iio: adis_lib: Initialize trigger before requesting interrupt
[linux/fpc-iii.git] / drivers / iommu / dmar.c
blobe913a930ac8093dc856fba2fb71b8664e37452b0
1 /*
2 * Copyright (c) 2006, Intel Corporation.
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.
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.
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.
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>
22 * This file implements early detection/parsing of Remapping Devices
23 * reported to OS through BIOS via DMA remapping reporting (DMAR) ACPI
24 * tables.
26 * These routines are used by both DMA-remapping and Interrupt-remapping
29 #define pr_fmt(fmt) "DMAR: " fmt
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 <linux/iommu.h>
42 #include <asm/irq_remapping.h>
43 #include <asm/iommu_table.h>
45 #include "irq_remapping.h"
47 typedef int (*dmar_res_handler_t)(struct acpi_dmar_header *, void *);
48 struct dmar_res_callback {
49 dmar_res_handler_t cb[ACPI_DMAR_TYPE_RESERVED];
50 void *arg[ACPI_DMAR_TYPE_RESERVED];
51 bool ignore_unhandled;
52 bool print_entry;
56 * Assumptions:
57 * 1) The hotplug framework guarentees that DMAR unit will be hot-added
58 * before IO devices managed by that unit.
59 * 2) The hotplug framework guarantees that DMAR unit will be hot-removed
60 * after IO devices managed by that unit.
61 * 3) Hotplug events are rare.
63 * Locking rules for DMA and interrupt remapping related global data structures:
64 * 1) Use dmar_global_lock in process context
65 * 2) Use RCU in interrupt context
67 DECLARE_RWSEM(dmar_global_lock);
68 LIST_HEAD(dmar_drhd_units);
70 struct acpi_table_header * __initdata dmar_tbl;
71 static acpi_size dmar_tbl_size;
72 static int dmar_dev_scope_status = 1;
73 static unsigned long dmar_seq_ids[BITS_TO_LONGS(DMAR_UNITS_SUPPORTED)];
75 static int alloc_iommu(struct dmar_drhd_unit *drhd);
76 static void free_iommu(struct intel_iommu *iommu);
78 static void dmar_register_drhd_unit(struct dmar_drhd_unit *drhd)
81 * add INCLUDE_ALL at the tail, so scan the list will find it at
82 * the very end.
84 if (drhd->include_all)
85 list_add_tail_rcu(&drhd->list, &dmar_drhd_units);
86 else
87 list_add_rcu(&drhd->list, &dmar_drhd_units);
90 void *dmar_alloc_dev_scope(void *start, void *end, int *cnt)
92 struct acpi_dmar_device_scope *scope;
94 *cnt = 0;
95 while (start < end) {
96 scope = start;
97 if (scope->entry_type == ACPI_DMAR_SCOPE_TYPE_NAMESPACE ||
98 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT ||
99 scope->entry_type == ACPI_DMAR_SCOPE_TYPE_BRIDGE)
100 (*cnt)++;
101 else if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_IOAPIC &&
102 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_HPET) {
103 pr_warn("Unsupported device scope\n");
105 start += scope->length;
107 if (*cnt == 0)
108 return NULL;
110 return kcalloc(*cnt, sizeof(struct dmar_dev_scope), GFP_KERNEL);
113 void dmar_free_dev_scope(struct dmar_dev_scope **devices, int *cnt)
115 int i;
116 struct device *tmp_dev;
118 if (*devices && *cnt) {
119 for_each_active_dev_scope(*devices, *cnt, i, tmp_dev)
120 put_device(tmp_dev);
121 kfree(*devices);
124 *devices = NULL;
125 *cnt = 0;
128 /* Optimize out kzalloc()/kfree() for normal cases */
129 static char dmar_pci_notify_info_buf[64];
131 static struct dmar_pci_notify_info *
132 dmar_alloc_pci_notify_info(struct pci_dev *dev, unsigned long event)
134 int level = 0;
135 size_t size;
136 struct pci_dev *tmp;
137 struct dmar_pci_notify_info *info;
139 BUG_ON(dev->is_virtfn);
141 /* Only generate path[] for device addition event */
142 if (event == BUS_NOTIFY_ADD_DEVICE)
143 for (tmp = dev; tmp; tmp = tmp->bus->self)
144 level++;
146 size = sizeof(*info) + level * sizeof(struct acpi_dmar_pci_path);
147 if (size <= sizeof(dmar_pci_notify_info_buf)) {
148 info = (struct dmar_pci_notify_info *)dmar_pci_notify_info_buf;
149 } else {
150 info = kzalloc(size, GFP_KERNEL);
151 if (!info) {
152 pr_warn("Out of memory when allocating notify_info "
153 "for %s.\n", pci_name(dev));
154 if (dmar_dev_scope_status == 0)
155 dmar_dev_scope_status = -ENOMEM;
156 return NULL;
160 info->event = event;
161 info->dev = dev;
162 info->seg = pci_domain_nr(dev->bus);
163 info->level = level;
164 if (event == BUS_NOTIFY_ADD_DEVICE) {
165 for (tmp = dev; tmp; tmp = tmp->bus->self) {
166 level--;
167 info->path[level].bus = tmp->bus->number;
168 info->path[level].device = PCI_SLOT(tmp->devfn);
169 info->path[level].function = PCI_FUNC(tmp->devfn);
170 if (pci_is_root_bus(tmp->bus))
171 info->bus = tmp->bus->number;
175 return info;
178 static inline void dmar_free_pci_notify_info(struct dmar_pci_notify_info *info)
180 if ((void *)info != dmar_pci_notify_info_buf)
181 kfree(info);
184 static bool dmar_match_pci_path(struct dmar_pci_notify_info *info, int bus,
185 struct acpi_dmar_pci_path *path, int count)
187 int i;
189 if (info->bus != bus)
190 goto fallback;
191 if (info->level != count)
192 goto fallback;
194 for (i = 0; i < count; i++) {
195 if (path[i].device != info->path[i].device ||
196 path[i].function != info->path[i].function)
197 goto fallback;
200 return true;
202 fallback:
204 if (count != 1)
205 return false;
207 i = info->level - 1;
208 if (bus == info->path[i].bus &&
209 path[0].device == info->path[i].device &&
210 path[0].function == info->path[i].function) {
211 pr_info(FW_BUG "RMRR entry for device %02x:%02x.%x is broken - applying workaround\n",
212 bus, path[0].device, path[0].function);
213 return true;
216 return false;
219 /* Return: > 0 if match found, 0 if no match found, < 0 if error happens */
220 int dmar_insert_dev_scope(struct dmar_pci_notify_info *info,
221 void *start, void*end, u16 segment,
222 struct dmar_dev_scope *devices,
223 int devices_cnt)
225 int i, level;
226 struct device *tmp, *dev = &info->dev->dev;
227 struct acpi_dmar_device_scope *scope;
228 struct acpi_dmar_pci_path *path;
230 if (segment != info->seg)
231 return 0;
233 for (; start < end; start += scope->length) {
234 scope = start;
235 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_ENDPOINT &&
236 scope->entry_type != ACPI_DMAR_SCOPE_TYPE_BRIDGE)
237 continue;
239 path = (struct acpi_dmar_pci_path *)(scope + 1);
240 level = (scope->length - sizeof(*scope)) / sizeof(*path);
241 if (!dmar_match_pci_path(info, scope->bus, path, level))
242 continue;
244 if ((scope->entry_type == ACPI_DMAR_SCOPE_TYPE_ENDPOINT) ^
245 (info->dev->hdr_type == PCI_HEADER_TYPE_NORMAL)) {
246 pr_warn("Device scope type does not match for %s\n",
247 pci_name(info->dev));
248 return -EINVAL;
251 for_each_dev_scope(devices, devices_cnt, i, tmp)
252 if (tmp == NULL) {
253 devices[i].bus = info->dev->bus->number;
254 devices[i].devfn = info->dev->devfn;
255 rcu_assign_pointer(devices[i].dev,
256 get_device(dev));
257 return 1;
259 BUG_ON(i >= devices_cnt);
262 return 0;
265 int dmar_remove_dev_scope(struct dmar_pci_notify_info *info, u16 segment,
266 struct dmar_dev_scope *devices, int count)
268 int index;
269 struct device *tmp;
271 if (info->seg != segment)
272 return 0;
274 for_each_active_dev_scope(devices, count, index, tmp)
275 if (tmp == &info->dev->dev) {
276 RCU_INIT_POINTER(devices[index].dev, NULL);
277 synchronize_rcu();
278 put_device(tmp);
279 return 1;
282 return 0;
285 static int dmar_pci_bus_add_dev(struct dmar_pci_notify_info *info)
287 int ret = 0;
288 struct dmar_drhd_unit *dmaru;
289 struct acpi_dmar_hardware_unit *drhd;
291 for_each_drhd_unit(dmaru) {
292 if (dmaru->include_all)
293 continue;
295 drhd = container_of(dmaru->hdr,
296 struct acpi_dmar_hardware_unit, header);
297 ret = dmar_insert_dev_scope(info, (void *)(drhd + 1),
298 ((void *)drhd) + drhd->header.length,
299 dmaru->segment,
300 dmaru->devices, dmaru->devices_cnt);
301 if (ret != 0)
302 break;
304 if (ret >= 0)
305 ret = dmar_iommu_notify_scope_dev(info);
306 if (ret < 0 && dmar_dev_scope_status == 0)
307 dmar_dev_scope_status = ret;
309 return ret;
312 static void dmar_pci_bus_del_dev(struct dmar_pci_notify_info *info)
314 struct dmar_drhd_unit *dmaru;
316 for_each_drhd_unit(dmaru)
317 if (dmar_remove_dev_scope(info, dmaru->segment,
318 dmaru->devices, dmaru->devices_cnt))
319 break;
320 dmar_iommu_notify_scope_dev(info);
323 static int dmar_pci_bus_notifier(struct notifier_block *nb,
324 unsigned long action, void *data)
326 struct pci_dev *pdev = to_pci_dev(data);
327 struct dmar_pci_notify_info *info;
329 /* Only care about add/remove events for physical functions.
330 * For VFs we actually do the lookup based on the corresponding
331 * PF in device_to_iommu() anyway. */
332 if (pdev->is_virtfn)
333 return NOTIFY_DONE;
334 if (action != BUS_NOTIFY_ADD_DEVICE &&
335 action != BUS_NOTIFY_REMOVED_DEVICE)
336 return NOTIFY_DONE;
338 info = dmar_alloc_pci_notify_info(pdev, action);
339 if (!info)
340 return NOTIFY_DONE;
342 down_write(&dmar_global_lock);
343 if (action == BUS_NOTIFY_ADD_DEVICE)
344 dmar_pci_bus_add_dev(info);
345 else if (action == BUS_NOTIFY_REMOVED_DEVICE)
346 dmar_pci_bus_del_dev(info);
347 up_write(&dmar_global_lock);
349 dmar_free_pci_notify_info(info);
351 return NOTIFY_OK;
354 static struct notifier_block dmar_pci_bus_nb = {
355 .notifier_call = dmar_pci_bus_notifier,
356 .priority = INT_MIN,
359 static struct dmar_drhd_unit *
360 dmar_find_dmaru(struct acpi_dmar_hardware_unit *drhd)
362 struct dmar_drhd_unit *dmaru;
364 list_for_each_entry_rcu(dmaru, &dmar_drhd_units, list)
365 if (dmaru->segment == drhd->segment &&
366 dmaru->reg_base_addr == drhd->address)
367 return dmaru;
369 return NULL;
373 * dmar_parse_one_drhd - parses exactly one DMA remapping hardware definition
374 * structure which uniquely represent one DMA remapping hardware unit
375 * present in the platform
377 static int dmar_parse_one_drhd(struct acpi_dmar_header *header, void *arg)
379 struct acpi_dmar_hardware_unit *drhd;
380 struct dmar_drhd_unit *dmaru;
381 int ret = 0;
383 drhd = (struct acpi_dmar_hardware_unit *)header;
384 dmaru = dmar_find_dmaru(drhd);
385 if (dmaru)
386 goto out;
388 dmaru = kzalloc(sizeof(*dmaru) + header->length, GFP_KERNEL);
389 if (!dmaru)
390 return -ENOMEM;
393 * If header is allocated from slab by ACPI _DSM method, we need to
394 * copy the content because the memory buffer will be freed on return.
396 dmaru->hdr = (void *)(dmaru + 1);
397 memcpy(dmaru->hdr, header, header->length);
398 dmaru->reg_base_addr = drhd->address;
399 dmaru->segment = drhd->segment;
400 dmaru->include_all = drhd->flags & 0x1; /* BIT0: INCLUDE_ALL */
401 dmaru->devices = dmar_alloc_dev_scope((void *)(drhd + 1),
402 ((void *)drhd) + drhd->header.length,
403 &dmaru->devices_cnt);
404 if (dmaru->devices_cnt && dmaru->devices == NULL) {
405 kfree(dmaru);
406 return -ENOMEM;
409 ret = alloc_iommu(dmaru);
410 if (ret) {
411 dmar_free_dev_scope(&dmaru->devices,
412 &dmaru->devices_cnt);
413 kfree(dmaru);
414 return ret;
416 dmar_register_drhd_unit(dmaru);
418 out:
419 if (arg)
420 (*(int *)arg)++;
422 return 0;
425 static void dmar_free_drhd(struct dmar_drhd_unit *dmaru)
427 if (dmaru->devices && dmaru->devices_cnt)
428 dmar_free_dev_scope(&dmaru->devices, &dmaru->devices_cnt);
429 if (dmaru->iommu)
430 free_iommu(dmaru->iommu);
431 kfree(dmaru);
434 static int __init dmar_parse_one_andd(struct acpi_dmar_header *header,
435 void *arg)
437 struct acpi_dmar_andd *andd = (void *)header;
439 /* Check for NUL termination within the designated length */
440 if (strnlen(andd->device_name, header->length - 8) == header->length - 8) {
441 WARN_TAINT(1, TAINT_FIRMWARE_WORKAROUND,
442 "Your BIOS is broken; ANDD object name is not NUL-terminated\n"
443 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
444 dmi_get_system_info(DMI_BIOS_VENDOR),
445 dmi_get_system_info(DMI_BIOS_VERSION),
446 dmi_get_system_info(DMI_PRODUCT_VERSION));
447 return -EINVAL;
449 pr_info("ANDD device: %x name: %s\n", andd->device_number,
450 andd->device_name);
452 return 0;
455 #ifdef CONFIG_ACPI_NUMA
456 static int dmar_parse_one_rhsa(struct acpi_dmar_header *header, void *arg)
458 struct acpi_dmar_rhsa *rhsa;
459 struct dmar_drhd_unit *drhd;
461 rhsa = (struct acpi_dmar_rhsa *)header;
462 for_each_drhd_unit(drhd) {
463 if (drhd->reg_base_addr == rhsa->base_address) {
464 int node = acpi_map_pxm_to_node(rhsa->proximity_domain);
466 if (!node_online(node))
467 node = -1;
468 drhd->iommu->node = node;
469 return 0;
472 WARN_TAINT(
473 1, TAINT_FIRMWARE_WORKAROUND,
474 "Your BIOS is broken; RHSA refers to non-existent DMAR unit at %llx\n"
475 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
476 drhd->reg_base_addr,
477 dmi_get_system_info(DMI_BIOS_VENDOR),
478 dmi_get_system_info(DMI_BIOS_VERSION),
479 dmi_get_system_info(DMI_PRODUCT_VERSION));
481 return 0;
483 #else
484 #define dmar_parse_one_rhsa dmar_res_noop
485 #endif
487 static void __init
488 dmar_table_print_dmar_entry(struct acpi_dmar_header *header)
490 struct acpi_dmar_hardware_unit *drhd;
491 struct acpi_dmar_reserved_memory *rmrr;
492 struct acpi_dmar_atsr *atsr;
493 struct acpi_dmar_rhsa *rhsa;
495 switch (header->type) {
496 case ACPI_DMAR_TYPE_HARDWARE_UNIT:
497 drhd = container_of(header, struct acpi_dmar_hardware_unit,
498 header);
499 pr_info("DRHD base: %#016Lx flags: %#x\n",
500 (unsigned long long)drhd->address, drhd->flags);
501 break;
502 case ACPI_DMAR_TYPE_RESERVED_MEMORY:
503 rmrr = container_of(header, struct acpi_dmar_reserved_memory,
504 header);
505 pr_info("RMRR base: %#016Lx end: %#016Lx\n",
506 (unsigned long long)rmrr->base_address,
507 (unsigned long long)rmrr->end_address);
508 break;
509 case ACPI_DMAR_TYPE_ROOT_ATS:
510 atsr = container_of(header, struct acpi_dmar_atsr, header);
511 pr_info("ATSR flags: %#x\n", atsr->flags);
512 break;
513 case ACPI_DMAR_TYPE_HARDWARE_AFFINITY:
514 rhsa = container_of(header, struct acpi_dmar_rhsa, header);
515 pr_info("RHSA base: %#016Lx proximity domain: %#x\n",
516 (unsigned long long)rhsa->base_address,
517 rhsa->proximity_domain);
518 break;
519 case ACPI_DMAR_TYPE_NAMESPACE:
520 /* We don't print this here because we need to sanity-check
521 it first. So print it in dmar_parse_one_andd() instead. */
522 break;
527 * dmar_table_detect - checks to see if the platform supports DMAR devices
529 static int __init dmar_table_detect(void)
531 acpi_status status = AE_OK;
533 /* if we could find DMAR table, then there are DMAR devices */
534 status = acpi_get_table_with_size(ACPI_SIG_DMAR, 0,
535 (struct acpi_table_header **)&dmar_tbl,
536 &dmar_tbl_size);
538 if (ACPI_SUCCESS(status) && !dmar_tbl) {
539 pr_warn("Unable to map DMAR\n");
540 status = AE_NOT_FOUND;
543 return (ACPI_SUCCESS(status) ? 1 : 0);
546 static int dmar_walk_remapping_entries(struct acpi_dmar_header *start,
547 size_t len, struct dmar_res_callback *cb)
549 int ret = 0;
550 struct acpi_dmar_header *iter, *next;
551 struct acpi_dmar_header *end = ((void *)start) + len;
553 for (iter = start; iter < end && ret == 0; iter = next) {
554 next = (void *)iter + iter->length;
555 if (iter->length == 0) {
556 /* Avoid looping forever on bad ACPI tables */
557 pr_debug(FW_BUG "Invalid 0-length structure\n");
558 break;
559 } else if (next > end) {
560 /* Avoid passing table end */
561 pr_warn(FW_BUG "Record passes table end\n");
562 ret = -EINVAL;
563 break;
566 if (cb->print_entry)
567 dmar_table_print_dmar_entry(iter);
569 if (iter->type >= ACPI_DMAR_TYPE_RESERVED) {
570 /* continue for forward compatibility */
571 pr_debug("Unknown DMAR structure type %d\n",
572 iter->type);
573 } else if (cb->cb[iter->type]) {
574 ret = cb->cb[iter->type](iter, cb->arg[iter->type]);
575 } else if (!cb->ignore_unhandled) {
576 pr_warn("No handler for DMAR structure type %d\n",
577 iter->type);
578 ret = -EINVAL;
582 return ret;
585 static inline int dmar_walk_dmar_table(struct acpi_table_dmar *dmar,
586 struct dmar_res_callback *cb)
588 return dmar_walk_remapping_entries((void *)(dmar + 1),
589 dmar->header.length - sizeof(*dmar), cb);
593 * parse_dmar_table - parses the DMA reporting table
595 static int __init
596 parse_dmar_table(void)
598 struct acpi_table_dmar *dmar;
599 int ret = 0;
600 int drhd_count = 0;
601 struct dmar_res_callback cb = {
602 .print_entry = true,
603 .ignore_unhandled = true,
604 .arg[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &drhd_count,
605 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_parse_one_drhd,
606 .cb[ACPI_DMAR_TYPE_RESERVED_MEMORY] = &dmar_parse_one_rmrr,
607 .cb[ACPI_DMAR_TYPE_ROOT_ATS] = &dmar_parse_one_atsr,
608 .cb[ACPI_DMAR_TYPE_HARDWARE_AFFINITY] = &dmar_parse_one_rhsa,
609 .cb[ACPI_DMAR_TYPE_NAMESPACE] = &dmar_parse_one_andd,
613 * Do it again, earlier dmar_tbl mapping could be mapped with
614 * fixed map.
616 dmar_table_detect();
619 * ACPI tables may not be DMA protected by tboot, so use DMAR copy
620 * SINIT saved in SinitMleData in TXT heap (which is DMA protected)
622 dmar_tbl = tboot_get_dmar_table(dmar_tbl);
624 dmar = (struct acpi_table_dmar *)dmar_tbl;
625 if (!dmar)
626 return -ENODEV;
628 if (dmar->width < PAGE_SHIFT - 1) {
629 pr_warn("Invalid DMAR haw\n");
630 return -EINVAL;
633 pr_info("Host address width %d\n", dmar->width + 1);
634 ret = dmar_walk_dmar_table(dmar, &cb);
635 if (ret == 0 && drhd_count == 0)
636 pr_warn(FW_BUG "No DRHD structure found in DMAR table\n");
638 return ret;
641 static int dmar_pci_device_match(struct dmar_dev_scope devices[],
642 int cnt, struct pci_dev *dev)
644 int index;
645 struct device *tmp;
647 while (dev) {
648 for_each_active_dev_scope(devices, cnt, index, tmp)
649 if (dev_is_pci(tmp) && dev == to_pci_dev(tmp))
650 return 1;
652 /* Check our parent */
653 dev = dev->bus->self;
656 return 0;
659 struct dmar_drhd_unit *
660 dmar_find_matched_drhd_unit(struct pci_dev *dev)
662 struct dmar_drhd_unit *dmaru;
663 struct acpi_dmar_hardware_unit *drhd;
665 dev = pci_physfn(dev);
667 rcu_read_lock();
668 for_each_drhd_unit(dmaru) {
669 drhd = container_of(dmaru->hdr,
670 struct acpi_dmar_hardware_unit,
671 header);
673 if (dmaru->include_all &&
674 drhd->segment == pci_domain_nr(dev->bus))
675 goto out;
677 if (dmar_pci_device_match(dmaru->devices,
678 dmaru->devices_cnt, dev))
679 goto out;
681 dmaru = NULL;
682 out:
683 rcu_read_unlock();
685 return dmaru;
688 static void __init dmar_acpi_insert_dev_scope(u8 device_number,
689 struct acpi_device *adev)
691 struct dmar_drhd_unit *dmaru;
692 struct acpi_dmar_hardware_unit *drhd;
693 struct acpi_dmar_device_scope *scope;
694 struct device *tmp;
695 int i;
696 struct acpi_dmar_pci_path *path;
698 for_each_drhd_unit(dmaru) {
699 drhd = container_of(dmaru->hdr,
700 struct acpi_dmar_hardware_unit,
701 header);
703 for (scope = (void *)(drhd + 1);
704 (unsigned long)scope < ((unsigned long)drhd) + drhd->header.length;
705 scope = ((void *)scope) + scope->length) {
706 if (scope->entry_type != ACPI_DMAR_SCOPE_TYPE_NAMESPACE)
707 continue;
708 if (scope->enumeration_id != device_number)
709 continue;
711 path = (void *)(scope + 1);
712 pr_info("ACPI device \"%s\" under DMAR at %llx as %02x:%02x.%d\n",
713 dev_name(&adev->dev), dmaru->reg_base_addr,
714 scope->bus, path->device, path->function);
715 for_each_dev_scope(dmaru->devices, dmaru->devices_cnt, i, tmp)
716 if (tmp == NULL) {
717 dmaru->devices[i].bus = scope->bus;
718 dmaru->devices[i].devfn = PCI_DEVFN(path->device,
719 path->function);
720 rcu_assign_pointer(dmaru->devices[i].dev,
721 get_device(&adev->dev));
722 return;
724 BUG_ON(i >= dmaru->devices_cnt);
727 pr_warn("No IOMMU scope found for ANDD enumeration ID %d (%s)\n",
728 device_number, dev_name(&adev->dev));
731 static int __init dmar_acpi_dev_scope_init(void)
733 struct acpi_dmar_andd *andd;
735 if (dmar_tbl == NULL)
736 return -ENODEV;
738 for (andd = (void *)dmar_tbl + sizeof(struct acpi_table_dmar);
739 ((unsigned long)andd) < ((unsigned long)dmar_tbl) + dmar_tbl->length;
740 andd = ((void *)andd) + andd->header.length) {
741 if (andd->header.type == ACPI_DMAR_TYPE_NAMESPACE) {
742 acpi_handle h;
743 struct acpi_device *adev;
745 if (!ACPI_SUCCESS(acpi_get_handle(ACPI_ROOT_OBJECT,
746 andd->device_name,
747 &h))) {
748 pr_err("Failed to find handle for ACPI object %s\n",
749 andd->device_name);
750 continue;
752 if (acpi_bus_get_device(h, &adev)) {
753 pr_err("Failed to get device for ACPI object %s\n",
754 andd->device_name);
755 continue;
757 dmar_acpi_insert_dev_scope(andd->device_number, adev);
760 return 0;
763 int __init dmar_dev_scope_init(void)
765 struct pci_dev *dev = NULL;
766 struct dmar_pci_notify_info *info;
768 if (dmar_dev_scope_status != 1)
769 return dmar_dev_scope_status;
771 if (list_empty(&dmar_drhd_units)) {
772 dmar_dev_scope_status = -ENODEV;
773 } else {
774 dmar_dev_scope_status = 0;
776 dmar_acpi_dev_scope_init();
778 for_each_pci_dev(dev) {
779 if (dev->is_virtfn)
780 continue;
782 info = dmar_alloc_pci_notify_info(dev,
783 BUS_NOTIFY_ADD_DEVICE);
784 if (!info) {
785 return dmar_dev_scope_status;
786 } else {
787 dmar_pci_bus_add_dev(info);
788 dmar_free_pci_notify_info(info);
792 bus_register_notifier(&pci_bus_type, &dmar_pci_bus_nb);
795 return dmar_dev_scope_status;
799 int __init dmar_table_init(void)
801 static int dmar_table_initialized;
802 int ret;
804 if (dmar_table_initialized == 0) {
805 ret = parse_dmar_table();
806 if (ret < 0) {
807 if (ret != -ENODEV)
808 pr_info("Parse DMAR table failure.\n");
809 } else if (list_empty(&dmar_drhd_units)) {
810 pr_info("No DMAR devices found\n");
811 ret = -ENODEV;
814 if (ret < 0)
815 dmar_table_initialized = ret;
816 else
817 dmar_table_initialized = 1;
820 return dmar_table_initialized < 0 ? dmar_table_initialized : 0;
823 static void warn_invalid_dmar(u64 addr, const char *message)
825 WARN_TAINT_ONCE(
826 1, TAINT_FIRMWARE_WORKAROUND,
827 "Your BIOS is broken; DMAR reported at address %llx%s!\n"
828 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
829 addr, message,
830 dmi_get_system_info(DMI_BIOS_VENDOR),
831 dmi_get_system_info(DMI_BIOS_VERSION),
832 dmi_get_system_info(DMI_PRODUCT_VERSION));
835 static int __ref
836 dmar_validate_one_drhd(struct acpi_dmar_header *entry, void *arg)
838 struct acpi_dmar_hardware_unit *drhd;
839 void __iomem *addr;
840 u64 cap, ecap;
842 drhd = (void *)entry;
843 if (!drhd->address) {
844 warn_invalid_dmar(0, "");
845 return -EINVAL;
848 if (arg)
849 addr = ioremap(drhd->address, VTD_PAGE_SIZE);
850 else
851 addr = early_ioremap(drhd->address, VTD_PAGE_SIZE);
852 if (!addr) {
853 pr_warn("Can't validate DRHD address: %llx\n", drhd->address);
854 return -EINVAL;
857 cap = dmar_readq(addr + DMAR_CAP_REG);
858 ecap = dmar_readq(addr + DMAR_ECAP_REG);
860 if (arg)
861 iounmap(addr);
862 else
863 early_iounmap(addr, VTD_PAGE_SIZE);
865 if (cap == (uint64_t)-1 && ecap == (uint64_t)-1) {
866 warn_invalid_dmar(drhd->address, " returns all ones");
867 return -EINVAL;
870 return 0;
873 int __init detect_intel_iommu(void)
875 int ret;
876 struct dmar_res_callback validate_drhd_cb = {
877 .cb[ACPI_DMAR_TYPE_HARDWARE_UNIT] = &dmar_validate_one_drhd,
878 .ignore_unhandled = true,
881 down_write(&dmar_global_lock);
882 ret = dmar_table_detect();
883 if (ret)
884 ret = !dmar_walk_dmar_table((struct acpi_table_dmar *)dmar_tbl,
885 &validate_drhd_cb);
886 if (ret && !no_iommu && !iommu_detected && !dmar_disabled) {
887 iommu_detected = 1;
888 /* Make sure ACS will be enabled */
889 pci_request_acs();
892 #ifdef CONFIG_X86
893 if (ret)
894 x86_init.iommu.iommu_init = intel_iommu_init;
895 #endif
897 early_acpi_os_unmap_memory((void __iomem *)dmar_tbl, dmar_tbl_size);
898 dmar_tbl = NULL;
899 up_write(&dmar_global_lock);
901 return ret ? 1 : -ENODEV;
905 static void unmap_iommu(struct intel_iommu *iommu)
907 iounmap(iommu->reg);
908 release_mem_region(iommu->reg_phys, iommu->reg_size);
912 * map_iommu: map the iommu's registers
913 * @iommu: the iommu to map
914 * @phys_addr: the physical address of the base resgister
916 * Memory map the iommu's registers. Start w/ a single page, and
917 * possibly expand if that turns out to be insufficent.
919 static int map_iommu(struct intel_iommu *iommu, u64 phys_addr)
921 int map_size, err=0;
923 iommu->reg_phys = phys_addr;
924 iommu->reg_size = VTD_PAGE_SIZE;
926 if (!request_mem_region(iommu->reg_phys, iommu->reg_size, iommu->name)) {
927 pr_err("Can't reserve memory\n");
928 err = -EBUSY;
929 goto out;
932 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size);
933 if (!iommu->reg) {
934 pr_err("Can't map the region\n");
935 err = -ENOMEM;
936 goto release;
939 iommu->cap = dmar_readq(iommu->reg + DMAR_CAP_REG);
940 iommu->ecap = dmar_readq(iommu->reg + DMAR_ECAP_REG);
942 if (iommu->cap == (uint64_t)-1 && iommu->ecap == (uint64_t)-1) {
943 err = -EINVAL;
944 warn_invalid_dmar(phys_addr, " returns all ones");
945 goto unmap;
948 /* the registers might be more than one page */
949 map_size = max_t(int, ecap_max_iotlb_offset(iommu->ecap),
950 cap_max_fault_reg_offset(iommu->cap));
951 map_size = VTD_PAGE_ALIGN(map_size);
952 if (map_size > iommu->reg_size) {
953 iounmap(iommu->reg);
954 release_mem_region(iommu->reg_phys, iommu->reg_size);
955 iommu->reg_size = map_size;
956 if (!request_mem_region(iommu->reg_phys, iommu->reg_size,
957 iommu->name)) {
958 pr_err("Can't reserve memory\n");
959 err = -EBUSY;
960 goto out;
962 iommu->reg = ioremap(iommu->reg_phys, iommu->reg_size);
963 if (!iommu->reg) {
964 pr_err("Can't map the region\n");
965 err = -ENOMEM;
966 goto release;
969 err = 0;
970 goto out;
972 unmap:
973 iounmap(iommu->reg);
974 release:
975 release_mem_region(iommu->reg_phys, iommu->reg_size);
976 out:
977 return err;
980 static int dmar_alloc_seq_id(struct intel_iommu *iommu)
982 iommu->seq_id = find_first_zero_bit(dmar_seq_ids,
983 DMAR_UNITS_SUPPORTED);
984 if (iommu->seq_id >= DMAR_UNITS_SUPPORTED) {
985 iommu->seq_id = -1;
986 } else {
987 set_bit(iommu->seq_id, dmar_seq_ids);
988 sprintf(iommu->name, "dmar%d", iommu->seq_id);
991 return iommu->seq_id;
994 static void dmar_free_seq_id(struct intel_iommu *iommu)
996 if (iommu->seq_id >= 0) {
997 clear_bit(iommu->seq_id, dmar_seq_ids);
998 iommu->seq_id = -1;
1002 static int alloc_iommu(struct dmar_drhd_unit *drhd)
1004 struct intel_iommu *iommu;
1005 u32 ver, sts;
1006 int agaw = 0;
1007 int msagaw = 0;
1008 int err;
1010 if (!drhd->reg_base_addr) {
1011 warn_invalid_dmar(0, "");
1012 return -EINVAL;
1015 iommu = kzalloc(sizeof(*iommu), GFP_KERNEL);
1016 if (!iommu)
1017 return -ENOMEM;
1019 if (dmar_alloc_seq_id(iommu) < 0) {
1020 pr_err("Failed to allocate seq_id\n");
1021 err = -ENOSPC;
1022 goto error;
1025 err = map_iommu(iommu, drhd->reg_base_addr);
1026 if (err) {
1027 pr_err("Failed to map %s\n", iommu->name);
1028 goto error_free_seq_id;
1031 err = -EINVAL;
1032 agaw = iommu_calculate_agaw(iommu);
1033 if (agaw < 0) {
1034 pr_err("Cannot get a valid agaw for iommu (seq_id = %d)\n",
1035 iommu->seq_id);
1036 goto err_unmap;
1038 msagaw = iommu_calculate_max_sagaw(iommu);
1039 if (msagaw < 0) {
1040 pr_err("Cannot get a valid max agaw for iommu (seq_id = %d)\n",
1041 iommu->seq_id);
1042 goto err_unmap;
1044 iommu->agaw = agaw;
1045 iommu->msagaw = msagaw;
1046 iommu->segment = drhd->segment;
1048 iommu->node = -1;
1050 ver = readl(iommu->reg + DMAR_VER_REG);
1051 pr_info("%s: reg_base_addr %llx ver %d:%d cap %llx ecap %llx\n",
1052 iommu->name,
1053 (unsigned long long)drhd->reg_base_addr,
1054 DMAR_VER_MAJOR(ver), DMAR_VER_MINOR(ver),
1055 (unsigned long long)iommu->cap,
1056 (unsigned long long)iommu->ecap);
1058 /* Reflect status in gcmd */
1059 sts = readl(iommu->reg + DMAR_GSTS_REG);
1060 if (sts & DMA_GSTS_IRES)
1061 iommu->gcmd |= DMA_GCMD_IRE;
1062 if (sts & DMA_GSTS_TES)
1063 iommu->gcmd |= DMA_GCMD_TE;
1064 if (sts & DMA_GSTS_QIES)
1065 iommu->gcmd |= DMA_GCMD_QIE;
1067 raw_spin_lock_init(&iommu->register_lock);
1069 drhd->iommu = iommu;
1071 if (intel_iommu_enabled)
1072 iommu->iommu_dev = iommu_device_create(NULL, iommu,
1073 intel_iommu_groups,
1074 "%s", iommu->name);
1076 return 0;
1078 err_unmap:
1079 unmap_iommu(iommu);
1080 error_free_seq_id:
1081 dmar_free_seq_id(iommu);
1082 error:
1083 kfree(iommu);
1084 return err;
1087 static void free_iommu(struct intel_iommu *iommu)
1089 iommu_device_destroy(iommu->iommu_dev);
1091 if (iommu->irq) {
1092 if (iommu->pr_irq) {
1093 free_irq(iommu->pr_irq, iommu);
1094 dmar_free_hwirq(iommu->pr_irq);
1095 iommu->pr_irq = 0;
1097 free_irq(iommu->irq, iommu);
1098 dmar_free_hwirq(iommu->irq);
1099 iommu->irq = 0;
1102 if (iommu->qi) {
1103 free_page((unsigned long)iommu->qi->desc);
1104 kfree(iommu->qi->desc_status);
1105 kfree(iommu->qi);
1108 if (iommu->reg)
1109 unmap_iommu(iommu);
1111 dmar_free_seq_id(iommu);
1112 kfree(iommu);
1116 * Reclaim all the submitted descriptors which have completed its work.
1118 static inline void reclaim_free_desc(struct q_inval *qi)
1120 while (qi->desc_status[qi->free_tail] == QI_DONE ||
1121 qi->desc_status[qi->free_tail] == QI_ABORT) {
1122 qi->desc_status[qi->free_tail] = QI_FREE;
1123 qi->free_tail = (qi->free_tail + 1) % QI_LENGTH;
1124 qi->free_cnt++;
1128 static int qi_check_fault(struct intel_iommu *iommu, int index)
1130 u32 fault;
1131 int head, tail;
1132 struct q_inval *qi = iommu->qi;
1133 int wait_index = (index + 1) % QI_LENGTH;
1135 if (qi->desc_status[wait_index] == QI_ABORT)
1136 return -EAGAIN;
1138 fault = readl(iommu->reg + DMAR_FSTS_REG);
1141 * If IQE happens, the head points to the descriptor associated
1142 * with the error. No new descriptors are fetched until the IQE
1143 * is cleared.
1145 if (fault & DMA_FSTS_IQE) {
1146 head = readl(iommu->reg + DMAR_IQH_REG);
1147 if ((head >> DMAR_IQ_SHIFT) == index) {
1148 pr_err("VT-d detected invalid descriptor: "
1149 "low=%llx, high=%llx\n",
1150 (unsigned long long)qi->desc[index].low,
1151 (unsigned long long)qi->desc[index].high);
1152 memcpy(&qi->desc[index], &qi->desc[wait_index],
1153 sizeof(struct qi_desc));
1154 __iommu_flush_cache(iommu, &qi->desc[index],
1155 sizeof(struct qi_desc));
1156 writel(DMA_FSTS_IQE, iommu->reg + DMAR_FSTS_REG);
1157 return -EINVAL;
1162 * If ITE happens, all pending wait_desc commands are aborted.
1163 * No new descriptors are fetched until the ITE is cleared.
1165 if (fault & DMA_FSTS_ITE) {
1166 head = readl(iommu->reg + DMAR_IQH_REG);
1167 head = ((head >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
1168 head |= 1;
1169 tail = readl(iommu->reg + DMAR_IQT_REG);
1170 tail = ((tail >> DMAR_IQ_SHIFT) - 1 + QI_LENGTH) % QI_LENGTH;
1172 writel(DMA_FSTS_ITE, iommu->reg + DMAR_FSTS_REG);
1174 do {
1175 if (qi->desc_status[head] == QI_IN_USE)
1176 qi->desc_status[head] = QI_ABORT;
1177 head = (head - 2 + QI_LENGTH) % QI_LENGTH;
1178 } while (head != tail);
1180 if (qi->desc_status[wait_index] == QI_ABORT)
1181 return -EAGAIN;
1184 if (fault & DMA_FSTS_ICE)
1185 writel(DMA_FSTS_ICE, iommu->reg + DMAR_FSTS_REG);
1187 return 0;
1191 * Submit the queued invalidation descriptor to the remapping
1192 * hardware unit and wait for its completion.
1194 int qi_submit_sync(struct qi_desc *desc, struct intel_iommu *iommu)
1196 int rc;
1197 struct q_inval *qi = iommu->qi;
1198 struct qi_desc *hw, wait_desc;
1199 int wait_index, index;
1200 unsigned long flags;
1202 if (!qi)
1203 return 0;
1205 hw = qi->desc;
1207 restart:
1208 rc = 0;
1210 raw_spin_lock_irqsave(&qi->q_lock, flags);
1211 while (qi->free_cnt < 3) {
1212 raw_spin_unlock_irqrestore(&qi->q_lock, flags);
1213 cpu_relax();
1214 raw_spin_lock_irqsave(&qi->q_lock, flags);
1217 index = qi->free_head;
1218 wait_index = (index + 1) % QI_LENGTH;
1220 qi->desc_status[index] = qi->desc_status[wait_index] = QI_IN_USE;
1222 hw[index] = *desc;
1224 wait_desc.low = QI_IWD_STATUS_DATA(QI_DONE) |
1225 QI_IWD_STATUS_WRITE | QI_IWD_TYPE;
1226 wait_desc.high = virt_to_phys(&qi->desc_status[wait_index]);
1228 hw[wait_index] = wait_desc;
1230 __iommu_flush_cache(iommu, &hw[index], sizeof(struct qi_desc));
1231 __iommu_flush_cache(iommu, &hw[wait_index], sizeof(struct qi_desc));
1233 qi->free_head = (qi->free_head + 2) % QI_LENGTH;
1234 qi->free_cnt -= 2;
1237 * update the HW tail register indicating the presence of
1238 * new descriptors.
1240 writel(qi->free_head << DMAR_IQ_SHIFT, iommu->reg + DMAR_IQT_REG);
1242 while (qi->desc_status[wait_index] != QI_DONE) {
1244 * We will leave the interrupts disabled, to prevent interrupt
1245 * context to queue another cmd while a cmd is already submitted
1246 * and waiting for completion on this cpu. This is to avoid
1247 * a deadlock where the interrupt context can wait indefinitely
1248 * for free slots in the queue.
1250 rc = qi_check_fault(iommu, index);
1251 if (rc)
1252 break;
1254 raw_spin_unlock(&qi->q_lock);
1255 cpu_relax();
1256 raw_spin_lock(&qi->q_lock);
1259 qi->desc_status[index] = QI_DONE;
1261 reclaim_free_desc(qi);
1262 raw_spin_unlock_irqrestore(&qi->q_lock, flags);
1264 if (rc == -EAGAIN)
1265 goto restart;
1267 return rc;
1271 * Flush the global interrupt entry cache.
1273 void qi_global_iec(struct intel_iommu *iommu)
1275 struct qi_desc desc;
1277 desc.low = QI_IEC_TYPE;
1278 desc.high = 0;
1280 /* should never fail */
1281 qi_submit_sync(&desc, iommu);
1284 void qi_flush_context(struct intel_iommu *iommu, u16 did, u16 sid, u8 fm,
1285 u64 type)
1287 struct qi_desc desc;
1289 desc.low = QI_CC_FM(fm) | QI_CC_SID(sid) | QI_CC_DID(did)
1290 | QI_CC_GRAN(type) | QI_CC_TYPE;
1291 desc.high = 0;
1293 qi_submit_sync(&desc, iommu);
1296 void qi_flush_iotlb(struct intel_iommu *iommu, u16 did, u64 addr,
1297 unsigned int size_order, u64 type)
1299 u8 dw = 0, dr = 0;
1301 struct qi_desc desc;
1302 int ih = 0;
1304 if (cap_write_drain(iommu->cap))
1305 dw = 1;
1307 if (cap_read_drain(iommu->cap))
1308 dr = 1;
1310 desc.low = QI_IOTLB_DID(did) | QI_IOTLB_DR(dr) | QI_IOTLB_DW(dw)
1311 | QI_IOTLB_GRAN(type) | QI_IOTLB_TYPE;
1312 desc.high = QI_IOTLB_ADDR(addr) | QI_IOTLB_IH(ih)
1313 | QI_IOTLB_AM(size_order);
1315 qi_submit_sync(&desc, iommu);
1318 void qi_flush_dev_iotlb(struct intel_iommu *iommu, u16 sid, u16 qdep,
1319 u64 addr, unsigned mask)
1321 struct qi_desc desc;
1323 if (mask) {
1324 BUG_ON(addr & ((1 << (VTD_PAGE_SHIFT + mask)) - 1));
1325 addr |= (1 << (VTD_PAGE_SHIFT + mask - 1)) - 1;
1326 desc.high = QI_DEV_IOTLB_ADDR(addr) | QI_DEV_IOTLB_SIZE;
1327 } else
1328 desc.high = QI_DEV_IOTLB_ADDR(addr);
1330 if (qdep >= QI_DEV_IOTLB_MAX_INVS)
1331 qdep = 0;
1333 desc.low = QI_DEV_IOTLB_SID(sid) | QI_DEV_IOTLB_QDEP(qdep) |
1334 QI_DIOTLB_TYPE;
1336 qi_submit_sync(&desc, iommu);
1340 * Disable Queued Invalidation interface.
1342 void dmar_disable_qi(struct intel_iommu *iommu)
1344 unsigned long flags;
1345 u32 sts;
1346 cycles_t start_time = get_cycles();
1348 if (!ecap_qis(iommu->ecap))
1349 return;
1351 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1353 sts = readl(iommu->reg + DMAR_GSTS_REG);
1354 if (!(sts & DMA_GSTS_QIES))
1355 goto end;
1358 * Give a chance to HW to complete the pending invalidation requests.
1360 while ((readl(iommu->reg + DMAR_IQT_REG) !=
1361 readl(iommu->reg + DMAR_IQH_REG)) &&
1362 (DMAR_OPERATION_TIMEOUT > (get_cycles() - start_time)))
1363 cpu_relax();
1365 iommu->gcmd &= ~DMA_GCMD_QIE;
1366 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1368 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl,
1369 !(sts & DMA_GSTS_QIES), sts);
1370 end:
1371 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1375 * Enable queued invalidation.
1377 static void __dmar_enable_qi(struct intel_iommu *iommu)
1379 u32 sts;
1380 unsigned long flags;
1381 struct q_inval *qi = iommu->qi;
1383 qi->free_head = qi->free_tail = 0;
1384 qi->free_cnt = QI_LENGTH;
1386 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1388 /* write zero to the tail reg */
1389 writel(0, iommu->reg + DMAR_IQT_REG);
1391 dmar_writeq(iommu->reg + DMAR_IQA_REG, virt_to_phys(qi->desc));
1393 iommu->gcmd |= DMA_GCMD_QIE;
1394 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1396 /* Make sure hardware complete it */
1397 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, readl, (sts & DMA_GSTS_QIES), sts);
1399 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1403 * Enable Queued Invalidation interface. This is a must to support
1404 * interrupt-remapping. Also used by DMA-remapping, which replaces
1405 * register based IOTLB invalidation.
1407 int dmar_enable_qi(struct intel_iommu *iommu)
1409 struct q_inval *qi;
1410 struct page *desc_page;
1412 if (!ecap_qis(iommu->ecap))
1413 return -ENOENT;
1416 * queued invalidation is already setup and enabled.
1418 if (iommu->qi)
1419 return 0;
1421 iommu->qi = kmalloc(sizeof(*qi), GFP_ATOMIC);
1422 if (!iommu->qi)
1423 return -ENOMEM;
1425 qi = iommu->qi;
1428 desc_page = alloc_pages_node(iommu->node, GFP_ATOMIC | __GFP_ZERO, 0);
1429 if (!desc_page) {
1430 kfree(qi);
1431 iommu->qi = NULL;
1432 return -ENOMEM;
1435 qi->desc = page_address(desc_page);
1437 qi->desc_status = kzalloc(QI_LENGTH * sizeof(int), GFP_ATOMIC);
1438 if (!qi->desc_status) {
1439 free_page((unsigned long) qi->desc);
1440 kfree(qi);
1441 iommu->qi = NULL;
1442 return -ENOMEM;
1445 raw_spin_lock_init(&qi->q_lock);
1447 __dmar_enable_qi(iommu);
1449 return 0;
1452 /* iommu interrupt handling. Most stuff are MSI-like. */
1454 enum faulttype {
1455 DMA_REMAP,
1456 INTR_REMAP,
1457 UNKNOWN,
1460 static const char *dma_remap_fault_reasons[] =
1462 "Software",
1463 "Present bit in root entry is clear",
1464 "Present bit in context entry is clear",
1465 "Invalid context entry",
1466 "Access beyond MGAW",
1467 "PTE Write access is not set",
1468 "PTE Read access is not set",
1469 "Next page table ptr is invalid",
1470 "Root table address invalid",
1471 "Context table ptr is invalid",
1472 "non-zero reserved fields in RTP",
1473 "non-zero reserved fields in CTP",
1474 "non-zero reserved fields in PTE",
1475 "PCE for translation request specifies blocking",
1478 static const char *irq_remap_fault_reasons[] =
1480 "Detected reserved fields in the decoded interrupt-remapped request",
1481 "Interrupt index exceeded the interrupt-remapping table size",
1482 "Present field in the IRTE entry is clear",
1483 "Error accessing interrupt-remapping table pointed by IRTA_REG",
1484 "Detected reserved fields in the IRTE entry",
1485 "Blocked a compatibility format interrupt request",
1486 "Blocked an interrupt request due to source-id verification failure",
1489 static const char *dmar_get_fault_reason(u8 fault_reason, int *fault_type)
1491 if (fault_reason >= 0x20 && (fault_reason - 0x20 <
1492 ARRAY_SIZE(irq_remap_fault_reasons))) {
1493 *fault_type = INTR_REMAP;
1494 return irq_remap_fault_reasons[fault_reason - 0x20];
1495 } else if (fault_reason < ARRAY_SIZE(dma_remap_fault_reasons)) {
1496 *fault_type = DMA_REMAP;
1497 return dma_remap_fault_reasons[fault_reason];
1498 } else {
1499 *fault_type = UNKNOWN;
1500 return "Unknown";
1505 static inline int dmar_msi_reg(struct intel_iommu *iommu, int irq)
1507 if (iommu->irq == irq)
1508 return DMAR_FECTL_REG;
1509 else if (iommu->pr_irq == irq)
1510 return DMAR_PECTL_REG;
1511 else
1512 BUG();
1515 void dmar_msi_unmask(struct irq_data *data)
1517 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data);
1518 int reg = dmar_msi_reg(iommu, data->irq);
1519 unsigned long flag;
1521 /* unmask it */
1522 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1523 writel(0, iommu->reg + reg);
1524 /* Read a reg to force flush the post write */
1525 readl(iommu->reg + reg);
1526 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1529 void dmar_msi_mask(struct irq_data *data)
1531 struct intel_iommu *iommu = irq_data_get_irq_handler_data(data);
1532 int reg = dmar_msi_reg(iommu, data->irq);
1533 unsigned long flag;
1535 /* mask it */
1536 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1537 writel(DMA_FECTL_IM, iommu->reg + reg);
1538 /* Read a reg to force flush the post write */
1539 readl(iommu->reg + reg);
1540 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1543 void dmar_msi_write(int irq, struct msi_msg *msg)
1545 struct intel_iommu *iommu = irq_get_handler_data(irq);
1546 int reg = dmar_msi_reg(iommu, irq);
1547 unsigned long flag;
1549 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1550 writel(msg->data, iommu->reg + reg + 4);
1551 writel(msg->address_lo, iommu->reg + reg + 8);
1552 writel(msg->address_hi, iommu->reg + reg + 12);
1553 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1556 void dmar_msi_read(int irq, struct msi_msg *msg)
1558 struct intel_iommu *iommu = irq_get_handler_data(irq);
1559 int reg = dmar_msi_reg(iommu, irq);
1560 unsigned long flag;
1562 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1563 msg->data = readl(iommu->reg + reg + 4);
1564 msg->address_lo = readl(iommu->reg + reg + 8);
1565 msg->address_hi = readl(iommu->reg + reg + 12);
1566 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1569 static int dmar_fault_do_one(struct intel_iommu *iommu, int type,
1570 u8 fault_reason, u16 source_id, unsigned long long addr)
1572 const char *reason;
1573 int fault_type;
1575 reason = dmar_get_fault_reason(fault_reason, &fault_type);
1577 if (fault_type == INTR_REMAP)
1578 pr_err("INTR-REMAP: Request device [[%02x:%02x.%d] "
1579 "fault index %llx\n"
1580 "INTR-REMAP:[fault reason %02d] %s\n",
1581 (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1582 PCI_FUNC(source_id & 0xFF), addr >> 48,
1583 fault_reason, reason);
1584 else
1585 pr_err("DMAR:[%s] Request device [%02x:%02x.%d] "
1586 "fault addr %llx \n"
1587 "DMAR:[fault reason %02d] %s\n",
1588 (type ? "DMA Read" : "DMA Write"),
1589 (source_id >> 8), PCI_SLOT(source_id & 0xFF),
1590 PCI_FUNC(source_id & 0xFF), addr, fault_reason, reason);
1591 return 0;
1594 #define PRIMARY_FAULT_REG_LEN (16)
1595 irqreturn_t dmar_fault(int irq, void *dev_id)
1597 struct intel_iommu *iommu = dev_id;
1598 int reg, fault_index;
1599 u32 fault_status;
1600 unsigned long flag;
1602 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1603 fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1604 if (fault_status)
1605 pr_err("DRHD: handling fault status reg %x\n", fault_status);
1607 /* TBD: ignore advanced fault log currently */
1608 if (!(fault_status & DMA_FSTS_PPF))
1609 goto unlock_exit;
1611 fault_index = dma_fsts_fault_record_index(fault_status);
1612 reg = cap_fault_reg_offset(iommu->cap);
1613 while (1) {
1614 u8 fault_reason;
1615 u16 source_id;
1616 u64 guest_addr;
1617 int type;
1618 u32 data;
1620 /* highest 32 bits */
1621 data = readl(iommu->reg + reg +
1622 fault_index * PRIMARY_FAULT_REG_LEN + 12);
1623 if (!(data & DMA_FRCD_F))
1624 break;
1626 fault_reason = dma_frcd_fault_reason(data);
1627 type = dma_frcd_type(data);
1629 data = readl(iommu->reg + reg +
1630 fault_index * PRIMARY_FAULT_REG_LEN + 8);
1631 source_id = dma_frcd_source_id(data);
1633 guest_addr = dmar_readq(iommu->reg + reg +
1634 fault_index * PRIMARY_FAULT_REG_LEN);
1635 guest_addr = dma_frcd_page_addr(guest_addr);
1636 /* clear the fault */
1637 writel(DMA_FRCD_F, iommu->reg + reg +
1638 fault_index * PRIMARY_FAULT_REG_LEN + 12);
1640 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1642 dmar_fault_do_one(iommu, type, fault_reason,
1643 source_id, guest_addr);
1645 fault_index++;
1646 if (fault_index >= cap_num_fault_regs(iommu->cap))
1647 fault_index = 0;
1648 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1651 writel(DMA_FSTS_PFO | DMA_FSTS_PPF, iommu->reg + DMAR_FSTS_REG);
1653 unlock_exit:
1654 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1655 return IRQ_HANDLED;
1658 int dmar_set_interrupt(struct intel_iommu *iommu)
1660 int irq, ret;
1663 * Check if the fault interrupt is already initialized.
1665 if (iommu->irq)
1666 return 0;
1668 irq = dmar_alloc_hwirq(iommu->seq_id, iommu->node, iommu);
1669 if (irq > 0) {
1670 iommu->irq = irq;
1671 } else {
1672 pr_err("No free IRQ vectors\n");
1673 return -EINVAL;
1676 ret = request_irq(irq, dmar_fault, IRQF_NO_THREAD, iommu->name, iommu);
1677 if (ret)
1678 pr_err("Can't request irq\n");
1679 return ret;
1682 int __init enable_drhd_fault_handling(void)
1684 struct dmar_drhd_unit *drhd;
1685 struct intel_iommu *iommu;
1688 * Enable fault control interrupt.
1690 for_each_iommu(iommu, drhd) {
1691 u32 fault_status;
1692 int ret = dmar_set_interrupt(iommu);
1694 if (ret) {
1695 pr_err("DRHD %Lx: failed to enable fault, interrupt, ret %d\n",
1696 (unsigned long long)drhd->reg_base_addr, ret);
1697 return -1;
1701 * Clear any previous faults.
1703 dmar_fault(iommu->irq, iommu);
1704 fault_status = readl(iommu->reg + DMAR_FSTS_REG);
1705 writel(fault_status, iommu->reg + DMAR_FSTS_REG);
1708 return 0;
1712 * Re-enable Queued Invalidation interface.
1714 int dmar_reenable_qi(struct intel_iommu *iommu)
1716 if (!ecap_qis(iommu->ecap))
1717 return -ENOENT;
1719 if (!iommu->qi)
1720 return -ENOENT;
1723 * First disable queued invalidation.
1725 dmar_disable_qi(iommu);
1727 * Then enable queued invalidation again. Since there is no pending
1728 * invalidation requests now, it's safe to re-enable queued
1729 * invalidation.
1731 __dmar_enable_qi(iommu);
1733 return 0;
1737 * Check interrupt remapping support in DMAR table description.
1739 int __init dmar_ir_support(void)
1741 struct acpi_table_dmar *dmar;
1742 dmar = (struct acpi_table_dmar *)dmar_tbl;
1743 if (!dmar)
1744 return 0;
1745 return dmar->flags & 0x1;
1748 /* Check whether DMAR units are in use */
1749 static inline bool dmar_in_use(void)
1751 return irq_remapping_enabled || intel_iommu_enabled;
1754 static int __init dmar_free_unused_resources(void)
1756 struct dmar_drhd_unit *dmaru, *dmaru_n;
1758 if (dmar_in_use())
1759 return 0;
1761 if (dmar_dev_scope_status != 1 && !list_empty(&dmar_drhd_units))
1762 bus_unregister_notifier(&pci_bus_type, &dmar_pci_bus_nb);
1764 down_write(&dmar_global_lock);
1765 list_for_each_entry_safe(dmaru, dmaru_n, &dmar_drhd_units, list) {
1766 list_del(&dmaru->list);
1767 dmar_free_drhd(dmaru);
1769 up_write(&dmar_global_lock);
1771 return 0;
1774 late_initcall(dmar_free_unused_resources);
1775 IOMMU_INIT_POST(detect_intel_iommu);
1778 * DMAR Hotplug Support
1779 * For more details, please refer to Intel(R) Virtualization Technology
1780 * for Directed-IO Architecture Specifiction, Rev 2.2, Section 8.8
1781 * "Remapping Hardware Unit Hot Plug".
1783 static u8 dmar_hp_uuid[] = {
1784 /* 0000 */ 0xA6, 0xA3, 0xC1, 0xD8, 0x9B, 0xBE, 0x9B, 0x4C,
1785 /* 0008 */ 0x91, 0xBF, 0xC3, 0xCB, 0x81, 0xFC, 0x5D, 0xAF
1789 * Currently there's only one revision and BIOS will not check the revision id,
1790 * so use 0 for safety.
1792 #define DMAR_DSM_REV_ID 0
1793 #define DMAR_DSM_FUNC_DRHD 1
1794 #define DMAR_DSM_FUNC_ATSR 2
1795 #define DMAR_DSM_FUNC_RHSA 3
1797 static inline bool dmar_detect_dsm(acpi_handle handle, int func)
1799 return acpi_check_dsm(handle, dmar_hp_uuid, DMAR_DSM_REV_ID, 1 << func);
1802 static int dmar_walk_dsm_resource(acpi_handle handle, int func,
1803 dmar_res_handler_t handler, void *arg)
1805 int ret = -ENODEV;
1806 union acpi_object *obj;
1807 struct acpi_dmar_header *start;
1808 struct dmar_res_callback callback;
1809 static int res_type[] = {
1810 [DMAR_DSM_FUNC_DRHD] = ACPI_DMAR_TYPE_HARDWARE_UNIT,
1811 [DMAR_DSM_FUNC_ATSR] = ACPI_DMAR_TYPE_ROOT_ATS,
1812 [DMAR_DSM_FUNC_RHSA] = ACPI_DMAR_TYPE_HARDWARE_AFFINITY,
1815 if (!dmar_detect_dsm(handle, func))
1816 return 0;
1818 obj = acpi_evaluate_dsm_typed(handle, dmar_hp_uuid, DMAR_DSM_REV_ID,
1819 func, NULL, ACPI_TYPE_BUFFER);
1820 if (!obj)
1821 return -ENODEV;
1823 memset(&callback, 0, sizeof(callback));
1824 callback.cb[res_type[func]] = handler;
1825 callback.arg[res_type[func]] = arg;
1826 start = (struct acpi_dmar_header *)obj->buffer.pointer;
1827 ret = dmar_walk_remapping_entries(start, obj->buffer.length, &callback);
1829 ACPI_FREE(obj);
1831 return ret;
1834 static int dmar_hp_add_drhd(struct acpi_dmar_header *header, void *arg)
1836 int ret;
1837 struct dmar_drhd_unit *dmaru;
1839 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header);
1840 if (!dmaru)
1841 return -ENODEV;
1843 ret = dmar_ir_hotplug(dmaru, true);
1844 if (ret == 0)
1845 ret = dmar_iommu_hotplug(dmaru, true);
1847 return ret;
1850 static int dmar_hp_remove_drhd(struct acpi_dmar_header *header, void *arg)
1852 int i, ret;
1853 struct device *dev;
1854 struct dmar_drhd_unit *dmaru;
1856 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header);
1857 if (!dmaru)
1858 return 0;
1861 * All PCI devices managed by this unit should have been destroyed.
1863 if (!dmaru->include_all && dmaru->devices && dmaru->devices_cnt) {
1864 for_each_active_dev_scope(dmaru->devices,
1865 dmaru->devices_cnt, i, dev)
1866 return -EBUSY;
1869 ret = dmar_ir_hotplug(dmaru, false);
1870 if (ret == 0)
1871 ret = dmar_iommu_hotplug(dmaru, false);
1873 return ret;
1876 static int dmar_hp_release_drhd(struct acpi_dmar_header *header, void *arg)
1878 struct dmar_drhd_unit *dmaru;
1880 dmaru = dmar_find_dmaru((struct acpi_dmar_hardware_unit *)header);
1881 if (dmaru) {
1882 list_del_rcu(&dmaru->list);
1883 synchronize_rcu();
1884 dmar_free_drhd(dmaru);
1887 return 0;
1890 static int dmar_hotplug_insert(acpi_handle handle)
1892 int ret;
1893 int drhd_count = 0;
1895 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1896 &dmar_validate_one_drhd, (void *)1);
1897 if (ret)
1898 goto out;
1900 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1901 &dmar_parse_one_drhd, (void *)&drhd_count);
1902 if (ret == 0 && drhd_count == 0) {
1903 pr_warn(FW_BUG "No DRHD structures in buffer returned by _DSM method\n");
1904 goto out;
1905 } else if (ret) {
1906 goto release_drhd;
1909 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_RHSA,
1910 &dmar_parse_one_rhsa, NULL);
1911 if (ret)
1912 goto release_drhd;
1914 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR,
1915 &dmar_parse_one_atsr, NULL);
1916 if (ret)
1917 goto release_atsr;
1919 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1920 &dmar_hp_add_drhd, NULL);
1921 if (!ret)
1922 return 0;
1924 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1925 &dmar_hp_remove_drhd, NULL);
1926 release_atsr:
1927 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR,
1928 &dmar_release_one_atsr, NULL);
1929 release_drhd:
1930 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1931 &dmar_hp_release_drhd, NULL);
1932 out:
1933 return ret;
1936 static int dmar_hotplug_remove(acpi_handle handle)
1938 int ret;
1940 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR,
1941 &dmar_check_one_atsr, NULL);
1942 if (ret)
1943 return ret;
1945 ret = dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1946 &dmar_hp_remove_drhd, NULL);
1947 if (ret == 0) {
1948 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_ATSR,
1949 &dmar_release_one_atsr, NULL));
1950 WARN_ON(dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1951 &dmar_hp_release_drhd, NULL));
1952 } else {
1953 dmar_walk_dsm_resource(handle, DMAR_DSM_FUNC_DRHD,
1954 &dmar_hp_add_drhd, NULL);
1957 return ret;
1960 static acpi_status dmar_get_dsm_handle(acpi_handle handle, u32 lvl,
1961 void *context, void **retval)
1963 acpi_handle *phdl = retval;
1965 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) {
1966 *phdl = handle;
1967 return AE_CTRL_TERMINATE;
1970 return AE_OK;
1973 static int dmar_device_hotplug(acpi_handle handle, bool insert)
1975 int ret;
1976 acpi_handle tmp = NULL;
1977 acpi_status status;
1979 if (!dmar_in_use())
1980 return 0;
1982 if (dmar_detect_dsm(handle, DMAR_DSM_FUNC_DRHD)) {
1983 tmp = handle;
1984 } else {
1985 status = acpi_walk_namespace(ACPI_TYPE_DEVICE, handle,
1986 ACPI_UINT32_MAX,
1987 dmar_get_dsm_handle,
1988 NULL, NULL, &tmp);
1989 if (ACPI_FAILURE(status)) {
1990 pr_warn("Failed to locate _DSM method.\n");
1991 return -ENXIO;
1994 if (tmp == NULL)
1995 return 0;
1997 down_write(&dmar_global_lock);
1998 if (insert)
1999 ret = dmar_hotplug_insert(tmp);
2000 else
2001 ret = dmar_hotplug_remove(tmp);
2002 up_write(&dmar_global_lock);
2004 return ret;
2007 int dmar_device_add(acpi_handle handle)
2009 return dmar_device_hotplug(handle, true);
2012 int dmar_device_remove(acpi_handle handle)
2014 return dmar_device_hotplug(handle, false);