pat.c 24.9 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 * Handle caching attributes in page tables (PAT)
 *
 * Authors: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
 *          Suresh B Siddha <suresh.b.siddha@intel.com>
 *
 * Loosely based on earlier PAT patchset from Eric Biederman and Andi Kleen.
 */

I
Ingo Molnar 已提交
10 11 12
#include <linux/seq_file.h>
#include <linux/bootmem.h>
#include <linux/debugfs.h>
13
#include <linux/kernel.h>
14
#include <linux/module.h>
15
#include <linux/gfp.h>
I
Ingo Molnar 已提交
16
#include <linux/mm.h>
17
#include <linux/fs.h>
18
#include <linux/rbtree.h>
19

I
Ingo Molnar 已提交
20
#include <asm/cacheflush.h>
21
#include <asm/processor.h>
I
Ingo Molnar 已提交
22
#include <asm/tlbflush.h>
23
#include <asm/x86_init.h>
24 25
#include <asm/pgtable.h>
#include <asm/fcntl.h>
I
Ingo Molnar 已提交
26
#include <asm/e820.h>
27
#include <asm/mtrr.h>
I
Ingo Molnar 已提交
28 29 30
#include <asm/page.h>
#include <asm/msr.h>
#include <asm/pat.h>
31
#include <asm/io.h>
32

33
#ifdef CONFIG_X86_PAT
34
int __read_mostly pat_enabled = 1;
35

36
static inline void pat_disable(const char *reason)
37
{
38
	pat_enabled = 0;
39
	printk(KERN_INFO "%s\n", reason);
40 41
}

A
Andrew Morton 已提交
42
static int __init nopat(char *str)
43
{
44
	pat_disable("PAT support disabled.");
45 46
	return 0;
}
47
early_param("nopat", nopat);
48 49 50 51 52
#else
static inline void pat_disable(const char *reason)
{
	(void)reason;
}
53 54
#endif

55 56

static int debug_enable;
I
Ingo Molnar 已提交
57

58 59 60 61 62 63 64 65 66 67 68
static int __init pat_debug_setup(char *str)
{
	debug_enable = 1;
	return 0;
}
__setup("debugpat", pat_debug_setup);

#define dprintk(fmt, arg...) \
	do { if (debug_enable) printk(KERN_INFO fmt, ##arg); } while (0)


69
static u64 __read_mostly boot_pat_state;
70 71 72 73 74 75 76 77 78 79

enum {
	PAT_UC = 0,		/* uncached */
	PAT_WC = 1,		/* Write combining */
	PAT_WT = 4,		/* Write Through */
	PAT_WP = 5,		/* Write Protected */
	PAT_WB = 6,		/* Write Back (default) */
	PAT_UC_MINUS = 7,	/* UC, but can be overriden by MTRR */
};

80
#define PAT(x, y)	((u64)PAT_ ## y << ((x)*8))
81 82 83 84

void pat_init(void)
{
	u64 pat;
85
	bool boot_cpu = !boot_pat_state;
86

87
	if (!pat_enabled)
88 89
		return;

90 91 92 93 94 95 96 97 98 99 100 101 102 103
	if (!cpu_has_pat) {
		if (!boot_pat_state) {
			pat_disable("PAT not supported by CPU.");
			return;
		} else {
			/*
			 * If this happens we are on a secondary CPU, but
			 * switched to PAT on the boot CPU. We have no way to
			 * undo PAT.
			 */
			printk(KERN_ERR "PAT enabled, "
			       "but not supported by secondary CPU\n");
			BUG();
		}
104
	}
105 106 107 108 109 110 111 112 113 114 115 116 117 118

	/* Set PWT to Write-Combining. All other bits stay the same */
	/*
	 * PTE encoding used in Linux:
	 *      PAT
	 *      |PCD
	 *      ||PWT
	 *      |||
	 *      000 WB		_PAGE_CACHE_WB
	 *      001 WC		_PAGE_CACHE_WC
	 *      010 UC-		_PAGE_CACHE_UC_MINUS
	 *      011 UC		_PAGE_CACHE_UC
	 * PAT bit unused
	 */
119 120
	pat = PAT(0, WB) | PAT(1, WC) | PAT(2, UC_MINUS) | PAT(3, UC) |
	      PAT(4, WB) | PAT(5, WC) | PAT(6, UC_MINUS) | PAT(7, UC);
121 122

	/* Boot CPU check */
123
	if (!boot_pat_state)
124 125 126
		rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);

	wrmsrl(MSR_IA32_CR_PAT, pat);
127 128 129 130

	if (boot_cpu)
		printk(KERN_INFO "x86 PAT enabled: cpu %d, old 0x%Lx, new 0x%Lx\n",
		       smp_processor_id(), boot_pat_state, pat);
131 132 133 134 135 136 137
}

#undef PAT

static char *cattr_name(unsigned long flags)
{
	switch (flags & _PAGE_CACHE_MASK) {
138 139 140 141 142
	case _PAGE_CACHE_UC:		return "uncached";
	case _PAGE_CACHE_UC_MINUS:	return "uncached-minus";
	case _PAGE_CACHE_WB:		return "write-back";
	case _PAGE_CACHE_WC:		return "write-combining";
	default:			return "broken";
143 144 145 146 147 148 149 150 151 152 153 154 155
	}
}

/*
 * The global memtype list keeps track of memory type for specific
 * physical memory areas. Conflicting memory types in different
 * mappings can cause CPU cache corruption. To avoid this we keep track.
 *
 * The list is sorted based on starting address and can contain multiple
 * entries for each address (this allows reference counting for overlapping
 * areas). All the aliases have the same cache attributes of course.
 * Zero attributes are represented as holes.
 *
156 157
 * The data structure is a list that is also organized as an rbtree
 * sorted on the start address of memtype range.
158
 *
159
 * memtype_lock protects both the linear list and rbtree.
160 161 162
 */

struct memtype {
I
Ingo Molnar 已提交
163 164 165 166
	u64			start;
	u64			end;
	unsigned long		type;
	struct list_head	nd;
167
	struct rb_node		rb;
168 169
};

170
static struct rb_root memtype_rbroot = RB_ROOT;
171
static LIST_HEAD(memtype_list);
I
Ingo Molnar 已提交
172
static DEFINE_SPINLOCK(memtype_lock);	/* protects memtype list */
173

174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
static struct memtype *memtype_rb_search(struct rb_root *root, u64 start)
{
	struct rb_node *node = root->rb_node;
	struct memtype *last_lower = NULL;

	while (node) {
		struct memtype *data = container_of(node, struct memtype, rb);

		if (data->start < start) {
			last_lower = data;
			node = node->rb_right;
		} else if (data->start > start) {
			node = node->rb_left;
		} else
			return data;
	}

	/* Will return NULL if there is no entry with its start <= start */
	return last_lower;
}

static void memtype_rb_insert(struct rb_root *root, struct memtype *data)
{
	struct rb_node **new = &(root->rb_node);
	struct rb_node *parent = NULL;

	while (*new) {
		struct memtype *this = container_of(*new, struct memtype, rb);

		parent = *new;
		if (data->start <= this->start)
			new = &((*new)->rb_left);
		else if (data->start > this->start)
			new = &((*new)->rb_right);
	}

	rb_link_node(&data->rb, parent, new);
	rb_insert_color(&data->rb, root);
}

214 215 216 217 218 219 220
/*
 * Does intersection of PAT memory type and MTRR memory type and returns
 * the resulting memory type as PAT understands it.
 * (Type in pat and mtrr will not have same value)
 * The intersection is based on "Effective Memory Type" tables in IA-32
 * SDM vol 3a
 */
221
static unsigned long pat_x_mtrr_type(u64 start, u64 end, unsigned long req_type)
222
{
223 224 225 226
	/*
	 * Look for MTRR hint to get the effective type in case where PAT
	 * request is for WB.
	 */
227 228 229 230
	if (req_type == _PAGE_CACHE_WB) {
		u8 mtrr_type;

		mtrr_type = mtrr_type_lookup(start, end);
231 232 233 234
		if (mtrr_type != MTRR_TYPE_WRBACK)
			return _PAGE_CACHE_UC_MINUS;

		return _PAGE_CACHE_WB;
235 236 237
	}

	return req_type;
238 239
}

I
Ingo Molnar 已提交
240 241
static int
chk_conflict(struct memtype *new, struct memtype *entry, unsigned long *type)
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
{
	if (new->type != entry->type) {
		if (type) {
			new->type = entry->type;
			*type = entry->type;
		} else
			goto conflict;
	}

	 /* check overlaps with more than one entry in the list */
	list_for_each_entry_continue(entry, &memtype_list, nd) {
		if (new->end <= entry->start)
			break;
		else if (new->type != entry->type)
			goto conflict;
	}
	return 0;

 conflict:
	printk(KERN_INFO "%s:%d conflicting memory types "
	       "%Lx-%Lx %s<->%s\n", current->comm, current->pid, new->start,
	       new->end, cattr_name(new->type), cattr_name(entry->type));
	return -EBUSY;
}

267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
static int pat_pagerange_is_ram(unsigned long start, unsigned long end)
{
	int ram_page = 0, not_rampage = 0;
	unsigned long page_nr;

	for (page_nr = (start >> PAGE_SHIFT); page_nr < (end >> PAGE_SHIFT);
	     ++page_nr) {
		/*
		 * For legacy reasons, physical address range in the legacy ISA
		 * region is tracked as non-RAM. This will allow users of
		 * /dev/mem to map portions of legacy ISA region, even when
		 * some of those portions are listed(or not even listed) with
		 * different e820 types(RAM/reserved/..)
		 */
		if (page_nr >= (ISA_END_ADDRESS >> PAGE_SHIFT) &&
		    page_is_ram(page_nr))
			ram_page = 1;
		else
			not_rampage = 1;

		if (ram_page == not_rampage)
			return -1;
	}

	return ram_page;
}

294
/*
295 296 297 298
 * For RAM pages, we use page flags to mark the pages with appropriate type.
 * Here we do two pass:
 * - Find the memtype of all the pages in the range, look for any conflicts
 * - In case of no conflicts, set the new memtype for pages in the range
299
 *
300
 * Caller must hold memtype_lock for atomicity.
301 302
 */
static int reserve_ram_pages_type(u64 start, u64 end, unsigned long req_type,
I
Ingo Molnar 已提交
303
				  unsigned long *new_type)
304 305
{
	struct page *page;
306 307 308 309 310 311 312
	u64 pfn;

	if (req_type == _PAGE_CACHE_UC) {
		/* We do not support strong UC */
		WARN_ON_ONCE(1);
		req_type = _PAGE_CACHE_UC_MINUS;
	}
313 314

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
315
		unsigned long type;
316

317 318 319 320 321 322 323 324 325 326 327
		page = pfn_to_page(pfn);
		type = get_page_memtype(page);
		if (type != -1) {
			printk(KERN_INFO "reserve_ram_pages_type failed "
				"0x%Lx-0x%Lx, track 0x%lx, req 0x%lx\n",
				start, end, type, req_type);
			if (new_type)
				*new_type = type;

			return -EBUSY;
		}
328 329
	}

330 331 332 333
	if (new_type)
		*new_type = req_type;

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
334
		page = pfn_to_page(pfn);
335
		set_page_memtype(page, req_type);
336
	}
337
	return 0;
338 339 340 341 342
}

static int free_ram_pages_type(u64 start, u64 end)
{
	struct page *page;
343
	u64 pfn;
344 345 346

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
		page = pfn_to_page(pfn);
347
		set_page_memtype(page, -1);
348 349 350 351
	}
	return 0;
}

352 353 354 355 356 357 358 359 360 361
/*
 * req_type typically has one of the:
 * - _PAGE_CACHE_WB
 * - _PAGE_CACHE_WC
 * - _PAGE_CACHE_UC_MINUS
 * - _PAGE_CACHE_UC
 *
 * req_type will have a special case value '-1', when requester want to inherit
 * the memory type from mtrr (if WB), existing PAT, defaulting to UC_MINUS.
 *
362 363 364
 * If new_type is NULL, function will return an error if it cannot reserve the
 * region with req_type. If new_type is non-NULL, function will return
 * available type in new_type in case of no error. In case of any error
365 366
 * it will return a negative return value.
 */
367
int reserve_memtype(u64 start, u64 end, unsigned long req_type,
I
Ingo Molnar 已提交
368
		    unsigned long *new_type)
369
{
370
	struct memtype *new, *entry;
371
	unsigned long actual_type;
372
	struct list_head *where;
373
	int is_range_ram;
I
Ingo Molnar 已提交
374
	int err = 0;
375

I
Ingo Molnar 已提交
376
	BUG_ON(start >= end); /* end is exclusive */
377

378
	if (!pat_enabled) {
379
		/* This is identical to page table setting without PAT */
380 381 382
		if (new_type) {
			if (req_type == -1)
				*new_type = _PAGE_CACHE_WB;
383 384
			else if (req_type == _PAGE_CACHE_WC)
				*new_type = _PAGE_CACHE_UC_MINUS;
385 386
			else
				*new_type = req_type & _PAGE_CACHE_MASK;
387
		}
388 389 390 391
		return 0;
	}

	/* Low ISA region is always mapped WB in page table. No need to track */
392
	if (x86_platform.is_untracked_pat_range(start, end)) {
393 394
		if (new_type)
			*new_type = _PAGE_CACHE_WB;
395 396 397
		return 0;
	}

398 399 400 401 402 403 404
	/*
	 * Call mtrr_lookup to get the type hint. This is an
	 * optimization for /dev/mem mmap'ers into WB memory (BIOS
	 * tools and ACPI tools). Use WB request for WB memory and use
	 * UC_MINUS otherwise.
	 */
	actual_type = pat_x_mtrr_type(start, end, req_type & _PAGE_CACHE_MASK);
405

406 407 408
	if (new_type)
		*new_type = actual_type;

409
	is_range_ram = pat_pagerange_is_ram(start, end);
410 411 412 413 414 415 416 417
	if (is_range_ram == 1) {

		spin_lock(&memtype_lock);
		err = reserve_ram_pages_type(start, end, req_type, new_type);
		spin_unlock(&memtype_lock);

		return err;
	} else if (is_range_ram < 0) {
418
		return -EINVAL;
419
	}
420

421 422
	new  = kmalloc(sizeof(struct memtype), GFP_KERNEL);
	if (!new)
423 424
		return -ENOMEM;

I
Ingo Molnar 已提交
425 426 427
	new->start	= start;
	new->end	= end;
	new->type	= actual_type;
428 429 430 431

	spin_lock(&memtype_lock);

	/* Search for existing mapping that overlaps the current range */
432
	where = NULL;
433
	list_for_each_entry(entry, &memtype_list, nd) {
434
		if (end <= entry->start) {
435
			where = entry->nd.prev;
436
			break;
437
		} else if (start <= entry->start) { /* end > entry->start */
438
			err = chk_conflict(new, entry, new_type);
439 440 441 442
			if (!err) {
				dprintk("Overlap at 0x%Lx-0x%Lx\n",
					entry->start, entry->end);
				where = entry->nd.prev;
443 444
			}
			break;
445
		} else if (start < entry->end) { /* start > entry->start */
446
			err = chk_conflict(new, entry, new_type);
447 448 449
			if (!err) {
				dprintk("Overlap at 0x%Lx-0x%Lx\n",
					entry->start, entry->end);
450 451 452 453 454 455 456 457 458 459 460 461

				/*
				 * Move to right position in the linked
				 * list to add this new entry
				 */
				list_for_each_entry_continue(entry,
							&memtype_list, nd) {
					if (start <= entry->start) {
						where = entry->nd.prev;
						break;
					}
				}
462 463 464 465 466 467
			}
			break;
		}
	}

	if (err) {
468 469 470
		printk(KERN_INFO "reserve_memtype failed 0x%Lx-0x%Lx, "
		       "track %s, req %s\n",
		       start, end, cattr_name(new->type), cattr_name(req_type));
471
		kfree(new);
472
		spin_unlock(&memtype_lock);
I
Ingo Molnar 已提交
473

474 475 476
		return err;
	}

477 478 479
	if (where)
		list_add(&new->nd, where);
	else
480
		list_add_tail(&new->nd, &memtype_list);
481

482 483
	memtype_rb_insert(&memtype_rbroot, new);

484
	spin_unlock(&memtype_lock);
485 486 487 488 489

	dprintk("reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n",
		start, end, cattr_name(new->type), cattr_name(req_type),
		new_type ? cattr_name(*new_type) : "-");

490 491 492 493 494
	return err;
}

int free_memtype(u64 start, u64 end)
{
495
	struct memtype *entry, *saved_entry;
496
	int err = -EINVAL;
497
	int is_range_ram;
498

499
	if (!pat_enabled)
500 501 502
		return 0;

	/* Low ISA region is always mapped WB. No need to track */
503
	if (x86_platform.is_untracked_pat_range(start, end))
504 505
		return 0;

506
	is_range_ram = pat_pagerange_is_ram(start, end);
507 508 509 510 511 512 513 514
	if (is_range_ram == 1) {

		spin_lock(&memtype_lock);
		err = free_ram_pages_type(start, end);
		spin_unlock(&memtype_lock);

		return err;
	} else if (is_range_ram < 0) {
515
		return -EINVAL;
516
	}
517

518
	spin_lock(&memtype_lock);
519 520 521 522 523 524 525 526 527 528 529 530

	entry = memtype_rb_search(&memtype_rbroot, start);
	if (unlikely(entry == NULL))
		goto unlock_ret;

	/*
	 * Saved entry points to an entry with start same or less than what
	 * we searched for. Now go through the list in both directions to look
	 * for the entry that matches with both start and end, with list stored
	 * in sorted start address
	 */
	saved_entry = entry;
531
	list_for_each_entry_from(entry, &memtype_list, nd) {
532
		if (entry->start == start && entry->end == end) {
533 534 535 536 537 538 539 540 541 542 543 544
			rb_erase(&entry->rb, &memtype_rbroot);
			list_del(&entry->nd);
			kfree(entry);
			err = 0;
			break;
		} else if (entry->start > start) {
			break;
		}
	}

	if (!err)
		goto unlock_ret;
545

546 547 548 549
	entry = saved_entry;
	list_for_each_entry_reverse(entry, &memtype_list, nd) {
		if (entry->start == start && entry->end == end) {
			rb_erase(&entry->rb, &memtype_rbroot);
550 551
			list_del(&entry->nd);
			kfree(entry);
552 553
			err = 0;
			break;
554 555
		} else if (entry->start < start) {
			break;
556 557
		}
	}
558
unlock_ret:
559 560 561
	spin_unlock(&memtype_lock);

	if (err) {
I
Ingo Molnar 已提交
562
		printk(KERN_INFO "%s:%d freeing invalid memtype %Lx-%Lx\n",
563 564
			current->comm, current->pid, start, end);
	}
565

566
	dprintk("free_memtype request 0x%Lx-0x%Lx\n", start, end);
I
Ingo Molnar 已提交
567

568 569 570
	return err;
}

571

572 573 574 575 576 577 578 579 580 581 582 583 584 585
/**
 * lookup_memtype - Looksup the memory type for a physical address
 * @paddr: physical address of which memory type needs to be looked up
 *
 * Only to be called when PAT is enabled
 *
 * Returns _PAGE_CACHE_WB, _PAGE_CACHE_WC, _PAGE_CACHE_UC_MINUS or
 * _PAGE_CACHE_UC
 */
static unsigned long lookup_memtype(u64 paddr)
{
	int rettype = _PAGE_CACHE_WB;
	struct memtype *entry;

586
	if (x86_platform.is_untracked_pat_range(paddr, paddr + PAGE_SIZE))
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
		return rettype;

	if (pat_pagerange_is_ram(paddr, paddr + PAGE_SIZE)) {
		struct page *page;
		spin_lock(&memtype_lock);
		page = pfn_to_page(paddr >> PAGE_SHIFT);
		rettype = get_page_memtype(page);
		spin_unlock(&memtype_lock);
		/*
		 * -1 from get_page_memtype() implies RAM page is in its
		 * default state and not reserved, and hence of type WB
		 */
		if (rettype == -1)
			rettype = _PAGE_CACHE_WB;

		return rettype;
	}

	spin_lock(&memtype_lock);

	entry = memtype_rb_search(&memtype_rbroot, paddr);
	if (entry != NULL)
		rettype = entry->type;
	else
		rettype = _PAGE_CACHE_UC_MINUS;

	spin_unlock(&memtype_lock);
	return rettype;
}

617 618 619 620 621 622 623 624 625 626 627 628 629
/**
 * io_reserve_memtype - Request a memory type mapping for a region of memory
 * @start: start (physical address) of the region
 * @end: end (physical address) of the region
 * @type: A pointer to memtype, with requested type. On success, requested
 * or any other compatible type that was available for the region is returned
 *
 * On success, returns 0
 * On failure, returns non-zero
 */
int io_reserve_memtype(resource_size_t start, resource_size_t end,
			unsigned long *type)
{
630
	resource_size_t size = end - start;
631 632 633 634
	unsigned long req_type = *type;
	unsigned long new_type;
	int ret;

635
	WARN_ON_ONCE(iomem_map_sanity_check(start, size));
636 637 638 639 640

	ret = reserve_memtype(start, end, req_type, &new_type);
	if (ret)
		goto out_err;

641
	if (!is_new_memtype_allowed(start, size, req_type, new_type))
642 643
		goto out_free;

644
	if (kernel_map_sync_memtype(start, size, new_type) < 0)
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
		goto out_free;

	*type = new_type;
	return 0;

out_free:
	free_memtype(start, end);
	ret = -EBUSY;
out_err:
	return ret;
}

/**
 * io_free_memtype - Release a memory type mapping for a region of memory
 * @start: start (physical address) of the region
 * @end: end (physical address) of the region
 */
void io_free_memtype(resource_size_t start, resource_size_t end)
{
	free_memtype(start, end);
}

667 668 669 670 671 672
pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t vma_prot)
{
	return vma_prot;
}

673 674
#ifdef CONFIG_STRICT_DEVMEM
/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM*/
675 676 677 678 679
static inline int range_is_allowed(unsigned long pfn, unsigned long size)
{
	return 1;
}
#else
680
/* This check is needed to avoid cache aliasing when PAT is enabled */
681 682 683 684 685 686
static inline int range_is_allowed(unsigned long pfn, unsigned long size)
{
	u64 from = ((u64)pfn) << PAGE_SHIFT;
	u64 to = from + size;
	u64 cursor = from;

687 688 689
	if (!pat_enabled)
		return 1;

690 691 692 693 694 695 696 697 698 699 700 701
	while (cursor < to) {
		if (!devmem_is_allowed(pfn)) {
			printk(KERN_INFO
		"Program %s tried to access /dev/mem between %Lx->%Lx.\n",
				current->comm, from, to);
			return 0;
		}
		cursor += PAGE_SIZE;
		pfn++;
	}
	return 1;
}
702
#endif /* CONFIG_STRICT_DEVMEM */
703

704 705 706
int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t *vma_prot)
{
707
	unsigned long flags = _PAGE_CACHE_WB;
708

709 710 711
	if (!range_is_allowed(pfn, size))
		return 0;

712
	if (file->f_flags & O_SYNC) {
713
		flags = _PAGE_CACHE_UC_MINUS;
714 715 716 717 718 719 720 721 722 723 724
	}

#ifdef CONFIG_X86_32
	/*
	 * On the PPro and successors, the MTRRs are used to set
	 * memory types for physical addresses outside main memory,
	 * so blindly setting UC or PWT on those pages is wrong.
	 * For Pentiums and earlier, the surround logic should disable
	 * caching for the high addresses through the KEN pin, but
	 * we maintain the tradition of paranoia in this code.
	 */
725
	if (!pat_enabled &&
726 727 728 729 730
	    !(boot_cpu_has(X86_FEATURE_MTRR) ||
	      boot_cpu_has(X86_FEATURE_K6_MTRR) ||
	      boot_cpu_has(X86_FEATURE_CYRIX_ARR) ||
	      boot_cpu_has(X86_FEATURE_CENTAUR_MCR)) &&
	    (pfn << PAGE_SHIFT) >= __pa(high_memory)) {
731
		flags = _PAGE_CACHE_UC;
732 733 734
	}
#endif

735 736
	*vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) |
			     flags);
737 738
	return 1;
}
739

740 741 742 743 744 745 746 747
/*
 * Change the memory type for the physial address range in kernel identity
 * mapping space if that range is a part of identity map.
 */
int kernel_map_sync_memtype(u64 base, unsigned long size, unsigned long flags)
{
	unsigned long id_sz;

748
	if (base >= __pa(high_memory))
749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
		return 0;

	id_sz = (__pa(high_memory) < base + size) ?
				__pa(high_memory) - base :
				size;

	if (ioremap_change_attr((unsigned long)__va(base), id_sz, flags) < 0) {
		printk(KERN_INFO
			"%s:%d ioremap_change_attr failed %s "
			"for %Lx-%Lx\n",
			current->comm, current->pid,
			cattr_name(flags),
			base, (unsigned long long)(base + size));
		return -EINVAL;
	}
	return 0;
}

767 768 769 770 771
/*
 * Internal interface to reserve a range of physical memory with prot.
 * Reserved non RAM regions only and after successful reserve_memtype,
 * this func also keeps identity mapping (if any) in sync with this new prot.
 */
772 773
static int reserve_pfn_range(u64 paddr, unsigned long size, pgprot_t *vma_prot,
				int strict_prot)
774 775
{
	int is_ram = 0;
776
	int ret;
777
	unsigned long want_flags = (pgprot_val(*vma_prot) & _PAGE_CACHE_MASK);
778
	unsigned long flags = want_flags;
779

780
	is_ram = pat_pagerange_is_ram(paddr, paddr + size);
781

782
	/*
783 784 785
	 * reserve_pfn_range() for RAM pages. We do not refcount to keep
	 * track of number of mappings of RAM pages. We can assert that
	 * the type requested matches the type of first page in the range.
786
	 */
787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
	if (is_ram) {
		if (!pat_enabled)
			return 0;

		flags = lookup_memtype(paddr);
		if (want_flags != flags) {
			printk(KERN_WARNING
			"%s:%d map pfn RAM range req %s for %Lx-%Lx, got %s\n",
				current->comm, current->pid,
				cattr_name(want_flags),
				(unsigned long long)paddr,
				(unsigned long long)(paddr + size),
				cattr_name(flags));
			*vma_prot = __pgprot((pgprot_val(*vma_prot) &
					      (~_PAGE_CACHE_MASK)) |
					     flags);
		}
804
		return 0;
805
	}
806 807 808 809 810 811

	ret = reserve_memtype(paddr, paddr + size, want_flags, &flags);
	if (ret)
		return ret;

	if (flags != want_flags) {
812 813
		if (strict_prot ||
		    !is_new_memtype_allowed(paddr, size, want_flags, flags)) {
814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830
			free_memtype(paddr, paddr + size);
			printk(KERN_ERR "%s:%d map pfn expected mapping type %s"
				" for %Lx-%Lx, got %s\n",
				current->comm, current->pid,
				cattr_name(want_flags),
				(unsigned long long)paddr,
				(unsigned long long)(paddr + size),
				cattr_name(flags));
			return -EINVAL;
		}
		/*
		 * We allow returning different type than the one requested in
		 * non strict case.
		 */
		*vma_prot = __pgprot((pgprot_val(*vma_prot) &
				      (~_PAGE_CACHE_MASK)) |
				     flags);
831 832
	}

833
	if (kernel_map_sync_memtype(paddr, size, flags) < 0) {
834 835 836 837 838 839 840 841 842 843 844 845 846 847
		free_memtype(paddr, paddr + size);
		return -EINVAL;
	}
	return 0;
}

/*
 * Internal interface to free a range of physical memory.
 * Frees non RAM regions only.
 */
static void free_pfn_range(u64 paddr, unsigned long size)
{
	int is_ram;

848
	is_ram = pat_pagerange_is_ram(paddr, paddr + size);
849 850 851 852 853 854 855 856 857 858 859 860 861
	if (is_ram == 0)
		free_memtype(paddr, paddr + size);
}

/*
 * track_pfn_vma_copy is called when vma that is covering the pfnmap gets
 * copied through copy_page_range().
 *
 * If the vma has a linear pfn mapping for the entire range, we get the prot
 * from pte and reserve the entire vma range with single reserve_pfn_range call.
 */
int track_pfn_vma_copy(struct vm_area_struct *vma)
{
862
	resource_size_t paddr;
863
	unsigned long prot;
864
	unsigned long vma_size = vma->vm_end - vma->vm_start;
865
	pgprot_t pgprot;
866 867 868

	if (is_linear_pfn_mapping(vma)) {
		/*
869 870
		 * reserve the whole chunk covered by vma. We need the
		 * starting address and protection from pte.
871
		 */
872
		if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr)) {
873
			WARN_ON_ONCE(1);
874
			return -EINVAL;
875
		}
876 877
		pgprot = __pgprot(prot);
		return reserve_pfn_range(paddr, vma_size, &pgprot, 1);
878 879 880 881 882 883 884 885 886 887 888 889 890
	}

	return 0;
}

/*
 * track_pfn_vma_new is called when a _new_ pfn mapping is being established
 * for physical range indicated by pfn and size.
 *
 * prot is passed in as a parameter for the new mapping. If the vma has a
 * linear pfn mapping for the entire range reserve the entire vma range with
 * single reserve_pfn_range call.
 */
891
int track_pfn_vma_new(struct vm_area_struct *vma, pgprot_t *prot,
892 893
			unsigned long pfn, unsigned long size)
{
894
	unsigned long flags;
895
	resource_size_t paddr;
896
	unsigned long vma_size = vma->vm_end - vma->vm_start;
897 898 899

	if (is_linear_pfn_mapping(vma)) {
		/* reserve the whole chunk starting from vm_pgoff */
900
		paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT;
901
		return reserve_pfn_range(paddr, vma_size, prot, 0);
902 903
	}

904 905 906 907 908 909 910 911
	if (!pat_enabled)
		return 0;

	/* for vm_insert_pfn and friends, we set prot based on lookup */
	flags = lookup_memtype(pfn << PAGE_SHIFT);
	*prot = __pgprot((pgprot_val(vma->vm_page_prot) & (~_PAGE_CACHE_MASK)) |
			 flags);

912 913 914 915 916 917 918 919 920 921 922
	return 0;
}

/*
 * untrack_pfn_vma is called while unmapping a pfnmap for a region.
 * untrack can be called for a specific region indicated by pfn and size or
 * can be for the entire vma (in which case size can be zero).
 */
void untrack_pfn_vma(struct vm_area_struct *vma, unsigned long pfn,
			unsigned long size)
{
923
	resource_size_t paddr;
924
	unsigned long vma_size = vma->vm_end - vma->vm_start;
925 926 927

	if (is_linear_pfn_mapping(vma)) {
		/* free the whole chunk starting from vm_pgoff */
928
		paddr = (resource_size_t)vma->vm_pgoff << PAGE_SHIFT;
929 930 931 932 933
		free_pfn_range(paddr, vma_size);
		return;
	}
}

934 935 936 937 938 939 940
pgprot_t pgprot_writecombine(pgprot_t prot)
{
	if (pat_enabled)
		return __pgprot(pgprot_val(prot) | _PAGE_CACHE_WC);
	else
		return pgprot_noncached(prot);
}
941
EXPORT_SYMBOL_GPL(pgprot_writecombine);
942

943
#if defined(CONFIG_DEBUG_FS) && defined(CONFIG_X86_PAT)
944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965

/* get Nth element of the linked list */
static struct memtype *memtype_get_idx(loff_t pos)
{
	struct memtype *list_node, *print_entry;
	int i = 1;

	print_entry  = kmalloc(sizeof(struct memtype), GFP_KERNEL);
	if (!print_entry)
		return NULL;

	spin_lock(&memtype_lock);
	list_for_each_entry(list_node, &memtype_list, nd) {
		if (pos == i) {
			*print_entry = *list_node;
			spin_unlock(&memtype_lock);
			return print_entry;
		}
		++i;
	}
	spin_unlock(&memtype_lock);
	kfree(print_entry);
I
Ingo Molnar 已提交
966

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
	return NULL;
}

static void *memtype_seq_start(struct seq_file *seq, loff_t *pos)
{
	if (*pos == 0) {
		++*pos;
		seq_printf(seq, "PAT memtype list:\n");
	}

	return memtype_get_idx(*pos);
}

static void *memtype_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	++*pos;
	return memtype_get_idx(*pos);
}

static void memtype_seq_stop(struct seq_file *seq, void *v)
{
}

static int memtype_seq_show(struct seq_file *seq, void *v)
{
	struct memtype *print_entry = (struct memtype *)v;

	seq_printf(seq, "%s @ 0x%Lx-0x%Lx\n", cattr_name(print_entry->type),
			print_entry->start, print_entry->end);
	kfree(print_entry);
I
Ingo Molnar 已提交
997

998 999 1000
	return 0;
}

T
Tobias Klauser 已提交
1001
static const struct seq_operations memtype_seq_ops = {
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
	.start = memtype_seq_start,
	.next  = memtype_seq_next,
	.stop  = memtype_seq_stop,
	.show  = memtype_seq_show,
};

static int memtype_seq_open(struct inode *inode, struct file *file)
{
	return seq_open(file, &memtype_seq_ops);
}

static const struct file_operations memtype_fops = {
	.open    = memtype_seq_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
	.release = seq_release,
};

static int __init pat_memtype_list_init(void)
{
	debugfs_create_file("pat_memtype_list", S_IRUSR, arch_debugfs_dir,
				NULL, &memtype_fops);
	return 0;
}

late_initcall(pat_memtype_list_init);

1029
#endif /* CONFIG_DEBUG_FS && CONFIG_X86_PAT */