amd_iommu_init.c 57.4 KB
Newer Older
1
/*
2
 * Copyright (C) 2007-2010 Advanced Micro Devices, Inc.
J
Joerg Roedel 已提交
3
 * Author: Joerg Roedel <jroedel@suse.de>
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 *         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 <linux/export.h>
A
Alex Williamson 已提交
29
#include <linux/iommu.h>
30
#include <asm/pci-direct.h>
31
#include <asm/iommu.h>
32
#include <asm/gart.h>
33
#include <asm/x86_init.h>
34
#include <asm/iommu_table.h>
35
#include <asm/io_apic.h>
36
#include <asm/irq_remapping.h>
37 38 39

#include "amd_iommu_proto.h"
#include "amd_iommu_types.h"
40
#include "irq_remapping.h"
41

42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
/*
 * 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
60 61 62 63
#define IVHD_DEV_SPECIAL		0x48

#define IVHD_SPECIAL_IOAPIC		1
#define IVHD_SPECIAL_HPET		2
64

J
Joerg Roedel 已提交
65 66 67 68
#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
69 70 71 72 73 74 75 76 77 78 79 80 81

#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

82 83 84 85 86 87 88 89 90 91 92
/*
 * 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.
 */
93 94 95 96 97 98 99 100 101
struct ivhd_header {
	u8 type;
	u8 flags;
	u16 length;
	u16 devid;
	u16 cap_ptr;
	u64 mmio_phys;
	u16 pci_seg;
	u16 info;
102 103 104 105 106
	u32 efr_attr;

	/* Following only valid on IVHD type 11h and 40h */
	u64 efr_reg; /* Exact copy of MMIO_EXT_FEATURES */
	u64 res;
107 108
} __attribute__((packed));

109 110 111 112
/*
 * A device entry describing which devices a specific IOMMU translates and
 * which requestor ids they use.
 */
113 114 115 116 117 118 119
struct ivhd_entry {
	u8 type;
	u16 devid;
	u8 flags;
	u32 ext;
} __attribute__((packed));

120 121 122 123
/*
 * An AMD IOMMU memory definition structure. It defines things like exclusion
 * ranges for devices and regions that should be unity mapped.
 */
124 125 126 127 128 129 130 131 132 133 134
struct ivmd_header {
	u8 type;
	u8 flags;
	u16 length;
	u16 devid;
	u16 aux;
	u64 resv;
	u64 range_start;
	u64 range_length;
} __attribute__((packed));

135
bool amd_iommu_dump;
136
bool amd_iommu_irq_remap __read_mostly;
137

138
static bool amd_iommu_detected;
139
static bool __initdata amd_iommu_disabled;
140

141 142
u16 amd_iommu_last_bdf;			/* largest PCI device id we have
					   to handle */
143
LIST_HEAD(amd_iommu_unity_map);		/* a list of required unity mappings
144
					   we find in ACPI */
145
bool amd_iommu_unmap_flush;		/* if true, flush on every unmap */
146

147
LIST_HEAD(amd_iommu_list);		/* list of all AMD IOMMUs in the
148
					   system */
149

150 151 152 153
/* Array to assign indices to IOMMUs*/
struct amd_iommu *amd_iommus[MAX_IOMMUS];
int amd_iommus_present;

154 155
/* IOMMUs have a non-present cache? */
bool amd_iommu_np_cache __read_mostly;
156
bool amd_iommu_iotlb_sup __read_mostly = true;
157

158
u32 amd_iommu_max_pasid __read_mostly = ~0;
159

160
bool amd_iommu_v2_present __read_mostly;
J
Joerg Roedel 已提交
161
static bool amd_iommu_pc_present __read_mostly;
162

163 164
bool amd_iommu_force_isolation __read_mostly;

165 166 167 168 169 170
/*
 * List of protection domains - used during resume
 */
LIST_HEAD(amd_iommu_pd_list);
spinlock_t amd_iommu_pd_lock;

171 172 173 174 175 176
/*
 * 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.
 */
177
struct dev_table_entry *amd_iommu_dev_table;
178 179 180 181 182 183

/*
 * 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.
 */
184
u16 *amd_iommu_alias_table;
185 186 187 188 189

/*
 * 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.
 */
190
struct amd_iommu **amd_iommu_rlookup_table;
191 192

/*
193 194 195 196 197
 * This table is used to find the irq remapping table for a given device id
 * quickly.
 */
struct irq_remap_table **irq_lookup_table;

198
/*
F
Frank Arnold 已提交
199
 * AMD IOMMU allows up to 2^16 different protection domains. This is a bitmap
200 201
 * to know which ones are already in use.
 */
202 203
unsigned long *amd_iommu_pd_alloc_bitmap;

204 205 206
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 */
207

208 209 210 211 212 213 214 215 216 217 218 219 220
enum iommu_init_state {
	IOMMU_START_STATE,
	IOMMU_IVRS_DETECTED,
	IOMMU_ACPI_FINISHED,
	IOMMU_ENABLED,
	IOMMU_PCI_INIT,
	IOMMU_INTERRUPTS_EN,
	IOMMU_DMA_OPS,
	IOMMU_INITIALIZED,
	IOMMU_NOT_FOUND,
	IOMMU_INIT_ERROR,
};

221 222 223 224 225 226
/* Early ioapic and hpet maps from kernel command line */
#define EARLY_MAP_SIZE		4
static struct devid_map __initdata early_ioapic_map[EARLY_MAP_SIZE];
static struct devid_map __initdata early_hpet_map[EARLY_MAP_SIZE];
static int __initdata early_ioapic_map_size;
static int __initdata early_hpet_map_size;
227
static bool __initdata cmdline_maps;
228

229 230
static enum iommu_init_state init_state = IOMMU_START_STATE;

231
static int amd_iommu_enable_interrupts(void);
232
static int __init iommu_go_to_state(enum iommu_init_state state);
233
static void init_device_table_dma(void);
234

235 236 237 238
static int iommu_pc_get_set_reg_val(struct amd_iommu *iommu,
				    u8 bank, u8 cntr, u8 fxn,
				    u64 *value, bool is_write);

239 240 241 242 243 244
static inline void update_last_devid(u16 devid)
{
	if (devid > amd_iommu_last_bdf)
		amd_iommu_last_bdf = devid;
}

245 246 247
static inline unsigned long tbl_size(int entry_size)
{
	unsigned shift = PAGE_SHIFT +
248
			 get_order(((int)amd_iommu_last_bdf + 1) * entry_size);
249 250 251 252

	return 1UL << shift;
}

253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
/* 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);
}

286 287 288 289 290 291 292 293
/****************************************************************************
 *
 * AMD IOMMU MMIO register space handling functions
 *
 * These functions are used to program the IOMMU device registers in
 * MMIO space required for that driver.
 *
 ****************************************************************************/
294

295 296 297 298
/*
 * This function set the exclusion range in the IOMMU. DMA accesses to the
 * exclusion range are passed through untranslated
 */
299
static void iommu_set_exclusion_range(struct amd_iommu *iommu)
300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
{
	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));
}

317
/* Programs the physical address of the device table into the IOMMU hardware */
318
static void iommu_set_device_table(struct amd_iommu *iommu)
319
{
320
	u64 entry;
321 322 323 324 325 326 327 328 329

	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));
}

330
/* Generic functions to enable/disable certain features of the IOMMU. */
331
static void iommu_feature_enable(struct amd_iommu *iommu, u8 bit)
332 333 334 335 336 337 338 339
{
	u32 ctrl;

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

340
static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
341 342 343
{
	u32 ctrl;

344
	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
345 346 347 348
	ctrl &= ~(1 << bit);
	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
}

349 350 351 352 353 354 355 356 357 358
static void iommu_set_inv_tlb_timeout(struct amd_iommu *iommu, int timeout)
{
	u32 ctrl;

	ctrl = readl(iommu->mmio_base + MMIO_CONTROL_OFFSET);
	ctrl &= ~CTRL_INV_TO_MASK;
	ctrl |= (timeout << CONTROL_INV_TIMEOUT) & CTRL_INV_TO_MASK;
	writel(ctrl, iommu->mmio_base + MMIO_CONTROL_OFFSET);
}

359
/* Function to enable the hardware */
360
static void iommu_enable(struct amd_iommu *iommu)
361 362 363 364
{
	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
}

365
static void iommu_disable(struct amd_iommu *iommu)
J
Joerg Roedel 已提交
366
{
367 368 369 370 371 372 373 374
	/* 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 */
375
	iommu_feature_disable(iommu, CONTROL_IOMMU_EN);
J
Joerg Roedel 已提交
376 377
}

378 379 380 381
/*
 * mapping and unmapping functions for the IOMMU MMIO space. Each AMD IOMMU in
 * the system has one.
 */
382
static u8 __iomem * __init iommu_map_mmio_space(u64 address, u64 end)
383
{
384 385 386
	if (!request_mem_region(address, end, "amd_iommu")) {
		pr_err("AMD-Vi: Can not reserve memory region %llx-%llx for mmio\n",
			address, end);
387
		pr_err("AMD-Vi: This is a BIOS bug. Please contact your hardware vendor\n");
388
		return NULL;
389
	}
390

391
	return (u8 __iomem *)ioremap_nocache(address, end);
392 393 394 395 396 397
}

static void __init iommu_unmap_mmio_space(struct amd_iommu *iommu)
{
	if (iommu->mmio_base)
		iounmap(iommu->mmio_base);
398
	release_mem_region(iommu->mmio_phys, iommu->mmio_phys_end);
399 400
}

401 402 403 404 405 406 407 408 409
/****************************************************************************
 *
 * 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.
 *
 ****************************************************************************/

410 411 412 413 414 415 416 417
/*
 * This function calculates the length of a given IVHD entry
 */
static inline int ivhd_entry_length(u8 *ivhd)
{
	return 0x04 << (*ivhd >> 6);
}

418 419 420 421
/*
 * 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
 */
422 423 424 425 426 427 428 429 430 431 432
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;

	while (p < end) {
		dev = (struct ivhd_entry *)p;
		switch (dev->type) {
433 434 435 436
		case IVHD_DEV_ALL:
			/* Use maximum BDF value for DEV_ALL */
			update_last_devid(0xffff);
			break;
437 438 439 440
		case IVHD_DEV_SELECT:
		case IVHD_DEV_RANGE_END:
		case IVHD_DEV_ALIAS:
		case IVHD_DEV_EXT_SELECT:
441
			/* all the above subfield types refer to device ids */
442
			update_last_devid(dev->devid);
443 444 445 446
			break;
		default:
			break;
		}
447
		p += ivhd_entry_length(p);
448 449 450 451 452 453 454
	}

	WARN_ON(p != end);

	return 0;
}

455 456 457 458 459
/*
 * 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.
 */
460 461 462 463 464 465 466 467 468 469 470 471
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];
472
	if (checksum != 0)
473
		/* ACPI table corrupt */
474
		return -ENODEV;
475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494

	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;
}

495 496
/****************************************************************************
 *
F
Frank Arnold 已提交
497
 * The following functions belong to the code path which parses the ACPI table
498 499 500 501 502 503 504 505 506 507 508
 * 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
 */
509
static int __init alloc_command_buffer(struct amd_iommu *iommu)
510
{
511 512
	iommu->cmd_buf = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
						  get_order(CMD_BUFFER_SIZE));
513

514
	return iommu->cmd_buf ? 0 : -ENOMEM;
515 516
}

517 518 519 520 521 522 523 524 525 526 527 528 529 530
/*
 * 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);
}

531 532 533 534 535 536 537 538 539 540 541
/*
 * 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);
542
	entry |= MMIO_CMD_SIZE_512;
543

544
	memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET,
545
		    &entry, sizeof(entry));
546

547
	amd_iommu_reset_cmd_buffer(iommu);
548 549 550 551
}

static void __init free_command_buffer(struct amd_iommu *iommu)
{
552
	free_pages((unsigned long)iommu->cmd_buf, get_order(CMD_BUFFER_SIZE));
553 554
}

555
/* allocates the memory where the IOMMU will log its events to */
556
static int __init alloc_event_buffer(struct amd_iommu *iommu)
557
{
558 559
	iommu->evt_buf = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
						  get_order(EVT_BUFFER_SIZE));
560

561
	return iommu->evt_buf ? 0 : -ENOMEM;
562 563 564 565 566 567 568 569
}

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

	BUG_ON(iommu->evt_buf == NULL);

570
	entry = (u64)virt_to_phys(iommu->evt_buf) | EVT_LEN_MASK;
571

572 573 574
	memcpy_toio(iommu->mmio_base + MMIO_EVT_BUF_OFFSET,
		    &entry, sizeof(entry));

575 576 577 578
	/* 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);

579
	iommu_feature_enable(iommu, CONTROL_EVT_LOG_EN);
580 581 582 583 584 585 586
}

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

587
/* allocates the memory where the IOMMU will log its events to */
588
static int __init alloc_ppr_log(struct amd_iommu *iommu)
589
{
590 591
	iommu->ppr_log = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
						  get_order(PPR_LOG_SIZE));
592

593
	return iommu->ppr_log ? 0 : -ENOMEM;
594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
}

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

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

	entry = (u64)virt_to_phys(iommu->ppr_log) | PPR_LOG_SIZE_512;

	memcpy_toio(iommu->mmio_base + MMIO_PPR_LOG_OFFSET,
		    &entry, sizeof(entry));

	/* set head and tail to zero manually */
	writel(0x00, iommu->mmio_base + MMIO_PPR_HEAD_OFFSET);
	writel(0x00, iommu->mmio_base + MMIO_PPR_TAIL_OFFSET);

	iommu_feature_enable(iommu, CONTROL_PPFLOG_EN);
	iommu_feature_enable(iommu, CONTROL_PPR_EN);
}

static void __init free_ppr_log(struct amd_iommu *iommu)
{
	if (iommu->ppr_log == NULL)
		return;

	free_pages((unsigned long)iommu->ppr_log, get_order(PPR_LOG_SIZE));
}

624 625 626 627 628 629 630 631
static void iommu_enable_gt(struct amd_iommu *iommu)
{
	if (!iommu_feature(iommu, FEATURE_GT))
		return;

	iommu_feature_enable(iommu, CONTROL_GT_EN);
}

632
/* sets a specific bit in the device table entry. */
633 634
static void set_dev_entry_bit(u16 devid, u8 bit)
{
635 636
	int i = (bit >> 6) & 0x03;
	int _bit = bit & 0x3f;
637

638
	amd_iommu_dev_table[devid].data[i] |= (1UL << _bit);
639 640
}

641 642
static int get_dev_entry_bit(u16 devid, u8 bit)
{
643 644
	int i = (bit >> 6) & 0x03;
	int _bit = bit & 0x3f;
645

646
	return (amd_iommu_dev_table[devid].data[i] & (1UL << _bit)) >> _bit;
647 648 649 650 651 652 653 654 655 656 657 658 659 660
}


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);
}

661 662 663 664 665 666
/* 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;
}

667 668 669 670
/*
 * This function takes the device specific flags read from the ACPI
 * table and sets up the device table entry with that information
 */
671 672
static void __init set_dev_entry_from_acpi(struct amd_iommu *iommu,
					   u16 devid, u32 flags, u32 ext_flags)
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
{
	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);

689 690
	amd_iommu_apply_erratum_63(devid);

691
	set_iommu_for_device(iommu, devid);
692 693
}

694
static int __init add_special_device(u8 type, u8 id, u16 *devid, bool cmd_line)
695 696 697 698
{
	struct devid_map *entry;
	struct list_head *list;

699 700 701 702 703
	if (type == IVHD_SPECIAL_IOAPIC)
		list = &ioapic_map;
	else if (type == IVHD_SPECIAL_HPET)
		list = &hpet_map;
	else
704 705
		return -EINVAL;

706 707 708 709 710 711 712
	list_for_each_entry(entry, list, list) {
		if (!(entry->id == id && entry->cmd_line))
			continue;

		pr_info("AMD-Vi: Command-line override present for %s id %d - ignoring\n",
			type == IVHD_SPECIAL_IOAPIC ? "IOAPIC" : "HPET", id);

713 714
		*devid = entry->devid;

715 716 717
		return 0;
	}

718 719 720 721
	entry = kzalloc(sizeof(*entry), GFP_KERNEL);
	if (!entry)
		return -ENOMEM;

722
	entry->id	= id;
723
	entry->devid	= *devid;
724
	entry->cmd_line	= cmd_line;
725 726 727 728 729 730

	list_add_tail(&entry->list, list);

	return 0;
}

731 732 733 734 735 736 737
static int __init add_early_maps(void)
{
	int i, ret;

	for (i = 0; i < early_ioapic_map_size; ++i) {
		ret = add_special_device(IVHD_SPECIAL_IOAPIC,
					 early_ioapic_map[i].id,
738
					 &early_ioapic_map[i].devid,
739 740 741 742 743 744 745 746
					 early_ioapic_map[i].cmd_line);
		if (ret)
			return ret;
	}

	for (i = 0; i < early_hpet_map_size; ++i) {
		ret = add_special_device(IVHD_SPECIAL_HPET,
					 early_hpet_map[i].id,
747
					 &early_hpet_map[i].devid,
748 749 750 751 752 753 754 755
					 early_hpet_map[i].cmd_line);
		if (ret)
			return ret;
	}

	return 0;
}

756
/*
F
Frank Arnold 已提交
757
 * Reads the device exclusion range from ACPI and initializes the IOMMU with
758 759
 * it
 */
760 761 762 763 764 765 766 767
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) {
768 769 770 771 772
		/*
		 * We only can configure exclusion ranges per IOMMU, not
		 * per device. But we can enable the exclusion range per
		 * device. This is done here
		 */
773
		set_dev_entry_bit(devid, DEV_ENTRY_EX);
774 775 776 777 778
		iommu->exclusion_start = m->range_start;
		iommu->exclusion_length = m->range_length;
	}
}

779 780 781 782
/*
 * Takes a pointer to an AMD IOMMU entry in the ACPI table and
 * initializes the hardware and our data structures with it.
 */
783
static int __init init_iommu_from_acpi(struct amd_iommu *iommu,
784 785 786 787
					struct ivhd_header *h)
{
	u8 *p = (u8 *)h;
	u8 *end = p, flags = 0;
788 789
	u16 devid = 0, devid_start = 0, devid_to = 0;
	u32 dev_i, ext_flags = 0;
790
	bool alias = false;
791
	struct ivhd_entry *e;
792 793 794 795 796 797
	int ret;


	ret = add_early_maps();
	if (ret)
		return ret;
798 799

	/*
800
	 * First save the recommended feature enable bits from ACPI
801
	 */
802
	iommu->acpi_flags = h->flags;
803 804 805 806 807 808 809

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

810

811 812 813 814
	while (p < end) {
		e = (struct ivhd_entry *)p;
		switch (e->type) {
		case IVHD_DEV_ALL:
815

816
			DUMP_printk("  DEV_ALL\t\t\tflags: %02x\n", e->flags);
817

818 819
			for (dev_i = 0; dev_i <= amd_iommu_last_bdf; ++dev_i)
				set_dev_entry_from_acpi(iommu, dev_i, e->flags, 0);
820 821
			break;
		case IVHD_DEV_SELECT:
822 823 824

			DUMP_printk("  DEV_SELECT\t\t\t devid: %02x:%02x.%x "
				    "flags: %02x\n",
825
				    PCI_BUS_NUM(e->devid),
826 827 828 829
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags);

830
			devid = e->devid;
831
			set_dev_entry_from_acpi(iommu, devid, e->flags, 0);
832 833
			break;
		case IVHD_DEV_SELECT_RANGE_START:
834 835 836

			DUMP_printk("  DEV_SELECT_RANGE_START\t "
				    "devid: %02x:%02x.%x flags: %02x\n",
837
				    PCI_BUS_NUM(e->devid),
838 839 840 841
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags);

842 843 844
			devid_start = e->devid;
			flags = e->flags;
			ext_flags = 0;
845
			alias = false;
846 847
			break;
		case IVHD_DEV_ALIAS:
848 849 850

			DUMP_printk("  DEV_ALIAS\t\t\t devid: %02x:%02x.%x "
				    "flags: %02x devid_to: %02x:%02x.%x\n",
851
				    PCI_BUS_NUM(e->devid),
852 853 854
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags,
855
				    PCI_BUS_NUM(e->ext >> 8),
856 857 858
				    PCI_SLOT(e->ext >> 8),
				    PCI_FUNC(e->ext >> 8));

859 860
			devid = e->devid;
			devid_to = e->ext >> 8;
861
			set_dev_entry_from_acpi(iommu, devid   , e->flags, 0);
862
			set_dev_entry_from_acpi(iommu, devid_to, e->flags, 0);
863 864 865
			amd_iommu_alias_table[devid] = devid_to;
			break;
		case IVHD_DEV_ALIAS_RANGE:
866 867 868 869

			DUMP_printk("  DEV_ALIAS_RANGE\t\t "
				    "devid: %02x:%02x.%x flags: %02x "
				    "devid_to: %02x:%02x.%x\n",
870
				    PCI_BUS_NUM(e->devid),
871 872 873
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags,
874
				    PCI_BUS_NUM(e->ext >> 8),
875 876 877
				    PCI_SLOT(e->ext >> 8),
				    PCI_FUNC(e->ext >> 8));

878 879 880 881
			devid_start = e->devid;
			flags = e->flags;
			devid_to = e->ext >> 8;
			ext_flags = 0;
882
			alias = true;
883 884
			break;
		case IVHD_DEV_EXT_SELECT:
885 886 887

			DUMP_printk("  DEV_EXT_SELECT\t\t devid: %02x:%02x.%x "
				    "flags: %02x ext: %08x\n",
888
				    PCI_BUS_NUM(e->devid),
889 890 891 892
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags, e->ext);

893
			devid = e->devid;
894 895
			set_dev_entry_from_acpi(iommu, devid, e->flags,
						e->ext);
896 897
			break;
		case IVHD_DEV_EXT_SELECT_RANGE:
898 899 900

			DUMP_printk("  DEV_EXT_SELECT_RANGE\t devid: "
				    "%02x:%02x.%x flags: %02x ext: %08x\n",
901
				    PCI_BUS_NUM(e->devid),
902 903 904 905
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid),
				    e->flags, e->ext);

906 907 908
			devid_start = e->devid;
			flags = e->flags;
			ext_flags = e->ext;
909
			alias = false;
910 911
			break;
		case IVHD_DEV_RANGE_END:
912 913

			DUMP_printk("  DEV_RANGE_END\t\t devid: %02x:%02x.%x\n",
914
				    PCI_BUS_NUM(e->devid),
915 916 917
				    PCI_SLOT(e->devid),
				    PCI_FUNC(e->devid));

918 919
			devid = e->devid;
			for (dev_i = devid_start; dev_i <= devid; ++dev_i) {
920
				if (alias) {
921
					amd_iommu_alias_table[dev_i] = devid_to;
922 923 924 925 926
					set_dev_entry_from_acpi(iommu,
						devid_to, flags, ext_flags);
				}
				set_dev_entry_from_acpi(iommu, dev_i,
							flags, ext_flags);
927 928
			}
			break;
929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947
		case IVHD_DEV_SPECIAL: {
			u8 handle, type;
			const char *var;
			u16 devid;
			int ret;

			handle = e->ext & 0xff;
			devid  = (e->ext >>  8) & 0xffff;
			type   = (e->ext >> 24) & 0xff;

			if (type == IVHD_SPECIAL_IOAPIC)
				var = "IOAPIC";
			else if (type == IVHD_SPECIAL_HPET)
				var = "HPET";
			else
				var = "UNKNOWN";

			DUMP_printk("  DEV_SPECIAL(%s[%d])\t\tdevid: %02x:%02x.%x\n",
				    var, (int)handle,
948
				    PCI_BUS_NUM(devid),
949 950 951
				    PCI_SLOT(devid),
				    PCI_FUNC(devid));

952
			ret = add_special_device(type, handle, &devid, false);
953 954
			if (ret)
				return ret;
955 956 957 958 959 960 961 962

			/*
			 * add_special_device might update the devid in case a
			 * command-line override is present. So call
			 * set_dev_entry_from_acpi after add_special_device.
			 */
			set_dev_entry_from_acpi(iommu, devid, e->flags, 0);

963 964
			break;
		}
965 966 967 968
		default:
			break;
		}

969
		p += ivhd_entry_length(p);
970
	}
971 972

	return 0;
973 974
}

975 976 977
static void __init free_iommu_one(struct amd_iommu *iommu)
{
	free_command_buffer(iommu);
978
	free_event_buffer(iommu);
979
	free_ppr_log(iommu);
980 981 982 983 984 985 986
	iommu_unmap_mmio_space(iommu);
}

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

987
	for_each_iommu_safe(iommu, next) {
988 989 990 991 992 993
		list_del(&iommu->list);
		free_iommu_one(iommu);
		kfree(iommu);
	}
}

994 995 996 997 998 999
/*
 * Family15h Model 10h-1fh erratum 746 (IOMMU Logging May Stall Translations)
 * Workaround:
 *     BIOS should disable L2B micellaneous clock gating by setting
 *     L2_L2B_CK_GATE_CONTROL[CKGateL2BMiscDisable](D0F2xF4_x90[2]) = 1b
 */
1000
static void amd_iommu_erratum_746_workaround(struct amd_iommu *iommu)
1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
{
	u32 value;

	if ((boot_cpu_data.x86 != 0x15) ||
	    (boot_cpu_data.x86_model < 0x10) ||
	    (boot_cpu_data.x86_model > 0x1f))
		return;

	pci_write_config_dword(iommu->dev, 0xf0, 0x90);
	pci_read_config_dword(iommu->dev, 0xf4, &value);

	if (value & BIT(2))
		return;

	/* Select NB indirect register 0x90 and enable writing */
	pci_write_config_dword(iommu->dev, 0xf0, 0x90 | (1 << 8));

	pci_write_config_dword(iommu->dev, 0xf4, value | 0x4);
	pr_info("AMD-Vi: Applying erratum 746 workaround for IOMMU at %s\n",
		dev_name(&iommu->dev->dev));

	/* Clear the enable writing bit */
	pci_write_config_dword(iommu->dev, 0xf0, 0x90);
}

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
/*
 * Family15h Model 30h-3fh (IOMMU Mishandles ATS Write Permission)
 * Workaround:
 *     BIOS should enable ATS write permission check by setting
 *     L2_DEBUG_3[AtsIgnoreIWDis](D0F2xF4_x47[0]) = 1b
 */
static void amd_iommu_ats_write_check_workaround(struct amd_iommu *iommu)
{
	u32 value;

	if ((boot_cpu_data.x86 != 0x15) ||
	    (boot_cpu_data.x86_model < 0x30) ||
	    (boot_cpu_data.x86_model > 0x3f))
		return;

	/* Test L2_DEBUG_3[AtsIgnoreIWDis] == 1 */
	value = iommu_read_l2(iommu, 0x47);

	if (value & BIT(0))
		return;

	/* Set L2_DEBUG_3[AtsIgnoreIWDis] = 1 */
	iommu_write_l2(iommu, 0x47, value | BIT(0));

	pr_info("AMD-Vi: Applying ATS write check workaround for IOMMU at %s\n",
		dev_name(&iommu->dev->dev));
}

1054 1055 1056 1057 1058
/*
 * 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.
 */
1059 1060
static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h)
{
1061 1062
	int ret;

1063
	spin_lock_init(&iommu->lock);
1064 1065

	/* Add IOMMU to internal data structures */
1066
	list_add_tail(&iommu->list, &amd_iommu_list);
1067 1068 1069 1070 1071 1072 1073 1074 1075
	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;
1076 1077 1078 1079

	/*
	 * Copy data from ACPI table entry to the iommu struct
	 */
1080
	iommu->devid   = h->devid;
1081
	iommu->cap_ptr = h->cap_ptr;
1082
	iommu->pci_seg = h->pci_seg;
1083
	iommu->mmio_phys = h->mmio_phys;
1084

1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
	switch (h->type) {
	case 0x10:
		/* Check if IVHD EFR contains proper max banks/counters */
		if ((h->efr_attr != 0) &&
		    ((h->efr_attr & (0xF << 13)) != 0) &&
		    ((h->efr_attr & (0x3F << 17)) != 0))
			iommu->mmio_phys_end = MMIO_REG_END_OFFSET;
		else
			iommu->mmio_phys_end = MMIO_CNTR_CONF_OFFSET;
		break;
	case 0x11:
	case 0x40:
		if (h->efr_reg & (1 << 9))
			iommu->mmio_phys_end = MMIO_REG_END_OFFSET;
		else
			iommu->mmio_phys_end = MMIO_CNTR_CONF_OFFSET;
		break;
	default:
		return -EINVAL;
1104 1105 1106 1107
	}

	iommu->mmio_base = iommu_map_mmio_space(iommu->mmio_phys,
						iommu->mmio_phys_end);
1108 1109 1110
	if (!iommu->mmio_base)
		return -ENOMEM;

1111
	if (alloc_command_buffer(iommu))
1112 1113
		return -ENOMEM;

1114
	if (alloc_event_buffer(iommu))
1115 1116
		return -ENOMEM;

1117 1118
	iommu->int_enabled = false;

1119 1120 1121
	ret = init_iommu_from_acpi(iommu, h);
	if (ret)
		return ret;
1122

1123 1124 1125 1126
	ret = amd_iommu_create_irq_domain(iommu);
	if (ret)
		return ret;

1127 1128 1129 1130 1131 1132
	/*
	 * Make sure IOMMU is not considered to translate itself. The IVRS
	 * table tells us so, but this is a lie!
	 */
	amd_iommu_rlookup_table[iommu->devid] = NULL;

1133
	return 0;
1134 1135
}

1136 1137 1138 1139
/*
 * Iterates over all IOMMU entries in the ACPI table, allocates the
 * IOMMU structure and initializes it with init_iommu_one()
 */
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
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:
1154

1155
			DUMP_printk("device: %02x:%02x.%01x cap: %04x "
1156
				    "seg: %d flags: %01x info %04x\n",
1157
				    PCI_BUS_NUM(h->devid), PCI_SLOT(h->devid),
1158 1159 1160 1161 1162
				    PCI_FUNC(h->devid), h->cap_ptr,
				    h->pci_seg, h->flags, h->info);
			DUMP_printk("       mmio-addr: %016llx\n",
				    h->mmio_phys);

1163
			iommu = kzalloc(sizeof(struct amd_iommu), GFP_KERNEL);
1164 1165
			if (iommu == NULL)
				return -ENOMEM;
1166

1167
			ret = init_iommu_one(iommu, h);
1168 1169
			if (ret)
				return ret;
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181
			break;
		default:
			break;
		}
		p += h->length;

	}
	WARN_ON(p != end);

	return 0;
}

1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192

static void init_iommu_perf_ctr(struct amd_iommu *iommu)
{
	u64 val = 0xabcd, val2 = 0;

	if (!iommu_feature(iommu, FEATURE_PC))
		return;

	amd_iommu_pc_present = true;

	/* Check if the performance counters can be written to */
1193 1194
	if ((0 != iommu_pc_get_set_reg_val(iommu, 0, 0, 0, &val, true)) ||
	    (0 != iommu_pc_get_set_reg_val(iommu, 0, 0, 0, &val2, false)) ||
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207
	    (val != val2)) {
		pr_err("AMD-Vi: Unable to write to IOMMU perf counter.\n");
		amd_iommu_pc_present = false;
		return;
	}

	pr_info("AMD-Vi: IOMMU performance counters supported\n");

	val = readl(iommu->mmio_base + MMIO_CNTR_CONF_OFFSET);
	iommu->max_banks = (u8) ((val >> 12) & 0x3f);
	iommu->max_counters = (u8) ((val >> 7) & 0xf);
}

A
Alex Williamson 已提交
1208 1209 1210 1211 1212 1213 1214 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
static ssize_t amd_iommu_show_cap(struct device *dev,
				  struct device_attribute *attr,
				  char *buf)
{
	struct amd_iommu *iommu = dev_get_drvdata(dev);
	return sprintf(buf, "%x\n", iommu->cap);
}
static DEVICE_ATTR(cap, S_IRUGO, amd_iommu_show_cap, NULL);

static ssize_t amd_iommu_show_features(struct device *dev,
				       struct device_attribute *attr,
				       char *buf)
{
	struct amd_iommu *iommu = dev_get_drvdata(dev);
	return sprintf(buf, "%llx\n", iommu->features);
}
static DEVICE_ATTR(features, S_IRUGO, amd_iommu_show_features, NULL);

static struct attribute *amd_iommu_attrs[] = {
	&dev_attr_cap.attr,
	&dev_attr_features.attr,
	NULL,
};

static struct attribute_group amd_iommu_group = {
	.name = "amd-iommu",
	.attrs = amd_iommu_attrs,
};

static const struct attribute_group *amd_iommu_groups[] = {
	&amd_iommu_group,
	NULL,
};
1241

1242 1243 1244 1245 1246
static int iommu_init_pci(struct amd_iommu *iommu)
{
	int cap_ptr = iommu->cap_ptr;
	u32 range, misc, low, high;

1247
	iommu->dev = pci_get_bus_and_slot(PCI_BUS_NUM(iommu->devid),
1248 1249 1250 1251
					  iommu->devid & 0xff);
	if (!iommu->dev)
		return -ENODEV;

1252 1253 1254
	/* Prevent binding other PCI device drivers to IOMMU devices */
	iommu->dev->match_driver = false;

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
	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);
	pci_read_config_dword(iommu->dev, cap_ptr + MMIO_MISC_OFFSET,
			      &misc);

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

	/* 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;

	if (iommu_feature(iommu, FEATURE_GT)) {
		int glxval;
1273 1274
		u32 max_pasid;
		u64 pasmax;
1275

1276 1277 1278
		pasmax = iommu->features & FEATURE_PASID_MASK;
		pasmax >>= FEATURE_PASID_SHIFT;
		max_pasid  = (1 << (pasmax + 1)) - 1;
1279

1280 1281 1282
		amd_iommu_max_pasid = min(amd_iommu_max_pasid, max_pasid);

		BUG_ON(amd_iommu_max_pasid & ~PASID_MASK);
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298

		glxval   = iommu->features & FEATURE_GLXVAL_MASK;
		glxval >>= FEATURE_GLXVAL_SHIFT;

		if (amd_iommu_max_glx_val == -1)
			amd_iommu_max_glx_val = glxval;
		else
			amd_iommu_max_glx_val = min(amd_iommu_max_glx_val, glxval);
	}

	if (iommu_feature(iommu, FEATURE_GT) &&
	    iommu_feature(iommu, FEATURE_PPR)) {
		iommu->is_iommu_v2   = true;
		amd_iommu_v2_present = true;
	}

1299 1300
	if (iommu_feature(iommu, FEATURE_PPR) && alloc_ppr_log(iommu))
		return -ENOMEM;
1301 1302 1303 1304

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

1305 1306
	init_iommu_perf_ctr(iommu);

1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
	if (is_rd890_iommu(iommu->dev)) {
		int i, j;

		iommu->root_pdev = pci_get_bus_and_slot(iommu->dev->bus->number,
				PCI_DEVFN(0, 0));

		/*
		 * 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);
	}

1334
	amd_iommu_erratum_746_workaround(iommu);
1335
	amd_iommu_ats_write_check_workaround(iommu);
1336

A
Alex Williamson 已提交
1337 1338 1339 1340
	iommu->iommu_dev = iommu_device_create(&iommu->dev->dev, iommu,
					       amd_iommu_groups, "ivhd%d",
					       iommu->index);

1341 1342 1343
	return pci_enable_device(iommu->dev);
}

1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
static void print_iommu_info(void)
{
	static const char * const feat_str[] = {
		"PreF", "PPR", "X2APIC", "NX", "GT", "[5]",
		"IA", "GA", "HE", "PC"
	};
	struct amd_iommu *iommu;

	for_each_iommu(iommu) {
		int i;

		pr_info("AMD-Vi: Found IOMMU at %s cap 0x%hx\n",
			dev_name(&iommu->dev->dev), iommu->cap_ptr);

		if (iommu->cap & (1 << IOMMU_CAP_EFR)) {
			pr_info("AMD-Vi:  Extended features: ");
1360
			for (i = 0; i < ARRAY_SIZE(feat_str); ++i) {
1361 1362 1363
				if (iommu_feature(iommu, (1ULL << i)))
					pr_cont(" %s", feat_str[i]);
			}
1364
			pr_cont("\n");
1365
		}
1366
	}
1367 1368
	if (irq_remapping_enabled)
		pr_info("AMD-Vi: Interrupt remapping enabled\n");
1369 1370
}

1371
static int __init amd_iommu_init_pci(void)
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
{
	struct amd_iommu *iommu;
	int ret = 0;

	for_each_iommu(iommu) {
		ret = iommu_init_pci(iommu);
		if (ret)
			break;
	}

1382 1383 1384 1385 1386
	init_device_table_dma();

	for_each_iommu(iommu)
		iommu_flush_all_caches(iommu);

1387
	ret = amd_iommu_init_api();
1388

1389 1390
	if (!ret)
		print_iommu_info();
1391

1392 1393 1394
	return ret;
}

1395 1396 1397
/****************************************************************************
 *
 * The following functions initialize the MSI interrupts for all IOMMUs
F
Frank Arnold 已提交
1398
 * in the system. It's a bit challenging because there could be multiple
1399 1400 1401 1402 1403
 * IOMMUs per PCI BDF but we can call pci_enable_msi(x) only once per
 * pci_dev.
 *
 ****************************************************************************/

1404
static int iommu_setup_msi(struct amd_iommu *iommu)
1405 1406 1407
{
	int r;

1408 1409 1410
	r = pci_enable_msi(iommu->dev);
	if (r)
		return r;
1411

1412 1413 1414 1415
	r = request_threaded_irq(iommu->dev->irq,
				 amd_iommu_int_handler,
				 amd_iommu_int_thread,
				 0, "AMD-Vi",
1416
				 iommu);
1417 1418 1419

	if (r) {
		pci_disable_msi(iommu->dev);
1420
		return r;
1421 1422
	}

1423
	iommu->int_enabled = true;
1424

1425 1426 1427
	return 0;
}

1428
static int iommu_init_msi(struct amd_iommu *iommu)
1429
{
1430 1431
	int ret;

1432
	if (iommu->int_enabled)
1433
		goto enable_faults;
1434

1435
	if (iommu->dev->msi_cap)
1436 1437 1438 1439 1440 1441
		ret = iommu_setup_msi(iommu);
	else
		ret = -ENODEV;

	if (ret)
		return ret;
1442

1443 1444
enable_faults:
	iommu_feature_enable(iommu, CONTROL_EVT_INT_EN);
1445

1446 1447 1448 1449
	if (iommu->ppr_log != NULL)
		iommu_feature_enable(iommu, CONTROL_PPFINT_EN);

	return 0;
1450 1451
}

1452 1453 1454 1455
/****************************************************************************
 *
 * The next functions belong to the third pass of parsing the ACPI
 * table. In this last pass the memory mapping requirements are
F
Frank Arnold 已提交
1456
 * gathered (like exclusion and unity mapping ranges).
1457 1458 1459
 *
 ****************************************************************************/

1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
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);
	}
}

1470
/* called when we find an exclusion range definition in ACPI */
1471 1472 1473 1474 1475 1476 1477 1478 1479
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:
1480
		for (i = 0; i <= amd_iommu_last_bdf; ++i)
1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493
			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;
}

1494
/* called for unity map ACPI definition */
1495 1496
static int __init init_unity_map_range(struct ivmd_header *m)
{
J
Joerg Roedel 已提交
1497
	struct unity_map_entry *e = NULL;
1498
	char *s;
1499 1500 1501 1502 1503 1504 1505

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

	switch (m->type) {
	default:
1506 1507
		kfree(e);
		return 0;
1508
	case ACPI_IVMD_TYPE:
1509
		s = "IVMD_TYPEi\t\t\t";
1510 1511 1512
		e->devid_start = e->devid_end = m->devid;
		break;
	case ACPI_IVMD_TYPE_ALL:
1513
		s = "IVMD_TYPE_ALL\t\t";
1514 1515 1516 1517
		e->devid_start = 0;
		e->devid_end = amd_iommu_last_bdf;
		break;
	case ACPI_IVMD_TYPE_RANGE:
1518
		s = "IVMD_TYPE_RANGE\t\t";
1519 1520 1521 1522 1523 1524 1525 1526
		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;

1527 1528
	DUMP_printk("%s devid_start: %02x:%02x.%x devid_end: %02x:%02x.%x"
		    " range_start: %016llx range_end: %016llx flags: %x\n", s,
1529 1530
		    PCI_BUS_NUM(e->devid_start), PCI_SLOT(e->devid_start),
		    PCI_FUNC(e->devid_start), PCI_BUS_NUM(e->devid_end),
1531 1532 1533
		    PCI_SLOT(e->devid_end), PCI_FUNC(e->devid_end),
		    e->address_start, e->address_end, m->flags);

1534 1535 1536 1537 1538
	list_add_tail(&e->list, &amd_iommu_unity_map);

	return 0;
}

1539
/* iterates over all memory definitions we find in the ACPI table */
1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
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;
}

1561 1562 1563 1564
/*
 * Init the device table to not allow DMA access for devices and
 * suppress all page faults
 */
1565
static void init_device_table_dma(void)
1566
{
1567
	u32 devid;
1568 1569 1570 1571 1572 1573 1574

	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);
	}
}

1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
static void __init uninit_device_table_dma(void)
{
	u32 devid;

	for (devid = 0; devid <= amd_iommu_last_bdf; ++devid) {
		amd_iommu_dev_table[devid].data[0] = 0ULL;
		amd_iommu_dev_table[devid].data[1] = 0ULL;
	}
}

1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
static void init_device_table(void)
{
	u32 devid;

	if (!amd_iommu_irq_remap)
		return;

	for (devid = 0; devid <= amd_iommu_last_bdf; ++devid)
		set_dev_entry_bit(devid, DEV_ENTRY_IRQ_TBL_EN);
}

1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
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);
1618 1619 1620

	/* Set IOTLB invalidation timeout to 1s */
	iommu_set_inv_tlb_timeout(iommu, CTRL_INV_TO_1S);
1621 1622
}

1623
static void iommu_apply_resume_quirks(struct amd_iommu *iommu)
1624
{
1625 1626
	int i, j;
	u32 ioc_feature_control;
1627
	struct pci_dev *pdev = iommu->root_pdev;
1628 1629

	/* RD890 BIOSes may not have completely reconfigured the iommu */
1630
	if (!is_rd890_iommu(iommu->dev) || !pdev)
1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
		return;

	/*
	 * First, we need to ensure that the iommu is enabled. This is
	 * controlled by a register in the northbridge
	 */

	/* 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);

	/* 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);
1664 1665
}

1666 1667 1668 1669
/*
 * This function finally enables all IOMMUs found in the system after
 * they have been initialized
 */
1670
static void early_enable_iommus(void)
1671 1672 1673
{
	struct amd_iommu *iommu;

1674
	for_each_iommu(iommu) {
1675
		iommu_disable(iommu);
1676
		iommu_init_flags(iommu);
1677 1678 1679
		iommu_set_device_table(iommu);
		iommu_enable_command_buffer(iommu);
		iommu_enable_event_buffer(iommu);
1680 1681
		iommu_set_exclusion_range(iommu);
		iommu_enable(iommu);
1682
		iommu_flush_all_caches(iommu);
1683 1684 1685
	}
}

1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
static void enable_iommus_v2(void)
{
	struct amd_iommu *iommu;

	for_each_iommu(iommu) {
		iommu_enable_ppr_log(iommu);
		iommu_enable_gt(iommu);
	}
}

static void enable_iommus(void)
{
	early_enable_iommus();

	enable_iommus_v2();
}

1703 1704 1705 1706 1707 1708 1709 1710
static void disable_iommus(void)
{
	struct amd_iommu *iommu;

	for_each_iommu(iommu)
		iommu_disable(iommu);
}

1711 1712 1713 1714 1715
/*
 * Suspend/Resume support
 * disable suspend until real resume implemented
 */

1716
static void amd_iommu_resume(void)
1717
{
1718 1719 1720 1721 1722
	struct amd_iommu *iommu;

	for_each_iommu(iommu)
		iommu_apply_resume_quirks(iommu);

1723 1724
	/* re-load the hardware */
	enable_iommus();
1725 1726

	amd_iommu_enable_interrupts();
1727 1728
}

1729
static int amd_iommu_suspend(void)
1730
{
1731 1732 1733 1734
	/* disable IOMMUs to go out of the way for BIOS */
	disable_iommus();

	return 0;
1735 1736
}

1737
static struct syscore_ops amd_iommu_syscore_ops = {
1738 1739 1740 1741
	.suspend = amd_iommu_suspend,
	.resume = amd_iommu_resume,
};

1742 1743
static void __init free_on_init_error(void)
{
1744 1745
	free_pages((unsigned long)irq_lookup_table,
		   get_order(rlookup_table_size));
1746

1747 1748
	kmem_cache_destroy(amd_iommu_irq_cache);
	amd_iommu_irq_cache = NULL;
1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770

	free_pages((unsigned long)amd_iommu_rlookup_table,
		   get_order(rlookup_table_size));

	free_pages((unsigned long)amd_iommu_alias_table,
		   get_order(alias_table_size));

	free_pages((unsigned long)amd_iommu_dev_table,
		   get_order(dev_table_size));

	free_iommu_all();

#ifdef CONFIG_GART_IOMMU
	/*
	 * We failed to initialize the AMD IOMMU - try fallback to GART
	 * if possible.
	 */
	gart_iommu_init();

#endif
}

1771 1772 1773
/* SB IOAPIC is always on this device in AMD systems */
#define IOAPIC_SB_DEVID		((0x00 << 8) | PCI_DEVFN(0x14, 0))

1774 1775
static bool __init check_ioapic_information(void)
{
1776
	const char *fw_bug = FW_BUG;
1777
	bool ret, has_sb_ioapic;
1778 1779
	int idx;

1780 1781
	has_sb_ioapic = false;
	ret           = false;
1782

1783 1784 1785 1786 1787 1788 1789 1790
	/*
	 * If we have map overrides on the kernel command line the
	 * messages in this function might not describe firmware bugs
	 * anymore - so be careful
	 */
	if (cmdline_maps)
		fw_bug = "";

1791 1792 1793 1794 1795
	for (idx = 0; idx < nr_ioapics; idx++) {
		int devid, id = mpc_ioapic_id(idx);

		devid = get_ioapic_devid(id);
		if (devid < 0) {
1796 1797
			pr_err("%sAMD-Vi: IOAPIC[%d] not in IVRS table\n",
				fw_bug, id);
1798 1799 1800 1801
			ret = false;
		} else if (devid == IOAPIC_SB_DEVID) {
			has_sb_ioapic = true;
			ret           = true;
1802 1803 1804
		}
	}

1805 1806 1807 1808 1809 1810 1811 1812 1813
	if (!has_sb_ioapic) {
		/*
		 * We expect the SB IOAPIC to be listed in the IVRS
		 * table. The system timer is connected to the SB IOAPIC
		 * and if we don't have it in the list the system will
		 * panic at boot time.  This situation usually happens
		 * when the BIOS is buggy and provides us the wrong
		 * device id for the IOAPIC in the system.
		 */
1814
		pr_err("%sAMD-Vi: No southbridge IOAPIC found\n", fw_bug);
1815 1816 1817
	}

	if (!ret)
1818
		pr_err("AMD-Vi: Disabling interrupt remapping\n");
1819 1820

	return ret;
1821 1822
}

1823 1824 1825 1826 1827 1828 1829 1830
static void __init free_dma_resources(void)
{
	free_pages((unsigned long)amd_iommu_pd_alloc_bitmap,
		   get_order(MAX_DOMAIN_ID/8));

	free_unity_maps();
}

1831
/*
1832 1833 1834
 * This is the hardware init function for AMD IOMMU in the system.
 * This function is called either from amd_iommu_init or from the interrupt
 * remapping setup code.
1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
 *
 * 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.
 *
1853 1854
 * After everything is set up the IOMMUs are enabled and the necessary
 * hotplug and suspend notifiers are registered.
1855
 */
1856
static int __init early_amd_iommu_init(void)
1857
{
1858 1859 1860
	struct acpi_table_header *ivrs_base;
	acpi_size ivrs_size;
	acpi_status status;
1861 1862
	int i, ret = 0;

1863
	if (!amd_iommu_detected)
1864 1865
		return -ENODEV;

1866 1867 1868 1869 1870 1871 1872 1873 1874
	status = acpi_get_table_with_size("IVRS", 0, &ivrs_base, &ivrs_size);
	if (status == AE_NOT_FOUND)
		return -ENODEV;
	else if (ACPI_FAILURE(status)) {
		const char *err = acpi_format_exception(status);
		pr_err("AMD-Vi: IVRS table error: %s\n", err);
		return -EINVAL;
	}

1875 1876 1877 1878 1879
	/*
	 * 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
	 */
1880 1881
	ret = find_last_devid_acpi(ivrs_base);
	if (ret)
1882 1883
		goto out;

1884 1885 1886
	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);
1887 1888

	/* Device table - directly used by all IOMMUs */
1889
	ret = -ENOMEM;
1890
	amd_iommu_dev_table = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901
				      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)
1902
		goto out;
1903 1904

	/* IOMMU rlookup table - find the IOMMU for a specific device */
1905 1906
	amd_iommu_rlookup_table = (void *)__get_free_pages(
			GFP_KERNEL | __GFP_ZERO,
1907 1908
			get_order(rlookup_table_size));
	if (amd_iommu_rlookup_table == NULL)
1909
		goto out;
1910

1911 1912
	amd_iommu_pd_alloc_bitmap = (void *)__get_free_pages(
					    GFP_KERNEL | __GFP_ZERO,
1913 1914
					    get_order(MAX_DOMAIN_ID/8));
	if (amd_iommu_pd_alloc_bitmap == NULL)
1915
		goto out;
1916 1917

	/*
1918
	 * let all alias entries point to itself
1919
	 */
1920
	for (i = 0; i <= amd_iommu_last_bdf; ++i)
1921 1922 1923 1924 1925 1926 1927 1928
		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;

1929 1930
	spin_lock_init(&amd_iommu_pd_lock);

1931 1932 1933 1934
	/*
	 * now the data structures are allocated and basically initialized
	 * start the real acpi table scan
	 */
1935 1936
	ret = init_iommu_all(ivrs_base);
	if (ret)
1937
		goto out;
1938

1939 1940 1941
	if (amd_iommu_irq_remap)
		amd_iommu_irq_remap = check_ioapic_information();

1942 1943 1944 1945 1946
	if (amd_iommu_irq_remap) {
		/*
		 * Interrupt remapping enabled, create kmem_cache for the
		 * remapping tables.
		 */
1947
		ret = -ENOMEM;
1948 1949 1950 1951 1952 1953
		amd_iommu_irq_cache = kmem_cache_create("irq_remap_cache",
				MAX_IRQS_PER_TABLE * sizeof(u32),
				IRQ_TABLE_ALIGNMENT,
				0, NULL);
		if (!amd_iommu_irq_cache)
			goto out;
1954 1955 1956 1957 1958 1959

		irq_lookup_table = (void *)__get_free_pages(
				GFP_KERNEL | __GFP_ZERO,
				get_order(rlookup_table_size));
		if (!irq_lookup_table)
			goto out;
1960 1961
	}

1962 1963
	ret = init_memory_definitions(ivrs_base);
	if (ret)
1964
		goto out;
1965

1966 1967 1968
	/* init the device table */
	init_device_table();

1969
out:
1970 1971 1972 1973
	/* Don't leak any ACPI memory */
	early_acpi_os_unmap_memory((char __iomem *)ivrs_base, ivrs_size);
	ivrs_base = NULL;

1974 1975 1976
	return ret;
}

1977
static int amd_iommu_enable_interrupts(void)
1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
{
	struct amd_iommu *iommu;
	int ret = 0;

	for_each_iommu(iommu) {
		ret = iommu_init_msi(iommu);
		if (ret)
			goto out;
	}

out:
	return ret;
}

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
static bool detect_ivrs(void)
{
	struct acpi_table_header *ivrs_base;
	acpi_size ivrs_size;
	acpi_status status;

	status = acpi_get_table_with_size("IVRS", 0, &ivrs_base, &ivrs_size);
	if (status == AE_NOT_FOUND)
		return false;
	else if (ACPI_FAILURE(status)) {
		const char *err = acpi_format_exception(status);
		pr_err("AMD-Vi: IVRS table error: %s\n", err);
		return false;
	}

	early_acpi_os_unmap_memory((char __iomem *)ivrs_base, ivrs_size);

2009 2010 2011
	/* Make sure ACS will be enabled during PCI probe */
	pci_request_acs();

2012 2013 2014
	return true;
}

2015
/****************************************************************************
2016
 *
2017 2018 2019 2020 2021
 * AMD IOMMU Initialization State Machine
 *
 ****************************************************************************/

static int __init state_next(void)
2022 2023 2024
{
	int ret = 0;

2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
	switch (init_state) {
	case IOMMU_START_STATE:
		if (!detect_ivrs()) {
			init_state	= IOMMU_NOT_FOUND;
			ret		= -ENODEV;
		} else {
			init_state	= IOMMU_IVRS_DETECTED;
		}
		break;
	case IOMMU_IVRS_DETECTED:
		ret = early_amd_iommu_init();
		init_state = ret ? IOMMU_INIT_ERROR : IOMMU_ACPI_FINISHED;
		break;
	case IOMMU_ACPI_FINISHED:
		early_enable_iommus();
		register_syscore_ops(&amd_iommu_syscore_ops);
		x86_platform.iommu_shutdown = disable_iommus;
		init_state = IOMMU_ENABLED;
		break;
	case IOMMU_ENABLED:
		ret = amd_iommu_init_pci();
		init_state = ret ? IOMMU_INIT_ERROR : IOMMU_PCI_INIT;
		enable_iommus_v2();
		break;
	case IOMMU_PCI_INIT:
		ret = amd_iommu_enable_interrupts();
		init_state = ret ? IOMMU_INIT_ERROR : IOMMU_INTERRUPTS_EN;
		break;
	case IOMMU_INTERRUPTS_EN:
2054
		ret = amd_iommu_init_dma_ops();
2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
		init_state = ret ? IOMMU_INIT_ERROR : IOMMU_DMA_OPS;
		break;
	case IOMMU_DMA_OPS:
		init_state = IOMMU_INITIALIZED;
		break;
	case IOMMU_INITIALIZED:
		/* Nothing to do */
		break;
	case IOMMU_NOT_FOUND:
	case IOMMU_INIT_ERROR:
		/* Error states => do nothing */
		ret = -EINVAL;
		break;
	default:
		/* Unknown state */
		BUG();
	}
2072

2073 2074
	return ret;
}
2075

2076 2077 2078
static int __init iommu_go_to_state(enum iommu_init_state state)
{
	int ret = 0;
2079

2080 2081 2082 2083 2084 2085
	while (init_state != state) {
		ret = state_next();
		if (init_state == IOMMU_NOT_FOUND ||
		    init_state == IOMMU_INIT_ERROR)
			break;
	}
2086

2087
	return ret;
2088
}
2089

2090 2091 2092
#ifdef CONFIG_IRQ_REMAP
int __init amd_iommu_prepare(void)
{
2093 2094
	int ret;

2095
	amd_iommu_irq_remap = true;
2096

2097 2098 2099 2100
	ret = iommu_go_to_state(IOMMU_ACPI_FINISHED);
	if (ret)
		return ret;
	return amd_iommu_irq_remap ? 0 : -ENODEV;
2101
}
2102

2103 2104 2105 2106 2107 2108 2109
int __init amd_iommu_enable(void)
{
	int ret;

	ret = iommu_go_to_state(IOMMU_ENABLED);
	if (ret)
		return ret;
2110

2111
	irq_remapping_enabled = 1;
2112

2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
	return 0;
}

void amd_iommu_disable(void)
{
	amd_iommu_suspend();
}

int amd_iommu_reenable(int mode)
{
	amd_iommu_resume();

	return 0;
}
2127

2128 2129 2130 2131 2132 2133
int __init amd_iommu_enable_faulting(void)
{
	/* We enable MSI later when PCI is initialized */
	return 0;
}
#endif
2134

2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145
/*
 * 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.
 */
static int __init amd_iommu_init(void)
{
	int ret;

	ret = iommu_go_to_state(IOMMU_INITIALIZED);
	if (ret) {
2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156
		free_dma_resources();
		if (!irq_remapping_enabled) {
			disable_iommus();
			free_on_init_error();
		} else {
			struct amd_iommu *iommu;

			uninit_device_table_dma();
			for_each_iommu(iommu)
				iommu_flush_all_caches(iommu);
		}
2157 2158 2159
	}

	return ret;
2160 2161
}

2162 2163 2164 2165 2166 2167 2168
/****************************************************************************
 *
 * 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
 *
 ****************************************************************************/
2169
int __init amd_iommu_detect(void)
2170
{
2171
	int ret;
2172

2173
	if (no_iommu || (iommu_detected && !gart_iommu_aperture))
2174
		return -ENODEV;
2175

2176
	if (amd_iommu_disabled)
2177
		return -ENODEV;
2178

2179 2180 2181
	ret = iommu_go_to_state(IOMMU_IVRS_DETECTED);
	if (ret)
		return ret;
2182

2183 2184 2185 2186
	amd_iommu_detected = true;
	iommu_detected = 1;
	x86_init.iommu.iommu_init = amd_iommu_init;

2187
	return 1;
2188 2189
}

2190 2191 2192 2193 2194 2195 2196
/****************************************************************************
 *
 * Parsing functions for the AMD IOMMU specific kernel command line
 * options.
 *
 ****************************************************************************/

2197 2198 2199 2200 2201 2202 2203
static int __init parse_amd_iommu_dump(char *str)
{
	amd_iommu_dump = true;

	return 1;
}

2204 2205 2206
static int __init parse_amd_iommu_options(char *str)
{
	for (; *str; ++str) {
2207
		if (strncmp(str, "fullflush", 9) == 0)
2208
			amd_iommu_unmap_flush = true;
2209 2210
		if (strncmp(str, "off", 3) == 0)
			amd_iommu_disabled = true;
2211 2212
		if (strncmp(str, "force_isolation", 15) == 0)
			amd_iommu_force_isolation = true;
2213 2214 2215 2216 2217
	}

	return 1;
}

2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238
static int __init parse_ivrs_ioapic(char *str)
{
	unsigned int bus, dev, fn;
	int ret, id, i;
	u16 devid;

	ret = sscanf(str, "[%d]=%x:%x.%x", &id, &bus, &dev, &fn);

	if (ret != 4) {
		pr_err("AMD-Vi: Invalid command line: ivrs_ioapic%s\n", str);
		return 1;
	}

	if (early_ioapic_map_size == EARLY_MAP_SIZE) {
		pr_err("AMD-Vi: Early IOAPIC map overflow - ignoring ivrs_ioapic%s\n",
			str);
		return 1;
	}

	devid = ((bus & 0xff) << 8) | ((dev & 0x1f) << 3) | (fn & 0x7);

2239
	cmdline_maps			= true;
2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268
	i				= early_ioapic_map_size++;
	early_ioapic_map[i].id		= id;
	early_ioapic_map[i].devid	= devid;
	early_ioapic_map[i].cmd_line	= true;

	return 1;
}

static int __init parse_ivrs_hpet(char *str)
{
	unsigned int bus, dev, fn;
	int ret, id, i;
	u16 devid;

	ret = sscanf(str, "[%d]=%x:%x.%x", &id, &bus, &dev, &fn);

	if (ret != 4) {
		pr_err("AMD-Vi: Invalid command line: ivrs_hpet%s\n", str);
		return 1;
	}

	if (early_hpet_map_size == EARLY_MAP_SIZE) {
		pr_err("AMD-Vi: Early HPET map overflow - ignoring ivrs_hpet%s\n",
			str);
		return 1;
	}

	devid = ((bus & 0xff) << 8) | ((dev & 0x1f) << 3) | (fn & 0x7);

2269
	cmdline_maps			= true;
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
	i				= early_hpet_map_size++;
	early_hpet_map[i].id		= id;
	early_hpet_map[i].devid		= devid;
	early_hpet_map[i].cmd_line	= true;

	return 1;
}

__setup("amd_iommu_dump",	parse_amd_iommu_dump);
__setup("amd_iommu=",		parse_amd_iommu_options);
__setup("ivrs_ioapic",		parse_ivrs_ioapic);
__setup("ivrs_hpet",		parse_ivrs_hpet);
2282 2283 2284

IOMMU_INIT_FINISH(amd_iommu_detect,
		  gart_iommu_hole_init,
J
Joerg Roedel 已提交
2285 2286
		  NULL,
		  NULL);
2287 2288 2289 2290 2291 2292

bool amd_iommu_v2_supported(void)
{
	return amd_iommu_v2_present;
}
EXPORT_SYMBOL(amd_iommu_v2_supported);
2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334

/****************************************************************************
 *
 * IOMMU EFR Performance Counter support functionality. This code allows
 * access to the IOMMU PC functionality.
 *
 ****************************************************************************/

u8 amd_iommu_pc_get_max_banks(u16 devid)
{
	struct amd_iommu *iommu;
	u8 ret = 0;

	/* locate the iommu governing the devid */
	iommu = amd_iommu_rlookup_table[devid];
	if (iommu)
		ret = iommu->max_banks;

	return ret;
}
EXPORT_SYMBOL(amd_iommu_pc_get_max_banks);

bool amd_iommu_pc_supported(void)
{
	return amd_iommu_pc_present;
}
EXPORT_SYMBOL(amd_iommu_pc_supported);

u8 amd_iommu_pc_get_max_counters(u16 devid)
{
	struct amd_iommu *iommu;
	u8 ret = 0;

	/* locate the iommu governing the devid */
	iommu = amd_iommu_rlookup_table[devid];
	if (iommu)
		ret = iommu->max_counters;

	return ret;
}
EXPORT_SYMBOL(amd_iommu_pc_get_max_counters);

2335 2336
static int iommu_pc_get_set_reg_val(struct amd_iommu *iommu,
				    u8 bank, u8 cntr, u8 fxn,
2337 2338 2339 2340 2341 2342
				    u64 *value, bool is_write)
{
	u32 offset;
	u32 max_offset_lim;

	/* Check for valid iommu and pc register indexing */
2343
	if (WARN_ON((fxn > 0x28) || (fxn & 7)))
2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366
		return -ENODEV;

	offset = (u32)(((0x40|bank) << 12) | (cntr << 8) | fxn);

	/* Limit the offset to the hw defined mmio region aperture */
	max_offset_lim = (u32)(((0x40|iommu->max_banks) << 12) |
				(iommu->max_counters << 8) | 0x28);
	if ((offset < MMIO_CNTR_REG_OFFSET) ||
	    (offset > max_offset_lim))
		return -EINVAL;

	if (is_write) {
		writel((u32)*value, iommu->mmio_base + offset);
		writel((*value >> 32), iommu->mmio_base + offset + 4);
	} else {
		*value = readl(iommu->mmio_base + offset + 4);
		*value <<= 32;
		*value = readl(iommu->mmio_base + offset);
	}

	return 0;
}
EXPORT_SYMBOL(amd_iommu_pc_get_set_reg_val);
2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379

int amd_iommu_pc_get_set_reg_val(u16 devid, u8 bank, u8 cntr, u8 fxn,
				    u64 *value, bool is_write)
{
	struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];

	/* Make sure the IOMMU PC resource is available */
	if (!amd_iommu_pc_present || iommu == NULL)
		return -ENODEV;

	return iommu_pc_get_set_reg_val(iommu, bank, cntr, fxn,
					value, is_write);
}