intel-iommu.c 115.7 KB
Newer Older
1
/*
2
 * Copyright © 2006-2014 Intel Corporation.
3 4 5 6 7 8 9 10 11 12
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
13 14 15 16 17
 * Authors: David Woodhouse <dwmw2@infradead.org>,
 *          Ashok Raj <ashok.raj@intel.com>,
 *          Shaohua Li <shaohua.li@intel.com>,
 *          Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>,
 *          Fenghua Yu <fenghua.yu@intel.com>
18 19 20 21
 */

#include <linux/init.h>
#include <linux/bitmap.h>
M
mark gross 已提交
22
#include <linux/debugfs.h>
23
#include <linux/export.h>
24 25 26 27 28 29 30 31
#include <linux/slab.h>
#include <linux/irq.h>
#include <linux/interrupt.h>
#include <linux/spinlock.h>
#include <linux/pci.h>
#include <linux/dmar.h>
#include <linux/dma-mapping.h>
#include <linux/mempool.h>
32
#include <linux/memory.h>
M
mark gross 已提交
33
#include <linux/timer.h>
K
Kay, Allen M 已提交
34
#include <linux/iova.h>
35
#include <linux/iommu.h>
K
Kay, Allen M 已提交
36
#include <linux/intel-iommu.h>
37
#include <linux/syscore_ops.h>
38
#include <linux/tboot.h>
39
#include <linux/dmi.h>
40
#include <linux/pci-ats.h>
T
Tejun Heo 已提交
41
#include <linux/memblock.h>
42
#include <asm/irq_remapping.h>
43
#include <asm/cacheflush.h>
44
#include <asm/iommu.h>
45

46
#include "irq_remapping.h"
47
#include "pci.h"
48

F
Fenghua Yu 已提交
49 50 51
#define ROOT_SIZE		VTD_PAGE_SIZE
#define CONTEXT_SIZE		VTD_PAGE_SIZE

52 53
#define IS_GFX_DEVICE(pdev) ((pdev->class >> 16) == PCI_BASE_CLASS_DISPLAY)
#define IS_ISA_DEVICE(pdev) ((pdev->class >> 8) == PCI_CLASS_BRIDGE_ISA)
54
#define IS_AZALIA(pdev) ((pdev)->vendor == 0x8086 && (pdev)->device == 0x3a3e)
55 56 57 58 59 60 61

#define IOAPIC_RANGE_START	(0xfee00000)
#define IOAPIC_RANGE_END	(0xfeefffff)
#define IOVA_START_ADDR		(0x1000)

#define DEFAULT_DOMAIN_ADDRESS_WIDTH 48

F
Fenghua Yu 已提交
62
#define MAX_AGAW_WIDTH 64
63
#define MAX_AGAW_PFN_WIDTH	(MAX_AGAW_WIDTH - VTD_PAGE_SHIFT)
F
Fenghua Yu 已提交
64

65 66 67 68 69 70 71 72
#define __DOMAIN_MAX_PFN(gaw)  ((((uint64_t)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
#define __DOMAIN_MAX_ADDR(gaw) ((((uint64_t)1) << gaw) - 1)

/* We limit DOMAIN_MAX_PFN to fit in an unsigned long, and DOMAIN_MAX_ADDR
   to match. That way, we can use 'unsigned long' for PFNs with impunity. */
#define DOMAIN_MAX_PFN(gaw)	((unsigned long) min_t(uint64_t, \
				__DOMAIN_MAX_PFN(gaw), (unsigned long)-1))
#define DOMAIN_MAX_ADDR(gaw)	(((uint64_t)__DOMAIN_MAX_PFN(gaw)) << VTD_PAGE_SHIFT)
73

74
#define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
75
#define DMA_32BIT_PFN		IOVA_PFN(DMA_BIT_MASK(32))
76
#define DMA_64BIT_PFN		IOVA_PFN(DMA_BIT_MASK(64))
M
mark gross 已提交
77

78 79 80 81
/* page table handling */
#define LEVEL_STRIDE		(9)
#define LEVEL_MASK		(((u64)1 << LEVEL_STRIDE) - 1)

82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99
/*
 * This bitmap is used to advertise the page sizes our hardware support
 * to the IOMMU core, which will then use this information to split
 * physically contiguous memory regions it is mapping into page sizes
 * that we support.
 *
 * Traditionally the IOMMU core just handed us the mappings directly,
 * after making sure the size is an order of a 4KiB page and that the
 * mapping has natural alignment.
 *
 * To retain this behavior, we currently advertise that we support
 * all page sizes that are an order of 4KiB.
 *
 * If at some point we'd like to utilize the IOMMU core's new behavior,
 * we could change this to advertise the real page sizes we support.
 */
#define INTEL_IOMMU_PGSIZES	(~0xFFFUL)

100 101 102 103 104 105 106
static inline int agaw_to_level(int agaw)
{
	return agaw + 2;
}

static inline int agaw_to_width(int agaw)
{
107
	return min_t(int, 30 + agaw * LEVEL_STRIDE, MAX_AGAW_WIDTH);
108 109 110 111
}

static inline int width_to_agaw(int width)
{
112
	return DIV_ROUND_UP(width - 30, LEVEL_STRIDE);
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
}

static inline unsigned int level_to_offset_bits(int level)
{
	return (level - 1) * LEVEL_STRIDE;
}

static inline int pfn_level_offset(unsigned long pfn, int level)
{
	return (pfn >> level_to_offset_bits(level)) & LEVEL_MASK;
}

static inline unsigned long level_mask(int level)
{
	return -1UL << level_to_offset_bits(level);
}

static inline unsigned long level_size(int level)
{
	return 1UL << level_to_offset_bits(level);
}

static inline unsigned long align_to_level(unsigned long pfn, int level)
{
	return (pfn + level_size(level) - 1) & level_mask(level);
}
139

140 141
static inline unsigned long lvl_to_nr_pages(unsigned int lvl)
{
142
	return  1 << min_t(int, (lvl - 1) * LEVEL_STRIDE, MAX_AGAW_PFN_WIDTH);
143 144
}

145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
/* VT-d pages must always be _smaller_ than MM pages. Otherwise things
   are never going to work. */
static inline unsigned long dma_to_mm_pfn(unsigned long dma_pfn)
{
	return dma_pfn >> (PAGE_SHIFT - VTD_PAGE_SHIFT);
}

static inline unsigned long mm_to_dma_pfn(unsigned long mm_pfn)
{
	return mm_pfn << (PAGE_SHIFT - VTD_PAGE_SHIFT);
}
static inline unsigned long page_to_dma_pfn(struct page *pg)
{
	return mm_to_dma_pfn(page_to_pfn(pg));
}
static inline unsigned long virt_to_dma_pfn(void *p)
{
	return page_to_dma_pfn(virt_to_page(p));
}

W
Weidong Han 已提交
165 166 167
/* global iommu list, set NULL for ignored DMAR units */
static struct intel_iommu **g_iommus;

168
static void __init check_tylersburg_isoch(void);
169 170
static int rwbf_quirk;

171 172 173 174 175 176
/*
 * set to 1 to panic kernel if can't successfully enable VT-d
 * (used when kernel is launched w/ TXT)
 */
static int force_on = 0;

177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
/*
 * 0: Present
 * 1-11: Reserved
 * 12-63: Context Ptr (12 - (haw-1))
 * 64-127: Reserved
 */
struct root_entry {
	u64	val;
	u64	rsvd1;
};
#define ROOT_ENTRY_NR (VTD_PAGE_SIZE/sizeof(struct root_entry))
static inline bool root_present(struct root_entry *root)
{
	return (root->val & 1);
}
static inline void set_root_present(struct root_entry *root)
{
	root->val |= 1;
}
static inline void set_root_value(struct root_entry *root, unsigned long value)
{
	root->val |= value & VTD_PAGE_MASK;
}

static inline struct context_entry *
get_context_addr_from_root(struct root_entry *root)
{
	return (struct context_entry *)
		(root_present(root)?phys_to_virt(
		root->val & VTD_PAGE_MASK) :
		NULL);
}

210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
/*
 * low 64 bits:
 * 0: present
 * 1: fault processing disable
 * 2-3: translation type
 * 12-63: address space root
 * high 64 bits:
 * 0-2: address width
 * 3-6: aval
 * 8-23: domain id
 */
struct context_entry {
	u64 lo;
	u64 hi;
};
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269

static inline bool context_present(struct context_entry *context)
{
	return (context->lo & 1);
}
static inline void context_set_present(struct context_entry *context)
{
	context->lo |= 1;
}

static inline void context_set_fault_enable(struct context_entry *context)
{
	context->lo &= (((u64)-1) << 2) | 1;
}

static inline void context_set_translation_type(struct context_entry *context,
						unsigned long value)
{
	context->lo &= (((u64)-1) << 4) | 3;
	context->lo |= (value & 3) << 2;
}

static inline void context_set_address_root(struct context_entry *context,
					    unsigned long value)
{
	context->lo |= value & VTD_PAGE_MASK;
}

static inline void context_set_address_width(struct context_entry *context,
					     unsigned long value)
{
	context->hi |= value & 7;
}

static inline void context_set_domain_id(struct context_entry *context,
					 unsigned long value)
{
	context->hi |= (value & ((1 << 16) - 1)) << 8;
}

static inline void context_clear_entry(struct context_entry *context)
{
	context->lo = 0;
	context->hi = 0;
}
270

271 272 273 274 275
/*
 * 0: readable
 * 1: writable
 * 2-6: reserved
 * 7: super page
276 277
 * 8-10: available
 * 11: snoop behavior
278 279 280 281 282 283
 * 12-63: Host physcial address
 */
struct dma_pte {
	u64 val;
};

284 285 286 287 288 289 290
static inline void dma_clear_pte(struct dma_pte *pte)
{
	pte->val = 0;
}

static inline u64 dma_pte_addr(struct dma_pte *pte)
{
291 292 293 294
#ifdef CONFIG_64BIT
	return pte->val & VTD_PAGE_MASK;
#else
	/* Must have a full atomic 64-bit read */
295
	return  __cmpxchg64(&pte->val, 0ULL, 0ULL) & VTD_PAGE_MASK;
296
#endif
297 298 299 300 301 302
}

static inline bool dma_pte_present(struct dma_pte *pte)
{
	return (pte->val & 3) != 0;
}
303

304 305 306 307 308
static inline bool dma_pte_superpage(struct dma_pte *pte)
{
	return (pte->val & (1 << 7));
}

309 310 311 312 313
static inline int first_pte_in_page(struct dma_pte *pte)
{
	return !((unsigned long)pte & ~VTD_PAGE_MASK);
}

314 315 316 317 318 319
/*
 * This domain is a statically identity mapping domain.
 *	1. This domain creats a static 1:1 mapping to all usable memory.
 * 	2. It maps to each iommu if successful.
 *	3. Each iommu mapps to this domain if successful.
 */
320 321
static struct dmar_domain *si_domain;
static int hw_pass_through = 1;
322

W
Weidong Han 已提交
323
/* devices under the same p2p bridge are owned in one domain */
324
#define DOMAIN_FLAG_P2P_MULTIPLE_DEVICES (1 << 0)
W
Weidong Han 已提交
325

326 327 328 329 330
/* domain represents a virtual machine, more than one devices
 * across iommus may be owned in one domain, e.g. kvm guest.
 */
#define DOMAIN_FLAG_VIRTUAL_MACHINE	(1 << 1)

331 332 333
/* si_domain contains mulitple devices */
#define DOMAIN_FLAG_STATIC_IDENTITY	(1 << 2)

334 335 336 337 338 339 340
/* define the limit of IOMMUs supported in each domain */
#ifdef	CONFIG_X86
# define	IOMMU_UNITS_SUPPORTED	MAX_IO_APICS
#else
# define	IOMMU_UNITS_SUPPORTED	64
#endif

341 342
struct dmar_domain {
	int	id;			/* domain id */
343
	int	nid;			/* node id */
344 345
	DECLARE_BITMAP(iommu_bmp, IOMMU_UNITS_SUPPORTED);
					/* bitmap of iommus this domain uses*/
346 347 348 349 350 351 352 353 354 355

	struct list_head devices; 	/* all devices' list */
	struct iova_domain iovad;	/* iova's that belong to this domain */

	struct dma_pte	*pgd;		/* virtual address */
	int		gaw;		/* max guest address width */

	/* adjusted guest address width, 0 is level 2 30-bit */
	int		agaw;

W
Weidong Han 已提交
356
	int		flags;		/* flags to find out type of domain */
W
Weidong Han 已提交
357 358

	int		iommu_coherency;/* indicate coherency of iommu access */
359
	int		iommu_snooping; /* indicate snooping control feature*/
360
	int		iommu_count;	/* reference count of iommu */
361 362 363
	int		iommu_superpage;/* Level of superpages supported:
					   0 == 4KiB (no superpages), 1 == 2MiB,
					   2 == 1GiB, 3 == 512GiB, 4 == 1TiB */
364
	spinlock_t	iommu_lock;	/* protect iommu set in domain */
365
	u64		max_addr;	/* maximum mapped address */
366 367
};

368 369 370 371
/* PCI domain-device relationship */
struct device_domain_info {
	struct list_head link;	/* link to domain siblings */
	struct list_head global; /* link to global list */
372
	u8 bus;			/* PCI bus number */
373
	u8 devfn;		/* PCI devfn number */
374
	struct device *dev; /* it's NULL for PCIe-to-PCI bridge */
Y
Yu Zhao 已提交
375
	struct intel_iommu *iommu; /* IOMMU used by this device */
376 377 378
	struct dmar_domain *domain; /* pointer to domain */
};

379 380 381 382 383
struct dmar_rmrr_unit {
	struct list_head list;		/* list of rmrr units	*/
	struct acpi_dmar_header *hdr;	/* ACPI header		*/
	u64	base_address;		/* reserved base address*/
	u64	end_address;		/* reserved end address */
384
	struct dmar_dev_scope *devices;	/* target devices */
385 386 387 388 389 390
	int	devices_cnt;		/* target device count */
};

struct dmar_atsr_unit {
	struct list_head list;		/* list of ATSR units */
	struct acpi_dmar_header *hdr;	/* ACPI header */
391
	struct dmar_dev_scope *devices;	/* target devices */
392 393 394 395 396 397 398 399 400 401
	int devices_cnt;		/* target device count */
	u8 include_all:1;		/* include all ports */
};

static LIST_HEAD(dmar_atsr_units);
static LIST_HEAD(dmar_rmrr_units);

#define for_each_rmrr_units(rmrr) \
	list_for_each_entry(rmrr, &dmar_rmrr_units, list)

M
mark gross 已提交
402 403
static void flush_unmaps_timeout(unsigned long data);

404
static DEFINE_TIMER(unmap_timer,  flush_unmaps_timeout, 0, 0);
M
mark gross 已提交
405

406 407 408 409 410
#define HIGH_WATER_MARK 250
struct deferred_flush_tables {
	int next;
	struct iova *iova[HIGH_WATER_MARK];
	struct dmar_domain *domain[HIGH_WATER_MARK];
411
	struct page *freelist[HIGH_WATER_MARK];
412 413 414 415
};

static struct deferred_flush_tables *deferred_flush;

M
mark gross 已提交
416 417 418 419 420 421 422 423 424
/* bitmap for indexing intel_iommus */
static int g_num_of_iommus;

static DEFINE_SPINLOCK(async_umap_flush_lock);
static LIST_HEAD(unmaps_to_do);

static int timer_on;
static long list_size;

425
static void domain_exit(struct dmar_domain *domain);
426
static void domain_remove_dev_info(struct dmar_domain *domain);
427
static void domain_remove_one_dev_info(struct dmar_domain *domain,
428
				       struct device *dev);
429
static void iommu_detach_dependent_devices(struct intel_iommu *iommu,
430
					   struct device *dev);
431

432
#ifdef CONFIG_INTEL_IOMMU_DEFAULT_ON
433 434 435
int dmar_disabled = 0;
#else
int dmar_disabled = 1;
436
#endif /*CONFIG_INTEL_IOMMU_DEFAULT_ON*/
437

438 439 440
int intel_iommu_enabled = 0;
EXPORT_SYMBOL_GPL(intel_iommu_enabled);

441
static int dmar_map_gfx = 1;
442
static int dmar_forcedac;
M
mark gross 已提交
443
static int intel_iommu_strict;
444
static int intel_iommu_superpage = 1;
445

446 447 448
int intel_iommu_gfx_mapped;
EXPORT_SYMBOL_GPL(intel_iommu_gfx_mapped);

449 450 451 452
#define DUMMY_DEVICE_DOMAIN_INFO ((struct device_domain_info *)(-1))
static DEFINE_SPINLOCK(device_domain_lock);
static LIST_HEAD(device_domain_list);

453 454
static struct iommu_ops intel_iommu_ops;

455 456 457 458 459
static int __init intel_iommu_setup(char *str)
{
	if (!str)
		return -EINVAL;
	while (*str) {
460 461 462 463
		if (!strncmp(str, "on", 2)) {
			dmar_disabled = 0;
			printk(KERN_INFO "Intel-IOMMU: enabled\n");
		} else if (!strncmp(str, "off", 3)) {
464
			dmar_disabled = 1;
465
			printk(KERN_INFO "Intel-IOMMU: disabled\n");
466 467 468 469
		} else if (!strncmp(str, "igfx_off", 8)) {
			dmar_map_gfx = 0;
			printk(KERN_INFO
				"Intel-IOMMU: disable GFX device mapping\n");
470
		} else if (!strncmp(str, "forcedac", 8)) {
M
mark gross 已提交
471
			printk(KERN_INFO
472 473
				"Intel-IOMMU: Forcing DAC for PCI devices\n");
			dmar_forcedac = 1;
M
mark gross 已提交
474 475 476 477
		} else if (!strncmp(str, "strict", 6)) {
			printk(KERN_INFO
				"Intel-IOMMU: disable batched IOTLB flush\n");
			intel_iommu_strict = 1;
478 479 480 481
		} else if (!strncmp(str, "sp_off", 6)) {
			printk(KERN_INFO
				"Intel-IOMMU: disable supported super page\n");
			intel_iommu_superpage = 0;
482 483 484 485 486 487 488 489 490 491 492 493 494 495
		}

		str += strcspn(str, ",");
		while (*str == ',')
			str++;
	}
	return 0;
}
__setup("intel_iommu=", intel_iommu_setup);

static struct kmem_cache *iommu_domain_cache;
static struct kmem_cache *iommu_devinfo_cache;
static struct kmem_cache *iommu_iova_cache;

496
static inline void *alloc_pgtable_page(int node)
497
{
498 499
	struct page *page;
	void *vaddr = NULL;
500

501 502 503
	page = alloc_pages_node(node, GFP_ATOMIC | __GFP_ZERO, 0);
	if (page)
		vaddr = page_address(page);
504
	return vaddr;
505 506 507 508 509 510 511 512 513
}

static inline void free_pgtable_page(void *vaddr)
{
	free_page((unsigned long)vaddr);
}

static inline void *alloc_domain_mem(void)
{
514
	return kmem_cache_alloc(iommu_domain_cache, GFP_ATOMIC);
515 516
}

K
Kay, Allen M 已提交
517
static void free_domain_mem(void *vaddr)
518 519 520 521 522 523
{
	kmem_cache_free(iommu_domain_cache, vaddr);
}

static inline void * alloc_devinfo_mem(void)
{
524
	return kmem_cache_alloc(iommu_devinfo_cache, GFP_ATOMIC);
525 526 527 528 529 530 531 532 533
}

static inline void free_devinfo_mem(void *vaddr)
{
	kmem_cache_free(iommu_devinfo_cache, vaddr);
}

struct iova *alloc_iova_mem(void)
{
534
	return kmem_cache_alloc(iommu_iova_cache, GFP_ATOMIC);
535 536 537 538 539 540 541
}

void free_iova_mem(struct iova *iova)
{
	kmem_cache_free(iommu_iova_cache, iova);
}

W
Weidong Han 已提交
542

F
Fenghua Yu 已提交
543
static int __iommu_calculate_agaw(struct intel_iommu *iommu, int max_gaw)
W
Weidong Han 已提交
544 545 546 547 548
{
	unsigned long sagaw;
	int agaw = -1;

	sagaw = cap_sagaw(iommu->cap);
F
Fenghua Yu 已提交
549
	for (agaw = width_to_agaw(max_gaw);
W
Weidong Han 已提交
550 551 552 553 554 555 556 557
	     agaw >= 0; agaw--) {
		if (test_bit(agaw, &sagaw))
			break;
	}

	return agaw;
}

F
Fenghua Yu 已提交
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
/*
 * Calculate max SAGAW for each iommu.
 */
int iommu_calculate_max_sagaw(struct intel_iommu *iommu)
{
	return __iommu_calculate_agaw(iommu, MAX_AGAW_WIDTH);
}

/*
 * calculate agaw for each iommu.
 * "SAGAW" may be different across iommus, use a default agaw, and
 * get a supported less agaw for iommus that don't support the default agaw.
 */
int iommu_calculate_agaw(struct intel_iommu *iommu)
{
	return __iommu_calculate_agaw(iommu, DEFAULT_DOMAIN_ADDRESS_WIDTH);
}

576
/* This functionin only returns single iommu in a domain */
577 578 579 580
static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
{
	int iommu_id;

581
	/* si_domain and vm domain should not get here. */
582
	BUG_ON(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE);
583
	BUG_ON(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY);
584

585
	iommu_id = find_first_bit(domain->iommu_bmp, g_num_of_iommus);
586 587 588 589 590 591
	if (iommu_id < 0 || iommu_id >= g_num_of_iommus)
		return NULL;

	return g_iommus[iommu_id];
}

W
Weidong Han 已提交
592 593
static void domain_update_iommu_coherency(struct dmar_domain *domain)
{
594 595 596
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu;
	int i, found = 0;
597

598
	domain->iommu_coherency = 1;
W
Weidong Han 已提交
599

600
	for_each_set_bit(i, domain->iommu_bmp, g_num_of_iommus) {
601
		found = 1;
W
Weidong Han 已提交
602 603 604 605 606
		if (!ecap_coherent(g_iommus[i]->ecap)) {
			domain->iommu_coherency = 0;
			break;
		}
	}
607 608 609 610 611 612 613 614 615 616 617 618
	if (found)
		return;

	/* No hardware attached; use lowest common denominator */
	rcu_read_lock();
	for_each_active_iommu(iommu, drhd) {
		if (!ecap_coherent(iommu->ecap)) {
			domain->iommu_coherency = 0;
			break;
		}
	}
	rcu_read_unlock();
W
Weidong Han 已提交
619 620
}

621 622 623 624 625 626
static void domain_update_iommu_snooping(struct dmar_domain *domain)
{
	int i;

	domain->iommu_snooping = 1;

627
	for_each_set_bit(i, domain->iommu_bmp, g_num_of_iommus) {
628 629 630 631 632 633 634
		if (!ecap_sc_support(g_iommus[i]->ecap)) {
			domain->iommu_snooping = 0;
			break;
		}
	}
}

635 636
static void domain_update_iommu_superpage(struct dmar_domain *domain)
{
637 638 639
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu = NULL;
	int mask = 0xf;
640 641 642 643 644 645

	if (!intel_iommu_superpage) {
		domain->iommu_superpage = 0;
		return;
	}

646
	/* set iommu_superpage to the smallest common denominator */
647
	rcu_read_lock();
648 649
	for_each_active_iommu(iommu, drhd) {
		mask &= cap_super_page_val(iommu->cap);
650 651 652 653
		if (!mask) {
			break;
		}
	}
654 655
	rcu_read_unlock();

656 657 658
	domain->iommu_superpage = fls(mask);
}

659 660 661 662 663
/* Some capabilities may be different across iommus */
static void domain_update_iommu_cap(struct dmar_domain *domain)
{
	domain_update_iommu_coherency(domain);
	domain_update_iommu_snooping(domain);
664
	domain_update_iommu_superpage(domain);
665 666
}

667
static struct intel_iommu *device_to_iommu(struct device *dev, u8 *bus, u8 *devfn)
668 669
{
	struct dmar_drhd_unit *drhd = NULL;
670
	struct intel_iommu *iommu;
671 672 673
	struct device *tmp;
	struct pci_dev *ptmp, *pdev = NULL;
	u16 segment;
674 675
	int i;

676 677 678 679 680 681
	if (dev_is_pci(dev)) {
		pdev = to_pci_dev(dev);
		segment = pci_domain_nr(pdev->bus);
	} else if (ACPI_COMPANION(dev))
		dev = &ACPI_COMPANION(dev)->dev;

682
	rcu_read_lock();
683
	for_each_active_iommu(iommu, drhd) {
684
		if (pdev && segment != drhd->segment)
685
			continue;
686

687
		for_each_active_dev_scope(drhd->devices,
688 689 690 691
					  drhd->devices_cnt, i, tmp) {
			if (tmp == dev) {
				*bus = drhd->devices[i].bus;
				*devfn = drhd->devices[i].devfn;
692
				goto out;
693 694 695 696 697 698 699 700 701 702
			}

			if (!pdev || !dev_is_pci(tmp))
				continue;

			ptmp = to_pci_dev(tmp);
			if (ptmp->subordinate &&
			    ptmp->subordinate->number <= pdev->bus->number &&
			    ptmp->subordinate->busn_res.end >= pdev->bus->number)
				goto got_pdev;
703
		}
704

705 706 707 708
		if (pdev && drhd->include_all) {
		got_pdev:
			*bus = pdev->bus->number;
			*devfn = pdev->devfn;
709
			goto out;
710
		}
711
	}
712
	iommu = NULL;
713
 out:
714
	rcu_read_unlock();
715

716
	return iommu;
717 718
}

W
Weidong Han 已提交
719 720 721 722 723 724 725
static void domain_flush_cache(struct dmar_domain *domain,
			       void *addr, int size)
{
	if (!domain->iommu_coherency)
		clflush_cache_range(addr, size);
}

726 727 728 729 730 731 732 733 734 735 736 737 738
/* Gets context entry for a given bus and devfn */
static struct context_entry * device_to_context_entry(struct intel_iommu *iommu,
		u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	unsigned long phy_addr;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (!context) {
739 740
		context = (struct context_entry *)
				alloc_pgtable_page(iommu->node);
741 742 743 744
		if (!context) {
			spin_unlock_irqrestore(&iommu->lock, flags);
			return NULL;
		}
F
Fenghua Yu 已提交
745
		__iommu_flush_cache(iommu, (void *)context, CONTEXT_SIZE);
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768
		phy_addr = virt_to_phys((void *)context);
		set_root_value(root, phy_addr);
		set_root_present(root);
		__iommu_flush_cache(iommu, root, sizeof(*root));
	}
	spin_unlock_irqrestore(&iommu->lock, flags);
	return &context[devfn];
}

static int device_context_mapped(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	int ret;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (!context) {
		ret = 0;
		goto out;
	}
769
	ret = context_present(&context[devfn]);
770 771 772 773 774 775 776 777 778 779 780 781 782 783 784
out:
	spin_unlock_irqrestore(&iommu->lock, flags);
	return ret;
}

static void clear_context_table(struct intel_iommu *iommu, u8 bus, u8 devfn)
{
	struct root_entry *root;
	struct context_entry *context;
	unsigned long flags;

	spin_lock_irqsave(&iommu->lock, flags);
	root = &iommu->root_entry[bus];
	context = get_context_addr_from_root(root);
	if (context) {
785
		context_clear_entry(&context[devfn]);
786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814
		__iommu_flush_cache(iommu, &context[devfn], \
			sizeof(*context));
	}
	spin_unlock_irqrestore(&iommu->lock, flags);
}

static void free_context_table(struct intel_iommu *iommu)
{
	struct root_entry *root;
	int i;
	unsigned long flags;
	struct context_entry *context;

	spin_lock_irqsave(&iommu->lock, flags);
	if (!iommu->root_entry) {
		goto out;
	}
	for (i = 0; i < ROOT_ENTRY_NR; i++) {
		root = &iommu->root_entry[i];
		context = get_context_addr_from_root(root);
		if (context)
			free_pgtable_page(context);
	}
	free_pgtable_page(iommu->root_entry);
	iommu->root_entry = NULL;
out:
	spin_unlock_irqrestore(&iommu->lock, flags);
}

815
static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
816
				      unsigned long pfn, int *target_level)
817
{
818
	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
819 820
	struct dma_pte *parent, *pte = NULL;
	int level = agaw_to_level(domain->agaw);
821
	int offset;
822 823

	BUG_ON(!domain->pgd);
824 825 826 827 828

	if (addr_width < BITS_PER_LONG && pfn >> addr_width)
		/* Address beyond IOMMU's addressing capabilities. */
		return NULL;

829 830
	parent = domain->pgd;

831
	while (1) {
832 833
		void *tmp_page;

834
		offset = pfn_level_offset(pfn, level);
835
		pte = &parent[offset];
836
		if (!*target_level && (dma_pte_superpage(pte) || !dma_pte_present(pte)))
837
			break;
838
		if (level == *target_level)
839 840
			break;

841
		if (!dma_pte_present(pte)) {
842 843
			uint64_t pteval;

844
			tmp_page = alloc_pgtable_page(domain->nid);
845

846
			if (!tmp_page)
847
				return NULL;
848

849
			domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE);
850
			pteval = ((uint64_t)virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE;
851 852 853 854 855 856 857
			if (cmpxchg64(&pte->val, 0ULL, pteval)) {
				/* Someone else set it while we were thinking; use theirs. */
				free_pgtable_page(tmp_page);
			} else {
				dma_pte_addr(pte);
				domain_flush_cache(domain, pte, sizeof(*pte));
			}
858
		}
859 860 861
		if (level == 1)
			break;

862
		parent = phys_to_virt(dma_pte_addr(pte));
863 864 865
		level--;
	}

866 867 868
	if (!*target_level)
		*target_level = level;

869 870 871
	return pte;
}

872

873
/* return address's pte at specific level */
874 875
static struct dma_pte *dma_pfn_level_pte(struct dmar_domain *domain,
					 unsigned long pfn,
876
					 int level, int *large_page)
877 878 879 880 881 882 883
{
	struct dma_pte *parent, *pte = NULL;
	int total = agaw_to_level(domain->agaw);
	int offset;

	parent = domain->pgd;
	while (level <= total) {
884
		offset = pfn_level_offset(pfn, total);
885 886 887 888
		pte = &parent[offset];
		if (level == total)
			return pte;

889 890
		if (!dma_pte_present(pte)) {
			*large_page = total;
891
			break;
892 893 894 895 896 897 898
		}

		if (pte->val & DMA_PTE_LARGE_PAGE) {
			*large_page = total;
			return pte;
		}

899
		parent = phys_to_virt(dma_pte_addr(pte));
900 901 902 903 904 905
		total--;
	}
	return NULL;
}

/* clear last level pte, a tlb flush should be followed */
906
static void dma_pte_clear_range(struct dmar_domain *domain,
907 908
				unsigned long start_pfn,
				unsigned long last_pfn)
909
{
910
	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
911
	unsigned int large_page = 1;
912
	struct dma_pte *first_pte, *pte;
913

914
	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
915
	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
916
	BUG_ON(start_pfn > last_pfn);
917

918
	/* we don't need lock here; nobody else touches the iova range */
919
	do {
920 921
		large_page = 1;
		first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1, &large_page);
922
		if (!pte) {
923
			start_pfn = align_to_level(start_pfn + 1, large_page + 1);
924 925
			continue;
		}
926
		do {
927
			dma_clear_pte(pte);
928
			start_pfn += lvl_to_nr_pages(large_page);
929
			pte++;
930 931
		} while (start_pfn <= last_pfn && !first_pte_in_page(pte));

932 933
		domain_flush_cache(domain, first_pte,
				   (void *)pte - (void *)first_pte);
934 935

	} while (start_pfn && start_pfn <= last_pfn);
936 937
}

938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
static void dma_pte_free_level(struct dmar_domain *domain, int level,
			       struct dma_pte *pte, unsigned long pfn,
			       unsigned long start_pfn, unsigned long last_pfn)
{
	pfn = max(start_pfn, pfn);
	pte = &pte[pfn_level_offset(pfn, level)];

	do {
		unsigned long level_pfn;
		struct dma_pte *level_pte;

		if (!dma_pte_present(pte) || dma_pte_superpage(pte))
			goto next;

		level_pfn = pfn & level_mask(level - 1);
		level_pte = phys_to_virt(dma_pte_addr(pte));

		if (level > 2)
			dma_pte_free_level(domain, level - 1, level_pte,
					   level_pfn, start_pfn, last_pfn);

		/* If range covers entire pagetable, free it */
		if (!(start_pfn > level_pfn ||
961
		      last_pfn < level_pfn + level_size(level) - 1)) {
962 963 964 965 966 967 968 969 970
			dma_clear_pte(pte);
			domain_flush_cache(domain, pte, sizeof(*pte));
			free_pgtable_page(level_pte);
		}
next:
		pfn += level_size(level);
	} while (!first_pte_in_page(++pte) && pfn <= last_pfn);
}

971 972
/* free page table pages. last level pte should already be cleared */
static void dma_pte_free_pagetable(struct dmar_domain *domain,
973 974
				   unsigned long start_pfn,
				   unsigned long last_pfn)
975
{
976
	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
977

978 979
	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
980
	BUG_ON(start_pfn > last_pfn);
981

982
	/* We don't need lock here; nobody else touches the iova range */
983 984
	dma_pte_free_level(domain, agaw_to_level(domain->agaw),
			   domain->pgd, 0, start_pfn, last_pfn);
985

986
	/* free pgd */
987
	if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
988 989 990 991 992
		free_pgtable_page(domain->pgd);
		domain->pgd = NULL;
	}
}

993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
/* When a page at a given level is being unlinked from its parent, we don't
   need to *modify* it at all. All we need to do is make a list of all the
   pages which can be freed just as soon as we've flushed the IOTLB and we
   know the hardware page-walk will no longer touch them.
   The 'pte' argument is the *parent* PTE, pointing to the page that is to
   be freed. */
static struct page *dma_pte_list_pagetables(struct dmar_domain *domain,
					    int level, struct dma_pte *pte,
					    struct page *freelist)
{
	struct page *pg;

	pg = pfn_to_page(dma_pte_addr(pte) >> PAGE_SHIFT);
	pg->freelist = freelist;
	freelist = pg;

	if (level == 1)
		return freelist;

1012 1013
	pte = page_address(pg);
	do {
1014 1015 1016
		if (dma_pte_present(pte) && !dma_pte_superpage(pte))
			freelist = dma_pte_list_pagetables(domain, level - 1,
							   pte, freelist);
1017 1018
		pte++;
	} while (!first_pte_in_page(pte));
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111

	return freelist;
}

static struct page *dma_pte_clear_level(struct dmar_domain *domain, int level,
					struct dma_pte *pte, unsigned long pfn,
					unsigned long start_pfn,
					unsigned long last_pfn,
					struct page *freelist)
{
	struct dma_pte *first_pte = NULL, *last_pte = NULL;

	pfn = max(start_pfn, pfn);
	pte = &pte[pfn_level_offset(pfn, level)];

	do {
		unsigned long level_pfn;

		if (!dma_pte_present(pte))
			goto next;

		level_pfn = pfn & level_mask(level);

		/* If range covers entire pagetable, free it */
		if (start_pfn <= level_pfn &&
		    last_pfn >= level_pfn + level_size(level) - 1) {
			/* These suborbinate page tables are going away entirely. Don't
			   bother to clear them; we're just going to *free* them. */
			if (level > 1 && !dma_pte_superpage(pte))
				freelist = dma_pte_list_pagetables(domain, level - 1, pte, freelist);

			dma_clear_pte(pte);
			if (!first_pte)
				first_pte = pte;
			last_pte = pte;
		} else if (level > 1) {
			/* Recurse down into a level that isn't *entirely* obsolete */
			freelist = dma_pte_clear_level(domain, level - 1,
						       phys_to_virt(dma_pte_addr(pte)),
						       level_pfn, start_pfn, last_pfn,
						       freelist);
		}
next:
		pfn += level_size(level);
	} while (!first_pte_in_page(++pte) && pfn <= last_pfn);

	if (first_pte)
		domain_flush_cache(domain, first_pte,
				   (void *)++last_pte - (void *)first_pte);

	return freelist;
}

/* We can't just free the pages because the IOMMU may still be walking
   the page tables, and may have cached the intermediate levels. The
   pages can only be freed after the IOTLB flush has been done. */
struct page *domain_unmap(struct dmar_domain *domain,
			  unsigned long start_pfn,
			  unsigned long last_pfn)
{
	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
	struct page *freelist = NULL;

	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
	BUG_ON(start_pfn > last_pfn);

	/* we don't need lock here; nobody else touches the iova range */
	freelist = dma_pte_clear_level(domain, agaw_to_level(domain->agaw),
				       domain->pgd, 0, start_pfn, last_pfn, NULL);

	/* free pgd */
	if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
		struct page *pgd_page = virt_to_page(domain->pgd);
		pgd_page->freelist = freelist;
		freelist = pgd_page;

		domain->pgd = NULL;
	}

	return freelist;
}

void dma_free_pagelist(struct page *freelist)
{
	struct page *pg;

	while ((pg = freelist)) {
		freelist = pg->freelist;
		free_pgtable_page(page_address(pg));
	}
}

1112 1113 1114 1115 1116 1117
/* iommu handling */
static int iommu_alloc_root_entry(struct intel_iommu *iommu)
{
	struct root_entry *root;
	unsigned long flags;

1118
	root = (struct root_entry *)alloc_pgtable_page(iommu->node);
1119 1120 1121
	if (!root)
		return -ENOMEM;

F
Fenghua Yu 已提交
1122
	__iommu_flush_cache(iommu, root, ROOT_SIZE);
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133

	spin_lock_irqsave(&iommu->lock, flags);
	iommu->root_entry = root;
	spin_unlock_irqrestore(&iommu->lock, flags);

	return 0;
}

static void iommu_set_root_entry(struct intel_iommu *iommu)
{
	void *addr;
1134
	u32 sts;
1135 1136 1137 1138
	unsigned long flag;

	addr = iommu->root_entry;

1139
	raw_spin_lock_irqsave(&iommu->register_lock, flag);
1140 1141
	dmar_writeq(iommu->reg + DMAR_RTADDR_REG, virt_to_phys(addr));

1142
	writel(iommu->gcmd | DMA_GCMD_SRTP, iommu->reg + DMAR_GCMD_REG);
1143 1144 1145

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1146
		      readl, (sts & DMA_GSTS_RTPS), sts);
1147

1148
	raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1149 1150 1151 1152 1153 1154 1155
}

static void iommu_flush_write_buffer(struct intel_iommu *iommu)
{
	u32 val;
	unsigned long flag;

1156
	if (!rwbf_quirk && !cap_rwbf(iommu->cap))
1157 1158
		return;

1159
	raw_spin_lock_irqsave(&iommu->register_lock, flag);
1160
	writel(iommu->gcmd | DMA_GCMD_WBF, iommu->reg + DMAR_GCMD_REG);
1161 1162 1163

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1164
		      readl, (!(val & DMA_GSTS_WBFS)), val);
1165

1166
	raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1167 1168 1169
}

/* return value determine if we need a write buffer flush */
1170 1171 1172
static void __iommu_flush_context(struct intel_iommu *iommu,
				  u16 did, u16 source_id, u8 function_mask,
				  u64 type)
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
{
	u64 val = 0;
	unsigned long flag;

	switch (type) {
	case DMA_CCMD_GLOBAL_INVL:
		val = DMA_CCMD_GLOBAL_INVL;
		break;
	case DMA_CCMD_DOMAIN_INVL:
		val = DMA_CCMD_DOMAIN_INVL|DMA_CCMD_DID(did);
		break;
	case DMA_CCMD_DEVICE_INVL:
		val = DMA_CCMD_DEVICE_INVL|DMA_CCMD_DID(did)
			| DMA_CCMD_SID(source_id) | DMA_CCMD_FM(function_mask);
		break;
	default:
		BUG();
	}
	val |= DMA_CCMD_ICC;

1193
	raw_spin_lock_irqsave(&iommu->register_lock, flag);
1194 1195 1196 1197 1198 1199
	dmar_writeq(iommu->reg + DMAR_CCMD_REG, val);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_CCMD_REG,
		dmar_readq, (!(val & DMA_CCMD_ICC)), val);

1200
	raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1201 1202 1203
}

/* return value determine if we need a write buffer flush */
1204 1205
static void __iommu_flush_iotlb(struct intel_iommu *iommu, u16 did,
				u64 addr, unsigned int size_order, u64 type)
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
{
	int tlb_offset = ecap_iotlb_offset(iommu->ecap);
	u64 val = 0, val_iva = 0;
	unsigned long flag;

	switch (type) {
	case DMA_TLB_GLOBAL_FLUSH:
		/* global flush doesn't need set IVA_REG */
		val = DMA_TLB_GLOBAL_FLUSH|DMA_TLB_IVT;
		break;
	case DMA_TLB_DSI_FLUSH:
		val = DMA_TLB_DSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did);
		break;
	case DMA_TLB_PSI_FLUSH:
		val = DMA_TLB_PSI_FLUSH|DMA_TLB_IVT|DMA_TLB_DID(did);
1221
		/* IH bit is passed in as part of address */
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
		val_iva = size_order | addr;
		break;
	default:
		BUG();
	}
	/* Note: set drain read/write */
#if 0
	/*
	 * This is probably to be super secure.. Looks like we can
	 * ignore it without any impact.
	 */
	if (cap_read_drain(iommu->cap))
		val |= DMA_TLB_READ_DRAIN;
#endif
	if (cap_write_drain(iommu->cap))
		val |= DMA_TLB_WRITE_DRAIN;

1239
	raw_spin_lock_irqsave(&iommu->register_lock, flag);
1240 1241 1242 1243 1244 1245 1246 1247 1248
	/* Note: Only uses first TLB reg currently */
	if (val_iva)
		dmar_writeq(iommu->reg + tlb_offset, val_iva);
	dmar_writeq(iommu->reg + tlb_offset + 8, val);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, tlb_offset + 8,
		dmar_readq, (!(val & DMA_TLB_IVT)), val);

1249
	raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1250 1251 1252 1253 1254 1255

	/* check IOTLB invalidation granularity */
	if (DMA_TLB_IAIG(val) == 0)
		printk(KERN_ERR"IOMMU: flush IOTLB failed\n");
	if (DMA_TLB_IAIG(val) != DMA_TLB_IIRG(type))
		pr_debug("IOMMU: tlb flush request %Lx, actual %Lx\n",
F
Fenghua Yu 已提交
1256 1257
			(unsigned long long)DMA_TLB_IIRG(type),
			(unsigned long long)DMA_TLB_IAIG(val));
1258 1259
}

1260 1261 1262
static struct device_domain_info *
iommu_support_dev_iotlb (struct dmar_domain *domain, struct intel_iommu *iommu,
			 u8 bus, u8 devfn)
Y
Yu Zhao 已提交
1263 1264 1265 1266
{
	int found = 0;
	unsigned long flags;
	struct device_domain_info *info;
1267
	struct pci_dev *pdev;
Y
Yu Zhao 已提交
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282

	if (!ecap_dev_iotlb_support(iommu->ecap))
		return NULL;

	if (!iommu->qi)
		return NULL;

	spin_lock_irqsave(&device_domain_lock, flags);
	list_for_each_entry(info, &domain->devices, link)
		if (info->bus == bus && info->devfn == devfn) {
			found = 1;
			break;
		}
	spin_unlock_irqrestore(&device_domain_lock, flags);

1283
	if (!found || !info->dev || !dev_is_pci(info->dev))
Y
Yu Zhao 已提交
1284 1285
		return NULL;

1286 1287 1288
	pdev = to_pci_dev(info->dev);

	if (!pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ATS))
Y
Yu Zhao 已提交
1289 1290
		return NULL;

1291
	if (!dmar_find_matched_atsr_unit(pdev))
Y
Yu Zhao 已提交
1292 1293 1294 1295 1296 1297
		return NULL;

	return info;
}

static void iommu_enable_dev_iotlb(struct device_domain_info *info)
1298
{
1299
	if (!info || !dev_is_pci(info->dev))
Y
Yu Zhao 已提交
1300 1301
		return;

1302
	pci_enable_ats(to_pci_dev(info->dev), VTD_PAGE_SHIFT);
Y
Yu Zhao 已提交
1303 1304 1305 1306
}

static void iommu_disable_dev_iotlb(struct device_domain_info *info)
{
1307 1308
	if (!info->dev || !dev_is_pci(info->dev) ||
	    !pci_ats_enabled(to_pci_dev(info->dev)))
Y
Yu Zhao 已提交
1309 1310
		return;

1311
	pci_disable_ats(to_pci_dev(info->dev));
Y
Yu Zhao 已提交
1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322
}

static void iommu_flush_dev_iotlb(struct dmar_domain *domain,
				  u64 addr, unsigned mask)
{
	u16 sid, qdep;
	unsigned long flags;
	struct device_domain_info *info;

	spin_lock_irqsave(&device_domain_lock, flags);
	list_for_each_entry(info, &domain->devices, link) {
1323 1324 1325 1326 1327 1328
		struct pci_dev *pdev;
		if (!info->dev || !dev_is_pci(info->dev))
			continue;

		pdev = to_pci_dev(info->dev);
		if (!pci_ats_enabled(pdev))
Y
Yu Zhao 已提交
1329 1330 1331
			continue;

		sid = info->bus << 8 | info->devfn;
1332
		qdep = pci_ats_queue_depth(pdev);
Y
Yu Zhao 已提交
1333 1334 1335 1336 1337
		qi_flush_dev_iotlb(info->iommu, sid, qdep, addr, mask);
	}
	spin_unlock_irqrestore(&device_domain_lock, flags);
}

1338
static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
1339
				  unsigned long pfn, unsigned int pages, int ih, int map)
1340
{
1341
	unsigned int mask = ilog2(__roundup_pow_of_two(pages));
1342
	uint64_t addr = (uint64_t)pfn << VTD_PAGE_SHIFT;
1343 1344 1345

	BUG_ON(pages == 0);

1346 1347
	if (ih)
		ih = 1 << 6;
1348
	/*
1349 1350
	 * Fallback to domain selective flush if no PSI support or the size is
	 * too big.
1351 1352 1353
	 * PSI requires page size to be 2 ^ x, and the base address is naturally
	 * aligned to the size
	 */
1354 1355
	if (!cap_pgsel_inv(iommu->cap) || mask > cap_max_amask_val(iommu->cap))
		iommu->flush.flush_iotlb(iommu, did, 0, 0,
1356
						DMA_TLB_DSI_FLUSH);
1357
	else
1358
		iommu->flush.flush_iotlb(iommu, did, addr | ih, mask,
1359
						DMA_TLB_PSI_FLUSH);
1360 1361

	/*
1362 1363
	 * In caching mode, changes of pages from non-present to present require
	 * flush. However, device IOTLB doesn't need to be flushed in this case.
1364
	 */
1365
	if (!cap_caching_mode(iommu->cap) || !map)
Y
Yu Zhao 已提交
1366
		iommu_flush_dev_iotlb(iommu->domains[did], addr, mask);
1367 1368
}

M
mark gross 已提交
1369 1370 1371 1372 1373
static void iommu_disable_protect_mem_regions(struct intel_iommu *iommu)
{
	u32 pmen;
	unsigned long flags;

1374
	raw_spin_lock_irqsave(&iommu->register_lock, flags);
M
mark gross 已提交
1375 1376 1377 1378 1379 1380 1381 1382
	pmen = readl(iommu->reg + DMAR_PMEN_REG);
	pmen &= ~DMA_PMEN_EPM;
	writel(pmen, iommu->reg + DMAR_PMEN_REG);

	/* wait for the protected region status bit to clear */
	IOMMU_WAIT_OP(iommu, DMAR_PMEN_REG,
		readl, !(pmen & DMA_PMEN_PRS), pmen);

1383
	raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
M
mark gross 已提交
1384 1385
}

1386 1387 1388 1389 1390
static int iommu_enable_translation(struct intel_iommu *iommu)
{
	u32 sts;
	unsigned long flags;

1391
	raw_spin_lock_irqsave(&iommu->register_lock, flags);
1392 1393
	iommu->gcmd |= DMA_GCMD_TE;
	writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);
1394 1395 1396

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1397
		      readl, (sts & DMA_GSTS_TES), sts);
1398

1399
	raw_spin_unlock_irqrestore(&iommu->register_lock, flags);
1400 1401 1402 1403 1404 1405 1406 1407
	return 0;
}

static int iommu_disable_translation(struct intel_iommu *iommu)
{
	u32 sts;
	unsigned long flag;

1408
	raw_spin_lock_irqsave(&iommu->register_lock, flag);
1409 1410 1411 1412 1413
	iommu->gcmd &= ~DMA_GCMD_TE;
	writel(iommu->gcmd, iommu->reg + DMAR_GCMD_REG);

	/* Make sure hardware complete it */
	IOMMU_WAIT_OP(iommu, DMAR_GSTS_REG,
1414
		      readl, (!(sts & DMA_GSTS_TES)), sts);
1415

1416
	raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
1417 1418 1419
	return 0;
}

1420

1421 1422 1423 1424 1425 1426
static int iommu_init_domains(struct intel_iommu *iommu)
{
	unsigned long ndomains;
	unsigned long nlongs;

	ndomains = cap_ndoms(iommu->cap);
1427 1428
	pr_debug("IOMMU%d: Number of Domains supported <%ld>\n",
		 iommu->seq_id, ndomains);
1429 1430
	nlongs = BITS_TO_LONGS(ndomains);

1431 1432
	spin_lock_init(&iommu->lock);

1433 1434 1435 1436 1437
	/* TBD: there might be 64K domains,
	 * consider other allocation for future chip
	 */
	iommu->domain_ids = kcalloc(nlongs, sizeof(unsigned long), GFP_KERNEL);
	if (!iommu->domain_ids) {
1438 1439
		pr_err("IOMMU%d: allocating domain id array failed\n",
		       iommu->seq_id);
1440 1441 1442 1443 1444
		return -ENOMEM;
	}
	iommu->domains = kcalloc(ndomains, sizeof(struct dmar_domain *),
			GFP_KERNEL);
	if (!iommu->domains) {
1445 1446 1447 1448
		pr_err("IOMMU%d: allocating domain array failed\n",
		       iommu->seq_id);
		kfree(iommu->domain_ids);
		iommu->domain_ids = NULL;
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460
		return -ENOMEM;
	}

	/*
	 * if Caching mode is set, then invalid translations are tagged
	 * with domainid 0. Hence we need to pre-allocate it.
	 */
	if (cap_caching_mode(iommu->cap))
		set_bit(0, iommu->domain_ids);
	return 0;
}

1461
static void free_dmar_iommu(struct intel_iommu *iommu)
1462 1463
{
	struct dmar_domain *domain;
1464
	int i, count;
1465
	unsigned long flags;
1466

1467
	if ((iommu->domains) && (iommu->domain_ids)) {
1468
		for_each_set_bit(i, iommu->domain_ids, cap_ndoms(iommu->cap)) {
1469 1470 1471 1472 1473 1474 1475
			/*
			 * Domain id 0 is reserved for invalid translation
			 * if hardware supports caching mode.
			 */
			if (cap_caching_mode(iommu->cap) && i == 0)
				continue;

1476 1477 1478 1479
			domain = iommu->domains[i];
			clear_bit(i, iommu->domain_ids);

			spin_lock_irqsave(&domain->iommu_lock, flags);
1480 1481
			count = --domain->iommu_count;
			spin_unlock_irqrestore(&domain->iommu_lock, flags);
1482 1483
			if (count == 0)
				domain_exit(domain);
1484
		}
1485 1486 1487 1488 1489 1490 1491
	}

	if (iommu->gcmd & DMA_GCMD_TE)
		iommu_disable_translation(iommu);

	kfree(iommu->domains);
	kfree(iommu->domain_ids);
1492 1493
	iommu->domains = NULL;
	iommu->domain_ids = NULL;
1494

W
Weidong Han 已提交
1495 1496
	g_iommus[iommu->seq_id] = NULL;

1497 1498 1499 1500
	/* free context mapping */
	free_context_table(iommu);
}

1501
static struct dmar_domain *alloc_domain(bool vm)
1502
{
1503 1504
	/* domain id for virtual machine, it won't be set in context */
	static atomic_t vm_domid = ATOMIC_INIT(0);
1505 1506 1507 1508 1509 1510
	struct dmar_domain *domain;

	domain = alloc_domain_mem();
	if (!domain)
		return NULL;

1511
	domain->nid = -1;
1512
	domain->iommu_count = 0;
1513
	memset(domain->iommu_bmp, 0, sizeof(domain->iommu_bmp));
1514
	domain->flags = 0;
1515 1516 1517 1518 1519 1520
	spin_lock_init(&domain->iommu_lock);
	INIT_LIST_HEAD(&domain->devices);
	if (vm) {
		domain->id = atomic_inc_return(&vm_domid);
		domain->flags = DOMAIN_FLAG_VIRTUAL_MACHINE;
	}
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531

	return domain;
}

static int iommu_attach_domain(struct dmar_domain *domain,
			       struct intel_iommu *iommu)
{
	int num;
	unsigned long ndomains;
	unsigned long flags;

1532 1533 1534
	ndomains = cap_ndoms(iommu->cap);

	spin_lock_irqsave(&iommu->lock, flags);
1535

1536 1537 1538 1539
	num = find_first_zero_bit(iommu->domain_ids, ndomains);
	if (num >= ndomains) {
		spin_unlock_irqrestore(&iommu->lock, flags);
		printk(KERN_ERR "IOMMU: no free domain ids\n");
1540
		return -ENOMEM;
1541 1542 1543
	}

	domain->id = num;
1544
	domain->iommu_count++;
1545
	set_bit(num, iommu->domain_ids);
1546
	set_bit(iommu->seq_id, domain->iommu_bmp);
1547 1548 1549
	iommu->domains[num] = domain;
	spin_unlock_irqrestore(&iommu->lock, flags);

1550
	return 0;
1551 1552
}

1553 1554
static void iommu_detach_domain(struct dmar_domain *domain,
				struct intel_iommu *iommu)
1555 1556
{
	unsigned long flags;
1557
	int num, ndomains;
1558

1559
	spin_lock_irqsave(&iommu->lock, flags);
1560
	ndomains = cap_ndoms(iommu->cap);
1561
	for_each_set_bit(num, iommu->domain_ids, ndomains) {
1562
		if (iommu->domains[num] == domain) {
1563 1564
			clear_bit(num, iommu->domain_ids);
			iommu->domains[num] = NULL;
1565 1566 1567
			break;
		}
	}
1568
	spin_unlock_irqrestore(&iommu->lock, flags);
1569 1570 1571
}

static struct iova_domain reserved_iova_list;
M
Mark Gross 已提交
1572
static struct lock_class_key reserved_rbtree_key;
1573

1574
static int dmar_init_reserved_ranges(void)
1575 1576 1577 1578 1579
{
	struct pci_dev *pdev = NULL;
	struct iova *iova;
	int i;

D
David Miller 已提交
1580
	init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
1581

M
Mark Gross 已提交
1582 1583 1584
	lockdep_set_class(&reserved_iova_list.iova_rbtree_lock,
		&reserved_rbtree_key);

1585 1586 1587
	/* IOAPIC ranges shouldn't be accessed by DMA */
	iova = reserve_iova(&reserved_iova_list, IOVA_PFN(IOAPIC_RANGE_START),
		IOVA_PFN(IOAPIC_RANGE_END));
1588
	if (!iova) {
1589
		printk(KERN_ERR "Reserve IOAPIC range failed\n");
1590 1591
		return -ENODEV;
	}
1592 1593 1594 1595 1596 1597 1598 1599 1600

	/* Reserve all PCI MMIO to avoid peer-to-peer access */
	for_each_pci_dev(pdev) {
		struct resource *r;

		for (i = 0; i < PCI_NUM_RESOURCES; i++) {
			r = &pdev->resource[i];
			if (!r->flags || !(r->flags & IORESOURCE_MEM))
				continue;
1601 1602 1603
			iova = reserve_iova(&reserved_iova_list,
					    IOVA_PFN(r->start),
					    IOVA_PFN(r->end));
1604
			if (!iova) {
1605
				printk(KERN_ERR "Reserve iova failed\n");
1606 1607
				return -ENODEV;
			}
1608 1609
		}
	}
1610
	return 0;
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
}

static void domain_reserve_special_ranges(struct dmar_domain *domain)
{
	copy_reserved_iova(&reserved_iova_list, &domain->iovad);
}

static inline int guestwidth_to_adjustwidth(int gaw)
{
	int agaw;
	int r = (gaw - 12) % 9;

	if (r == 0)
		agaw = gaw;
	else
		agaw = gaw + 9 - r;
	if (agaw > 64)
		agaw = 64;
	return agaw;
}

static int domain_init(struct dmar_domain *domain, int guest_width)
{
	struct intel_iommu *iommu;
	int adjust_width, agaw;
	unsigned long sagaw;

D
David Miller 已提交
1638
	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
1639 1640 1641
	domain_reserve_special_ranges(domain);

	/* calculate AGAW */
1642
	iommu = domain_get_iommu(domain);
1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
	if (guest_width > cap_mgaw(iommu->cap))
		guest_width = cap_mgaw(iommu->cap);
	domain->gaw = guest_width;
	adjust_width = guestwidth_to_adjustwidth(guest_width);
	agaw = width_to_agaw(adjust_width);
	sagaw = cap_sagaw(iommu->cap);
	if (!test_bit(agaw, &sagaw)) {
		/* hardware doesn't support it, choose a bigger one */
		pr_debug("IOMMU: hardware doesn't support agaw %d\n", agaw);
		agaw = find_next_bit(&sagaw, 5, agaw);
		if (agaw >= 5)
			return -ENODEV;
	}
	domain->agaw = agaw;

W
Weidong Han 已提交
1658 1659 1660 1661 1662
	if (ecap_coherent(iommu->ecap))
		domain->iommu_coherency = 1;
	else
		domain->iommu_coherency = 0;

1663 1664 1665 1666 1667
	if (ecap_sc_support(iommu->ecap))
		domain->iommu_snooping = 1;
	else
		domain->iommu_snooping = 0;

1668 1669 1670 1671 1672
	if (intel_iommu_superpage)
		domain->iommu_superpage = fls(cap_super_page_val(iommu->cap));
	else
		domain->iommu_superpage = 0;

1673
	domain->nid = iommu->node;
1674

1675
	/* always allocate the top pgd */
1676
	domain->pgd = (struct dma_pte *)alloc_pgtable_page(domain->nid);
1677 1678
	if (!domain->pgd)
		return -ENOMEM;
F
Fenghua Yu 已提交
1679
	__iommu_flush_cache(iommu, domain->pgd, PAGE_SIZE);
1680 1681 1682 1683 1684
	return 0;
}

static void domain_exit(struct dmar_domain *domain)
{
1685 1686
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu;
1687
	struct page *freelist = NULL;
1688 1689 1690 1691 1692

	/* Domain 0 is reserved, so dont process it */
	if (!domain)
		return;

1693 1694 1695 1696
	/* Flush any lazy unmaps that may reference this domain */
	if (!intel_iommu_strict)
		flush_unmaps_timeout(0);

1697
	/* remove associated devices */
1698
	domain_remove_dev_info(domain);
1699

1700 1701 1702
	/* destroy iovas */
	put_iova_domain(&domain->iovad);

1703
	freelist = domain_unmap(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
1704

1705
	/* clear attached or cached domains */
1706
	rcu_read_lock();
1707
	for_each_active_iommu(iommu, drhd)
1708 1709
		if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
		    test_bit(iommu->seq_id, domain->iommu_bmp))
1710
			iommu_detach_domain(domain, iommu);
1711
	rcu_read_unlock();
1712

1713 1714
	dma_free_pagelist(freelist);

1715 1716 1717
	free_domain_mem(domain);
}

1718 1719 1720
static int domain_context_mapping_one(struct dmar_domain *domain,
				      struct intel_iommu *iommu,
				      u8 bus, u8 devfn, int translation)
1721 1722 1723
{
	struct context_entry *context;
	unsigned long flags;
1724 1725 1726 1727 1728
	struct dma_pte *pgd;
	unsigned long num;
	unsigned long ndomains;
	int id;
	int agaw;
Y
Yu Zhao 已提交
1729
	struct device_domain_info *info = NULL;
1730 1731 1732

	pr_debug("Set context mapping for %02x:%02x.%d\n",
		bus, PCI_SLOT(devfn), PCI_FUNC(devfn));
F
Fenghua Yu 已提交
1733

1734
	BUG_ON(!domain->pgd);
F
Fenghua Yu 已提交
1735 1736
	BUG_ON(translation != CONTEXT_TT_PASS_THROUGH &&
	       translation != CONTEXT_TT_MULTI_LEVEL);
W
Weidong Han 已提交
1737

1738 1739 1740 1741
	context = device_to_context_entry(iommu, bus, devfn);
	if (!context)
		return -ENOMEM;
	spin_lock_irqsave(&iommu->lock, flags);
1742
	if (context_present(context)) {
1743 1744 1745 1746
		spin_unlock_irqrestore(&iommu->lock, flags);
		return 0;
	}

1747 1748 1749
	id = domain->id;
	pgd = domain->pgd;

1750 1751
	if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
	    domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) {
1752 1753 1754 1755
		int found = 0;

		/* find an available domain id for this device in iommu */
		ndomains = cap_ndoms(iommu->cap);
1756
		for_each_set_bit(num, iommu->domain_ids, ndomains) {
1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
			if (iommu->domains[num] == domain) {
				id = num;
				found = 1;
				break;
			}
		}

		if (found == 0) {
			num = find_first_zero_bit(iommu->domain_ids, ndomains);
			if (num >= ndomains) {
				spin_unlock_irqrestore(&iommu->lock, flags);
				printk(KERN_ERR "IOMMU: no free domain ids\n");
				return -EFAULT;
			}

			set_bit(num, iommu->domain_ids);
			iommu->domains[num] = domain;
			id = num;
		}

		/* Skip top levels of page tables for
		 * iommu which has less agaw than default.
1779
		 * Unnecessary for PT mode.
1780
		 */
1781 1782 1783 1784 1785 1786 1787
		if (translation != CONTEXT_TT_PASS_THROUGH) {
			for (agaw = domain->agaw; agaw != iommu->agaw; agaw--) {
				pgd = phys_to_virt(dma_pte_addr(pgd));
				if (!dma_pte_present(pgd)) {
					spin_unlock_irqrestore(&iommu->lock, flags);
					return -ENOMEM;
				}
1788 1789 1790 1791 1792
			}
		}
	}

	context_set_domain_id(context, id);
F
Fenghua Yu 已提交
1793

Y
Yu Zhao 已提交
1794
	if (translation != CONTEXT_TT_PASS_THROUGH) {
1795
		info = iommu_support_dev_iotlb(domain, iommu, bus, devfn);
Y
Yu Zhao 已提交
1796 1797 1798
		translation = info ? CONTEXT_TT_DEV_IOTLB :
				     CONTEXT_TT_MULTI_LEVEL;
	}
F
Fenghua Yu 已提交
1799 1800 1801 1802
	/*
	 * In pass through mode, AW must be programmed to indicate the largest
	 * AGAW value supported by hardware. And ASR is ignored by hardware.
	 */
Y
Yu Zhao 已提交
1803
	if (unlikely(translation == CONTEXT_TT_PASS_THROUGH))
F
Fenghua Yu 已提交
1804
		context_set_address_width(context, iommu->msagaw);
Y
Yu Zhao 已提交
1805 1806 1807 1808
	else {
		context_set_address_root(context, virt_to_phys(pgd));
		context_set_address_width(context, iommu->agaw);
	}
F
Fenghua Yu 已提交
1809 1810

	context_set_translation_type(context, translation);
1811 1812
	context_set_fault_enable(context);
	context_set_present(context);
W
Weidong Han 已提交
1813
	domain_flush_cache(domain, context, sizeof(*context));
1814

1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825
	/*
	 * It's a non-present to present mapping. If hardware doesn't cache
	 * non-present entry we only need to flush the write-buffer. If the
	 * _does_ cache non-present entries, then it does so in the special
	 * domain #0, which we have to flush:
	 */
	if (cap_caching_mode(iommu->cap)) {
		iommu->flush.flush_context(iommu, 0,
					   (((u16)bus) << 8) | devfn,
					   DMA_CCMD_MASK_NOBIT,
					   DMA_CCMD_DEVICE_INVL);
1826
		iommu->flush.flush_iotlb(iommu, domain->id, 0, 0, DMA_TLB_DSI_FLUSH);
1827
	} else {
1828
		iommu_flush_write_buffer(iommu);
1829
	}
Y
Yu Zhao 已提交
1830
	iommu_enable_dev_iotlb(info);
1831
	spin_unlock_irqrestore(&iommu->lock, flags);
1832 1833

	spin_lock_irqsave(&domain->iommu_lock, flags);
1834
	if (!test_and_set_bit(iommu->seq_id, domain->iommu_bmp)) {
1835
		domain->iommu_count++;
1836 1837
		if (domain->iommu_count == 1)
			domain->nid = iommu->node;
1838
		domain_update_iommu_cap(domain);
1839 1840
	}
	spin_unlock_irqrestore(&domain->iommu_lock, flags);
1841 1842 1843 1844
	return 0;
}

static int
1845 1846
domain_context_mapping(struct dmar_domain *domain, struct device *dev,
		       int translation)
1847 1848
{
	int ret;
1849
	struct pci_dev *pdev, *tmp, *parent;
1850
	struct intel_iommu *iommu;
1851
	u8 bus, devfn;
1852

1853
	iommu = device_to_iommu(dev, &bus, &devfn);
1854 1855
	if (!iommu)
		return -ENODEV;
1856

1857
	ret = domain_context_mapping_one(domain, iommu, bus, devfn,
F
Fenghua Yu 已提交
1858
					 translation);
1859
	if (ret || !dev_is_pci(dev))
1860 1861 1862
		return ret;

	/* dependent device mapping */
1863
	pdev = to_pci_dev(dev);
1864 1865 1866 1867 1868 1869
	tmp = pci_find_upstream_pcie_bridge(pdev);
	if (!tmp)
		return 0;
	/* Secondary interface's bus number and devfn 0 */
	parent = pdev->bus->self;
	while (parent != tmp) {
1870
		ret = domain_context_mapping_one(domain, iommu,
1871
						 parent->bus->number,
F
Fenghua Yu 已提交
1872
						 parent->devfn, translation);
1873 1874 1875 1876
		if (ret)
			return ret;
		parent = parent->bus->self;
	}
1877
	if (pci_is_pcie(tmp)) /* this is a PCIe-to-PCI bridge */
1878
		return domain_context_mapping_one(domain, iommu,
F
Fenghua Yu 已提交
1879 1880
					tmp->subordinate->number, 0,
					translation);
1881
	else /* this is a legacy PCI bridge */
1882
		return domain_context_mapping_one(domain, iommu,
1883
						  tmp->bus->number,
F
Fenghua Yu 已提交
1884 1885
						  tmp->devfn,
						  translation);
1886 1887
}

1888
static int domain_context_mapped(struct device *dev)
1889 1890
{
	int ret;
1891
	struct pci_dev *pdev, *tmp, *parent;
W
Weidong Han 已提交
1892
	struct intel_iommu *iommu;
1893
	u8 bus, devfn;
W
Weidong Han 已提交
1894

1895
	iommu = device_to_iommu(dev, &bus, &devfn);
W
Weidong Han 已提交
1896 1897
	if (!iommu)
		return -ENODEV;
1898

1899
	ret = device_context_mapped(iommu, bus, devfn);
1900
	if (!ret || !dev_is_pci(dev))
1901
		return ret;
1902

1903
	/* dependent device mapping */
1904
	pdev = to_pci_dev(dev);
1905 1906 1907 1908 1909 1910
	tmp = pci_find_upstream_pcie_bridge(pdev);
	if (!tmp)
		return ret;
	/* Secondary interface's bus number and devfn 0 */
	parent = pdev->bus->self;
	while (parent != tmp) {
1911
		ret = device_context_mapped(iommu, parent->bus->number,
1912
					    parent->devfn);
1913 1914 1915 1916
		if (!ret)
			return ret;
		parent = parent->bus->self;
	}
1917
	if (pci_is_pcie(tmp))
1918 1919
		return device_context_mapped(iommu, tmp->subordinate->number,
					     0);
1920
	else
1921 1922
		return device_context_mapped(iommu, tmp->bus->number,
					     tmp->devfn);
1923 1924
}

1925 1926 1927 1928 1929 1930 1931 1932
/* Returns a number of VTD pages, but aligned to MM page size */
static inline unsigned long aligned_nrpages(unsigned long host_addr,
					    size_t size)
{
	host_addr &= ~PAGE_MASK;
	return PAGE_ALIGN(host_addr + size) >> VTD_PAGE_SHIFT;
}

1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960
/* Return largest possible superpage level for a given mapping */
static inline int hardware_largepage_caps(struct dmar_domain *domain,
					  unsigned long iov_pfn,
					  unsigned long phy_pfn,
					  unsigned long pages)
{
	int support, level = 1;
	unsigned long pfnmerge;

	support = domain->iommu_superpage;

	/* To use a large page, the virtual *and* physical addresses
	   must be aligned to 2MiB/1GiB/etc. Lower bits set in either
	   of them will mean we have to use smaller pages. So just
	   merge them and check both at once. */
	pfnmerge = iov_pfn | phy_pfn;

	while (support && !(pfnmerge & ~VTD_STRIDE_MASK)) {
		pages >>= VTD_STRIDE_SHIFT;
		if (!pages)
			break;
		pfnmerge >>= VTD_STRIDE_SHIFT;
		level++;
		support--;
	}
	return level;
}

1961 1962 1963
static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
			    struct scatterlist *sg, unsigned long phys_pfn,
			    unsigned long nr_pages, int prot)
1964 1965
{
	struct dma_pte *first_pte = NULL, *pte = NULL;
1966
	phys_addr_t uninitialized_var(pteval);
1967
	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
1968
	unsigned long sg_res;
1969 1970
	unsigned int largepage_lvl = 0;
	unsigned long lvl_pages = 0;
1971 1972 1973 1974 1975 1976 1977 1978

	BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);

	if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
		return -EINVAL;

	prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;

1979 1980 1981 1982 1983 1984 1985
	if (sg)
		sg_res = 0;
	else {
		sg_res = nr_pages + 1;
		pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
	}

1986
	while (nr_pages > 0) {
1987 1988
		uint64_t tmp;

1989
		if (!sg_res) {
1990
			sg_res = aligned_nrpages(sg->offset, sg->length);
1991 1992 1993
			sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
			sg->dma_length = sg->length;
			pteval = page_to_phys(sg_page(sg)) | prot;
1994
			phys_pfn = pteval >> VTD_PAGE_SHIFT;
1995
		}
1996

1997
		if (!pte) {
1998 1999
			largepage_lvl = hardware_largepage_caps(domain, iov_pfn, phys_pfn, sg_res);

2000
			first_pte = pte = pfn_to_dma_pte(domain, iov_pfn, &largepage_lvl);
2001 2002
			if (!pte)
				return -ENOMEM;
2003
			/* It is large page*/
2004
			if (largepage_lvl > 1) {
2005
				pteval |= DMA_PTE_LARGE_PAGE;
2006 2007 2008 2009 2010 2011 2012
				/* Ensure that old small page tables are removed to make room
				   for superpage, if they exist. */
				dma_pte_clear_range(domain, iov_pfn,
						    iov_pfn + lvl_to_nr_pages(largepage_lvl) - 1);
				dma_pte_free_pagetable(domain, iov_pfn,
						       iov_pfn + lvl_to_nr_pages(largepage_lvl) - 1);
			} else {
2013
				pteval &= ~(uint64_t)DMA_PTE_LARGE_PAGE;
2014
			}
2015

2016 2017 2018 2019
		}
		/* We don't need lock here, nobody else
		 * touches the iova range
		 */
2020
		tmp = cmpxchg64_local(&pte->val, 0ULL, pteval);
2021
		if (tmp) {
2022
			static int dumps = 5;
2023 2024
			printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
			       iov_pfn, tmp, (unsigned long long)pteval);
2025 2026 2027 2028 2029 2030
			if (dumps) {
				dumps--;
				debug_dma_dump_mappings(NULL);
			}
			WARN_ON(1);
		}
2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053

		lvl_pages = lvl_to_nr_pages(largepage_lvl);

		BUG_ON(nr_pages < lvl_pages);
		BUG_ON(sg_res < lvl_pages);

		nr_pages -= lvl_pages;
		iov_pfn += lvl_pages;
		phys_pfn += lvl_pages;
		pteval += lvl_pages * VTD_PAGE_SIZE;
		sg_res -= lvl_pages;

		/* If the next PTE would be the first in a new page, then we
		   need to flush the cache on the entries we've just written.
		   And then we'll need to recalculate 'pte', so clear it and
		   let it get set again in the if (!pte) block above.

		   If we're done (!nr_pages) we need to flush the cache too.

		   Also if we've been setting superpages, we may need to
		   recalculate 'pte' and switch back to smaller pages for the
		   end of the mapping, if the trailing size is not enough to
		   use another superpage (i.e. sg_res < lvl_pages). */
2054
		pte++;
2055 2056
		if (!nr_pages || first_pte_in_page(pte) ||
		    (largepage_lvl > 1 && sg_res < lvl_pages)) {
2057 2058 2059 2060
			domain_flush_cache(domain, first_pte,
					   (void *)pte - (void *)first_pte);
			pte = NULL;
		}
2061 2062

		if (!sg_res && nr_pages)
2063 2064 2065 2066 2067
			sg = sg_next(sg);
	}
	return 0;
}

2068 2069 2070
static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
				    struct scatterlist *sg, unsigned long nr_pages,
				    int prot)
2071
{
2072 2073
	return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
}
2074

2075 2076 2077 2078 2079
static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
				     unsigned long phys_pfn, unsigned long nr_pages,
				     int prot)
{
	return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
2080 2081
}

2082
static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
2083
{
2084 2085
	if (!iommu)
		return;
2086 2087 2088

	clear_context_table(iommu, bus, devfn);
	iommu->flush.flush_context(iommu, 0, 0, 0,
2089
					   DMA_CCMD_GLOBAL_INVL);
2090
	iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
2091 2092
}

2093 2094 2095 2096 2097 2098
static inline void unlink_domain_info(struct device_domain_info *info)
{
	assert_spin_locked(&device_domain_lock);
	list_del(&info->link);
	list_del(&info->global);
	if (info->dev)
2099
		info->dev->archdata.iommu = NULL;
2100 2101
}

2102 2103 2104
static void domain_remove_dev_info(struct dmar_domain *domain)
{
	struct device_domain_info *info;
2105
	unsigned long flags, flags2;
2106 2107 2108 2109 2110

	spin_lock_irqsave(&device_domain_lock, flags);
	while (!list_empty(&domain->devices)) {
		info = list_entry(domain->devices.next,
			struct device_domain_info, link);
2111
		unlink_domain_info(info);
2112 2113
		spin_unlock_irqrestore(&device_domain_lock, flags);

Y
Yu Zhao 已提交
2114
		iommu_disable_dev_iotlb(info);
2115
		iommu_detach_dev(info->iommu, info->bus, info->devfn);
2116

2117
		if (domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) {
2118
			iommu_detach_dependent_devices(info->iommu, info->dev);
2119 2120 2121 2122
			/* clear this iommu in iommu_bmp, update iommu count
			 * and capabilities
			 */
			spin_lock_irqsave(&domain->iommu_lock, flags2);
2123
			if (test_and_clear_bit(info->iommu->seq_id,
2124 2125 2126 2127 2128 2129 2130 2131
					       domain->iommu_bmp)) {
				domain->iommu_count--;
				domain_update_iommu_cap(domain);
			}
			spin_unlock_irqrestore(&domain->iommu_lock, flags2);
		}

		free_devinfo_mem(info);
2132 2133 2134 2135 2136 2137 2138
		spin_lock_irqsave(&device_domain_lock, flags);
	}
	spin_unlock_irqrestore(&device_domain_lock, flags);
}

/*
 * find_domain
2139
 * Note: we use struct device->archdata.iommu stores the info
2140
 */
2141
static struct dmar_domain *find_domain(struct device *dev)
2142 2143 2144 2145
{
	struct device_domain_info *info;

	/* No lock here, assumes no domain exit in normal case */
2146
	info = dev->archdata.iommu;
2147 2148 2149 2150 2151
	if (info)
		return info->domain;
	return NULL;
}

2152
static inline struct device_domain_info *
2153 2154 2155 2156 2157
dmar_search_domain_by_dev_info(int segment, int bus, int devfn)
{
	struct device_domain_info *info;

	list_for_each_entry(info, &device_domain_list, global)
2158
		if (info->iommu->segment == segment && info->bus == bus &&
2159
		    info->devfn == devfn)
2160
			return info;
2161 2162 2163 2164

	return NULL;
}

2165
static struct dmar_domain *dmar_insert_dev_info(struct intel_iommu *iommu,
2166
						int bus, int devfn,
2167 2168
						struct device *dev,
						struct dmar_domain *domain)
2169
{
2170
	struct dmar_domain *found = NULL;
2171 2172 2173 2174 2175
	struct device_domain_info *info;
	unsigned long flags;

	info = alloc_devinfo_mem();
	if (!info)
2176
		return NULL;
2177 2178 2179 2180 2181

	info->bus = bus;
	info->devfn = devfn;
	info->dev = dev;
	info->domain = domain;
2182
	info->iommu = iommu;
2183 2184 2185 2186 2187
	if (!dev)
		domain->flags |= DOMAIN_FLAG_P2P_MULTIPLE_DEVICES;

	spin_lock_irqsave(&device_domain_lock, flags);
	if (dev)
2188
		found = find_domain(dev);
2189 2190
	else {
		struct device_domain_info *info2;
2191
		info2 = dmar_search_domain_by_dev_info(iommu->segment, bus, devfn);
2192 2193 2194
		if (info2)
			found = info2->domain;
	}
2195 2196 2197
	if (found) {
		spin_unlock_irqrestore(&device_domain_lock, flags);
		free_devinfo_mem(info);
2198 2199
		/* Caller must free the original domain */
		return found;
2200 2201
	}

2202 2203 2204 2205 2206 2207 2208
	list_add(&info->link, &domain->devices);
	list_add(&info->global, &device_domain_list);
	if (dev)
		dev->archdata.iommu = info;
	spin_unlock_irqrestore(&device_domain_lock, flags);

	return domain;
2209 2210
}

2211
/* domain is initialized */
2212
static struct dmar_domain *get_domain_for_dev(struct device *dev, int gaw)
2213
{
2214
	struct dmar_domain *domain, *free = NULL;
2215 2216
	struct intel_iommu *iommu = NULL;
	struct device_domain_info *info;
2217
	struct pci_dev *dev_tmp = NULL;
2218
	unsigned long flags;
2219
	u8 bus, devfn, bridge_bus, bridge_devfn;
2220

2221
	domain = find_domain(dev);
2222 2223 2224
	if (domain)
		return domain;

2225 2226 2227
	if (dev_is_pci(dev)) {
		struct pci_dev *pdev = to_pci_dev(dev);
		u16 segment;
2228

2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248
		segment = pci_domain_nr(pdev->bus);
		dev_tmp = pci_find_upstream_pcie_bridge(pdev);
		if (dev_tmp) {
			if (pci_is_pcie(dev_tmp)) {
				bridge_bus = dev_tmp->subordinate->number;
				bridge_devfn = 0;
			} else {
				bridge_bus = dev_tmp->bus->number;
				bridge_devfn = dev_tmp->devfn;
			}
			spin_lock_irqsave(&device_domain_lock, flags);
			info = dmar_search_domain_by_dev_info(segment, bus, devfn);
			if (info) {
				iommu = info->iommu;
				domain = info->domain;
			}
			spin_unlock_irqrestore(&device_domain_lock, flags);
			/* pcie-pci bridge already has a domain, uses it */
			if (info)
				goto found_domain;
2249
		}
2250 2251
	}

2252 2253 2254
	iommu = device_to_iommu(dev, &bus, &devfn);
	if (!iommu)
		goto error;
2255

2256
	/* Allocate and initialize new domain for the device */
2257
	domain = alloc_domain(false);
2258 2259 2260
	if (!domain)
		goto error;
	if (iommu_attach_domain(domain, iommu)) {
2261
		free_domain_mem(domain);
2262
		domain = NULL;
2263
		goto error;
2264
	}
2265 2266
	free = domain;
	if (domain_init(domain, gaw))
2267 2268 2269 2270
		goto error;

	/* register pcie-to-pci device */
	if (dev_tmp) {
2271 2272
		domain = dmar_insert_dev_info(iommu, bridge_bus, bridge_devfn,
					      NULL, domain);
2273
		if (!domain)
2274 2275 2276 2277
			goto error;
	}

found_domain:
2278
	domain = dmar_insert_dev_info(iommu, bus, devfn, dev, domain);
2279
error:
2280
	if (free != domain)
2281
		domain_exit(free);
2282 2283

	return domain;
2284 2285
}

2286
static int iommu_identity_mapping;
2287 2288 2289
#define IDENTMAP_ALL		1
#define IDENTMAP_GFX		2
#define IDENTMAP_AZALIA		4
2290

2291 2292 2293
static int iommu_domain_identity_map(struct dmar_domain *domain,
				     unsigned long long start,
				     unsigned long long end)
2294
{
2295 2296 2297 2298 2299
	unsigned long first_vpfn = start >> VTD_PAGE_SHIFT;
	unsigned long last_vpfn = end >> VTD_PAGE_SHIFT;

	if (!reserve_iova(&domain->iovad, dma_to_mm_pfn(first_vpfn),
			  dma_to_mm_pfn(last_vpfn))) {
2300
		printk(KERN_ERR "IOMMU: reserve iova failed\n");
2301
		return -ENOMEM;
2302 2303
	}

2304 2305
	pr_debug("Mapping reserved region %llx-%llx for domain %d\n",
		 start, end, domain->id);
2306 2307 2308 2309
	/*
	 * RMRR range might have overlap with physical memory range,
	 * clear it first
	 */
2310
	dma_pte_clear_range(domain, first_vpfn, last_vpfn);
2311

2312 2313
	return domain_pfn_mapping(domain, first_vpfn, first_vpfn,
				  last_vpfn - first_vpfn + 1,
2314
				  DMA_PTE_READ|DMA_PTE_WRITE);
2315 2316
}

2317
static int iommu_prepare_identity_map(struct device *dev,
2318 2319 2320 2321 2322 2323
				      unsigned long long start,
				      unsigned long long end)
{
	struct dmar_domain *domain;
	int ret;

2324
	domain = get_domain_for_dev(dev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
2325 2326 2327
	if (!domain)
		return -ENOMEM;

2328 2329 2330 2331 2332 2333
	/* For _hardware_ passthrough, don't bother. But for software
	   passthrough, we do it anyway -- it may indicate a memory
	   range which is reserved in E820, so which didn't get set
	   up to start with in si_domain */
	if (domain == si_domain && hw_pass_through) {
		printk("Ignoring identity map for HW passthrough device %s [0x%Lx - 0x%Lx]\n",
2334
		       dev_name(dev), start, end);
2335 2336 2337 2338 2339
		return 0;
	}

	printk(KERN_INFO
	       "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
2340
	       dev_name(dev), start, end);
2341
	
2342 2343 2344 2345 2346 2347 2348 2349 2350 2351
	if (end < start) {
		WARN(1, "Your BIOS is broken; RMRR ends before it starts!\n"
			"BIOS vendor: %s; Ver: %s; Product Version: %s\n",
			dmi_get_system_info(DMI_BIOS_VENDOR),
			dmi_get_system_info(DMI_BIOS_VERSION),
		     dmi_get_system_info(DMI_PRODUCT_VERSION));
		ret = -EIO;
		goto error;
	}

2352 2353 2354 2355 2356 2357 2358 2359 2360 2361
	if (end >> agaw_to_width(domain->agaw)) {
		WARN(1, "Your BIOS is broken; RMRR exceeds permitted address width (%d bits)\n"
		     "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
		     agaw_to_width(domain->agaw),
		     dmi_get_system_info(DMI_BIOS_VENDOR),
		     dmi_get_system_info(DMI_BIOS_VERSION),
		     dmi_get_system_info(DMI_PRODUCT_VERSION));
		ret = -EIO;
		goto error;
	}
2362

2363
	ret = iommu_domain_identity_map(domain, start, end);
2364 2365 2366 2367
	if (ret)
		goto error;

	/* context entry init */
2368
	ret = domain_context_mapping(domain, dev, CONTEXT_TT_MULTI_LEVEL);
2369 2370 2371 2372 2373 2374
	if (ret)
		goto error;

	return 0;

 error:
2375 2376 2377 2378 2379
	domain_exit(domain);
	return ret;
}

static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
2380
					 struct device *dev)
2381
{
2382
	if (dev->archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO)
2383
		return 0;
2384 2385
	return iommu_prepare_identity_map(dev, rmrr->base_address,
					  rmrr->end_address);
2386 2387
}

2388
#ifdef CONFIG_INTEL_IOMMU_FLOPPY_WA
2389 2390 2391 2392 2393 2394 2395 2396 2397
static inline void iommu_prepare_isa(void)
{
	struct pci_dev *pdev;
	int ret;

	pdev = pci_get_class(PCI_CLASS_BRIDGE_ISA << 8, NULL);
	if (!pdev)
		return;

2398
	printk(KERN_INFO "IOMMU: Prepare 0-16MiB unity mapping for LPC\n");
2399
	ret = iommu_prepare_identity_map(&pdev->dev, 0, 16*1024*1024 - 1);
2400 2401

	if (ret)
2402 2403
		printk(KERN_ERR "IOMMU: Failed to create 0-16MiB identity map; "
		       "floppy might not work\n");
2404 2405 2406 2407 2408 2409 2410

}
#else
static inline void iommu_prepare_isa(void)
{
	return;
}
2411
#endif /* !CONFIG_INTEL_IOMMU_FLPY_WA */
2412

2413
static int md_domain_init(struct dmar_domain *domain, int guest_width);
2414

2415
static int __init si_domain_init(int hw)
2416 2417 2418
{
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu;
2419
	int nid, ret = 0;
2420

2421
	si_domain = alloc_domain(false);
2422 2423 2424
	if (!si_domain)
		return -EFAULT;

2425 2426
	si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;

2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
	for_each_active_iommu(iommu, drhd) {
		ret = iommu_attach_domain(si_domain, iommu);
		if (ret) {
			domain_exit(si_domain);
			return -EFAULT;
		}
	}

	if (md_domain_init(si_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
		domain_exit(si_domain);
		return -EFAULT;
	}

2440 2441
	pr_debug("IOMMU: identity mapping domain is domain %d\n",
		 si_domain->id);
2442

2443 2444 2445
	if (hw)
		return 0;

2446
	for_each_online_node(nid) {
2447 2448 2449 2450 2451 2452 2453 2454 2455
		unsigned long start_pfn, end_pfn;
		int i;

		for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
			ret = iommu_domain_identity_map(si_domain,
					PFN_PHYS(start_pfn), PFN_PHYS(end_pfn));
			if (ret)
				return ret;
		}
2456 2457
	}

2458 2459 2460
	return 0;
}

2461
static int identity_mapping(struct device *dev)
2462 2463 2464 2465 2466 2467
{
	struct device_domain_info *info;

	if (likely(!iommu_identity_mapping))
		return 0;

2468
	info = dev->archdata.iommu;
2469 2470
	if (info && info != DUMMY_DEVICE_DOMAIN_INFO)
		return (info->domain == si_domain);
2471 2472 2473 2474 2475

	return 0;
}

static int domain_add_dev_info(struct dmar_domain *domain,
2476
			       struct device *dev, int translation)
2477
{
2478
	struct dmar_domain *ndomain;
2479
	struct intel_iommu *iommu;
2480
	u8 bus, devfn;
2481
	int ret;
2482

2483
	iommu = device_to_iommu(dev, &bus, &devfn);
2484 2485 2486
	if (!iommu)
		return -ENODEV;

2487
	ndomain = dmar_insert_dev_info(iommu, bus, devfn, dev, domain);
2488 2489
	if (ndomain != domain)
		return -EBUSY;
2490

2491
	ret = domain_context_mapping(domain, dev, translation);
2492
	if (ret) {
2493
		domain_remove_one_dev_info(domain, dev);
2494 2495 2496
		return ret;
	}

2497 2498 2499
	return 0;
}

2500
static bool device_has_rmrr(struct device *dev)
2501 2502
{
	struct dmar_rmrr_unit *rmrr;
2503
	struct device *tmp;
2504 2505
	int i;

2506
	rcu_read_lock();
2507
	for_each_rmrr_units(rmrr) {
2508 2509 2510 2511 2512 2513
		/*
		 * Return TRUE if this RMRR contains the device that
		 * is passed in.
		 */
		for_each_active_dev_scope(rmrr->devices,
					  rmrr->devices_cnt, i, tmp)
2514
			if (tmp == dev) {
2515
				rcu_read_unlock();
2516
				return true;
2517
			}
2518
	}
2519
	rcu_read_unlock();
2520 2521 2522
	return false;
}

2523
static int iommu_should_identity_map(struct device *dev, int startup)
2524
{
2525

2526 2527
	if (dev_is_pci(dev)) {
		struct pci_dev *pdev = to_pci_dev(dev);
2528

2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
		/*
		 * We want to prevent any device associated with an RMRR from
		 * getting placed into the SI Domain. This is done because
		 * problems exist when devices are moved in and out of domains
		 * and their respective RMRR info is lost. We exempt USB devices
		 * from this process due to their usage of RMRRs that are known
		 * to not be needed after BIOS hand-off to OS.
		 */
		if (device_has_rmrr(dev) &&
		    (pdev->class >> 8) != PCI_CLASS_SERIAL_USB)
			return 0;
2540

2541 2542
		if ((iommu_identity_mapping & IDENTMAP_AZALIA) && IS_AZALIA(pdev))
			return 1;
2543

2544 2545
		if ((iommu_identity_mapping & IDENTMAP_GFX) && IS_GFX_DEVICE(pdev))
			return 1;
2546

2547
		if (!(iommu_identity_mapping & IDENTMAP_ALL))
2548
			return 0;
2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572

		/*
		 * We want to start off with all devices in the 1:1 domain, and
		 * take them out later if we find they can't access all of memory.
		 *
		 * However, we can't do this for PCI devices behind bridges,
		 * because all PCI devices behind the same bridge will end up
		 * with the same source-id on their transactions.
		 *
		 * Practically speaking, we can't change things around for these
		 * devices at run-time, because we can't be sure there'll be no
		 * DMA transactions in flight for any of their siblings.
		 *
		 * So PCI devices (unless they're on the root bus) as well as
		 * their parent PCI-PCI or PCIe-PCI bridges must be left _out_ of
		 * the 1:1 domain, just in _case_ one of their siblings turns out
		 * not to be able to map all of memory.
		 */
		if (!pci_is_pcie(pdev)) {
			if (!pci_is_root_bus(pdev->bus))
				return 0;
			if (pdev->class >> 8 == PCI_CLASS_BRIDGE_PCI)
				return 0;
		} else if (pci_pcie_type(pdev) == PCI_EXP_TYPE_PCI_BRIDGE)
2573
			return 0;
2574 2575 2576 2577
	} else {
		if (device_has_rmrr(dev))
			return 0;
	}
2578

2579
	/*
2580
	 * At boot time, we don't yet know if devices will be 64-bit capable.
2581
	 * Assume that they will — if they turn out not to be, then we can
2582 2583
	 * take them out of the 1:1 domain later.
	 */
2584 2585 2586 2587 2588
	if (!startup) {
		/*
		 * If the device's dma_mask is less than the system's memory
		 * size then this is not a candidate for identity mapping.
		 */
2589
		u64 dma_mask = *dev->dma_mask;
2590

2591 2592 2593
		if (dev->coherent_dma_mask &&
		    dev->coherent_dma_mask < dma_mask)
			dma_mask = dev->coherent_dma_mask;
2594

2595
		return dma_mask >= dma_get_required_mask(dev);
2596
	}
2597 2598 2599 2600

	return 1;
}

2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621
static int __init dev_prepare_static_identity_mapping(struct device *dev, int hw)
{
	int ret;

	if (!iommu_should_identity_map(dev, 1))
		return 0;

	ret = domain_add_dev_info(si_domain, dev,
				  hw ? CONTEXT_TT_PASS_THROUGH :
				       CONTEXT_TT_MULTI_LEVEL);
	if (!ret)
		pr_info("IOMMU: %s identity mapping for device %s\n",
			hw ? "hardware" : "software", dev_name(dev));
	else if (ret == -ENODEV)
		/* device not associated with an iommu */
		ret = 0;

	return ret;
}


2622
static int __init iommu_prepare_static_identity_mapping(int hw)
2623 2624
{
	struct pci_dev *pdev = NULL;
2625 2626 2627 2628 2629
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu;
	struct device *dev;
	int i;
	int ret = 0;
2630

2631
	ret = si_domain_init(hw);
2632 2633 2634 2635
	if (ret)
		return -EFAULT;

	for_each_pci_dev(pdev) {
2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654
		ret = dev_prepare_static_identity_mapping(&pdev->dev, hw);
		if (ret)
			return ret;
	}

	for_each_active_iommu(iommu, drhd)
		for_each_active_dev_scope(drhd->devices, drhd->devices_cnt, i, dev) {
			struct acpi_device_physical_node *pn;
			struct acpi_device *adev;

			if (dev->bus != &acpi_bus_type)
				continue;
				
			adev= to_acpi_device(dev);
			mutex_lock(&adev->physical_node_lock);
			list_for_each_entry(pn, &adev->physical_node_list, node) {
				ret = dev_prepare_static_identity_mapping(pn->dev, hw);
				if (ret)
					break;
2655
			}
2656 2657 2658
			mutex_unlock(&adev->physical_node_lock);
			if (ret)
				return ret;
2659
		}
2660 2661 2662 2663

	return 0;
}

2664
static int __init init_dmars(void)
2665 2666 2667
{
	struct dmar_drhd_unit *drhd;
	struct dmar_rmrr_unit *rmrr;
2668
	struct device *dev;
2669
	struct intel_iommu *iommu;
2670
	int i, ret;
2671

2672 2673 2674 2675 2676 2677 2678
	/*
	 * for each drhd
	 *    allocate root
	 *    initialize and program root entry to not present
	 * endfor
	 */
	for_each_drhd_unit(drhd) {
M
mark gross 已提交
2679 2680 2681 2682 2683
		/*
		 * lock not needed as this is only incremented in the single
		 * threaded kernel __init code path all other access are read
		 * only
		 */
2684 2685 2686 2687 2688 2689
		if (g_num_of_iommus < IOMMU_UNITS_SUPPORTED) {
			g_num_of_iommus++;
			continue;
		}
		printk_once(KERN_ERR "intel-iommu: exceeded %d IOMMUs\n",
			  IOMMU_UNITS_SUPPORTED);
M
mark gross 已提交
2690 2691
	}

W
Weidong Han 已提交
2692 2693 2694 2695 2696 2697 2698 2699
	g_iommus = kcalloc(g_num_of_iommus, sizeof(struct intel_iommu *),
			GFP_KERNEL);
	if (!g_iommus) {
		printk(KERN_ERR "Allocating global iommu array failed\n");
		ret = -ENOMEM;
		goto error;
	}

2700 2701 2702
	deferred_flush = kzalloc(g_num_of_iommus *
		sizeof(struct deferred_flush_tables), GFP_KERNEL);
	if (!deferred_flush) {
M
mark gross 已提交
2703
		ret = -ENOMEM;
2704
		goto free_g_iommus;
M
mark gross 已提交
2705 2706
	}

2707
	for_each_active_iommu(iommu, drhd) {
W
Weidong Han 已提交
2708
		g_iommus[iommu->seq_id] = iommu;
2709

2710 2711
		ret = iommu_init_domains(iommu);
		if (ret)
2712
			goto free_iommu;
2713

2714 2715 2716
		/*
		 * TBD:
		 * we could share the same root & context tables
L
Lucas De Marchi 已提交
2717
		 * among all IOMMU's. Need to Split it later.
2718 2719 2720 2721
		 */
		ret = iommu_alloc_root_entry(iommu);
		if (ret) {
			printk(KERN_ERR "IOMMU: allocate root entry failed\n");
2722
			goto free_iommu;
2723
		}
F
Fenghua Yu 已提交
2724
		if (!ecap_pass_through(iommu->ecap))
2725
			hw_pass_through = 0;
2726 2727
	}

2728 2729 2730
	/*
	 * Start from the sane iommu hardware state.
	 */
2731
	for_each_active_iommu(iommu, drhd) {
2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750
		/*
		 * If the queued invalidation is already initialized by us
		 * (for example, while enabling interrupt-remapping) then
		 * we got the things already rolling from a sane state.
		 */
		if (iommu->qi)
			continue;

		/*
		 * Clear any previous faults.
		 */
		dmar_fault(-1, iommu);
		/*
		 * Disable queued invalidation if supported and already enabled
		 * before OS handover.
		 */
		dmar_disable_qi(iommu);
	}

2751
	for_each_active_iommu(iommu, drhd) {
2752 2753 2754 2755 2756 2757 2758
		if (dmar_enable_qi(iommu)) {
			/*
			 * Queued Invalidate not enabled, use Register Based
			 * Invalidate
			 */
			iommu->flush.flush_context = __iommu_flush_context;
			iommu->flush.flush_iotlb = __iommu_flush_iotlb;
Y
Yinghai Lu 已提交
2759
			printk(KERN_INFO "IOMMU %d 0x%Lx: using Register based "
2760
			       "invalidation\n",
Y
Yinghai Lu 已提交
2761
				iommu->seq_id,
2762
			       (unsigned long long)drhd->reg_base_addr);
2763 2764 2765
		} else {
			iommu->flush.flush_context = qi_flush_context;
			iommu->flush.flush_iotlb = qi_flush_iotlb;
Y
Yinghai Lu 已提交
2766
			printk(KERN_INFO "IOMMU %d 0x%Lx: using Queued "
2767
			       "invalidation\n",
Y
Yinghai Lu 已提交
2768
				iommu->seq_id,
2769
			       (unsigned long long)drhd->reg_base_addr);
2770 2771 2772
		}
	}

2773
	if (iommu_pass_through)
2774 2775
		iommu_identity_mapping |= IDENTMAP_ALL;

2776
#ifdef CONFIG_INTEL_IOMMU_BROKEN_GFX_WA
2777
	iommu_identity_mapping |= IDENTMAP_GFX;
2778
#endif
2779 2780 2781

	check_tylersburg_isoch();

2782
	/*
2783 2784 2785
	 * If pass through is not set or not enabled, setup context entries for
	 * identity mappings for rmrr, gfx, and isa and may fall back to static
	 * identity mapping if iommu_identity_mapping is set.
2786
	 */
2787 2788
	if (iommu_identity_mapping) {
		ret = iommu_prepare_static_identity_mapping(hw_pass_through);
F
Fenghua Yu 已提交
2789
		if (ret) {
2790
			printk(KERN_CRIT "Failed to setup IOMMU pass-through\n");
2791
			goto free_iommu;
2792 2793 2794
		}
	}
	/*
2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
	 * For each rmrr
	 *   for each dev attached to rmrr
	 *   do
	 *     locate drhd for dev, alloc domain for dev
	 *     allocate free domain
	 *     allocate page table entries for rmrr
	 *     if context not allocated for bus
	 *           allocate and init context
	 *           set present in root table for this bus
	 *     init context with domain, translation etc
	 *    endfor
	 * endfor
2807
	 */
2808 2809
	printk(KERN_INFO "IOMMU: Setting RMRR:\n");
	for_each_rmrr_units(rmrr) {
2810 2811
		/* some BIOS lists non-exist devices in DMAR table. */
		for_each_active_dev_scope(rmrr->devices, rmrr->devices_cnt,
2812
					  i, dev) {
2813
			ret = iommu_prepare_rmrr_dev(rmrr, dev);
2814 2815 2816
			if (ret)
				printk(KERN_ERR
				       "IOMMU: mapping reserved region failed\n");
2817
		}
F
Fenghua Yu 已提交
2818
	}
2819

2820 2821
	iommu_prepare_isa();

2822 2823 2824 2825 2826 2827 2828
	/*
	 * for each drhd
	 *   enable fault log
	 *   global invalidate context cache
	 *   global invalidate iotlb
	 *   enable translation
	 */
2829
	for_each_iommu(iommu, drhd) {
2830 2831 2832 2833 2834 2835
		if (drhd->ignored) {
			/*
			 * we always have to disable PMRs or DMA may fail on
			 * this device
			 */
			if (force_on)
2836
				iommu_disable_protect_mem_regions(iommu);
2837
			continue;
2838
		}
2839 2840 2841

		iommu_flush_write_buffer(iommu);

2842 2843
		ret = dmar_set_interrupt(iommu);
		if (ret)
2844
			goto free_iommu;
2845

2846 2847
		iommu_set_root_entry(iommu);

2848
		iommu->flush.flush_context(iommu, 0, 0, 0, DMA_CCMD_GLOBAL_INVL);
2849
		iommu->flush.flush_iotlb(iommu, 0, 0, 0, DMA_TLB_GLOBAL_FLUSH);
M
mark gross 已提交
2850

2851 2852
		ret = iommu_enable_translation(iommu);
		if (ret)
2853
			goto free_iommu;
2854 2855

		iommu_disable_protect_mem_regions(iommu);
2856 2857 2858
	}

	return 0;
2859 2860

free_iommu:
2861
	for_each_active_iommu(iommu, drhd)
2862
		free_dmar_iommu(iommu);
2863
	kfree(deferred_flush);
2864
free_g_iommus:
W
Weidong Han 已提交
2865
	kfree(g_iommus);
2866
error:
2867 2868 2869
	return ret;
}

2870
/* This takes a number of _MM_ pages, not VTD pages */
2871 2872 2873
static struct iova *intel_alloc_iova(struct device *dev,
				     struct dmar_domain *domain,
				     unsigned long nrpages, uint64_t dma_mask)
2874 2875 2876
{
	struct iova *iova = NULL;

2877 2878 2879 2880
	/* Restrict dma_mask to the width that the iommu can handle */
	dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);

	if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
2881 2882
		/*
		 * First try to allocate an io virtual address in
2883
		 * DMA_BIT_MASK(32) and if that fails then try allocating
J
Joe Perches 已提交
2884
		 * from higher range
2885
		 */
2886 2887 2888 2889 2890 2891 2892 2893
		iova = alloc_iova(&domain->iovad, nrpages,
				  IOVA_PFN(DMA_BIT_MASK(32)), 1);
		if (iova)
			return iova;
	}
	iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
	if (unlikely(!iova)) {
		printk(KERN_ERR "Allocating %ld-page iova for %s failed",
2894
		       nrpages, dev_name(dev));
2895 2896 2897 2898 2899 2900
		return NULL;
	}

	return iova;
}

2901
static struct dmar_domain *__get_valid_domain_for_dev(struct device *dev)
2902 2903 2904 2905
{
	struct dmar_domain *domain;
	int ret;

2906
	domain = get_domain_for_dev(dev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
2907
	if (!domain) {
2908 2909
		printk(KERN_ERR "Allocating domain for %s failed",
		       dev_name(dev));
A
Al Viro 已提交
2910
		return NULL;
2911 2912 2913
	}

	/* make sure context mapping is ok */
2914 2915
	if (unlikely(!domain_context_mapped(dev))) {
		ret = domain_context_mapping(domain, dev, CONTEXT_TT_MULTI_LEVEL);
2916
		if (ret) {
2917 2918
			printk(KERN_ERR "Domain context map for %s failed",
			       dev_name(dev));
A
Al Viro 已提交
2919
			return NULL;
2920
		}
2921 2922
	}

2923 2924 2925
	return domain;
}

2926
static inline struct dmar_domain *get_valid_domain_for_dev(struct device *dev)
2927 2928 2929 2930
{
	struct device_domain_info *info;

	/* No lock here, assumes no domain exit in normal case */
2931
	info = dev->archdata.iommu;
2932 2933 2934 2935 2936 2937
	if (likely(info))
		return info->domain;

	return __get_valid_domain_for_dev(dev);
}

2938
static int iommu_dummy(struct device *dev)
2939
{
2940
	return dev->archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO;
2941 2942
}

2943
/* Check if the dev needs to go through non-identity map and unmap process.*/
2944
static int iommu_no_mapping(struct device *dev)
2945 2946 2947
{
	int found;

2948
	if (iommu_dummy(dev))
2949 2950
		return 1;

2951
	if (!iommu_identity_mapping)
2952
		return 0;
2953

2954
	found = identity_mapping(dev);
2955
	if (found) {
2956
		if (iommu_should_identity_map(dev, 0))
2957 2958 2959 2960 2961 2962
			return 1;
		else {
			/*
			 * 32 bit DMA is removed from si_domain and fall back
			 * to non-identity mapping.
			 */
2963
			domain_remove_one_dev_info(si_domain, dev);
2964
			printk(KERN_INFO "32bit %s uses non-identity mapping\n",
2965
			       dev_name(dev));
2966 2967 2968 2969 2970 2971 2972
			return 0;
		}
	} else {
		/*
		 * In case of a detached 64 bit DMA device from vm, the device
		 * is put into si_domain for identity mapping.
		 */
2973
		if (iommu_should_identity_map(dev, 0)) {
2974
			int ret;
2975
			ret = domain_add_dev_info(si_domain, dev,
2976 2977 2978
						  hw_pass_through ?
						  CONTEXT_TT_PASS_THROUGH :
						  CONTEXT_TT_MULTI_LEVEL);
2979 2980
			if (!ret) {
				printk(KERN_INFO "64bit %s uses identity mapping\n",
2981
				       dev_name(dev));
2982 2983 2984 2985 2986
				return 1;
			}
		}
	}

2987
	return 0;
2988 2989
}

2990
static dma_addr_t __intel_map_single(struct device *dev, phys_addr_t paddr,
2991
				     size_t size, int dir, u64 dma_mask)
2992 2993
{
	struct dmar_domain *domain;
F
Fenghua Yu 已提交
2994
	phys_addr_t start_paddr;
2995 2996
	struct iova *iova;
	int prot = 0;
I
Ingo Molnar 已提交
2997
	int ret;
2998
	struct intel_iommu *iommu;
2999
	unsigned long paddr_pfn = paddr >> PAGE_SHIFT;
3000 3001

	BUG_ON(dir == DMA_NONE);
3002

3003
	if (iommu_no_mapping(dev))
I
Ingo Molnar 已提交
3004
		return paddr;
3005

3006
	domain = get_valid_domain_for_dev(dev);
3007 3008 3009
	if (!domain)
		return 0;

3010
	iommu = domain_get_iommu(domain);
3011
	size = aligned_nrpages(paddr, size);
3012

3013
	iova = intel_alloc_iova(dev, domain, dma_to_mm_pfn(size), dma_mask);
3014 3015 3016
	if (!iova)
		goto error;

3017 3018 3019 3020 3021
	/*
	 * Check if DMAR supports zero-length reads on write only
	 * mappings..
	 */
	if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
3022
			!cap_zlr(iommu->cap))
3023 3024 3025 3026
		prot |= DMA_PTE_READ;
	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
		prot |= DMA_PTE_WRITE;
	/*
I
Ingo Molnar 已提交
3027
	 * paddr - (paddr + size) might be partial page, we should map the whole
3028
	 * page.  Note: if two part of one page are separately mapped, we
I
Ingo Molnar 已提交
3029
	 * might have two guest_addr mapping to the same host paddr, but this
3030 3031
	 * is not a big problem
	 */
3032
	ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
3033
				 mm_to_dma_pfn(paddr_pfn), size, prot);
3034 3035 3036
	if (ret)
		goto error;

3037 3038
	/* it's a non-present to present mapping. Only flush if caching mode */
	if (cap_caching_mode(iommu->cap))
3039
		iommu_flush_iotlb_psi(iommu, domain->id, mm_to_dma_pfn(iova->pfn_lo), size, 0, 1);
3040
	else
3041
		iommu_flush_write_buffer(iommu);
3042

3043 3044 3045
	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
	start_paddr += paddr & ~PAGE_MASK;
	return start_paddr;
3046 3047

error:
3048 3049
	if (iova)
		__free_iova(&domain->iovad, iova);
3050
	printk(KERN_ERR"Device %s request: %zx@%llx dir %d --- failed\n",
3051
		dev_name(dev), size, (unsigned long long)paddr, dir);
3052 3053 3054
	return 0;
}

3055 3056 3057 3058
static dma_addr_t intel_map_page(struct device *dev, struct page *page,
				 unsigned long offset, size_t size,
				 enum dma_data_direction dir,
				 struct dma_attrs *attrs)
3059
{
3060
	return __intel_map_single(dev, page_to_phys(page) + offset, size,
3061
				  dir, *dev->dma_mask);
3062 3063
}

M
mark gross 已提交
3064 3065
static void flush_unmaps(void)
{
3066
	int i, j;
M
mark gross 已提交
3067 3068 3069 3070 3071

	timer_on = 0;

	/* just flush them all */
	for (i = 0; i < g_num_of_iommus; i++) {
3072 3073 3074
		struct intel_iommu *iommu = g_iommus[i];
		if (!iommu)
			continue;
3075

3076 3077 3078
		if (!deferred_flush[i].next)
			continue;

3079 3080 3081
		/* In caching mode, global flushes turn emulation expensive */
		if (!cap_caching_mode(iommu->cap))
			iommu->flush.flush_iotlb(iommu, 0, 0, 0,
Y
Yu Zhao 已提交
3082
					 DMA_TLB_GLOBAL_FLUSH);
3083
		for (j = 0; j < deferred_flush[i].next; j++) {
Y
Yu Zhao 已提交
3084 3085
			unsigned long mask;
			struct iova *iova = deferred_flush[i].iova[j];
3086 3087 3088 3089 3090
			struct dmar_domain *domain = deferred_flush[i].domain[j];

			/* On real hardware multiple invalidations are expensive */
			if (cap_caching_mode(iommu->cap))
				iommu_flush_iotlb_psi(iommu, domain->id,
3091 3092
					iova->pfn_lo, iova->pfn_hi - iova->pfn_lo + 1,
					!deferred_flush[i].freelist[j], 0);
3093 3094 3095 3096 3097
			else {
				mask = ilog2(mm_to_dma_pfn(iova->pfn_hi - iova->pfn_lo + 1));
				iommu_flush_dev_iotlb(deferred_flush[i].domain[j],
						(uint64_t)iova->pfn_lo << PAGE_SHIFT, mask);
			}
Y
Yu Zhao 已提交
3098
			__free_iova(&deferred_flush[i].domain[j]->iovad, iova);
3099 3100
			if (deferred_flush[i].freelist[j])
				dma_free_pagelist(deferred_flush[i].freelist[j]);
3101
		}
3102
		deferred_flush[i].next = 0;
M
mark gross 已提交
3103 3104 3105 3106 3107 3108 3109
	}

	list_size = 0;
}

static void flush_unmaps_timeout(unsigned long data)
{
3110 3111 3112
	unsigned long flags;

	spin_lock_irqsave(&async_umap_flush_lock, flags);
M
mark gross 已提交
3113
	flush_unmaps();
3114
	spin_unlock_irqrestore(&async_umap_flush_lock, flags);
M
mark gross 已提交
3115 3116
}

3117
static void add_unmap(struct dmar_domain *dom, struct iova *iova, struct page *freelist)
M
mark gross 已提交
3118 3119
{
	unsigned long flags;
3120
	int next, iommu_id;
3121
	struct intel_iommu *iommu;
M
mark gross 已提交
3122 3123

	spin_lock_irqsave(&async_umap_flush_lock, flags);
3124 3125 3126
	if (list_size == HIGH_WATER_MARK)
		flush_unmaps();

3127 3128
	iommu = domain_get_iommu(dom);
	iommu_id = iommu->seq_id;
3129

3130 3131 3132
	next = deferred_flush[iommu_id].next;
	deferred_flush[iommu_id].domain[next] = dom;
	deferred_flush[iommu_id].iova[next] = iova;
3133
	deferred_flush[iommu_id].freelist[next] = freelist;
3134
	deferred_flush[iommu_id].next++;
M
mark gross 已提交
3135 3136 3137 3138 3139 3140 3141 3142 3143

	if (!timer_on) {
		mod_timer(&unmap_timer, jiffies + msecs_to_jiffies(10));
		timer_on = 1;
	}
	list_size++;
	spin_unlock_irqrestore(&async_umap_flush_lock, flags);
}

3144 3145 3146
static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
			     size_t size, enum dma_data_direction dir,
			     struct dma_attrs *attrs)
3147
{
3148
	struct dmar_domain *domain;
3149
	unsigned long start_pfn, last_pfn;
3150
	struct iova *iova;
3151
	struct intel_iommu *iommu;
3152
	struct page *freelist;
3153

3154
	if (iommu_no_mapping(dev))
3155
		return;
3156

3157
	domain = find_domain(dev);
3158 3159
	BUG_ON(!domain);

3160 3161
	iommu = domain_get_iommu(domain);

3162
	iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
3163 3164
	if (WARN_ONCE(!iova, "Driver unmaps unmatched page at PFN %llx\n",
		      (unsigned long long)dev_addr))
3165 3166
		return;

3167 3168
	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
3169

3170
	pr_debug("Device %s unmapping: pfn %lx-%lx\n",
3171
		 dev_name(dev), start_pfn, last_pfn);
3172

3173
	freelist = domain_unmap(domain, start_pfn, last_pfn);
3174

M
mark gross 已提交
3175
	if (intel_iommu_strict) {
3176
		iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
3177
				      last_pfn - start_pfn + 1, !freelist, 0);
M
mark gross 已提交
3178 3179
		/* free iova */
		__free_iova(&domain->iovad, iova);
3180
		dma_free_pagelist(freelist);
M
mark gross 已提交
3181
	} else {
3182
		add_unmap(domain, iova, freelist);
M
mark gross 已提交
3183 3184 3185 3186 3187
		/*
		 * queue up the release of the unmap to save the 1/6th of the
		 * cpu used up by the iotlb flush operation...
		 */
	}
3188 3189
}

3190
static void *intel_alloc_coherent(struct device *dev, size_t size,
3191 3192
				  dma_addr_t *dma_handle, gfp_t flags,
				  struct dma_attrs *attrs)
3193 3194 3195 3196
{
	void *vaddr;
	int order;

F
Fenghua Yu 已提交
3197
	size = PAGE_ALIGN(size);
3198
	order = get_order(size);
3199

3200
	if (!iommu_no_mapping(dev))
3201
		flags &= ~(GFP_DMA | GFP_DMA32);
3202 3203
	else if (dev->coherent_dma_mask < dma_get_required_mask(dev)) {
		if (dev->coherent_dma_mask < DMA_BIT_MASK(32))
3204 3205 3206 3207
			flags |= GFP_DMA;
		else
			flags |= GFP_DMA32;
	}
3208 3209 3210 3211 3212 3213

	vaddr = (void *)__get_free_pages(flags, order);
	if (!vaddr)
		return NULL;
	memset(vaddr, 0, size);

3214
	*dma_handle = __intel_map_single(dev, virt_to_bus(vaddr), size,
3215
					 DMA_BIDIRECTIONAL,
3216
					 dev->coherent_dma_mask);
3217 3218 3219 3220 3221 3222
	if (*dma_handle)
		return vaddr;
	free_pages((unsigned long)vaddr, order);
	return NULL;
}

3223
static void intel_free_coherent(struct device *dev, size_t size, void *vaddr,
3224
				dma_addr_t dma_handle, struct dma_attrs *attrs)
3225 3226 3227
{
	int order;

F
Fenghua Yu 已提交
3228
	size = PAGE_ALIGN(size);
3229 3230
	order = get_order(size);

3231
	intel_unmap_page(dev, dma_handle, size, DMA_BIDIRECTIONAL, NULL);
3232 3233 3234
	free_pages((unsigned long)vaddr, order);
}

3235
static void intel_unmap_sg(struct device *dev, struct scatterlist *sglist,
3236 3237
			   int nelems, enum dma_data_direction dir,
			   struct dma_attrs *attrs)
3238 3239
{
	struct dmar_domain *domain;
3240
	unsigned long start_pfn, last_pfn;
3241
	struct iova *iova;
3242
	struct intel_iommu *iommu;
3243
	struct page *freelist;
3244

3245
	if (iommu_no_mapping(dev))
3246 3247
		return;

3248
	domain = find_domain(dev);
3249 3250 3251
	BUG_ON(!domain);

	iommu = domain_get_iommu(domain);
3252

F
FUJITA Tomonori 已提交
3253
	iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
3254 3255
	if (WARN_ONCE(!iova, "Driver unmaps unmatched sglist at PFN %llx\n",
		      (unsigned long long)sglist[0].dma_address))
3256 3257
		return;

3258 3259
	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
3260

3261
	freelist = domain_unmap(domain, start_pfn, last_pfn);
3262

3263 3264
	if (intel_iommu_strict) {
		iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
3265
				      last_pfn - start_pfn + 1, !freelist, 0);
3266 3267
		/* free iova */
		__free_iova(&domain->iovad, iova);
3268
		dma_free_pagelist(freelist);
3269
	} else {
3270
		add_unmap(domain, iova, freelist);
3271 3272 3273 3274 3275
		/*
		 * queue up the release of the unmap to save the 1/6th of the
		 * cpu used up by the iotlb flush operation...
		 */
	}
3276 3277 3278
}

static int intel_nontranslate_map_sg(struct device *hddev,
F
FUJITA Tomonori 已提交
3279
	struct scatterlist *sglist, int nelems, int dir)
3280 3281
{
	int i;
F
FUJITA Tomonori 已提交
3282
	struct scatterlist *sg;
3283

F
FUJITA Tomonori 已提交
3284
	for_each_sg(sglist, sg, nelems, i) {
F
FUJITA Tomonori 已提交
3285
		BUG_ON(!sg_page(sg));
3286
		sg->dma_address = page_to_phys(sg_page(sg)) + sg->offset;
F
FUJITA Tomonori 已提交
3287
		sg->dma_length = sg->length;
3288 3289 3290 3291
	}
	return nelems;
}

3292
static int intel_map_sg(struct device *dev, struct scatterlist *sglist, int nelems,
3293
			enum dma_data_direction dir, struct dma_attrs *attrs)
3294 3295 3296
{
	int i;
	struct dmar_domain *domain;
3297 3298 3299 3300
	size_t size = 0;
	int prot = 0;
	struct iova *iova = NULL;
	int ret;
F
FUJITA Tomonori 已提交
3301
	struct scatterlist *sg;
3302
	unsigned long start_vpfn;
3303
	struct intel_iommu *iommu;
3304 3305

	BUG_ON(dir == DMA_NONE);
3306 3307
	if (iommu_no_mapping(dev))
		return intel_nontranslate_map_sg(dev, sglist, nelems, dir);
3308

3309
	domain = get_valid_domain_for_dev(dev);
3310 3311 3312
	if (!domain)
		return 0;

3313 3314
	iommu = domain_get_iommu(domain);

3315
	for_each_sg(sglist, sg, nelems, i)
3316
		size += aligned_nrpages(sg->offset, sg->length);
3317

3318 3319
	iova = intel_alloc_iova(dev, domain, dma_to_mm_pfn(size),
				*dev->dma_mask);
3320
	if (!iova) {
F
FUJITA Tomonori 已提交
3321
		sglist->dma_length = 0;
3322 3323 3324 3325 3326 3327 3328 3329
		return 0;
	}

	/*
	 * Check if DMAR supports zero-length reads on write only
	 * mappings..
	 */
	if (dir == DMA_TO_DEVICE || dir == DMA_BIDIRECTIONAL || \
3330
			!cap_zlr(iommu->cap))
3331 3332 3333 3334
		prot |= DMA_PTE_READ;
	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
		prot |= DMA_PTE_WRITE;

3335
	start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
3336

3337
	ret = domain_sg_mapping(domain, start_vpfn, sglist, size, prot);
3338 3339 3340 3341 3342 3343 3344 3345 3346 3347
	if (unlikely(ret)) {
		/*  clear the page */
		dma_pte_clear_range(domain, start_vpfn,
				    start_vpfn + size - 1);
		/* free page tables */
		dma_pte_free_pagetable(domain, start_vpfn,
				       start_vpfn + size - 1);
		/* free iova */
		__free_iova(&domain->iovad, iova);
		return 0;
3348 3349
	}

3350 3351
	/* it's a non-present to present mapping. Only flush if caching mode */
	if (cap_caching_mode(iommu->cap))
3352
		iommu_flush_iotlb_psi(iommu, domain->id, start_vpfn, size, 0, 1);
3353
	else
3354
		iommu_flush_write_buffer(iommu);
3355

3356 3357 3358
	return nelems;
}

3359 3360 3361 3362 3363
static int intel_mapping_error(struct device *dev, dma_addr_t dma_addr)
{
	return !dma_addr;
}

3364
struct dma_map_ops intel_dma_ops = {
3365 3366
	.alloc = intel_alloc_coherent,
	.free = intel_free_coherent,
3367 3368
	.map_sg = intel_map_sg,
	.unmap_sg = intel_unmap_sg,
3369 3370
	.map_page = intel_map_page,
	.unmap_page = intel_unmap_page,
3371
	.mapping_error = intel_mapping_error,
3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455
};

static inline int iommu_domain_cache_init(void)
{
	int ret = 0;

	iommu_domain_cache = kmem_cache_create("iommu_domain",
					 sizeof(struct dmar_domain),
					 0,
					 SLAB_HWCACHE_ALIGN,

					 NULL);
	if (!iommu_domain_cache) {
		printk(KERN_ERR "Couldn't create iommu_domain cache\n");
		ret = -ENOMEM;
	}

	return ret;
}

static inline int iommu_devinfo_cache_init(void)
{
	int ret = 0;

	iommu_devinfo_cache = kmem_cache_create("iommu_devinfo",
					 sizeof(struct device_domain_info),
					 0,
					 SLAB_HWCACHE_ALIGN,
					 NULL);
	if (!iommu_devinfo_cache) {
		printk(KERN_ERR "Couldn't create devinfo cache\n");
		ret = -ENOMEM;
	}

	return ret;
}

static inline int iommu_iova_cache_init(void)
{
	int ret = 0;

	iommu_iova_cache = kmem_cache_create("iommu_iova",
					 sizeof(struct iova),
					 0,
					 SLAB_HWCACHE_ALIGN,
					 NULL);
	if (!iommu_iova_cache) {
		printk(KERN_ERR "Couldn't create iova cache\n");
		ret = -ENOMEM;
	}

	return ret;
}

static int __init iommu_init_mempool(void)
{
	int ret;
	ret = iommu_iova_cache_init();
	if (ret)
		return ret;

	ret = iommu_domain_cache_init();
	if (ret)
		goto domain_error;

	ret = iommu_devinfo_cache_init();
	if (!ret)
		return ret;

	kmem_cache_destroy(iommu_domain_cache);
domain_error:
	kmem_cache_destroy(iommu_iova_cache);

	return -ENOMEM;
}

static void __init iommu_exit_mempool(void)
{
	kmem_cache_destroy(iommu_devinfo_cache);
	kmem_cache_destroy(iommu_domain_cache);
	kmem_cache_destroy(iommu_iova_cache);

}

3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483
static void quirk_ioat_snb_local_iommu(struct pci_dev *pdev)
{
	struct dmar_drhd_unit *drhd;
	u32 vtbar;
	int rc;

	/* We know that this device on this chipset has its own IOMMU.
	 * If we find it under a different IOMMU, then the BIOS is lying
	 * to us. Hope that the IOMMU for this device is actually
	 * disabled, and it needs no translation...
	 */
	rc = pci_bus_read_config_dword(pdev->bus, PCI_DEVFN(0, 0), 0xb0, &vtbar);
	if (rc) {
		/* "can't" happen */
		dev_info(&pdev->dev, "failed to run vt-d quirk\n");
		return;
	}
	vtbar &= 0xffff0000;

	/* we know that the this iommu should be at offset 0xa000 from vtbar */
	drhd = dmar_find_matched_drhd_unit(pdev);
	if (WARN_TAINT_ONCE(!drhd || drhd->reg_base_addr - vtbar != 0xa000,
			    TAINT_FIRMWARE_WORKAROUND,
			    "BIOS assigned incorrect VT-d unit for Intel(R) QuickData Technology device\n"))
		pdev->dev.archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO;
}
DECLARE_PCI_FIXUP_ENABLE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL_IOAT_SNB, quirk_ioat_snb_local_iommu);

3484 3485 3486
static void __init init_no_remapping_devices(void)
{
	struct dmar_drhd_unit *drhd;
3487
	struct device *dev;
3488
	int i;
3489 3490 3491

	for_each_drhd_unit(drhd) {
		if (!drhd->include_all) {
3492 3493 3494
			for_each_active_dev_scope(drhd->devices,
						  drhd->devices_cnt, i, dev)
				break;
3495
			/* ignore DMAR unit if no devices exist */
3496 3497 3498 3499 3500
			if (i == drhd->devices_cnt)
				drhd->ignored = 1;
		}
	}

3501 3502
	for_each_active_drhd_unit(drhd) {
		if (drhd->include_all)
3503 3504
			continue;

3505 3506
		for_each_active_dev_scope(drhd->devices,
					  drhd->devices_cnt, i, dev)
3507
			if (!dev_is_pci(dev) || !IS_GFX_DEVICE(to_pci_dev(dev)))
3508 3509 3510 3511
				break;
		if (i < drhd->devices_cnt)
			continue;

3512 3513 3514 3515 3516 3517
		/* This IOMMU has *only* gfx devices. Either bypass it or
		   set the gfx_mapped flag, as appropriate */
		if (dmar_map_gfx) {
			intel_iommu_gfx_mapped = 1;
		} else {
			drhd->ignored = 1;
3518 3519
			for_each_active_dev_scope(drhd->devices,
						  drhd->devices_cnt, i, dev)
3520
				dev->archdata.iommu = DUMMY_DEVICE_DOMAIN_INFO;
3521 3522 3523 3524
		}
	}
}

3525 3526 3527 3528 3529 3530 3531 3532 3533 3534
#ifdef CONFIG_SUSPEND
static int init_iommu_hw(void)
{
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu = NULL;

	for_each_active_iommu(iommu, drhd)
		if (iommu->qi)
			dmar_reenable_qi(iommu);

3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545
	for_each_iommu(iommu, drhd) {
		if (drhd->ignored) {
			/*
			 * we always have to disable PMRs or DMA may fail on
			 * this device
			 */
			if (force_on)
				iommu_disable_protect_mem_regions(iommu);
			continue;
		}
	
3546 3547 3548 3549 3550
		iommu_flush_write_buffer(iommu);

		iommu_set_root_entry(iommu);

		iommu->flush.flush_context(iommu, 0, 0, 0,
3551
					   DMA_CCMD_GLOBAL_INVL);
3552
		iommu->flush.flush_iotlb(iommu, 0, 0, 0,
3553
					 DMA_TLB_GLOBAL_FLUSH);
3554 3555
		if (iommu_enable_translation(iommu))
			return 1;
3556
		iommu_disable_protect_mem_regions(iommu);
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
	}

	return 0;
}

static void iommu_flush_all(void)
{
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu;

	for_each_active_iommu(iommu, drhd) {
		iommu->flush.flush_context(iommu, 0, 0, 0,
3569
					   DMA_CCMD_GLOBAL_INVL);
3570
		iommu->flush.flush_iotlb(iommu, 0, 0, 0,
3571
					 DMA_TLB_GLOBAL_FLUSH);
3572 3573 3574
	}
}

3575
static int iommu_suspend(void)
3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592
{
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu = NULL;
	unsigned long flag;

	for_each_active_iommu(iommu, drhd) {
		iommu->iommu_state = kzalloc(sizeof(u32) * MAX_SR_DMAR_REGS,
						 GFP_ATOMIC);
		if (!iommu->iommu_state)
			goto nomem;
	}

	iommu_flush_all();

	for_each_active_iommu(iommu, drhd) {
		iommu_disable_translation(iommu);

3593
		raw_spin_lock_irqsave(&iommu->register_lock, flag);
3594 3595 3596 3597 3598 3599 3600 3601 3602 3603

		iommu->iommu_state[SR_DMAR_FECTL_REG] =
			readl(iommu->reg + DMAR_FECTL_REG);
		iommu->iommu_state[SR_DMAR_FEDATA_REG] =
			readl(iommu->reg + DMAR_FEDATA_REG);
		iommu->iommu_state[SR_DMAR_FEADDR_REG] =
			readl(iommu->reg + DMAR_FEADDR_REG);
		iommu->iommu_state[SR_DMAR_FEUADDR_REG] =
			readl(iommu->reg + DMAR_FEUADDR_REG);

3604
		raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
3605 3606 3607 3608 3609 3610 3611 3612 3613 3614
	}
	return 0;

nomem:
	for_each_active_iommu(iommu, drhd)
		kfree(iommu->iommu_state);

	return -ENOMEM;
}

3615
static void iommu_resume(void)
3616 3617 3618 3619 3620 3621
{
	struct dmar_drhd_unit *drhd;
	struct intel_iommu *iommu = NULL;
	unsigned long flag;

	if (init_iommu_hw()) {
3622 3623 3624 3625
		if (force_on)
			panic("tboot: IOMMU setup failed, DMAR can not resume!\n");
		else
			WARN(1, "IOMMU setup failed, DMAR can not resume!\n");
3626
		return;
3627 3628 3629 3630
	}

	for_each_active_iommu(iommu, drhd) {

3631
		raw_spin_lock_irqsave(&iommu->register_lock, flag);
3632 3633 3634 3635 3636 3637 3638 3639 3640 3641

		writel(iommu->iommu_state[SR_DMAR_FECTL_REG],
			iommu->reg + DMAR_FECTL_REG);
		writel(iommu->iommu_state[SR_DMAR_FEDATA_REG],
			iommu->reg + DMAR_FEDATA_REG);
		writel(iommu->iommu_state[SR_DMAR_FEADDR_REG],
			iommu->reg + DMAR_FEADDR_REG);
		writel(iommu->iommu_state[SR_DMAR_FEUADDR_REG],
			iommu->reg + DMAR_FEUADDR_REG);

3642
		raw_spin_unlock_irqrestore(&iommu->register_lock, flag);
3643 3644 3645 3646 3647 3648
	}

	for_each_active_iommu(iommu, drhd)
		kfree(iommu->iommu_state);
}

3649
static struct syscore_ops iommu_syscore_ops = {
3650 3651 3652 3653
	.resume		= iommu_resume,
	.suspend	= iommu_suspend,
};

3654
static void __init init_iommu_pm_ops(void)
3655
{
3656
	register_syscore_ops(&iommu_syscore_ops);
3657 3658 3659
}

#else
3660
static inline void init_iommu_pm_ops(void) {}
3661 3662
#endif	/* CONFIG_PM */

3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676

int __init dmar_parse_one_rmrr(struct acpi_dmar_header *header)
{
	struct acpi_dmar_reserved_memory *rmrr;
	struct dmar_rmrr_unit *rmrru;

	rmrru = kzalloc(sizeof(*rmrru), GFP_KERNEL);
	if (!rmrru)
		return -ENOMEM;

	rmrru->hdr = header;
	rmrr = (struct acpi_dmar_reserved_memory *)header;
	rmrru->base_address = rmrr->base_address;
	rmrru->end_address = rmrr->end_address;
3677 3678 3679 3680 3681 3682 3683
	rmrru->devices = dmar_alloc_dev_scope((void *)(rmrr + 1),
				((void *)rmrr) + rmrr->header.length,
				&rmrru->devices_cnt);
	if (rmrru->devices_cnt && rmrru->devices == NULL) {
		kfree(rmrru);
		return -ENOMEM;
	}
3684

3685
	list_add(&rmrru->list, &dmar_rmrr_units);
3686

3687
	return 0;
3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701
}

int __init dmar_parse_one_atsr(struct acpi_dmar_header *hdr)
{
	struct acpi_dmar_atsr *atsr;
	struct dmar_atsr_unit *atsru;

	atsr = container_of(hdr, struct acpi_dmar_atsr, header);
	atsru = kzalloc(sizeof(*atsru), GFP_KERNEL);
	if (!atsru)
		return -ENOMEM;

	atsru->hdr = hdr;
	atsru->include_all = atsr->flags & 0x1;
3702 3703 3704 3705 3706 3707 3708 3709 3710
	if (!atsru->include_all) {
		atsru->devices = dmar_alloc_dev_scope((void *)(atsr + 1),
				(void *)atsr + atsr->header.length,
				&atsru->devices_cnt);
		if (atsru->devices_cnt && atsru->devices == NULL) {
			kfree(atsru);
			return -ENOMEM;
		}
	}
3711

3712
	list_add_rcu(&atsru->list, &dmar_atsr_units);
3713 3714 3715 3716

	return 0;
}

3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
static void intel_iommu_free_atsr(struct dmar_atsr_unit *atsru)
{
	dmar_free_dev_scope(&atsru->devices, &atsru->devices_cnt);
	kfree(atsru);
}

static void intel_iommu_free_dmars(void)
{
	struct dmar_rmrr_unit *rmrru, *rmrr_n;
	struct dmar_atsr_unit *atsru, *atsr_n;

	list_for_each_entry_safe(rmrru, rmrr_n, &dmar_rmrr_units, list) {
		list_del(&rmrru->list);
		dmar_free_dev_scope(&rmrru->devices, &rmrru->devices_cnt);
		kfree(rmrru);
3732 3733
	}

3734 3735 3736 3737
	list_for_each_entry_safe(atsru, atsr_n, &dmar_atsr_units, list) {
		list_del(&atsru->list);
		intel_iommu_free_atsr(atsru);
	}
3738 3739 3740 3741
}

int dmar_find_matched_atsr_unit(struct pci_dev *dev)
{
3742
	int i, ret = 1;
3743
	struct pci_bus *bus;
3744 3745
	struct pci_dev *bridge = NULL;
	struct device *tmp;
3746 3747 3748 3749 3750
	struct acpi_dmar_atsr *atsr;
	struct dmar_atsr_unit *atsru;

	dev = pci_physfn(dev);
	for (bus = dev->bus; bus; bus = bus->parent) {
3751
		bridge = bus->self;
3752
		if (!bridge || !pci_is_pcie(bridge) ||
3753
		    pci_pcie_type(bridge) == PCI_EXP_TYPE_PCI_BRIDGE)
3754
			return 0;
3755
		if (pci_pcie_type(bridge) == PCI_EXP_TYPE_ROOT_PORT)
3756 3757
			break;
	}
3758 3759
	if (!bridge)
		return 0;
3760

3761
	rcu_read_lock();
3762 3763 3764 3765 3766
	list_for_each_entry_rcu(atsru, &dmar_atsr_units, list) {
		atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
		if (atsr->segment != pci_domain_nr(dev->bus))
			continue;

3767
		for_each_dev_scope(atsru->devices, atsru->devices_cnt, i, tmp)
3768
			if (tmp == &bridge->dev)
3769
				goto out;
3770 3771

		if (atsru->include_all)
3772
			goto out;
3773
	}
3774 3775
	ret = 0;
out:
3776
	rcu_read_unlock();
3777

3778
	return ret;
3779 3780
}

3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834
int dmar_iommu_notify_scope_dev(struct dmar_pci_notify_info *info)
{
	int ret = 0;
	struct dmar_rmrr_unit *rmrru;
	struct dmar_atsr_unit *atsru;
	struct acpi_dmar_atsr *atsr;
	struct acpi_dmar_reserved_memory *rmrr;

	if (!intel_iommu_enabled && system_state != SYSTEM_BOOTING)
		return 0;

	list_for_each_entry(rmrru, &dmar_rmrr_units, list) {
		rmrr = container_of(rmrru->hdr,
				    struct acpi_dmar_reserved_memory, header);
		if (info->event == BUS_NOTIFY_ADD_DEVICE) {
			ret = dmar_insert_dev_scope(info, (void *)(rmrr + 1),
				((void *)rmrr) + rmrr->header.length,
				rmrr->segment, rmrru->devices,
				rmrru->devices_cnt);
			if (ret > 0)
				break;
			else if(ret < 0)
				return ret;
		} else if (info->event == BUS_NOTIFY_DEL_DEVICE) {
			if (dmar_remove_dev_scope(info, rmrr->segment,
				rmrru->devices, rmrru->devices_cnt))
				break;
		}
	}

	list_for_each_entry(atsru, &dmar_atsr_units, list) {
		if (atsru->include_all)
			continue;

		atsr = container_of(atsru->hdr, struct acpi_dmar_atsr, header);
		if (info->event == BUS_NOTIFY_ADD_DEVICE) {
			ret = dmar_insert_dev_scope(info, (void *)(atsr + 1),
					(void *)atsr + atsr->header.length,
					atsr->segment, atsru->devices,
					atsru->devices_cnt);
			if (ret > 0)
				break;
			else if(ret < 0)
				return ret;
		} else if (info->event == BUS_NOTIFY_DEL_DEVICE) {
			if (dmar_remove_dev_scope(info, atsr->segment,
					atsru->devices, atsru->devices_cnt))
				break;
		}
	}

	return 0;
}

F
Fenghua Yu 已提交
3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846
/*
 * Here we only respond to action of unbound device from driver.
 *
 * Added device is not attached to its DMAR domain here yet. That will happen
 * when mapping the device to iova.
 */
static int device_notifier(struct notifier_block *nb,
				  unsigned long action, void *data)
{
	struct device *dev = data;
	struct dmar_domain *domain;

3847
	if (iommu_dummy(dev))
3848 3849
		return 0;

3850 3851 3852 3853
	if (action != BUS_NOTIFY_UNBOUND_DRIVER &&
	    action != BUS_NOTIFY_DEL_DEVICE)
		return 0;

3854
	domain = find_domain(dev);
F
Fenghua Yu 已提交
3855 3856 3857
	if (!domain)
		return 0;

3858
	down_read(&dmar_global_lock);
3859
	domain_remove_one_dev_info(domain, dev);
3860 3861 3862 3863
	if (!(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) &&
	    !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY) &&
	    list_empty(&domain->devices))
		domain_exit(domain);
3864
	up_read(&dmar_global_lock);
3865

F
Fenghua Yu 已提交
3866 3867 3868 3869 3870 3871 3872
	return 0;
}

static struct notifier_block device_nb = {
	.notifier_call = device_notifier,
};

3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898
static int intel_iommu_memory_notifier(struct notifier_block *nb,
				       unsigned long val, void *v)
{
	struct memory_notify *mhp = v;
	unsigned long long start, end;
	unsigned long start_vpfn, last_vpfn;

	switch (val) {
	case MEM_GOING_ONLINE:
		start = mhp->start_pfn << PAGE_SHIFT;
		end = ((mhp->start_pfn + mhp->nr_pages) << PAGE_SHIFT) - 1;
		if (iommu_domain_identity_map(si_domain, start, end)) {
			pr_warn("dmar: failed to build identity map for [%llx-%llx]\n",
				start, end);
			return NOTIFY_BAD;
		}
		break;

	case MEM_OFFLINE:
	case MEM_CANCEL_ONLINE:
		start_vpfn = mm_to_dma_pfn(mhp->start_pfn);
		last_vpfn = mm_to_dma_pfn(mhp->start_pfn + mhp->nr_pages - 1);
		while (start_vpfn <= last_vpfn) {
			struct iova *iova;
			struct dmar_drhd_unit *drhd;
			struct intel_iommu *iommu;
3899
			struct page *freelist;
3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915

			iova = find_iova(&si_domain->iovad, start_vpfn);
			if (iova == NULL) {
				pr_debug("dmar: failed get IOVA for PFN %lx\n",
					 start_vpfn);
				break;
			}

			iova = split_and_remove_iova(&si_domain->iovad, iova,
						     start_vpfn, last_vpfn);
			if (iova == NULL) {
				pr_warn("dmar: failed to split IOVA PFN [%lx-%lx]\n",
					start_vpfn, last_vpfn);
				return NOTIFY_BAD;
			}

3916 3917 3918
			freelist = domain_unmap(si_domain, iova->pfn_lo,
					       iova->pfn_hi);

3919 3920 3921 3922
			rcu_read_lock();
			for_each_active_iommu(iommu, drhd)
				iommu_flush_iotlb_psi(iommu, si_domain->id,
					iova->pfn_lo,
3923 3924
					iova->pfn_hi - iova->pfn_lo + 1,
					!freelist, 0);
3925
			rcu_read_unlock();
3926
			dma_free_pagelist(freelist);
3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941

			start_vpfn = iova->pfn_hi + 1;
			free_iova_mem(iova);
		}
		break;
	}

	return NOTIFY_OK;
}

static struct notifier_block intel_iommu_memory_nb = {
	.notifier_call = intel_iommu_memory_notifier,
	.priority = 0
};

3942 3943
int __init intel_iommu_init(void)
{
3944
	int ret = -ENODEV;
3945
	struct dmar_drhd_unit *drhd;
3946
	struct intel_iommu *iommu;
3947

3948 3949 3950
	/* VT-d is required for a TXT/tboot launch, so enforce that */
	force_on = tboot_force_iommu();

3951 3952 3953 3954 3955 3956 3957
	if (iommu_init_mempool()) {
		if (force_on)
			panic("tboot: Failed to initialize iommu memory\n");
		return -ENOMEM;
	}

	down_write(&dmar_global_lock);
3958 3959 3960
	if (dmar_table_init()) {
		if (force_on)
			panic("tboot: Failed to initialize DMAR table\n");
3961
		goto out_free_dmar;
3962
	}
3963

3964 3965 3966
	/*
	 * Disable translation if already enabled prior to OS handover.
	 */
3967
	for_each_active_iommu(iommu, drhd)
3968 3969 3970
		if (iommu->gcmd & DMA_GCMD_TE)
			iommu_disable_translation(iommu);

3971
	if (dmar_dev_scope_init() < 0) {
3972 3973
		if (force_on)
			panic("tboot: Failed to initialize DMAR device scope\n");
3974
		goto out_free_dmar;
3975
	}
3976

3977
	if (no_iommu || dmar_disabled)
3978
		goto out_free_dmar;
3979

3980 3981 3982 3983 3984 3985
	if (list_empty(&dmar_rmrr_units))
		printk(KERN_INFO "DMAR: No RMRR found\n");

	if (list_empty(&dmar_atsr_units))
		printk(KERN_INFO "DMAR: No ATSR found\n");

3986 3987 3988
	if (dmar_init_reserved_ranges()) {
		if (force_on)
			panic("tboot: Failed to reserve iommu ranges\n");
3989
		goto out_free_reserved_range;
3990
	}
3991 3992 3993

	init_no_remapping_devices();

3994
	ret = init_dmars();
3995
	if (ret) {
3996 3997
		if (force_on)
			panic("tboot: Failed to initialize DMARs\n");
3998
		printk(KERN_ERR "IOMMU: dmar init failed\n");
3999
		goto out_free_reserved_range;
4000
	}
4001
	up_write(&dmar_global_lock);
4002 4003 4004
	printk(KERN_INFO
	"PCI-DMA: Intel(R) Virtualization Technology for Directed I/O\n");

M
mark gross 已提交
4005
	init_timer(&unmap_timer);
4006 4007 4008
#ifdef CONFIG_SWIOTLB
	swiotlb = 0;
#endif
4009
	dma_ops = &intel_dma_ops;
F
Fenghua Yu 已提交
4010

4011
	init_iommu_pm_ops();
4012

4013
	bus_set_iommu(&pci_bus_type, &intel_iommu_ops);
F
Fenghua Yu 已提交
4014
	bus_register_notifier(&pci_bus_type, &device_nb);
4015 4016
	if (si_domain && !hw_pass_through)
		register_memory_notifier(&intel_iommu_memory_nb);
F
Fenghua Yu 已提交
4017

4018 4019
	intel_iommu_enabled = 1;

4020
	return 0;
4021 4022 4023 4024 4025

out_free_reserved_range:
	put_iova_domain(&reserved_iova_list);
out_free_dmar:
	intel_iommu_free_dmars();
4026 4027
	up_write(&dmar_global_lock);
	iommu_exit_mempool();
4028
	return ret;
4029
}
4030

4031
static void iommu_detach_dependent_devices(struct intel_iommu *iommu,
4032
					   struct device *dev)
4033
{
4034
	struct pci_dev *tmp, *parent, *pdev;
4035

4036
	if (!iommu || !dev || !dev_is_pci(dev))
4037 4038
		return;

4039 4040
	pdev = to_pci_dev(dev);

4041 4042 4043 4044 4045 4046 4047
	/* dependent device detach */
	tmp = pci_find_upstream_pcie_bridge(pdev);
	/* Secondary interface's bus number and devfn 0 */
	if (tmp) {
		parent = pdev->bus->self;
		while (parent != tmp) {
			iommu_detach_dev(iommu, parent->bus->number,
4048
					 parent->devfn);
4049 4050
			parent = parent->bus->self;
		}
4051
		if (pci_is_pcie(tmp)) /* this is a PCIe-to-PCI bridge */
4052 4053 4054
			iommu_detach_dev(iommu,
				tmp->subordinate->number, 0);
		else /* this is a legacy PCI bridge */
4055 4056
			iommu_detach_dev(iommu, tmp->bus->number,
					 tmp->devfn);
4057 4058 4059
	}
}

4060
static void domain_remove_one_dev_info(struct dmar_domain *domain,
4061
				       struct device *dev)
4062
{
4063
	struct device_domain_info *info, *tmp;
4064 4065 4066
	struct intel_iommu *iommu;
	unsigned long flags;
	int found = 0;
4067
	u8 bus, devfn;
4068

4069
	iommu = device_to_iommu(dev, &bus, &devfn);
4070 4071 4072 4073
	if (!iommu)
		return;

	spin_lock_irqsave(&device_domain_lock, flags);
4074
	list_for_each_entry_safe(info, tmp, &domain->devices, link) {
4075 4076
		if (info->iommu == iommu && info->bus == bus &&
		    info->devfn == devfn) {
4077
			unlink_domain_info(info);
4078 4079
			spin_unlock_irqrestore(&device_domain_lock, flags);

Y
Yu Zhao 已提交
4080
			iommu_disable_dev_iotlb(info);
4081
			iommu_detach_dev(iommu, info->bus, info->devfn);
4082
			iommu_detach_dependent_devices(iommu, dev);
4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096
			free_devinfo_mem(info);

			spin_lock_irqsave(&device_domain_lock, flags);

			if (found)
				break;
			else
				continue;
		}

		/* if there is no other devices under the same iommu
		 * owned by this domain, clear this iommu in iommu_bmp
		 * update iommu count and coherency
		 */
4097
		if (info->iommu == iommu)
4098 4099 4100
			found = 1;
	}

4101 4102
	spin_unlock_irqrestore(&device_domain_lock, flags);

4103 4104 4105
	if (found == 0) {
		unsigned long tmp_flags;
		spin_lock_irqsave(&domain->iommu_lock, tmp_flags);
4106
		clear_bit(iommu->seq_id, domain->iommu_bmp);
4107
		domain->iommu_count--;
4108
		domain_update_iommu_cap(domain);
4109
		spin_unlock_irqrestore(&domain->iommu_lock, tmp_flags);
4110

4111 4112 4113 4114 4115 4116 4117
		if (!(domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE) &&
		    !(domain->flags & DOMAIN_FLAG_STATIC_IDENTITY)) {
			spin_lock_irqsave(&iommu->lock, tmp_flags);
			clear_bit(domain->id, iommu->domain_ids);
			iommu->domains[domain->id] = NULL;
			spin_unlock_irqrestore(&iommu->lock, tmp_flags);
		}
4118 4119 4120
	}
}

4121
static int md_domain_init(struct dmar_domain *domain, int guest_width)
4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133
{
	int adjust_width;

	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
	domain_reserve_special_ranges(domain);

	/* calculate AGAW */
	domain->gaw = guest_width;
	adjust_width = guestwidth_to_adjustwidth(guest_width);
	domain->agaw = width_to_agaw(adjust_width);

	domain->iommu_coherency = 0;
4134
	domain->iommu_snooping = 0;
4135
	domain->iommu_superpage = 0;
4136
	domain->max_addr = 0;
4137
	domain->nid = -1;
4138 4139

	/* always allocate the top pgd */
4140
	domain->pgd = (struct dma_pte *)alloc_pgtable_page(domain->nid);
4141 4142 4143 4144 4145 4146
	if (!domain->pgd)
		return -ENOMEM;
	domain_flush_cache(domain, domain->pgd, PAGE_SIZE);
	return 0;
}

4147
static int intel_iommu_domain_init(struct iommu_domain *domain)
K
Kay, Allen M 已提交
4148
{
4149
	struct dmar_domain *dmar_domain;
K
Kay, Allen M 已提交
4150

4151
	dmar_domain = alloc_domain(true);
4152
	if (!dmar_domain) {
K
Kay, Allen M 已提交
4153
		printk(KERN_ERR
4154 4155
			"intel_iommu_domain_init: dmar_domain == NULL\n");
		return -ENOMEM;
K
Kay, Allen M 已提交
4156
	}
4157
	if (md_domain_init(dmar_domain, DEFAULT_DOMAIN_ADDRESS_WIDTH)) {
K
Kay, Allen M 已提交
4158
		printk(KERN_ERR
4159
			"intel_iommu_domain_init() failed\n");
4160
		domain_exit(dmar_domain);
4161
		return -ENOMEM;
K
Kay, Allen M 已提交
4162
	}
4163
	domain_update_iommu_cap(dmar_domain);
4164
	domain->priv = dmar_domain;
4165

4166 4167 4168 4169
	domain->geometry.aperture_start = 0;
	domain->geometry.aperture_end   = __DOMAIN_MAX_ADDR(dmar_domain->gaw);
	domain->geometry.force_aperture = true;

4170
	return 0;
K
Kay, Allen M 已提交
4171 4172
}

4173
static void intel_iommu_domain_destroy(struct iommu_domain *domain)
K
Kay, Allen M 已提交
4174
{
4175 4176 4177
	struct dmar_domain *dmar_domain = domain->priv;

	domain->priv = NULL;
4178
	domain_exit(dmar_domain);
K
Kay, Allen M 已提交
4179 4180
}

4181 4182
static int intel_iommu_attach_device(struct iommu_domain *domain,
				     struct device *dev)
K
Kay, Allen M 已提交
4183
{
4184
	struct dmar_domain *dmar_domain = domain->priv;
4185 4186
	struct intel_iommu *iommu;
	int addr_width;
4187
	u8 bus, devfn;
4188

4189 4190
	/* normally dev is not mapped */
	if (unlikely(domain_context_mapped(dev))) {
4191 4192
		struct dmar_domain *old_domain;

4193
		old_domain = find_domain(dev);
4194
		if (old_domain) {
4195 4196
			if (dmar_domain->flags & DOMAIN_FLAG_VIRTUAL_MACHINE ||
			    dmar_domain->flags & DOMAIN_FLAG_STATIC_IDENTITY)
4197
				domain_remove_one_dev_info(old_domain, dev);
4198 4199 4200 4201 4202
			else
				domain_remove_dev_info(old_domain);
		}
	}

4203
	iommu = device_to_iommu(dev, &bus, &devfn);
4204 4205 4206 4207 4208
	if (!iommu)
		return -ENODEV;

	/* check if this iommu agaw is sufficient for max mapped address */
	addr_width = agaw_to_width(iommu->agaw);
4209 4210 4211 4212 4213
	if (addr_width > cap_mgaw(iommu->cap))
		addr_width = cap_mgaw(iommu->cap);

	if (dmar_domain->max_addr > (1LL << addr_width)) {
		printk(KERN_ERR "%s: iommu width (%d) is not "
4214
		       "sufficient for the mapped address (%llx)\n",
4215
		       __func__, addr_width, dmar_domain->max_addr);
4216 4217
		return -EFAULT;
	}
4218 4219 4220 4221 4222 4223 4224 4225 4226 4227
	dmar_domain->gaw = addr_width;

	/*
	 * Knock out extra levels of page tables if necessary
	 */
	while (iommu->agaw < dmar_domain->agaw) {
		struct dma_pte *pte;

		pte = dmar_domain->pgd;
		if (dma_pte_present(pte)) {
4228 4229
			dmar_domain->pgd = (struct dma_pte *)
				phys_to_virt(dma_pte_addr(pte));
4230
			free_pgtable_page(pte);
4231 4232 4233
		}
		dmar_domain->agaw--;
	}
4234

4235
	return domain_add_dev_info(dmar_domain, dev, CONTEXT_TT_MULTI_LEVEL);
K
Kay, Allen M 已提交
4236 4237
}

4238 4239
static void intel_iommu_detach_device(struct iommu_domain *domain,
				      struct device *dev)
K
Kay, Allen M 已提交
4240
{
4241 4242
	struct dmar_domain *dmar_domain = domain->priv;

4243
	domain_remove_one_dev_info(dmar_domain, dev);
4244
}
4245

4246 4247
static int intel_iommu_map(struct iommu_domain *domain,
			   unsigned long iova, phys_addr_t hpa,
4248
			   size_t size, int iommu_prot)
4249
{
4250
	struct dmar_domain *dmar_domain = domain->priv;
4251
	u64 max_addr;
4252
	int prot = 0;
4253
	int ret;
4254

4255 4256 4257 4258
	if (iommu_prot & IOMMU_READ)
		prot |= DMA_PTE_READ;
	if (iommu_prot & IOMMU_WRITE)
		prot |= DMA_PTE_WRITE;
4259 4260
	if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
		prot |= DMA_PTE_SNP;
4261

4262
	max_addr = iova + size;
4263
	if (dmar_domain->max_addr < max_addr) {
4264 4265 4266
		u64 end;

		/* check if minimum agaw is sufficient for mapped address */
4267
		end = __DOMAIN_MAX_ADDR(dmar_domain->gaw) + 1;
4268
		if (end < max_addr) {
4269
			printk(KERN_ERR "%s: iommu width (%d) is not "
4270
			       "sufficient for the mapped address (%llx)\n",
4271
			       __func__, dmar_domain->gaw, max_addr);
4272 4273
			return -EFAULT;
		}
4274
		dmar_domain->max_addr = max_addr;
4275
	}
4276 4277
	/* Round up size to next multiple of PAGE_SIZE, if it and
	   the low bits of hpa would take us onto the next page */
4278
	size = aligned_nrpages(hpa, size);
4279 4280
	ret = domain_pfn_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
				 hpa >> VTD_PAGE_SHIFT, size, prot);
4281
	return ret;
K
Kay, Allen M 已提交
4282 4283
}

4284
static size_t intel_iommu_unmap(struct iommu_domain *domain,
4285
				unsigned long iova, size_t size)
K
Kay, Allen M 已提交
4286
{
4287
	struct dmar_domain *dmar_domain = domain->priv;
4288 4289 4290 4291 4292
	struct page *freelist = NULL;
	struct intel_iommu *iommu;
	unsigned long start_pfn, last_pfn;
	unsigned int npages;
	int iommu_id, num, ndomains, level = 0;
4293 4294 4295 4296 4297 4298 4299 4300

	/* Cope with horrid API which requires us to unmap more than the
	   size argument if it happens to be a large-page mapping. */
	if (!pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level))
		BUG();

	if (size < VTD_PAGE_SIZE << level_to_offset_bits(level))
		size = VTD_PAGE_SIZE << level_to_offset_bits(level);
4301

4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322 4323 4324
	start_pfn = iova >> VTD_PAGE_SHIFT;
	last_pfn = (iova + size - 1) >> VTD_PAGE_SHIFT;

	freelist = domain_unmap(dmar_domain, start_pfn, last_pfn);

	npages = last_pfn - start_pfn + 1;

	for_each_set_bit(iommu_id, dmar_domain->iommu_bmp, g_num_of_iommus) {
               iommu = g_iommus[iommu_id];

               /*
                * find bit position of dmar_domain
                */
               ndomains = cap_ndoms(iommu->cap);
               for_each_set_bit(num, iommu->domain_ids, ndomains) {
                       if (iommu->domains[num] == dmar_domain)
                               iommu_flush_iotlb_psi(iommu, num, start_pfn,
						     npages, !freelist, 0);
	       }

	}

	dma_free_pagelist(freelist);
4325

4326 4327
	if (dmar_domain->max_addr == iova + size)
		dmar_domain->max_addr = iova;
4328

4329
	return size;
K
Kay, Allen M 已提交
4330 4331
}

4332
static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
4333
					    dma_addr_t iova)
K
Kay, Allen M 已提交
4334
{
4335
	struct dmar_domain *dmar_domain = domain->priv;
K
Kay, Allen M 已提交
4336
	struct dma_pte *pte;
4337
	int level = 0;
4338
	u64 phys = 0;
K
Kay, Allen M 已提交
4339

4340
	pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT, &level);
K
Kay, Allen M 已提交
4341
	if (pte)
4342
		phys = dma_pte_addr(pte);
K
Kay, Allen M 已提交
4343

4344
	return phys;
K
Kay, Allen M 已提交
4345
}
4346

S
Sheng Yang 已提交
4347 4348 4349 4350 4351 4352 4353
static int intel_iommu_domain_has_cap(struct iommu_domain *domain,
				      unsigned long cap)
{
	struct dmar_domain *dmar_domain = domain->priv;

	if (cap == IOMMU_CAP_CACHE_COHERENCY)
		return dmar_domain->iommu_snooping;
4354
	if (cap == IOMMU_CAP_INTR_REMAP)
4355
		return irq_remapping_enabled;
S
Sheng Yang 已提交
4356 4357 4358 4359

	return 0;
}

4360
#define REQ_ACS_FLAGS	(PCI_ACS_SV | PCI_ACS_RR | PCI_ACS_CR | PCI_ACS_UF)
4361

4362 4363 4364
static int intel_iommu_add_device(struct device *dev)
{
	struct pci_dev *pdev = to_pci_dev(dev);
4365
	struct pci_dev *bridge, *dma_pdev = NULL;
4366 4367
	struct iommu_group *group;
	int ret;
4368
	u8 bus, devfn;
4369

4370
	if (!device_to_iommu(dev, &bus, &devfn))
4371 4372 4373 4374
		return -ENODEV;

	bridge = pci_find_upstream_pcie_bridge(pdev);
	if (bridge) {
4375 4376 4377 4378
		if (pci_is_pcie(bridge))
			dma_pdev = pci_get_domain_bus_and_slot(
						pci_domain_nr(pdev->bus),
						bridge->subordinate->number, 0);
4379
		if (!dma_pdev)
4380 4381 4382 4383
			dma_pdev = pci_dev_get(bridge);
	} else
		dma_pdev = pci_dev_get(pdev);

4384
	/* Account for quirked devices */
4385 4386
	swap_pci_ref(&dma_pdev, pci_get_dma_source(dma_pdev));

4387 4388
	/*
	 * If it's a multifunction device that does not support our
4389 4390
	 * required ACS flags, add to the same group as lowest numbered
	 * function that also does not suport the required ACS flags.
4391
	 */
4392
	if (dma_pdev->multifunction &&
4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409
	    !pci_acs_enabled(dma_pdev, REQ_ACS_FLAGS)) {
		u8 i, slot = PCI_SLOT(dma_pdev->devfn);

		for (i = 0; i < 8; i++) {
			struct pci_dev *tmp;

			tmp = pci_get_slot(dma_pdev->bus, PCI_DEVFN(slot, i));
			if (!tmp)
				continue;

			if (!pci_acs_enabled(tmp, REQ_ACS_FLAGS)) {
				swap_pci_ref(&dma_pdev, tmp);
				break;
			}
			pci_dev_put(tmp);
		}
	}
4410

4411 4412 4413 4414 4415
	/*
	 * Devices on the root bus go through the iommu.  If that's not us,
	 * find the next upstream device and test ACS up to the root bus.
	 * Finding the next device may require skipping virtual buses.
	 */
4416
	while (!pci_is_root_bus(dma_pdev->bus)) {
4417 4418 4419 4420 4421 4422 4423 4424 4425 4426
		struct pci_bus *bus = dma_pdev->bus;

		while (!bus->self) {
			if (!pci_is_root_bus(bus))
				bus = bus->parent;
			else
				goto root_bus;
		}

		if (pci_acs_path_enabled(bus->self, NULL, REQ_ACS_FLAGS))
4427 4428
			break;

4429
		swap_pci_ref(&dma_pdev, pci_dev_get(bus->self));
4430 4431
	}

4432
root_bus:
4433 4434 4435 4436 4437 4438
	group = iommu_group_get(&dma_pdev->dev);
	pci_dev_put(dma_pdev);
	if (!group) {
		group = iommu_group_alloc();
		if (IS_ERR(group))
			return PTR_ERR(group);
4439 4440
	}

4441
	ret = iommu_group_add_device(group, dev);
4442

4443 4444 4445
	iommu_group_put(group);
	return ret;
}
4446

4447 4448 4449
static void intel_iommu_remove_device(struct device *dev)
{
	iommu_group_remove_device(dev);
4450 4451
}

4452 4453 4454 4455 4456
static struct iommu_ops intel_iommu_ops = {
	.domain_init	= intel_iommu_domain_init,
	.domain_destroy = intel_iommu_domain_destroy,
	.attach_dev	= intel_iommu_attach_device,
	.detach_dev	= intel_iommu_detach_device,
4457 4458
	.map		= intel_iommu_map,
	.unmap		= intel_iommu_unmap,
4459
	.iova_to_phys	= intel_iommu_iova_to_phys,
S
Sheng Yang 已提交
4460
	.domain_has_cap = intel_iommu_domain_has_cap,
4461 4462
	.add_device	= intel_iommu_add_device,
	.remove_device	= intel_iommu_remove_device,
4463
	.pgsize_bitmap	= INTEL_IOMMU_PGSIZES,
4464
};
4465

4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480
static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
{
	/* G4x/GM45 integrated gfx dmar support is totally busted. */
	printk(KERN_INFO "DMAR: Disabling IOMMU for graphics on this chipset\n");
	dmar_map_gfx = 0;
}

DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e00, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e10, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e20, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e30, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e40, quirk_iommu_g4x_gfx);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e90, quirk_iommu_g4x_gfx);

4481
static void quirk_iommu_rwbf(struct pci_dev *dev)
4482 4483 4484
{
	/*
	 * Mobile 4 Series Chipset neglects to set RWBF capability,
4485
	 * but needs it. Same seems to hold for the desktop versions.
4486 4487 4488 4489 4490 4491
	 */
	printk(KERN_INFO "DMAR: Forcing write-buffer flush capability\n");
	rwbf_quirk = 1;
}

DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2a40, quirk_iommu_rwbf);
4492 4493 4494 4495 4496 4497
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e00, quirk_iommu_rwbf);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e10, quirk_iommu_rwbf);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e20, quirk_iommu_rwbf);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e30, quirk_iommu_rwbf);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e40, quirk_iommu_rwbf);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x2e90, quirk_iommu_rwbf);
4498

4499 4500 4501 4502 4503 4504 4505 4506 4507 4508
#define GGC 0x52
#define GGC_MEMORY_SIZE_MASK	(0xf << 8)
#define GGC_MEMORY_SIZE_NONE	(0x0 << 8)
#define GGC_MEMORY_SIZE_1M	(0x1 << 8)
#define GGC_MEMORY_SIZE_2M	(0x3 << 8)
#define GGC_MEMORY_VT_ENABLED	(0x8 << 8)
#define GGC_MEMORY_SIZE_2M_VT	(0x9 << 8)
#define GGC_MEMORY_SIZE_3M_VT	(0xa << 8)
#define GGC_MEMORY_SIZE_4M_VT	(0xb << 8)

4509
static void quirk_calpella_no_shadow_gtt(struct pci_dev *dev)
4510 4511 4512
{
	unsigned short ggc;

4513
	if (pci_read_config_word(dev, GGC, &ggc))
4514 4515
		return;

4516
	if (!(ggc & GGC_MEMORY_VT_ENABLED)) {
4517 4518
		printk(KERN_INFO "DMAR: BIOS has allocated no shadow GTT; disabling IOMMU for graphics\n");
		dmar_map_gfx = 0;
4519 4520 4521 4522 4523
	} else if (dmar_map_gfx) {
		/* we have to ensure the gfx device is idle before we flush */
		printk(KERN_INFO "DMAR: Disabling batched IOTLB flush on Ironlake\n");
		intel_iommu_strict = 1;
       }
4524 4525 4526 4527 4528 4529
}
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0040, quirk_calpella_no_shadow_gtt);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0044, quirk_calpella_no_shadow_gtt);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x0062, quirk_calpella_no_shadow_gtt);
DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_INTEL, 0x006a, quirk_calpella_no_shadow_gtt);

4530 4531 4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586
/* On Tylersburg chipsets, some BIOSes have been known to enable the
   ISOCH DMAR unit for the Azalia sound device, but not give it any
   TLB entries, which causes it to deadlock. Check for that.  We do
   this in a function called from init_dmars(), instead of in a PCI
   quirk, because we don't want to print the obnoxious "BIOS broken"
   message if VT-d is actually disabled.
*/
static void __init check_tylersburg_isoch(void)
{
	struct pci_dev *pdev;
	uint32_t vtisochctrl;

	/* If there's no Azalia in the system anyway, forget it. */
	pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x3a3e, NULL);
	if (!pdev)
		return;
	pci_dev_put(pdev);

	/* System Management Registers. Might be hidden, in which case
	   we can't do the sanity check. But that's OK, because the
	   known-broken BIOSes _don't_ actually hide it, so far. */
	pdev = pci_get_device(PCI_VENDOR_ID_INTEL, 0x342e, NULL);
	if (!pdev)
		return;

	if (pci_read_config_dword(pdev, 0x188, &vtisochctrl)) {
		pci_dev_put(pdev);
		return;
	}

	pci_dev_put(pdev);

	/* If Azalia DMA is routed to the non-isoch DMAR unit, fine. */
	if (vtisochctrl & 1)
		return;

	/* Drop all bits other than the number of TLB entries */
	vtisochctrl &= 0x1c;

	/* If we have the recommended number of TLB entries (16), fine. */
	if (vtisochctrl == 0x10)
		return;

	/* Zero TLB entries? You get to ride the short bus to school. */
	if (!vtisochctrl) {
		WARN(1, "Your BIOS is broken; DMA routed to ISOCH DMAR unit but no TLB space.\n"
		     "BIOS vendor: %s; Ver: %s; Product Version: %s\n",
		     dmi_get_system_info(DMI_BIOS_VENDOR),
		     dmi_get_system_info(DMI_BIOS_VERSION),
		     dmi_get_system_info(DMI_PRODUCT_VERSION));
		iommu_identity_mapping |= IDENTMAP_AZALIA;
		return;
	}
	
	printk(KERN_WARNING "DMAR: Recommended TLB entries for ISOCH unit is 16; your BIOS set %d\n",
	       vtisochctrl);
}