amd_iommu_init.c 39.5 KB
Newer Older
1
/*
2
 * Copyright (C) 2007-2010 Advanced Micro Devices, Inc.
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 * Author: Joerg Roedel <joerg.roedel@amd.com>
 *         Leo Duran <leo.duran@amd.com>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 as published
 * by the Free Software Foundation.
 *
 * This program is distributed in the hope that 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.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
 */

#include <linux/pci.h>
#include <linux/acpi.h>
#include <linux/list.h>
23
#include <linux/slab.h>
24
#include <linux/syscore_ops.h>
25 26
#include <linux/interrupt.h>
#include <linux/msi.h>
27
#include <linux/amd-iommu.h>
28
#include <asm/pci-direct.h>
29
#include <asm/iommu.h>
30
#include <asm/gart.h>
31
#include <asm/x86_init.h>
32
#include <asm/iommu_table.h>
33 34 35 36

#include "amd_iommu_proto.h"
#include "amd_iommu_types.h"

37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55
/*
 * definitions for the ACPI scanning code
 */
#define IVRS_HEADER_LENGTH 48

#define ACPI_IVHD_TYPE                  0x10
#define ACPI_IVMD_TYPE_ALL              0x20
#define ACPI_IVMD_TYPE                  0x21
#define ACPI_IVMD_TYPE_RANGE            0x22

#define IVHD_DEV_ALL                    0x01
#define IVHD_DEV_SELECT                 0x02
#define IVHD_DEV_SELECT_RANGE_START     0x03
#define IVHD_DEV_RANGE_END              0x04
#define IVHD_DEV_ALIAS                  0x42
#define IVHD_DEV_ALIAS_RANGE            0x43
#define IVHD_DEV_EXT_SELECT             0x46
#define IVHD_DEV_EXT_SELECT_RANGE       0x47

J
Joerg Roedel 已提交
56 57 58 59
#define IVHD_FLAG_HT_TUN_EN_MASK        0x01
#define IVHD_FLAG_PASSPW_EN_MASK        0x02
#define IVHD_FLAG_RESPASSPW_EN_MASK     0x04
#define IVHD_FLAG_ISOC_EN_MASK          0x08
60 61 62 63 64 65 66 67 68 69 70 71 72

#define IVMD_FLAG_EXCL_RANGE            0x08
#define IVMD_FLAG_UNITY_MAP             0x01

#define ACPI_DEVFLAG_INITPASS           0x01
#define ACPI_DEVFLAG_EXTINT             0x02
#define ACPI_DEVFLAG_NMI                0x04
#define ACPI_DEVFLAG_SYSMGT1            0x10
#define ACPI_DEVFLAG_SYSMGT2            0x20
#define ACPI_DEVFLAG_LINT0              0x40
#define ACPI_DEVFLAG_LINT1              0x80
#define ACPI_DEVFLAG_ATSDIS             0x10000000

73 74 75 76 77 78 79 80 81 82 83
/*
 * ACPI table definitions
 *
 * These data structures are laid over the table to parse the important values
 * out of it.
 */

/*
 * structure describing one IOMMU in the ACPI table. Typically followed by one
 * or more ivhd_entrys.
 */
84 85 86 87 88 89 90 91 92 93 94 95
struct ivhd_header {
	u8 type;
	u8 flags;
	u16 length;
	u16 devid;
	u16 cap_ptr;
	u64 mmio_phys;
	u16 pci_seg;
	u16 info;
	u32 reserved;
} __attribute__((packed));

96 97 98 99
/*
 * A device entry describing which devices a specific IOMMU translates and
 * which requestor ids they use.
 */
100 101 102 103 104 105 106
struct ivhd_entry {
	u8 type;
	u16 devid;
	u8 flags;
	u32 ext;
} __attribute__((packed));

107 108 109 110
/*
 * An AMD IOMMU memory definition structure. It defines things like exclusion
 * ranges for devices and regions that should be unity mapped.
 */
111 112 113 114 115 116 117 118 119 120 121
struct ivmd_header {
	u8 type;
	u8 flags;
	u16 length;
	u16 devid;
	u16 aux;
	u64 resv;
	u64 range_start;
	u64 range_length;
} __attribute__((packed));

122 123
bool amd_iommu_dump;

124
static int __initdata amd_iommu_detected;
125
static bool __initdata amd_iommu_disabled;
126

127 128
u16 amd_iommu_last_bdf;			/* largest PCI device id we have
					   to handle */
129
LIST_HEAD(amd_iommu_unity_map);		/* a list of required unity mappings
130
					   we find in ACPI */
131
bool amd_iommu_unmap_flush;		/* if true, flush on every unmap */
132

133
LIST_HEAD(amd_iommu_list);		/* list of all AMD IOMMUs in the
134
					   system */
135

136 137 138 139
/* Array to assign indices to IOMMUs*/
struct amd_iommu *amd_iommus[MAX_IOMMUS];
int amd_iommus_present;

140 141
/* IOMMUs have a non-present cache? */
bool amd_iommu_np_cache __read_mostly;
142
bool amd_iommu_iotlb_sup __read_mostly = true;
143

144
/*
145
 * The ACPI table parsing functions set this variable on an error
146
 */
147
static int __initdata amd_iommu_init_err;
148

149 150 151 152 153 154
/*
 * List of protection domains - used during resume
 */
LIST_HEAD(amd_iommu_pd_list);
spinlock_t amd_iommu_pd_lock;

155 156 157 158 159 160
/*
 * Pointer to the device table which is shared by all AMD IOMMUs
 * it is indexed by the PCI device id or the HT unit id and contains
 * information about the domain the device belongs to as well as the
 * page table root pointer.
 */
161
struct dev_table_entry *amd_iommu_dev_table;
162 163 164 165 166 167

/*
 * The alias table is a driver specific data structure which contains the
 * mappings of the PCI device ids to the actual requestor ids on the IOMMU.
 * More than one device can share the same requestor id.
 */
168
u16 *amd_iommu_alias_table;
169 170 171 172 173

/*
 * The rlookup table is used to find the IOMMU which is responsible
 * for a specific device. It is also indexed by the PCI device id.
 */
174
struct amd_iommu **amd_iommu_rlookup_table;
175 176 177 178 179

/*
 * AMD IOMMU allows up to 2^16 differend protection domains. This is a bitmap
 * to know which ones are already in use.
 */
180 181
unsigned long *amd_iommu_pd_alloc_bitmap;

182 183 184
static u32 dev_table_size;	/* size of the device table */
static u32 alias_table_size;	/* size of the alias table */
static u32 rlookup_table_size;	/* size if the rlookup table */
185

186 187 188 189 190 191
/*
 * This function flushes all internal caches of
 * the IOMMU used by this driver.
 */
extern void iommu_flush_all_caches(struct amd_iommu *iommu);

192 193 194 195 196 197
static inline void update_last_devid(u16 devid)
{
	if (devid > amd_iommu_last_bdf)
		amd_iommu_last_bdf = devid;
}

198 199 200
static inline unsigned long tbl_size(int entry_size)
{
	unsigned shift = PAGE_SHIFT +
201
			 get_order(((int)amd_iommu_last_bdf + 1) * entry_size);
202 203 204 205

	return 1UL << shift;
}

206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238
/* Access to l1 and l2 indexed register spaces */

static u32 iommu_read_l1(struct amd_iommu *iommu, u16 l1, u8 address)
{
	u32 val;

	pci_write_config_dword(iommu->dev, 0xf8, (address | l1 << 16));
	pci_read_config_dword(iommu->dev, 0xfc, &val);
	return val;
}

static void iommu_write_l1(struct amd_iommu *iommu, u16 l1, u8 address, u32 val)
{
	pci_write_config_dword(iommu->dev, 0xf8, (address | l1 << 16 | 1 << 31));
	pci_write_config_dword(iommu->dev, 0xfc, val);
	pci_write_config_dword(iommu->dev, 0xf8, (address | l1 << 16));
}

static u32 iommu_read_l2(struct amd_iommu *iommu, u8 address)
{
	u32 val;

	pci_write_config_dword(iommu->dev, 0xf0, address);
	pci_read_config_dword(iommu->dev, 0xf4, &val);
	return val;
}

static void iommu_write_l2(struct amd_iommu *iommu, u8 address, u32 val)
{
	pci_write_config_dword(iommu->dev, 0xf0, (address | 1 << 8));
	pci_write_config_dword(iommu->dev, 0xf4, val);
}

239 240 241 242 243 244 245 246
/****************************************************************************
 *
 * AMD IOMMU MMIO register space handling functions
 *
 * These functions are used to program the IOMMU device registers in
 * MMIO space required for that driver.
 *
 ****************************************************************************/
247

248 249 250 251
/*
 * This function set the exclusion range in the IOMMU. DMA accesses to the
 * exclusion range are passed through untranslated
 */
252
static void iommu_set_exclusion_range(struct amd_iommu *iommu)
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269
{
	u64 start = iommu->exclusion_start & PAGE_MASK;
	u64 limit = (start + iommu->exclusion_length) & PAGE_MASK;
	u64 entry;

	if (!iommu->exclusion_start)
		return;

	entry = start | MMIO_EXCL_ENABLE_MASK;
	memcpy_toio(iommu->mmio_base + MMIO_EXCL_BASE_OFFSET,
			&entry, sizeof(entry));

	entry = limit;
	memcpy_toio(iommu->mmio_base + MMIO_EXCL_LIMIT_OFFSET,
			&entry, sizeof(entry));
}

270
/* Programs the physical address of the device table into the IOMMU hardware */
271 272
static void __init iommu_set_device_table(struct amd_iommu *iommu)
{
273
	u64 entry;
274 275 276 277 278 279 280 281 282

	BUG_ON(iommu->mmio_base == NULL);

	entry = virt_to_phys(amd_iommu_dev_table);
	entry |= (dev_table_size >> 12) - 1;
	memcpy_toio(iommu->mmio_base + MMIO_DEV_TABLE_OFFSET,
			&entry, sizeof(entry));
}

283
/* Generic functions to enable/disable certain features of the IOMMU. */
284
static void iommu_feature_enable(struct amd_iommu *iommu, u8 bit)
285 286 287 288 289 290 291 292
{
	u32 ctrl;

	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
	ctrl |= (1 << bit);
	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
}

293
static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
294 295 296
{
	u32 ctrl;

297
	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
298 299 300 301
	ctrl &= ~(1 << bit);
	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
}

302
/* Function to enable the hardware */
303
static void iommu_enable(struct amd_iommu *iommu)
304
{
305 306 307 308 309 310 311
	static const char * const feat_str[] = {
		"PreF", "PPR", "X2APIC", "NX", "GT", "[5]",
		"IA", "GA", "HE", "PC", NULL
	};
	int i;

	printk(KERN_INFO "AMD-Vi: Enabling IOMMU at %s cap 0x%hx",
312
	       dev_name(&iommu->dev->dev), iommu->cap_ptr);
313

314 315 316 317 318 319 320 321
	if (iommu->cap & (1 << IOMMU_CAP_EFR)) {
		printk(KERN_CONT " extended features: ");
		for (i = 0; feat_str[i]; ++i)
			if (iommu_feature(iommu, (1ULL << i)))
				printk(KERN_CONT " %s", feat_str[i]);
	}
	printk(KERN_CONT "\n");

322 323 324
	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
}

325
static void iommu_disable(struct amd_iommu *iommu)
J
Joerg Roedel 已提交
326
{
327 328 329 330 331 332 333 334
	/* Disable command buffer */
	iommu_feature_disable(iommu, CONTROL_CMDBUF_EN);

	/* Disable event logging and event interrupts */
	iommu_feature_disable(iommu, CONTROL_EVT_INT_EN);
	iommu_feature_disable(iommu, CONTROL_EVT_LOG_EN);

	/* Disable IOMMU hardware itself */
335
	iommu_feature_disable(iommu, CONTROL_IOMMU_EN);
J
Joerg Roedel 已提交
336 337
}

338 339 340 341
/*
 * mapping and unmapping functions for the IOMMU MMIO space. Each AMD IOMMU in
 * the system has one.
 */
342 343 344 345
static u8 * __init iommu_map_mmio_space(u64 address)
{
	u8 *ret;

346 347 348 349
	if (!request_mem_region(address, MMIO_REGION_LENGTH, "amd_iommu")) {
		pr_err("AMD-Vi: Can not reserve memory region %llx for mmio\n",
			address);
		pr_err("AMD-Vi: This is a BIOS bug. Please contact your hardware vendor\n");
350
		return NULL;
351
	}
352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368

	ret = ioremap_nocache(address, MMIO_REGION_LENGTH);
	if (ret != NULL)
		return ret;

	release_mem_region(address, MMIO_REGION_LENGTH);

	return NULL;
}

static void __init iommu_unmap_mmio_space(struct amd_iommu *iommu)
{
	if (iommu->mmio_base)
		iounmap(iommu->mmio_base);
	release_mem_region(iommu->mmio_phys, MMIO_REGION_LENGTH);
}

369 370 371 372 373 374 375 376 377
/****************************************************************************
 *
 * The functions below belong to the first pass of AMD IOMMU ACPI table
 * parsing. In this pass we try to find out the highest device id this
 * code has to handle. Upon this information the size of the shared data
 * structures is determined later.
 *
 ****************************************************************************/

378 379 380 381 382 383 384 385
/*
 * This function calculates the length of a given IVHD entry
 */
static inline int ivhd_entry_length(u8 *ivhd)
{
	return 0x04 << (*ivhd >> 6);
}

386 387 388 389
/*
 * This function reads the last device id the IOMMU has to handle from the PCI
 * capability header for this IOMMU
 */
390 391 392 393 394
static int __init find_last_devid_on_pci(int bus, int dev, int fn, int cap_ptr)
{
	u32 cap;

	cap = read_pci_config(bus, dev, fn, cap_ptr+MMIO_RANGE_OFFSET);
395
	update_last_devid(calc_devid(MMIO_GET_BUS(cap), MMIO_GET_LD(cap)));
396 397 398 399

	return 0;
}

400 401 402 403
/*
 * After reading the highest device id from the IOMMU PCI capability header
 * this function looks if there is a higher device id defined in the ACPI table
 */
404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
static int __init find_last_devid_from_ivhd(struct ivhd_header *h)
{
	u8 *p = (void *)h, *end = (void *)h;
	struct ivhd_entry *dev;

	p += sizeof(*h);
	end += h->length;

	find_last_devid_on_pci(PCI_BUS(h->devid),
			PCI_SLOT(h->devid),
			PCI_FUNC(h->devid),
			h->cap_ptr);

	while (p < end) {
		dev = (struct ivhd_entry *)p;
		switch (dev->type) {
		case IVHD_DEV_SELECT:
		case IVHD_DEV_RANGE_END:
		case IVHD_DEV_ALIAS:
		case IVHD_DEV_EXT_SELECT:
424
			/* all the above subfield types refer to device ids */
425
			update_last_devid(dev->devid);
426 427 428 429
			break;
		default:
			break;
		}
430
		p += ivhd_entry_length(p);
431 432 433 434 435 436 437
	}

	WARN_ON(p != end);

	return 0;
}

438 439 440 441 442
/*
 * Iterate over all IVHD entries in the ACPI table and find the highest device
 * id which we need to handle. This is the first of three functions which parse
 * the ACPI table. So we check the checksum here.
 */
443 444 445 446 447 448 449 450 451 452 453 454
static int __init find_last_devid_acpi(struct acpi_table_header *table)
{
	int i;
	u8 checksum = 0, *p = (u8 *)table, *end = (u8 *)table;
	struct ivhd_header *h;

	/*
	 * Validate checksum here so we don't need to do it when
	 * we actually parse the table
	 */
	for (i = 0; i < table->length; ++i)
		checksum += p[i];
455
	if (checksum != 0) {
456
		/* ACPI table corrupt */
457 458 459
		amd_iommu_init_err = -ENODEV;
		return 0;
	}
460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479

	p += IVRS_HEADER_LENGTH;

	end += table->length;
	while (p < end) {
		h = (struct ivhd_header *)p;
		switch (h->type) {
		case ACPI_IVHD_TYPE:
			find_last_devid_from_ivhd(h);
			break;
		default:
			break;
		}
		p += h->length;
	}
	WARN_ON(p != end);

	return 0;
}

480 481 482 483 484 485 486 487 488 489 490 491 492 493
/****************************************************************************
 *
 * The following functions belong the the code path which parses the ACPI table
 * the second time. In this ACPI parsing iteration we allocate IOMMU specific
 * data structures, initialize the device/alias/rlookup table and also
 * basically initialize the hardware.
 *
 ****************************************************************************/

/*
 * Allocates the command buffer. This buffer is per AMD IOMMU. We can
 * write commands to that buffer later and the IOMMU will execute them
 * asynchronously
 */
494 495
static u8 * __init alloc_command_buffer(struct amd_iommu *iommu)
{
496
	u8 *cmd_buf = (u8 *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
497 498 499 500 501
			get_order(CMD_BUFFER_SIZE));

	if (cmd_buf == NULL)
		return NULL;

502
	iommu->cmd_buf_size = CMD_BUFFER_SIZE | CMD_BUFFER_UNINITIALIZED;
503

504 505 506
	return cmd_buf;
}

507 508 509 510 511 512 513 514 515 516 517 518 519 520
/*
 * This function resets the command buffer if the IOMMU stopped fetching
 * commands from it.
 */
void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu)
{
	iommu_feature_disable(iommu, CONTROL_CMDBUF_EN);

	writel(0x00, iommu->mmio_base + MMIO_CMD_HEAD_OFFSET);
	writel(0x00, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);

	iommu_feature_enable(iommu, CONTROL_CMDBUF_EN);
}

521 522 523 524 525 526 527 528 529 530 531
/*
 * This function writes the command buffer address to the hardware and
 * enables it.
 */
static void iommu_enable_command_buffer(struct amd_iommu *iommu)
{
	u64 entry;

	BUG_ON(iommu->cmd_buf == NULL);

	entry = (u64)virt_to_phys(iommu->cmd_buf);
532
	entry |= MMIO_CMD_SIZE_512;
533

534
	memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET,
535
		    &entry, sizeof(entry));
536

537
	amd_iommu_reset_cmd_buffer(iommu);
538
	iommu->cmd_buf_size &= ~(CMD_BUFFER_UNINITIALIZED);
539 540 541 542
}

static void __init free_command_buffer(struct amd_iommu *iommu)
{
543
	free_pages((unsigned long)iommu->cmd_buf,
544
		   get_order(iommu->cmd_buf_size & ~(CMD_BUFFER_UNINITIALIZED)));
545 546
}

547 548 549 550 551 552 553 554 555
/* allocates the memory where the IOMMU will log its events to */
static u8 * __init alloc_event_buffer(struct amd_iommu *iommu)
{
	iommu->evt_buf = (u8 *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
						get_order(EVT_BUFFER_SIZE));

	if (iommu->evt_buf == NULL)
		return NULL;

556 557
	iommu->evt_buf_size = EVT_BUFFER_SIZE;

558 559 560 561 562 563 564 565 566
	return iommu->evt_buf;
}

static void iommu_enable_event_buffer(struct amd_iommu *iommu)
{
	u64 entry;

	BUG_ON(iommu->evt_buf == NULL);

567
	entry = (u64)virt_to_phys(iommu->evt_buf) | EVT_LEN_MASK;
568

569 570 571
	memcpy_toio(iommu->mmio_base + MMIO_EVT_BUF_OFFSET,
		    &entry, sizeof(entry));

572 573 574 575
	/* set head and tail to zero manually */
	writel(0x00, iommu->mmio_base + MMIO_EVT_HEAD_OFFSET);
	writel(0x00, iommu->mmio_base + MMIO_EVT_TAIL_OFFSET);

576
	iommu_feature_enable(iommu, CONTROL_EVT_LOG_EN);
577 578 579 580 581 582 583
}

static void __init free_event_buffer(struct amd_iommu *iommu)
{
	free_pages((unsigned long)iommu->evt_buf, get_order(EVT_BUFFER_SIZE));
}

584
/* sets a specific bit in the device table entry. */
585 586 587 588 589 590 591 592
static void set_dev_entry_bit(u16 devid, u8 bit)
{
	int i = (bit >> 5) & 0x07;
	int _bit = bit & 0x1f;

	amd_iommu_dev_table[devid].data[i] |= (1 << _bit);
}

593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612
static int get_dev_entry_bit(u16 devid, u8 bit)
{
	int i = (bit >> 5) & 0x07;
	int _bit = bit & 0x1f;

	return (amd_iommu_dev_table[devid].data[i] & (1 << _bit)) >> _bit;
}


void amd_iommu_apply_erratum_63(u16 devid)
{
	int sysmgt;

	sysmgt = get_dev_entry_bit(devid, DEV_ENTRY_SYSMGT1) |
		 (get_dev_entry_bit(devid, DEV_ENTRY_SYSMGT2) << 1);

	if (sysmgt == 0x01)
		set_dev_entry_bit(devid, DEV_ENTRY_IW);
}

613 614 615 616 617 618
/* Writes the specific IOMMU for a device into the rlookup table */
static void __init set_iommu_for_device(struct amd_iommu *iommu, u16 devid)
{
	amd_iommu_rlookup_table[devid] = iommu;
}

619 620 621 622
/*
 * This function takes the device specific flags read from the ACPI
 * table and sets up the device table entry with that information
 */
623 624
static void __init set_dev_entry_from_acpi(struct amd_iommu *iommu,
					   u16 devid, u32 flags, u32 ext_flags)
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
{
	if (flags & ACPI_DEVFLAG_INITPASS)
		set_dev_entry_bit(devid, DEV_ENTRY_INIT_PASS);
	if (flags & ACPI_DEVFLAG_EXTINT)
		set_dev_entry_bit(devid, DEV_ENTRY_EINT_PASS);
	if (flags & ACPI_DEVFLAG_NMI)
		set_dev_entry_bit(devid, DEV_ENTRY_NMI_PASS);
	if (flags & ACPI_DEVFLAG_SYSMGT1)
		set_dev_entry_bit(devid, DEV_ENTRY_SYSMGT1);
	if (flags & ACPI_DEVFLAG_SYSMGT2)
		set_dev_entry_bit(devid, DEV_ENTRY_SYSMGT2);
	if (flags & ACPI_DEVFLAG_LINT0)
		set_dev_entry_bit(devid, DEV_ENTRY_LINT0_PASS);
	if (flags & ACPI_DEVFLAG_LINT1)
		set_dev_entry_bit(devid, DEV_ENTRY_LINT1_PASS);

641 642
	amd_iommu_apply_erratum_63(devid);

643
	set_iommu_for_device(iommu, devid);
644 645
}

646 647 648 649
/*
 * Reads the device exclusion range from ACPI and initialize IOMMU with
 * it
 */
650 651 652 653 654 655 656 657
static void __init set_device_exclusion_range(u16 devid, struct ivmd_header *m)
{
	struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];

	if (!(m->flags & IVMD_FLAG_EXCL_RANGE))
		return;

	if (iommu) {
658 659 660 661 662
		/*
		 * We only can configure exclusion ranges per IOMMU, not
		 * per device. But we can enable the exclusion range per
		 * device. This is done here
		 */
663 664 665 666 667 668
		set_dev_entry_bit(m->devid, DEV_ENTRY_EX);
		iommu->exclusion_start = m->range_start;
		iommu->exclusion_length = m->range_length;
	}
}

669 670 671 672 673
/*
 * This function reads some important data from the IOMMU PCI space and
 * initializes the driver data structure with it. It reads the hardware
 * capabilities and the first/last device entries
 */
674 675 676
static void __init init_iommu_from_pci(struct amd_iommu *iommu)
{
	int cap_ptr = iommu->cap_ptr;
677
	u32 range, misc, low, high;
678
	int i, j;
679

680 681 682 683
	pci_read_config_dword(iommu->dev, cap_ptr + MMIO_CAP_HDR_OFFSET,
			      &iommu->cap);
	pci_read_config_dword(iommu->dev, cap_ptr + MMIO_RANGE_OFFSET,
			      &range);
684 685
	pci_read_config_dword(iommu->dev, cap_ptr + MMIO_MISC_OFFSET,
			      &misc);
686

687 688 689 690
	iommu->first_device = calc_devid(MMIO_GET_BUS(range),
					 MMIO_GET_FD(range));
	iommu->last_device = calc_devid(MMIO_GET_BUS(range),
					MMIO_GET_LD(range));
691
	iommu->evt_msi_num = MMIO_MSI_NUM(misc);
692

693 694 695
	if (!(iommu->cap & (1 << IOMMU_CAP_IOTLB)))
		amd_iommu_iotlb_sup = false;

696 697 698 699 700 701
	/* read extended feature bits */
	low  = readl(iommu->mmio_base + MMIO_EXT_FEATURES);
	high = readl(iommu->mmio_base + MMIO_EXT_FEATURES + 4);

	iommu->features = ((u64)high << 32) | low;

702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
	if (!is_rd890_iommu(iommu->dev))
		return;

	/*
	 * Some rd890 systems may not be fully reconfigured by the BIOS, so
	 * it's necessary for us to store this information so it can be
	 * reprogrammed on resume
	 */

	pci_read_config_dword(iommu->dev, iommu->cap_ptr + 4,
			      &iommu->stored_addr_lo);
	pci_read_config_dword(iommu->dev, iommu->cap_ptr + 8,
			      &iommu->stored_addr_hi);

	/* Low bit locks writes to configuration space */
	iommu->stored_addr_lo &= ~1;

	for (i = 0; i < 6; i++)
		for (j = 0; j < 0x12; j++)
			iommu->stored_l1[i][j] = iommu_read_l1(iommu, i, j);

	for (i = 0; i < 0x83; i++)
		iommu->stored_l2[i] = iommu_read_l2(iommu, i);
725 726
}

727 728 729 730
/*
 * Takes a pointer to an AMD IOMMU entry in the ACPI table and
 * initializes the hardware and our data structures with it.
 */
731 732 733 734 735
static void __init init_iommu_from_acpi(struct amd_iommu *iommu,
					struct ivhd_header *h)
{
	u8 *p = (u8 *)h;
	u8 *end = p, flags = 0;
736 737
	u16 devid = 0, devid_start = 0, devid_to = 0;
	u32 dev_i, ext_flags = 0;
738
	bool alias = false;
739 740 741
	struct ivhd_entry *e;

	/*
742
	 * First save the recommended feature enable bits from ACPI
743
	 */
744
	iommu->acpi_flags = h->flags;
745 746 747 748 749 750 751

	/*
	 * Done. Now parse the device entries
	 */
	p += sizeof(struct ivhd_header);
	end += h->length;

752

753 754 755 756
	while (p < end) {
		e = (struct ivhd_entry *)p;
		switch (e->type) {
		case IVHD_DEV_ALL:
757 758 759 760 761 762 763 764 765 766 767

			DUMP_printk("  DEV_ALL\t\t\t first devid: %02x:%02x.%x"
				    " last device %02x:%02x.%x flags: %02x\n",
				    PCI_BUS(iommu->first_device),
				    PCI_SLOT(iommu->first_device),
				    PCI_FUNC(iommu->first_device),
				    PCI_BUS(iommu->last_device),
				    PCI_SLOT(iommu->last_device),
				    PCI_FUNC(iommu->last_device),
				    e->flags);

768 769
			for (dev_i = iommu->first_device;
					dev_i <= iommu->last_device; ++dev_i)
770 771
				set_dev_entry_from_acpi(iommu, dev_i,
							e->flags, 0);
772 773
			break;
		case IVHD_DEV_SELECT:
774 775 776 777 778 779 780 781

			DUMP_printk("  DEV_SELECT\t\t\t devid: %02x:%02x.%x "
				    "flags: %02x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags);

782
			devid = e->devid;
783
			set_dev_entry_from_acpi(iommu, devid, e->flags, 0);
784 785
			break;
		case IVHD_DEV_SELECT_RANGE_START:
786 787 788 789 790 791 792 793

			DUMP_printk("  DEV_SELECT_RANGE_START\t "
				    "devid: %02x:%02x.%x flags: %02x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags);

794 795 796
			devid_start = e->devid;
			flags = e->flags;
			ext_flags = 0;
797
			alias = false;
798 799
			break;
		case IVHD_DEV_ALIAS:
800 801 802 803 804 805 806 807 808 809 810

			DUMP_printk("  DEV_ALIAS\t\t\t devid: %02x:%02x.%x "
				    "flags: %02x devid_to: %02x:%02x.%x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags,
				    PCI_BUS(e->ext >> 8),
				    PCI_SLOT(e->ext >> 8),
				    PCI_FUNC(e->ext >> 8));

811 812
			devid = e->devid;
			devid_to = e->ext >> 8;
813
			set_dev_entry_from_acpi(iommu, devid   , e->flags, 0);
814
			set_dev_entry_from_acpi(iommu, devid_to, e->flags, 0);
815 816 817
			amd_iommu_alias_table[devid] = devid_to;
			break;
		case IVHD_DEV_ALIAS_RANGE:
818 819 820 821 822 823 824 825 826 827 828 829

			DUMP_printk("  DEV_ALIAS_RANGE\t\t "
				    "devid: %02x:%02x.%x flags: %02x "
				    "devid_to: %02x:%02x.%x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags,
				    PCI_BUS(e->ext >> 8),
				    PCI_SLOT(e->ext >> 8),
				    PCI_FUNC(e->ext >> 8));

830 831 832 833
			devid_start = e->devid;
			flags = e->flags;
			devid_to = e->ext >> 8;
			ext_flags = 0;
834
			alias = true;
835 836
			break;
		case IVHD_DEV_EXT_SELECT:
837 838 839 840 841 842 843 844

			DUMP_printk("  DEV_EXT_SELECT\t\t devid: %02x:%02x.%x "
				    "flags: %02x ext: %08x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags, e->ext);

845
			devid = e->devid;
846 847
			set_dev_entry_from_acpi(iommu, devid, e->flags,
						e->ext);
848 849
			break;
		case IVHD_DEV_EXT_SELECT_RANGE:
850 851 852 853 854 855 856 857

			DUMP_printk("  DEV_EXT_SELECT_RANGE\t devid: "
				    "%02x:%02x.%x flags: %02x ext: %08x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags, e->ext);

858 859 860
			devid_start = e->devid;
			flags = e->flags;
			ext_flags = e->ext;
861
			alias = false;
862 863
			break;
		case IVHD_DEV_RANGE_END:
864 865 866 867 868 869

			DUMP_printk("  DEV_RANGE_END\t\t devid: %02x:%02x.%x\n",
				    PCI_BUS(e->devid),
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid));

870 871
			devid = e->devid;
			for (dev_i = devid_start; dev_i <= devid; ++dev_i) {
872
				if (alias) {
873
					amd_iommu_alias_table[dev_i] = devid_to;
874 875 876 877 878
					set_dev_entry_from_acpi(iommu,
						devid_to, flags, ext_flags);
				}
				set_dev_entry_from_acpi(iommu, dev_i,
							flags, ext_flags);
879 880 881 882 883 884
			}
			break;
		default:
			break;
		}

885
		p += ivhd_entry_length(p);
886 887 888
	}
}

889
/* Initializes the device->iommu mapping for the driver */
890 891
static int __init init_iommu_devices(struct amd_iommu *iommu)
{
892
	u32 i;
893 894 895 896 897 898 899

	for (i = iommu->first_device; i <= iommu->last_device; ++i)
		set_iommu_for_device(iommu, i);

	return 0;
}

900 901 902
static void __init free_iommu_one(struct amd_iommu *iommu)
{
	free_command_buffer(iommu);
903
	free_event_buffer(iommu);
904 905 906 907 908 909 910
	iommu_unmap_mmio_space(iommu);
}

static void __init free_iommu_all(void)
{
	struct amd_iommu *iommu, *next;

911
	for_each_iommu_safe(iommu, next) {
912 913 914 915 916 917
		list_del(&iommu->list);
		free_iommu_one(iommu);
		kfree(iommu);
	}
}

918 919 920 921 922
/*
 * This function clues the initialization function for one IOMMU
 * together and also allocates the command buffer and programs the
 * hardware. It does NOT enable the IOMMU. This is done afterwards.
 */
923 924 925
static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h)
{
	spin_lock_init(&iommu->lock);
926 927

	/* Add IOMMU to internal data structures */
928
	list_add_tail(&iommu->list, &amd_iommu_list);
929 930 931 932 933 934 935 936 937
	iommu->index             = amd_iommus_present++;

	if (unlikely(iommu->index >= MAX_IOMMUS)) {
		WARN(1, "AMD-Vi: System has more IOMMUs than supported by this driver\n");
		return -ENOSYS;
	}

	/* Index is fine - add IOMMU to the array */
	amd_iommus[iommu->index] = iommu;
938 939 940 941

	/*
	 * Copy data from ACPI table entry to the iommu struct
	 */
942 943 944 945
	iommu->dev = pci_get_bus_and_slot(PCI_BUS(h->devid), h->devid & 0xff);
	if (!iommu->dev)
		return 1;

946
	iommu->cap_ptr = h->cap_ptr;
947
	iommu->pci_seg = h->pci_seg;
948 949 950 951 952 953 954 955 956
	iommu->mmio_phys = h->mmio_phys;
	iommu->mmio_base = iommu_map_mmio_space(h->mmio_phys);
	if (!iommu->mmio_base)
		return -ENOMEM;

	iommu->cmd_buf = alloc_command_buffer(iommu);
	if (!iommu->cmd_buf)
		return -ENOMEM;

957 958 959 960
	iommu->evt_buf = alloc_event_buffer(iommu);
	if (!iommu->evt_buf)
		return -ENOMEM;

961 962
	iommu->int_enabled = false;

963 964 965 966
	init_iommu_from_pci(iommu);
	init_iommu_from_acpi(iommu, h);
	init_iommu_devices(iommu);

967 968 969
	if (iommu->cap & (1UL << IOMMU_CAP_NPCACHE))
		amd_iommu_np_cache = true;

970
	return pci_enable_device(iommu->dev);
971 972
}

973 974 975 976
/*
 * Iterates over all IOMMU entries in the ACPI table, allocates the
 * IOMMU structure and initializes it with init_iommu_one()
 */
977 978 979 980 981 982 983 984 985 986 987 988 989 990
static int __init init_iommu_all(struct acpi_table_header *table)
{
	u8 *p = (u8 *)table, *end = (u8 *)table;
	struct ivhd_header *h;
	struct amd_iommu *iommu;
	int ret;

	end += table->length;
	p += IVRS_HEADER_LENGTH;

	while (p < end) {
		h = (struct ivhd_header *)p;
		switch (*p) {
		case ACPI_IVHD_TYPE:
991

992
			DUMP_printk("device: %02x:%02x.%01x cap: %04x "
993 994 995 996 997 998 999
				    "seg: %d flags: %01x info %04x\n",
				    PCI_BUS(h->devid), PCI_SLOT(h->devid),
				    PCI_FUNC(h->devid), h->cap_ptr,
				    h->pci_seg, h->flags, h->info);
			DUMP_printk("       mmio-addr: %016llx\n",
				    h->mmio_phys);

1000
			iommu = kzalloc(sizeof(struct amd_iommu), GFP_KERNEL);
1001 1002 1003 1004 1005
			if (iommu == NULL) {
				amd_iommu_init_err = -ENOMEM;
				return 0;
			}

1006
			ret = init_iommu_one(iommu, h);
1007 1008 1009 1010
			if (ret) {
				amd_iommu_init_err = ret;
				return 0;
			}
1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
			break;
		default:
			break;
		}
		p += h->length;

	}
	WARN_ON(p != end);

	return 0;
}

1023 1024 1025 1026 1027 1028 1029 1030 1031
/****************************************************************************
 *
 * The following functions initialize the MSI interrupts for all IOMMUs
 * in the system. Its a bit challenging because there could be multiple
 * IOMMUs per PCI BDF but we can call pci_enable_msi(x) only once per
 * pci_dev.
 *
 ****************************************************************************/

1032
static int iommu_setup_msi(struct amd_iommu *iommu)
1033 1034 1035 1036 1037 1038
{
	int r;

	if (pci_enable_msi(iommu->dev))
		return 1;

1039 1040 1041 1042 1043
	r = request_threaded_irq(iommu->dev->irq,
				 amd_iommu_int_handler,
				 amd_iommu_int_thread,
				 0, "AMD-Vi",
				 iommu->dev);
1044 1045 1046 1047 1048 1049

	if (r) {
		pci_disable_msi(iommu->dev);
		return 1;
	}

1050
	iommu->int_enabled = true;
1051 1052
	iommu_feature_enable(iommu, CONTROL_EVT_INT_EN);

1053 1054 1055
	return 0;
}

1056
static int iommu_init_msi(struct amd_iommu *iommu)
1057 1058 1059 1060
{
	if (iommu->int_enabled)
		return 0;

1061
	if (pci_find_capability(iommu->dev, PCI_CAP_ID_MSI))
1062 1063 1064 1065 1066
		return iommu_setup_msi(iommu);

	return 1;
}

1067 1068 1069 1070 1071 1072 1073 1074
/****************************************************************************
 *
 * The next functions belong to the third pass of parsing the ACPI
 * table. In this last pass the memory mapping requirements are
 * gathered (like exclusion and unity mapping reanges).
 *
 ****************************************************************************/

1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
static void __init free_unity_maps(void)
{
	struct unity_map_entry *entry, *next;

	list_for_each_entry_safe(entry, next, &amd_iommu_unity_map, list) {
		list_del(&entry->list);
		kfree(entry);
	}
}

1085
/* called when we find an exclusion range definition in ACPI */
1086 1087 1088 1089 1090 1091 1092 1093 1094
static int __init init_exclusion_range(struct ivmd_header *m)
{
	int i;

	switch (m->type) {
	case ACPI_IVMD_TYPE:
		set_device_exclusion_range(m->devid, m);
		break;
	case ACPI_IVMD_TYPE_ALL:
1095
		for (i = 0; i <= amd_iommu_last_bdf; ++i)
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108
			set_device_exclusion_range(i, m);
		break;
	case ACPI_IVMD_TYPE_RANGE:
		for (i = m->devid; i <= m->aux; ++i)
			set_device_exclusion_range(i, m);
		break;
	default:
		break;
	}

	return 0;
}

1109
/* called for unity map ACPI definition */
1110 1111 1112
static int __init init_unity_map_range(struct ivmd_header *m)
{
	struct unity_map_entry *e = 0;
1113
	char *s;
1114 1115 1116 1117 1118 1119 1120

	e = kzalloc(sizeof(*e), GFP_KERNEL);
	if (e == NULL)
		return -ENOMEM;

	switch (m->type) {
	default:
1121 1122
		kfree(e);
		return 0;
1123
	case ACPI_IVMD_TYPE:
1124
		s = "IVMD_TYPEi\t\t\t";
1125 1126 1127
		e->devid_start = e->devid_end = m->devid;
		break;
	case ACPI_IVMD_TYPE_ALL:
1128
		s = "IVMD_TYPE_ALL\t\t";
1129 1130 1131 1132
		e->devid_start = 0;
		e->devid_end = amd_iommu_last_bdf;
		break;
	case ACPI_IVMD_TYPE_RANGE:
1133
		s = "IVMD_TYPE_RANGE\t\t";
1134 1135 1136 1137 1138 1139 1140 1141
		e->devid_start = m->devid;
		e->devid_end = m->aux;
		break;
	}
	e->address_start = PAGE_ALIGN(m->range_start);
	e->address_end = e->address_start + PAGE_ALIGN(m->range_length);
	e->prot = m->flags >> 1;

1142 1143 1144 1145 1146 1147 1148
	DUMP_printk("%s devid_start: %02x:%02x.%x devid_end: %02x:%02x.%x"
		    " range_start: %016llx range_end: %016llx flags: %x\n", s,
		    PCI_BUS(e->devid_start), PCI_SLOT(e->devid_start),
		    PCI_FUNC(e->devid_start), PCI_BUS(e->devid_end),
		    PCI_SLOT(e->devid_end), PCI_FUNC(e->devid_end),
		    e->address_start, e->address_end, m->flags);

1149 1150 1151 1152 1153
	list_add_tail(&e->list, &amd_iommu_unity_map);

	return 0;
}

1154
/* iterates over all memory definitions we find in the ACPI table */
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
static int __init init_memory_definitions(struct acpi_table_header *table)
{
	u8 *p = (u8 *)table, *end = (u8 *)table;
	struct ivmd_header *m;

	end += table->length;
	p += IVRS_HEADER_LENGTH;

	while (p < end) {
		m = (struct ivmd_header *)p;
		if (m->flags & IVMD_FLAG_EXCL_RANGE)
			init_exclusion_range(m);
		else if (m->flags & IVMD_FLAG_UNITY_MAP)
			init_unity_map_range(m);

		p += m->length;
	}

	return 0;
}

1176 1177 1178 1179 1180 1181
/*
 * Init the device table to not allow DMA access for devices and
 * suppress all page faults
 */
static void init_device_table(void)
{
1182
	u32 devid;
1183 1184 1185 1186 1187 1188 1189

	for (devid = 0; devid <= amd_iommu_last_bdf; ++devid) {
		set_dev_entry_bit(devid, DEV_ENTRY_VALID);
		set_dev_entry_bit(devid, DEV_ENTRY_TRANSLATION);
	}
}

1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
static void iommu_init_flags(struct amd_iommu *iommu)
{
	iommu->acpi_flags & IVHD_FLAG_HT_TUN_EN_MASK ?
		iommu_feature_enable(iommu, CONTROL_HT_TUN_EN) :
		iommu_feature_disable(iommu, CONTROL_HT_TUN_EN);

	iommu->acpi_flags & IVHD_FLAG_PASSPW_EN_MASK ?
		iommu_feature_enable(iommu, CONTROL_PASSPW_EN) :
		iommu_feature_disable(iommu, CONTROL_PASSPW_EN);

	iommu->acpi_flags & IVHD_FLAG_RESPASSPW_EN_MASK ?
		iommu_feature_enable(iommu, CONTROL_RESPASSPW_EN) :
		iommu_feature_disable(iommu, CONTROL_RESPASSPW_EN);

	iommu->acpi_flags & IVHD_FLAG_ISOC_EN_MASK ?
		iommu_feature_enable(iommu, CONTROL_ISOC_EN) :
		iommu_feature_disable(iommu, CONTROL_ISOC_EN);

	/*
	 * make IOMMU memory accesses cache coherent
	 */
	iommu_feature_enable(iommu, CONTROL_COHERENT_EN);
}

1214
static void iommu_apply_resume_quirks(struct amd_iommu *iommu)
1215
{
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
	int i, j;
	u32 ioc_feature_control;
	struct pci_dev *pdev = NULL;

	/* RD890 BIOSes may not have completely reconfigured the iommu */
	if (!is_rd890_iommu(iommu->dev))
		return;

	/*
	 * First, we need to ensure that the iommu is enabled. This is
	 * controlled by a register in the northbridge
	 */
	pdev = pci_get_bus_and_slot(iommu->dev->bus->number, PCI_DEVFN(0, 0));

	if (!pdev)
		return;

	/* Select Northbridge indirect register 0x75 and enable writing */
	pci_write_config_dword(pdev, 0x60, 0x75 | (1 << 7));
	pci_read_config_dword(pdev, 0x64, &ioc_feature_control);

	/* Enable the iommu */
	if (!(ioc_feature_control & 0x1))
		pci_write_config_dword(pdev, 0x64, ioc_feature_control | 1);

	pci_dev_put(pdev);

	/* Restore the iommu BAR */
	pci_write_config_dword(iommu->dev, iommu->cap_ptr + 4,
			       iommu->stored_addr_lo);
	pci_write_config_dword(iommu->dev, iommu->cap_ptr + 8,
			       iommu->stored_addr_hi);

	/* Restore the l1 indirect regs for each of the 6 l1s */
	for (i = 0; i < 6; i++)
		for (j = 0; j < 0x12; j++)
			iommu_write_l1(iommu, i, j, iommu->stored_l1[i][j]);

	/* Restore the l2 indirect regs */
	for (i = 0; i < 0x83; i++)
		iommu_write_l2(iommu, i, iommu->stored_l2[i]);

	/* Lock PCI setup registers */
	pci_write_config_dword(iommu->dev, iommu->cap_ptr + 4,
			       iommu->stored_addr_lo | 1);
1261 1262
}

1263 1264 1265 1266
/*
 * This function finally enables all IOMMUs found in the system after
 * they have been initialized
 */
1267
static void enable_iommus(void)
1268 1269 1270
{
	struct amd_iommu *iommu;

1271
	for_each_iommu(iommu) {
1272
		iommu_disable(iommu);
1273
		iommu_init_flags(iommu);
1274 1275 1276
		iommu_set_device_table(iommu);
		iommu_enable_command_buffer(iommu);
		iommu_enable_event_buffer(iommu);
1277
		iommu_set_exclusion_range(iommu);
1278
		iommu_init_msi(iommu);
1279
		iommu_enable(iommu);
1280
		iommu_flush_all_caches(iommu);
1281 1282 1283
	}
}

1284 1285 1286 1287 1288 1289 1290 1291
static void disable_iommus(void)
{
	struct amd_iommu *iommu;

	for_each_iommu(iommu)
		iommu_disable(iommu);
}

1292 1293 1294 1295 1296
/*
 * Suspend/Resume support
 * disable suspend until real resume implemented
 */

1297
static void amd_iommu_resume(void)
1298
{
1299 1300 1301 1302 1303
	struct amd_iommu *iommu;

	for_each_iommu(iommu)
		iommu_apply_resume_quirks(iommu);

1304 1305 1306 1307 1308 1309 1310
	/* re-load the hardware */
	enable_iommus();

	/*
	 * we have to flush after the IOMMUs are enabled because a
	 * disabled IOMMU will never execute the commands we send
	 */
1311 1312
	for_each_iommu(iommu)
		iommu_flush_all_caches(iommu);
1313 1314
}

1315
static int amd_iommu_suspend(void)
1316
{
1317 1318 1319 1320
	/* disable IOMMUs to go out of the way for BIOS */
	disable_iommus();

	return 0;
1321 1322
}

1323
static struct syscore_ops amd_iommu_syscore_ops = {
1324 1325 1326 1327
	.suspend = amd_iommu_suspend,
	.resume = amd_iommu_resume,
};

1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
/*
 * This is the core init function for AMD IOMMU hardware in the system.
 * This function is called from the generic x86 DMA layer initialization
 * code.
 *
 * This function basically parses the ACPI table for AMD IOMMU (IVRS)
 * three times:
 *
 *	1 pass) Find the highest PCI device id the driver has to handle.
 *		Upon this information the size of the data structures is
 *		determined that needs to be allocated.
 *
 *	2 pass) Initialize the data structures just allocated with the
 *		information in the ACPI table about available AMD IOMMUs
 *		in the system. It also maps the PCI devices in the
 *		system to specific IOMMUs
 *
 *	3 pass) After the basic data structures are allocated and
 *		initialized we update them with information about memory
 *		remapping requirements parsed out of the ACPI table in
 *		this last pass.
 *
 * After that the hardware is initialized and ready to go. In the last
 * step we do some Linux specific things like registering the driver in
 * the dma_ops interface and initializing the suspend/resume support
 * functions. Finally it prints some information about AMD IOMMUs and
 * the driver state and enables the hardware.
 */
1356
static int __init amd_iommu_init(void)
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367
{
	int i, ret = 0;

	/*
	 * First parse ACPI tables to find the largest Bus/Dev/Func
	 * we need to handle. Upon this information the shared data
	 * structures for the IOMMUs in the system will be allocated
	 */
	if (acpi_table_parse("IVRS", find_last_devid_acpi) != 0)
		return -ENODEV;

1368 1369 1370 1371
	ret = amd_iommu_init_err;
	if (ret)
		goto out;

1372 1373 1374
	dev_table_size     = tbl_size(DEV_TABLE_ENTRY_SIZE);
	alias_table_size   = tbl_size(ALIAS_TABLE_ENTRY_SIZE);
	rlookup_table_size = tbl_size(RLOOKUP_TABLE_ENTRY_SIZE);
1375 1376 1377 1378

	ret = -ENOMEM;

	/* Device table - directly used by all IOMMUs */
1379
	amd_iommu_dev_table = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
				      get_order(dev_table_size));
	if (amd_iommu_dev_table == NULL)
		goto out;

	/*
	 * Alias table - map PCI Bus/Dev/Func to Bus/Dev/Func the
	 * IOMMU see for that device
	 */
	amd_iommu_alias_table = (void *)__get_free_pages(GFP_KERNEL,
			get_order(alias_table_size));
	if (amd_iommu_alias_table == NULL)
		goto free;

	/* IOMMU rlookup table - find the IOMMU for a specific device */
1394 1395
	amd_iommu_rlookup_table = (void *)__get_free_pages(
			GFP_KERNEL | __GFP_ZERO,
1396 1397 1398 1399
			get_order(rlookup_table_size));
	if (amd_iommu_rlookup_table == NULL)
		goto free;

1400 1401
	amd_iommu_pd_alloc_bitmap = (void *)__get_free_pages(
					    GFP_KERNEL | __GFP_ZERO,
1402 1403 1404 1405
					    get_order(MAX_DOMAIN_ID/8));
	if (amd_iommu_pd_alloc_bitmap == NULL)
		goto free;

1406 1407 1408
	/* init the device table */
	init_device_table();

1409
	/*
1410
	 * let all alias entries point to itself
1411
	 */
1412
	for (i = 0; i <= amd_iommu_last_bdf; ++i)
1413 1414 1415 1416 1417 1418 1419 1420
		amd_iommu_alias_table[i] = i;

	/*
	 * never allocate domain 0 because its used as the non-allocated and
	 * error value placeholder
	 */
	amd_iommu_pd_alloc_bitmap[0] = 1;

1421 1422
	spin_lock_init(&amd_iommu_pd_lock);

1423 1424 1425 1426 1427 1428 1429 1430
	/*
	 * now the data structures are allocated and basically initialized
	 * start the real acpi table scan
	 */
	ret = -ENODEV;
	if (acpi_table_parse("IVRS", init_iommu_all) != 0)
		goto free;

1431 1432
	if (amd_iommu_init_err) {
		ret = amd_iommu_init_err;
1433
		goto free;
1434
	}
1435

1436 1437 1438
	if (acpi_table_parse("IVRS", init_memory_definitions) != 0)
		goto free;

1439 1440 1441 1442 1443
	if (amd_iommu_init_err) {
		ret = amd_iommu_init_err;
		goto free;
	}

J
Joerg Roedel 已提交
1444 1445 1446 1447
	ret = amd_iommu_init_devices();
	if (ret)
		goto free;

1448 1449
	enable_iommus();

1450 1451 1452 1453
	if (iommu_pass_through)
		ret = amd_iommu_init_passthrough();
	else
		ret = amd_iommu_init_dma_ops();
1454

1455
	if (ret)
1456
		goto free_disable;
1457

1458 1459
	amd_iommu_init_api();

1460 1461
	amd_iommu_init_notifier();

1462 1463
	register_syscore_ops(&amd_iommu_syscore_ops);

1464 1465 1466
	if (iommu_pass_through)
		goto out;

1467
	if (amd_iommu_unmap_flush)
1468
		printk(KERN_INFO "AMD-Vi: IO/TLB flush on unmap enabled\n");
1469
	else
1470
		printk(KERN_INFO "AMD-Vi: Lazy IO/TLB flushing enabled\n");
1471

1472
	x86_platform.iommu_shutdown = disable_iommus;
1473 1474 1475
out:
	return ret;

1476
free_disable:
1477
	disable_iommus();
J
Joerg Roedel 已提交
1478

1479
free:
J
Joerg Roedel 已提交
1480 1481
	amd_iommu_uninit_devices();

1482 1483
	free_pages((unsigned long)amd_iommu_pd_alloc_bitmap,
		   get_order(MAX_DOMAIN_ID/8));
1484

1485 1486
	free_pages((unsigned long)amd_iommu_rlookup_table,
		   get_order(rlookup_table_size));
1487

1488 1489
	free_pages((unsigned long)amd_iommu_alias_table,
		   get_order(alias_table_size));
1490

1491 1492
	free_pages((unsigned long)amd_iommu_dev_table,
		   get_order(dev_table_size));
1493 1494 1495 1496 1497

	free_iommu_all();

	free_unity_maps();

1498 1499 1500 1501 1502 1503 1504 1505 1506
#ifdef CONFIG_GART_IOMMU
	/*
	 * We failed to initialize the AMD IOMMU - try fallback to GART
	 * if possible.
	 */
	gart_iommu_init();

#endif

1507 1508 1509
	goto out;
}

1510 1511 1512 1513 1514 1515 1516
/****************************************************************************
 *
 * Early detect code. This code runs at IOMMU detection time in the DMA
 * layer. It just looks if there is an IVRS ACPI table to detect AMD
 * IOMMUs
 *
 ****************************************************************************/
1517 1518 1519 1520 1521
static int __init early_amd_iommu_detect(struct acpi_table_header *table)
{
	return 0;
}

1522
int __init amd_iommu_detect(void)
1523
{
1524
	if (no_iommu || (iommu_detected && !gart_iommu_aperture))
1525
		return -ENODEV;
1526

1527
	if (amd_iommu_disabled)
1528
		return -ENODEV;
1529

1530 1531
	if (acpi_table_parse("IVRS", early_amd_iommu_detect) == 0) {
		iommu_detected = 1;
1532
		amd_iommu_detected = 1;
1533
		x86_init.iommu.iommu_init = amd_iommu_init;
1534

C
Chris Wright 已提交
1535 1536
		/* Make sure ACS will be enabled */
		pci_request_acs();
1537
		return 1;
1538
	}
1539
	return -ENODEV;
1540 1541
}

1542 1543 1544 1545 1546 1547 1548
/****************************************************************************
 *
 * Parsing functions for the AMD IOMMU specific kernel command line
 * options.
 *
 ****************************************************************************/

1549 1550 1551 1552 1553 1554 1555
static int __init parse_amd_iommu_dump(char *str)
{
	amd_iommu_dump = true;

	return 1;
}

1556 1557 1558
static int __init parse_amd_iommu_options(char *str)
{
	for (; *str; ++str) {
1559
		if (strncmp(str, "fullflush", 9) == 0)
1560
			amd_iommu_unmap_flush = true;
1561 1562
		if (strncmp(str, "off", 3) == 0)
			amd_iommu_disabled = true;
1563 1564 1565 1566 1567
	}

	return 1;
}

1568
__setup("amd_iommu_dump", parse_amd_iommu_dump);
1569
__setup("amd_iommu=", parse_amd_iommu_options);
1570 1571 1572 1573 1574

IOMMU_INIT_FINISH(amd_iommu_detect,
		  gart_iommu_hole_init,
		  0,
		  0);