pat.c 21.0 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/slab.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 34
#include "pat_internal.h"

35
#ifdef CONFIG_X86_PAT
36
int __read_mostly pat_enabled = 1;
37

38
static inline void pat_disable(const char *reason)
39
{
40
	pat_enabled = 0;
41
	printk(KERN_INFO "%s\n", reason);
42 43
}

A
Andrew Morton 已提交
44
static int __init nopat(char *str)
45
{
46
	pat_disable("PAT support disabled.");
47 48
	return 0;
}
49
early_param("nopat", nopat);
50 51 52 53 54
#else
static inline void pat_disable(const char *reason)
{
	(void)reason;
}
55 56
#endif

57

58
int pat_debug_enable;
I
Ingo Molnar 已提交
59

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

67
static u64 __read_mostly boot_pat_state;
68 69 70 71 72 73 74 75 76 77

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 */
};

78
#define PAT(x, y)	((u64)PAT_ ## y << ((x)*8))
79 80 81 82

void pat_init(void)
{
	u64 pat;
83
	bool boot_cpu = !boot_pat_state;
84

85
	if (!pat_enabled)
86 87
		return;

88 89 90 91 92 93 94 95 96 97 98 99 100 101
	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();
		}
102
	}
103 104 105 106 107 108 109 110 111 112 113 114 115 116

	/* 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
	 */
117 118
	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);
119 120

	/* Boot CPU check */
121
	if (!boot_pat_state)
122 123 124
		rdmsrl(MSR_IA32_CR_PAT, boot_pat_state);

	wrmsrl(MSR_IA32_CR_PAT, pat);
125 126 127 128

	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);
129 130 131 132
}

#undef PAT

133
static DEFINE_SPINLOCK(memtype_lock);	/* protects memtype accesses */
134

135 136 137 138 139 140 141
/*
 * 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
 */
142
static unsigned long pat_x_mtrr_type(u64 start, u64 end, unsigned long req_type)
143
{
144 145 146 147
	/*
	 * Look for MTRR hint to get the effective type in case where PAT
	 * request is for WB.
	 */
148 149 150 151
	if (req_type == _PAGE_CACHE_WB) {
		u8 mtrr_type;

		mtrr_type = mtrr_type_lookup(start, end);
152 153 154 155
		if (mtrr_type != MTRR_TYPE_WRBACK)
			return _PAGE_CACHE_UC_MINUS;

		return _PAGE_CACHE_WB;
156 157 158
	}

	return req_type;
159 160
}

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
struct pagerange_state {
	unsigned long		cur_pfn;
	int			ram;
	int			not_ram;
};

static int
pagerange_is_ram_callback(unsigned long initial_pfn, unsigned long total_nr_pages, void *arg)
{
	struct pagerange_state *state = arg;

	state->not_ram	|= initial_pfn > state->cur_pfn;
	state->ram	|= total_nr_pages > 0;
	state->cur_pfn	 = initial_pfn + total_nr_pages;

	return state->ram && state->not_ram;
}

179
static int pat_pagerange_is_ram(resource_size_t start, resource_size_t end)
180
{
181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
	int ret = 0;
	unsigned long start_pfn = start >> PAGE_SHIFT;
	unsigned long end_pfn = (end + PAGE_SIZE - 1) >> PAGE_SHIFT;
	struct pagerange_state state = {start_pfn, 0, 0};

	/*
	 * 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 (start_pfn < ISA_END_ADDRESS >> PAGE_SHIFT)
		start_pfn = ISA_END_ADDRESS >> PAGE_SHIFT;

	if (start_pfn < end_pfn) {
		ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn,
				&state, pagerange_is_ram_callback);
199 200
	}

201
	return (ret > 0) ? -1 : (state.ram ? 1 : 0);
202 203
}

204
/*
205 206 207 208
 * 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
209 210
 */
static int reserve_ram_pages_type(u64 start, u64 end, unsigned long req_type,
I
Ingo Molnar 已提交
211
				  unsigned long *new_type)
212 213
{
	struct page *page;
214 215 216 217 218 219 220
	u64 pfn;

	if (req_type == _PAGE_CACHE_UC) {
		/* We do not support strong UC */
		WARN_ON_ONCE(1);
		req_type = _PAGE_CACHE_UC_MINUS;
	}
221 222

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
223
		unsigned long type;
224

225 226 227
		page = pfn_to_page(pfn);
		type = get_page_memtype(page);
		if (type != -1) {
228 229
			printk(KERN_INFO "reserve_ram_pages_type failed [mem %#010Lx-%#010Lx], track 0x%lx, req 0x%lx\n",
				start, end - 1, type, req_type);
230 231 232 233 234
			if (new_type)
				*new_type = type;

			return -EBUSY;
		}
235 236
	}

237 238 239 240
	if (new_type)
		*new_type = req_type;

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
241
		page = pfn_to_page(pfn);
242
		set_page_memtype(page, req_type);
243
	}
244
	return 0;
245 246 247 248 249
}

static int free_ram_pages_type(u64 start, u64 end)
{
	struct page *page;
250
	u64 pfn;
251 252 253

	for (pfn = (start >> PAGE_SHIFT); pfn < (end >> PAGE_SHIFT); ++pfn) {
		page = pfn_to_page(pfn);
254
		set_page_memtype(page, -1);
255 256 257 258
	}
	return 0;
}

259 260 261 262 263 264 265
/*
 * req_type typically has one of the:
 * - _PAGE_CACHE_WB
 * - _PAGE_CACHE_WC
 * - _PAGE_CACHE_UC_MINUS
 * - _PAGE_CACHE_UC
 *
266 267 268
 * 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
269 270
 * it will return a negative return value.
 */
271
int reserve_memtype(u64 start, u64 end, unsigned long req_type,
I
Ingo Molnar 已提交
272
		    unsigned long *new_type)
273
{
274
	struct memtype *new;
275
	unsigned long actual_type;
276
	int is_range_ram;
I
Ingo Molnar 已提交
277
	int err = 0;
278

I
Ingo Molnar 已提交
279
	BUG_ON(start >= end); /* end is exclusive */
280

281
	if (!pat_enabled) {
282
		/* This is identical to page table setting without PAT */
283
		if (new_type) {
284
			if (req_type == _PAGE_CACHE_WC)
285
				*new_type = _PAGE_CACHE_UC_MINUS;
286 287
			else
				*new_type = req_type & _PAGE_CACHE_MASK;
288
		}
289 290 291 292
		return 0;
	}

	/* Low ISA region is always mapped WB in page table. No need to track */
293
	if (x86_platform.is_untracked_pat_range(start, end)) {
294 295
		if (new_type)
			*new_type = _PAGE_CACHE_WB;
296 297 298
		return 0;
	}

299 300 301 302 303 304 305
	/*
	 * 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);
306

307 308 309
	if (new_type)
		*new_type = actual_type;

310
	is_range_ram = pat_pagerange_is_ram(start, end);
311 312 313 314 315 316
	if (is_range_ram == 1) {

		err = reserve_ram_pages_type(start, end, req_type, new_type);

		return err;
	} else if (is_range_ram < 0) {
317
		return -EINVAL;
318
	}
319

320
	new  = kzalloc(sizeof(struct memtype), GFP_KERNEL);
321
	if (!new)
322 323
		return -ENOMEM;

I
Ingo Molnar 已提交
324 325 326
	new->start	= start;
	new->end	= end;
	new->type	= actual_type;
327 328 329

	spin_lock(&memtype_lock);

330
	err = rbt_memtype_check_insert(new, new_type);
331
	if (err) {
332 333 334
		printk(KERN_INFO "reserve_memtype failed [mem %#010Lx-%#010Lx], track %s, req %s\n",
		       start, end - 1,
		       cattr_name(new->type), cattr_name(req_type));
335
		kfree(new);
336
		spin_unlock(&memtype_lock);
I
Ingo Molnar 已提交
337

338 339 340 341
		return err;
	}

	spin_unlock(&memtype_lock);
342

343 344
	dprintk("reserve_memtype added [mem %#010Lx-%#010Lx], track %s, req %s, ret %s\n",
		start, end - 1, cattr_name(new->type), cattr_name(req_type),
345 346
		new_type ? cattr_name(*new_type) : "-");

347 348 349 350 351 352
	return err;
}

int free_memtype(u64 start, u64 end)
{
	int err = -EINVAL;
353
	int is_range_ram;
354
	struct memtype *entry;
355

356
	if (!pat_enabled)
357 358 359
		return 0;

	/* Low ISA region is always mapped WB. No need to track */
360
	if (x86_platform.is_untracked_pat_range(start, end))
361 362
		return 0;

363
	is_range_ram = pat_pagerange_is_ram(start, end);
364 365 366 367 368 369
	if (is_range_ram == 1) {

		err = free_ram_pages_type(start, end);

		return err;
	} else if (is_range_ram < 0) {
370
		return -EINVAL;
371
	}
372

373
	spin_lock(&memtype_lock);
374
	entry = rbt_memtype_erase(start, end);
375 376
	spin_unlock(&memtype_lock);

377
	if (!entry) {
378 379
		printk(KERN_INFO "%s:%d freeing invalid memtype [mem %#010Lx-%#010Lx]\n",
		       current->comm, current->pid, start, end - 1);
380
		return -EINVAL;
381
	}
382

383 384
	kfree(entry);

385
	dprintk("free_memtype request [mem %#010Lx-%#010Lx]\n", start, end - 1);
I
Ingo Molnar 已提交
386

387
	return 0;
388 389
}

390

391 392 393 394 395 396 397 398 399 400 401 402 403 404
/**
 * 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;

405
	if (x86_platform.is_untracked_pat_range(paddr, paddr + PAGE_SIZE))
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
		return rettype;

	if (pat_pagerange_is_ram(paddr, paddr + PAGE_SIZE)) {
		struct page *page;
		page = pfn_to_page(paddr >> PAGE_SHIFT);
		rettype = get_page_memtype(page);
		/*
		 * -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);

424
	entry = rbt_memtype_lookup(paddr);
425 426 427 428 429 430 431 432 433
	if (entry != NULL)
		rettype = entry->type;
	else
		rettype = _PAGE_CACHE_UC_MINUS;

	spin_unlock(&memtype_lock);
	return rettype;
}

434 435 436 437 438 439 440 441 442 443 444 445 446
/**
 * 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)
{
447
	resource_size_t size = end - start;
448 449 450 451
	unsigned long req_type = *type;
	unsigned long new_type;
	int ret;

452
	WARN_ON_ONCE(iomem_map_sanity_check(start, size));
453 454 455 456 457

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

458
	if (!is_new_memtype_allowed(start, size, req_type, new_type))
459 460
		goto out_free;

461
	if (kernel_map_sync_memtype(start, size, new_type) < 0)
462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
		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);
}

484 485 486 487 488 489
pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t vma_prot)
{
	return vma_prot;
}

490 491
#ifdef CONFIG_STRICT_DEVMEM
/* This check is done in drivers/char/mem.c in case of STRICT_DEVMEM*/
492 493 494 495 496
static inline int range_is_allowed(unsigned long pfn, unsigned long size)
{
	return 1;
}
#else
497
/* This check is needed to avoid cache aliasing when PAT is enabled */
498 499 500 501 502 503
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;

504 505 506
	if (!pat_enabled)
		return 1;

507 508
	while (cursor < to) {
		if (!devmem_is_allowed(pfn)) {
509 510
			printk(KERN_INFO "Program %s tried to access /dev/mem between [mem %#010Lx-%#010Lx]\n",
				current->comm, from, to - 1);
511 512 513 514 515 516 517
			return 0;
		}
		cursor += PAGE_SIZE;
		pfn++;
	}
	return 1;
}
518
#endif /* CONFIG_STRICT_DEVMEM */
519

520 521 522
int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn,
				unsigned long size, pgprot_t *vma_prot)
{
523
	unsigned long flags = _PAGE_CACHE_WB;
524

525 526 527
	if (!range_is_allowed(pfn, size))
		return 0;

528
	if (file->f_flags & O_DSYNC)
529
		flags = _PAGE_CACHE_UC_MINUS;
530 531 532 533 534 535 536 537 538 539

#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.
	 */
540
	if (!pat_enabled &&
541 542 543 544 545
	    !(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)) {
546
		flags = _PAGE_CACHE_UC;
547 548 549
	}
#endif

550 551
	*vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) |
			     flags);
552 553
	return 1;
}
554

555 556 557 558 559 560 561 562
/*
 * 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;

563
	if (base > __pa(high_memory-1))
564 565
		return 0;

566 567 568 569 570 571 572
	/*
	 * some areas in the middle of the kernel identity range
	 * are not mapped, like the PCI space.
	 */
	if (!page_is_ram(base >> PAGE_SHIFT))
		return 0;

573
	id_sz = (__pa(high_memory-1) <= base + size) ?
574 575 576 577
				__pa(high_memory) - base :
				size;

	if (ioremap_change_attr((unsigned long)__va(base), id_sz, flags) < 0) {
578 579
		printk(KERN_INFO "%s:%d ioremap_change_attr failed %s "
			"for [mem %#010Lx-%#010Lx]\n",
580 581
			current->comm, current->pid,
			cattr_name(flags),
582
			base, (unsigned long long)(base + size-1));
583 584 585 586 587
		return -EINVAL;
	}
	return 0;
}

588 589 590 591 592
/*
 * 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.
 */
593 594
static int reserve_pfn_range(u64 paddr, unsigned long size, pgprot_t *vma_prot,
				int strict_prot)
595 596
{
	int is_ram = 0;
597
	int ret;
598
	unsigned long want_flags = (pgprot_val(*vma_prot) & _PAGE_CACHE_MASK);
599
	unsigned long flags = want_flags;
600

601
	is_ram = pat_pagerange_is_ram(paddr, paddr + size);
602

603
	/*
604 605 606
	 * 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.
607
	 */
608 609 610 611 612 613
	if (is_ram) {
		if (!pat_enabled)
			return 0;

		flags = lookup_memtype(paddr);
		if (want_flags != flags) {
614
			printk(KERN_WARNING "%s:%d map pfn RAM range req %s for [mem %#010Lx-%#010Lx], got %s\n",
615 616 617
				current->comm, current->pid,
				cattr_name(want_flags),
				(unsigned long long)paddr,
618
				(unsigned long long)(paddr + size - 1),
619 620 621 622 623
				cattr_name(flags));
			*vma_prot = __pgprot((pgprot_val(*vma_prot) &
					      (~_PAGE_CACHE_MASK)) |
					     flags);
		}
624
		return 0;
625
	}
626 627 628 629 630 631

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

	if (flags != want_flags) {
632 633
		if (strict_prot ||
		    !is_new_memtype_allowed(paddr, size, want_flags, flags)) {
634 635
			free_memtype(paddr, paddr + size);
			printk(KERN_ERR "%s:%d map pfn expected mapping type %s"
636
				" for [mem %#010Lx-%#010Lx], got %s\n",
637 638 639
				current->comm, current->pid,
				cattr_name(want_flags),
				(unsigned long long)paddr,
640
				(unsigned long long)(paddr + size - 1),
641 642 643 644 645 646 647 648 649 650
				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);
651 652
	}

653
	if (kernel_map_sync_memtype(paddr, size, flags) < 0) {
654 655 656 657 658 659 660 661 662 663 664 665 666 667
		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;

668
	is_ram = pat_pagerange_is_ram(paddr, paddr + size);
669 670 671 672 673
	if (is_ram == 0)
		free_memtype(paddr, paddr + size);
}

/*
674
 * track_pfn_copy is called when vma that is covering the pfnmap gets
675 676 677 678 679
 * 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.
 */
680
int track_pfn_copy(struct vm_area_struct *vma)
681
{
682
	resource_size_t paddr;
683
	unsigned long prot;
684
	unsigned long vma_size = vma->vm_end - vma->vm_start;
685
	pgprot_t pgprot;
686

687
	if (vma->vm_flags & VM_PAT) {
688
		/*
689 690
		 * reserve the whole chunk covered by vma. We need the
		 * starting address and protection from pte.
691
		 */
692
		if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr)) {
693
			WARN_ON_ONCE(1);
694
			return -EINVAL;
695
		}
696 697
		pgprot = __pgprot(prot);
		return reserve_pfn_range(paddr, vma_size, &pgprot, 1);
698 699 700 701 702 703 704 705 706 707
	}

	return 0;
}

/*
 * 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.
 */
708
int track_pfn_remap(struct vm_area_struct *vma, pgprot_t *prot,
709
		    unsigned long pfn, unsigned long addr, unsigned long size)
710
{
711
	resource_size_t paddr = (resource_size_t)pfn << PAGE_SHIFT;
712
	unsigned long flags;
713

714
	/* reserve the whole chunk starting from paddr */
715 716 717 718 719 720 721 722
	if (addr == vma->vm_start && size == (vma->vm_end - vma->vm_start)) {
		int ret;

		ret = reserve_pfn_range(paddr, size, prot, 0);
		if (!ret)
			vma->vm_flags |= VM_PAT;
		return ret;
	}
723

724 725 726
	if (!pat_enabled)
		return 0;

727 728 729 730
	/*
	 * For anything smaller than the vma size we set prot based on the
	 * lookup.
	 */
731
	flags = lookup_memtype(paddr);
732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756

	/* Check memtype for the remaining pages */
	while (size > PAGE_SIZE) {
		size -= PAGE_SIZE;
		paddr += PAGE_SIZE;
		if (flags != lookup_memtype(paddr))
			return -EINVAL;
	}

	*prot = __pgprot((pgprot_val(vma->vm_page_prot) & (~_PAGE_CACHE_MASK)) |
			 flags);

	return 0;
}

int track_pfn_insert(struct vm_area_struct *vma, pgprot_t *prot,
		     unsigned long pfn)
{
	unsigned long flags;

	if (!pat_enabled)
		return 0;

	/* Set prot based on lookup */
	flags = lookup_memtype((resource_size_t)pfn << PAGE_SHIFT);
757 758 759
	*prot = __pgprot((pgprot_val(vma->vm_page_prot) & (~_PAGE_CACHE_MASK)) |
			 flags);

760 761 762 763
	return 0;
}

/*
764
 * untrack_pfn is called while unmapping a pfnmap for a region.
765
 * untrack can be called for a specific region indicated by pfn and size or
766
 * can be for the entire vma (in which case pfn, size are zero).
767
 */
768 769
void untrack_pfn(struct vm_area_struct *vma, unsigned long pfn,
		 unsigned long size)
770
{
771
	resource_size_t paddr;
772
	unsigned long prot;
773

774
	if (!(vma->vm_flags & VM_PAT))
775
		return;
776 777 778 779 780 781 782 783 784 785

	/* free the chunk starting from pfn or the whole chunk */
	paddr = (resource_size_t)pfn << PAGE_SHIFT;
	if (!paddr && !size) {
		if (follow_phys(vma, vma->vm_start, 0, &prot, &paddr)) {
			WARN_ON_ONCE(1);
			return;
		}

		size = vma->vm_end - vma->vm_start;
786
	}
787
	free_pfn_range(paddr, size);
788
	vma->vm_flags &= ~VM_PAT;
789 790
}

791 792 793 794 795 796 797
pgprot_t pgprot_writecombine(pgprot_t prot)
{
	if (pat_enabled)
		return __pgprot(pgprot_val(prot) | _PAGE_CACHE_WC);
	else
		return pgprot_noncached(prot);
}
798
EXPORT_SYMBOL_GPL(pgprot_writecombine);
799

800
#if defined(CONFIG_DEBUG_FS) && defined(CONFIG_X86_PAT)
801 802 803

static struct memtype *memtype_get_idx(loff_t pos)
{
804 805
	struct memtype *print_entry;
	int ret;
806

807
	print_entry  = kzalloc(sizeof(struct memtype), GFP_KERNEL);
808 809 810 811
	if (!print_entry)
		return NULL;

	spin_lock(&memtype_lock);
812
	ret = rbt_memtype_copy_nth_element(print_entry, pos);
813
	spin_unlock(&memtype_lock);
I
Ingo Molnar 已提交
814

815 816 817 818 819 820
	if (!ret) {
		return print_entry;
	} else {
		kfree(print_entry);
		return NULL;
	}
821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849
}

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
850

851 852 853
	return 0;
}

T
Tobias Klauser 已提交
854
static const struct seq_operations memtype_seq_ops = {
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874
	.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)
{
875 876 877 878
	if (pat_enabled) {
		debugfs_create_file("pat_memtype_list", S_IRUSR,
				    arch_debugfs_dir, NULL, &memtype_fops);
	}
879 880 881 882 883
	return 0;
}

late_initcall(pat_memtype_list_init);

884
#endif /* CONFIG_DEBUG_FS && CONFIG_X86_PAT */