Commit | Line | Data |
---|---|---|
ba395927 KA |
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 | * | |
98bcef56 | 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> | |
5b6985ce | 21 | * Author: Fenghua Yu <fenghua.yu@intel.com> |
ba395927 KA |
22 | */ |
23 | ||
24 | #include <linux/init.h> | |
25 | #include <linux/bitmap.h> | |
5e0d2a6f | 26 | #include <linux/debugfs.h> |
ba395927 KA |
27 | #include <linux/slab.h> |
28 | #include <linux/irq.h> | |
29 | #include <linux/interrupt.h> | |
ba395927 KA |
30 | #include <linux/spinlock.h> |
31 | #include <linux/pci.h> | |
32 | #include <linux/dmar.h> | |
33 | #include <linux/dma-mapping.h> | |
34 | #include <linux/mempool.h> | |
5e0d2a6f | 35 | #include <linux/timer.h> |
38717946 | 36 | #include <linux/iova.h> |
5d450806 | 37 | #include <linux/iommu.h> |
38717946 | 38 | #include <linux/intel-iommu.h> |
f59c7b69 | 39 | #include <linux/sysdev.h> |
ba395927 | 40 | #include <asm/cacheflush.h> |
46a7fa27 | 41 | #include <asm/iommu.h> |
ba395927 KA |
42 | #include "pci.h" |
43 | ||
5b6985ce FY |
44 | #define ROOT_SIZE VTD_PAGE_SIZE |
45 | #define CONTEXT_SIZE VTD_PAGE_SIZE | |
46 | ||
ba395927 KA |
47 | #define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY) |
48 | #define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA) | |
49 | ||
50 | #define IOAPIC_RANGE_START (0xfee00000) | |
51 | #define IOAPIC_RANGE_END (0xfeefffff) | |
52 | #define IOVA_START_ADDR (0x1000) | |
53 | ||
54 | #define DEFAULT_DOMAIN_ADDRESS_WIDTH 48 | |
55 | ||
ba395927 KA |
56 | #define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1) |
57 | ||
f27be03b | 58 | #define IOVA_PFN(addr) ((addr) >> PAGE_SHIFT) |
284901a9 | 59 | #define DMA_32BIT_PFN IOVA_PFN(DMA_BIT_MASK(32)) |
6a35528a | 60 | #define DMA_64BIT_PFN IOVA_PFN(DMA_BIT_MASK(64)) |
5e0d2a6f | 61 | |
d9630fe9 WH |
62 | /* global iommu list, set NULL for ignored DMAR units */ |
63 | static struct intel_iommu **g_iommus; | |
64 | ||
9af88143 DW |
65 | static int rwbf_quirk; |
66 | ||
46b08e1a MM |
67 | /* |
68 | * 0: Present | |
69 | * 1-11: Reserved | |
70 | * 12-63: Context Ptr (12 - (haw-1)) | |
71 | * 64-127: Reserved | |
72 | */ | |
73 | struct root_entry { | |
74 | u64 val; | |
75 | u64 rsvd1; | |
76 | }; | |
77 | #define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry)) | |
78 | static inline bool root_present(struct root_entry *root) | |
79 | { | |
80 | return (root->val & 1); | |
81 | } | |
82 | static inline void set_root_present(struct root_entry *root) | |
83 | { | |
84 | root->val |= 1; | |
85 | } | |
86 | static inline void set_root_value(struct root_entry *root, unsigned long value) | |
87 | { | |
88 | root->val |= value & VTD_PAGE_MASK; | |
89 | } | |
90 | ||
91 | static inline struct context_entry * | |
92 | get_context_addr_from_root(struct root_entry *root) | |
93 | { | |
94 | return (struct context_entry *) | |
95 | (root_present(root)?phys_to_virt( | |
96 | root->val & VTD_PAGE_MASK) : | |
97 | NULL); | |
98 | } | |
99 | ||
7a8fc25e MM |
100 | /* |
101 | * low 64 bits: | |
102 | * 0: present | |
103 | * 1: fault processing disable | |
104 | * 2-3: translation type | |
105 | * 12-63: address space root | |
106 | * high 64 bits: | |
107 | * 0-2: address width | |
108 | * 3-6: aval | |
109 | * 8-23: domain id | |
110 | */ | |
111 | struct context_entry { | |
112 | u64 lo; | |
113 | u64 hi; | |
114 | }; | |
c07e7d21 MM |
115 | |
116 | static inline bool context_present(struct context_entry *context) | |
117 | { | |
118 | return (context->lo & 1); | |
119 | } | |
120 | static inline void context_set_present(struct context_entry *context) | |
121 | { | |
122 | context->lo |= 1; | |
123 | } | |
124 | ||
125 | static inline void context_set_fault_enable(struct context_entry *context) | |
126 | { | |
127 | context->lo &= (((u64)-1) << 2) | 1; | |
128 | } | |
129 | ||
7a8fc25e | 130 | #define CONTEXT_TT_MULTI_LEVEL 0 |
c07e7d21 MM |
131 | |
132 | static inline void context_set_translation_type(struct context_entry *context, | |
133 | unsigned long value) | |
134 | { | |
135 | context->lo &= (((u64)-1) << 4) | 3; | |
136 | context->lo |= (value & 3) << 2; | |
137 | } | |
138 | ||
139 | static inline void context_set_address_root(struct context_entry *context, | |
140 | unsigned long value) | |
141 | { | |
142 | context->lo |= value & VTD_PAGE_MASK; | |
143 | } | |
144 | ||
145 | static inline void context_set_address_width(struct context_entry *context, | |
146 | unsigned long value) | |
147 | { | |
148 | context->hi |= value & 7; | |
149 | } | |
150 | ||
151 | static inline void context_set_domain_id(struct context_entry *context, | |
152 | unsigned long value) | |
153 | { | |
154 | context->hi |= (value & ((1 << 16) - 1)) << 8; | |
155 | } | |
156 | ||
157 | static inline void context_clear_entry(struct context_entry *context) | |
158 | { | |
159 | context->lo = 0; | |
160 | context->hi = 0; | |
161 | } | |
7a8fc25e | 162 | |
622ba12a MM |
163 | /* |
164 | * 0: readable | |
165 | * 1: writable | |
166 | * 2-6: reserved | |
167 | * 7: super page | |
9cf06697 SY |
168 | * 8-10: available |
169 | * 11: snoop behavior | |
622ba12a MM |
170 | * 12-63: Host physcial address |
171 | */ | |
172 | struct dma_pte { | |
173 | u64 val; | |
174 | }; | |
622ba12a | 175 | |
19c239ce MM |
176 | static inline void dma_clear_pte(struct dma_pte *pte) |
177 | { | |
178 | pte->val = 0; | |
179 | } | |
180 | ||
181 | static inline void dma_set_pte_readable(struct dma_pte *pte) | |
182 | { | |
183 | pte->val |= DMA_PTE_READ; | |
184 | } | |
185 | ||
186 | static inline void dma_set_pte_writable(struct dma_pte *pte) | |
187 | { | |
188 | pte->val |= DMA_PTE_WRITE; | |
189 | } | |
190 | ||
9cf06697 SY |
191 | static inline void dma_set_pte_snp(struct dma_pte *pte) |
192 | { | |
193 | pte->val |= DMA_PTE_SNP; | |
194 | } | |
195 | ||
19c239ce MM |
196 | static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot) |
197 | { | |
198 | pte->val = (pte->val & ~3) | (prot & 3); | |
199 | } | |
200 | ||
201 | static inline u64 dma_pte_addr(struct dma_pte *pte) | |
202 | { | |
203 | return (pte->val & VTD_PAGE_MASK); | |
204 | } | |
205 | ||
206 | static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr) | |
207 | { | |
208 | pte->val |= (addr & VTD_PAGE_MASK); | |
209 | } | |
210 | ||
211 | static inline bool dma_pte_present(struct dma_pte *pte) | |
212 | { | |
213 | return (pte->val & 3) != 0; | |
214 | } | |
622ba12a | 215 | |
3b5410e7 | 216 | /* devices under the same p2p bridge are owned in one domain */ |
cdc7b837 | 217 | #define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0) |
3b5410e7 | 218 | |
1ce28feb WH |
219 | /* domain represents a virtual machine, more than one devices |
220 | * across iommus may be owned in one domain, e.g. kvm guest. | |
221 | */ | |
222 | #define DOMAIN_FLAG_VIRTUAL_MACHINE (1 << 1) | |
223 | ||
99126f7c MM |
224 | struct dmar_domain { |
225 | int id; /* domain id */ | |
8c11e798 | 226 | unsigned long iommu_bmp; /* bitmap of iommus this domain uses*/ |
99126f7c MM |
227 | |
228 | struct list_head devices; /* all devices' list */ | |
229 | struct iova_domain iovad; /* iova's that belong to this domain */ | |
230 | ||
231 | struct dma_pte *pgd; /* virtual address */ | |
232 | spinlock_t mapping_lock; /* page table lock */ | |
233 | int gaw; /* max guest address width */ | |
234 | ||
235 | /* adjusted guest address width, 0 is level 2 30-bit */ | |
236 | int agaw; | |
237 | ||
3b5410e7 | 238 | int flags; /* flags to find out type of domain */ |
8e604097 WH |
239 | |
240 | int iommu_coherency;/* indicate coherency of iommu access */ | |
58c610bd | 241 | int iommu_snooping; /* indicate snooping control feature*/ |
c7151a8d WH |
242 | int iommu_count; /* reference count of iommu */ |
243 | spinlock_t iommu_lock; /* protect iommu set in domain */ | |
fe40f1e0 | 244 | u64 max_addr; /* maximum mapped address */ |
99126f7c MM |
245 | }; |
246 | ||
a647dacb MM |
247 | /* PCI domain-device relationship */ |
248 | struct device_domain_info { | |
249 | struct list_head link; /* link to domain siblings */ | |
250 | struct list_head global; /* link to global list */ | |
276dbf99 DW |
251 | int segment; /* PCI domain */ |
252 | u8 bus; /* PCI bus number */ | |
a647dacb MM |
253 | u8 devfn; /* PCI devfn number */ |
254 | struct pci_dev *dev; /* it's NULL for PCIE-to-PCI bridge */ | |
255 | struct dmar_domain *domain; /* pointer to domain */ | |
256 | }; | |
257 | ||
5e0d2a6f | 258 | static void flush_unmaps_timeout(unsigned long data); |
259 | ||
260 | DEFINE_TIMER(unmap_timer, flush_unmaps_timeout, 0, 0); | |
261 | ||
80b20dd8 | 262 | #define HIGH_WATER_MARK 250 |
263 | struct deferred_flush_tables { | |
264 | int next; | |
265 | struct iova *iova[HIGH_WATER_MARK]; | |
266 | struct dmar_domain *domain[HIGH_WATER_MARK]; | |
267 | }; | |
268 | ||
269 | static struct deferred_flush_tables *deferred_flush; | |
270 | ||
5e0d2a6f | 271 | /* bitmap for indexing intel_iommus */ |
5e0d2a6f | 272 | static int g_num_of_iommus; |
273 | ||
274 | static DEFINE_SPINLOCK(async_umap_flush_lock); | |
275 | static LIST_HEAD(unmaps_to_do); | |
276 | ||
277 | static int timer_on; | |
278 | static long list_size; | |
5e0d2a6f | 279 | |
ba395927 KA |
280 | static void domain_remove_dev_info(struct dmar_domain *domain); |
281 | ||
0cd5c3c8 KM |
282 | #ifdef CONFIG_DMAR_DEFAULT_ON |
283 | int dmar_disabled = 0; | |
284 | #else | |
285 | int dmar_disabled = 1; | |
286 | #endif /*CONFIG_DMAR_DEFAULT_ON*/ | |
287 | ||
ba395927 | 288 | static int __initdata dmar_map_gfx = 1; |
7d3b03ce | 289 | static int dmar_forcedac; |
5e0d2a6f | 290 | static int intel_iommu_strict; |
ba395927 KA |
291 | |
292 | #define DUMMY_DEVICE_DOMAIN_INFO ((struct device_domain_info *)(-1)) | |
293 | static DEFINE_SPINLOCK(device_domain_lock); | |
294 | static LIST_HEAD(device_domain_list); | |
295 | ||
a8bcbb0d JR |
296 | static struct iommu_ops intel_iommu_ops; |
297 | ||
ba395927 KA |
298 | static int __init intel_iommu_setup(char *str) |
299 | { | |
300 | if (!str) | |
301 | return -EINVAL; | |
302 | while (*str) { | |
0cd5c3c8 KM |
303 | if (!strncmp(str, "on", 2)) { |
304 | dmar_disabled = 0; | |
305 | printk(KERN_INFO "Intel-IOMMU: enabled\n"); | |
306 | } else if (!strncmp(str, "off", 3)) { | |
ba395927 | 307 | dmar_disabled = 1; |
0cd5c3c8 | 308 | printk(KERN_INFO "Intel-IOMMU: disabled\n"); |
ba395927 KA |
309 | } else if (!strncmp(str, "igfx_off", 8)) { |
310 | dmar_map_gfx = 0; | |
311 | printk(KERN_INFO | |
312 | "Intel-IOMMU: disable GFX device mapping\n"); | |
7d3b03ce | 313 | } else if (!strncmp(str, "forcedac", 8)) { |
5e0d2a6f | 314 | printk(KERN_INFO |
7d3b03ce KA |
315 | "Intel-IOMMU: Forcing DAC for PCI devices\n"); |
316 | dmar_forcedac = 1; | |
5e0d2a6f | 317 | } else if (!strncmp(str, "strict", 6)) { |
318 | printk(KERN_INFO | |
319 | "Intel-IOMMU: disable batched IOTLB flush\n"); | |
320 | intel_iommu_strict = 1; | |
ba395927 KA |
321 | } |
322 | ||
323 | str += strcspn(str, ","); | |
324 | while (*str == ',') | |
325 | str++; | |
326 | } | |
327 | return 0; | |
328 | } | |
329 | __setup("intel_iommu=", intel_iommu_setup); | |
330 | ||
331 | static struct kmem_cache *iommu_domain_cache; | |
332 | static struct kmem_cache *iommu_devinfo_cache; | |
333 | static struct kmem_cache *iommu_iova_cache; | |
334 | ||
eb3fa7cb KA |
335 | static inline void *iommu_kmem_cache_alloc(struct kmem_cache *cachep) |
336 | { | |
337 | unsigned int flags; | |
338 | void *vaddr; | |
339 | ||
340 | /* trying to avoid low memory issues */ | |
341 | flags = current->flags & PF_MEMALLOC; | |
342 | current->flags |= PF_MEMALLOC; | |
343 | vaddr = kmem_cache_alloc(cachep, GFP_ATOMIC); | |
344 | current->flags &= (~PF_MEMALLOC | flags); | |
345 | return vaddr; | |
346 | } | |
347 | ||
348 | ||
ba395927 KA |
349 | static inline void *alloc_pgtable_page(void) |
350 | { | |
eb3fa7cb KA |
351 | unsigned int flags; |
352 | void *vaddr; | |
353 | ||
354 | /* trying to avoid low memory issues */ | |
355 | flags = current->flags & PF_MEMALLOC; | |
356 | current->flags |= PF_MEMALLOC; | |
357 | vaddr = (void *)get_zeroed_page(GFP_ATOMIC); | |
358 | current->flags &= (~PF_MEMALLOC | flags); | |
359 | return vaddr; | |
ba395927 KA |
360 | } |
361 | ||
362 | static inline void free_pgtable_page(void *vaddr) | |
363 | { | |
364 | free_page((unsigned long)vaddr); | |
365 | } | |
366 | ||
367 | static inline void *alloc_domain_mem(void) | |
368 | { | |
eb3fa7cb | 369 | return iommu_kmem_cache_alloc(iommu_domain_cache); |
ba395927 KA |
370 | } |
371 | ||
38717946 | 372 | static void free_domain_mem(void *vaddr) |
ba395927 KA |
373 | { |
374 | kmem_cache_free(iommu_domain_cache, vaddr); | |
375 | } | |
376 | ||
377 | static inline void * alloc_devinfo_mem(void) | |
378 | { | |
eb3fa7cb | 379 | return iommu_kmem_cache_alloc(iommu_devinfo_cache); |
ba395927 KA |
380 | } |
381 | ||
382 | static inline void free_devinfo_mem(void *vaddr) | |
383 | { | |
384 | kmem_cache_free(iommu_devinfo_cache, vaddr); | |
385 | } | |
386 | ||
387 | struct iova *alloc_iova_mem(void) | |
388 | { | |
eb3fa7cb | 389 | return iommu_kmem_cache_alloc(iommu_iova_cache); |
ba395927 KA |
390 | } |
391 | ||
392 | void free_iova_mem(struct iova *iova) | |
393 | { | |
394 | kmem_cache_free(iommu_iova_cache, iova); | |
395 | } | |
396 | ||
1b573683 WH |
397 | |
398 | static inline int width_to_agaw(int width); | |
399 | ||
400 | /* calculate agaw for each iommu. | |
401 | * "SAGAW" may be different across iommus, use a default agaw, and | |
402 | * get a supported less agaw for iommus that don't support the default agaw. | |
403 | */ | |
404 | int iommu_calculate_agaw(struct intel_iommu *iommu) | |
405 | { | |
406 | unsigned long sagaw; | |
407 | int agaw = -1; | |
408 | ||
409 | sagaw = cap_sagaw(iommu->cap); | |
410 | for (agaw = width_to_agaw(DEFAULT_DOMAIN_ADDRESS_WIDTH); | |
411 | agaw >= 0; agaw--) { | |
412 | if (test_bit(agaw, &sagaw)) | |
413 | break; | |
414 | } | |
415 | ||
416 | return agaw; | |
417 | } | |
418 | ||
8c11e798 WH |
419 | /* in native case, each domain is related to only one iommu */ |
420 | static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain) | |
421 | { | |
422 | int iommu_id; | |
423 | ||
1ce28feb WH |
424 | BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE); |
425 | ||
8c11e798 WH |
426 | iommu_id = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); |
427 | if (iommu_id < 0 || iommu_id >= g_num_of_iommus) | |
428 | return NULL; | |
429 | ||
430 | return g_iommus[iommu_id]; | |
431 | } | |
432 | ||
8e604097 WH |
433 | static void domain_update_iommu_coherency(struct dmar_domain *domain) |
434 | { | |
435 | int i; | |
436 | ||
437 | domain->iommu_coherency = 1; | |
438 | ||
439 | i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); | |
440 | for (; i < g_num_of_iommus; ) { | |
441 | if (!ecap_coherent(g_iommus[i]->ecap)) { | |
442 | domain->iommu_coherency = 0; | |
443 | break; | |
444 | } | |
445 | i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1); | |
446 | } | |
447 | } | |
448 | ||
58c610bd SY |
449 | static void domain_update_iommu_snooping(struct dmar_domain *domain) |
450 | { | |
451 | int i; | |
452 | ||
453 | domain->iommu_snooping = 1; | |
454 | ||
455 | i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); | |
456 | for (; i < g_num_of_iommus; ) { | |
457 | if (!ecap_sc_support(g_iommus[i]->ecap)) { | |
458 | domain->iommu_snooping = 0; | |
459 | break; | |
460 | } | |
461 | i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1); | |
462 | } | |
463 | } | |
464 | ||
465 | /* Some capabilities may be different across iommus */ | |
466 | static void domain_update_iommu_cap(struct dmar_domain *domain) | |
467 | { | |
468 | domain_update_iommu_coherency(domain); | |
469 | domain_update_iommu_snooping(domain); | |
470 | } | |
471 | ||
276dbf99 | 472 | static struct intel_iommu *device_to_iommu(int segment, u8 bus, u8 devfn) |
c7151a8d WH |
473 | { |
474 | struct dmar_drhd_unit *drhd = NULL; | |
475 | int i; | |
476 | ||
477 | for_each_drhd_unit(drhd) { | |
478 | if (drhd->ignored) | |
479 | continue; | |
276dbf99 DW |
480 | if (segment != drhd->segment) |
481 | continue; | |
c7151a8d | 482 | |
924b6231 | 483 | for (i = 0; i < drhd->devices_cnt; i++) { |
288e4877 DH |
484 | if (drhd->devices[i] && |
485 | drhd->devices[i]->bus->number == bus && | |
c7151a8d WH |
486 | drhd->devices[i]->devfn == devfn) |
487 | return drhd->iommu; | |
4958c5dc DW |
488 | if (drhd->devices[i] && |
489 | drhd->devices[i]->subordinate && | |
924b6231 DW |
490 | drhd->devices[i]->subordinate->number <= bus && |
491 | drhd->devices[i]->subordinate->subordinate >= bus) | |
492 | return drhd->iommu; | |
493 | } | |
c7151a8d WH |
494 | |
495 | if (drhd->include_all) | |
496 | return drhd->iommu; | |
497 | } | |
498 | ||
499 | return NULL; | |
500 | } | |
501 | ||
5331fe6f WH |
502 | static void domain_flush_cache(struct dmar_domain *domain, |
503 | void *addr, int size) | |
504 | { | |
505 | if (!domain->iommu_coherency) | |
506 | clflush_cache_range(addr, size); | |
507 | } | |
508 | ||
ba395927 KA |
509 | /* Gets context entry for a given bus and devfn */ |
510 | static struct context_entry * device_to_context_entry(struct intel_iommu *iommu, | |
511 | u8 bus, u8 devfn) | |
512 | { | |
513 | struct root_entry *root; | |
514 | struct context_entry *context; | |
515 | unsigned long phy_addr; | |
516 | unsigned long flags; | |
517 | ||
518 | spin_lock_irqsave(&iommu->lock, flags); | |
519 | root = &iommu->root_entry[bus]; | |
520 | context = get_context_addr_from_root(root); | |
521 | if (!context) { | |
522 | context = (struct context_entry *)alloc_pgtable_page(); | |
523 | if (!context) { | |
524 | spin_unlock_irqrestore(&iommu->lock, flags); | |
525 | return NULL; | |
526 | } | |
5b6985ce | 527 | __iommu_flush_cache(iommu, (void *)context, CONTEXT_SIZE); |
ba395927 KA |
528 | phy_addr = virt_to_phys((void *)context); |
529 | set_root_value(root, phy_addr); | |
530 | set_root_present(root); | |
531 | __iommu_flush_cache(iommu, root, sizeof(*root)); | |
532 | } | |
533 | spin_unlock_irqrestore(&iommu->lock, flags); | |
534 | return &context[devfn]; | |
535 | } | |
536 | ||
537 | static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn) | |
538 | { | |
539 | struct root_entry *root; | |
540 | struct context_entry *context; | |
541 | int ret; | |
542 | unsigned long flags; | |
543 | ||
544 | spin_lock_irqsave(&iommu->lock, flags); | |
545 | root = &iommu->root_entry[bus]; | |
546 | context = get_context_addr_from_root(root); | |
547 | if (!context) { | |
548 | ret = 0; | |
549 | goto out; | |
550 | } | |
c07e7d21 | 551 | ret = context_present(&context[devfn]); |
ba395927 KA |
552 | out: |
553 | spin_unlock_irqrestore(&iommu->lock, flags); | |
554 | return ret; | |
555 | } | |
556 | ||
557 | static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn) | |
558 | { | |
559 | struct root_entry *root; | |
560 | struct context_entry *context; | |
561 | unsigned long flags; | |
562 | ||
563 | spin_lock_irqsave(&iommu->lock, flags); | |
564 | root = &iommu->root_entry[bus]; | |
565 | context = get_context_addr_from_root(root); | |
566 | if (context) { | |
c07e7d21 | 567 | context_clear_entry(&context[devfn]); |
ba395927 KA |
568 | __iommu_flush_cache(iommu, &context[devfn], \ |
569 | sizeof(*context)); | |
570 | } | |
571 | spin_unlock_irqrestore(&iommu->lock, flags); | |
572 | } | |
573 | ||
574 | static void free_context_table(struct intel_iommu *iommu) | |
575 | { | |
576 | struct root_entry *root; | |
577 | int i; | |
578 | unsigned long flags; | |
579 | struct context_entry *context; | |
580 | ||
581 | spin_lock_irqsave(&iommu->lock, flags); | |
582 | if (!iommu->root_entry) { | |
583 | goto out; | |
584 | } | |
585 | for (i = 0; i < ROOT_ENTRY_NR; i++) { | |
586 | root = &iommu->root_entry[i]; | |
587 | context = get_context_addr_from_root(root); | |
588 | if (context) | |
589 | free_pgtable_page(context); | |
590 | } | |
591 | free_pgtable_page(iommu->root_entry); | |
592 | iommu->root_entry = NULL; | |
593 | out: | |
594 | spin_unlock_irqrestore(&iommu->lock, flags); | |
595 | } | |
596 | ||
597 | /* page table handling */ | |
598 | #define LEVEL_STRIDE (9) | |
599 | #define LEVEL_MASK (((u64)1 << LEVEL_STRIDE) - 1) | |
600 | ||
601 | static inline int agaw_to_level(int agaw) | |
602 | { | |
603 | return agaw + 2; | |
604 | } | |
605 | ||
606 | static inline int agaw_to_width(int agaw) | |
607 | { | |
608 | return 30 + agaw * LEVEL_STRIDE; | |
609 | ||
610 | } | |
611 | ||
612 | static inline int width_to_agaw(int width) | |
613 | { | |
614 | return (width - 30) / LEVEL_STRIDE; | |
615 | } | |
616 | ||
617 | static inline unsigned int level_to_offset_bits(int level) | |
618 | { | |
619 | return (12 + (level - 1) * LEVEL_STRIDE); | |
620 | } | |
621 | ||
622 | static inline int address_level_offset(u64 addr, int level) | |
623 | { | |
624 | return ((addr >> level_to_offset_bits(level)) & LEVEL_MASK); | |
625 | } | |
626 | ||
627 | static inline u64 level_mask(int level) | |
628 | { | |
629 | return ((u64)-1 << level_to_offset_bits(level)); | |
630 | } | |
631 | ||
632 | static inline u64 level_size(int level) | |
633 | { | |
634 | return ((u64)1 << level_to_offset_bits(level)); | |
635 | } | |
636 | ||
637 | static inline u64 align_to_level(u64 addr, int level) | |
638 | { | |
639 | return ((addr + level_size(level) - 1) & level_mask(level)); | |
640 | } | |
641 | ||
642 | static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr) | |
643 | { | |
644 | int addr_width = agaw_to_width(domain->agaw); | |
645 | struct dma_pte *parent, *pte = NULL; | |
646 | int level = agaw_to_level(domain->agaw); | |
647 | int offset; | |
648 | unsigned long flags; | |
649 | ||
650 | BUG_ON(!domain->pgd); | |
651 | ||
652 | addr &= (((u64)1) << addr_width) - 1; | |
653 | parent = domain->pgd; | |
654 | ||
655 | spin_lock_irqsave(&domain->mapping_lock, flags); | |
656 | while (level > 0) { | |
657 | void *tmp_page; | |
658 | ||
659 | offset = address_level_offset(addr, level); | |
660 | pte = &parent[offset]; | |
661 | if (level == 1) | |
662 | break; | |
663 | ||
19c239ce | 664 | if (!dma_pte_present(pte)) { |
ba395927 KA |
665 | tmp_page = alloc_pgtable_page(); |
666 | ||
667 | if (!tmp_page) { | |
668 | spin_unlock_irqrestore(&domain->mapping_lock, | |
669 | flags); | |
670 | return NULL; | |
671 | } | |
5331fe6f | 672 | domain_flush_cache(domain, tmp_page, PAGE_SIZE); |
19c239ce | 673 | dma_set_pte_addr(pte, virt_to_phys(tmp_page)); |
ba395927 KA |
674 | /* |
675 | * high level table always sets r/w, last level page | |
676 | * table control read/write | |
677 | */ | |
19c239ce MM |
678 | dma_set_pte_readable(pte); |
679 | dma_set_pte_writable(pte); | |
5331fe6f | 680 | domain_flush_cache(domain, pte, sizeof(*pte)); |
ba395927 | 681 | } |
19c239ce | 682 | parent = phys_to_virt(dma_pte_addr(pte)); |
ba395927 KA |
683 | level--; |
684 | } | |
685 | ||
686 | spin_unlock_irqrestore(&domain->mapping_lock, flags); | |
687 | return pte; | |
688 | } | |
689 | ||
690 | /* return address's pte at specific level */ | |
691 | static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr, | |
692 | int level) | |
693 | { | |
694 | struct dma_pte *parent, *pte = NULL; | |
695 | int total = agaw_to_level(domain->agaw); | |
696 | int offset; | |
697 | ||
698 | parent = domain->pgd; | |
699 | while (level <= total) { | |
700 | offset = address_level_offset(addr, total); | |
701 | pte = &parent[offset]; | |
702 | if (level == total) | |
703 | return pte; | |
704 | ||
19c239ce | 705 | if (!dma_pte_present(pte)) |
ba395927 | 706 | break; |
19c239ce | 707 | parent = phys_to_virt(dma_pte_addr(pte)); |
ba395927 KA |
708 | total--; |
709 | } | |
710 | return NULL; | |
711 | } | |
712 | ||
713 | /* clear one page's page table */ | |
714 | static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr) | |
715 | { | |
716 | struct dma_pte *pte = NULL; | |
717 | ||
718 | /* get last level pte */ | |
719 | pte = dma_addr_level_pte(domain, addr, 1); | |
720 | ||
721 | if (pte) { | |
19c239ce | 722 | dma_clear_pte(pte); |
5331fe6f | 723 | domain_flush_cache(domain, pte, sizeof(*pte)); |
ba395927 KA |
724 | } |
725 | } | |
726 | ||
727 | /* clear last level pte, a tlb flush should be followed */ | |
728 | static void dma_pte_clear_range(struct dmar_domain *domain, u64 start, u64 end) | |
729 | { | |
730 | int addr_width = agaw_to_width(domain->agaw); | |
afeeb7ce | 731 | int npages; |
ba395927 KA |
732 | |
733 | start &= (((u64)1) << addr_width) - 1; | |
734 | end &= (((u64)1) << addr_width) - 1; | |
735 | /* in case it's partial page */ | |
31d3568d FY |
736 | start &= PAGE_MASK; |
737 | end = PAGE_ALIGN(end); | |
afeeb7ce | 738 | npages = (end - start) / VTD_PAGE_SIZE; |
ba395927 KA |
739 | |
740 | /* we don't need lock here, nobody else touches the iova range */ | |
afeeb7ce | 741 | while (npages--) { |
ba395927 | 742 | dma_pte_clear_one(domain, start); |
5b6985ce | 743 | start += VTD_PAGE_SIZE; |
ba395927 KA |
744 | } |
745 | } | |
746 | ||
747 | /* free page table pages. last level pte should already be cleared */ | |
748 | static void dma_pte_free_pagetable(struct dmar_domain *domain, | |
749 | u64 start, u64 end) | |
750 | { | |
751 | int addr_width = agaw_to_width(domain->agaw); | |
752 | struct dma_pte *pte; | |
753 | int total = agaw_to_level(domain->agaw); | |
754 | int level; | |
755 | u64 tmp; | |
756 | ||
757 | start &= (((u64)1) << addr_width) - 1; | |
758 | end &= (((u64)1) << addr_width) - 1; | |
759 | ||
760 | /* we don't need lock here, nobody else touches the iova range */ | |
761 | level = 2; | |
762 | while (level <= total) { | |
763 | tmp = align_to_level(start, level); | |
764 | if (tmp >= end || (tmp + level_size(level) > end)) | |
765 | return; | |
766 | ||
767 | while (tmp < end) { | |
768 | pte = dma_addr_level_pte(domain, tmp, level); | |
769 | if (pte) { | |
770 | free_pgtable_page( | |
19c239ce MM |
771 | phys_to_virt(dma_pte_addr(pte))); |
772 | dma_clear_pte(pte); | |
5331fe6f | 773 | domain_flush_cache(domain, pte, sizeof(*pte)); |
ba395927 KA |
774 | } |
775 | tmp += level_size(level); | |
776 | } | |
777 | level++; | |
778 | } | |
779 | /* free pgd */ | |
780 | if (start == 0 && end >= ((((u64)1) << addr_width) - 1)) { | |
781 | free_pgtable_page(domain->pgd); | |
782 | domain->pgd = NULL; | |
783 | } | |
784 | } | |
785 | ||
786 | /* iommu handling */ | |
787 | static int iommu_alloc_root_entry(struct intel_iommu *iommu) | |
788 | { | |
789 | struct root_entry *root; | |
790 | unsigned long flags; | |
791 | ||
792 | root = (struct root_entry *)alloc_pgtable_page(); | |
793 | if (!root) | |
794 | return -ENOMEM; | |
795 | ||
5b6985ce | 796 | __iommu_flush_cache(iommu, root, ROOT_SIZE); |
ba395927 KA |
797 | |
798 | spin_lock_irqsave(&iommu->lock, flags); | |
799 | iommu->root_entry = root; | |
800 | spin_unlock_irqrestore(&iommu->lock, flags); | |
801 | ||
802 | return 0; | |
803 | } | |
804 | ||
ba395927 KA |
805 | static void iommu_set_root_entry(struct intel_iommu *iommu) |
806 | { | |
807 | void *addr; | |
808 | u32 cmd, sts; | |
809 | unsigned long flag; | |
810 | ||
811 | addr = iommu->root_entry; | |
812 | ||
813 | spin_lock_irqsave(&iommu->register_lock, flag); | |
814 | dmar_writeq(iommu->reg + DMAR_RTADDR_REG, virt_to_phys(addr)); | |
815 | ||
816 | cmd = iommu->gcmd | DMA_GCMD_SRTP; | |
817 | writel(cmd, iommu->reg + DMAR_GCMD_REG); | |
818 | ||
819 | /* Make sure hardware complete it */ | |
820 | IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, | |
821 | readl, (sts & DMA_GSTS_RTPS), sts); | |
822 | ||
823 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
824 | } | |
825 | ||
826 | static void iommu_flush_write_buffer(struct intel_iommu *iommu) | |
827 | { | |
828 | u32 val; | |
829 | unsigned long flag; | |
830 | ||
9af88143 | 831 | if (!rwbf_quirk && !cap_rwbf(iommu->cap)) |
ba395927 KA |
832 | return; |
833 | val = iommu->gcmd | DMA_GCMD_WBF; | |
834 | ||
835 | spin_lock_irqsave(&iommu->register_lock, flag); | |
836 | writel(val, iommu->reg + DMAR_GCMD_REG); | |
837 | ||
838 | /* Make sure hardware complete it */ | |
839 | IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, | |
840 | readl, (!(val & DMA_GSTS_WBFS)), val); | |
841 | ||
842 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
843 | } | |
844 | ||
845 | /* return value determine if we need a write buffer flush */ | |
846 | static int __iommu_flush_context(struct intel_iommu *iommu, | |
847 | u16 did, u16 source_id, u8 function_mask, u64 type, | |
848 | int non_present_entry_flush) | |
849 | { | |
850 | u64 val = 0; | |
851 | unsigned long flag; | |
852 | ||
853 | /* | |
854 | * In the non-present entry flush case, if hardware doesn't cache | |
855 | * non-present entry we do nothing and if hardware cache non-present | |
856 | * entry, we flush entries of domain 0 (the domain id is used to cache | |
857 | * any non-present entries) | |
858 | */ | |
859 | if (non_present_entry_flush) { | |
860 | if (!cap_caching_mode(iommu->cap)) | |
861 | return 1; | |
862 | else | |
863 | did = 0; | |
864 | } | |
865 | ||
866 | switch (type) { | |
867 | case DMA_CCMD_GLOBAL_INVL: | |
868 | val = DMA_CCMD_GLOBAL_INVL; | |
869 | break; | |
870 | case DMA_CCMD_DOMAIN_INVL: | |
871 | val = DMA_CCMD_DOMAIN_INVL|DMA_CCMD_DID(did); | |
872 | break; | |
873 | case DMA_CCMD_DEVICE_INVL: | |
874 | val = DMA_CCMD_DEVICE_INVL|DMA_CCMD_DID(did) | |
875 | | DMA_CCMD_SID(source_id) | DMA_CCMD_FM(function_mask); | |
876 | break; | |
877 | default: | |
878 | BUG(); | |
879 | } | |
880 | val |= DMA_CCMD_ICC; | |
881 | ||
882 | spin_lock_irqsave(&iommu->register_lock, flag); | |
883 | dmar_writeq(iommu->reg + DMAR_CCMD_REG, val); | |
884 | ||
885 | /* Make sure hardware complete it */ | |
886 | IOMMU_WAIT_OP(iommu, DMAR_CCMD_REG, | |
887 | dmar_readq, (!(val & DMA_CCMD_ICC)), val); | |
888 | ||
889 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
890 | ||
4d235ba6 | 891 | /* flush context entry will implicitly flush write buffer */ |
ba395927 KA |
892 | return 0; |
893 | } | |
894 | ||
ba395927 KA |
895 | /* return value determine if we need a write buffer flush */ |
896 | static int __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did, | |
897 | u64 addr, unsigned int size_order, u64 type, | |
898 | int non_present_entry_flush) | |
899 | { | |
900 | int tlb_offset = ecap_iotlb_offset(iommu->ecap); | |
901 | u64 val = 0, val_iva = 0; | |
902 | unsigned long flag; | |
903 | ||
904 | /* | |
905 | * In the non-present entry flush case, if hardware doesn't cache | |
906 | * non-present entry we do nothing and if hardware cache non-present | |
907 | * entry, we flush entries of domain 0 (the domain id is used to cache | |
908 | * any non-present entries) | |
909 | */ | |
910 | if (non_present_entry_flush) { | |
911 | if (!cap_caching_mode(iommu->cap)) | |
912 | return 1; | |
913 | else | |
914 | did = 0; | |
915 | } | |
916 | ||
917 | switch (type) { | |
918 | case DMA_TLB_GLOBAL_FLUSH: | |
919 | /* global flush doesn't need set IVA_REG */ | |
920 | val = DMA_TLB_GLOBAL_FLUSH|DMA_TLB_IVT; | |
921 | break; | |
922 | case DMA_TLB_DSI_FLUSH: | |
923 | val = DMA_TLB_DSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did); | |
924 | break; | |
925 | case DMA_TLB_PSI_FLUSH: | |
926 | val = DMA_TLB_PSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did); | |
927 | /* Note: always flush non-leaf currently */ | |
928 | val_iva = size_order | addr; | |
929 | break; | |
930 | default: | |
931 | BUG(); | |
932 | } | |
933 | /* Note: set drain read/write */ | |
934 | #if 0 | |
935 | /* | |
936 | * This is probably to be super secure.. Looks like we can | |
937 | * ignore it without any impact. | |
938 | */ | |
939 | if (cap_read_drain(iommu->cap)) | |
940 | val |= DMA_TLB_READ_DRAIN; | |
941 | #endif | |
942 | if (cap_write_drain(iommu->cap)) | |
943 | val |= DMA_TLB_WRITE_DRAIN; | |
944 | ||
945 | spin_lock_irqsave(&iommu->register_lock, flag); | |
946 | /* Note: Only uses first TLB reg currently */ | |
947 | if (val_iva) | |
948 | dmar_writeq(iommu->reg + tlb_offset, val_iva); | |
949 | dmar_writeq(iommu->reg + tlb_offset + 8, val); | |
950 | ||
951 | /* Make sure hardware complete it */ | |
952 | IOMMU_WAIT_OP(iommu, tlb_offset + 8, | |
953 | dmar_readq, (!(val & DMA_TLB_IVT)), val); | |
954 | ||
955 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
956 | ||
957 | /* check IOTLB invalidation granularity */ | |
958 | if (DMA_TLB_IAIG(val) == 0) | |
959 | printk(KERN_ERR"IOMMU: flush IOTLB failed\n"); | |
960 | if (DMA_TLB_IAIG(val) != DMA_TLB_IIRG(type)) | |
961 | pr_debug("IOMMU: tlb flush request %Lx, actual %Lx\n", | |
5b6985ce FY |
962 | (unsigned long long)DMA_TLB_IIRG(type), |
963 | (unsigned long long)DMA_TLB_IAIG(val)); | |
4d235ba6 | 964 | /* flush iotlb entry will implicitly flush write buffer */ |
ba395927 KA |
965 | return 0; |
966 | } | |
967 | ||
ba395927 KA |
968 | static int iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did, |
969 | u64 addr, unsigned int pages, int non_present_entry_flush) | |
970 | { | |
f76aec76 | 971 | unsigned int mask; |
ba395927 | 972 | |
5b6985ce | 973 | BUG_ON(addr & (~VTD_PAGE_MASK)); |
ba395927 KA |
974 | BUG_ON(pages == 0); |
975 | ||
976 | /* Fallback to domain selective flush if no PSI support */ | |
977 | if (!cap_pgsel_inv(iommu->cap)) | |
a77b67d4 YS |
978 | return iommu->flush.flush_iotlb(iommu, did, 0, 0, |
979 | DMA_TLB_DSI_FLUSH, | |
980 | non_present_entry_flush); | |
ba395927 KA |
981 | |
982 | /* | |
983 | * PSI requires page size to be 2 ^ x, and the base address is naturally | |
984 | * aligned to the size | |
985 | */ | |
f76aec76 | 986 | mask = ilog2(__roundup_pow_of_two(pages)); |
ba395927 | 987 | /* Fallback to domain selective flush if size is too big */ |
f76aec76 | 988 | if (mask > cap_max_amask_val(iommu->cap)) |
a77b67d4 YS |
989 | return iommu->flush.flush_iotlb(iommu, did, 0, 0, |
990 | DMA_TLB_DSI_FLUSH, non_present_entry_flush); | |
ba395927 | 991 | |
a77b67d4 YS |
992 | return iommu->flush.flush_iotlb(iommu, did, addr, mask, |
993 | DMA_TLB_PSI_FLUSH, | |
994 | non_present_entry_flush); | |
ba395927 KA |
995 | } |
996 | ||
f8bab735 | 997 | static void iommu_disable_protect_mem_regions(struct intel_iommu *iommu) |
998 | { | |
999 | u32 pmen; | |
1000 | unsigned long flags; | |
1001 | ||
1002 | spin_lock_irqsave(&iommu->register_lock, flags); | |
1003 | pmen = readl(iommu->reg + DMAR_PMEN_REG); | |
1004 | pmen &= ~DMA_PMEN_EPM; | |
1005 | writel(pmen, iommu->reg + DMAR_PMEN_REG); | |
1006 | ||
1007 | /* wait for the protected region status bit to clear */ | |
1008 | IOMMU_WAIT_OP(iommu, DMAR_PMEN_REG, | |
1009 | readl, !(pmen & DMA_PMEN_PRS), pmen); | |
1010 | ||
1011 | spin_unlock_irqrestore(&iommu->register_lock, flags); | |
1012 | } | |
1013 | ||
ba395927 KA |
1014 | static int iommu_enable_translation(struct intel_iommu *iommu) |
1015 | { | |
1016 | u32 sts; | |
1017 | unsigned long flags; | |
1018 | ||
1019 | spin_lock_irqsave(&iommu->register_lock, flags); | |
1020 | writel(iommu->gcmd|DMA_GCMD_TE, iommu->reg + DMAR_GCMD_REG); | |
1021 | ||
1022 | /* Make sure hardware complete it */ | |
1023 | IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, | |
1024 | readl, (sts & DMA_GSTS_TES), sts); | |
1025 | ||
1026 | iommu->gcmd |= DMA_GCMD_TE; | |
1027 | spin_unlock_irqrestore(&iommu->register_lock, flags); | |
1028 | return 0; | |
1029 | } | |
1030 | ||
1031 | static int iommu_disable_translation(struct intel_iommu *iommu) | |
1032 | { | |
1033 | u32 sts; | |
1034 | unsigned long flag; | |
1035 | ||
1036 | spin_lock_irqsave(&iommu->register_lock, flag); | |
1037 | iommu->gcmd &= ~DMA_GCMD_TE; | |
1038 | writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG); | |
1039 | ||
1040 | /* Make sure hardware complete it */ | |
1041 | IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG, | |
1042 | readl, (!(sts & DMA_GSTS_TES)), sts); | |
1043 | ||
1044 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
1045 | return 0; | |
1046 | } | |
1047 | ||
3460a6d9 | 1048 | |
ba395927 KA |
1049 | static int iommu_init_domains(struct intel_iommu *iommu) |
1050 | { | |
1051 | unsigned long ndomains; | |
1052 | unsigned long nlongs; | |
1053 | ||
1054 | ndomains = cap_ndoms(iommu->cap); | |
1055 | pr_debug("Number of Domains supportd <%ld>\n", ndomains); | |
1056 | nlongs = BITS_TO_LONGS(ndomains); | |
1057 | ||
1058 | /* TBD: there might be 64K domains, | |
1059 | * consider other allocation for future chip | |
1060 | */ | |
1061 | iommu->domain_ids = kcalloc(nlongs, sizeof(unsigned long), GFP_KERNEL); | |
1062 | if (!iommu->domain_ids) { | |
1063 | printk(KERN_ERR "Allocating domain id array failed\n"); | |
1064 | return -ENOMEM; | |
1065 | } | |
1066 | iommu->domains = kcalloc(ndomains, sizeof(struct dmar_domain *), | |
1067 | GFP_KERNEL); | |
1068 | if (!iommu->domains) { | |
1069 | printk(KERN_ERR "Allocating domain array failed\n"); | |
1070 | kfree(iommu->domain_ids); | |
1071 | return -ENOMEM; | |
1072 | } | |
1073 | ||
e61d98d8 SS |
1074 | spin_lock_init(&iommu->lock); |
1075 | ||
ba395927 KA |
1076 | /* |
1077 | * if Caching mode is set, then invalid translations are tagged | |
1078 | * with domainid 0. Hence we need to pre-allocate it. | |
1079 | */ | |
1080 | if (cap_caching_mode(iommu->cap)) | |
1081 | set_bit(0, iommu->domain_ids); | |
1082 | return 0; | |
1083 | } | |
ba395927 | 1084 | |
ba395927 KA |
1085 | |
1086 | static void domain_exit(struct dmar_domain *domain); | |
5e98c4b1 | 1087 | static void vm_domain_exit(struct dmar_domain *domain); |
e61d98d8 SS |
1088 | |
1089 | void free_dmar_iommu(struct intel_iommu *iommu) | |
ba395927 KA |
1090 | { |
1091 | struct dmar_domain *domain; | |
1092 | int i; | |
c7151a8d | 1093 | unsigned long flags; |
ba395927 | 1094 | |
ba395927 KA |
1095 | i = find_first_bit(iommu->domain_ids, cap_ndoms(iommu->cap)); |
1096 | for (; i < cap_ndoms(iommu->cap); ) { | |
1097 | domain = iommu->domains[i]; | |
1098 | clear_bit(i, iommu->domain_ids); | |
c7151a8d WH |
1099 | |
1100 | spin_lock_irqsave(&domain->iommu_lock, flags); | |
5e98c4b1 WH |
1101 | if (--domain->iommu_count == 0) { |
1102 | if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) | |
1103 | vm_domain_exit(domain); | |
1104 | else | |
1105 | domain_exit(domain); | |
1106 | } | |
c7151a8d WH |
1107 | spin_unlock_irqrestore(&domain->iommu_lock, flags); |
1108 | ||
ba395927 KA |
1109 | i = find_next_bit(iommu->domain_ids, |
1110 | cap_ndoms(iommu->cap), i+1); | |
1111 | } | |
1112 | ||
1113 | if (iommu->gcmd & DMA_GCMD_TE) | |
1114 | iommu_disable_translation(iommu); | |
1115 | ||
1116 | if (iommu->irq) { | |
1117 | set_irq_data(iommu->irq, NULL); | |
1118 | /* This will mask the irq */ | |
1119 | free_irq(iommu->irq, iommu); | |
1120 | destroy_irq(iommu->irq); | |
1121 | } | |
1122 | ||
1123 | kfree(iommu->domains); | |
1124 | kfree(iommu->domain_ids); | |
1125 | ||
d9630fe9 WH |
1126 | g_iommus[iommu->seq_id] = NULL; |
1127 | ||
1128 | /* if all iommus are freed, free g_iommus */ | |
1129 | for (i = 0; i < g_num_of_iommus; i++) { | |
1130 | if (g_iommus[i]) | |
1131 | break; | |
1132 | } | |
1133 | ||
1134 | if (i == g_num_of_iommus) | |
1135 | kfree(g_iommus); | |
1136 | ||
ba395927 KA |
1137 | /* free context mapping */ |
1138 | free_context_table(iommu); | |
ba395927 KA |
1139 | } |
1140 | ||
1141 | static struct dmar_domain * iommu_alloc_domain(struct intel_iommu *iommu) | |
1142 | { | |
1143 | unsigned long num; | |
1144 | unsigned long ndomains; | |
1145 | struct dmar_domain *domain; | |
1146 | unsigned long flags; | |
1147 | ||
1148 | domain = alloc_domain_mem(); | |
1149 | if (!domain) | |
1150 | return NULL; | |
1151 | ||
1152 | ndomains = cap_ndoms(iommu->cap); | |
1153 | ||
1154 | spin_lock_irqsave(&iommu->lock, flags); | |
1155 | num = find_first_zero_bit(iommu->domain_ids, ndomains); | |
1156 | if (num >= ndomains) { | |
1157 | spin_unlock_irqrestore(&iommu->lock, flags); | |
1158 | free_domain_mem(domain); | |
1159 | printk(KERN_ERR "IOMMU: no free domain ids\n"); | |
1160 | return NULL; | |
1161 | } | |
1162 | ||
1163 | set_bit(num, iommu->domain_ids); | |
1164 | domain->id = num; | |
8c11e798 WH |
1165 | memset(&domain->iommu_bmp, 0, sizeof(unsigned long)); |
1166 | set_bit(iommu->seq_id, &domain->iommu_bmp); | |
d71a2f33 | 1167 | domain->flags = 0; |
ba395927 KA |
1168 | iommu->domains[num] = domain; |
1169 | spin_unlock_irqrestore(&iommu->lock, flags); | |
1170 | ||
1171 | return domain; | |
1172 | } | |
1173 | ||
1174 | static void iommu_free_domain(struct dmar_domain *domain) | |
1175 | { | |
1176 | unsigned long flags; | |
8c11e798 WH |
1177 | struct intel_iommu *iommu; |
1178 | ||
1179 | iommu = domain_get_iommu(domain); | |
ba395927 | 1180 | |
8c11e798 WH |
1181 | spin_lock_irqsave(&iommu->lock, flags); |
1182 | clear_bit(domain->id, iommu->domain_ids); | |
1183 | spin_unlock_irqrestore(&iommu->lock, flags); | |
ba395927 KA |
1184 | } |
1185 | ||
1186 | static struct iova_domain reserved_iova_list; | |
8a443df4 MG |
1187 | static struct lock_class_key reserved_alloc_key; |
1188 | static struct lock_class_key reserved_rbtree_key; | |
ba395927 KA |
1189 | |
1190 | static void dmar_init_reserved_ranges(void) | |
1191 | { | |
1192 | struct pci_dev *pdev = NULL; | |
1193 | struct iova *iova; | |
1194 | int i; | |
1195 | u64 addr, size; | |
1196 | ||
f661197e | 1197 | init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN); |
ba395927 | 1198 | |
8a443df4 MG |
1199 | lockdep_set_class(&reserved_iova_list.iova_alloc_lock, |
1200 | &reserved_alloc_key); | |
1201 | lockdep_set_class(&reserved_iova_list.iova_rbtree_lock, | |
1202 | &reserved_rbtree_key); | |
1203 | ||
ba395927 KA |
1204 | /* IOAPIC ranges shouldn't be accessed by DMA */ |
1205 | iova = reserve_iova(&reserved_iova_list, IOVA_PFN(IOAPIC_RANGE_START), | |
1206 | IOVA_PFN(IOAPIC_RANGE_END)); | |
1207 | if (!iova) | |
1208 | printk(KERN_ERR "Reserve IOAPIC range failed\n"); | |
1209 | ||
1210 | /* Reserve all PCI MMIO to avoid peer-to-peer access */ | |
1211 | for_each_pci_dev(pdev) { | |
1212 | struct resource *r; | |
1213 | ||
1214 | for (i = 0; i < PCI_NUM_RESOURCES; i++) { | |
1215 | r = &pdev->resource[i]; | |
1216 | if (!r->flags || !(r->flags & IORESOURCE_MEM)) | |
1217 | continue; | |
1218 | addr = r->start; | |
5b6985ce | 1219 | addr &= PAGE_MASK; |
ba395927 | 1220 | size = r->end - addr; |
5b6985ce | 1221 | size = PAGE_ALIGN(size); |
ba395927 KA |
1222 | iova = reserve_iova(&reserved_iova_list, IOVA_PFN(addr), |
1223 | IOVA_PFN(size + addr) - 1); | |
1224 | if (!iova) | |
1225 | printk(KERN_ERR "Reserve iova failed\n"); | |
1226 | } | |
1227 | } | |
1228 | ||
1229 | } | |
1230 | ||
1231 | static void domain_reserve_special_ranges(struct dmar_domain *domain) | |
1232 | { | |
1233 | copy_reserved_iova(&reserved_iova_list, &domain->iovad); | |
1234 | } | |
1235 | ||
1236 | static inline int guestwidth_to_adjustwidth(int gaw) | |
1237 | { | |
1238 | int agaw; | |
1239 | int r = (gaw - 12) % 9; | |
1240 | ||
1241 | if (r == 0) | |
1242 | agaw = gaw; | |
1243 | else | |
1244 | agaw = gaw + 9 - r; | |
1245 | if (agaw > 64) | |
1246 | agaw = 64; | |
1247 | return agaw; | |
1248 | } | |
1249 | ||
1250 | static int domain_init(struct dmar_domain *domain, int guest_width) | |
1251 | { | |
1252 | struct intel_iommu *iommu; | |
1253 | int adjust_width, agaw; | |
1254 | unsigned long sagaw; | |
1255 | ||
f661197e | 1256 | init_iova_domain(&domain->iovad, DMA_32BIT_PFN); |
ba395927 | 1257 | spin_lock_init(&domain->mapping_lock); |
c7151a8d | 1258 | spin_lock_init(&domain->iommu_lock); |
ba395927 KA |
1259 | |
1260 | domain_reserve_special_ranges(domain); | |
1261 | ||
1262 | /* calculate AGAW */ | |
8c11e798 | 1263 | iommu = domain_get_iommu(domain); |
ba395927 KA |
1264 | if (guest_width > cap_mgaw(iommu->cap)) |
1265 | guest_width = cap_mgaw(iommu->cap); | |
1266 | domain->gaw = guest_width; | |
1267 | adjust_width = guestwidth_to_adjustwidth(guest_width); | |
1268 | agaw = width_to_agaw(adjust_width); | |
1269 | sagaw = cap_sagaw(iommu->cap); | |
1270 | if (!test_bit(agaw, &sagaw)) { | |
1271 | /* hardware doesn't support it, choose a bigger one */ | |
1272 | pr_debug("IOMMU: hardware doesn't support agaw %d\n", agaw); | |
1273 | agaw = find_next_bit(&sagaw, 5, agaw); | |
1274 | if (agaw >= 5) | |
1275 | return -ENODEV; | |
1276 | } | |
1277 | domain->agaw = agaw; | |
1278 | INIT_LIST_HEAD(&domain->devices); | |
1279 | ||
8e604097 WH |
1280 | if (ecap_coherent(iommu->ecap)) |
1281 | domain->iommu_coherency = 1; | |
1282 | else | |
1283 | domain->iommu_coherency = 0; | |
1284 | ||
58c610bd SY |
1285 | if (ecap_sc_support(iommu->ecap)) |
1286 | domain->iommu_snooping = 1; | |
1287 | else | |
1288 | domain->iommu_snooping = 0; | |
1289 | ||
c7151a8d WH |
1290 | domain->iommu_count = 1; |
1291 | ||
ba395927 KA |
1292 | /* always allocate the top pgd */ |
1293 | domain->pgd = (struct dma_pte *)alloc_pgtable_page(); | |
1294 | if (!domain->pgd) | |
1295 | return -ENOMEM; | |
5b6985ce | 1296 | __iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE); |
ba395927 KA |
1297 | return 0; |
1298 | } | |
1299 | ||
1300 | static void domain_exit(struct dmar_domain *domain) | |
1301 | { | |
1302 | u64 end; | |
1303 | ||
1304 | /* Domain 0 is reserved, so dont process it */ | |
1305 | if (!domain) | |
1306 | return; | |
1307 | ||
1308 | domain_remove_dev_info(domain); | |
1309 | /* destroy iovas */ | |
1310 | put_iova_domain(&domain->iovad); | |
1311 | end = DOMAIN_MAX_ADDR(domain->gaw); | |
5b6985ce | 1312 | end = end & (~PAGE_MASK); |
ba395927 KA |
1313 | |
1314 | /* clear ptes */ | |
1315 | dma_pte_clear_range(domain, 0, end); | |
1316 | ||
1317 | /* free page tables */ | |
1318 | dma_pte_free_pagetable(domain, 0, end); | |
1319 | ||
1320 | iommu_free_domain(domain); | |
1321 | free_domain_mem(domain); | |
1322 | } | |
1323 | ||
1324 | static int domain_context_mapping_one(struct dmar_domain *domain, | |
276dbf99 | 1325 | int segment, u8 bus, u8 devfn) |
ba395927 KA |
1326 | { |
1327 | struct context_entry *context; | |
ba395927 | 1328 | unsigned long flags; |
5331fe6f | 1329 | struct intel_iommu *iommu; |
ea6606b0 WH |
1330 | struct dma_pte *pgd; |
1331 | unsigned long num; | |
1332 | unsigned long ndomains; | |
1333 | int id; | |
1334 | int agaw; | |
ba395927 KA |
1335 | |
1336 | pr_debug("Set context mapping for %02x:%02x.%d\n", | |
1337 | bus, PCI_SLOT(devfn), PCI_FUNC(devfn)); | |
1338 | BUG_ON(!domain->pgd); | |
5331fe6f | 1339 | |
276dbf99 | 1340 | iommu = device_to_iommu(segment, bus, devfn); |
5331fe6f WH |
1341 | if (!iommu) |
1342 | return -ENODEV; | |
1343 | ||
ba395927 KA |
1344 | context = device_to_context_entry(iommu, bus, devfn); |
1345 | if (!context) | |
1346 | return -ENOMEM; | |
1347 | spin_lock_irqsave(&iommu->lock, flags); | |
c07e7d21 | 1348 | if (context_present(context)) { |
ba395927 KA |
1349 | spin_unlock_irqrestore(&iommu->lock, flags); |
1350 | return 0; | |
1351 | } | |
1352 | ||
ea6606b0 WH |
1353 | id = domain->id; |
1354 | pgd = domain->pgd; | |
1355 | ||
1356 | if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) { | |
1357 | int found = 0; | |
1358 | ||
1359 | /* find an available domain id for this device in iommu */ | |
1360 | ndomains = cap_ndoms(iommu->cap); | |
1361 | num = find_first_bit(iommu->domain_ids, ndomains); | |
1362 | for (; num < ndomains; ) { | |
1363 | if (iommu->domains[num] == domain) { | |
1364 | id = num; | |
1365 | found = 1; | |
1366 | break; | |
1367 | } | |
1368 | num = find_next_bit(iommu->domain_ids, | |
1369 | cap_ndoms(iommu->cap), num+1); | |
1370 | } | |
1371 | ||
1372 | if (found == 0) { | |
1373 | num = find_first_zero_bit(iommu->domain_ids, ndomains); | |
1374 | if (num >= ndomains) { | |
1375 | spin_unlock_irqrestore(&iommu->lock, flags); | |
1376 | printk(KERN_ERR "IOMMU: no free domain ids\n"); | |
1377 | return -EFAULT; | |
1378 | } | |
1379 | ||
1380 | set_bit(num, iommu->domain_ids); | |
1381 | iommu->domains[num] = domain; | |
1382 | id = num; | |
1383 | } | |
1384 | ||
1385 | /* Skip top levels of page tables for | |
1386 | * iommu which has less agaw than default. | |
1387 | */ | |
1388 | for (agaw = domain->agaw; agaw != iommu->agaw; agaw--) { | |
1389 | pgd = phys_to_virt(dma_pte_addr(pgd)); | |
1390 | if (!dma_pte_present(pgd)) { | |
1391 | spin_unlock_irqrestore(&iommu->lock, flags); | |
1392 | return -ENOMEM; | |
1393 | } | |
1394 | } | |
1395 | } | |
1396 | ||
1397 | context_set_domain_id(context, id); | |
1398 | context_set_address_width(context, iommu->agaw); | |
1399 | context_set_address_root(context, virt_to_phys(pgd)); | |
c07e7d21 MM |
1400 | context_set_translation_type(context, CONTEXT_TT_MULTI_LEVEL); |
1401 | context_set_fault_enable(context); | |
1402 | context_set_present(context); | |
5331fe6f | 1403 | domain_flush_cache(domain, context, sizeof(*context)); |
ba395927 KA |
1404 | |
1405 | /* it's a non-present to present mapping */ | |
a77b67d4 YS |
1406 | if (iommu->flush.flush_context(iommu, domain->id, |
1407 | (((u16)bus) << 8) | devfn, DMA_CCMD_MASK_NOBIT, | |
1408 | DMA_CCMD_DEVICE_INVL, 1)) | |
ba395927 KA |
1409 | iommu_flush_write_buffer(iommu); |
1410 | else | |
a77b67d4 YS |
1411 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_DSI_FLUSH, 0); |
1412 | ||
ba395927 | 1413 | spin_unlock_irqrestore(&iommu->lock, flags); |
c7151a8d WH |
1414 | |
1415 | spin_lock_irqsave(&domain->iommu_lock, flags); | |
1416 | if (!test_and_set_bit(iommu->seq_id, &domain->iommu_bmp)) { | |
1417 | domain->iommu_count++; | |
58c610bd | 1418 | domain_update_iommu_cap(domain); |
c7151a8d WH |
1419 | } |
1420 | spin_unlock_irqrestore(&domain->iommu_lock, flags); | |
ba395927 KA |
1421 | return 0; |
1422 | } | |
1423 | ||
1424 | static int | |
1425 | domain_context_mapping(struct dmar_domain *domain, struct pci_dev *pdev) | |
1426 | { | |
1427 | int ret; | |
1428 | struct pci_dev *tmp, *parent; | |
1429 | ||
276dbf99 DW |
1430 | ret = domain_context_mapping_one(domain, pci_domain_nr(pdev->bus), |
1431 | pdev->bus->number, pdev->devfn); | |
ba395927 KA |
1432 | if (ret) |
1433 | return ret; | |
1434 | ||
1435 | /* dependent device mapping */ | |
1436 | tmp = pci_find_upstream_pcie_bridge(pdev); | |
1437 | if (!tmp) | |
1438 | return 0; | |
1439 | /* Secondary interface's bus number and devfn 0 */ | |
1440 | parent = pdev->bus->self; | |
1441 | while (parent != tmp) { | |
276dbf99 DW |
1442 | ret = domain_context_mapping_one(domain, |
1443 | pci_domain_nr(parent->bus), | |
1444 | parent->bus->number, | |
1445 | parent->devfn); | |
ba395927 KA |
1446 | if (ret) |
1447 | return ret; | |
1448 | parent = parent->bus->self; | |
1449 | } | |
1450 | if (tmp->is_pcie) /* this is a PCIE-to-PCI bridge */ | |
1451 | return domain_context_mapping_one(domain, | |
276dbf99 DW |
1452 | pci_domain_nr(tmp->subordinate), |
1453 | tmp->subordinate->number, 0); | |
ba395927 KA |
1454 | else /* this is a legacy PCI bridge */ |
1455 | return domain_context_mapping_one(domain, | |
276dbf99 DW |
1456 | pci_domain_nr(tmp->bus), |
1457 | tmp->bus->number, | |
1458 | tmp->devfn); | |
ba395927 KA |
1459 | } |
1460 | ||
5331fe6f | 1461 | static int domain_context_mapped(struct pci_dev *pdev) |
ba395927 KA |
1462 | { |
1463 | int ret; | |
1464 | struct pci_dev *tmp, *parent; | |
5331fe6f WH |
1465 | struct intel_iommu *iommu; |
1466 | ||
276dbf99 DW |
1467 | iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number, |
1468 | pdev->devfn); | |
5331fe6f WH |
1469 | if (!iommu) |
1470 | return -ENODEV; | |
ba395927 | 1471 | |
276dbf99 | 1472 | ret = device_context_mapped(iommu, pdev->bus->number, pdev->devfn); |
ba395927 KA |
1473 | if (!ret) |
1474 | return ret; | |
1475 | /* dependent device mapping */ | |
1476 | tmp = pci_find_upstream_pcie_bridge(pdev); | |
1477 | if (!tmp) | |
1478 | return ret; | |
1479 | /* Secondary interface's bus number and devfn 0 */ | |
1480 | parent = pdev->bus->self; | |
1481 | while (parent != tmp) { | |
8c11e798 | 1482 | ret = device_context_mapped(iommu, parent->bus->number, |
276dbf99 | 1483 | parent->devfn); |
ba395927 KA |
1484 | if (!ret) |
1485 | return ret; | |
1486 | parent = parent->bus->self; | |
1487 | } | |
1488 | if (tmp->is_pcie) | |
276dbf99 DW |
1489 | return device_context_mapped(iommu, tmp->subordinate->number, |
1490 | 0); | |
ba395927 | 1491 | else |
276dbf99 DW |
1492 | return device_context_mapped(iommu, tmp->bus->number, |
1493 | tmp->devfn); | |
ba395927 KA |
1494 | } |
1495 | ||
1496 | static int | |
1497 | domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova, | |
1498 | u64 hpa, size_t size, int prot) | |
1499 | { | |
1500 | u64 start_pfn, end_pfn; | |
1501 | struct dma_pte *pte; | |
1502 | int index; | |
5b6985ce FY |
1503 | int addr_width = agaw_to_width(domain->agaw); |
1504 | ||
1505 | hpa &= (((u64)1) << addr_width) - 1; | |
ba395927 KA |
1506 | |
1507 | if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0) | |
1508 | return -EINVAL; | |
5b6985ce FY |
1509 | iova &= PAGE_MASK; |
1510 | start_pfn = ((u64)hpa) >> VTD_PAGE_SHIFT; | |
1511 | end_pfn = (VTD_PAGE_ALIGN(((u64)hpa) + size)) >> VTD_PAGE_SHIFT; | |
ba395927 KA |
1512 | index = 0; |
1513 | while (start_pfn < end_pfn) { | |
5b6985ce | 1514 | pte = addr_to_dma_pte(domain, iova + VTD_PAGE_SIZE * index); |
ba395927 KA |
1515 | if (!pte) |
1516 | return -ENOMEM; | |
1517 | /* We don't need lock here, nobody else | |
1518 | * touches the iova range | |
1519 | */ | |
19c239ce MM |
1520 | BUG_ON(dma_pte_addr(pte)); |
1521 | dma_set_pte_addr(pte, start_pfn << VTD_PAGE_SHIFT); | |
1522 | dma_set_pte_prot(pte, prot); | |
9cf06697 SY |
1523 | if (prot & DMA_PTE_SNP) |
1524 | dma_set_pte_snp(pte); | |
5331fe6f | 1525 | domain_flush_cache(domain, pte, sizeof(*pte)); |
ba395927 KA |
1526 | start_pfn++; |
1527 | index++; | |
1528 | } | |
1529 | return 0; | |
1530 | } | |
1531 | ||
c7151a8d | 1532 | static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn) |
ba395927 | 1533 | { |
c7151a8d WH |
1534 | if (!iommu) |
1535 | return; | |
8c11e798 WH |
1536 | |
1537 | clear_context_table(iommu, bus, devfn); | |
1538 | iommu->flush.flush_context(iommu, 0, 0, 0, | |
a77b67d4 | 1539 | DMA_CCMD_GLOBAL_INVL, 0); |
8c11e798 | 1540 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, |
a77b67d4 | 1541 | DMA_TLB_GLOBAL_FLUSH, 0); |
ba395927 KA |
1542 | } |
1543 | ||
1544 | static void domain_remove_dev_info(struct dmar_domain *domain) | |
1545 | { | |
1546 | struct device_domain_info *info; | |
1547 | unsigned long flags; | |
c7151a8d | 1548 | struct intel_iommu *iommu; |
ba395927 KA |
1549 | |
1550 | spin_lock_irqsave(&device_domain_lock, flags); | |
1551 | while (!list_empty(&domain->devices)) { | |
1552 | info = list_entry(domain->devices.next, | |
1553 | struct device_domain_info, link); | |
1554 | list_del(&info->link); | |
1555 | list_del(&info->global); | |
1556 | if (info->dev) | |
358dd8ac | 1557 | info->dev->dev.archdata.iommu = NULL; |
ba395927 KA |
1558 | spin_unlock_irqrestore(&device_domain_lock, flags); |
1559 | ||
276dbf99 | 1560 | iommu = device_to_iommu(info->segment, info->bus, info->devfn); |
c7151a8d | 1561 | iommu_detach_dev(iommu, info->bus, info->devfn); |
ba395927 KA |
1562 | free_devinfo_mem(info); |
1563 | ||
1564 | spin_lock_irqsave(&device_domain_lock, flags); | |
1565 | } | |
1566 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
1567 | } | |
1568 | ||
1569 | /* | |
1570 | * find_domain | |
358dd8ac | 1571 | * Note: we use struct pci_dev->dev.archdata.iommu stores the info |
ba395927 | 1572 | */ |
38717946 | 1573 | static struct dmar_domain * |
ba395927 KA |
1574 | find_domain(struct pci_dev *pdev) |
1575 | { | |
1576 | struct device_domain_info *info; | |
1577 | ||
1578 | /* No lock here, assumes no domain exit in normal case */ | |
358dd8ac | 1579 | info = pdev->dev.archdata.iommu; |
ba395927 KA |
1580 | if (info) |
1581 | return info->domain; | |
1582 | return NULL; | |
1583 | } | |
1584 | ||
ba395927 KA |
1585 | /* domain is initialized */ |
1586 | static struct dmar_domain *get_domain_for_dev(struct pci_dev *pdev, int gaw) | |
1587 | { | |
1588 | struct dmar_domain *domain, *found = NULL; | |
1589 | struct intel_iommu *iommu; | |
1590 | struct dmar_drhd_unit *drhd; | |
1591 | struct device_domain_info *info, *tmp; | |
1592 | struct pci_dev *dev_tmp; | |
1593 | unsigned long flags; | |
1594 | int bus = 0, devfn = 0; | |
276dbf99 | 1595 | int segment; |
ba395927 KA |
1596 | |
1597 | domain = find_domain(pdev); | |
1598 | if (domain) | |
1599 | return domain; | |
1600 | ||
276dbf99 DW |
1601 | segment = pci_domain_nr(pdev->bus); |
1602 | ||
ba395927 KA |
1603 | dev_tmp = pci_find_upstream_pcie_bridge(pdev); |
1604 | if (dev_tmp) { | |
1605 | if (dev_tmp->is_pcie) { | |
1606 | bus = dev_tmp->subordinate->number; | |
1607 | devfn = 0; | |
1608 | } else { | |
1609 | bus = dev_tmp->bus->number; | |
1610 | devfn = dev_tmp->devfn; | |
1611 | } | |
1612 | spin_lock_irqsave(&device_domain_lock, flags); | |
1613 | list_for_each_entry(info, &device_domain_list, global) { | |
276dbf99 DW |
1614 | if (info->segment == segment && |
1615 | info->bus == bus && info->devfn == devfn) { | |
ba395927 KA |
1616 | found = info->domain; |
1617 | break; | |
1618 | } | |
1619 | } | |
1620 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
1621 | /* pcie-pci bridge already has a domain, uses it */ | |
1622 | if (found) { | |
1623 | domain = found; | |
1624 | goto found_domain; | |
1625 | } | |
1626 | } | |
1627 | ||
1628 | /* Allocate new domain for the device */ | |
1629 | drhd = dmar_find_matched_drhd_unit(pdev); | |
1630 | if (!drhd) { | |
1631 | printk(KERN_ERR "IOMMU: can't find DMAR for device %s\n", | |
1632 | pci_name(pdev)); | |
1633 | return NULL; | |
1634 | } | |
1635 | iommu = drhd->iommu; | |
1636 | ||
1637 | domain = iommu_alloc_domain(iommu); | |
1638 | if (!domain) | |
1639 | goto error; | |
1640 | ||
1641 | if (domain_init(domain, gaw)) { | |
1642 | domain_exit(domain); | |
1643 | goto error; | |
1644 | } | |
1645 | ||
1646 | /* register pcie-to-pci device */ | |
1647 | if (dev_tmp) { | |
1648 | info = alloc_devinfo_mem(); | |
1649 | if (!info) { | |
1650 | domain_exit(domain); | |
1651 | goto error; | |
1652 | } | |
276dbf99 | 1653 | info->segment = segment; |
ba395927 KA |
1654 | info->bus = bus; |
1655 | info->devfn = devfn; | |
1656 | info->dev = NULL; | |
1657 | info->domain = domain; | |
1658 | /* This domain is shared by devices under p2p bridge */ | |
3b5410e7 | 1659 | domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES; |
ba395927 KA |
1660 | |
1661 | /* pcie-to-pci bridge already has a domain, uses it */ | |
1662 | found = NULL; | |
1663 | spin_lock_irqsave(&device_domain_lock, flags); | |
1664 | list_for_each_entry(tmp, &device_domain_list, global) { | |
276dbf99 DW |
1665 | if (tmp->segment == segment && |
1666 | tmp->bus == bus && tmp->devfn == devfn) { | |
ba395927 KA |
1667 | found = tmp->domain; |
1668 | break; | |
1669 | } | |
1670 | } | |
1671 | if (found) { | |
1672 | free_devinfo_mem(info); | |
1673 | domain_exit(domain); | |
1674 | domain = found; | |
1675 | } else { | |
1676 | list_add(&info->link, &domain->devices); | |
1677 | list_add(&info->global, &device_domain_list); | |
1678 | } | |
1679 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
1680 | } | |
1681 | ||
1682 | found_domain: | |
1683 | info = alloc_devinfo_mem(); | |
1684 | if (!info) | |
1685 | goto error; | |
276dbf99 | 1686 | info->segment = segment; |
ba395927 KA |
1687 | info->bus = pdev->bus->number; |
1688 | info->devfn = pdev->devfn; | |
1689 | info->dev = pdev; | |
1690 | info->domain = domain; | |
1691 | spin_lock_irqsave(&device_domain_lock, flags); | |
1692 | /* somebody is fast */ | |
1693 | found = find_domain(pdev); | |
1694 | if (found != NULL) { | |
1695 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
1696 | if (found != domain) { | |
1697 | domain_exit(domain); | |
1698 | domain = found; | |
1699 | } | |
1700 | free_devinfo_mem(info); | |
1701 | return domain; | |
1702 | } | |
1703 | list_add(&info->link, &domain->devices); | |
1704 | list_add(&info->global, &device_domain_list); | |
358dd8ac | 1705 | pdev->dev.archdata.iommu = info; |
ba395927 KA |
1706 | spin_unlock_irqrestore(&device_domain_lock, flags); |
1707 | return domain; | |
1708 | error: | |
1709 | /* recheck it here, maybe others set it */ | |
1710 | return find_domain(pdev); | |
1711 | } | |
1712 | ||
5b6985ce FY |
1713 | static int iommu_prepare_identity_map(struct pci_dev *pdev, |
1714 | unsigned long long start, | |
1715 | unsigned long long end) | |
ba395927 KA |
1716 | { |
1717 | struct dmar_domain *domain; | |
1718 | unsigned long size; | |
5b6985ce | 1719 | unsigned long long base; |
ba395927 KA |
1720 | int ret; |
1721 | ||
1722 | printk(KERN_INFO | |
1723 | "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n", | |
1724 | pci_name(pdev), start, end); | |
1725 | /* page table init */ | |
1726 | domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH); | |
1727 | if (!domain) | |
1728 | return -ENOMEM; | |
1729 | ||
1730 | /* The address might not be aligned */ | |
5b6985ce | 1731 | base = start & PAGE_MASK; |
ba395927 | 1732 | size = end - base; |
5b6985ce | 1733 | size = PAGE_ALIGN(size); |
ba395927 KA |
1734 | if (!reserve_iova(&domain->iovad, IOVA_PFN(base), |
1735 | IOVA_PFN(base + size) - 1)) { | |
1736 | printk(KERN_ERR "IOMMU: reserve iova failed\n"); | |
1737 | ret = -ENOMEM; | |
1738 | goto error; | |
1739 | } | |
1740 | ||
1741 | pr_debug("Mapping reserved region %lx@%llx for %s\n", | |
1742 | size, base, pci_name(pdev)); | |
1743 | /* | |
1744 | * RMRR range might have overlap with physical memory range, | |
1745 | * clear it first | |
1746 | */ | |
1747 | dma_pte_clear_range(domain, base, base + size); | |
1748 | ||
1749 | ret = domain_page_mapping(domain, base, base, size, | |
1750 | DMA_PTE_READ|DMA_PTE_WRITE); | |
1751 | if (ret) | |
1752 | goto error; | |
1753 | ||
1754 | /* context entry init */ | |
1755 | ret = domain_context_mapping(domain, pdev); | |
1756 | if (!ret) | |
1757 | return 0; | |
1758 | error: | |
1759 | domain_exit(domain); | |
1760 | return ret; | |
1761 | ||
1762 | } | |
1763 | ||
1764 | static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr, | |
1765 | struct pci_dev *pdev) | |
1766 | { | |
358dd8ac | 1767 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) |
ba395927 KA |
1768 | return 0; |
1769 | return iommu_prepare_identity_map(pdev, rmrr->base_address, | |
1770 | rmrr->end_address + 1); | |
1771 | } | |
1772 | ||
e820482c | 1773 | #ifdef CONFIG_DMAR_GFX_WA |
d52d53b8 YL |
1774 | struct iommu_prepare_data { |
1775 | struct pci_dev *pdev; | |
1776 | int ret; | |
1777 | }; | |
1778 | ||
1779 | static int __init iommu_prepare_work_fn(unsigned long start_pfn, | |
1780 | unsigned long end_pfn, void *datax) | |
1781 | { | |
1782 | struct iommu_prepare_data *data; | |
1783 | ||
1784 | data = (struct iommu_prepare_data *)datax; | |
1785 | ||
1786 | data->ret = iommu_prepare_identity_map(data->pdev, | |
1787 | start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT); | |
1788 | return data->ret; | |
1789 | ||
1790 | } | |
1791 | ||
1792 | static int __init iommu_prepare_with_active_regions(struct pci_dev *pdev) | |
1793 | { | |
1794 | int nid; | |
1795 | struct iommu_prepare_data data; | |
1796 | ||
1797 | data.pdev = pdev; | |
1798 | data.ret = 0; | |
1799 | ||
1800 | for_each_online_node(nid) { | |
1801 | work_with_active_regions(nid, iommu_prepare_work_fn, &data); | |
1802 | if (data.ret) | |
1803 | return data.ret; | |
1804 | } | |
1805 | return data.ret; | |
1806 | } | |
1807 | ||
e820482c KA |
1808 | static void __init iommu_prepare_gfx_mapping(void) |
1809 | { | |
1810 | struct pci_dev *pdev = NULL; | |
e820482c KA |
1811 | int ret; |
1812 | ||
1813 | for_each_pci_dev(pdev) { | |
358dd8ac | 1814 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO || |
e820482c KA |
1815 | !IS_GFX_DEVICE(pdev)) |
1816 | continue; | |
1817 | printk(KERN_INFO "IOMMU: gfx device %s 1-1 mapping\n", | |
1818 | pci_name(pdev)); | |
d52d53b8 YL |
1819 | ret = iommu_prepare_with_active_regions(pdev); |
1820 | if (ret) | |
1821 | printk(KERN_ERR "IOMMU: mapping reserved region failed\n"); | |
e820482c KA |
1822 | } |
1823 | } | |
2abd7e16 MM |
1824 | #else /* !CONFIG_DMAR_GFX_WA */ |
1825 | static inline void iommu_prepare_gfx_mapping(void) | |
1826 | { | |
1827 | return; | |
1828 | } | |
e820482c KA |
1829 | #endif |
1830 | ||
49a0429e KA |
1831 | #ifdef CONFIG_DMAR_FLOPPY_WA |
1832 | static inline void iommu_prepare_isa(void) | |
1833 | { | |
1834 | struct pci_dev *pdev; | |
1835 | int ret; | |
1836 | ||
1837 | pdev = pci_get_class(PCI_CLASS_BRIDGE_ISA << 8, NULL); | |
1838 | if (!pdev) | |
1839 | return; | |
1840 | ||
1841 | printk(KERN_INFO "IOMMU: Prepare 0-16M unity mapping for LPC\n"); | |
1842 | ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024); | |
1843 | ||
1844 | if (ret) | |
1c35b8e5 | 1845 | printk(KERN_ERR "IOMMU: Failed to create 0-64M identity map, " |
49a0429e KA |
1846 | "floppy might not work\n"); |
1847 | ||
1848 | } | |
1849 | #else | |
1850 | static inline void iommu_prepare_isa(void) | |
1851 | { | |
1852 | return; | |
1853 | } | |
1854 | #endif /* !CONFIG_DMAR_FLPY_WA */ | |
1855 | ||
519a0549 | 1856 | static int __init init_dmars(void) |
ba395927 KA |
1857 | { |
1858 | struct dmar_drhd_unit *drhd; | |
1859 | struct dmar_rmrr_unit *rmrr; | |
1860 | struct pci_dev *pdev; | |
1861 | struct intel_iommu *iommu; | |
9d783ba0 | 1862 | int i, ret; |
ba395927 KA |
1863 | |
1864 | /* | |
1865 | * for each drhd | |
1866 | * allocate root | |
1867 | * initialize and program root entry to not present | |
1868 | * endfor | |
1869 | */ | |
1870 | for_each_drhd_unit(drhd) { | |
5e0d2a6f | 1871 | g_num_of_iommus++; |
1872 | /* | |
1873 | * lock not needed as this is only incremented in the single | |
1874 | * threaded kernel __init code path all other access are read | |
1875 | * only | |
1876 | */ | |
1877 | } | |
1878 | ||
d9630fe9 WH |
1879 | g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *), |
1880 | GFP_KERNEL); | |
1881 | if (!g_iommus) { | |
1882 | printk(KERN_ERR "Allocating global iommu array failed\n"); | |
1883 | ret = -ENOMEM; | |
1884 | goto error; | |
1885 | } | |
1886 | ||
80b20dd8 | 1887 | deferred_flush = kzalloc(g_num_of_iommus * |
1888 | sizeof(struct deferred_flush_tables), GFP_KERNEL); | |
1889 | if (!deferred_flush) { | |
d9630fe9 | 1890 | kfree(g_iommus); |
5e0d2a6f | 1891 | ret = -ENOMEM; |
1892 | goto error; | |
1893 | } | |
1894 | ||
5e0d2a6f | 1895 | for_each_drhd_unit(drhd) { |
1896 | if (drhd->ignored) | |
1897 | continue; | |
1886e8a9 SS |
1898 | |
1899 | iommu = drhd->iommu; | |
d9630fe9 | 1900 | g_iommus[iommu->seq_id] = iommu; |
ba395927 | 1901 | |
e61d98d8 SS |
1902 | ret = iommu_init_domains(iommu); |
1903 | if (ret) | |
1904 | goto error; | |
1905 | ||
ba395927 KA |
1906 | /* |
1907 | * TBD: | |
1908 | * we could share the same root & context tables | |
1909 | * amoung all IOMMU's. Need to Split it later. | |
1910 | */ | |
1911 | ret = iommu_alloc_root_entry(iommu); | |
1912 | if (ret) { | |
1913 | printk(KERN_ERR "IOMMU: allocate root entry failed\n"); | |
1914 | goto error; | |
1915 | } | |
1916 | } | |
1917 | ||
1531a6a6 SS |
1918 | /* |
1919 | * Start from the sane iommu hardware state. | |
1920 | */ | |
a77b67d4 YS |
1921 | for_each_drhd_unit(drhd) { |
1922 | if (drhd->ignored) | |
1923 | continue; | |
1924 | ||
1925 | iommu = drhd->iommu; | |
1531a6a6 SS |
1926 | |
1927 | /* | |
1928 | * If the queued invalidation is already initialized by us | |
1929 | * (for example, while enabling interrupt-remapping) then | |
1930 | * we got the things already rolling from a sane state. | |
1931 | */ | |
1932 | if (iommu->qi) | |
1933 | continue; | |
1934 | ||
1935 | /* | |
1936 | * Clear any previous faults. | |
1937 | */ | |
1938 | dmar_fault(-1, iommu); | |
1939 | /* | |
1940 | * Disable queued invalidation if supported and already enabled | |
1941 | * before OS handover. | |
1942 | */ | |
1943 | dmar_disable_qi(iommu); | |
1944 | } | |
1945 | ||
1946 | for_each_drhd_unit(drhd) { | |
1947 | if (drhd->ignored) | |
1948 | continue; | |
1949 | ||
1950 | iommu = drhd->iommu; | |
1951 | ||
a77b67d4 YS |
1952 | if (dmar_enable_qi(iommu)) { |
1953 | /* | |
1954 | * Queued Invalidate not enabled, use Register Based | |
1955 | * Invalidate | |
1956 | */ | |
1957 | iommu->flush.flush_context = __iommu_flush_context; | |
1958 | iommu->flush.flush_iotlb = __iommu_flush_iotlb; | |
1959 | printk(KERN_INFO "IOMMU 0x%Lx: using Register based " | |
b4e0f9eb FT |
1960 | "invalidation\n", |
1961 | (unsigned long long)drhd->reg_base_addr); | |
a77b67d4 YS |
1962 | } else { |
1963 | iommu->flush.flush_context = qi_flush_context; | |
1964 | iommu->flush.flush_iotlb = qi_flush_iotlb; | |
1965 | printk(KERN_INFO "IOMMU 0x%Lx: using Queued " | |
b4e0f9eb FT |
1966 | "invalidation\n", |
1967 | (unsigned long long)drhd->reg_base_addr); | |
a77b67d4 YS |
1968 | } |
1969 | } | |
1970 | ||
d0b03bd1 HW |
1971 | #ifdef CONFIG_INTR_REMAP |
1972 | if (!intr_remapping_enabled) { | |
1973 | ret = enable_intr_remapping(0); | |
1974 | if (ret) | |
1975 | printk(KERN_ERR | |
1976 | "IOMMU: enable interrupt remapping failed\n"); | |
1977 | } | |
1978 | #endif | |
1979 | ||
ba395927 KA |
1980 | /* |
1981 | * For each rmrr | |
1982 | * for each dev attached to rmrr | |
1983 | * do | |
1984 | * locate drhd for dev, alloc domain for dev | |
1985 | * allocate free domain | |
1986 | * allocate page table entries for rmrr | |
1987 | * if context not allocated for bus | |
1988 | * allocate and init context | |
1989 | * set present in root table for this bus | |
1990 | * init context with domain, translation etc | |
1991 | * endfor | |
1992 | * endfor | |
1993 | */ | |
1994 | for_each_rmrr_units(rmrr) { | |
ba395927 KA |
1995 | for (i = 0; i < rmrr->devices_cnt; i++) { |
1996 | pdev = rmrr->devices[i]; | |
1997 | /* some BIOS lists non-exist devices in DMAR table */ | |
1998 | if (!pdev) | |
1999 | continue; | |
2000 | ret = iommu_prepare_rmrr_dev(rmrr, pdev); | |
2001 | if (ret) | |
2002 | printk(KERN_ERR | |
2003 | "IOMMU: mapping reserved region failed\n"); | |
2004 | } | |
2005 | } | |
2006 | ||
e820482c KA |
2007 | iommu_prepare_gfx_mapping(); |
2008 | ||
49a0429e KA |
2009 | iommu_prepare_isa(); |
2010 | ||
ba395927 KA |
2011 | /* |
2012 | * for each drhd | |
2013 | * enable fault log | |
2014 | * global invalidate context cache | |
2015 | * global invalidate iotlb | |
2016 | * enable translation | |
2017 | */ | |
2018 | for_each_drhd_unit(drhd) { | |
2019 | if (drhd->ignored) | |
2020 | continue; | |
2021 | iommu = drhd->iommu; | |
ba395927 KA |
2022 | |
2023 | iommu_flush_write_buffer(iommu); | |
2024 | ||
3460a6d9 KA |
2025 | ret = dmar_set_interrupt(iommu); |
2026 | if (ret) | |
2027 | goto error; | |
2028 | ||
ba395927 KA |
2029 | iommu_set_root_entry(iommu); |
2030 | ||
a77b67d4 YS |
2031 | iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL, |
2032 | 0); | |
2033 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH, | |
2034 | 0); | |
f8bab735 | 2035 | iommu_disable_protect_mem_regions(iommu); |
2036 | ||
ba395927 KA |
2037 | ret = iommu_enable_translation(iommu); |
2038 | if (ret) | |
2039 | goto error; | |
2040 | } | |
2041 | ||
2042 | return 0; | |
2043 | error: | |
2044 | for_each_drhd_unit(drhd) { | |
2045 | if (drhd->ignored) | |
2046 | continue; | |
2047 | iommu = drhd->iommu; | |
2048 | free_iommu(iommu); | |
2049 | } | |
d9630fe9 | 2050 | kfree(g_iommus); |
ba395927 KA |
2051 | return ret; |
2052 | } | |
2053 | ||
2054 | static inline u64 aligned_size(u64 host_addr, size_t size) | |
2055 | { | |
2056 | u64 addr; | |
5b6985ce FY |
2057 | addr = (host_addr & (~PAGE_MASK)) + size; |
2058 | return PAGE_ALIGN(addr); | |
ba395927 KA |
2059 | } |
2060 | ||
2061 | struct iova * | |
f76aec76 | 2062 | iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end) |
ba395927 | 2063 | { |
ba395927 KA |
2064 | struct iova *piova; |
2065 | ||
2066 | /* Make sure it's in range */ | |
ba395927 | 2067 | end = min_t(u64, DOMAIN_MAX_ADDR(domain->gaw), end); |
f76aec76 | 2068 | if (!size || (IOVA_START_ADDR + size > end)) |
ba395927 KA |
2069 | return NULL; |
2070 | ||
2071 | piova = alloc_iova(&domain->iovad, | |
5b6985ce | 2072 | size >> PAGE_SHIFT, IOVA_PFN(end), 1); |
ba395927 KA |
2073 | return piova; |
2074 | } | |
2075 | ||
f76aec76 KA |
2076 | static struct iova * |
2077 | __intel_alloc_iova(struct device *dev, struct dmar_domain *domain, | |
bb9e6d65 | 2078 | size_t size, u64 dma_mask) |
ba395927 | 2079 | { |
ba395927 | 2080 | struct pci_dev *pdev = to_pci_dev(dev); |
ba395927 | 2081 | struct iova *iova = NULL; |
ba395927 | 2082 | |
284901a9 | 2083 | if (dma_mask <= DMA_BIT_MASK(32) || dmar_forcedac) |
bb9e6d65 FT |
2084 | iova = iommu_alloc_iova(domain, size, dma_mask); |
2085 | else { | |
ba395927 KA |
2086 | /* |
2087 | * First try to allocate an io virtual address in | |
284901a9 | 2088 | * DMA_BIT_MASK(32) and if that fails then try allocating |
3609801e | 2089 | * from higher range |
ba395927 | 2090 | */ |
284901a9 | 2091 | iova = iommu_alloc_iova(domain, size, DMA_BIT_MASK(32)); |
ba395927 | 2092 | if (!iova) |
bb9e6d65 | 2093 | iova = iommu_alloc_iova(domain, size, dma_mask); |
ba395927 KA |
2094 | } |
2095 | ||
2096 | if (!iova) { | |
2097 | printk(KERN_ERR"Allocating iova for %s failed", pci_name(pdev)); | |
f76aec76 KA |
2098 | return NULL; |
2099 | } | |
2100 | ||
2101 | return iova; | |
2102 | } | |
2103 | ||
2104 | static struct dmar_domain * | |
2105 | get_valid_domain_for_dev(struct pci_dev *pdev) | |
2106 | { | |
2107 | struct dmar_domain *domain; | |
2108 | int ret; | |
2109 | ||
2110 | domain = get_domain_for_dev(pdev, | |
2111 | DEFAULT_DOMAIN_ADDRESS_WIDTH); | |
2112 | if (!domain) { | |
2113 | printk(KERN_ERR | |
2114 | "Allocating domain for %s failed", pci_name(pdev)); | |
4fe05bbc | 2115 | return NULL; |
ba395927 KA |
2116 | } |
2117 | ||
2118 | /* make sure context mapping is ok */ | |
5331fe6f | 2119 | if (unlikely(!domain_context_mapped(pdev))) { |
ba395927 | 2120 | ret = domain_context_mapping(domain, pdev); |
f76aec76 KA |
2121 | if (ret) { |
2122 | printk(KERN_ERR | |
2123 | "Domain context map for %s failed", | |
2124 | pci_name(pdev)); | |
4fe05bbc | 2125 | return NULL; |
f76aec76 | 2126 | } |
ba395927 KA |
2127 | } |
2128 | ||
f76aec76 KA |
2129 | return domain; |
2130 | } | |
2131 | ||
bb9e6d65 FT |
2132 | static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr, |
2133 | size_t size, int dir, u64 dma_mask) | |
f76aec76 KA |
2134 | { |
2135 | struct pci_dev *pdev = to_pci_dev(hwdev); | |
f76aec76 | 2136 | struct dmar_domain *domain; |
5b6985ce | 2137 | phys_addr_t start_paddr; |
f76aec76 KA |
2138 | struct iova *iova; |
2139 | int prot = 0; | |
6865f0d1 | 2140 | int ret; |
8c11e798 | 2141 | struct intel_iommu *iommu; |
f76aec76 KA |
2142 | |
2143 | BUG_ON(dir == DMA_NONE); | |
358dd8ac | 2144 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) |
6865f0d1 | 2145 | return paddr; |
f76aec76 KA |
2146 | |
2147 | domain = get_valid_domain_for_dev(pdev); | |
2148 | if (!domain) | |
2149 | return 0; | |
2150 | ||
8c11e798 | 2151 | iommu = domain_get_iommu(domain); |
6865f0d1 | 2152 | size = aligned_size((u64)paddr, size); |
f76aec76 | 2153 | |
bb9e6d65 | 2154 | iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask); |
f76aec76 KA |
2155 | if (!iova) |
2156 | goto error; | |
2157 | ||
5b6985ce | 2158 | start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT; |
f76aec76 | 2159 | |
ba395927 KA |
2160 | /* |
2161 | * Check if DMAR supports zero-length reads on write only | |
2162 | * mappings.. | |
2163 | */ | |
2164 | if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \ | |
8c11e798 | 2165 | !cap_zlr(iommu->cap)) |
ba395927 KA |
2166 | prot |= DMA_PTE_READ; |
2167 | if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL) | |
2168 | prot |= DMA_PTE_WRITE; | |
2169 | /* | |
6865f0d1 | 2170 | * paddr - (paddr + size) might be partial page, we should map the whole |
ba395927 | 2171 | * page. Note: if two part of one page are separately mapped, we |
6865f0d1 | 2172 | * might have two guest_addr mapping to the same host paddr, but this |
ba395927 KA |
2173 | * is not a big problem |
2174 | */ | |
6865f0d1 | 2175 | ret = domain_page_mapping(domain, start_paddr, |
5b6985ce | 2176 | ((u64)paddr) & PAGE_MASK, size, prot); |
ba395927 KA |
2177 | if (ret) |
2178 | goto error; | |
2179 | ||
f76aec76 | 2180 | /* it's a non-present to present mapping */ |
8c11e798 | 2181 | ret = iommu_flush_iotlb_psi(iommu, domain->id, |
5b6985ce | 2182 | start_paddr, size >> VTD_PAGE_SHIFT, 1); |
f76aec76 | 2183 | if (ret) |
8c11e798 | 2184 | iommu_flush_write_buffer(iommu); |
f76aec76 | 2185 | |
5b6985ce | 2186 | return start_paddr + ((u64)paddr & (~PAGE_MASK)); |
ba395927 | 2187 | |
ba395927 | 2188 | error: |
f76aec76 KA |
2189 | if (iova) |
2190 | __free_iova(&domain->iovad, iova); | |
4cf2e75d | 2191 | printk(KERN_ERR"Device %s request: %zx@%llx dir %d --- failed\n", |
5b6985ce | 2192 | pci_name(pdev), size, (unsigned long long)paddr, dir); |
ba395927 KA |
2193 | return 0; |
2194 | } | |
2195 | ||
ffbbef5c FT |
2196 | static dma_addr_t intel_map_page(struct device *dev, struct page *page, |
2197 | unsigned long offset, size_t size, | |
2198 | enum dma_data_direction dir, | |
2199 | struct dma_attrs *attrs) | |
bb9e6d65 | 2200 | { |
ffbbef5c FT |
2201 | return __intel_map_single(dev, page_to_phys(page) + offset, size, |
2202 | dir, to_pci_dev(dev)->dma_mask); | |
bb9e6d65 FT |
2203 | } |
2204 | ||
5e0d2a6f | 2205 | static void flush_unmaps(void) |
2206 | { | |
80b20dd8 | 2207 | int i, j; |
5e0d2a6f | 2208 | |
5e0d2a6f | 2209 | timer_on = 0; |
2210 | ||
2211 | /* just flush them all */ | |
2212 | for (i = 0; i < g_num_of_iommus; i++) { | |
a2bb8459 WH |
2213 | struct intel_iommu *iommu = g_iommus[i]; |
2214 | if (!iommu) | |
2215 | continue; | |
c42d9f32 | 2216 | |
a2bb8459 | 2217 | if (deferred_flush[i].next) { |
a77b67d4 YS |
2218 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, |
2219 | DMA_TLB_GLOBAL_FLUSH, 0); | |
80b20dd8 | 2220 | for (j = 0; j < deferred_flush[i].next; j++) { |
2221 | __free_iova(&deferred_flush[i].domain[j]->iovad, | |
2222 | deferred_flush[i].iova[j]); | |
2223 | } | |
2224 | deferred_flush[i].next = 0; | |
2225 | } | |
5e0d2a6f | 2226 | } |
2227 | ||
5e0d2a6f | 2228 | list_size = 0; |
5e0d2a6f | 2229 | } |
2230 | ||
2231 | static void flush_unmaps_timeout(unsigned long data) | |
2232 | { | |
80b20dd8 | 2233 | unsigned long flags; |
2234 | ||
2235 | spin_lock_irqsave(&async_umap_flush_lock, flags); | |
5e0d2a6f | 2236 | flush_unmaps(); |
80b20dd8 | 2237 | spin_unlock_irqrestore(&async_umap_flush_lock, flags); |
5e0d2a6f | 2238 | } |
2239 | ||
2240 | static void add_unmap(struct dmar_domain *dom, struct iova *iova) | |
2241 | { | |
2242 | unsigned long flags; | |
80b20dd8 | 2243 | int next, iommu_id; |
8c11e798 | 2244 | struct intel_iommu *iommu; |
5e0d2a6f | 2245 | |
2246 | spin_lock_irqsave(&async_umap_flush_lock, flags); | |
80b20dd8 | 2247 | if (list_size == HIGH_WATER_MARK) |
2248 | flush_unmaps(); | |
2249 | ||
8c11e798 WH |
2250 | iommu = domain_get_iommu(dom); |
2251 | iommu_id = iommu->seq_id; | |
c42d9f32 | 2252 | |
80b20dd8 | 2253 | next = deferred_flush[iommu_id].next; |
2254 | deferred_flush[iommu_id].domain[next] = dom; | |
2255 | deferred_flush[iommu_id].iova[next] = iova; | |
2256 | deferred_flush[iommu_id].next++; | |
5e0d2a6f | 2257 | |
2258 | if (!timer_on) { | |
2259 | mod_timer(&unmap_timer, jiffies + msecs_to_jiffies(10)); | |
2260 | timer_on = 1; | |
2261 | } | |
2262 | list_size++; | |
2263 | spin_unlock_irqrestore(&async_umap_flush_lock, flags); | |
2264 | } | |
2265 | ||
ffbbef5c FT |
2266 | static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr, |
2267 | size_t size, enum dma_data_direction dir, | |
2268 | struct dma_attrs *attrs) | |
ba395927 | 2269 | { |
ba395927 | 2270 | struct pci_dev *pdev = to_pci_dev(dev); |
f76aec76 KA |
2271 | struct dmar_domain *domain; |
2272 | unsigned long start_addr; | |
ba395927 | 2273 | struct iova *iova; |
8c11e798 | 2274 | struct intel_iommu *iommu; |
ba395927 | 2275 | |
358dd8ac | 2276 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) |
f76aec76 | 2277 | return; |
ba395927 KA |
2278 | domain = find_domain(pdev); |
2279 | BUG_ON(!domain); | |
2280 | ||
8c11e798 WH |
2281 | iommu = domain_get_iommu(domain); |
2282 | ||
ba395927 | 2283 | iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr)); |
f76aec76 | 2284 | if (!iova) |
ba395927 | 2285 | return; |
ba395927 | 2286 | |
5b6985ce | 2287 | start_addr = iova->pfn_lo << PAGE_SHIFT; |
f76aec76 | 2288 | size = aligned_size((u64)dev_addr, size); |
ba395927 | 2289 | |
4cf2e75d | 2290 | pr_debug("Device %s unmapping: %zx@%llx\n", |
5b6985ce | 2291 | pci_name(pdev), size, (unsigned long long)start_addr); |
ba395927 | 2292 | |
f76aec76 KA |
2293 | /* clear the whole page */ |
2294 | dma_pte_clear_range(domain, start_addr, start_addr + size); | |
2295 | /* free page tables */ | |
2296 | dma_pte_free_pagetable(domain, start_addr, start_addr + size); | |
5e0d2a6f | 2297 | if (intel_iommu_strict) { |
8c11e798 | 2298 | if (iommu_flush_iotlb_psi(iommu, |
5b6985ce | 2299 | domain->id, start_addr, size >> VTD_PAGE_SHIFT, 0)) |
8c11e798 | 2300 | iommu_flush_write_buffer(iommu); |
5e0d2a6f | 2301 | /* free iova */ |
2302 | __free_iova(&domain->iovad, iova); | |
2303 | } else { | |
2304 | add_unmap(domain, iova); | |
2305 | /* | |
2306 | * queue up the release of the unmap to save the 1/6th of the | |
2307 | * cpu used up by the iotlb flush operation... | |
2308 | */ | |
5e0d2a6f | 2309 | } |
ba395927 KA |
2310 | } |
2311 | ||
d7ab5c46 FT |
2312 | static void intel_unmap_single(struct device *dev, dma_addr_t dev_addr, size_t size, |
2313 | int dir) | |
ffbbef5c FT |
2314 | { |
2315 | intel_unmap_page(dev, dev_addr, size, dir, NULL); | |
2316 | } | |
2317 | ||
d7ab5c46 FT |
2318 | static void *intel_alloc_coherent(struct device *hwdev, size_t size, |
2319 | dma_addr_t *dma_handle, gfp_t flags) | |
ba395927 KA |
2320 | { |
2321 | void *vaddr; | |
2322 | int order; | |
2323 | ||
5b6985ce | 2324 | size = PAGE_ALIGN(size); |
ba395927 KA |
2325 | order = get_order(size); |
2326 | flags &= ~(GFP_DMA | GFP_DMA32); | |
2327 | ||
2328 | vaddr = (void *)__get_free_pages(flags, order); | |
2329 | if (!vaddr) | |
2330 | return NULL; | |
2331 | memset(vaddr, 0, size); | |
2332 | ||
bb9e6d65 FT |
2333 | *dma_handle = __intel_map_single(hwdev, virt_to_bus(vaddr), size, |
2334 | DMA_BIDIRECTIONAL, | |
2335 | hwdev->coherent_dma_mask); | |
ba395927 KA |
2336 | if (*dma_handle) |
2337 | return vaddr; | |
2338 | free_pages((unsigned long)vaddr, order); | |
2339 | return NULL; | |
2340 | } | |
2341 | ||
d7ab5c46 FT |
2342 | static void intel_free_coherent(struct device *hwdev, size_t size, void *vaddr, |
2343 | dma_addr_t dma_handle) | |
ba395927 KA |
2344 | { |
2345 | int order; | |
2346 | ||
5b6985ce | 2347 | size = PAGE_ALIGN(size); |
ba395927 KA |
2348 | order = get_order(size); |
2349 | ||
2350 | intel_unmap_single(hwdev, dma_handle, size, DMA_BIDIRECTIONAL); | |
2351 | free_pages((unsigned long)vaddr, order); | |
2352 | } | |
2353 | ||
d7ab5c46 FT |
2354 | static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist, |
2355 | int nelems, enum dma_data_direction dir, | |
2356 | struct dma_attrs *attrs) | |
ba395927 KA |
2357 | { |
2358 | int i; | |
2359 | struct pci_dev *pdev = to_pci_dev(hwdev); | |
2360 | struct dmar_domain *domain; | |
f76aec76 KA |
2361 | unsigned long start_addr; |
2362 | struct iova *iova; | |
2363 | size_t size = 0; | |
4cf2e75d | 2364 | phys_addr_t addr; |
c03ab37c | 2365 | struct scatterlist *sg; |
8c11e798 | 2366 | struct intel_iommu *iommu; |
ba395927 | 2367 | |
358dd8ac | 2368 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) |
ba395927 KA |
2369 | return; |
2370 | ||
2371 | domain = find_domain(pdev); | |
8c11e798 WH |
2372 | BUG_ON(!domain); |
2373 | ||
2374 | iommu = domain_get_iommu(domain); | |
ba395927 | 2375 | |
c03ab37c | 2376 | iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address)); |
f76aec76 KA |
2377 | if (!iova) |
2378 | return; | |
c03ab37c | 2379 | for_each_sg(sglist, sg, nelems, i) { |
4cf2e75d | 2380 | addr = page_to_phys(sg_page(sg)) + sg->offset; |
f76aec76 KA |
2381 | size += aligned_size((u64)addr, sg->length); |
2382 | } | |
2383 | ||
5b6985ce | 2384 | start_addr = iova->pfn_lo << PAGE_SHIFT; |
f76aec76 KA |
2385 | |
2386 | /* clear the whole page */ | |
2387 | dma_pte_clear_range(domain, start_addr, start_addr + size); | |
2388 | /* free page tables */ | |
2389 | dma_pte_free_pagetable(domain, start_addr, start_addr + size); | |
2390 | ||
8c11e798 | 2391 | if (iommu_flush_iotlb_psi(iommu, domain->id, start_addr, |
5b6985ce | 2392 | size >> VTD_PAGE_SHIFT, 0)) |
8c11e798 | 2393 | iommu_flush_write_buffer(iommu); |
f76aec76 KA |
2394 | |
2395 | /* free iova */ | |
2396 | __free_iova(&domain->iovad, iova); | |
ba395927 KA |
2397 | } |
2398 | ||
ba395927 | 2399 | static int intel_nontranslate_map_sg(struct device *hddev, |
c03ab37c | 2400 | struct scatterlist *sglist, int nelems, int dir) |
ba395927 KA |
2401 | { |
2402 | int i; | |
c03ab37c | 2403 | struct scatterlist *sg; |
ba395927 | 2404 | |
c03ab37c | 2405 | for_each_sg(sglist, sg, nelems, i) { |
12d4d40e | 2406 | BUG_ON(!sg_page(sg)); |
4cf2e75d | 2407 | sg->dma_address = page_to_phys(sg_page(sg)) + sg->offset; |
c03ab37c | 2408 | sg->dma_length = sg->length; |
ba395927 KA |
2409 | } |
2410 | return nelems; | |
2411 | } | |
2412 | ||
d7ab5c46 FT |
2413 | static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems, |
2414 | enum dma_data_direction dir, struct dma_attrs *attrs) | |
ba395927 | 2415 | { |
4cf2e75d | 2416 | phys_addr_t addr; |
ba395927 | 2417 | int i; |
ba395927 KA |
2418 | struct pci_dev *pdev = to_pci_dev(hwdev); |
2419 | struct dmar_domain *domain; | |
f76aec76 KA |
2420 | size_t size = 0; |
2421 | int prot = 0; | |
2422 | size_t offset = 0; | |
2423 | struct iova *iova = NULL; | |
2424 | int ret; | |
c03ab37c | 2425 | struct scatterlist *sg; |
f76aec76 | 2426 | unsigned long start_addr; |
8c11e798 | 2427 | struct intel_iommu *iommu; |
ba395927 KA |
2428 | |
2429 | BUG_ON(dir == DMA_NONE); | |
358dd8ac | 2430 | if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO) |
c03ab37c | 2431 | return intel_nontranslate_map_sg(hwdev, sglist, nelems, dir); |
ba395927 | 2432 | |
f76aec76 KA |
2433 | domain = get_valid_domain_for_dev(pdev); |
2434 | if (!domain) | |
2435 | return 0; | |
2436 | ||
8c11e798 WH |
2437 | iommu = domain_get_iommu(domain); |
2438 | ||
c03ab37c | 2439 | for_each_sg(sglist, sg, nelems, i) { |
4cf2e75d | 2440 | addr = page_to_phys(sg_page(sg)) + sg->offset; |
f76aec76 KA |
2441 | size += aligned_size((u64)addr, sg->length); |
2442 | } | |
2443 | ||
bb9e6d65 | 2444 | iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask); |
f76aec76 | 2445 | if (!iova) { |
c03ab37c | 2446 | sglist->dma_length = 0; |
f76aec76 KA |
2447 | return 0; |
2448 | } | |
2449 | ||
2450 | /* | |
2451 | * Check if DMAR supports zero-length reads on write only | |
2452 | * mappings.. | |
2453 | */ | |
2454 | if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \ | |
8c11e798 | 2455 | !cap_zlr(iommu->cap)) |
f76aec76 KA |
2456 | prot |= DMA_PTE_READ; |
2457 | if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL) | |
2458 | prot |= DMA_PTE_WRITE; | |
2459 | ||
5b6985ce | 2460 | start_addr = iova->pfn_lo << PAGE_SHIFT; |
f76aec76 | 2461 | offset = 0; |
c03ab37c | 2462 | for_each_sg(sglist, sg, nelems, i) { |
4cf2e75d | 2463 | addr = page_to_phys(sg_page(sg)) + sg->offset; |
f76aec76 KA |
2464 | size = aligned_size((u64)addr, sg->length); |
2465 | ret = domain_page_mapping(domain, start_addr + offset, | |
5b6985ce | 2466 | ((u64)addr) & PAGE_MASK, |
f76aec76 KA |
2467 | size, prot); |
2468 | if (ret) { | |
2469 | /* clear the page */ | |
2470 | dma_pte_clear_range(domain, start_addr, | |
2471 | start_addr + offset); | |
2472 | /* free page tables */ | |
2473 | dma_pte_free_pagetable(domain, start_addr, | |
2474 | start_addr + offset); | |
2475 | /* free iova */ | |
2476 | __free_iova(&domain->iovad, iova); | |
ba395927 KA |
2477 | return 0; |
2478 | } | |
f76aec76 | 2479 | sg->dma_address = start_addr + offset + |
5b6985ce | 2480 | ((u64)addr & (~PAGE_MASK)); |
ba395927 | 2481 | sg->dma_length = sg->length; |
f76aec76 | 2482 | offset += size; |
ba395927 KA |
2483 | } |
2484 | ||
ba395927 | 2485 | /* it's a non-present to present mapping */ |
8c11e798 | 2486 | if (iommu_flush_iotlb_psi(iommu, domain->id, |
5b6985ce | 2487 | start_addr, offset >> VTD_PAGE_SHIFT, 1)) |
8c11e798 | 2488 | iommu_flush_write_buffer(iommu); |
ba395927 KA |
2489 | return nelems; |
2490 | } | |
2491 | ||
dfb805e8 FT |
2492 | static int intel_mapping_error(struct device *dev, dma_addr_t dma_addr) |
2493 | { | |
2494 | return !dma_addr; | |
2495 | } | |
2496 | ||
160c1d8e | 2497 | struct dma_map_ops intel_dma_ops = { |
ba395927 KA |
2498 | .alloc_coherent = intel_alloc_coherent, |
2499 | .free_coherent = intel_free_coherent, | |
ba395927 KA |
2500 | .map_sg = intel_map_sg, |
2501 | .unmap_sg = intel_unmap_sg, | |
ffbbef5c FT |
2502 | .map_page = intel_map_page, |
2503 | .unmap_page = intel_unmap_page, | |
dfb805e8 | 2504 | .mapping_error = intel_mapping_error, |
ba395927 KA |
2505 | }; |
2506 | ||
2507 | static inline int iommu_domain_cache_init(void) | |
2508 | { | |
2509 | int ret = 0; | |
2510 | ||
2511 | iommu_domain_cache = kmem_cache_create("iommu_domain", | |
2512 | sizeof(struct dmar_domain), | |
2513 | 0, | |
2514 | SLAB_HWCACHE_ALIGN, | |
2515 | ||
2516 | NULL); | |
2517 | if (!iommu_domain_cache) { | |
2518 | printk(KERN_ERR "Couldn't create iommu_domain cache\n"); | |
2519 | ret = -ENOMEM; | |
2520 | } | |
2521 | ||
2522 | return ret; | |
2523 | } | |
2524 | ||
2525 | static inline int iommu_devinfo_cache_init(void) | |
2526 | { | |
2527 | int ret = 0; | |
2528 | ||
2529 | iommu_devinfo_cache = kmem_cache_create("iommu_devinfo", | |
2530 | sizeof(struct device_domain_info), | |
2531 | 0, | |
2532 | SLAB_HWCACHE_ALIGN, | |
ba395927 KA |
2533 | NULL); |
2534 | if (!iommu_devinfo_cache) { | |
2535 | printk(KERN_ERR "Couldn't create devinfo cache\n"); | |
2536 | ret = -ENOMEM; | |
2537 | } | |
2538 | ||
2539 | return ret; | |
2540 | } | |
2541 | ||
2542 | static inline int iommu_iova_cache_init(void) | |
2543 | { | |
2544 | int ret = 0; | |
2545 | ||
2546 | iommu_iova_cache = kmem_cache_create("iommu_iova", | |
2547 | sizeof(struct iova), | |
2548 | 0, | |
2549 | SLAB_HWCACHE_ALIGN, | |
ba395927 KA |
2550 | NULL); |
2551 | if (!iommu_iova_cache) { | |
2552 | printk(KERN_ERR "Couldn't create iova cache\n"); | |
2553 | ret = -ENOMEM; | |
2554 | } | |
2555 | ||
2556 | return ret; | |
2557 | } | |
2558 | ||
2559 | static int __init iommu_init_mempool(void) | |
2560 | { | |
2561 | int ret; | |
2562 | ret = iommu_iova_cache_init(); | |
2563 | if (ret) | |
2564 | return ret; | |
2565 | ||
2566 | ret = iommu_domain_cache_init(); | |
2567 | if (ret) | |
2568 | goto domain_error; | |
2569 | ||
2570 | ret = iommu_devinfo_cache_init(); | |
2571 | if (!ret) | |
2572 | return ret; | |
2573 | ||
2574 | kmem_cache_destroy(iommu_domain_cache); | |
2575 | domain_error: | |
2576 | kmem_cache_destroy(iommu_iova_cache); | |
2577 | ||
2578 | return -ENOMEM; | |
2579 | } | |
2580 | ||
2581 | static void __init iommu_exit_mempool(void) | |
2582 | { | |
2583 | kmem_cache_destroy(iommu_devinfo_cache); | |
2584 | kmem_cache_destroy(iommu_domain_cache); | |
2585 | kmem_cache_destroy(iommu_iova_cache); | |
2586 | ||
2587 | } | |
2588 | ||
ba395927 KA |
2589 | static void __init init_no_remapping_devices(void) |
2590 | { | |
2591 | struct dmar_drhd_unit *drhd; | |
2592 | ||
2593 | for_each_drhd_unit(drhd) { | |
2594 | if (!drhd->include_all) { | |
2595 | int i; | |
2596 | for (i = 0; i < drhd->devices_cnt; i++) | |
2597 | if (drhd->devices[i] != NULL) | |
2598 | break; | |
2599 | /* ignore DMAR unit if no pci devices exist */ | |
2600 | if (i == drhd->devices_cnt) | |
2601 | drhd->ignored = 1; | |
2602 | } | |
2603 | } | |
2604 | ||
2605 | if (dmar_map_gfx) | |
2606 | return; | |
2607 | ||
2608 | for_each_drhd_unit(drhd) { | |
2609 | int i; | |
2610 | if (drhd->ignored || drhd->include_all) | |
2611 | continue; | |
2612 | ||
2613 | for (i = 0; i < drhd->devices_cnt; i++) | |
2614 | if (drhd->devices[i] && | |
2615 | !IS_GFX_DEVICE(drhd->devices[i])) | |
2616 | break; | |
2617 | ||
2618 | if (i < drhd->devices_cnt) | |
2619 | continue; | |
2620 | ||
2621 | /* bypass IOMMU if it is just for gfx devices */ | |
2622 | drhd->ignored = 1; | |
2623 | for (i = 0; i < drhd->devices_cnt; i++) { | |
2624 | if (!drhd->devices[i]) | |
2625 | continue; | |
358dd8ac | 2626 | drhd->devices[i]->dev.archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO; |
ba395927 KA |
2627 | } |
2628 | } | |
2629 | } | |
2630 | ||
f59c7b69 FY |
2631 | #ifdef CONFIG_SUSPEND |
2632 | static int init_iommu_hw(void) | |
2633 | { | |
2634 | struct dmar_drhd_unit *drhd; | |
2635 | struct intel_iommu *iommu = NULL; | |
2636 | ||
2637 | for_each_active_iommu(iommu, drhd) | |
2638 | if (iommu->qi) | |
2639 | dmar_reenable_qi(iommu); | |
2640 | ||
2641 | for_each_active_iommu(iommu, drhd) { | |
2642 | iommu_flush_write_buffer(iommu); | |
2643 | ||
2644 | iommu_set_root_entry(iommu); | |
2645 | ||
2646 | iommu->flush.flush_context(iommu, 0, 0, 0, | |
2647 | DMA_CCMD_GLOBAL_INVL, 0); | |
2648 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, | |
2649 | DMA_TLB_GLOBAL_FLUSH, 0); | |
2650 | iommu_disable_protect_mem_regions(iommu); | |
2651 | iommu_enable_translation(iommu); | |
2652 | } | |
2653 | ||
2654 | return 0; | |
2655 | } | |
2656 | ||
2657 | static void iommu_flush_all(void) | |
2658 | { | |
2659 | struct dmar_drhd_unit *drhd; | |
2660 | struct intel_iommu *iommu; | |
2661 | ||
2662 | for_each_active_iommu(iommu, drhd) { | |
2663 | iommu->flush.flush_context(iommu, 0, 0, 0, | |
2664 | DMA_CCMD_GLOBAL_INVL, 0); | |
2665 | iommu->flush.flush_iotlb(iommu, 0, 0, 0, | |
2666 | DMA_TLB_GLOBAL_FLUSH, 0); | |
2667 | } | |
2668 | } | |
2669 | ||
2670 | static int iommu_suspend(struct sys_device *dev, pm_message_t state) | |
2671 | { | |
2672 | struct dmar_drhd_unit *drhd; | |
2673 | struct intel_iommu *iommu = NULL; | |
2674 | unsigned long flag; | |
2675 | ||
2676 | for_each_active_iommu(iommu, drhd) { | |
2677 | iommu->iommu_state = kzalloc(sizeof(u32) * MAX_SR_DMAR_REGS, | |
2678 | GFP_ATOMIC); | |
2679 | if (!iommu->iommu_state) | |
2680 | goto nomem; | |
2681 | } | |
2682 | ||
2683 | iommu_flush_all(); | |
2684 | ||
2685 | for_each_active_iommu(iommu, drhd) { | |
2686 | iommu_disable_translation(iommu); | |
2687 | ||
2688 | spin_lock_irqsave(&iommu->register_lock, flag); | |
2689 | ||
2690 | iommu->iommu_state[SR_DMAR_FECTL_REG] = | |
2691 | readl(iommu->reg + DMAR_FECTL_REG); | |
2692 | iommu->iommu_state[SR_DMAR_FEDATA_REG] = | |
2693 | readl(iommu->reg + DMAR_FEDATA_REG); | |
2694 | iommu->iommu_state[SR_DMAR_FEADDR_REG] = | |
2695 | readl(iommu->reg + DMAR_FEADDR_REG); | |
2696 | iommu->iommu_state[SR_DMAR_FEUADDR_REG] = | |
2697 | readl(iommu->reg + DMAR_FEUADDR_REG); | |
2698 | ||
2699 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
2700 | } | |
2701 | return 0; | |
2702 | ||
2703 | nomem: | |
2704 | for_each_active_iommu(iommu, drhd) | |
2705 | kfree(iommu->iommu_state); | |
2706 | ||
2707 | return -ENOMEM; | |
2708 | } | |
2709 | ||
2710 | static int iommu_resume(struct sys_device *dev) | |
2711 | { | |
2712 | struct dmar_drhd_unit *drhd; | |
2713 | struct intel_iommu *iommu = NULL; | |
2714 | unsigned long flag; | |
2715 | ||
2716 | if (init_iommu_hw()) { | |
2717 | WARN(1, "IOMMU setup failed, DMAR can not resume!\n"); | |
2718 | return -EIO; | |
2719 | } | |
2720 | ||
2721 | for_each_active_iommu(iommu, drhd) { | |
2722 | ||
2723 | spin_lock_irqsave(&iommu->register_lock, flag); | |
2724 | ||
2725 | writel(iommu->iommu_state[SR_DMAR_FECTL_REG], | |
2726 | iommu->reg + DMAR_FECTL_REG); | |
2727 | writel(iommu->iommu_state[SR_DMAR_FEDATA_REG], | |
2728 | iommu->reg + DMAR_FEDATA_REG); | |
2729 | writel(iommu->iommu_state[SR_DMAR_FEADDR_REG], | |
2730 | iommu->reg + DMAR_FEADDR_REG); | |
2731 | writel(iommu->iommu_state[SR_DMAR_FEUADDR_REG], | |
2732 | iommu->reg + DMAR_FEUADDR_REG); | |
2733 | ||
2734 | spin_unlock_irqrestore(&iommu->register_lock, flag); | |
2735 | } | |
2736 | ||
2737 | for_each_active_iommu(iommu, drhd) | |
2738 | kfree(iommu->iommu_state); | |
2739 | ||
2740 | return 0; | |
2741 | } | |
2742 | ||
2743 | static struct sysdev_class iommu_sysclass = { | |
2744 | .name = "iommu", | |
2745 | .resume = iommu_resume, | |
2746 | .suspend = iommu_suspend, | |
2747 | }; | |
2748 | ||
2749 | static struct sys_device device_iommu = { | |
2750 | .cls = &iommu_sysclass, | |
2751 | }; | |
2752 | ||
2753 | static int __init init_iommu_sysfs(void) | |
2754 | { | |
2755 | int error; | |
2756 | ||
2757 | error = sysdev_class_register(&iommu_sysclass); | |
2758 | if (error) | |
2759 | return error; | |
2760 | ||
2761 | error = sysdev_register(&device_iommu); | |
2762 | if (error) | |
2763 | sysdev_class_unregister(&iommu_sysclass); | |
2764 | ||
2765 | return error; | |
2766 | } | |
2767 | ||
2768 | #else | |
2769 | static int __init init_iommu_sysfs(void) | |
2770 | { | |
2771 | return 0; | |
2772 | } | |
2773 | #endif /* CONFIG_PM */ | |
2774 | ||
ba395927 KA |
2775 | int __init intel_iommu_init(void) |
2776 | { | |
2777 | int ret = 0; | |
2778 | ||
ba395927 KA |
2779 | if (dmar_table_init()) |
2780 | return -ENODEV; | |
2781 | ||
1886e8a9 SS |
2782 | if (dmar_dev_scope_init()) |
2783 | return -ENODEV; | |
2784 | ||
2ae21010 SS |
2785 | /* |
2786 | * Check the need for DMA-remapping initialization now. | |
2787 | * Above initialization will also be used by Interrupt-remapping. | |
2788 | */ | |
2789 | if (no_iommu || swiotlb || dmar_disabled) | |
2790 | return -ENODEV; | |
2791 | ||
ba395927 KA |
2792 | iommu_init_mempool(); |
2793 | dmar_init_reserved_ranges(); | |
2794 | ||
2795 | init_no_remapping_devices(); | |
2796 | ||
2797 | ret = init_dmars(); | |
2798 | if (ret) { | |
2799 | printk(KERN_ERR "IOMMU: dmar init failed\n"); | |
2800 | put_iova_domain(&reserved_iova_list); | |
2801 | iommu_exit_mempool(); | |
2802 | return ret; | |
2803 | } | |
2804 | printk(KERN_INFO | |
2805 | "PCI-DMA: Intel(R) Virtualization Technology for Directed I/O\n"); | |
2806 | ||
5e0d2a6f | 2807 | init_timer(&unmap_timer); |
ba395927 KA |
2808 | force_iommu = 1; |
2809 | dma_ops = &intel_dma_ops; | |
f59c7b69 | 2810 | init_iommu_sysfs(); |
a8bcbb0d JR |
2811 | |
2812 | register_iommu(&intel_iommu_ops); | |
2813 | ||
ba395927 KA |
2814 | return 0; |
2815 | } | |
e820482c | 2816 | |
c7151a8d WH |
2817 | static int vm_domain_add_dev_info(struct dmar_domain *domain, |
2818 | struct pci_dev *pdev) | |
2819 | { | |
2820 | struct device_domain_info *info; | |
2821 | unsigned long flags; | |
2822 | ||
2823 | info = alloc_devinfo_mem(); | |
2824 | if (!info) | |
2825 | return -ENOMEM; | |
2826 | ||
276dbf99 | 2827 | info->segment = pci_domain_nr(pdev->bus); |
c7151a8d WH |
2828 | info->bus = pdev->bus->number; |
2829 | info->devfn = pdev->devfn; | |
2830 | info->dev = pdev; | |
2831 | info->domain = domain; | |
2832 | ||
2833 | spin_lock_irqsave(&device_domain_lock, flags); | |
2834 | list_add(&info->link, &domain->devices); | |
2835 | list_add(&info->global, &device_domain_list); | |
2836 | pdev->dev.archdata.iommu = info; | |
2837 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
2838 | ||
2839 | return 0; | |
2840 | } | |
2841 | ||
3199aa6b HW |
2842 | static void iommu_detach_dependent_devices(struct intel_iommu *iommu, |
2843 | struct pci_dev *pdev) | |
2844 | { | |
2845 | struct pci_dev *tmp, *parent; | |
2846 | ||
2847 | if (!iommu || !pdev) | |
2848 | return; | |
2849 | ||
2850 | /* dependent device detach */ | |
2851 | tmp = pci_find_upstream_pcie_bridge(pdev); | |
2852 | /* Secondary interface's bus number and devfn 0 */ | |
2853 | if (tmp) { | |
2854 | parent = pdev->bus->self; | |
2855 | while (parent != tmp) { | |
2856 | iommu_detach_dev(iommu, parent->bus->number, | |
276dbf99 | 2857 | parent->devfn); |
3199aa6b HW |
2858 | parent = parent->bus->self; |
2859 | } | |
2860 | if (tmp->is_pcie) /* this is a PCIE-to-PCI bridge */ | |
2861 | iommu_detach_dev(iommu, | |
2862 | tmp->subordinate->number, 0); | |
2863 | else /* this is a legacy PCI bridge */ | |
276dbf99 DW |
2864 | iommu_detach_dev(iommu, tmp->bus->number, |
2865 | tmp->devfn); | |
3199aa6b HW |
2866 | } |
2867 | } | |
2868 | ||
c7151a8d WH |
2869 | static void vm_domain_remove_one_dev_info(struct dmar_domain *domain, |
2870 | struct pci_dev *pdev) | |
2871 | { | |
2872 | struct device_domain_info *info; | |
2873 | struct intel_iommu *iommu; | |
2874 | unsigned long flags; | |
2875 | int found = 0; | |
2876 | struct list_head *entry, *tmp; | |
2877 | ||
276dbf99 DW |
2878 | iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number, |
2879 | pdev->devfn); | |
c7151a8d WH |
2880 | if (!iommu) |
2881 | return; | |
2882 | ||
2883 | spin_lock_irqsave(&device_domain_lock, flags); | |
2884 | list_for_each_safe(entry, tmp, &domain->devices) { | |
2885 | info = list_entry(entry, struct device_domain_info, link); | |
276dbf99 | 2886 | /* No need to compare PCI domain; it has to be the same */ |
c7151a8d WH |
2887 | if (info->bus == pdev->bus->number && |
2888 | info->devfn == pdev->devfn) { | |
2889 | list_del(&info->link); | |
2890 | list_del(&info->global); | |
2891 | if (info->dev) | |
2892 | info->dev->dev.archdata.iommu = NULL; | |
2893 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
2894 | ||
2895 | iommu_detach_dev(iommu, info->bus, info->devfn); | |
3199aa6b | 2896 | iommu_detach_dependent_devices(iommu, pdev); |
c7151a8d WH |
2897 | free_devinfo_mem(info); |
2898 | ||
2899 | spin_lock_irqsave(&device_domain_lock, flags); | |
2900 | ||
2901 | if (found) | |
2902 | break; | |
2903 | else | |
2904 | continue; | |
2905 | } | |
2906 | ||
2907 | /* if there is no other devices under the same iommu | |
2908 | * owned by this domain, clear this iommu in iommu_bmp | |
2909 | * update iommu count and coherency | |
2910 | */ | |
276dbf99 DW |
2911 | if (iommu == device_to_iommu(info->segment, info->bus, |
2912 | info->devfn)) | |
c7151a8d WH |
2913 | found = 1; |
2914 | } | |
2915 | ||
2916 | if (found == 0) { | |
2917 | unsigned long tmp_flags; | |
2918 | spin_lock_irqsave(&domain->iommu_lock, tmp_flags); | |
2919 | clear_bit(iommu->seq_id, &domain->iommu_bmp); | |
2920 | domain->iommu_count--; | |
58c610bd | 2921 | domain_update_iommu_cap(domain); |
c7151a8d WH |
2922 | spin_unlock_irqrestore(&domain->iommu_lock, tmp_flags); |
2923 | } | |
2924 | ||
2925 | spin_unlock_irqrestore(&device_domain_lock, flags); | |
2926 | } | |
2927 | ||
2928 | static void vm_domain_remove_all_dev_info(struct dmar_domain *domain) | |
2929 | { | |
2930 | struct device_domain_info *info; | |
2931 | struct intel_iommu *iommu; | |
2932 | unsigned long flags1, flags2; | |
2933 | ||
2934 | spin_lock_irqsave(&device_domain_lock, flags1); | |
2935 | while (!list_empty(&domain->devices)) { | |
2936 | info = list_entry(domain->devices.next, | |
2937 | struct device_domain_info, link); | |
2938 | list_del(&info->link); | |
2939 | list_del(&info->global); | |
2940 | if (info->dev) | |
2941 | info->dev->dev.archdata.iommu = NULL; | |
2942 | ||
2943 | spin_unlock_irqrestore(&device_domain_lock, flags1); | |
2944 | ||
276dbf99 | 2945 | iommu = device_to_iommu(info->segment, info->bus, info->devfn); |
c7151a8d | 2946 | iommu_detach_dev(iommu, info->bus, info->devfn); |
3199aa6b | 2947 | iommu_detach_dependent_devices(iommu, info->dev); |
c7151a8d WH |
2948 | |
2949 | /* clear this iommu in iommu_bmp, update iommu count | |
58c610bd | 2950 | * and capabilities |
c7151a8d WH |
2951 | */ |
2952 | spin_lock_irqsave(&domain->iommu_lock, flags2); | |
2953 | if (test_and_clear_bit(iommu->seq_id, | |
2954 | &domain->iommu_bmp)) { | |
2955 | domain->iommu_count--; | |
58c610bd | 2956 | domain_update_iommu_cap(domain); |
c7151a8d WH |
2957 | } |
2958 | spin_unlock_irqrestore(&domain->iommu_lock, flags2); | |
2959 | ||
2960 | free_devinfo_mem(info); | |
2961 | spin_lock_irqsave(&device_domain_lock, flags1); | |
2962 | } | |
2963 | spin_unlock_irqrestore(&device_domain_lock, flags1); | |
2964 | } | |
2965 | ||
5e98c4b1 WH |
2966 | /* domain id for virtual machine, it won't be set in context */ |
2967 | static unsigned long vm_domid; | |
2968 | ||
fe40f1e0 WH |
2969 | static int vm_domain_min_agaw(struct dmar_domain *domain) |
2970 | { | |
2971 | int i; | |
2972 | int min_agaw = domain->agaw; | |
2973 | ||
2974 | i = find_first_bit(&domain->iommu_bmp, g_num_of_iommus); | |
2975 | for (; i < g_num_of_iommus; ) { | |
2976 | if (min_agaw > g_iommus[i]->agaw) | |
2977 | min_agaw = g_iommus[i]->agaw; | |
2978 | ||
2979 | i = find_next_bit(&domain->iommu_bmp, g_num_of_iommus, i+1); | |
2980 | } | |
2981 | ||
2982 | return min_agaw; | |
2983 | } | |
2984 | ||
5e98c4b1 WH |
2985 | static struct dmar_domain *iommu_alloc_vm_domain(void) |
2986 | { | |
2987 | struct dmar_domain *domain; | |
2988 | ||
2989 | domain = alloc_domain_mem(); | |
2990 | if (!domain) | |
2991 | return NULL; | |
2992 | ||
2993 | domain->id = vm_domid++; | |
2994 | memset(&domain->iommu_bmp, 0, sizeof(unsigned long)); | |
2995 | domain->flags = DOMAIN_FLAG_VIRTUAL_MACHINE; | |
2996 | ||
2997 | return domain; | |
2998 | } | |
2999 | ||
3000 | static int vm_domain_init(struct dmar_domain *domain, int guest_width) | |
3001 | { | |
3002 | int adjust_width; | |
3003 | ||
3004 | init_iova_domain(&domain->iovad, DMA_32BIT_PFN); | |
3005 | spin_lock_init(&domain->mapping_lock); | |
3006 | spin_lock_init(&domain->iommu_lock); | |
3007 | ||
3008 | domain_reserve_special_ranges(domain); | |
3009 | ||
3010 | /* calculate AGAW */ | |
3011 | domain->gaw = guest_width; | |
3012 | adjust_width = guestwidth_to_adjustwidth(guest_width); | |
3013 | domain->agaw = width_to_agaw(adjust_width); | |
3014 | ||
3015 | INIT_LIST_HEAD(&domain->devices); | |
3016 | ||
3017 | domain->iommu_count = 0; | |
3018 | domain->iommu_coherency = 0; | |
fe40f1e0 | 3019 | domain->max_addr = 0; |
5e98c4b1 WH |
3020 | |
3021 | /* always allocate the top pgd */ | |
3022 | domain->pgd = (struct dma_pte *)alloc_pgtable_page(); | |
3023 | if (!domain->pgd) | |
3024 | return -ENOMEM; | |
3025 | domain_flush_cache(domain, domain->pgd, PAGE_SIZE); | |
3026 | return 0; | |
3027 | } | |
3028 | ||
3029 | static void iommu_free_vm_domain(struct dmar_domain *domain) | |
3030 | { | |
3031 | unsigned long flags; | |
3032 | struct dmar_drhd_unit *drhd; | |
3033 | struct intel_iommu *iommu; | |
3034 | unsigned long i; | |
3035 | unsigned long ndomains; | |
3036 | ||
3037 | for_each_drhd_unit(drhd) { | |
3038 | if (drhd->ignored) | |
3039 | continue; | |
3040 | iommu = drhd->iommu; | |
3041 | ||
3042 | ndomains = cap_ndoms(iommu->cap); | |
3043 | i = find_first_bit(iommu->domain_ids, ndomains); | |
3044 | for (; i < ndomains; ) { | |
3045 | if (iommu->domains[i] == domain) { | |
3046 | spin_lock_irqsave(&iommu->lock, flags); | |
3047 | clear_bit(i, iommu->domain_ids); | |
3048 | iommu->domains[i] = NULL; | |
3049 | spin_unlock_irqrestore(&iommu->lock, flags); | |
3050 | break; | |
3051 | } | |
3052 | i = find_next_bit(iommu->domain_ids, ndomains, i+1); | |
3053 | } | |
3054 | } | |
3055 | } | |
3056 | ||
3057 | static void vm_domain_exit(struct dmar_domain *domain) | |
3058 | { | |
3059 | u64 end; | |
3060 | ||
3061 | /* Domain 0 is reserved, so dont process it */ | |
3062 | if (!domain) | |
3063 | return; | |
3064 | ||
3065 | vm_domain_remove_all_dev_info(domain); | |
3066 | /* destroy iovas */ | |
3067 | put_iova_domain(&domain->iovad); | |
3068 | end = DOMAIN_MAX_ADDR(domain->gaw); | |
3069 | end = end & (~VTD_PAGE_MASK); | |
3070 | ||
3071 | /* clear ptes */ | |
3072 | dma_pte_clear_range(domain, 0, end); | |
3073 | ||
3074 | /* free page tables */ | |
3075 | dma_pte_free_pagetable(domain, 0, end); | |
3076 | ||
3077 | iommu_free_vm_domain(domain); | |
3078 | free_domain_mem(domain); | |
3079 | } | |
3080 | ||
5d450806 | 3081 | static int intel_iommu_domain_init(struct iommu_domain *domain) |
38717946 | 3082 | { |
5d450806 | 3083 | struct dmar_domain *dmar_domain; |
38717946 | 3084 | |
5d450806 JR |
3085 | dmar_domain = iommu_alloc_vm_domain(); |
3086 | if (!dmar_domain) { | |
38717946 | 3087 | printk(KERN_ERR |
5d450806 JR |
3088 | "intel_iommu_domain_init: dmar_domain == NULL\n"); |
3089 | return -ENOMEM; | |
38717946 | 3090 | } |
5d450806 | 3091 | if (vm_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) { |
38717946 | 3092 | printk(KERN_ERR |
5d450806 JR |
3093 | "intel_iommu_domain_init() failed\n"); |
3094 | vm_domain_exit(dmar_domain); | |
3095 | return -ENOMEM; | |
38717946 | 3096 | } |
5d450806 | 3097 | domain->priv = dmar_domain; |
faa3d6f5 | 3098 | |
5d450806 | 3099 | return 0; |
38717946 | 3100 | } |
38717946 | 3101 | |
5d450806 | 3102 | static void intel_iommu_domain_destroy(struct iommu_domain *domain) |
38717946 | 3103 | { |
5d450806 JR |
3104 | struct dmar_domain *dmar_domain = domain->priv; |
3105 | ||
3106 | domain->priv = NULL; | |
3107 | vm_domain_exit(dmar_domain); | |
38717946 | 3108 | } |
38717946 | 3109 | |
4c5478c9 JR |
3110 | static int intel_iommu_attach_device(struct iommu_domain *domain, |
3111 | struct device *dev) | |
38717946 | 3112 | { |
4c5478c9 JR |
3113 | struct dmar_domain *dmar_domain = domain->priv; |
3114 | struct pci_dev *pdev = to_pci_dev(dev); | |
fe40f1e0 WH |
3115 | struct intel_iommu *iommu; |
3116 | int addr_width; | |
3117 | u64 end; | |
faa3d6f5 WH |
3118 | int ret; |
3119 | ||
3120 | /* normally pdev is not mapped */ | |
3121 | if (unlikely(domain_context_mapped(pdev))) { | |
3122 | struct dmar_domain *old_domain; | |
3123 | ||
3124 | old_domain = find_domain(pdev); | |
3125 | if (old_domain) { | |
4c5478c9 | 3126 | if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) |
faa3d6f5 WH |
3127 | vm_domain_remove_one_dev_info(old_domain, pdev); |
3128 | else | |
3129 | domain_remove_dev_info(old_domain); | |
3130 | } | |
3131 | } | |
3132 | ||
276dbf99 DW |
3133 | iommu = device_to_iommu(pci_domain_nr(pdev->bus), pdev->bus->number, |
3134 | pdev->devfn); | |
fe40f1e0 WH |
3135 | if (!iommu) |
3136 | return -ENODEV; | |
3137 | ||
3138 | /* check if this iommu agaw is sufficient for max mapped address */ | |
3139 | addr_width = agaw_to_width(iommu->agaw); | |
3140 | end = DOMAIN_MAX_ADDR(addr_width); | |
3141 | end = end & VTD_PAGE_MASK; | |
4c5478c9 | 3142 | if (end < dmar_domain->max_addr) { |
fe40f1e0 WH |
3143 | printk(KERN_ERR "%s: iommu agaw (%d) is not " |
3144 | "sufficient for the mapped address (%llx)\n", | |
4c5478c9 | 3145 | __func__, iommu->agaw, dmar_domain->max_addr); |
fe40f1e0 WH |
3146 | return -EFAULT; |
3147 | } | |
3148 | ||
4c5478c9 | 3149 | ret = domain_context_mapping(dmar_domain, pdev); |
faa3d6f5 WH |
3150 | if (ret) |
3151 | return ret; | |
3152 | ||
4c5478c9 | 3153 | ret = vm_domain_add_dev_info(dmar_domain, pdev); |
faa3d6f5 | 3154 | return ret; |
38717946 | 3155 | } |
38717946 | 3156 | |
4c5478c9 JR |
3157 | static void intel_iommu_detach_device(struct iommu_domain *domain, |
3158 | struct device *dev) | |
38717946 | 3159 | { |
4c5478c9 JR |
3160 | struct dmar_domain *dmar_domain = domain->priv; |
3161 | struct pci_dev *pdev = to_pci_dev(dev); | |
3162 | ||
3163 | vm_domain_remove_one_dev_info(dmar_domain, pdev); | |
faa3d6f5 | 3164 | } |
c7151a8d | 3165 | |
dde57a21 JR |
3166 | static int intel_iommu_map_range(struct iommu_domain *domain, |
3167 | unsigned long iova, phys_addr_t hpa, | |
3168 | size_t size, int iommu_prot) | |
faa3d6f5 | 3169 | { |
dde57a21 | 3170 | struct dmar_domain *dmar_domain = domain->priv; |
fe40f1e0 WH |
3171 | u64 max_addr; |
3172 | int addr_width; | |
dde57a21 | 3173 | int prot = 0; |
faa3d6f5 | 3174 | int ret; |
fe40f1e0 | 3175 | |
dde57a21 JR |
3176 | if (iommu_prot & IOMMU_READ) |
3177 | prot |= DMA_PTE_READ; | |
3178 | if (iommu_prot & IOMMU_WRITE) | |
3179 | prot |= DMA_PTE_WRITE; | |
9cf06697 SY |
3180 | if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping) |
3181 | prot |= DMA_PTE_SNP; | |
dde57a21 | 3182 | |
fe40f1e0 | 3183 | max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size); |
dde57a21 | 3184 | if (dmar_domain->max_addr < max_addr) { |
fe40f1e0 WH |
3185 | int min_agaw; |
3186 | u64 end; | |
3187 | ||
3188 | /* check if minimum agaw is sufficient for mapped address */ | |
dde57a21 | 3189 | min_agaw = vm_domain_min_agaw(dmar_domain); |
fe40f1e0 WH |
3190 | addr_width = agaw_to_width(min_agaw); |
3191 | end = DOMAIN_MAX_ADDR(addr_width); | |
3192 | end = end & VTD_PAGE_MASK; | |
3193 | if (end < max_addr) { | |
3194 | printk(KERN_ERR "%s: iommu agaw (%d) is not " | |
3195 | "sufficient for the mapped address (%llx)\n", | |
3196 | __func__, min_agaw, max_addr); | |
3197 | return -EFAULT; | |
3198 | } | |
dde57a21 | 3199 | dmar_domain->max_addr = max_addr; |
fe40f1e0 WH |
3200 | } |
3201 | ||
dde57a21 | 3202 | ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot); |
faa3d6f5 | 3203 | return ret; |
38717946 | 3204 | } |
38717946 | 3205 | |
dde57a21 JR |
3206 | static void intel_iommu_unmap_range(struct iommu_domain *domain, |
3207 | unsigned long iova, size_t size) | |
38717946 | 3208 | { |
dde57a21 | 3209 | struct dmar_domain *dmar_domain = domain->priv; |
faa3d6f5 WH |
3210 | dma_addr_t base; |
3211 | ||
3212 | /* The address might not be aligned */ | |
3213 | base = iova & VTD_PAGE_MASK; | |
3214 | size = VTD_PAGE_ALIGN(size); | |
dde57a21 | 3215 | dma_pte_clear_range(dmar_domain, base, base + size); |
fe40f1e0 | 3216 | |
dde57a21 JR |
3217 | if (dmar_domain->max_addr == base + size) |
3218 | dmar_domain->max_addr = base; | |
38717946 | 3219 | } |
38717946 | 3220 | |
d14d6577 JR |
3221 | static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain, |
3222 | unsigned long iova) | |
38717946 | 3223 | { |
d14d6577 | 3224 | struct dmar_domain *dmar_domain = domain->priv; |
38717946 | 3225 | struct dma_pte *pte; |
faa3d6f5 | 3226 | u64 phys = 0; |
38717946 | 3227 | |
d14d6577 | 3228 | pte = addr_to_dma_pte(dmar_domain, iova); |
38717946 | 3229 | if (pte) |
faa3d6f5 | 3230 | phys = dma_pte_addr(pte); |
38717946 | 3231 | |
faa3d6f5 | 3232 | return phys; |
38717946 | 3233 | } |
a8bcbb0d | 3234 | |
dbb9fd86 SY |
3235 | static int intel_iommu_domain_has_cap(struct iommu_domain *domain, |
3236 | unsigned long cap) | |
3237 | { | |
3238 | struct dmar_domain *dmar_domain = domain->priv; | |
3239 | ||
3240 | if (cap == IOMMU_CAP_CACHE_COHERENCY) | |
3241 | return dmar_domain->iommu_snooping; | |
3242 | ||
3243 | return 0; | |
3244 | } | |
3245 | ||
a8bcbb0d JR |
3246 | static struct iommu_ops intel_iommu_ops = { |
3247 | .domain_init = intel_iommu_domain_init, | |
3248 | .domain_destroy = intel_iommu_domain_destroy, | |
3249 | .attach_dev = intel_iommu_attach_device, | |
3250 | .detach_dev = intel_iommu_detach_device, | |
3251 | .map = intel_iommu_map_range, | |
3252 | .unmap = intel_iommu_unmap_range, | |
3253 | .iova_to_phys = intel_iommu_iova_to_phys, | |
dbb9fd86 | 3254 | .domain_has_cap = intel_iommu_domain_has_cap, |
a8bcbb0d | 3255 | }; |
9af88143 DW |
3256 | |
3257 | static void __devinit quirk_iommu_rwbf(struct pci_dev *dev) | |
3258 | { | |
3259 | /* | |
3260 | * Mobile 4 Series Chipset neglects to set RWBF capability, | |
3261 | * but needs it: | |
3262 | */ | |
3263 | printk(KERN_INFO "DMAR: Forcing write-buffer flush capability\n"); | |
3264 | rwbf_quirk = 1; | |
3265 | } | |
3266 | ||
3267 | DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf); |