aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/iommu/intel-iommu.c
diff options
context:
space:
mode:
authorTejun Heo <tj@kernel.org>2011-11-28 12:46:22 -0500
committerTejun Heo <tj@kernel.org>2011-11-28 12:46:22 -0500
commitd4bbf7e7759afc172e2bfbc5c416324590049cdd (patch)
tree7eab5ee5481cd3dcf1162329fec827177640018a /drivers/iommu/intel-iommu.c
parenta150439c4a97db379f0ed6faa46fbbb6e7bf3cb2 (diff)
parent401d0069cb344f401bc9d264c31db55876ff78c0 (diff)
Merge branch 'master' into x86/memblock
Conflicts & resolutions: * arch/x86/xen/setup.c dc91c728fd "xen: allow extra memory to be in multiple regions" 24aa07882b "memblock, x86: Replace memblock_x86_reserve/free..." conflicted on xen_add_extra_mem() updates. The resolution is trivial as the latter just want to replace memblock_x86_reserve_range() with memblock_reserve(). * drivers/pci/intel-iommu.c 166e9278a3f "x86/ia64: intel-iommu: move to drivers/iommu/" 5dfe8660a3d "bootmem: Replace work_with_active_regions() with..." conflicted as the former moved the file under drivers/iommu/. Resolved by applying the chnages from the latter on the moved file. * mm/Kconfig 6661672053a "memblock: add NO_BOOTMEM config symbol" c378ddd53f9 "memblock, x86: Make ARCH_DISCARD_MEMBLOCK a config option" conflicted trivially. Both added config options. Just letting both add their own options resolves the conflict. * mm/memblock.c d1f0ece6cdc "mm/memblock.c: small function definition fixes" ed7b56a799c "memblock: Remove memblock_memory_can_coalesce()" confliected. The former updates function removed by the latter. Resolution is trivial. Signed-off-by: Tejun Heo <tj@kernel.org>
Diffstat (limited to 'drivers/iommu/intel-iommu.c')
-rw-r--r--drivers/iommu/intel-iommu.c4173
1 files changed, 4173 insertions, 0 deletions
diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
new file mode 100644
index 000000000000..bcbd693b351a
--- /dev/null
+++ b/drivers/iommu/intel-iommu.c
@@ -0,0 +1,4173 @@
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 * Author: Fenghua Yu <fenghua.yu@intel.com>
22 */
23
24#include <linux/init.h>
25#include <linux/bitmap.h>
26#include <linux/debugfs.h>
27#include <linux/export.h>
28#include <linux/slab.h>
29#include <linux/irq.h>
30#include <linux/interrupt.h>
31#include <linux/spinlock.h>
32#include <linux/pci.h>
33#include <linux/dmar.h>
34#include <linux/dma-mapping.h>
35#include <linux/mempool.h>
36#include <linux/timer.h>
37#include <linux/iova.h>
38#include <linux/iommu.h>
39#include <linux/intel-iommu.h>
40#include <linux/syscore_ops.h>
41#include <linux/tboot.h>
42#include <linux/dmi.h>
43#include <linux/pci-ats.h>
44#include <asm/cacheflush.h>
45#include <asm/iommu.h>
46
47#define ROOT_SIZE VTD_PAGE_SIZE
48#define CONTEXT_SIZE VTD_PAGE_SIZE
49
50#define IS_BRIDGE_HOST_DEVICE(pdev) \
51 ((pdev->class >> 8) == PCI_CLASS_BRIDGE_HOST)
52#define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY)
53#define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA)
54#define IS_AZALIA(pdev) ((pdev)->vendor == 0x8086 && (pdev)->device == 0x3a3e)
55
56#define IOAPIC_RANGE_START (0xfee00000)
57#define IOAPIC_RANGE_END (0xfeefffff)
58#define IOVA_START_ADDR (0x1000)
59
60#define DEFAULT_DOMAIN_ADDRESS_WIDTH 48
61
62#define MAX_AGAW_WIDTH 64
63
64#define __DOMAIN_MAX_PFN(gaw) ((((uint64_t)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
65#define __DOMAIN_MAX_ADDR(gaw) ((((uint64_t)1) << gaw) - 1)
66
67/* We limit DOMAIN_MAX_PFN to fit in an unsigned long, and DOMAIN_MAX_ADDR
68 to match. That way, we can use 'unsigned long' for PFNs with impunity. */
69#define DOMAIN_MAX_PFN(gaw) ((unsigned long) min_t(uint64_t, \
70 __DOMAIN_MAX_PFN(gaw), (unsigned long)-1))
71#define DOMAIN_MAX_ADDR(gaw) (((uint64_t)__DOMAIN_MAX_PFN(gaw)) << VTD_PAGE_SHIFT)
72
73#define IOVA_PFN(addr) ((addr) >> PAGE_SHIFT)
74#define DMA_32BIT_PFN IOVA_PFN(DMA_BIT_MASK(32))
75#define DMA_64BIT_PFN IOVA_PFN(DMA_BIT_MASK(64))
76
77/* page table handling */
78#define LEVEL_STRIDE (9)
79#define LEVEL_MASK (((u64)1 << LEVEL_STRIDE) - 1)
80
81static inline int agaw_to_level(int agaw)
82{
83 return agaw + 2;
84}
85
86static inline int agaw_to_width(int agaw)
87{
88 return 30 + agaw * LEVEL_STRIDE;
89}
90
91static inline int width_to_agaw(int width)
92{
93 return (width - 30) / LEVEL_STRIDE;
94}
95
96static inline unsigned int level_to_offset_bits(int level)
97{
98 return (level - 1) * LEVEL_STRIDE;
99}
100
101static inline int pfn_level_offset(unsigned long pfn, int level)
102{
103 return (pfn >> level_to_offset_bits(level)) & LEVEL_MASK;
104}
105
106static inline unsigned long level_mask(int level)
107{
108 return -1UL << level_to_offset_bits(level);
109}
110
111static inline unsigned long level_size(int level)
112{
113 return 1UL << level_to_offset_bits(level);
114}
115
116static inline unsigned long align_to_level(unsigned long pfn, int level)
117{
118 return (pfn + level_size(level) - 1) & level_mask(level);
119}
120
121static inline unsigned long lvl_to_nr_pages(unsigned int lvl)
122{
123 return 1 << ((lvl - 1) * LEVEL_STRIDE);
124}
125
126/* VT-d pages must always be _smaller_ than MM pages. Otherwise things
127 are never going to work. */
128static inline unsigned long dma_to_mm_pfn(unsigned long dma_pfn)
129{
130 return dma_pfn >> (PAGE_SHIFT - VTD_PAGE_SHIFT);
131}
132
133static inline unsigned long mm_to_dma_pfn(unsigned long mm_pfn)
134{
135 return mm_pfn << (PAGE_SHIFT - VTD_PAGE_SHIFT);
136}
137static inline unsigned long page_to_dma_pfn(struct page *pg)
138{
139 return mm_to_dma_pfn(page_to_pfn(pg));
140}
141static inline unsigned long virt_to_dma_pfn(void *p)
142{
143 return page_to_dma_pfn(virt_to_page(p));
144}
145
146/* global iommu list, set NULL for ignored DMAR units */
147static struct intel_iommu **g_iommus;
148
149static void __init check_tylersburg_isoch(void);
150static int rwbf_quirk;
151
152/*
153 * set to 1 to panic kernel if can't successfully enable VT-d
154 * (used when kernel is launched w/ TXT)
155 */
156static int force_on = 0;
157
158/*
159 * 0: Present
160 * 1-11: Reserved
161 * 12-63: Context Ptr (12 - (haw-1))
162 * 64-127: Reserved
163 */
164struct root_entry {
165 u64 val;
166 u64 rsvd1;
167};
168#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry))
169static inline bool root_present(struct root_entry *root)
170{
171 return (root->val & 1);
172}
173static inline void set_root_present(struct root_entry *root)
174{
175 root->val |= 1;
176}
177static inline void set_root_value(struct root_entry *root, unsigned long value)
178{
179 root->val |= value & VTD_PAGE_MASK;
180}
181
182static inline struct context_entry *
183get_context_addr_from_root(struct root_entry *root)
184{
185 return (struct context_entry *)
186 (root_present(root)?phys_to_virt(
187 root->val & VTD_PAGE_MASK) :
188 NULL);
189}
190
191/*
192 * low 64 bits:
193 * 0: present
194 * 1: fault processing disable
195 * 2-3: translation type
196 * 12-63: address space root
197 * high 64 bits:
198 * 0-2: address width
199 * 3-6: aval
200 * 8-23: domain id
201 */
202struct context_entry {
203 u64 lo;
204 u64 hi;
205};
206
207static inline bool context_present(struct context_entry *context)
208{
209 return (context->lo & 1);
210}
211static inline void context_set_present(struct context_entry *context)
212{
213 context->lo |= 1;
214}
215
216static inline void context_set_fault_enable(struct context_entry *context)
217{
218 context->lo &= (((u64)-1) << 2) | 1;
219}
220
221static inline void context_set_translation_type(struct context_entry *context,
222 unsigned long value)
223{
224 context->lo &= (((u64)-1) << 4) | 3;
225 context->lo |= (value & 3) << 2;
226}
227
228static inline void context_set_address_root(struct context_entry *context,
229 unsigned long value)
230{
231 context->lo |= value & VTD_PAGE_MASK;
232}
233
234static inline void context_set_address_width(struct context_entry *context,
235 unsigned long value)
236{
237 context->hi |= value & 7;
238}
239
240static inline void context_set_domain_id(struct context_entry *context,
241 unsigned long value)
242{
243 context->hi |= (value & ((1 << 16) - 1)) << 8;
244}
245
246static inline void context_clear_entry(struct context_entry *context)
247{
248 context->lo = 0;
249 context->hi = 0;
250}
251
252/*
253 * 0: readable
254 * 1: writable
255 * 2-6: reserved
256 * 7: super page
257 * 8-10: available
258 * 11: snoop behavior
259 * 12-63: Host physcial address
260 */
261struct dma_pte {
262 u64 val;
263};
264
265static inline void dma_clear_pte(struct dma_pte *pte)
266{
267 pte->val = 0;
268}
269
270static inline void dma_set_pte_readable(struct dma_pte *pte)
271{
272 pte->val |= DMA_PTE_READ;
273}
274
275static inline void dma_set_pte_writable(struct dma_pte *pte)
276{
277 pte->val |= DMA_PTE_WRITE;
278}
279
280static inline void dma_set_pte_snp(struct dma_pte *pte)
281{
282 pte->val |= DMA_PTE_SNP;
283}
284
285static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot)
286{
287 pte->val = (pte->val & ~3) | (prot & 3);
288}
289
290static inline u64 dma_pte_addr(struct dma_pte *pte)
291{
292#ifdef CONFIG_64BIT
293 return pte->val & VTD_PAGE_MASK;
294#else
295 /* Must have a full atomic 64-bit read */
296 return __cmpxchg64(&pte->val, 0ULL, 0ULL) & VTD_PAGE_MASK;
297#endif
298}
299
300static inline void dma_set_pte_pfn(struct dma_pte *pte, unsigned long pfn)
301{
302 pte->val |= (uint64_t)pfn << VTD_PAGE_SHIFT;
303}
304
305static inline bool dma_pte_present(struct dma_pte *pte)
306{
307 return (pte->val & 3) != 0;
308}
309
310static inline bool dma_pte_superpage(struct dma_pte *pte)
311{
312 return (pte->val & (1 << 7));
313}
314
315static inline int first_pte_in_page(struct dma_pte *pte)
316{
317 return !((unsigned long)pte & ~VTD_PAGE_MASK);
318}
319
320/*
321 * This domain is a statically identity mapping domain.
322 * 1. This domain creats a static 1:1 mapping to all usable memory.
323 * 2. It maps to each iommu if successful.
324 * 3. Each iommu mapps to this domain if successful.
325 */
326static struct dmar_domain *si_domain;
327static int hw_pass_through = 1;
328
329/* devices under the same p2p bridge are owned in one domain */
330#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0)
331
332/* domain represents a virtual machine, more than one devices
333 * across iommus may be owned in one domain, e.g. kvm guest.
334 */
335#define DOMAIN_FLAG_VIRTUAL_MACHINE (1 << 1)
336
337/* si_domain contains mulitple devices */
338#define DOMAIN_FLAG_STATIC_IDENTITY (1 << 2)
339
340struct dmar_domain {
341 int id; /* domain id */
342 int nid; /* node id */
343 unsigned long iommu_bmp; /* bitmap of iommus this domain uses*/
344
345 struct list_head devices; /* all devices' list */
346 struct iova_domain iovad; /* iova's that belong to this domain */
347
348 struct dma_pte *pgd; /* virtual address */
349 int gaw; /* max guest address width */
350
351 /* adjusted guest address width, 0 is level 2 30-bit */
352 int agaw;
353
354 int flags; /* flags to find out type of domain */
355
356 int iommu_coherency;/* indicate coherency of iommu access */
357 int iommu_snooping; /* indicate snooping control feature*/
358 int iommu_count; /* reference count of iommu */
359 int iommu_superpage;/* Level of superpages supported:
360 0 == 4KiB (no superpages), 1 == 2MiB,
361 2 == 1GiB, 3 == 512GiB, 4 == 1TiB */
362 spinlock_t iommu_lock; /* protect iommu set in domain */
363 u64 max_addr; /* maximum mapped address */
364};
365
366/* PCI domain-device relationship */
367struct device_domain_info {
368 struct list_head link; /* link to domain siblings */
369 struct list_head global; /* link to global list */
370 int segment; /* PCI domain */
371 u8 bus; /* PCI bus number */
372 u8 devfn; /* PCI devfn number */
373 struct pci_dev *dev; /* it's NULL for PCIe-to-PCI bridge */
374 struct intel_iommu *iommu; /* IOMMU used by this device */
375 struct dmar_domain *domain; /* pointer to domain */
376};
377
378static void flush_unmaps_timeout(unsigned long data);
379
380DEFINE_TIMER(unmap_timer, flush_unmaps_timeout, 0, 0);
381
382#define HIGH_WATER_MARK 250
383struct deferred_flush_tables {
384 int next;
385 struct iova *iova[HIGH_WATER_MARK];
386 struct dmar_domain *domain[HIGH_WATER_MARK];
387};
388
389static struct deferred_flush_tables *deferred_flush;
390
391/* bitmap for indexing intel_iommus */
392static int g_num_of_iommus;
393
394static DEFINE_SPINLOCK(async_umap_flush_lock);
395static LIST_HEAD(unmaps_to_do);
396
397static int timer_on;
398static long list_size;
399
400static void domain_remove_dev_info(struct dmar_domain *domain);
401
402#ifdef CONFIG_INTEL_IOMMU_DEFAULT_ON
403int dmar_disabled = 0;
404#else
405int dmar_disabled = 1;
406#endif /*CONFIG_INTEL_IOMMU_DEFAULT_ON*/
407
408static int dmar_map_gfx = 1;
409static int dmar_forcedac;
410static int intel_iommu_strict;
411static int intel_iommu_superpage = 1;
412
413int intel_iommu_gfx_mapped;
414EXPORT_SYMBOL_GPL(intel_iommu_gfx_mapped);
415
416#define DUMMY_DEVICE_DOMAIN_INFO ((struct device_domain_info *)(-1))
417static DEFINE_SPINLOCK(device_domain_lock);
418static LIST_HEAD(device_domain_list);
419
420static struct iommu_ops intel_iommu_ops;
421
422static int __init intel_iommu_setup(char *str)
423{
424 if (!str)
425 return -EINVAL;
426 while (*str) {
427 if (!strncmp(str, "on", 2)) {
428 dmar_disabled = 0;
429 printk(KERN_INFO "Intel-IOMMU: enabled\n");
430 } else if (!strncmp(str, "off", 3)) {
431 dmar_disabled = 1;
432 printk(KERN_INFO "Intel-IOMMU: disabled\n");
433 } else if (!strncmp(str, "igfx_off", 8)) {
434 dmar_map_gfx = 0;
435 printk(KERN_INFO
436 "Intel-IOMMU: disable GFX device mapping\n");
437 } else if (!strncmp(str, "forcedac", 8)) {
438 printk(KERN_INFO
439 "Intel-IOMMU: Forcing DAC for PCI devices\n");
440 dmar_forcedac = 1;
441 } else if (!strncmp(str, "strict", 6)) {
442 printk(KERN_INFO
443 "Intel-IOMMU: disable batched IOTLB flush\n");
444 intel_iommu_strict = 1;
445 } else if (!strncmp(str, "sp_off", 6)) {
446 printk(KERN_INFO
447 "Intel-IOMMU: disable supported super page\n");
448 intel_iommu_superpage = 0;
449 }
450
451 str += strcspn(str, ",");
452 while (*str == ',')
453 str++;
454 }
455 return 0;
456}
457__setup("intel_iommu=", intel_iommu_setup);
458
459static struct kmem_cache *iommu_domain_cache;
460static struct kmem_cache *iommu_devinfo_cache;
461static struct kmem_cache *iommu_iova_cache;
462
463static inline void *alloc_pgtable_page(int node)
464{
465 struct page *page;
466 void *vaddr = NULL;
467
468 page = alloc_pages_node(node, GFP_ATOMIC | __GFP_ZERO, 0);
469 if (page)
470 vaddr = page_address(page);
471 return vaddr;
472}
473
474static inline void free_pgtable_page(void *vaddr)
475{
476 free_page((unsigned long)vaddr);
477}
478
479static inline void *alloc_domain_mem(void)
480{
481 return kmem_cache_alloc(iommu_domain_cache, GFP_ATOMIC);
482}
483
484static void free_domain_mem(void *vaddr)
485{
486 kmem_cache_free(iommu_domain_cache, vaddr);
487}
488
489static inline void * alloc_devinfo_mem(void)
490{
491 return kmem_cache_alloc(iommu_devinfo_cache, GFP_ATOMIC);
492}
493
494static inline void free_devinfo_mem(void *vaddr)
495{
496 kmem_cache_free(iommu_devinfo_cache, vaddr);
497}
498
499struct iova *alloc_iova_mem(void)
500{
501 return kmem_cache_alloc(iommu_iova_cache, GFP_ATOMIC);
502}
503
504void free_iova_mem(struct iova *iova)
505{
506 kmem_cache_free(iommu_iova_cache, iova);
507}
508
509
510static int __iommu_calculate_agaw(struct intel_iommu *iommu, int max_gaw)
511{
512 unsigned long sagaw;
513 int agaw = -1;
514
515 sagaw = cap_sagaw(iommu->cap);
516 for (agaw = width_to_agaw(max_gaw);
517 agaw >= 0; agaw--) {
518 if (test_bit(agaw, &sagaw))
519 break;
520 }
521
522 return agaw;
523}
524
525/*
526 * Calculate max SAGAW for each iommu.
527 */
528int iommu_calculate_max_sagaw(struct intel_iommu *iommu)
529{
530 return __iommu_calculate_agaw(iommu, MAX_AGAW_WIDTH);
531}
532
533/*
534 * calculate agaw for each iommu.
535 * "SAGAW" may be different across iommus, use a default agaw, and
536 * get a supported less agaw for iommus that don't support the default agaw.
537 */
538int iommu_calculate_agaw(struct intel_iommu *iommu)
539{
540 return __iommu_calculate_agaw(iommu, DEFAULT_DOMAIN_ADDRESS_WIDTH);
541}
542
543/* This functionin only returns single iommu in a domain */
544static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
545{
546 int iommu_id;
547
548 /* si_domain and vm domain should not get here. */
549 BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE);
550 BUG_ON(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY);
551
552 iommu_id = find_first_bit(&domain->iommu_bmp, g_num_of_iommus);
553 if (iommu_id < 0 || iommu_id >= g_num_of_iommus)
554 return NULL;
555
556 return g_iommus[iommu_id];
557}
558
559static void domain_update_iommu_coherency(struct dmar_domain *domain)
560{
561 int i;
562
563 domain->iommu_coherency = 1;
564
565 for_each_set_bit(i, &domain->iommu_bmp, g_num_of_iommus) {
566 if (!ecap_coherent(g_iommus[i]->ecap)) {
567 domain->iommu_coherency = 0;
568 break;
569 }
570 }
571}
572
573static void domain_update_iommu_snooping(struct dmar_domain *domain)
574{
575 int i;
576
577 domain->iommu_snooping = 1;
578
579 for_each_set_bit(i, &domain->iommu_bmp, g_num_of_iommus) {
580 if (!ecap_sc_support(g_iommus[i]->ecap)) {
581 domain->iommu_snooping = 0;
582 break;
583 }
584 }
585}
586
587static void domain_update_iommu_superpage(struct dmar_domain *domain)
588{
589 struct dmar_drhd_unit *drhd;
590 struct intel_iommu *iommu = NULL;
591 int mask = 0xf;
592
593 if (!intel_iommu_superpage) {
594 domain->iommu_superpage = 0;
595 return;
596 }
597
598 /* set iommu_superpage to the smallest common denominator */
599 for_each_active_iommu(iommu, drhd) {
600 mask &= cap_super_page_val(iommu->cap);
601 if (!mask) {
602 break;
603 }
604 }
605 domain->iommu_superpage = fls(mask);
606}
607
608/* Some capabilities may be different across iommus */
609static void domain_update_iommu_cap(struct dmar_domain *domain)
610{
611 domain_update_iommu_coherency(domain);
612 domain_update_iommu_snooping(domain);
613 domain_update_iommu_superpage(domain);
614}
615
616static struct intel_iommu *device_to_iommu(int segment, u8 bus, u8 devfn)
617{
618 struct dmar_drhd_unit *drhd = NULL;
619 int i;
620
621 for_each_drhd_unit(drhd) {
622 if (drhd->ignored)
623 continue;
624 if (segment != drhd->segment)
625 continue;
626
627 for (i = 0; i < drhd->devices_cnt; i++) {
628 if (drhd->devices[i] &&
629 drhd->devices[i]->bus->number == bus &&
630 drhd->devices[i]->devfn == devfn)
631 return drhd->iommu;
632 if (drhd->devices[i] &&
633 drhd->devices[i]->subordinate &&
634 drhd->devices[i]->subordinate->number <= bus &&
635 drhd->devices[i]->subordinate->subordinate >= bus)
636 return drhd->iommu;
637 }
638
639 if (drhd->include_all)
640 return drhd->iommu;
641 }
642
643 return NULL;
644}
645
646static void domain_flush_cache(struct dmar_domain *domain,
647 void *addr, int size)
648{
649 if (!domain->iommu_coherency)
650 clflush_cache_range(addr, size);
651}
652
653/* Gets context entry for a given bus and devfn */
654static struct context_entry * device_to_context_entry(struct intel_iommu *iommu,
655 u8 bus, u8 devfn)
656{
657 struct root_entry *root;
658 struct context_entry *context;
659 unsigned long phy_addr;
660 unsigned long flags;
661
662 spin_lock_irqsave(&iommu->lock, flags);
663 root = &iommu->root_entry[bus];
664 context = get_context_addr_from_root(root);
665 if (!context) {
666 context = (struct context_entry *)
667 alloc_pgtable_page(iommu->node);
668 if (!context) {
669 spin_unlock_irqrestore(&iommu->lock, flags);
670 return NULL;
671 }
672 __iommu_flush_cache(iommu, (void *)context, CONTEXT_SIZE);
673 phy_addr = virt_to_phys((void *)context);
674 set_root_value(root, phy_addr);
675 set_root_present(root);
676 __iommu_flush_cache(iommu, root, sizeof(*root));
677 }
678 spin_unlock_irqrestore(&iommu->lock, flags);
679 return &context[devfn];
680}
681
682static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn)
683{
684 struct root_entry *root;
685 struct context_entry *context;
686 int ret;
687 unsigned long flags;
688
689 spin_lock_irqsave(&iommu->lock, flags);
690 root = &iommu->root_entry[bus];
691 context = get_context_addr_from_root(root);
692 if (!context) {
693 ret = 0;
694 goto out;
695 }
696 ret = context_present(&context[devfn]);
697out:
698 spin_unlock_irqrestore(&iommu->lock, flags);
699 return ret;
700}
701
702static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn)
703{
704 struct root_entry *root;
705 struct context_entry *context;
706 unsigned long flags;
707
708 spin_lock_irqsave(&iommu->lock, flags);
709 root = &iommu->root_entry[bus];
710 context = get_context_addr_from_root(root);
711 if (context) {
712 context_clear_entry(&context[devfn]);
713 __iommu_flush_cache(iommu, &context[devfn], \
714 sizeof(*context));
715 }
716 spin_unlock_irqrestore(&iommu->lock, flags);
717}
718
719static void free_context_table(struct intel_iommu *iommu)
720{
721 struct root_entry *root;
722 int i;
723 unsigned long flags;
724 struct context_entry *context;
725
726 spin_lock_irqsave(&iommu->lock, flags);
727 if (!iommu->root_entry) {
728 goto out;
729 }
730 for (i = 0; i < ROOT_ENTRY_NR; i++) {
731 root = &iommu->root_entry[i];
732 context = get_context_addr_from_root(root);
733 if (context)
734 free_pgtable_page(context);
735 }
736 free_pgtable_page(iommu->root_entry);
737 iommu->root_entry = NULL;
738out:
739 spin_unlock_irqrestore(&iommu->lock, flags);
740}
741
742static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
743 unsigned long pfn, int target_level)
744{
745 int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
746 struct dma_pte *parent, *pte = NULL;
747 int level = agaw_to_level(domain->agaw);
748 int offset;
749
750 BUG_ON(!domain->pgd);
751 BUG_ON(addr_width < BITS_PER_LONG && pfn >> addr_width);
752 parent = domain->pgd;
753
754 while (level > 0) {
755 void *tmp_page;
756
757 offset = pfn_level_offset(pfn, level);
758 pte = &parent[offset];
759 if (!target_level && (dma_pte_superpage(pte) || !dma_pte_present(pte)))
760 break;
761 if (level == target_level)
762 break;
763
764 if (!dma_pte_present(pte)) {
765 uint64_t pteval;
766
767 tmp_page = alloc_pgtable_page(domain->nid);
768
769 if (!tmp_page)
770 return NULL;
771
772 domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE);
773 pteval = ((uint64_t)virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE;
774 if (cmpxchg64(&pte->val, 0ULL, pteval)) {
775 /* Someone else set it while we were thinking; use theirs. */
776 free_pgtable_page(tmp_page);
777 } else {
778 dma_pte_addr(pte);
779 domain_flush_cache(domain, pte, sizeof(*pte));
780 }
781 }
782 parent = phys_to_virt(dma_pte_addr(pte));
783 level--;
784 }
785
786 return pte;
787}
788
789
790/* return address's pte at specific level */
791static struct dma_pte *dma_pfn_level_pte(struct dmar_domain *domain,
792 unsigned long pfn,
793 int level, int *large_page)
794{
795 struct dma_pte *parent, *pte = NULL;
796 int total = agaw_to_level(domain->agaw);
797 int offset;
798
799 parent = domain->pgd;
800 while (level <= total) {
801 offset = pfn_level_offset(pfn, total);
802 pte = &parent[offset];
803 if (level == total)
804 return pte;
805
806 if (!dma_pte_present(pte)) {
807 *large_page = total;
808 break;
809 }
810
811 if (pte->val & DMA_PTE_LARGE_PAGE) {
812 *large_page = total;
813 return pte;
814 }
815
816 parent = phys_to_virt(dma_pte_addr(pte));
817 total--;
818 }
819 return NULL;
820}
821
822/* clear last level pte, a tlb flush should be followed */
823static int dma_pte_clear_range(struct dmar_domain *domain,
824 unsigned long start_pfn,
825 unsigned long last_pfn)
826{
827 int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
828 unsigned int large_page = 1;
829 struct dma_pte *first_pte, *pte;
830 int order;
831
832 BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
833 BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
834 BUG_ON(start_pfn > last_pfn);
835
836 /* we don't need lock here; nobody else touches the iova range */
837 do {
838 large_page = 1;
839 first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1, &large_page);
840 if (!pte) {
841 start_pfn = align_to_level(start_pfn + 1, large_page + 1);
842 continue;
843 }
844 do {
845 dma_clear_pte(pte);
846 start_pfn += lvl_to_nr_pages(large_page);
847 pte++;
848 } while (start_pfn <= last_pfn && !first_pte_in_page(pte));
849
850 domain_flush_cache(domain, first_pte,
851 (void *)pte - (void *)first_pte);
852
853 } while (start_pfn && start_pfn <= last_pfn);
854
855 order = (large_page - 1) * 9;
856 return order;
857}
858
859/* free page table pages. last level pte should already be cleared */
860static void dma_pte_free_pagetable(struct dmar_domain *domain,
861 unsigned long start_pfn,
862 unsigned long last_pfn)
863{
864 int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
865 struct dma_pte *first_pte, *pte;
866 int total = agaw_to_level(domain->agaw);
867 int level;
868 unsigned long tmp;
869 int large_page = 2;
870
871 BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
872 BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
873 BUG_ON(start_pfn > last_pfn);
874
875 /* We don't need lock here; nobody else touches the iova range */
876 level = 2;
877 while (level <= total) {
878 tmp = align_to_level(start_pfn, level);
879
880 /* If we can't even clear one PTE at this level, we're done */
881 if (tmp + level_size(level) - 1 > last_pfn)
882 return;
883
884 do {
885 large_page = level;
886 first_pte = pte = dma_pfn_level_pte(domain, tmp, level, &large_page);
887 if (large_page > level)
888 level = large_page + 1;
889 if (!pte) {
890 tmp = align_to_level(tmp + 1, level + 1);
891 continue;
892 }
893 do {
894 if (dma_pte_present(pte)) {
895 free_pgtable_page(phys_to_virt(dma_pte_addr(pte)));
896 dma_clear_pte(pte);
897 }
898 pte++;
899 tmp += level_size(level);
900 } while (!first_pte_in_page(pte) &&
901 tmp + level_size(level) - 1 <= last_pfn);
902
903 domain_flush_cache(domain, first_pte,
904 (void *)pte - (void *)first_pte);
905
906 } while (tmp && tmp + level_size(level) - 1 <= last_pfn);
907 level++;
908 }
909 /* free pgd */
910 if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
911 free_pgtable_page(domain->pgd);
912 domain->pgd = NULL;
913 }
914}
915
916/* iommu handling */
917static int iommu_alloc_root_entry(struct intel_iommu *iommu)
918{
919 struct root_entry *root;
920 unsigned long flags;
921
922 root = (struct root_entry *)alloc_pgtable_page(iommu->node);
923 if (!root)
924 return -ENOMEM;
925
926 __iommu_flush_cache(iommu, root, ROOT_SIZE);
927
928 spin_lock_irqsave(&iommu->lock, flags);
929 iommu->root_entry = root;
930 spin_unlock_irqrestore(&iommu->lock, flags);
931
932 return 0;
933}
934
935static void iommu_set_root_entry(struct intel_iommu *iommu)
936{
937 void *addr;
938 u32 sts;
939 unsigned long flag;
940
941 addr = iommu->root_entry;
942
943 raw_spin_lock_irqsave(&iommu->register_lock, flag);
944 dmar_writeq(iommu->reg + DMAR_RTADDR_REG, virt_to_phys(addr));
945
946 writel(iommu->gcmd | DMA_GCMD_SRTP, iommu->reg + DMAR_GCMD_REG);
947
948 /* Make sure hardware complete it */
949 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
950 readl, (sts & DMA_GSTS_RTPS), sts);
951
952 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
953}
954
955static void iommu_flush_write_buffer(struct intel_iommu *iommu)
956{
957 u32 val;
958 unsigned long flag;
959
960 if (!rwbf_quirk && !cap_rwbf(iommu->cap))
961 return;
962
963 raw_spin_lock_irqsave(&iommu->register_lock, flag);
964 writel(iommu->gcmd | DMA_GCMD_WBF, iommu->reg + DMAR_GCMD_REG);
965
966 /* Make sure hardware complete it */
967 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
968 readl, (!(val & DMA_GSTS_WBFS)), val);
969
970 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
971}
972
973/* return value determine if we need a write buffer flush */
974static void __iommu_flush_context(struct intel_iommu *iommu,
975 u16 did, u16 source_id, u8 function_mask,
976 u64 type)
977{
978 u64 val = 0;
979 unsigned long flag;
980
981 switch (type) {
982 case DMA_CCMD_GLOBAL_INVL:
983 val = DMA_CCMD_GLOBAL_INVL;
984 break;
985 case DMA_CCMD_DOMAIN_INVL:
986 val = DMA_CCMD_DOMAIN_INVL|DMA_CCMD_DID(did);
987 break;
988 case DMA_CCMD_DEVICE_INVL:
989 val = DMA_CCMD_DEVICE_INVL|DMA_CCMD_DID(did)
990 | DMA_CCMD_SID(source_id) | DMA_CCMD_FM(function_mask);
991 break;
992 default:
993 BUG();
994 }
995 val |= DMA_CCMD_ICC;
996
997 raw_spin_lock_irqsave(&iommu->register_lock, flag);
998 dmar_writeq(iommu->reg + DMAR_CCMD_REG, val);
999
1000 /* Make sure hardware complete it */
1001 IOMMU_WAIT_OP(iommu, DMAR_CCMD_REG,
1002 dmar_readq, (!(val & DMA_CCMD_ICC)), val);
1003
1004 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1005}
1006
1007/* return value determine if we need a write buffer flush */
1008static void __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
1009 u64 addr, unsigned int size_order, u64 type)
1010{
1011 int tlb_offset = ecap_iotlb_offset(iommu->ecap);
1012 u64 val = 0, val_iva = 0;
1013 unsigned long flag;
1014
1015 switch (type) {
1016 case DMA_TLB_GLOBAL_FLUSH:
1017 /* global flush doesn't need set IVA_REG */
1018 val = DMA_TLB_GLOBAL_FLUSH|DMA_TLB_IVT;
1019 break;
1020 case DMA_TLB_DSI_FLUSH:
1021 val = DMA_TLB_DSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did);
1022 break;
1023 case DMA_TLB_PSI_FLUSH:
1024 val = DMA_TLB_PSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did);
1025 /* Note: always flush non-leaf currently */
1026 val_iva = size_order | addr;
1027 break;
1028 default:
1029 BUG();
1030 }
1031 /* Note: set drain read/write */
1032#if 0
1033 /*
1034 * This is probably to be super secure.. Looks like we can
1035 * ignore it without any impact.
1036 */
1037 if (cap_read_drain(iommu->cap))
1038 val |= DMA_TLB_READ_DRAIN;
1039#endif
1040 if (cap_write_drain(iommu->cap))
1041 val |= DMA_TLB_WRITE_DRAIN;
1042
1043 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1044 /* Note: Only uses first TLB reg currently */
1045 if (val_iva)
1046 dmar_writeq(iommu->reg + tlb_offset, val_iva);
1047 dmar_writeq(iommu->reg + tlb_offset + 8, val);
1048
1049 /* Make sure hardware complete it */
1050 IOMMU_WAIT_OP(iommu, tlb_offset + 8,
1051 dmar_readq, (!(val & DMA_TLB_IVT)), val);
1052
1053 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1054
1055 /* check IOTLB invalidation granularity */
1056 if (DMA_TLB_IAIG(val) == 0)
1057 printk(KERN_ERR"IOMMU: flush IOTLB failed\n");
1058 if (DMA_TLB_IAIG(val) != DMA_TLB_IIRG(type))
1059 pr_debug("IOMMU: tlb flush request %Lx, actual %Lx\n",
1060 (unsigned long long)DMA_TLB_IIRG(type),
1061 (unsigned long long)DMA_TLB_IAIG(val));
1062}
1063
1064static struct device_domain_info *iommu_support_dev_iotlb(
1065 struct dmar_domain *domain, int segment, u8 bus, u8 devfn)
1066{
1067 int found = 0;
1068 unsigned long flags;
1069 struct device_domain_info *info;
1070 struct intel_iommu *iommu = device_to_iommu(segment, bus, devfn);
1071
1072 if (!ecap_dev_iotlb_support(iommu->ecap))
1073 return NULL;
1074
1075 if (!iommu->qi)
1076 return NULL;
1077
1078 spin_lock_irqsave(&device_domain_lock, flags);
1079 list_for_each_entry(info, &domain->devices, link)
1080 if (info->bus == bus && info->devfn == devfn) {
1081 found = 1;
1082 break;
1083 }
1084 spin_unlock_irqrestore(&device_domain_lock, flags);
1085
1086 if (!found || !info->dev)
1087 return NULL;
1088
1089 if (!pci_find_ext_capability(info->dev, PCI_EXT_CAP_ID_ATS))
1090 return NULL;
1091
1092 if (!dmar_find_matched_atsr_unit(info->dev))
1093 return NULL;
1094
1095 info->iommu = iommu;
1096
1097 return info;
1098}
1099
1100static void iommu_enable_dev_iotlb(struct device_domain_info *info)
1101{
1102 if (!info)
1103 return;
1104
1105 pci_enable_ats(info->dev, VTD_PAGE_SHIFT);
1106}
1107
1108static void iommu_disable_dev_iotlb(struct device_domain_info *info)
1109{
1110 if (!info->dev || !pci_ats_enabled(info->dev))
1111 return;
1112
1113 pci_disable_ats(info->dev);
1114}
1115
1116static void iommu_flush_dev_iotlb(struct dmar_domain *domain,
1117 u64 addr, unsigned mask)
1118{
1119 u16 sid, qdep;
1120 unsigned long flags;
1121 struct device_domain_info *info;
1122
1123 spin_lock_irqsave(&device_domain_lock, flags);
1124 list_for_each_entry(info, &domain->devices, link) {
1125 if (!info->dev || !pci_ats_enabled(info->dev))
1126 continue;
1127
1128 sid = info->bus << 8 | info->devfn;
1129 qdep = pci_ats_queue_depth(info->dev);
1130 qi_flush_dev_iotlb(info->iommu, sid, qdep, addr, mask);
1131 }
1132 spin_unlock_irqrestore(&device_domain_lock, flags);
1133}
1134
1135static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
1136 unsigned long pfn, unsigned int pages, int map)
1137{
1138 unsigned int mask = ilog2(__roundup_pow_of_two(pages));
1139 uint64_t addr = (uint64_t)pfn << VTD_PAGE_SHIFT;
1140
1141 BUG_ON(pages == 0);
1142
1143 /*
1144 * Fallback to domain selective flush if no PSI support or the size is
1145 * too big.
1146 * PSI requires page size to be 2 ^ x, and the base address is naturally
1147 * aligned to the size
1148 */
1149 if (!cap_pgsel_inv(iommu->cap) || mask > cap_max_amask_val(iommu->cap))
1150 iommu->flush.flush_iotlb(iommu, did, 0, 0,
1151 DMA_TLB_DSI_FLUSH);
1152 else
1153 iommu->flush.flush_iotlb(iommu, did, addr, mask,
1154 DMA_TLB_PSI_FLUSH);
1155
1156 /*
1157 * In caching mode, changes of pages from non-present to present require
1158 * flush. However, device IOTLB doesn't need to be flushed in this case.
1159 */
1160 if (!cap_caching_mode(iommu->cap) || !map)
1161 iommu_flush_dev_iotlb(iommu->domains[did], addr, mask);
1162}
1163
1164static void iommu_disable_protect_mem_regions(struct intel_iommu *iommu)
1165{
1166 u32 pmen;
1167 unsigned long flags;
1168
1169 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1170 pmen = readl(iommu->reg + DMAR_PMEN_REG);
1171 pmen &= ~DMA_PMEN_EPM;
1172 writel(pmen, iommu->reg + DMAR_PMEN_REG);
1173
1174 /* wait for the protected region status bit to clear */
1175 IOMMU_WAIT_OP(iommu, DMAR_PMEN_REG,
1176 readl, !(pmen & DMA_PMEN_PRS), pmen);
1177
1178 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1179}
1180
1181static int iommu_enable_translation(struct intel_iommu *iommu)
1182{
1183 u32 sts;
1184 unsigned long flags;
1185
1186 raw_spin_lock_irqsave(&iommu->register_lock, flags);
1187 iommu->gcmd |= DMA_GCMD_TE;
1188 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1189
1190 /* Make sure hardware complete it */
1191 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1192 readl, (sts & DMA_GSTS_TES), sts);
1193
1194 raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1195 return 0;
1196}
1197
1198static int iommu_disable_translation(struct intel_iommu *iommu)
1199{
1200 u32 sts;
1201 unsigned long flag;
1202
1203 raw_spin_lock_irqsave(&iommu->register_lock, flag);
1204 iommu->gcmd &= ~DMA_GCMD_TE;
1205 writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1206
1207 /* Make sure hardware complete it */
1208 IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1209 readl, (!(sts & DMA_GSTS_TES)), sts);
1210
1211 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1212 return 0;
1213}
1214
1215
1216static int iommu_init_domains(struct intel_iommu *iommu)
1217{
1218 unsigned long ndomains;
1219 unsigned long nlongs;
1220
1221 ndomains = cap_ndoms(iommu->cap);
1222 pr_debug("IOMMU %d: Number of Domains supportd <%ld>\n", iommu->seq_id,
1223 ndomains);
1224 nlongs = BITS_TO_LONGS(ndomains);
1225
1226 spin_lock_init(&iommu->lock);
1227
1228 /* TBD: there might be 64K domains,
1229 * consider other allocation for future chip
1230 */
1231 iommu->domain_ids = kcalloc(nlongs, sizeof(unsigned long), GFP_KERNEL);
1232 if (!iommu->domain_ids) {
1233 printk(KERN_ERR "Allocating domain id array failed\n");
1234 return -ENOMEM;
1235 }
1236 iommu->domains = kcalloc(ndomains, sizeof(struct dmar_domain *),
1237 GFP_KERNEL);
1238 if (!iommu->domains) {
1239 printk(KERN_ERR "Allocating domain array failed\n");
1240 return -ENOMEM;
1241 }
1242
1243 /*
1244 * if Caching mode is set, then invalid translations are tagged
1245 * with domainid 0. Hence we need to pre-allocate it.
1246 */
1247 if (cap_caching_mode(iommu->cap))
1248 set_bit(0, iommu->domain_ids);
1249 return 0;
1250}
1251
1252
1253static void domain_exit(struct dmar_domain *domain);
1254static void vm_domain_exit(struct dmar_domain *domain);
1255
1256void free_dmar_iommu(struct intel_iommu *iommu)
1257{
1258 struct dmar_domain *domain;
1259 int i;
1260 unsigned long flags;
1261
1262 if ((iommu->domains) && (iommu->domain_ids)) {
1263 for_each_set_bit(i, iommu->domain_ids, cap_ndoms(iommu->cap)) {
1264 domain = iommu->domains[i];
1265 clear_bit(i, iommu->domain_ids);
1266
1267 spin_lock_irqsave(&domain->iommu_lock, flags);
1268 if (--domain->iommu_count == 0) {
1269 if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE)
1270 vm_domain_exit(domain);
1271 else
1272 domain_exit(domain);
1273 }
1274 spin_unlock_irqrestore(&domain->iommu_lock, flags);
1275 }
1276 }
1277
1278 if (iommu->gcmd & DMA_GCMD_TE)
1279 iommu_disable_translation(iommu);
1280
1281 if (iommu->irq) {
1282 irq_set_handler_data(iommu->irq, NULL);
1283 /* This will mask the irq */
1284 free_irq(iommu->irq, iommu);
1285 destroy_irq(iommu->irq);
1286 }
1287
1288 kfree(iommu->domains);
1289 kfree(iommu->domain_ids);
1290
1291 g_iommus[iommu->seq_id] = NULL;
1292
1293 /* if all iommus are freed, free g_iommus */
1294 for (i = 0; i < g_num_of_iommus; i++) {
1295 if (g_iommus[i])
1296 break;
1297 }
1298
1299 if (i == g_num_of_iommus)
1300 kfree(g_iommus);
1301
1302 /* free context mapping */
1303 free_context_table(iommu);
1304}
1305
1306static struct dmar_domain *alloc_domain(void)
1307{
1308 struct dmar_domain *domain;
1309
1310 domain = alloc_domain_mem();
1311 if (!domain)
1312 return NULL;
1313
1314 domain->nid = -1;
1315 memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
1316 domain->flags = 0;
1317
1318 return domain;
1319}
1320
1321static int iommu_attach_domain(struct dmar_domain *domain,
1322 struct intel_iommu *iommu)
1323{
1324 int num;
1325 unsigned long ndomains;
1326 unsigned long flags;
1327
1328 ndomains = cap_ndoms(iommu->cap);
1329
1330 spin_lock_irqsave(&iommu->lock, flags);
1331
1332 num = find_first_zero_bit(iommu->domain_ids, ndomains);
1333 if (num >= ndomains) {
1334 spin_unlock_irqrestore(&iommu->lock, flags);
1335 printk(KERN_ERR "IOMMU: no free domain ids\n");
1336 return -ENOMEM;
1337 }
1338
1339 domain->id = num;
1340 set_bit(num, iommu->domain_ids);
1341 set_bit(iommu->seq_id, &domain->iommu_bmp);
1342 iommu->domains[num] = domain;
1343 spin_unlock_irqrestore(&iommu->lock, flags);
1344
1345 return 0;
1346}
1347
1348static void iommu_detach_domain(struct dmar_domain *domain,
1349 struct intel_iommu *iommu)
1350{
1351 unsigned long flags;
1352 int num, ndomains;
1353 int found = 0;
1354
1355 spin_lock_irqsave(&iommu->lock, flags);
1356 ndomains = cap_ndoms(iommu->cap);
1357 for_each_set_bit(num, iommu->domain_ids, ndomains) {
1358 if (iommu->domains[num] == domain) {
1359 found = 1;
1360 break;
1361 }
1362 }
1363
1364 if (found) {
1365 clear_bit(num, iommu->domain_ids);
1366 clear_bit(iommu->seq_id, &domain->iommu_bmp);
1367 iommu->domains[num] = NULL;
1368 }
1369 spin_unlock_irqrestore(&iommu->lock, flags);
1370}
1371
1372static struct iova_domain reserved_iova_list;
1373static struct lock_class_key reserved_rbtree_key;
1374
1375static int dmar_init_reserved_ranges(void)
1376{
1377 struct pci_dev *pdev = NULL;
1378 struct iova *iova;
1379 int i;
1380
1381 init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
1382
1383 lockdep_set_class(&reserved_iova_list.iova_rbtree_lock,
1384 &reserved_rbtree_key);
1385
1386 /* IOAPIC ranges shouldn't be accessed by DMA */
1387 iova = reserve_iova(&reserved_iova_list, IOVA_PFN(IOAPIC_RANGE_START),
1388 IOVA_PFN(IOAPIC_RANGE_END));
1389 if (!iova) {
1390 printk(KERN_ERR "Reserve IOAPIC range failed\n");
1391 return -ENODEV;
1392 }
1393
1394 /* Reserve all PCI MMIO to avoid peer-to-peer access */
1395 for_each_pci_dev(pdev) {
1396 struct resource *r;
1397
1398 for (i = 0; i < PCI_NUM_RESOURCES; i++) {
1399 r = &pdev->resource[i];
1400 if (!r->flags || !(r->flags & IORESOURCE_MEM))
1401 continue;
1402 iova = reserve_iova(&reserved_iova_list,
1403 IOVA_PFN(r->start),
1404 IOVA_PFN(r->end));
1405 if (!iova) {
1406 printk(KERN_ERR "Reserve iova failed\n");
1407 return -ENODEV;
1408 }
1409 }
1410 }
1411 return 0;
1412}
1413
1414static void domain_reserve_special_ranges(struct dmar_domain *domain)
1415{
1416 copy_reserved_iova(&reserved_iova_list, &domain->iovad);
1417}
1418
1419static inline int guestwidth_to_adjustwidth(int gaw)
1420{
1421 int agaw;
1422 int r = (gaw - 12) % 9;
1423
1424 if (r == 0)
1425 agaw = gaw;
1426 else
1427 agaw = gaw + 9 - r;
1428 if (agaw > 64)
1429 agaw = 64;
1430 return agaw;
1431}
1432
1433static int domain_init(struct dmar_domain *domain, int guest_width)
1434{
1435 struct intel_iommu *iommu;
1436 int adjust_width, agaw;
1437 unsigned long sagaw;
1438
1439 init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
1440 spin_lock_init(&domain->iommu_lock);
1441
1442 domain_reserve_special_ranges(domain);
1443
1444 /* calculate AGAW */
1445 iommu = domain_get_iommu(domain);
1446 if (guest_width > cap_mgaw(iommu->cap))
1447 guest_width = cap_mgaw(iommu->cap);
1448 domain->gaw = guest_width;
1449 adjust_width = guestwidth_to_adjustwidth(guest_width);
1450 agaw = width_to_agaw(adjust_width);
1451 sagaw = cap_sagaw(iommu->cap);
1452 if (!test_bit(agaw, &sagaw)) {
1453 /* hardware doesn't support it, choose a bigger one */
1454 pr_debug("IOMMU: hardware doesn't support agaw %d\n", agaw);
1455 agaw = find_next_bit(&sagaw, 5, agaw);
1456 if (agaw >= 5)
1457 return -ENODEV;
1458 }
1459 domain->agaw = agaw;
1460 INIT_LIST_HEAD(&domain->devices);
1461
1462 if (ecap_coherent(iommu->ecap))
1463 domain->iommu_coherency = 1;
1464 else
1465 domain->iommu_coherency = 0;
1466
1467 if (ecap_sc_support(iommu->ecap))
1468 domain->iommu_snooping = 1;
1469 else
1470 domain->iommu_snooping = 0;
1471
1472 domain->iommu_superpage = fls(cap_super_page_val(iommu->cap));
1473 domain->iommu_count = 1;
1474 domain->nid = iommu->node;
1475
1476 /* always allocate the top pgd */
1477 domain->pgd = (struct dma_pte *)alloc_pgtable_page(domain->nid);
1478 if (!domain->pgd)
1479 return -ENOMEM;
1480 __iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE);
1481 return 0;
1482}
1483
1484static void domain_exit(struct dmar_domain *domain)
1485{
1486 struct dmar_drhd_unit *drhd;
1487 struct intel_iommu *iommu;
1488
1489 /* Domain 0 is reserved, so dont process it */
1490 if (!domain)
1491 return;
1492
1493 /* Flush any lazy unmaps that may reference this domain */
1494 if (!intel_iommu_strict)
1495 flush_unmaps_timeout(0);
1496
1497 domain_remove_dev_info(domain);
1498 /* destroy iovas */
1499 put_iova_domain(&domain->iovad);
1500
1501 /* clear ptes */
1502 dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
1503
1504 /* free page tables */
1505 dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
1506
1507 for_each_active_iommu(iommu, drhd)
1508 if (test_bit(iommu->seq_id, &domain->iommu_bmp))
1509 iommu_detach_domain(domain, iommu);
1510
1511 free_domain_mem(domain);
1512}
1513
1514static int domain_context_mapping_one(struct dmar_domain *domain, int segment,
1515 u8 bus, u8 devfn, int translation)
1516{
1517 struct context_entry *context;
1518 unsigned long flags;
1519 struct intel_iommu *iommu;
1520 struct dma_pte *pgd;
1521 unsigned long num;
1522 unsigned long ndomains;
1523 int id;
1524 int agaw;
1525 struct device_domain_info *info = NULL;
1526
1527 pr_debug("Set context mapping for %02x:%02x.%d\n",
1528 bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
1529
1530 BUG_ON(!domain->pgd);
1531 BUG_ON(translation != CONTEXT_TT_PASS_THROUGH &&
1532 translation != CONTEXT_TT_MULTI_LEVEL);
1533
1534 iommu = device_to_iommu(segment, bus, devfn);
1535 if (!iommu)
1536 return -ENODEV;
1537
1538 context = device_to_context_entry(iommu, bus, devfn);
1539 if (!context)
1540 return -ENOMEM;
1541 spin_lock_irqsave(&iommu->lock, flags);
1542 if (context_present(context)) {
1543 spin_unlock_irqrestore(&iommu->lock, flags);
1544 return 0;
1545 }
1546
1547 id = domain->id;
1548 pgd = domain->pgd;
1549
1550 if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
1551 domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) {
1552 int found = 0;
1553
1554 /* find an available domain id for this device in iommu */
1555 ndomains = cap_ndoms(iommu->cap);
1556 for_each_set_bit(num, iommu->domain_ids, ndomains) {
1557 if (iommu->domains[num] == domain) {
1558 id = num;
1559 found = 1;
1560 break;
1561 }
1562 }
1563
1564 if (found == 0) {
1565 num = find_first_zero_bit(iommu->domain_ids, ndomains);
1566 if (num >= ndomains) {
1567 spin_unlock_irqrestore(&iommu->lock, flags);
1568 printk(KERN_ERR "IOMMU: no free domain ids\n");
1569 return -EFAULT;
1570 }
1571
1572 set_bit(num, iommu->domain_ids);
1573 iommu->domains[num] = domain;
1574 id = num;
1575 }
1576
1577 /* Skip top levels of page tables for
1578 * iommu which has less agaw than default.
1579 * Unnecessary for PT mode.
1580 */
1581 if (translation != CONTEXT_TT_PASS_THROUGH) {
1582 for (agaw = domain->agaw; agaw != iommu->agaw; agaw--) {
1583 pgd = phys_to_virt(dma_pte_addr(pgd));
1584 if (!dma_pte_present(pgd)) {
1585 spin_unlock_irqrestore(&iommu->lock, flags);
1586 return -ENOMEM;
1587 }
1588 }
1589 }
1590 }
1591
1592 context_set_domain_id(context, id);
1593
1594 if (translation != CONTEXT_TT_PASS_THROUGH) {
1595 info = iommu_support_dev_iotlb(domain, segment, bus, devfn);
1596 translation = info ? CONTEXT_TT_DEV_IOTLB :
1597 CONTEXT_TT_MULTI_LEVEL;
1598 }
1599 /*
1600 * In pass through mode, AW must be programmed to indicate the largest
1601 * AGAW value supported by hardware. And ASR is ignored by hardware.
1602 */
1603 if (unlikely(translation == CONTEXT_TT_PASS_THROUGH))
1604 context_set_address_width(context, iommu->msagaw);
1605 else {
1606 context_set_address_root(context, virt_to_phys(pgd));
1607 context_set_address_width(context, iommu->agaw);
1608 }
1609
1610 context_set_translation_type(context, translation);
1611 context_set_fault_enable(context);
1612 context_set_present(context);
1613 domain_flush_cache(domain, context, sizeof(*context));
1614
1615 /*
1616 * It's a non-present to present mapping. If hardware doesn't cache
1617 * non-present entry we only need to flush the write-buffer. If the
1618 * _does_ cache non-present entries, then it does so in the special
1619 * domain #0, which we have to flush:
1620 */
1621 if (cap_caching_mode(iommu->cap)) {
1622 iommu->flush.flush_context(iommu, 0,
1623 (((u16)bus) << 8) | devfn,
1624 DMA_CCMD_MASK_NOBIT,
1625 DMA_CCMD_DEVICE_INVL);
1626 iommu->flush.flush_iotlb(iommu, domain->id, 0, 0, DMA_TLB_DSI_FLUSH);
1627 } else {
1628 iommu_flush_write_buffer(iommu);
1629 }
1630 iommu_enable_dev_iotlb(info);
1631 spin_unlock_irqrestore(&iommu->lock, flags);
1632
1633 spin_lock_irqsave(&domain->iommu_lock, flags);
1634 if (!test_and_set_bit(iommu->seq_id, &domain->iommu_bmp)) {
1635 domain->iommu_count++;
1636 if (domain->iommu_count == 1)
1637 domain->nid = iommu->node;
1638 domain_update_iommu_cap(domain);
1639 }
1640 spin_unlock_irqrestore(&domain->iommu_lock, flags);
1641 return 0;
1642}
1643
1644static int
1645domain_context_mapping(struct dmar_domain *domain, struct pci_dev *pdev,
1646 int translation)
1647{
1648 int ret;
1649 struct pci_dev *tmp, *parent;
1650
1651 ret = domain_context_mapping_one(domain, pci_domain_nr(pdev->bus),
1652 pdev->bus->number, pdev->devfn,
1653 translation);
1654 if (ret)
1655 return ret;
1656
1657 /* dependent device mapping */
1658 tmp = pci_find_upstream_pcie_bridge(pdev);
1659 if (!tmp)
1660 return 0;
1661 /* Secondary interface's bus number and devfn 0 */
1662 parent = pdev->bus->self;
1663 while (parent != tmp) {
1664 ret = domain_context_mapping_one(domain,
1665 pci_domain_nr(parent->bus),
1666 parent->bus->number,
1667 parent->devfn, translation);
1668 if (ret)
1669 return ret;
1670 parent = parent->bus->self;
1671 }
1672 if (pci_is_pcie(tmp)) /* this is a PCIe-to-PCI bridge */
1673 return domain_context_mapping_one(domain,
1674 pci_domain_nr(tmp->subordinate),
1675 tmp->subordinate->number, 0,
1676 translation);
1677 else /* this is a legacy PCI bridge */
1678 return domain_context_mapping_one(domain,
1679 pci_domain_nr(tmp->bus),
1680 tmp->bus->number,
1681 tmp->devfn,
1682 translation);
1683}
1684
1685static int domain_context_mapped(struct pci_dev *pdev)
1686{
1687 int ret;
1688 struct pci_dev *tmp, *parent;
1689 struct intel_iommu *iommu;
1690
1691 iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number,
1692 pdev->devfn);
1693 if (!iommu)
1694 return -ENODEV;
1695
1696 ret = device_context_mapped(iommu, pdev->bus->number, pdev->devfn);
1697 if (!ret)
1698 return ret;
1699 /* dependent device mapping */
1700 tmp = pci_find_upstream_pcie_bridge(pdev);
1701 if (!tmp)
1702 return ret;
1703 /* Secondary interface's bus number and devfn 0 */
1704 parent = pdev->bus->self;
1705 while (parent != tmp) {
1706 ret = device_context_mapped(iommu, parent->bus->number,
1707 parent->devfn);
1708 if (!ret)
1709 return ret;
1710 parent = parent->bus->self;
1711 }
1712 if (pci_is_pcie(tmp))
1713 return device_context_mapped(iommu, tmp->subordinate->number,
1714 0);
1715 else
1716 return device_context_mapped(iommu, tmp->bus->number,
1717 tmp->devfn);
1718}
1719
1720/* Returns a number of VTD pages, but aligned to MM page size */
1721static inline unsigned long aligned_nrpages(unsigned long host_addr,
1722 size_t size)
1723{
1724 host_addr &= ~PAGE_MASK;
1725 return PAGE_ALIGN(host_addr + size) >> VTD_PAGE_SHIFT;
1726}
1727
1728/* Return largest possible superpage level for a given mapping */
1729static inline int hardware_largepage_caps(struct dmar_domain *domain,
1730 unsigned long iov_pfn,
1731 unsigned long phy_pfn,
1732 unsigned long pages)
1733{
1734 int support, level = 1;
1735 unsigned long pfnmerge;
1736
1737 support = domain->iommu_superpage;
1738
1739 /* To use a large page, the virtual *and* physical addresses
1740 must be aligned to 2MiB/1GiB/etc. Lower bits set in either
1741 of them will mean we have to use smaller pages. So just
1742 merge them and check both at once. */
1743 pfnmerge = iov_pfn | phy_pfn;
1744
1745 while (support && !(pfnmerge & ~VTD_STRIDE_MASK)) {
1746 pages >>= VTD_STRIDE_SHIFT;
1747 if (!pages)
1748 break;
1749 pfnmerge >>= VTD_STRIDE_SHIFT;
1750 level++;
1751 support--;
1752 }
1753 return level;
1754}
1755
1756static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
1757 struct scatterlist *sg, unsigned long phys_pfn,
1758 unsigned long nr_pages, int prot)
1759{
1760 struct dma_pte *first_pte = NULL, *pte = NULL;
1761 phys_addr_t uninitialized_var(pteval);
1762 int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
1763 unsigned long sg_res;
1764 unsigned int largepage_lvl = 0;
1765 unsigned long lvl_pages = 0;
1766
1767 BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);
1768
1769 if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
1770 return -EINVAL;
1771
1772 prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;
1773
1774 if (sg)
1775 sg_res = 0;
1776 else {
1777 sg_res = nr_pages + 1;
1778 pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
1779 }
1780
1781 while (nr_pages > 0) {
1782 uint64_t tmp;
1783
1784 if (!sg_res) {
1785 sg_res = aligned_nrpages(sg->offset, sg->length);
1786 sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
1787 sg->dma_length = sg->length;
1788 pteval = page_to_phys(sg_page(sg)) | prot;
1789 phys_pfn = pteval >> VTD_PAGE_SHIFT;
1790 }
1791
1792 if (!pte) {
1793 largepage_lvl = hardware_largepage_caps(domain, iov_pfn, phys_pfn, sg_res);
1794
1795 first_pte = pte = pfn_to_dma_pte(domain, iov_pfn, largepage_lvl);
1796 if (!pte)
1797 return -ENOMEM;
1798 /* It is large page*/
1799 if (largepage_lvl > 1)
1800 pteval |= DMA_PTE_LARGE_PAGE;
1801 else
1802 pteval &= ~(uint64_t)DMA_PTE_LARGE_PAGE;
1803
1804 }
1805 /* We don't need lock here, nobody else
1806 * touches the iova range
1807 */
1808 tmp = cmpxchg64_local(&pte->val, 0ULL, pteval);
1809 if (tmp) {
1810 static int dumps = 5;
1811 printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
1812 iov_pfn, tmp, (unsigned long long)pteval);
1813 if (dumps) {
1814 dumps--;
1815 debug_dma_dump_mappings(NULL);
1816 }
1817 WARN_ON(1);
1818 }
1819
1820 lvl_pages = lvl_to_nr_pages(largepage_lvl);
1821
1822 BUG_ON(nr_pages < lvl_pages);
1823 BUG_ON(sg_res < lvl_pages);
1824
1825 nr_pages -= lvl_pages;
1826 iov_pfn += lvl_pages;
1827 phys_pfn += lvl_pages;
1828 pteval += lvl_pages * VTD_PAGE_SIZE;
1829 sg_res -= lvl_pages;
1830
1831 /* If the next PTE would be the first in a new page, then we
1832 need to flush the cache on the entries we've just written.
1833 And then we'll need to recalculate 'pte', so clear it and
1834 let it get set again in the if (!pte) block above.
1835
1836 If we're done (!nr_pages) we need to flush the cache too.
1837
1838 Also if we've been setting superpages, we may need to
1839 recalculate 'pte' and switch back to smaller pages for the
1840 end of the mapping, if the trailing size is not enough to
1841 use another superpage (i.e. sg_res < lvl_pages). */
1842 pte++;
1843 if (!nr_pages || first_pte_in_page(pte) ||
1844 (largepage_lvl > 1 && sg_res < lvl_pages)) {
1845 domain_flush_cache(domain, first_pte,
1846 (void *)pte - (void *)first_pte);
1847 pte = NULL;
1848 }
1849
1850 if (!sg_res && nr_pages)
1851 sg = sg_next(sg);
1852 }
1853 return 0;
1854}
1855
1856static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
1857 struct scatterlist *sg, unsigned long nr_pages,
1858 int prot)
1859{
1860 return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
1861}
1862
1863static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
1864 unsigned long phys_pfn, unsigned long nr_pages,
1865 int prot)
1866{
1867 return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
1868}
1869
1870static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
1871{
1872 if (!iommu)
1873 return;
1874
1875 clear_context_table(iommu, bus, devfn);
1876 iommu->flush.flush_context(iommu, 0, 0, 0,
1877 DMA_CCMD_GLOBAL_INVL);
1878 iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
1879}
1880
1881static void domain_remove_dev_info(struct dmar_domain *domain)
1882{
1883 struct device_domain_info *info;
1884 unsigned long flags;
1885 struct intel_iommu *iommu;
1886
1887 spin_lock_irqsave(&device_domain_lock, flags);
1888 while (!list_empty(&domain->devices)) {
1889 info = list_entry(domain->devices.next,
1890 struct device_domain_info, link);
1891 list_del(&info->link);
1892 list_del(&info->global);
1893 if (info->dev)
1894 info->dev->dev.archdata.iommu = NULL;
1895 spin_unlock_irqrestore(&device_domain_lock, flags);
1896
1897 iommu_disable_dev_iotlb(info);
1898 iommu = device_to_iommu(info->segment, info->bus, info->devfn);
1899 iommu_detach_dev(iommu, info->bus, info->devfn);
1900 free_devinfo_mem(info);
1901
1902 spin_lock_irqsave(&device_domain_lock, flags);
1903 }
1904 spin_unlock_irqrestore(&device_domain_lock, flags);
1905}
1906
1907/*
1908 * find_domain
1909 * Note: we use struct pci_dev->dev.archdata.iommu stores the info
1910 */
1911static struct dmar_domain *
1912find_domain(struct pci_dev *pdev)
1913{
1914 struct device_domain_info *info;
1915
1916 /* No lock here, assumes no domain exit in normal case */
1917 info = pdev->dev.archdata.iommu;
1918 if (info)
1919 return info->domain;
1920 return NULL;
1921}
1922
1923/* domain is initialized */
1924static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw)
1925{
1926 struct dmar_domain *domain, *found = NULL;
1927 struct intel_iommu *iommu;
1928 struct dmar_drhd_unit *drhd;
1929 struct device_domain_info *info, *tmp;
1930 struct pci_dev *dev_tmp;
1931 unsigned long flags;
1932 int bus = 0, devfn = 0;
1933 int segment;
1934 int ret;
1935
1936 domain = find_domain(pdev);
1937 if (domain)
1938 return domain;
1939
1940 segment = pci_domain_nr(pdev->bus);
1941
1942 dev_tmp = pci_find_upstream_pcie_bridge(pdev);
1943 if (dev_tmp) {
1944 if (pci_is_pcie(dev_tmp)) {
1945 bus = dev_tmp->subordinate->number;
1946 devfn = 0;
1947 } else {
1948 bus = dev_tmp->bus->number;
1949 devfn = dev_tmp->devfn;
1950 }
1951 spin_lock_irqsave(&device_domain_lock, flags);
1952 list_for_each_entry(info, &device_domain_list, global) {
1953 if (info->segment == segment &&
1954 info->bus == bus && info->devfn == devfn) {
1955 found = info->domain;
1956 break;
1957 }
1958 }
1959 spin_unlock_irqrestore(&device_domain_lock, flags);
1960 /* pcie-pci bridge already has a domain, uses it */
1961 if (found) {
1962 domain = found;
1963 goto found_domain;
1964 }
1965 }
1966
1967 domain = alloc_domain();
1968 if (!domain)
1969 goto error;
1970
1971 /* Allocate new domain for the device */
1972 drhd = dmar_find_matched_drhd_unit(pdev);
1973 if (!drhd) {
1974 printk(KERN_ERR "IOMMU: can't find DMAR for device %s\n",
1975 pci_name(pdev));
1976 return NULL;
1977 }
1978 iommu = drhd->iommu;
1979
1980 ret = iommu_attach_domain(domain, iommu);
1981 if (ret) {
1982 free_domain_mem(domain);
1983 goto error;
1984 }
1985
1986 if (domain_init(domain, gaw)) {
1987 domain_exit(domain);
1988 goto error;
1989 }
1990
1991 /* register pcie-to-pci device */
1992 if (dev_tmp) {
1993 info = alloc_devinfo_mem();
1994 if (!info) {
1995 domain_exit(domain);
1996 goto error;
1997 }
1998 info->segment = segment;
1999 info->bus = bus;
2000 info->devfn = devfn;
2001 info->dev = NULL;
2002 info->domain = domain;
2003 /* This domain is shared by devices under p2p bridge */
2004 domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES;
2005
2006 /* pcie-to-pci bridge already has a domain, uses it */
2007 found = NULL;
2008 spin_lock_irqsave(&device_domain_lock, flags);
2009 list_for_each_entry(tmp, &device_domain_list, global) {
2010 if (tmp->segment == segment &&
2011 tmp->bus == bus && tmp->devfn == devfn) {
2012 found = tmp->domain;
2013 break;
2014 }
2015 }
2016 if (found) {
2017 spin_unlock_irqrestore(&device_domain_lock, flags);
2018 free_devinfo_mem(info);
2019 domain_exit(domain);
2020 domain = found;
2021 } else {
2022 list_add(&info->link, &domain->devices);
2023 list_add(&info->global, &device_domain_list);
2024 spin_unlock_irqrestore(&device_domain_lock, flags);
2025 }
2026 }
2027
2028found_domain:
2029 info = alloc_devinfo_mem();
2030 if (!info)
2031 goto error;
2032 info->segment = segment;
2033 info->bus = pdev->bus->number;
2034 info->devfn = pdev->devfn;
2035 info->dev = pdev;
2036 info->domain = domain;
2037 spin_lock_irqsave(&device_domain_lock, flags);
2038 /* somebody is fast */
2039 found = find_domain(pdev);
2040 if (found != NULL) {
2041 spin_unlock_irqrestore(&device_domain_lock, flags);
2042 if (found != domain) {
2043 domain_exit(domain);
2044 domain = found;
2045 }
2046 free_devinfo_mem(info);
2047 return domain;
2048 }
2049 list_add(&info->link, &domain->devices);
2050 list_add(&info->global, &device_domain_list);
2051 pdev->dev.archdata.iommu = info;
2052 spin_unlock_irqrestore(&device_domain_lock, flags);
2053 return domain;
2054error:
2055 /* recheck it here, maybe others set it */
2056 return find_domain(pdev);
2057}
2058
2059static int iommu_identity_mapping;
2060#define IDENTMAP_ALL 1
2061#define IDENTMAP_GFX 2
2062#define IDENTMAP_AZALIA 4
2063
2064static int iommu_domain_identity_map(struct dmar_domain *domain,
2065 unsigned long long start,
2066 unsigned long long end)
2067{
2068 unsigned long first_vpfn = start >> VTD_PAGE_SHIFT;
2069 unsigned long last_vpfn = end >> VTD_PAGE_SHIFT;
2070
2071 if (!reserve_iova(&domain->iovad, dma_to_mm_pfn(first_vpfn),
2072 dma_to_mm_pfn(last_vpfn))) {
2073 printk(KERN_ERR "IOMMU: reserve iova failed\n");
2074 return -ENOMEM;
2075 }
2076
2077 pr_debug("Mapping reserved region %llx-%llx for domain %d\n",
2078 start, end, domain->id);
2079 /*
2080 * RMRR range might have overlap with physical memory range,
2081 * clear it first
2082 */
2083 dma_pte_clear_range(domain, first_vpfn, last_vpfn);
2084
2085 return domain_pfn_mapping(domain, first_vpfn, first_vpfn,
2086 last_vpfn - first_vpfn + 1,
2087 DMA_PTE_READ|DMA_PTE_WRITE);
2088}
2089
2090static int iommu_prepare_identity_map(struct pci_dev *pdev,
2091 unsigned long long start,
2092 unsigned long long end)
2093{
2094 struct dmar_domain *domain;
2095 int ret;
2096
2097 domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
2098 if (!domain)
2099 return -ENOMEM;
2100
2101 /* For _hardware_ passthrough, don't bother. But for software
2102 passthrough, we do it anyway -- it may indicate a memory
2103 range which is reserved in E820, so which didn't get set
2104 up to start with in si_domain */
2105 if (domain == si_domain && hw_pass_through) {
2106 printk("Ignoring identity map for HW passthrough device %s [0x%Lx - 0x%Lx]\n",
2107 pci_name(pdev), start, end);
2108 return 0;
2109 }
2110
2111 printk(KERN_INFO
2112 "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
2113 pci_name(pdev), start, end);
2114
2115 if (end < start) {
2116 WARN(1, "Your BIOS is broken; RMRR ends before it starts!\n"
2117 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
2118 dmi_get_system_info(DMI_BIOS_VENDOR),
2119 dmi_get_system_info(DMI_BIOS_VERSION),
2120 dmi_get_system_info(DMI_PRODUCT_VERSION));
2121 ret = -EIO;
2122 goto error;
2123 }
2124
2125 if (end >> agaw_to_width(domain->agaw)) {
2126 WARN(1, "Your BIOS is broken; RMRR exceeds permitted address width (%d bits)\n"
2127 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
2128 agaw_to_width(domain->agaw),
2129 dmi_get_system_info(DMI_BIOS_VENDOR),
2130 dmi_get_system_info(DMI_BIOS_VERSION),
2131 dmi_get_system_info(DMI_PRODUCT_VERSION));
2132 ret = -EIO;
2133 goto error;
2134 }
2135
2136 ret = iommu_domain_identity_map(domain, start, end);
2137 if (ret)
2138 goto error;
2139
2140 /* context entry init */
2141 ret = domain_context_mapping(domain, pdev, CONTEXT_TT_MULTI_LEVEL);
2142 if (ret)
2143 goto error;
2144
2145 return 0;
2146
2147 error:
2148 domain_exit(domain);
2149 return ret;
2150}
2151
2152static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
2153 struct pci_dev *pdev)
2154{
2155 if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
2156 return 0;
2157 return iommu_prepare_identity_map(pdev, rmrr->base_address,
2158 rmrr->end_address);
2159}
2160
2161#ifdef CONFIG_INTEL_IOMMU_FLOPPY_WA
2162static inline void iommu_prepare_isa(void)
2163{
2164 struct pci_dev *pdev;
2165 int ret;
2166
2167 pdev = pci_get_class(PCI_CLASS_BRIDGE_ISA << 8, NULL);
2168 if (!pdev)
2169 return;
2170
2171 printk(KERN_INFO "IOMMU: Prepare 0-16MiB unity mapping for LPC\n");
2172 ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024 - 1);
2173
2174 if (ret)
2175 printk(KERN_ERR "IOMMU: Failed to create 0-16MiB identity map; "
2176 "floppy might not work\n");
2177
2178}
2179#else
2180static inline void iommu_prepare_isa(void)
2181{
2182 return;
2183}
2184#endif /* !CONFIG_INTEL_IOMMU_FLPY_WA */
2185
2186static int md_domain_init(struct dmar_domain *domain, int guest_width);
2187
2188static int __init si_domain_init(int hw)
2189{
2190 struct dmar_drhd_unit *drhd;
2191 struct intel_iommu *iommu;
2192 int nid, ret = 0;
2193
2194 si_domain = alloc_domain();
2195 if (!si_domain)
2196 return -EFAULT;
2197
2198 pr_debug("Identity mapping domain is domain %d\n", si_domain->id);
2199
2200 for_each_active_iommu(iommu, drhd) {
2201 ret = iommu_attach_domain(si_domain, iommu);
2202 if (ret) {
2203 domain_exit(si_domain);
2204 return -EFAULT;
2205 }
2206 }
2207
2208 if (md_domain_init(si_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
2209 domain_exit(si_domain);
2210 return -EFAULT;
2211 }
2212
2213 si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;
2214
2215 if (hw)
2216 return 0;
2217
2218 for_each_online_node(nid) {
2219 unsigned long start_pfn, end_pfn;
2220 int i;
2221
2222 for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
2223 ret = iommu_domain_identity_map(si_domain,
2224 PFN_PHYS(start_pfn), PFN_PHYS(end_pfn));
2225 if (ret)
2226 return ret;
2227 }
2228 }
2229
2230 return 0;
2231}
2232
2233static void domain_remove_one_dev_info(struct dmar_domain *domain,
2234 struct pci_dev *pdev);
2235static int identity_mapping(struct pci_dev *pdev)
2236{
2237 struct device_domain_info *info;
2238
2239 if (likely(!iommu_identity_mapping))
2240 return 0;
2241
2242 info = pdev->dev.archdata.iommu;
2243 if (info && info != DUMMY_DEVICE_DOMAIN_INFO)
2244 return (info->domain == si_domain);
2245
2246 return 0;
2247}
2248
2249static int domain_add_dev_info(struct dmar_domain *domain,
2250 struct pci_dev *pdev,
2251 int translation)
2252{
2253 struct device_domain_info *info;
2254 unsigned long flags;
2255 int ret;
2256
2257 info = alloc_devinfo_mem();
2258 if (!info)
2259 return -ENOMEM;
2260
2261 ret = domain_context_mapping(domain, pdev, translation);
2262 if (ret) {
2263 free_devinfo_mem(info);
2264 return ret;
2265 }
2266
2267 info->segment = pci_domain_nr(pdev->bus);
2268 info->bus = pdev->bus->number;
2269 info->devfn = pdev->devfn;
2270 info->dev = pdev;
2271 info->domain = domain;
2272
2273 spin_lock_irqsave(&device_domain_lock, flags);
2274 list_add(&info->link, &domain->devices);
2275 list_add(&info->global, &device_domain_list);
2276 pdev->dev.archdata.iommu = info;
2277 spin_unlock_irqrestore(&device_domain_lock, flags);
2278
2279 return 0;
2280}
2281
2282static int iommu_should_identity_map(struct pci_dev *pdev, int startup)
2283{
2284 if ((iommu_identity_mapping & IDENTMAP_AZALIA) && IS_AZALIA(pdev))
2285 return 1;
2286
2287 if ((iommu_identity_mapping & IDENTMAP_GFX) && IS_GFX_DEVICE(pdev))
2288 return 1;
2289
2290 if (!(iommu_identity_mapping & IDENTMAP_ALL))
2291 return 0;
2292
2293 /*
2294 * We want to start off with all devices in the 1:1 domain, and
2295 * take them out later if we find they can't access all of memory.
2296 *
2297 * However, we can't do this for PCI devices behind bridges,
2298 * because all PCI devices behind the same bridge will end up
2299 * with the same source-id on their transactions.
2300 *
2301 * Practically speaking, we can't change things around for these
2302 * devices at run-time, because we can't be sure there'll be no
2303 * DMA transactions in flight for any of their siblings.
2304 *
2305 * So PCI devices (unless they're on the root bus) as well as
2306 * their parent PCI-PCI or PCIe-PCI bridges must be left _out_ of
2307 * the 1:1 domain, just in _case_ one of their siblings turns out
2308 * not to be able to map all of memory.
2309 */
2310 if (!pci_is_pcie(pdev)) {
2311 if (!pci_is_root_bus(pdev->bus))
2312 return 0;
2313 if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI)
2314 return 0;
2315 } else if (pdev->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
2316 return 0;
2317
2318 /*
2319 * At boot time, we don't yet know if devices will be 64-bit capable.
2320 * Assume that they will -- if they turn out not to be, then we can
2321 * take them out of the 1:1 domain later.
2322 */
2323 if (!startup) {
2324 /*
2325 * If the device's dma_mask is less than the system's memory
2326 * size then this is not a candidate for identity mapping.
2327 */
2328 u64 dma_mask = pdev->dma_mask;
2329
2330 if (pdev->dev.coherent_dma_mask &&
2331 pdev->dev.coherent_dma_mask < dma_mask)
2332 dma_mask = pdev->dev.coherent_dma_mask;
2333
2334 return dma_mask >= dma_get_required_mask(&pdev->dev);
2335 }
2336
2337 return 1;
2338}
2339
2340static int __init iommu_prepare_static_identity_mapping(int hw)
2341{
2342 struct pci_dev *pdev = NULL;
2343 int ret;
2344
2345 ret = si_domain_init(hw);
2346 if (ret)
2347 return -EFAULT;
2348
2349 for_each_pci_dev(pdev) {
2350 /* Skip Host/PCI Bridge devices */
2351 if (IS_BRIDGE_HOST_DEVICE(pdev))
2352 continue;
2353 if (iommu_should_identity_map(pdev, 1)) {
2354 printk(KERN_INFO "IOMMU: %s identity mapping for device %s\n",
2355 hw ? "hardware" : "software", pci_name(pdev));
2356
2357 ret = domain_add_dev_info(si_domain, pdev,
2358 hw ? CONTEXT_TT_PASS_THROUGH :
2359 CONTEXT_TT_MULTI_LEVEL);
2360 if (ret)
2361 return ret;
2362 }
2363 }
2364
2365 return 0;
2366}
2367
2368static int __init init_dmars(void)
2369{
2370 struct dmar_drhd_unit *drhd;
2371 struct dmar_rmrr_unit *rmrr;
2372 struct pci_dev *pdev;
2373 struct intel_iommu *iommu;
2374 int i, ret;
2375
2376 /*
2377 * for each drhd
2378 * allocate root
2379 * initialize and program root entry to not present
2380 * endfor
2381 */
2382 for_each_drhd_unit(drhd) {
2383 g_num_of_iommus++;
2384 /*
2385 * lock not needed as this is only incremented in the single
2386 * threaded kernel __init code path all other access are read
2387 * only
2388 */
2389 }
2390
2391 g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *),
2392 GFP_KERNEL);
2393 if (!g_iommus) {
2394 printk(KERN_ERR "Allocating global iommu array failed\n");
2395 ret = -ENOMEM;
2396 goto error;
2397 }
2398
2399 deferred_flush = kzalloc(g_num_of_iommus *
2400 sizeof(struct deferred_flush_tables), GFP_KERNEL);
2401 if (!deferred_flush) {
2402 ret = -ENOMEM;
2403 goto error;
2404 }
2405
2406 for_each_drhd_unit(drhd) {
2407 if (drhd->ignored)
2408 continue;
2409
2410 iommu = drhd->iommu;
2411 g_iommus[iommu->seq_id] = iommu;
2412
2413 ret = iommu_init_domains(iommu);
2414 if (ret)
2415 goto error;
2416
2417 /*
2418 * TBD:
2419 * we could share the same root & context tables
2420 * among all IOMMU's. Need to Split it later.
2421 */
2422 ret = iommu_alloc_root_entry(iommu);
2423 if (ret) {
2424 printk(KERN_ERR "IOMMU: allocate root entry failed\n");
2425 goto error;
2426 }
2427 if (!ecap_pass_through(iommu->ecap))
2428 hw_pass_through = 0;
2429 }
2430
2431 /*
2432 * Start from the sane iommu hardware state.
2433 */
2434 for_each_drhd_unit(drhd) {
2435 if (drhd->ignored)
2436 continue;
2437
2438 iommu = drhd->iommu;
2439
2440 /*
2441 * If the queued invalidation is already initialized by us
2442 * (for example, while enabling interrupt-remapping) then
2443 * we got the things already rolling from a sane state.
2444 */
2445 if (iommu->qi)
2446 continue;
2447
2448 /*
2449 * Clear any previous faults.
2450 */
2451 dmar_fault(-1, iommu);
2452 /*
2453 * Disable queued invalidation if supported and already enabled
2454 * before OS handover.
2455 */
2456 dmar_disable_qi(iommu);
2457 }
2458
2459 for_each_drhd_unit(drhd) {
2460 if (drhd->ignored)
2461 continue;
2462
2463 iommu = drhd->iommu;
2464
2465 if (dmar_enable_qi(iommu)) {
2466 /*
2467 * Queued Invalidate not enabled, use Register Based
2468 * Invalidate
2469 */
2470 iommu->flush.flush_context = __iommu_flush_context;
2471 iommu->flush.flush_iotlb = __iommu_flush_iotlb;
2472 printk(KERN_INFO "IOMMU %d 0x%Lx: using Register based "
2473 "invalidation\n",
2474 iommu->seq_id,
2475 (unsigned long long)drhd->reg_base_addr);
2476 } else {
2477 iommu->flush.flush_context = qi_flush_context;
2478 iommu->flush.flush_iotlb = qi_flush_iotlb;
2479 printk(KERN_INFO "IOMMU %d 0x%Lx: using Queued "
2480 "invalidation\n",
2481 iommu->seq_id,
2482 (unsigned long long)drhd->reg_base_addr);
2483 }
2484 }
2485
2486 if (iommu_pass_through)
2487 iommu_identity_mapping |= IDENTMAP_ALL;
2488
2489#ifdef CONFIG_INTEL_IOMMU_BROKEN_GFX_WA
2490 iommu_identity_mapping |= IDENTMAP_GFX;
2491#endif
2492
2493 check_tylersburg_isoch();
2494
2495 /*
2496 * If pass through is not set or not enabled, setup context entries for
2497 * identity mappings for rmrr, gfx, and isa and may fall back to static
2498 * identity mapping if iommu_identity_mapping is set.
2499 */
2500 if (iommu_identity_mapping) {
2501 ret = iommu_prepare_static_identity_mapping(hw_pass_through);
2502 if (ret) {
2503 printk(KERN_CRIT "Failed to setup IOMMU pass-through\n");
2504 goto error;
2505 }
2506 }
2507 /*
2508 * For each rmrr
2509 * for each dev attached to rmrr
2510 * do
2511 * locate drhd for dev, alloc domain for dev
2512 * allocate free domain
2513 * allocate page table entries for rmrr
2514 * if context not allocated for bus
2515 * allocate and init context
2516 * set present in root table for this bus
2517 * init context with domain, translation etc
2518 * endfor
2519 * endfor
2520 */
2521 printk(KERN_INFO "IOMMU: Setting RMRR:\n");
2522 for_each_rmrr_units(rmrr) {
2523 for (i = 0; i < rmrr->devices_cnt; i++) {
2524 pdev = rmrr->devices[i];
2525 /*
2526 * some BIOS lists non-exist devices in DMAR
2527 * table.
2528 */
2529 if (!pdev)
2530 continue;
2531 ret = iommu_prepare_rmrr_dev(rmrr, pdev);
2532 if (ret)
2533 printk(KERN_ERR
2534 "IOMMU: mapping reserved region failed\n");
2535 }
2536 }
2537
2538 iommu_prepare_isa();
2539
2540 /*
2541 * for each drhd
2542 * enable fault log
2543 * global invalidate context cache
2544 * global invalidate iotlb
2545 * enable translation
2546 */
2547 for_each_drhd_unit(drhd) {
2548 if (drhd->ignored) {
2549 /*
2550 * we always have to disable PMRs or DMA may fail on
2551 * this device
2552 */
2553 if (force_on)
2554 iommu_disable_protect_mem_regions(drhd->iommu);
2555 continue;
2556 }
2557 iommu = drhd->iommu;
2558
2559 iommu_flush_write_buffer(iommu);
2560
2561 ret = dmar_set_interrupt(iommu);
2562 if (ret)
2563 goto error;
2564
2565 iommu_set_root_entry(iommu);
2566
2567 iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL);
2568 iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
2569
2570 ret = iommu_enable_translation(iommu);
2571 if (ret)
2572 goto error;
2573
2574 iommu_disable_protect_mem_regions(iommu);
2575 }
2576
2577 return 0;
2578error:
2579 for_each_drhd_unit(drhd) {
2580 if (drhd->ignored)
2581 continue;
2582 iommu = drhd->iommu;
2583 free_iommu(iommu);
2584 }
2585 kfree(g_iommus);
2586 return ret;
2587}
2588
2589/* This takes a number of _MM_ pages, not VTD pages */
2590static struct iova *intel_alloc_iova(struct device *dev,
2591 struct dmar_domain *domain,
2592 unsigned long nrpages, uint64_t dma_mask)
2593{
2594 struct pci_dev *pdev = to_pci_dev(dev);
2595 struct iova *iova = NULL;
2596
2597 /* Restrict dma_mask to the width that the iommu can handle */
2598 dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);
2599
2600 if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
2601 /*
2602 * First try to allocate an io virtual address in
2603 * DMA_BIT_MASK(32) and if that fails then try allocating
2604 * from higher range
2605 */
2606 iova = alloc_iova(&domain->iovad, nrpages,
2607 IOVA_PFN(DMA_BIT_MASK(32)), 1);
2608 if (iova)
2609 return iova;
2610 }
2611 iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
2612 if (unlikely(!iova)) {
2613 printk(KERN_ERR "Allocating %ld-page iova for %s failed",
2614 nrpages, pci_name(pdev));
2615 return NULL;
2616 }
2617
2618 return iova;
2619}
2620
2621static struct dmar_domain *__get_valid_domain_for_dev(struct pci_dev *pdev)
2622{
2623 struct dmar_domain *domain;
2624 int ret;
2625
2626 domain = get_domain_for_dev(pdev,
2627 DEFAULT_DOMAIN_ADDRESS_WIDTH);
2628 if (!domain) {
2629 printk(KERN_ERR
2630 "Allocating domain for %s failed", pci_name(pdev));
2631 return NULL;
2632 }
2633
2634 /* make sure context mapping is ok */
2635 if (unlikely(!domain_context_mapped(pdev))) {
2636 ret = domain_context_mapping(domain, pdev,
2637 CONTEXT_TT_MULTI_LEVEL);
2638 if (ret) {
2639 printk(KERN_ERR
2640 "Domain context map for %s failed",
2641 pci_name(pdev));
2642 return NULL;
2643 }
2644 }
2645
2646 return domain;
2647}
2648
2649static inline struct dmar_domain *get_valid_domain_for_dev(struct pci_dev *dev)
2650{
2651 struct device_domain_info *info;
2652
2653 /* No lock here, assumes no domain exit in normal case */
2654 info = dev->dev.archdata.iommu;
2655 if (likely(info))
2656 return info->domain;
2657
2658 return __get_valid_domain_for_dev(dev);
2659}
2660
2661static int iommu_dummy(struct pci_dev *pdev)
2662{
2663 return pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO;
2664}
2665
2666/* Check if the pdev needs to go through non-identity map and unmap process.*/
2667static int iommu_no_mapping(struct device *dev)
2668{
2669 struct pci_dev *pdev;
2670 int found;
2671
2672 if (unlikely(dev->bus != &pci_bus_type))
2673 return 1;
2674
2675 pdev = to_pci_dev(dev);
2676 if (iommu_dummy(pdev))
2677 return 1;
2678
2679 if (!iommu_identity_mapping)
2680 return 0;
2681
2682 found = identity_mapping(pdev);
2683 if (found) {
2684 if (iommu_should_identity_map(pdev, 0))
2685 return 1;
2686 else {
2687 /*
2688 * 32 bit DMA is removed from si_domain and fall back
2689 * to non-identity mapping.
2690 */
2691 domain_remove_one_dev_info(si_domain, pdev);
2692 printk(KERN_INFO "32bit %s uses non-identity mapping\n",
2693 pci_name(pdev));
2694 return 0;
2695 }
2696 } else {
2697 /*
2698 * In case of a detached 64 bit DMA device from vm, the device
2699 * is put into si_domain for identity mapping.
2700 */
2701 if (iommu_should_identity_map(pdev, 0)) {
2702 int ret;
2703 ret = domain_add_dev_info(si_domain, pdev,
2704 hw_pass_through ?
2705 CONTEXT_TT_PASS_THROUGH :
2706 CONTEXT_TT_MULTI_LEVEL);
2707 if (!ret) {
2708 printk(KERN_INFO "64bit %s uses identity mapping\n",
2709 pci_name(pdev));
2710 return 1;
2711 }
2712 }
2713 }
2714
2715 return 0;
2716}
2717
2718static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
2719 size_t size, int dir, u64 dma_mask)
2720{
2721 struct pci_dev *pdev = to_pci_dev(hwdev);
2722 struct dmar_domain *domain;
2723 phys_addr_t start_paddr;
2724 struct iova *iova;
2725 int prot = 0;
2726 int ret;
2727 struct intel_iommu *iommu;
2728 unsigned long paddr_pfn = paddr >> PAGE_SHIFT;
2729
2730 BUG_ON(dir == DMA_NONE);
2731
2732 if (iommu_no_mapping(hwdev))
2733 return paddr;
2734
2735 domain = get_valid_domain_for_dev(pdev);
2736 if (!domain)
2737 return 0;
2738
2739 iommu = domain_get_iommu(domain);
2740 size = aligned_nrpages(paddr, size);
2741
2742 iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size), dma_mask);
2743 if (!iova)
2744 goto error;
2745
2746 /*
2747 * Check if DMAR supports zero-length reads on write only
2748 * mappings..
2749 */
2750 if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
2751 !cap_zlr(iommu->cap))
2752 prot |= DMA_PTE_READ;
2753 if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
2754 prot |= DMA_PTE_WRITE;
2755 /*
2756 * paddr - (paddr + size) might be partial page, we should map the whole
2757 * page. Note: if two part of one page are separately mapped, we
2758 * might have two guest_addr mapping to the same host paddr, but this
2759 * is not a big problem
2760 */
2761 ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
2762 mm_to_dma_pfn(paddr_pfn), size, prot);
2763 if (ret)
2764 goto error;
2765
2766 /* it's a non-present to present mapping. Only flush if caching mode */
2767 if (cap_caching_mode(iommu->cap))
2768 iommu_flush_iotlb_psi(iommu, domain->id, mm_to_dma_pfn(iova->pfn_lo), size, 1);
2769 else
2770 iommu_flush_write_buffer(iommu);
2771
2772 start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
2773 start_paddr += paddr & ~PAGE_MASK;
2774 return start_paddr;
2775
2776error:
2777 if (iova)
2778 __free_iova(&domain->iovad, iova);
2779 printk(KERN_ERR"Device %s request: %zx@%llx dir %d --- failed\n",
2780 pci_name(pdev), size, (unsigned long long)paddr, dir);
2781 return 0;
2782}
2783
2784static dma_addr_t intel_map_page(struct device *dev, struct page *page,
2785 unsigned long offset, size_t size,
2786 enum dma_data_direction dir,
2787 struct dma_attrs *attrs)
2788{
2789 return __intel_map_single(dev, page_to_phys(page) + offset, size,
2790 dir, to_pci_dev(dev)->dma_mask);
2791}
2792
2793static void flush_unmaps(void)
2794{
2795 int i, j;
2796
2797 timer_on = 0;
2798
2799 /* just flush them all */
2800 for (i = 0; i < g_num_of_iommus; i++) {
2801 struct intel_iommu *iommu = g_iommus[i];
2802 if (!iommu)
2803 continue;
2804
2805 if (!deferred_flush[i].next)
2806 continue;
2807
2808 /* In caching mode, global flushes turn emulation expensive */
2809 if (!cap_caching_mode(iommu->cap))
2810 iommu->flush.flush_iotlb(iommu, 0, 0, 0,
2811 DMA_TLB_GLOBAL_FLUSH);
2812 for (j = 0; j < deferred_flush[i].next; j++) {
2813 unsigned long mask;
2814 struct iova *iova = deferred_flush[i].iova[j];
2815 struct dmar_domain *domain = deferred_flush[i].domain[j];
2816
2817 /* On real hardware multiple invalidations are expensive */
2818 if (cap_caching_mode(iommu->cap))
2819 iommu_flush_iotlb_psi(iommu, domain->id,
2820 iova->pfn_lo, iova->pfn_hi - iova->pfn_lo + 1, 0);
2821 else {
2822 mask = ilog2(mm_to_dma_pfn(iova->pfn_hi - iova->pfn_lo + 1));
2823 iommu_flush_dev_iotlb(deferred_flush[i].domain[j],
2824 (uint64_t)iova->pfn_lo << PAGE_SHIFT, mask);
2825 }
2826 __free_iova(&deferred_flush[i].domain[j]->iovad, iova);
2827 }
2828 deferred_flush[i].next = 0;
2829 }
2830
2831 list_size = 0;
2832}
2833
2834static void flush_unmaps_timeout(unsigned long data)
2835{
2836 unsigned long flags;
2837
2838 spin_lock_irqsave(&async_umap_flush_lock, flags);
2839 flush_unmaps();
2840 spin_unlock_irqrestore(&async_umap_flush_lock, flags);
2841}
2842
2843static void add_unmap(struct dmar_domain *dom, struct iova *iova)
2844{
2845 unsigned long flags;
2846 int next, iommu_id;
2847 struct intel_iommu *iommu;
2848
2849 spin_lock_irqsave(&async_umap_flush_lock, flags);
2850 if (list_size == HIGH_WATER_MARK)
2851 flush_unmaps();
2852
2853 iommu = domain_get_iommu(dom);
2854 iommu_id = iommu->seq_id;
2855
2856 next = deferred_flush[iommu_id].next;
2857 deferred_flush[iommu_id].domain[next] = dom;
2858 deferred_flush[iommu_id].iova[next] = iova;
2859 deferred_flush[iommu_id].next++;
2860
2861 if (!timer_on) {
2862 mod_timer(&unmap_timer, jiffies + msecs_to_jiffies(10));
2863 timer_on = 1;
2864 }
2865 list_size++;
2866 spin_unlock_irqrestore(&async_umap_flush_lock, flags);
2867}
2868
2869static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
2870 size_t size, enum dma_data_direction dir,
2871 struct dma_attrs *attrs)
2872{
2873 struct pci_dev *pdev = to_pci_dev(dev);
2874 struct dmar_domain *domain;
2875 unsigned long start_pfn, last_pfn;
2876 struct iova *iova;
2877 struct intel_iommu *iommu;
2878
2879 if (iommu_no_mapping(dev))
2880 return;
2881
2882 domain = find_domain(pdev);
2883 BUG_ON(!domain);
2884
2885 iommu = domain_get_iommu(domain);
2886
2887 iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
2888 if (WARN_ONCE(!iova, "Driver unmaps unmatched page at PFN %llx\n",
2889 (unsigned long long)dev_addr))
2890 return;
2891
2892 start_pfn = mm_to_dma_pfn(iova->pfn_lo);
2893 last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
2894
2895 pr_debug("Device %s unmapping: pfn %lx-%lx\n",
2896 pci_name(pdev), start_pfn, last_pfn);
2897
2898 /* clear the whole page */
2899 dma_pte_clear_range(domain, start_pfn, last_pfn);
2900
2901 /* free page tables */
2902 dma_pte_free_pagetable(domain, start_pfn, last_pfn);
2903
2904 if (intel_iommu_strict) {
2905 iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
2906 last_pfn - start_pfn + 1, 0);
2907 /* free iova */
2908 __free_iova(&domain->iovad, iova);
2909 } else {
2910 add_unmap(domain, iova);
2911 /*
2912 * queue up the release of the unmap to save the 1/6th of the
2913 * cpu used up by the iotlb flush operation...
2914 */
2915 }
2916}
2917
2918static void *intel_alloc_coherent(struct device *hwdev, size_t size,
2919 dma_addr_t *dma_handle, gfp_t flags)
2920{
2921 void *vaddr;
2922 int order;
2923
2924 size = PAGE_ALIGN(size);
2925 order = get_order(size);
2926
2927 if (!iommu_no_mapping(hwdev))
2928 flags &= ~(GFP_DMA | GFP_DMA32);
2929 else if (hwdev->coherent_dma_mask < dma_get_required_mask(hwdev)) {
2930 if (hwdev->coherent_dma_mask < DMA_BIT_MASK(32))
2931 flags |= GFP_DMA;
2932 else
2933 flags |= GFP_DMA32;
2934 }
2935
2936 vaddr = (void *)__get_free_pages(flags, order);
2937 if (!vaddr)
2938 return NULL;
2939 memset(vaddr, 0, size);
2940
2941 *dma_handle = __intel_map_single(hwdev, virt_to_bus(vaddr), size,
2942 DMA_BIDIRECTIONAL,
2943 hwdev->coherent_dma_mask);
2944 if (*dma_handle)
2945 return vaddr;
2946 free_pages((unsigned long)vaddr, order);
2947 return NULL;
2948}
2949
2950static void intel_free_coherent(struct device *hwdev, size_t size, void *vaddr,
2951 dma_addr_t dma_handle)
2952{
2953 int order;
2954
2955 size = PAGE_ALIGN(size);
2956 order = get_order(size);
2957
2958 intel_unmap_page(hwdev, dma_handle, size, DMA_BIDIRECTIONAL, NULL);
2959 free_pages((unsigned long)vaddr, order);
2960}
2961
2962static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
2963 int nelems, enum dma_data_direction dir,
2964 struct dma_attrs *attrs)
2965{
2966 struct pci_dev *pdev = to_pci_dev(hwdev);
2967 struct dmar_domain *domain;
2968 unsigned long start_pfn, last_pfn;
2969 struct iova *iova;
2970 struct intel_iommu *iommu;
2971
2972 if (iommu_no_mapping(hwdev))
2973 return;
2974
2975 domain = find_domain(pdev);
2976 BUG_ON(!domain);
2977
2978 iommu = domain_get_iommu(domain);
2979
2980 iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
2981 if (WARN_ONCE(!iova, "Driver unmaps unmatched sglist at PFN %llx\n",
2982 (unsigned long long)sglist[0].dma_address))
2983 return;
2984
2985 start_pfn = mm_to_dma_pfn(iova->pfn_lo);
2986 last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
2987
2988 /* clear the whole page */
2989 dma_pte_clear_range(domain, start_pfn, last_pfn);
2990
2991 /* free page tables */
2992 dma_pte_free_pagetable(domain, start_pfn, last_pfn);
2993
2994 if (intel_iommu_strict) {
2995 iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
2996 last_pfn - start_pfn + 1, 0);
2997 /* free iova */
2998 __free_iova(&domain->iovad, iova);
2999 } else {
3000 add_unmap(domain, iova);
3001 /*
3002 * queue up the release of the unmap to save the 1/6th of the
3003 * cpu used up by the iotlb flush operation...
3004 */
3005 }
3006}
3007
3008static int intel_nontranslate_map_sg(struct device *hddev,
3009 struct scatterlist *sglist, int nelems, int dir)
3010{
3011 int i;
3012 struct scatterlist *sg;
3013
3014 for_each_sg(sglist, sg, nelems, i) {
3015 BUG_ON(!sg_page(sg));
3016 sg->dma_address = page_to_phys(sg_page(sg)) + sg->offset;
3017 sg->dma_length = sg->length;
3018 }
3019 return nelems;
3020}
3021
3022static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
3023 enum dma_data_direction dir, struct dma_attrs *attrs)
3024{
3025 int i;
3026 struct pci_dev *pdev = to_pci_dev(hwdev);
3027 struct dmar_domain *domain;
3028 size_t size = 0;
3029 int prot = 0;
3030 struct iova *iova = NULL;
3031 int ret;
3032 struct scatterlist *sg;
3033 unsigned long start_vpfn;
3034 struct intel_iommu *iommu;
3035
3036 BUG_ON(dir == DMA_NONE);
3037 if (iommu_no_mapping(hwdev))
3038 return intel_nontranslate_map_sg(hwdev, sglist, nelems, dir);
3039
3040 domain = get_valid_domain_for_dev(pdev);
3041 if (!domain)
3042 return 0;
3043
3044 iommu = domain_get_iommu(domain);
3045
3046 for_each_sg(sglist, sg, nelems, i)
3047 size += aligned_nrpages(sg->offset, sg->length);
3048
3049 iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
3050 pdev->dma_mask);
3051 if (!iova) {
3052 sglist->dma_length = 0;
3053 return 0;
3054 }
3055
3056 /*
3057 * Check if DMAR supports zero-length reads on write only
3058 * mappings..
3059 */
3060 if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
3061 !cap_zlr(iommu->cap))
3062 prot |= DMA_PTE_READ;
3063 if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
3064 prot |= DMA_PTE_WRITE;
3065
3066 start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
3067
3068 ret = domain_sg_mapping(domain, start_vpfn, sglist, size, prot);
3069 if (unlikely(ret)) {
3070 /* clear the page */
3071 dma_pte_clear_range(domain, start_vpfn,
3072 start_vpfn + size - 1);
3073 /* free page tables */
3074 dma_pte_free_pagetable(domain, start_vpfn,
3075 start_vpfn + size - 1);
3076 /* free iova */
3077 __free_iova(&domain->iovad, iova);
3078 return 0;
3079 }
3080
3081 /* it's a non-present to present mapping. Only flush if caching mode */
3082 if (cap_caching_mode(iommu->cap))
3083 iommu_flush_iotlb_psi(iommu, domain->id, start_vpfn, size, 1);
3084 else
3085 iommu_flush_write_buffer(iommu);
3086
3087 return nelems;
3088}
3089
3090static int intel_mapping_error(struct device *dev, dma_addr_t dma_addr)
3091{
3092 return !dma_addr;
3093}
3094
3095struct dma_map_ops intel_dma_ops = {
3096 .alloc_coherent = intel_alloc_coherent,
3097 .free_coherent = intel_free_coherent,
3098 .map_sg = intel_map_sg,
3099 .unmap_sg = intel_unmap_sg,
3100 .map_page = intel_map_page,
3101 .unmap_page = intel_unmap_page,
3102 .mapping_error = intel_mapping_error,
3103};
3104
3105static inline int iommu_domain_cache_init(void)
3106{
3107 int ret = 0;
3108
3109 iommu_domain_cache = kmem_cache_create("iommu_domain",
3110 sizeof(struct dmar_domain),
3111 0,
3112 SLAB_HWCACHE_ALIGN,
3113
3114 NULL);
3115 if (!iommu_domain_cache) {
3116 printk(KERN_ERR "Couldn't create iommu_domain cache\n");
3117 ret = -ENOMEM;
3118 }
3119
3120 return ret;
3121}
3122
3123static inline int iommu_devinfo_cache_init(void)
3124{
3125 int ret = 0;
3126
3127 iommu_devinfo_cache = kmem_cache_create("iommu_devinfo",
3128 sizeof(struct device_domain_info),
3129 0,
3130 SLAB_HWCACHE_ALIGN,
3131 NULL);
3132 if (!iommu_devinfo_cache) {
3133 printk(KERN_ERR "Couldn't create devinfo cache\n");
3134 ret = -ENOMEM;
3135 }
3136
3137 return ret;
3138}
3139
3140static inline int iommu_iova_cache_init(void)
3141{
3142 int ret = 0;
3143
3144 iommu_iova_cache = kmem_cache_create("iommu_iova",
3145 sizeof(struct iova),
3146 0,
3147 SLAB_HWCACHE_ALIGN,
3148 NULL);
3149 if (!iommu_iova_cache) {
3150 printk(KERN_ERR "Couldn't create iova cache\n");
3151 ret = -ENOMEM;
3152 }
3153
3154 return ret;
3155}
3156
3157static int __init iommu_init_mempool(void)
3158{
3159 int ret;
3160 ret = iommu_iova_cache_init();
3161 if (ret)
3162 return ret;
3163
3164 ret = iommu_domain_cache_init();
3165 if (ret)
3166 goto domain_error;
3167
3168 ret = iommu_devinfo_cache_init();
3169 if (!ret)
3170 return ret;
3171
3172 kmem_cache_destroy(iommu_domain_cache);
3173domain_error:
3174 kmem_cache_destroy(iommu_iova_cache);
3175
3176 return -ENOMEM;
3177}
3178
3179static void __init iommu_exit_mempool(void)
3180{
3181 kmem_cache_destroy(iommu_devinfo_cache);
3182 kmem_cache_destroy(iommu_domain_cache);
3183 kmem_cache_destroy(iommu_iova_cache);
3184
3185}
3186
3187static void quirk_ioat_snb_local_iommu(struct pci_dev *pdev)
3188{
3189 struct dmar_drhd_unit *drhd;
3190 u32 vtbar;
3191 int rc;
3192
3193 /* We know that this device on this chipset has its own IOMMU.
3194 * If we find it under a different IOMMU, then the BIOS is lying
3195 * to us. Hope that the IOMMU for this device is actually
3196 * disabled, and it needs no translation...
3197 */
3198 rc = pci_bus_read_config_dword(pdev->bus, PCI_DEVFN(0, 0), 0xb0, &vtbar);
3199 if (rc) {
3200 /* "can't" happen */
3201 dev_info(&pdev->dev, "failed to run vt-d quirk\n");
3202 return;
3203 }
3204 vtbar &= 0xffff0000;
3205
3206 /* we know that the this iommu should be at offset 0xa000 from vtbar */
3207 drhd = dmar_find_matched_drhd_unit(pdev);
3208 if (WARN_TAINT_ONCE(!drhd || drhd->reg_base_addr - vtbar != 0xa000,
3209 TAINT_FIRMWARE_WORKAROUND,
3210 "BIOS assigned incorrect VT-d unit for Intel(R) QuickData Technology device\n"))
3211 pdev->dev.archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO;
3212}
3213DECLARE_PCI_FIXUP_ENABLE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IOAT_SNB, quirk_ioat_snb_local_iommu);
3214
3215static void __init init_no_remapping_devices(void)
3216{
3217 struct dmar_drhd_unit *drhd;
3218
3219 for_each_drhd_unit(drhd) {
3220 if (!drhd->include_all) {
3221 int i;
3222 for (i = 0; i < drhd->devices_cnt; i++)
3223 if (drhd->devices[i] != NULL)
3224 break;
3225 /* ignore DMAR unit if no pci devices exist */
3226 if (i == drhd->devices_cnt)
3227 drhd->ignored = 1;
3228 }
3229 }
3230
3231 for_each_drhd_unit(drhd) {
3232 int i;
3233 if (drhd->ignored || drhd->include_all)
3234 continue;
3235
3236 for (i = 0; i < drhd->devices_cnt; i++)
3237 if (drhd->devices[i] &&
3238 !IS_GFX_DEVICE(drhd->devices[i]))
3239 break;
3240
3241 if (i < drhd->devices_cnt)
3242 continue;
3243
3244 /* This IOMMU has *only* gfx devices. Either bypass it or
3245 set the gfx_mapped flag, as appropriate */
3246 if (dmar_map_gfx) {
3247 intel_iommu_gfx_mapped = 1;
3248 } else {
3249 drhd->ignored = 1;
3250 for (i = 0; i < drhd->devices_cnt; i++) {
3251 if (!drhd->devices[i])
3252 continue;
3253 drhd->devices[i]->dev.archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO;
3254 }
3255 }
3256 }
3257}
3258
3259#ifdef CONFIG_SUSPEND
3260static int init_iommu_hw(void)
3261{
3262 struct dmar_drhd_unit *drhd;
3263 struct intel_iommu *iommu = NULL;
3264
3265 for_each_active_iommu(iommu, drhd)
3266 if (iommu->qi)
3267 dmar_reenable_qi(iommu);
3268
3269 for_each_iommu(iommu, drhd) {
3270 if (drhd->ignored) {
3271 /*
3272 * we always have to disable PMRs or DMA may fail on
3273 * this device
3274 */
3275 if (force_on)
3276 iommu_disable_protect_mem_regions(iommu);
3277 continue;
3278 }
3279
3280 iommu_flush_write_buffer(iommu);
3281
3282 iommu_set_root_entry(iommu);
3283
3284 iommu->flush.flush_context(iommu, 0, 0, 0,
3285 DMA_CCMD_GLOBAL_INVL);
3286 iommu->flush.flush_iotlb(iommu, 0, 0, 0,
3287 DMA_TLB_GLOBAL_FLUSH);
3288 if (iommu_enable_translation(iommu))
3289 return 1;
3290 iommu_disable_protect_mem_regions(iommu);
3291 }
3292
3293 return 0;
3294}
3295
3296static void iommu_flush_all(void)
3297{
3298 struct dmar_drhd_unit *drhd;
3299 struct intel_iommu *iommu;
3300
3301 for_each_active_iommu(iommu, drhd) {
3302 iommu->flush.flush_context(iommu, 0, 0, 0,
3303 DMA_CCMD_GLOBAL_INVL);
3304 iommu->flush.flush_iotlb(iommu, 0, 0, 0,
3305 DMA_TLB_GLOBAL_FLUSH);
3306 }
3307}
3308
3309static int iommu_suspend(void)
3310{
3311 struct dmar_drhd_unit *drhd;
3312 struct intel_iommu *iommu = NULL;
3313 unsigned long flag;
3314
3315 for_each_active_iommu(iommu, drhd) {
3316 iommu->iommu_state = kzalloc(sizeof(u32) * MAX_SR_DMAR_REGS,
3317 GFP_ATOMIC);
3318 if (!iommu->iommu_state)
3319 goto nomem;
3320 }
3321
3322 iommu_flush_all();
3323
3324 for_each_active_iommu(iommu, drhd) {
3325 iommu_disable_translation(iommu);
3326
3327 raw_spin_lock_irqsave(&iommu->register_lock, flag);
3328
3329 iommu->iommu_state[SR_DMAR_FECTL_REG] =
3330 readl(iommu->reg + DMAR_FECTL_REG);
3331 iommu->iommu_state[SR_DMAR_FEDATA_REG] =
3332 readl(iommu->reg + DMAR_FEDATA_REG);
3333 iommu->iommu_state[SR_DMAR_FEADDR_REG] =
3334 readl(iommu->reg + DMAR_FEADDR_REG);
3335 iommu->iommu_state[SR_DMAR_FEUADDR_REG] =
3336 readl(iommu->reg + DMAR_FEUADDR_REG);
3337
3338 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
3339 }
3340 return 0;
3341
3342nomem:
3343 for_each_active_iommu(iommu, drhd)
3344 kfree(iommu->iommu_state);
3345
3346 return -ENOMEM;
3347}
3348
3349static void iommu_resume(void)
3350{
3351 struct dmar_drhd_unit *drhd;
3352 struct intel_iommu *iommu = NULL;
3353 unsigned long flag;
3354
3355 if (init_iommu_hw()) {
3356 if (force_on)
3357 panic("tboot: IOMMU setup failed, DMAR can not resume!\n");
3358 else
3359 WARN(1, "IOMMU setup failed, DMAR can not resume!\n");
3360 return;
3361 }
3362
3363 for_each_active_iommu(iommu, drhd) {
3364
3365 raw_spin_lock_irqsave(&iommu->register_lock, flag);
3366
3367 writel(iommu->iommu_state[SR_DMAR_FECTL_REG],
3368 iommu->reg + DMAR_FECTL_REG);
3369 writel(iommu->iommu_state[SR_DMAR_FEDATA_REG],
3370 iommu->reg + DMAR_FEDATA_REG);
3371 writel(iommu->iommu_state[SR_DMAR_FEADDR_REG],
3372 iommu->reg + DMAR_FEADDR_REG);
3373 writel(iommu->iommu_state[SR_DMAR_FEUADDR_REG],
3374 iommu->reg + DMAR_FEUADDR_REG);
3375
3376 raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
3377 }
3378
3379 for_each_active_iommu(iommu, drhd)
3380 kfree(iommu->iommu_state);
3381}
3382
3383static struct syscore_ops iommu_syscore_ops = {
3384 .resume = iommu_resume,
3385 .suspend = iommu_suspend,
3386};
3387
3388static void __init init_iommu_pm_ops(void)
3389{
3390 register_syscore_ops(&iommu_syscore_ops);
3391}
3392
3393#else
3394static inline void init_iommu_pm_ops(void) {}
3395#endif /* CONFIG_PM */
3396
3397LIST_HEAD(dmar_rmrr_units);
3398
3399static void __init dmar_register_rmrr_unit(struct dmar_rmrr_unit *rmrr)
3400{
3401 list_add(&rmrr->list, &dmar_rmrr_units);
3402}
3403
3404
3405int __init dmar_parse_one_rmrr(struct acpi_dmar_header *header)
3406{
3407 struct acpi_dmar_reserved_memory *rmrr;
3408 struct dmar_rmrr_unit *rmrru;
3409
3410 rmrru = kzalloc(sizeof(*rmrru), GFP_KERNEL);
3411 if (!rmrru)
3412 return -ENOMEM;
3413
3414 rmrru->hdr = header;
3415 rmrr = (struct acpi_dmar_reserved_memory *)header;
3416 rmrru->base_address = rmrr->base_address;
3417 rmrru->end_address = rmrr->end_address;
3418
3419 dmar_register_rmrr_unit(rmrru);
3420 return 0;
3421}
3422
3423static int __init
3424rmrr_parse_dev(struct dmar_rmrr_unit *rmrru)
3425{
3426 struct acpi_dmar_reserved_memory *rmrr;
3427 int ret;
3428
3429 rmrr = (struct acpi_dmar_reserved_memory *) rmrru->hdr;
3430 ret = dmar_parse_dev_scope((void *)(rmrr + 1),
3431 ((void *)rmrr) + rmrr->header.length,
3432 &rmrru->devices_cnt, &rmrru->devices, rmrr->segment);
3433
3434 if (ret || (rmrru->devices_cnt == 0)) {
3435 list_del(&rmrru->list);
3436 kfree(rmrru);
3437 }
3438 return ret;
3439}
3440
3441static LIST_HEAD(dmar_atsr_units);
3442
3443int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
3444{
3445 struct acpi_dmar_atsr *atsr;
3446 struct dmar_atsr_unit *atsru;
3447
3448 atsr = container_of(hdr, struct acpi_dmar_atsr, header);
3449 atsru = kzalloc(sizeof(*atsru), GFP_KERNEL);
3450 if (!atsru)
3451 return -ENOMEM;
3452
3453 atsru->hdr = hdr;
3454 atsru->include_all = atsr->flags & 0x1;
3455
3456 list_add(&atsru->list, &dmar_atsr_units);
3457
3458 return 0;
3459}
3460
3461static int __init atsr_parse_dev(struct dmar_atsr_unit *atsru)
3462{
3463 int rc;
3464 struct acpi_dmar_atsr *atsr;
3465
3466 if (atsru->include_all)
3467 return 0;
3468
3469 atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
3470 rc = dmar_parse_dev_scope((void *)(atsr + 1),
3471 (void *)atsr + atsr->header.length,
3472 &atsru->devices_cnt, &atsru->devices,
3473 atsr->segment);
3474 if (rc || !atsru->devices_cnt) {
3475 list_del(&atsru->list);
3476 kfree(atsru);
3477 }
3478
3479 return rc;
3480}
3481
3482int dmar_find_matched_atsr_unit(struct pci_dev *dev)
3483{
3484 int i;
3485 struct pci_bus *bus;
3486 struct acpi_dmar_atsr *atsr;
3487 struct dmar_atsr_unit *atsru;
3488
3489 dev = pci_physfn(dev);
3490
3491 list_for_each_entry(atsru, &dmar_atsr_units, list) {
3492 atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
3493 if (atsr->segment == pci_domain_nr(dev->bus))
3494 goto found;
3495 }
3496
3497 return 0;
3498
3499found:
3500 for (bus = dev->bus; bus; bus = bus->parent) {
3501 struct pci_dev *bridge = bus->self;
3502
3503 if (!bridge || !pci_is_pcie(bridge) ||
3504 bridge->pcie_type == PCI_EXP_TYPE_PCI_BRIDGE)
3505 return 0;
3506
3507 if (bridge->pcie_type == PCI_EXP_TYPE_ROOT_PORT) {
3508 for (i = 0; i < atsru->devices_cnt; i++)
3509 if (atsru->devices[i] == bridge)
3510 return 1;
3511 break;
3512 }
3513 }
3514
3515 if (atsru->include_all)
3516 return 1;
3517
3518 return 0;
3519}
3520
3521int dmar_parse_rmrr_atsr_dev(void)
3522{
3523 struct dmar_rmrr_unit *rmrr, *rmrr_n;
3524 struct dmar_atsr_unit *atsr, *atsr_n;
3525 int ret = 0;
3526
3527 list_for_each_entry_safe(rmrr, rmrr_n, &dmar_rmrr_units, list) {
3528 ret = rmrr_parse_dev(rmrr);
3529 if (ret)
3530 return ret;
3531 }
3532
3533 list_for_each_entry_safe(atsr, atsr_n, &dmar_atsr_units, list) {
3534 ret = atsr_parse_dev(atsr);
3535 if (ret)
3536 return ret;
3537 }
3538
3539 return ret;
3540}
3541
3542/*
3543 * Here we only respond to action of unbound device from driver.
3544 *
3545 * Added device is not attached to its DMAR domain here yet. That will happen
3546 * when mapping the device to iova.
3547 */
3548static int device_notifier(struct notifier_block *nb,
3549 unsigned long action, void *data)
3550{
3551 struct device *dev = data;
3552 struct pci_dev *pdev = to_pci_dev(dev);
3553 struct dmar_domain *domain;
3554
3555 if (iommu_no_mapping(dev))
3556 return 0;
3557
3558 domain = find_domain(pdev);
3559 if (!domain)
3560 return 0;
3561
3562 if (action == BUS_NOTIFY_UNBOUND_DRIVER && !iommu_pass_through) {
3563 domain_remove_one_dev_info(domain, pdev);
3564
3565 if (!(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) &&
3566 !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) &&
3567 list_empty(&domain->devices))
3568 domain_exit(domain);
3569 }
3570
3571 return 0;
3572}
3573
3574static struct notifier_block device_nb = {
3575 .notifier_call = device_notifier,
3576};
3577
3578int __init intel_iommu_init(void)
3579{
3580 int ret = 0;
3581
3582 /* VT-d is required for a TXT/tboot launch, so enforce that */
3583 force_on = tboot_force_iommu();
3584
3585 if (dmar_table_init()) {
3586 if (force_on)
3587 panic("tboot: Failed to initialize DMAR table\n");
3588 return -ENODEV;
3589 }
3590
3591 if (dmar_dev_scope_init() < 0) {
3592 if (force_on)
3593 panic("tboot: Failed to initialize DMAR device scope\n");
3594 return -ENODEV;
3595 }
3596
3597 if (no_iommu || dmar_disabled)
3598 return -ENODEV;
3599
3600 if (iommu_init_mempool()) {
3601 if (force_on)
3602 panic("tboot: Failed to initialize iommu memory\n");
3603 return -ENODEV;
3604 }
3605
3606 if (list_empty(&dmar_rmrr_units))
3607 printk(KERN_INFO "DMAR: No RMRR found\n");
3608
3609 if (list_empty(&dmar_atsr_units))
3610 printk(KERN_INFO "DMAR: No ATSR found\n");
3611
3612 if (dmar_init_reserved_ranges()) {
3613 if (force_on)
3614 panic("tboot: Failed to reserve iommu ranges\n");
3615 return -ENODEV;
3616 }
3617
3618 init_no_remapping_devices();
3619
3620 ret = init_dmars();
3621 if (ret) {
3622 if (force_on)
3623 panic("tboot: Failed to initialize DMARs\n");
3624 printk(KERN_ERR "IOMMU: dmar init failed\n");
3625 put_iova_domain(&reserved_iova_list);
3626 iommu_exit_mempool();
3627 return ret;
3628 }
3629 printk(KERN_INFO
3630 "PCI-DMA: Intel(R) Virtualization Technology for Directed I/O\n");
3631
3632 init_timer(&unmap_timer);
3633#ifdef CONFIG_SWIOTLB
3634 swiotlb = 0;
3635#endif
3636 dma_ops = &intel_dma_ops;
3637
3638 init_iommu_pm_ops();
3639
3640 bus_set_iommu(&pci_bus_type, &intel_iommu_ops);
3641
3642 bus_register_notifier(&pci_bus_type, &device_nb);
3643
3644 return 0;
3645}
3646
3647static void iommu_detach_dependent_devices(struct intel_iommu *iommu,
3648 struct pci_dev *pdev)
3649{
3650 struct pci_dev *tmp, *parent;
3651
3652 if (!iommu || !pdev)
3653 return;
3654
3655 /* dependent device detach */
3656 tmp = pci_find_upstream_pcie_bridge(pdev);
3657 /* Secondary interface's bus number and devfn 0 */
3658 if (tmp) {
3659 parent = pdev->bus->self;
3660 while (parent != tmp) {
3661 iommu_detach_dev(iommu, parent->bus->number,
3662 parent->devfn);
3663 parent = parent->bus->self;
3664 }
3665 if (pci_is_pcie(tmp)) /* this is a PCIe-to-PCI bridge */
3666 iommu_detach_dev(iommu,
3667 tmp->subordinate->number, 0);
3668 else /* this is a legacy PCI bridge */
3669 iommu_detach_dev(iommu, tmp->bus->number,
3670 tmp->devfn);
3671 }
3672}
3673
3674static void domain_remove_one_dev_info(struct dmar_domain *domain,
3675 struct pci_dev *pdev)
3676{
3677 struct device_domain_info *info;
3678 struct intel_iommu *iommu;
3679 unsigned long flags;
3680 int found = 0;
3681 struct list_head *entry, *tmp;
3682
3683 iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number,
3684 pdev->devfn);
3685 if (!iommu)
3686 return;
3687
3688 spin_lock_irqsave(&device_domain_lock, flags);
3689 list_for_each_safe(entry, tmp, &domain->devices) {
3690 info = list_entry(entry, struct device_domain_info, link);
3691 if (info->segment == pci_domain_nr(pdev->bus) &&
3692 info->bus == pdev->bus->number &&
3693 info->devfn == pdev->devfn) {
3694 list_del(&info->link);
3695 list_del(&info->global);
3696 if (info->dev)
3697 info->dev->dev.archdata.iommu = NULL;
3698 spin_unlock_irqrestore(&device_domain_lock, flags);
3699
3700 iommu_disable_dev_iotlb(info);
3701 iommu_detach_dev(iommu, info->bus, info->devfn);
3702 iommu_detach_dependent_devices(iommu, pdev);
3703 free_devinfo_mem(info);
3704
3705 spin_lock_irqsave(&device_domain_lock, flags);
3706
3707 if (found)
3708 break;
3709 else
3710 continue;
3711 }
3712
3713 /* if there is no other devices under the same iommu
3714 * owned by this domain, clear this iommu in iommu_bmp
3715 * update iommu count and coherency
3716 */
3717 if (iommu == device_to_iommu(info->segment, info->bus,
3718 info->devfn))
3719 found = 1;
3720 }
3721
3722 spin_unlock_irqrestore(&device_domain_lock, flags);
3723
3724 if (found == 0) {
3725 unsigned long tmp_flags;
3726 spin_lock_irqsave(&domain->iommu_lock, tmp_flags);
3727 clear_bit(iommu->seq_id, &domain->iommu_bmp);
3728 domain->iommu_count--;
3729 domain_update_iommu_cap(domain);
3730 spin_unlock_irqrestore(&domain->iommu_lock, tmp_flags);
3731
3732 if (!(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) &&
3733 !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY)) {
3734 spin_lock_irqsave(&iommu->lock, tmp_flags);
3735 clear_bit(domain->id, iommu->domain_ids);
3736 iommu->domains[domain->id] = NULL;
3737 spin_unlock_irqrestore(&iommu->lock, tmp_flags);
3738 }
3739 }
3740}
3741
3742static void vm_domain_remove_all_dev_info(struct dmar_domain *domain)
3743{
3744 struct device_domain_info *info;
3745 struct intel_iommu *iommu;
3746 unsigned long flags1, flags2;
3747
3748 spin_lock_irqsave(&device_domain_lock, flags1);
3749 while (!list_empty(&domain->devices)) {
3750 info = list_entry(domain->devices.next,
3751 struct device_domain_info, link);
3752 list_del(&info->link);
3753 list_del(&info->global);
3754 if (info->dev)
3755 info->dev->dev.archdata.iommu = NULL;
3756
3757 spin_unlock_irqrestore(&device_domain_lock, flags1);
3758
3759 iommu_disable_dev_iotlb(info);
3760 iommu = device_to_iommu(info->segment, info->bus, info->devfn);
3761 iommu_detach_dev(iommu, info->bus, info->devfn);
3762 iommu_detach_dependent_devices(iommu, info->dev);
3763
3764 /* clear this iommu in iommu_bmp, update iommu count
3765 * and capabilities
3766 */
3767 spin_lock_irqsave(&domain->iommu_lock, flags2);
3768 if (test_and_clear_bit(iommu->seq_id,
3769 &domain->iommu_bmp)) {
3770 domain->iommu_count--;
3771 domain_update_iommu_cap(domain);
3772 }
3773 spin_unlock_irqrestore(&domain->iommu_lock, flags2);
3774
3775 free_devinfo_mem(info);
3776 spin_lock_irqsave(&device_domain_lock, flags1);
3777 }
3778 spin_unlock_irqrestore(&device_domain_lock, flags1);
3779}
3780
3781/* domain id for virtual machine, it won't be set in context */
3782static unsigned long vm_domid;
3783
3784static struct dmar_domain *iommu_alloc_vm_domain(void)
3785{
3786 struct dmar_domain *domain;
3787
3788 domain = alloc_domain_mem();
3789 if (!domain)
3790 return NULL;
3791
3792 domain->id = vm_domid++;
3793 domain->nid = -1;
3794 memset(&domain->iommu_bmp, 0, sizeof(unsigned long));
3795 domain->flags = DOMAIN_FLAG_VIRTUAL_MACHINE;
3796
3797 return domain;
3798}
3799
3800static int md_domain_init(struct dmar_domain *domain, int guest_width)
3801{
3802 int adjust_width;
3803
3804 init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
3805 spin_lock_init(&domain->iommu_lock);
3806
3807 domain_reserve_special_ranges(domain);
3808
3809 /* calculate AGAW */
3810 domain->gaw = guest_width;
3811 adjust_width = guestwidth_to_adjustwidth(guest_width);
3812 domain->agaw = width_to_agaw(adjust_width);
3813
3814 INIT_LIST_HEAD(&domain->devices);
3815
3816 domain->iommu_count = 0;
3817 domain->iommu_coherency = 0;
3818 domain->iommu_snooping = 0;
3819 domain->iommu_superpage = 0;
3820 domain->max_addr = 0;
3821 domain->nid = -1;
3822
3823 /* always allocate the top pgd */
3824 domain->pgd = (struct dma_pte *)alloc_pgtable_page(domain->nid);
3825 if (!domain->pgd)
3826 return -ENOMEM;
3827 domain_flush_cache(domain, domain->pgd, PAGE_SIZE);
3828 return 0;
3829}
3830
3831static void iommu_free_vm_domain(struct dmar_domain *domain)
3832{
3833 unsigned long flags;
3834 struct dmar_drhd_unit *drhd;
3835 struct intel_iommu *iommu;
3836 unsigned long i;
3837 unsigned long ndomains;
3838
3839 for_each_drhd_unit(drhd) {
3840 if (drhd->ignored)
3841 continue;
3842 iommu = drhd->iommu;
3843
3844 ndomains = cap_ndoms(iommu->cap);
3845 for_each_set_bit(i, iommu->domain_ids, ndomains) {
3846 if (iommu->domains[i] == domain) {
3847 spin_lock_irqsave(&iommu->lock, flags);
3848 clear_bit(i, iommu->domain_ids);
3849 iommu->domains[i] = NULL;
3850 spin_unlock_irqrestore(&iommu->lock, flags);
3851 break;
3852 }
3853 }
3854 }
3855}
3856
3857static void vm_domain_exit(struct dmar_domain *domain)
3858{
3859 /* Domain 0 is reserved, so dont process it */
3860 if (!domain)
3861 return;
3862
3863 vm_domain_remove_all_dev_info(domain);
3864 /* destroy iovas */
3865 put_iova_domain(&domain->iovad);
3866
3867 /* clear ptes */
3868 dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
3869
3870 /* free page tables */
3871 dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
3872
3873 iommu_free_vm_domain(domain);
3874 free_domain_mem(domain);
3875}
3876
3877static int intel_iommu_domain_init(struct iommu_domain *domain)
3878{
3879 struct dmar_domain *dmar_domain;
3880
3881 dmar_domain = iommu_alloc_vm_domain();
3882 if (!dmar_domain) {
3883 printk(KERN_ERR
3884 "intel_iommu_domain_init: dmar_domain == NULL\n");
3885 return -ENOMEM;
3886 }
3887 if (md_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
3888 printk(KERN_ERR
3889 "intel_iommu_domain_init() failed\n");
3890 vm_domain_exit(dmar_domain);
3891 return -ENOMEM;
3892 }
3893 domain_update_iommu_cap(dmar_domain);
3894 domain->priv = dmar_domain;
3895
3896 return 0;
3897}
3898
3899static void intel_iommu_domain_destroy(struct iommu_domain *domain)
3900{
3901 struct dmar_domain *dmar_domain = domain->priv;
3902
3903 domain->priv = NULL;
3904 vm_domain_exit(dmar_domain);
3905}
3906
3907static int intel_iommu_attach_device(struct iommu_domain *domain,
3908 struct device *dev)
3909{
3910 struct dmar_domain *dmar_domain = domain->priv;
3911 struct pci_dev *pdev = to_pci_dev(dev);
3912 struct intel_iommu *iommu;
3913 int addr_width;
3914
3915 /* normally pdev is not mapped */
3916 if (unlikely(domain_context_mapped(pdev))) {
3917 struct dmar_domain *old_domain;
3918
3919 old_domain = find_domain(pdev);
3920 if (old_domain) {
3921 if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
3922 dmar_domain->flags & DOMAIN_FLAG_STATIC_IDENTITY)
3923 domain_remove_one_dev_info(old_domain, pdev);
3924 else
3925 domain_remove_dev_info(old_domain);
3926 }
3927 }
3928
3929 iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number,
3930 pdev->devfn);
3931 if (!iommu)
3932 return -ENODEV;
3933
3934 /* check if this iommu agaw is sufficient for max mapped address */
3935 addr_width = agaw_to_width(iommu->agaw);
3936 if (addr_width > cap_mgaw(iommu->cap))
3937 addr_width = cap_mgaw(iommu->cap);
3938
3939 if (dmar_domain->max_addr > (1LL << addr_width)) {
3940 printk(KERN_ERR "%s: iommu width (%d) is not "
3941 "sufficient for the mapped address (%llx)\n",
3942 __func__, addr_width, dmar_domain->max_addr);
3943 return -EFAULT;
3944 }
3945 dmar_domain->gaw = addr_width;
3946
3947 /*
3948 * Knock out extra levels of page tables if necessary
3949 */
3950 while (iommu->agaw < dmar_domain->agaw) {
3951 struct dma_pte *pte;
3952
3953 pte = dmar_domain->pgd;
3954 if (dma_pte_present(pte)) {
3955 dmar_domain->pgd = (struct dma_pte *)
3956 phys_to_virt(dma_pte_addr(pte));
3957 free_pgtable_page(pte);
3958 }
3959 dmar_domain->agaw--;
3960 }
3961
3962 return domain_add_dev_info(dmar_domain, pdev, CONTEXT_TT_MULTI_LEVEL);
3963}
3964
3965static void intel_iommu_detach_device(struct iommu_domain *domain,
3966 struct device *dev)
3967{
3968 struct dmar_domain *dmar_domain = domain->priv;
3969 struct pci_dev *pdev = to_pci_dev(dev);
3970
3971 domain_remove_one_dev_info(dmar_domain, pdev);
3972}
3973
3974static int intel_iommu_map(struct iommu_domain *domain,
3975 unsigned long iova, phys_addr_t hpa,
3976 int gfp_order, int iommu_prot)
3977{
3978 struct dmar_domain *dmar_domain = domain->priv;
3979 u64 max_addr;
3980 int prot = 0;
3981 size_t size;
3982 int ret;
3983
3984 if (iommu_prot & IOMMU_READ)
3985 prot |= DMA_PTE_READ;
3986 if (iommu_prot & IOMMU_WRITE)
3987 prot |= DMA_PTE_WRITE;
3988 if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
3989 prot |= DMA_PTE_SNP;
3990
3991 size = PAGE_SIZE << gfp_order;
3992 max_addr = iova + size;
3993 if (dmar_domain->max_addr < max_addr) {
3994 u64 end;
3995
3996 /* check if minimum agaw is sufficient for mapped address */
3997 end = __DOMAIN_MAX_ADDR(dmar_domain->gaw) + 1;
3998 if (end < max_addr) {
3999 printk(KERN_ERR "%s: iommu width (%d) is not "
4000 "sufficient for the mapped address (%llx)\n",
4001 __func__, dmar_domain->gaw, max_addr);
4002 return -EFAULT;
4003 }
4004 dmar_domain->max_addr = max_addr;
4005 }
4006 /* Round up size to next multiple of PAGE_SIZE, if it and
4007 the low bits of hpa would take us onto the next page */
4008 size = aligned_nrpages(hpa, size);
4009 ret = domain_pfn_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
4010 hpa >> VTD_PAGE_SHIFT, size, prot);
4011 return ret;
4012}
4013
4014static int intel_iommu_unmap(struct iommu_domain *domain,
4015 unsigned long iova, int gfp_order)
4016{
4017 struct dmar_domain *dmar_domain = domain->priv;
4018 size_t size = PAGE_SIZE << gfp_order;
4019 int order;
4020
4021 order = dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
4022 (iova + size - 1) >> VTD_PAGE_SHIFT);
4023
4024 if (dmar_domain->max_addr == iova + size)
4025 dmar_domain->max_addr = iova;
4026
4027 return order;
4028}
4029
4030static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
4031 unsigned long iova)
4032{
4033 struct dmar_domain *dmar_domain = domain->priv;
4034 struct dma_pte *pte;
4035 u64 phys = 0;
4036
4037 pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, 0);
4038 if (pte)
4039 phys = dma_pte_addr(pte);
4040
4041 return phys;
4042}
4043
4044static int intel_iommu_domain_has_cap(struct iommu_domain *domain,
4045 unsigned long cap)
4046{
4047 struct dmar_domain *dmar_domain = domain->priv;
4048
4049 if (cap == IOMMU_CAP_CACHE_COHERENCY)
4050 return dmar_domain->iommu_snooping;
4051 if (cap == IOMMU_CAP_INTR_REMAP)
4052 return intr_remapping_enabled;
4053
4054 return 0;
4055}
4056
4057static struct iommu_ops intel_iommu_ops = {
4058 .domain_init = intel_iommu_domain_init,
4059 .domain_destroy = intel_iommu_domain_destroy,
4060 .attach_dev = intel_iommu_attach_device,
4061 .detach_dev = intel_iommu_detach_device,
4062 .map = intel_iommu_map,
4063 .unmap = intel_iommu_unmap,
4064 .iova_to_phys = intel_iommu_iova_to_phys,
4065 .domain_has_cap = intel_iommu_domain_has_cap,
4066};
4067
4068static void __devinit quirk_iommu_rwbf(struct pci_dev *dev)
4069{
4070 /*
4071 * Mobile 4 Series Chipset neglects to set RWBF capability,
4072 * but needs it:
4073 */
4074 printk(KERN_INFO "DMAR: Forcing write-buffer flush capability\n");
4075 rwbf_quirk = 1;
4076
4077 /* https://bugzilla.redhat.com/show_bug.cgi?id=538163 */
4078 if (dev->revision == 0x07) {
4079 printk(KERN_INFO "DMAR: Disabling IOMMU for graphics on this chipset\n");
4080 dmar_map_gfx = 0;
4081 }
4082}
4083
4084DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf);
4085
4086#define GGC 0x52
4087#define GGC_MEMORY_SIZE_MASK (0xf << 8)
4088#define GGC_MEMORY_SIZE_NONE (0x0 << 8)
4089#define GGC_MEMORY_SIZE_1M (0x1 << 8)
4090#define GGC_MEMORY_SIZE_2M (0x3 << 8)
4091#define GGC_MEMORY_VT_ENABLED (0x8 << 8)
4092#define GGC_MEMORY_SIZE_2M_VT (0x9 << 8)
4093#define GGC_MEMORY_SIZE_3M_VT (0xa << 8)
4094#define GGC_MEMORY_SIZE_4M_VT (0xb << 8)
4095
4096static void __devinit quirk_calpella_no_shadow_gtt(struct pci_dev *dev)
4097{
4098 unsigned short ggc;
4099
4100 if (pci_read_config_word(dev, GGC, &ggc))
4101 return;
4102
4103 if (!(ggc & GGC_MEMORY_VT_ENABLED)) {
4104 printk(KERN_INFO "DMAR: BIOS has allocated no shadow GTT; disabling IOMMU for graphics\n");
4105 dmar_map_gfx = 0;
4106 } else if (dmar_map_gfx) {
4107 /* we have to ensure the gfx device is idle before we flush */
4108 printk(KERN_INFO "DMAR: Disabling batched IOTLB flush on Ironlake\n");
4109 intel_iommu_strict = 1;
4110 }
4111}
4112DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0040, quirk_calpella_no_shadow_gtt);
4113DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0044, quirk_calpella_no_shadow_gtt);
4114DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0062, quirk_calpella_no_shadow_gtt);
4115DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x006a, quirk_calpella_no_shadow_gtt);
4116
4117/* On Tylersburg chipsets, some BIOSes have been known to enable the
4118 ISOCH DMAR unit for the Azalia sound device, but not give it any
4119 TLB entries, which causes it to deadlock. Check for that. We do
4120 this in a function called from init_dmars(), instead of in a PCI
4121 quirk, because we don't want to print the obnoxious "BIOS broken"
4122 message if VT-d is actually disabled.
4123*/
4124static void __init check_tylersburg_isoch(void)
4125{
4126 struct pci_dev *pdev;
4127 uint32_t vtisochctrl;
4128
4129 /* If there's no Azalia in the system anyway, forget it. */
4130 pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x3a3e, NULL);
4131 if (!pdev)
4132 return;
4133 pci_dev_put(pdev);
4134
4135 /* System Management Registers. Might be hidden, in which case
4136 we can't do the sanity check. But that's OK, because the
4137 known-broken BIOSes _don't_ actually hide it, so far. */
4138 pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x342e, NULL);
4139 if (!pdev)
4140 return;
4141
4142 if (pci_read_config_dword(pdev, 0x188, &vtisochctrl)) {
4143 pci_dev_put(pdev);
4144 return;
4145 }
4146
4147 pci_dev_put(pdev);
4148
4149 /* If Azalia DMA is routed to the non-isoch DMAR unit, fine. */
4150 if (vtisochctrl & 1)
4151 return;
4152
4153 /* Drop all bits other than the number of TLB entries */
4154 vtisochctrl &= 0x1c;
4155
4156 /* If we have the recommended number of TLB entries (16), fine. */
4157 if (vtisochctrl == 0x10)
4158 return;
4159
4160 /* Zero TLB entries? You get to ride the short bus to school. */
4161 if (!vtisochctrl) {
4162 WARN(1, "Your BIOS is broken; DMA routed to ISOCH DMAR unit but no TLB space.\n"
4163 "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
4164 dmi_get_system_info(DMI_BIOS_VENDOR),
4165 dmi_get_system_info(DMI_BIOS_VERSION),
4166 dmi_get_system_info(DMI_PRODUCT_VERSION));
4167 iommu_identity_mapping |= IDENTMAP_AZALIA;
4168 return;
4169 }
4170
4171 printk(KERN_WARNING "DMAR: Recommended TLB entries for ISOCH unit is 16; your BIOS set %d\n",
4172 vtisochctrl);
4173}
generated by cgit v1.2.2 (git 2.25.0) at 2025-01-01 08:58:49 -0500