_unit,
629 header);
630
631 for (scope = (void *)(drhd + 1);
632 (unsigned long)scope < ((unsigned long)drhd) + drhd->header.length;
633 scope = ((void *)scope) + scope->length) {
634 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_ACPI)
635 continue;
636 if (scope->enumeration_id != device_number)
637 continue;
638
639 path = (void *)(scope + 1);
640 pr_info("ACPI device \"%s\" under DMAR at %llx as %02x:%02x.%d\n",
641 dev_name(&adev->dev), dmaru->reg_base_addr,
642 scope->bus, path->device, path->function);
643 for_each_dev_scope(dmaru->devices, dmaru->devices_cnt, i, tmp)
644 if (tmp == NULL) {
645 dmaru->devices[i].bus = scope->bus;
646 dmaru->devices[i].devfn = PCI_DEVFN(path->device,
647 path->function);
648 rcu_assign_pointer(dmaru->devices[i].dev,
649 get_device(&adev->dev));
650 return;
651 }
652 BUG_ON(i >= dmaru->devices_cnt);
653 }
654 }
655 pr_warn("No IOMMU scope found for ANDD enumeration ID %d (%s)\n",
656 device_number, dev_name(&adev->dev));
657 }
658
659 static int __init dmar_acpi_dev_scope_init(void)
660 {
661 struct acpi_dmar_andd *andd;
662
663 if (dmar_tbl == NULL)
664 return -ENODEV;
665
666 for (andd = (void *)dmar_tbl + sizeof(struct acpi_table_dmar);
667 ((unsigned long)andd) < ((unsigned long)dmar_tbl) + dmar_tbl->length;
668 andd = ((void *)andd) + andd->header.length) {
669 if (andd->header.type == ACPI_DMAR_TYPE_ANDD) {
670 acpi_handle h;
671 struct acpi_device *adev;
672
673 if (!ACPI_SUCCESS(acpi_get_handle(ACPI_ROOT_OBJECT,
674 andd->object_name,
675 &h))) {
676 pr_err("Failed to find handle for ACPI object %s\n",
677 andd->object_name);
678 continue;
679 }
680 acpi_bus_get_device(h, &adev);
681 if (!adev) {
682 pr_err("Failed to get device for ACPI object %s\n",
683 andd->object_name);
684 continue;
685 }
686 dmar_acpi_insert_dev_scope(andd->device_number, adev);
687 }
688 }
689 return 0;
690 }
691
692 int __init dmar_dev_scope_init(void)
693 {
694 struct pci_dev *dev = NULL;
695 struct dmar_pci_notify_info *info;
696
697 if (dmar_dev_scope_status != 1)
698 return dmar_dev_scope_status;
699
700 if (list_empty(&dmar_drhd_units)) {
701 dmar_dev_scope_status = -ENODEV;
702 } else {
703 dmar_dev_scope_status = 0;
704
705 dmar_acpi_dev_scope_init();
706
707 for_each_pci_dev(dev) {
708 if (dev->is_virtfn)
709 continue;
710
711 info = dmar_alloc_pci_notify_info(dev,
712 BUS_NOTIFY_ADD_DEVICE);
713 if (!info) {
714 return dmar_dev_scope_status;
715 } else {
716 dmar_pci_bus_add_dev(info);
717 dmar_free_pci_notify_info(info);
718 }
719 }
720
721 bus_register_notifier(&pci_bus_type, &dmar_pci_bus_nb);
722 }
723
724 return dmar_dev_scope_status;
725 }
726
727
728 int __init dmar_table_init(void)
729 {
730 static int dmar_table_initialized;
731 int ret;
732
733 if (dmar_table_initialized == 0) {
734 ret = parse_dmar_table();
735 if (ret < 0) {
736 if (ret != -ENODEV)
737 pr_info("parse DMAR table failure.\n");
738 } else if (list_empty(&dmar_drhd_units)) {
739 pr_info("No DMAR devices found\n");
740 ret = -ENODEV;
741 }
742
743 if (ret < 0)
744 dmar_table_initialized = ret;
745 else
746 dmar_table_initialized = 1;
747 }
748
749 return dmar_table_initialized < 0 ? dmar_table_initialized : 0;
750 }
751
752 static void warn_invalid_dmar(u64 addr, const char *message)
753 {
754 WARN_TAINT_ONCE(
755 1, TAINT_FIRMWARE_WORKAROUND,
756 "Your BIOS is broken; DMAR reported at address %llx%s!\n"
757 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
758 addr, message,
759 dmi_get_system_info(DMI_BIOS_VENDOR),
760 dmi_get_system_info(DMI_BIOS_VERSION),
761 dmi_get_system_info(DMI_PRODUCT_VERSION));
762 }
763
764 static int __init check_zero_address(void)
765 {
766 struct acpi_table_dmar *dmar;
767 struct acpi_dmar_header *entry_header;
768 struct acpi_dmar_hardware_unit *drhd;
769
770 dmar = (struct acpi_table_dmar *)dmar_tbl;
771 entry_header = (struct acpi_dmar_header *)(dmar + 1);
772
773 while (((unsigned long)entry_header) <
774 (((unsigned long)dmar) + dmar_tbl->length)) {
775 /* Avoid looping forever on bad ACPI tables */
776 if (entry_header->length == 0) {
777 pr_warn("Invalid 0-length structure\n");
778 return 0;
779 }
780
781 if (entry_header->type == ACPI_DMAR_TYPE_HARDWARE_UNIT) {
782 void __iomem *addr;
783 u64 cap, ecap;
784
785 drhd = (void *)entry_header;
786 if (!drhd->address) {
787 warn_invalid_dmar(0, "");
788 goto failed;
789 }
790
791 addr = early_ioremap(drhd->address, VTD_PAGE_SIZE);
792 if (!addr ) {
793 printk("IOMMU: can't validate: %llx\n", drhd->address);
794 goto failed;
795 }
796 cap = dmar_readq(addr + DMAR_CAP_REG);
797 ecap = dmar_readq(addr + DMAR_ECAP_REG);
798 early_iounmap(addr, VTD_PAGE_SIZE);
799 if (cap == (uint64_t)-1 && ecap == (uint64_t)-1) {
800 warn_invalid_dmar(drhd->address,
801 " returns all ones");
802 goto failed;
803 }
804 }
805
806 entry_header = ((void *)entry_header + entry_header->length);
807 }
808 return 1;
809
810 failed:
811 return 0;
812 }
813
814 int __init detect_intel_iommu(void)
815 {
816 int ret;
817
818 down_write(&dmar_global_lock);
819 ret = dmar_table_detect();
820 if (ret)
821 ret = check_zero_address();
822 {
823 if (ret && !no_iommu && !iommu_detected && !dmar_disabled) {
824 iommu_detected = 1;
825 /* Make sure ACS will be enabled */
826 pci_request_acs();
827 }
828
829 #ifdef CONFIG_X86
830 if (ret)
831 x86_init.iommu.iommu_init = intel_iommu_init;
832 #endif
833 }
834 early_acpi_os_unmap_memory((void __iomem *)dmar_tbl, dmar_tbl_size);
835 dmar_tbl = NULL;
836 up_write(&dmar_global_lock);
837
838 return ret ? 1 : -ENODEV;
839 }
840
841
842 static void unmap_iommu(struct intel_iommu *iommu)
843 {
844 iounmap(iommu->reg);
845 release_mem_region(iommu->reg_phys, iommu->reg_size);
846 }
847
848 /**
849 * map_iommu: map the iommu's registers
850 * @iommu: the iommu to map
851 * @phys_addr: the physical address of the base resgister
852 *
853 * Memory map the iommu's registers. Start w/ a single page, and
854 * possibly expand if that turns out to be insufficent.
855 */
856 static int map_iommu(struct intel_iommu *iommu, u64 phys_addr)
857 {
858 int map_size, err=0;
859
860 iommu->reg_phys = phys_addr;
861 iommu->reg_size = VTD_PAGE_SIZE;
862
863 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, iommu->name)) {
864 pr_err("IOMMU: can't reserve memory\n");
865 err = -EBUSY;
866 goto out;
867 }
868
869 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size);
870 if (!iommu->reg) {
871 pr_err("IOMMU: can't map the region\n");
872 err = -ENOMEM;
873 goto release;
874 }
875
876 iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
877 iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
878
879 if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) {
880 err = -EINVAL;
881 warn_invalid_dmar(phys_addr, " returns all ones");
882 goto unmap;
883 }
884
885 /* the registers might be more than one page */
886 map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
887 cap_max_fault_reg_offset(iommu->cap));
888 map_size = VTD_PAGE_ALIGN(map_size);
889 if (map_size > iommu->reg_size) {
890 iounmap(iommu->reg);
891 release_mem_region(iommu->reg_phys, iommu->reg_size);
892 iommu->reg_size = map_size;
893 if (!request_mem_region(iommu->reg_phys, iommu->reg_size,
894 iommu->name)) {
895 pr_err("IOMMU: can't reserve memory\n");
896 err = -EBUSY;
897 goto out;
898 }
899 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size);
900 if (!iommu->reg) {
901 pr_err("IOMMU: can't map the region\n");
902 err = -ENOMEM;
903 goto release;
904 }
905 }
906 err = 0;
907 goto out;
908
909 unmap:
910 iounmap(iommu->reg);
911 release:
912 release_mem_region(iommu->reg_phys, iommu->reg_size);
913 out:
914 return err;
915 }
916
917 static int alloc_iommu(struct dmar_drhd_unit *drhd)
918 {
919 struct intel_iommu *iommu;
920 u32 ver, sts;
921 static int iommu_allocated = 0;
922 int agaw = 0;
923 int msagaw = 0;
924 int err;
925
926 if (!drhd->reg_base_addr) {
927 warn_invalid_dmar(0, "");
928 return -EINVAL;
929 }
930
931 iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
932 if (!iommu)
933 return -ENOMEM;
934
935 iommu->seq_id = iommu_allocated++;
936 sprintf (iommu->name, "dmar%d", iommu->seq_id);
937
938 err = map_iommu(iommu, drhd->reg_base_addr);
939 if (err) {
940 pr_err("IOMMU: failed to map %s\n", iommu->name);
941 goto error;
942 }
943
944 err = -EINVAL;
945 agaw = iommu_calculate_agaw(iommu);
946 if (agaw < 0) {
947 pr_err("Cannot get a valid agaw for iommu (seq_id = %d)\n",
948 iommu->seq_id);
949 goto err_unmap;
950 }
951 msagaw = iommu_calculate_max_sagaw(iommu);
952 if (msagaw < 0) {
953 pr_err("Cannot get a valid max agaw for iommu (seq_id = %d)\n",
954 iommu->seq_id);
955 goto err_unmap;
956 }
957 iommu->agaw = agaw;
958 iommu->msagaw = msagaw;
959 iommu->segment = drhd->segment;
960
961 iommu->node = -1;
962
963 ver = readl(iommu->reg + DMAR_VER_REG);
964 pr_info("IOMMU %d: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n",
965 iommu->seq_id,
966 (unsigned long long)drhd->reg_base_addr,
967 DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
968 (unsigned long long)iommu->cap,
969 (unsigned long long)iommu->ecap);
970
971 /* Reflect status in gcmd */
972 sts = readl(iommu->reg + DMAR_GSTS_REG);
973 if (sts & DMA_GSTS_IRES)
974 iommu->gcmd |= DMA_GCMD_IRE;
975 if (sts & DMA_GSTS_TES)
976 iommu->gcmd |= DMA_GCMD_TE;
977 if (sts & DMA_GSTS_QIES)
978 iommu->gcmd |= DMA_GCMD_QIE;
979
980 raw_spin_lock_init(&iommu->register_lock);
981
982 drhd->iommu = iommu;
983 return 0;
984
985 err_unmap:
986 unmap_iommu(iommu);
987 error:
988 kfree(iommu);
989 return err;
990 }
991
992 static void free_iommu(struct intel_iommu *iommu)
993 {
994 if (iommu->irq) {
995 free_irq(iommu->irq, iommu);
996 irq_set_handler_data(iommu->irq, NULL);
997 dmar_free_hwirq(iommu->irq);
998 }
999
1000 if (iommu->qi) {
1001 free_page((unsigned long)iommu->qi->desc);
1002 kfree(iommu->qi->desc_status);
1003 kfree(iommu->qi);
1004 }
1005
1006 if (iommu->reg)
1007 unmap_iommu(iommu);
1008
1009 kfree(iommu);
1010 }
1011
1012 /*
1013 * Reclaim all the submitted descriptors which have completed its work.
1014 */
1015 static inline void reclaim_free_desc(struct q_inval *qi)
1016 {
1017 while (qi->desc_status[qi->free_tail] == QI_DONE ||
1018 qi->desc_status[qi->free_tail] == QI_ABORT) {
1019 qi->desc_status[qi->free_tail] = QI_FREE;
1020 qi->free_tail = (qi->free_tail + 1) % QI_LENGTH;
1021 qi->free_cnt++;
1022 }
1023 }
1024
1025 static int qi_check_fault(struct intel_iommu *iommu, int index)
1026 {
1027 u32 fault;
1028 int head, tail;
1029 struct q_inval *qi = iommu->qi;
1030 int wait_index = (index + 1) % QI_LENGTH;
1031
1032 if (qi->desc_status[wait_index] == QI_ABORT)
1033 return -EAGAIN;
1034
1035 fault = readl(iommu->reg + DMAR_FSTS_REG);
1036
1037 /*
1038 * If IQE happens, the head points to the descriptor associated
1039 * with the error. No new descriptors are fetched until the IQE
1040 * is cleared.
1041 */
1042 if (fault & DMA_FSTS_IQE) {
1043 head = readl(iommu->reg + DMAR_IQH_REG);
1044 if ((head >> DMAR_IQ_SHIFT) == index) {
1045 pr_err("VT-d detected invalid descriptor: "
1046 "low=%llx, high=%llx\n",
1047 (unsigned long long)qi->desc[index].low,
1048 (unsigned long long)qi->desc[index].high);
1049 memcpy(&qi->desc[index], &qi->desc[wait_index],
1050 sizeof(struct qi_desc));
1051 __iommu_flush_cache(iommu, &qi->desc[index],
1052 sizeof(struct qi_desc));
1053 writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG);
1054 return -EINVAL;
1055 }
1056 }
1057
1058 /*
1059 * If ITE happens, all pending wait_desc commands are aborted.
1060 * No new descriptors are fetched until the ITE is cleared.
1061 */
1062 if (fault & DMA_FSTS_ITE) {
1063 head = readl(iommu->reg + DMAR_IQH_REG);
1064 head = ((head >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
1065 head |= 1;
1066 tail = readl(iommu->reg + DMAR_IQT_REG);
1067 tail = ((tail >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
1068
1069 writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG);
1070
1071 do {
1072 if (qi->desc_status[head] == QI_IN_USE)
1073 qi->desc_status[head] = QI_ABORT;
1074 head = (head - 2 + QI_LENGTH) % QI_LENGTH;
1075 } while (head != tail);
1076
1077 if (qi->desc_status[wait_index] == QI_ABORT)
1078 return -EAGAIN;
1079 }
1080
1081 if (fault & DMA_FSTS_ICE)
1082 writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG);
1083
1084 return 0;
1085 }
1086
1087 /*
1088 * Submit the queued invalidation descriptor to the remapping
1089 * hardware unit and wait for its completion.
1090 */
1091 int qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
1092 {
1093 int rc;
1094 struct q_inval *qi = iommu->qi;
1095 struct qi_desc *hw, wait_desc;
1096 int wait_index, index;
1097 unsigned long flags;
1098
1099 if (!qi)
1100 return 0;
1101
1102 hw = qi->desc;
1103
1104 restart:
1105 rc = 0;
1106
1107 raw_spin_lock_irqsave(&qi->q_lock, flags);
1108 while (qi->free_cnt < 3) {
1109 raw_spin_unlock_irqrestore(&qi->q_lock, flags);
1110 cpu_relax();
1111 raw_spin_lock_irqsave(&qi->q_lock, flags);
1112 }
1113
1114 index = qi->free_head;
1115 wait_index = (index + 1) % QI_LENGTH;
1116
1117 qi->desc_status[index] = qi->desc_status[wait_index] = QI_IN_USE;
1118
1119 hw[index] = *desc;
1120
1121 wait_desc.low = QI_IWD_STATUS_DATA(QI_DONE) |
1122 QI_IWD_STATUS_WRITE | QI_IWD_TYPE;
1123 wait_desc.high = virt_to_phys(&qi->desc_status[wait_index]);
1124
1125 hw[wait_index] = wait_desc;
1126
1127 __iommu_flush_cache(iommu, &hw[index], sizeof(struct qi_desc));
1128 __iommu_flush_cache(iommu, &hw[wait_index], sizeof(struct qi_desc));
1129
1130 qi->free_head = (qi->free_head + 2) % QI_LENGTH;
1131 qi->free_cnt -= 2;
1132
1133 /*
1134 * update the HW tail register indicating the presence of
1135 * new descriptors.
1136 */
1137 writel(qi->free_head << DMAR_IQ_SHIFT, iommu->reg + DMAR_IQT_REG);
1138
1139 while (qi->desc_status[wait_index] != QI_DONE) {
1140 /*
1141 * We will leave the interrupts disabled, to prevent interrupt
1142 * context to queue another cmd while a cmd is already submitted
1143 * and waiting for completion on this cpu. This is to avoid
1144 * a deadlock where the interrupt context can wait indefinitely
1145 * for free slots in the queue.
1146 */
1147 rc = qi_check_fault(iommu, index);
1148 if (rc)
1149 break;
1150
1151 raw_spin_unlock(&qi->q_lock);
1152 cpu_relax();
1153 raw_spin_lock(&qi->q_lock);
1154 }
1155
1156 qi->desc_status[index] = QI_DONE;
1157
1158 reclaim_free_desc(qi);
1159 raw_spin_unlock_irqrestore(&qi->q_lock, flags);
1160
1161 if (rc == -EAGAIN)
1162 goto restart;
1163
1164 return rc;
1165 }
1166
1167 /*
1168 * Flush the global interrupt entry cache.
1169 */
1170 void qi_global_iec(struct intel_iommu *iommu)
1171 {
1172 struct qi_desc desc;
1173
1174 desc.low = QI_IEC_TYPE;
1175 desc.high = 0;
1176
1177 /* should never fail */
1178 qi_submit_sync(&desc, iommu);
1179 }
1180
1181 void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
1182 u64 type)
1183 {
1184 struct qi_desc desc;
1185
1186 desc.low = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did)
1187 | QI_CC_GRAN(type) | QI_CC_TYPE;
1188 desc.high = 0;
1189
1190 qi_submit_sync(&desc, iommu);
1191 }
1192
1193 void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
1194 unsigned int size_order, u64 type)
1195 {
1196 u8 dw = 0, dr = 0;
1197
1198 struct qi_desc desc;
1199 int ih = 0;
1200
1201 if (cap_write_drain(iommu->cap))
1202 dw = 1;
1203
1204 if (cap_read_drain(iommu->cap))
1205 dr = 1;
1206
1207 desc.low = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw)
1208 | QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE;
1209 desc.high = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih)
1210 | QI_IOTLB_AM(size_order);
1211
1212 qi_submit_sync(&desc, iommu);
1213 }
1214
1215 void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 qdep,
1216 u64 addr, unsigned mask)
1217 {
1218 struct qi_desc desc;
1219
1220 if (mask) {
1221 BUG_ON(addr & ((1 << (VTD_PAGE_SHIFT + mask)) - 1));
1222 addr |= (1 << (VTD_PAGE_SHIFT + mask - 1)) - 1;
1223 desc.high = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE;
1224 } else
1225 desc.high = QI_DEV_IOTLB_ADDR(addr);
1226
1227 if (qdep >= QI_DEV_IOTLB_MAX_INVS)
1228 qdep = 0;
1229
1230 desc.low = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) |
1231 QI_DIOTLB_TYPE;
1232
1233 qi_submit_sync(&desc, iommu);
1234 }
1235
1236 /*
1237 * Disable Queued Invalidation interface.
1238 */
1239 void dmar_disable_qi(struct intel_iommu *iommu)
1240 {
1241 unsigned long flags;
1242 u32 sts;
1243 cycles_t start_time = get_cycles();
1244
1245 if (!ecap_qis(iommu->ecap))
1246 return;
1247
1248 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1249
1250 sts = dmar_readq(iommu->reg + DMAR_GSTS_REG);
1251 if (!(sts & DMA_GSTS_QIES))
1252 goto end;
1253
1254 /*
1255 * Give a chance to HW to complete the pending invalidation requests.
1256 */
1257 while ((readl(iommu->reg + DMAR_IQT_REG) !=
1258 readl(iommu->reg + DMAR_IQH_REG)) &&
1259 (DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time)))
1260 cpu_relax();
1261
1262 iommu->gcmd &= ~DMA_GCMD_QIE;
1263 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1264
1265 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl,
1266 !(sts & DMA_GSTS_QIES), sts);
1267 end:
1268 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1269 }
1270
1271 /*
1272 * Enable queued invalidation.
1273 */
1274 static void __dmar_enable_qi(struct intel_iommu *iommu)
1275 {
1276 u32 sts;
1277 unsigned long flags;
1278 struct q_inval *qi = iommu->qi;
1279
1280 qi->free_head = qi->free_tail = 0;
1281 qi->free_cnt = QI_LENGTH;
1282
1283 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1284
1285 /* write zero to the tail reg */
1286 writel(0, iommu->reg + DMAR_IQT_REG);
1287
1288 dmar_writeq(iommu->reg + DMAR_IQA_REG, virt_to_phys(qi->desc));
1289
1290 iommu->gcmd |= DMA_GCMD_QIE;
1291 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1292
1293 /* Make sure hardware complete it */
1294 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts);
1295
1296 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1297 }
1298
1299 /*
1300 * Enable Queued Invalidation interface. This is a must to support
1301 * interrupt-remapping. Also used by DMA-remapping, which replaces
1302 * register based IOTLB invalidation.
1303 */
1304 int dmar_enable_qi(struct intel_iommu *iommu)
1305 {
1306 struct q_inval *qi;
1307 struct page *desc_page;
1308
1309 if (!ecap_qis(iommu->ecap))
1310 return -ENOENT;
1311
1312 /*
1313 * queued invalidation is already setup and enabled.
1314 */
1315 if (iommu->qi)
1316 return 0;
1317
1318 iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC);
1319 if (!iommu->qi)
1320 return -ENOMEM;
1321
1322 qi = iommu->qi;
1323
1324
1325 desc_page = alloc_pages_node(iommu->node, GFP_ATOMIC | __GFP_ZERO, 0);
1326 if (!desc_page) {
1327 kfree(qi);
1328 iommu->qi = NULL;
1329 return -ENOMEM;
1330 }
1331
1332 qi->desc = page_address(desc_page);
1333
1334 qi->desc_status = kzalloc(QI_LENGTH * sizeof(int), GFP_ATOMIC);
1335 if (!qi->desc_status) {
1336 free_page((unsigned long) qi->desc);
1337 kfree(qi);
1338 iommu->qi = NULL;
1339 return -ENOMEM;
1340 }
1341
1342 qi->free_head = qi->free_tail = 0;
1343 qi->free_cnt = QI_LENGTH;
1344
1345 raw_spin_lock_init(&qi->q_lock);
1346
1347 __dmar_enable_qi(iommu);
1348
1349 return 0;
1350 }
1351
1352 /* iommu interrupt handling. Most stuff are MSI-like. */
1353
1354 enum faulttype {
1355 DMA_REMAP,
1356 INTR_REMAP,
1357 UNKNOWN,
1358 };
1359
1360 static const char *dma_remap_fault_reasons[] =
1361 {
1362 "Software",
1363 "Present bit in root entry is clear",
1364 "Present bit in context entry is clear",
1365 "Invalid context entry",
1366 "Access beyond MGAW",
1367 "PTE Write access is not set",
1368 "PTE Read access is not set",
1369 "Next page table ptr is invalid",
1370 "Root table address invalid",
1371 "Context table ptr is invalid",
1372 "non-zero reserved fields in RTP",
1373 "non-zero reserved fields in CTP",
1374 "non-zero reserved fields in PTE",
1375 "PCE for translation request specifies blocking",
1376 };
1377
1378 static const char *irq_remap_fault_reasons[] =
1379 {
1380 "Detected reserved fields in the decoded interrupt-remapped request",
1381 "Interrupt index exceeded the interrupt-remapping table size",
1382 "Present field in the IRTE entry is clear",
1383 "Error accessing interrupt-remapping table pointed by IRTA_REG",
1384 "Detected reserved fields in the IRTE entry",
1385 "Blocked a compatibility format interrupt request",
1386 "Blocked an interrupt request due to source-id verification failure",
1387 };
1388
1389 static const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type)
1390 {
1391 if (fault_reason >= 0x20 && (fault_reason - 0x20 <
1392 ARRAY_SIZE(irq_remap_fault_reasons))) {
1393 *fault_type = INTR_REMAP;
1394 return irq_remap_fault_reasons[fault_reason - 0x20];
1395 } else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) {
1396 *fault_type = DMA_REMAP;
1397 return dma_remap_fault_reasons[fault_reason];
1398 } else {
1399 *fault_type = UNKNOWN;
1400 return "Unknown";
1401 }
1402 }
1403
1404 void dmar_msi_unmask(struct irq_data *data)
1405 {
1406 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data);
1407 unsigned long flag;
1408
1409 /* unmask it */
1410 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1411 writel(0, iommu->reg + DMAR_FECTL_REG);
1412 /* Read a reg to force flush the post write */
1413 readl(iommu->reg + DMAR_FECTL_REG);
1414 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1415 }
1416
1417 void dmar_msi_mask(struct irq_data *data)
1418 {
1419 unsigned long flag;
1420 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data);
1421
1422 /* mask it */
1423 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1424 writel(DMA_FECTL_IM, iommu->reg + DMAR_FECTL_REG);
1425 /* Read a reg to force flush the post write */
1426 readl(iommu->reg + DMAR_FECTL_REG);
1427 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1428 }
1429
1430 void dmar_msi_write(int irq, struct msi_msg *msg)
1431 {
1432 struct intel_iommu *iommu = irq_get_handler_data(irq);
1433 unsigned long flag;
1434
1435 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1436 writel(msg->data, iommu->reg + DMAR_FEDATA_REG);
1437 writel(msg->address_lo, iommu->reg + DMAR_FEADDR_REG);
1438 writel(msg->address_hi, iommu->reg + DMAR_FEUADDR_REG);
1439 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1440 }
1441
1442 void dmar_msi_read(int irq, struct msi_msg *msg)
1443 {
1444 struct intel_iommu *iommu = irq_get_handler_data(irq);
1445 unsigned long flag;
1446
1447 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1448 msg->data = readl(iommu->reg + DMAR_FEDATA_REG);
1449 msg->address_lo = readl(iommu->reg + DMAR_FEADDR_REG);
1450 msg->address_hi = readl(iommu->reg + DMAR_FEUADDR_REG);
1451 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1452 }
1453
1454 static int dmar_fault_do_one(struct intel_iommu *iommu, int type,
1455 u8 fault_reason, u16 source_id, unsigned long long addr)
1456 {
1457 const char *reason;
1458 int fault_type;
1459
1460 reason = dmar_get_fault_reason(fault_reason, &fault_type);
1461
1462 if (fault_type == INTR_REMAP)
1463 pr_err("INTR-REMAP: Request device [[%02x:%02x.%d] "
1464 "fault index %llx\n"
1465 "INTR-REMAP:[fault reason %02d] %s\n",
1466 (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1467 PCI_FUNC(source_id & 0xFF), addr >> 48,
1468 fault_reason, reason);
1469 else
1470 pr_err("DMAR:[%s] Request device [%02x:%02x.%d] "
1471 "fault addr %llx \n"
1472 "DMAR:[fault reason %02d] %s\n",
1473 (type ? "DMA Read" : "DMA Write"),
1474 (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1475 PCI_FUNC(source_id & 0xFF), addr, fault_reason, reason);
1476 return 0;
1477 }
1478
1479 #define PRIMARY_FAULT_REG_LEN (16)
1480 irqreturn_t dmar_fault(int irq, void *dev_id)
1481 {
1482 struct intel_iommu *iommu = dev_id;
1483 int reg, fault_index;
1484 u32 fault_status;
1485 unsigned long flag;
1486
1487 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1488 fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1489 if (fault_status)
1490 pr_err("DRHD: handling fault status reg %x\n", fault_status);
1491
1492 /* TBD: ignore advanced fault log currently */
1493 if (!(fault_status & DMA_FSTS_PPF))
1494 goto unlock_exit;
1495
1496 fault_index = dma_fsts_fault_record_index(fault_status);
1497 reg = cap_fault_reg_offset(iommu->cap);
1498 while (1) {
1499 u8 fault_reason;
1500 u16 source_id;
1501 u64 guest_addr;
1502 int type;
1503 u32 data;
1504
1505 /* highest 32 bits */
1506 data = readl(iommu->reg + reg +
1507 fault_index * PRIMARY_FAULT_REG_LEN + 12);
1508 if (!(data & DMA_FRCD_F))
1509 break;
1510
1511 fault_reason = dma_frcd_fault_reason(data);
1512 type = dma_frcd_type(data);
1513
1514 data = readl(iommu->reg + reg +
1515 fault_index * PRIMARY_FAULT_REG_LEN + 8);
1516 source_id = dma_frcd_source_id(data);
1517
1518 guest_addr = dmar_readq(iommu->reg + reg +
1519 fault_index * PRIMARY_FAULT_REG_LEN);
1520 guest_addr = dma_frcd_page_addr(guest_addr);
1521 /* clear the fault */
1522 writel(DMA_FRCD_F, iommu->reg + reg +
1523 fault_index * PRIMARY_FAULT_REG_LEN + 12);
1524
1525 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1526
1527 dmar_fault_do_one(iommu, type, fault_reason,
1528 source_id, guest_addr);
1529
1530 fault_index++;
1531 if (fault_index >= cap_num_fault_regs(iommu->cap))
1532 fault_index = 0;
1533 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1534 }
1535
1536 writel(DMA_FSTS_PFO | DMA_FSTS_PPF, iommu->reg + DMAR_FSTS_REG);
1537
1538 unlock_exit:
1539 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1540 return IRQ_HANDLED;
1541 }
1542
1543 int dmar_set_interrupt(struct intel_iommu *iommu)
1544 {
1545 int irq, ret;
1546
1547 /*
1548 * Check if the fault interrupt is already initialized.
1549 */
1550 if (iommu->irq)
1551 return 0;
1552
1553 irq = dmar_alloc_hwirq();
1554 if (irq <= 0) {
1555 pr_err("IOMMU: no free vectors\n");
1556 return -EINVAL;
1557 }
1558
1559 irq_set_handler_data(irq, iommu);
1560 iommu->irq = irq;
1561
1562 ret = arch_setup_dmar_msi(irq);
1563 if (ret) {
1564 irq_set_handler_data(irq, NULL);
1565 iommu->irq = 0;
1566 dmar_free_hwirq(irq);
1567 return ret;
1568 }
1569
1570 ret = request_irq(irq, dmar_fault, IRQF_NO_THREAD, iommu->name, iommu);
1571 if (ret)
1572 pr_err("IOMMU: can't request irq\n");
1573 return ret;
1574 }
1575
1576 int __init enable_drhd_fault_handling(void)
1577 {
1578 struct dmar_drhd_unit *drhd;
1579 struct intel_iommu *iommu;
1580
1581 /*
1582 * Enable fault control interrupt.
1583 */
1584 for_each_iommu(iommu, drhd) {
1585 u32 fault_status;
1586 int ret = dmar_set_interrupt(iommu);
1587
1588 if (ret) {
1589 pr_err("DRHD %Lx: failed to enable fault, interrupt, ret %d\n",
1590 (unsigned long long)drhd->reg_base_addr, ret);
1591 return -1;
1592 }
1593
1594 /*
1595 * Clear any previous faults.
1596 */
1597 dmar_fault(iommu->irq, iommu);
1598 fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1599 writel(fault_status, iommu->reg + DMAR_FSTS_REG);
1600 }
1601
1602 return 0;
1603 }
1604
1605 /*
1606 * Re-enable Queued Invalidation interface.
1607 */
1608 int dmar_reenable_qi(struct intel_iommu *iommu)
1609 {
1610 if (!ecap_qis(iommu->ecap))
1611 return -ENOENT;
1612
1613 if (!iommu->qi)
1614 return -ENOENT;
1615
1616 /*
1617 * First disable queued invalidation.
1618 */
1619 dmar_disable_qi(iommu);
1620 /*
1621 * Then enable queued invalidation again. Since there is no pending
1622 * invalidation requests now, it's safe to re-enable queued
1623 * invalidation.
1624 */
1625 __dmar_enable_qi(iommu);
1626
1627 return 0;
1628 }
1629
1630 /*
1631 * Check interrupt remapping support in DMAR table description.
1632 */
1633 int __init dmar_ir_support(void)
1634 {
1635 struct acpi_table_dmar *dmar;
1636 dmar = (struct acpi_table_dmar *)dmar_tbl;
1637 if (!dmar)
1638 return 0;
1639 return dmar->flags & 0x1;
1640 }
1641
1642 static int __init dmar_free_unused_resources(void)
1643 {
1644 struct dmar_drhd_unit *dmaru, *dmaru_n;
1645
1646 /* DMAR units are in use */
1647 if (irq_remapping_enabled || intel_iommu_enabled)
1648 return 0;
1649
1650 if (dmar_dev_scope_status != 1 && !list_empty(&dmar_drhd_units))
1651 bus_unregister_notifier(&pci_bus_type, &dmar_pci_bus_nb);
1652
1653 down_write(&dmar_global_lock);
1654 list_for_each_entry_safe(dmaru, dmaru_n, &dmar_drhd_units, list) {
1655 list_del(&dmaru->list);
1656 dmar_free_drhd(dmaru);
1657 }
1658 up_write(&dmar_global_lock);
1659
1660 return 0;
1661 }
1662
1663 late_initcall(dmar_free_unused_resources);
1664 IOMMU_INIT_POST(detect_intel_iommu);
(deprecated) Btrfs devel tree