pageattr.c 33.4 KB
Newer Older
1 2
/*
 * Copyright 2002 Andi Kleen, SuSE Labs.
L
Linus Torvalds 已提交
3
 * Thanks to Ben LaHaise for precious feedback.
4
 */
L
Linus Torvalds 已提交
5
#include <linux/highmem.h>
I
Ingo Molnar 已提交
6
#include <linux/bootmem.h>
L
Linus Torvalds 已提交
7
#include <linux/module.h>
8 9
#include <linux/sched.h>
#include <linux/mm.h>
10
#include <linux/interrupt.h>
11 12
#include <linux/seq_file.h>
#include <linux/debugfs.h>
13
#include <linux/pfn.h>
14
#include <linux/percpu.h>
15
#include <linux/gfp.h>
16

17
#include <asm/e820.h>
L
Linus Torvalds 已提交
18 19
#include <asm/processor.h>
#include <asm/tlbflush.h>
D
Dave Jones 已提交
20
#include <asm/sections.h>
21
#include <asm/setup.h>
22 23
#include <asm/uaccess.h>
#include <asm/pgalloc.h>
T
Thomas Gleixner 已提交
24
#include <asm/proto.h>
25
#include <asm/pat.h>
L
Linus Torvalds 已提交
26

I
Ingo Molnar 已提交
27 28 29
/*
 * The current flushing context - we pass it instead of 5 arguments:
 */
T
Thomas Gleixner 已提交
30
struct cpa_data {
31
	unsigned long	*vaddr;
T
Thomas Gleixner 已提交
32 33
	pgprot_t	mask_set;
	pgprot_t	mask_clr;
34
	int		numpages;
35
	int		flags;
T
Thomas Gleixner 已提交
36
	unsigned long	pfn;
37
	unsigned	force_split : 1;
38
	int		curpage;
39
	struct page	**pages;
T
Thomas Gleixner 已提交
40 41
};

42 43 44 45 46 47 48 49
/*
 * Serialize cpa() (for !DEBUG_PAGEALLOC which uses large identity mappings)
 * using cpa_lock. So that we don't allow any other cpu, with stale large tlb
 * entries change the page attribute in parallel to some other cpu
 * splitting a large page entry along with changing the attribute.
 */
static DEFINE_SPINLOCK(cpa_lock);

50 51
#define CPA_FLUSHTLB 1
#define CPA_ARRAY 2
52
#define CPA_PAGES_ARRAY 4
53

54
#ifdef CONFIG_PROC_FS
55 56
static unsigned long direct_pages_count[PG_LEVEL_NUM];

57
void update_page_count(int level, unsigned long pages)
58 59
{
	unsigned long flags;
60

61 62 63 64
	/* Protect against CPA */
	spin_lock_irqsave(&pgd_lock, flags);
	direct_pages_count[level] += pages;
	spin_unlock_irqrestore(&pgd_lock, flags);
65 66 67 68 69 70 71 72
}

static void split_page_count(int level)
{
	direct_pages_count[level]--;
	direct_pages_count[level - 1] += PTRS_PER_PTE;
}

73
void arch_report_meminfo(struct seq_file *m)
74
{
75
	seq_printf(m, "DirectMap4k:    %8lu kB\n",
H
Hugh Dickins 已提交
76 77
			direct_pages_count[PG_LEVEL_4K] << 2);
#if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
78
	seq_printf(m, "DirectMap2M:    %8lu kB\n",
H
Hugh Dickins 已提交
79 80
			direct_pages_count[PG_LEVEL_2M] << 11);
#else
81
	seq_printf(m, "DirectMap4M:    %8lu kB\n",
H
Hugh Dickins 已提交
82 83
			direct_pages_count[PG_LEVEL_2M] << 12);
#endif
84
#ifdef CONFIG_X86_64
H
Hugh Dickins 已提交
85
	if (direct_gbpages)
86
		seq_printf(m, "DirectMap1G:    %8lu kB\n",
H
Hugh Dickins 已提交
87
			direct_pages_count[PG_LEVEL_1G] << 20);
88 89
#endif
}
90 91 92
#else
static inline void split_page_count(int level) { }
#endif
93

T
Thomas Gleixner 已提交
94 95 96 97 98 99 100 101 102
#ifdef CONFIG_X86_64

static inline unsigned long highmap_start_pfn(void)
{
	return __pa(_text) >> PAGE_SHIFT;
}

static inline unsigned long highmap_end_pfn(void)
{
103
	return __pa(roundup(_brk_end, PMD_SIZE)) >> PAGE_SHIFT;
T
Thomas Gleixner 已提交
104 105 106 107
}

#endif

108 109 110 111 112 113
#ifdef CONFIG_DEBUG_PAGEALLOC
# define debug_pagealloc 1
#else
# define debug_pagealloc 0
#endif

114 115
static inline int
within(unsigned long addr, unsigned long start, unsigned long end)
I
Ingo Molnar 已提交
116
{
117 118 119
	return addr >= start && addr < end;
}

T
Thomas Gleixner 已提交
120 121 122
/*
 * Flushing functions
 */
123 124 125 126 127 128 129 130 131

/**
 * clflush_cache_range - flush a cache range with clflush
 * @addr:	virtual start address
 * @size:	number of bytes to flush
 *
 * clflush is an unordered instruction which needs fencing with mfence
 * to avoid ordering issues.
 */
I
Ingo Molnar 已提交
132
void clflush_cache_range(void *vaddr, unsigned int size)
T
Thomas Gleixner 已提交
133
{
I
Ingo Molnar 已提交
134
	void *vend = vaddr + size - 1;
T
Thomas Gleixner 已提交
135

136
	mb();
I
Ingo Molnar 已提交
137 138 139 140 141 142 143 144

	for (; vaddr < vend; vaddr += boot_cpu_data.x86_clflush_size)
		clflush(vaddr);
	/*
	 * Flush any possible final partial cacheline:
	 */
	clflush(vend);

145
	mb();
T
Thomas Gleixner 已提交
146
}
147
EXPORT_SYMBOL_GPL(clflush_cache_range);
T
Thomas Gleixner 已提交
148

149
static void __cpa_flush_all(void *arg)
T
Thomas Gleixner 已提交
150
{
151 152
	unsigned long cache = (unsigned long)arg;

T
Thomas Gleixner 已提交
153 154 155 156 157 158
	/*
	 * Flush all to work around Errata in early athlons regarding
	 * large page flushing.
	 */
	__flush_tlb_all();

159
	if (cache && boot_cpu_data.x86 >= 4)
T
Thomas Gleixner 已提交
160 161 162
		wbinvd();
}

163
static void cpa_flush_all(unsigned long cache)
T
Thomas Gleixner 已提交
164 165 166
{
	BUG_ON(irqs_disabled());

167
	on_each_cpu(__cpa_flush_all, (void *) cache, 1);
T
Thomas Gleixner 已提交
168 169
}

170 171 172 173 174 175 176 177 178 179
static void __cpa_flush_range(void *arg)
{
	/*
	 * We could optimize that further and do individual per page
	 * tlb invalidates for a low number of pages. Caveat: we must
	 * flush the high aliases on 64bit as well.
	 */
	__flush_tlb_all();
}

180
static void cpa_flush_range(unsigned long start, int numpages, int cache)
181
{
I
Ingo Molnar 已提交
182 183 184
	unsigned int i, level;
	unsigned long addr;

185
	BUG_ON(irqs_disabled());
I
Ingo Molnar 已提交
186
	WARN_ON(PAGE_ALIGN(start) != start);
187

188
	on_each_cpu(__cpa_flush_range, NULL, 1);
189

190 191 192
	if (!cache)
		return;

T
Thomas Gleixner 已提交
193 194 195 196 197 198
	/*
	 * We only need to flush on one CPU,
	 * clflush is a MESI-coherent instruction that
	 * will cause all other CPUs to flush the same
	 * cachelines:
	 */
I
Ingo Molnar 已提交
199 200 201 202 203 204
	for (i = 0, addr = start; i < numpages; i++, addr += PAGE_SIZE) {
		pte_t *pte = lookup_address(addr, &level);

		/*
		 * Only flush present addresses:
		 */
205
		if (pte && (pte_val(*pte) & _PAGE_PRESENT))
I
Ingo Molnar 已提交
206 207
			clflush_cache_range((void *) addr, PAGE_SIZE);
	}
208 209
}

210 211
static void cpa_flush_array(unsigned long *start, int numpages, int cache,
			    int in_flags, struct page **pages)
212 213
{
	unsigned int i, level;
214
	unsigned long do_wbinvd = cache && numpages >= 1024; /* 4M threshold */
215 216 217

	BUG_ON(irqs_disabled());

218
	on_each_cpu(__cpa_flush_all, (void *) do_wbinvd, 1);
219

220
	if (!cache || do_wbinvd)
221 222 223 224 225 226 227 228
		return;

	/*
	 * We only need to flush on one CPU,
	 * clflush is a MESI-coherent instruction that
	 * will cause all other CPUs to flush the same
	 * cachelines:
	 */
229 230 231 232 233 234 235 236 237 238
	for (i = 0; i < numpages; i++) {
		unsigned long addr;
		pte_t *pte;

		if (in_flags & CPA_PAGES_ARRAY)
			addr = (unsigned long)page_address(pages[i]);
		else
			addr = start[i];

		pte = lookup_address(addr, &level);
239 240 241 242 243

		/*
		 * Only flush present addresses:
		 */
		if (pte && (pte_val(*pte) & _PAGE_PRESENT))
244
			clflush_cache_range((void *)addr, PAGE_SIZE);
245 246 247
	}
}

248 249 250 251 252 253
/*
 * Certain areas of memory on x86 require very specific protection flags,
 * for example the BIOS area or kernel text. Callers don't always get this
 * right (again, ioremap() on BIOS memory is not uncommon) so this function
 * checks and fixes these known static required protection bits.
 */
T
Thomas Gleixner 已提交
254 255
static inline pgprot_t static_protections(pgprot_t prot, unsigned long address,
				   unsigned long pfn)
256 257 258
{
	pgprot_t forbidden = __pgprot(0);

I
Ingo Molnar 已提交
259
	/*
260 261
	 * The BIOS area between 640k and 1Mb needs to be executable for
	 * PCI BIOS based config access (CONFIG_PCI_GOBIOS) support.
I
Ingo Molnar 已提交
262
	 */
T
Thomas Gleixner 已提交
263
	if (within(pfn, BIOS_BEGIN >> PAGE_SHIFT, BIOS_END >> PAGE_SHIFT))
264 265 266 267
		pgprot_val(forbidden) |= _PAGE_NX;

	/*
	 * The kernel text needs to be executable for obvious reasons
T
Thomas Gleixner 已提交
268 269
	 * Does not cover __inittext since that is gone later on. On
	 * 64bit we do not enforce !NX on the low mapping
270 271 272
	 */
	if (within(address, (unsigned long)_text, (unsigned long)_etext))
		pgprot_val(forbidden) |= _PAGE_NX;
273 274

	/*
T
Thomas Gleixner 已提交
275 276
	 * The .rodata section needs to be read-only. Using the pfn
	 * catches all aliases.
277
	 */
T
Thomas Gleixner 已提交
278 279
	if (within(pfn, __pa((unsigned long)__start_rodata) >> PAGE_SHIFT,
		   __pa((unsigned long)__end_rodata) >> PAGE_SHIFT))
280
		pgprot_val(forbidden) |= _PAGE_RW;
281

282
#if defined(CONFIG_X86_64) && defined(CONFIG_DEBUG_RODATA)
283
	/*
284 285 286 287
	 * Once the kernel maps the text as RO (kernel_set_to_readonly is set),
	 * kernel text mappings for the large page aligned text, rodata sections
	 * will be always read-only. For the kernel identity mappings covering
	 * the holes caused by this alignment can be anything that user asks.
288 289 290 291
	 *
	 * This will preserve the large page mappings for kernel text/data
	 * at no extra cost.
	 */
292 293
	if (kernel_set_to_readonly &&
	    within(address, (unsigned long)_text,
294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
		   (unsigned long)__end_rodata_hpage_align)) {
		unsigned int level;

		/*
		 * Don't enforce the !RW mapping for the kernel text mapping,
		 * if the current mapping is already using small page mapping.
		 * No need to work hard to preserve large page mappings in this
		 * case.
		 *
		 * This also fixes the Linux Xen paravirt guest boot failure
		 * (because of unexpected read-only mappings for kernel identity
		 * mappings). In this paravirt guest case, the kernel text
		 * mapping and the kernel identity mapping share the same
		 * page-table pages. Thus we can't really use different
		 * protections for the kernel text and identity mappings. Also,
		 * these shared mappings are made of small page mappings.
		 * Thus this don't enforce !RW mapping for small page kernel
		 * text mapping logic will help Linux Xen parvirt guest boot
		 * aswell.
		 */
		if (lookup_address(address, &level) && (level != PG_LEVEL_4K))
			pgprot_val(forbidden) |= _PAGE_RW;
	}
317 318
#endif

319
	prot = __pgprot(pgprot_val(prot) & ~pgprot_val(forbidden));
I
Ingo Molnar 已提交
320 321 322 323

	return prot;
}

T
Thomas Gleixner 已提交
324 325 326 327 328 329 330 331
/*
 * Lookup the page table entry for a virtual address. Return a pointer
 * to the entry and the level of the mapping.
 *
 * Note: We return pud and pmd either when the entry is marked large
 * or when the present bit is not set. Otherwise we would return a
 * pointer to a nonexisting mapping.
 */
332
pte_t *lookup_address(unsigned long address, unsigned int *level)
333
{
L
Linus Torvalds 已提交
334 335 336
	pgd_t *pgd = pgd_offset_k(address);
	pud_t *pud;
	pmd_t *pmd;
337

T
Thomas Gleixner 已提交
338 339
	*level = PG_LEVEL_NONE;

L
Linus Torvalds 已提交
340 341
	if (pgd_none(*pgd))
		return NULL;
I
Ingo Molnar 已提交
342

L
Linus Torvalds 已提交
343 344 345
	pud = pud_offset(pgd, address);
	if (pud_none(*pud))
		return NULL;
346 347 348 349 350

	*level = PG_LEVEL_1G;
	if (pud_large(*pud) || !pud_present(*pud))
		return (pte_t *)pud;

L
Linus Torvalds 已提交
351 352 353
	pmd = pmd_offset(pud, address);
	if (pmd_none(*pmd))
		return NULL;
T
Thomas Gleixner 已提交
354 355

	*level = PG_LEVEL_2M;
T
Thomas Gleixner 已提交
356
	if (pmd_large(*pmd) || !pmd_present(*pmd))
L
Linus Torvalds 已提交
357 358
		return (pte_t *)pmd;

T
Thomas Gleixner 已提交
359
	*level = PG_LEVEL_4K;
I
Ingo Molnar 已提交
360

361 362
	return pte_offset_kernel(pmd, address);
}
363
EXPORT_SYMBOL_GPL(lookup_address);
364

I
Ingo Molnar 已提交
365 366 367
/*
 * Set the new pmd in all the pgds we know about:
 */
I
Ingo Molnar 已提交
368
static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
369 370 371
{
	/* change init_mm */
	set_pte_atomic(kpte, pte);
372
#ifdef CONFIG_X86_32
373
	if (!SHARED_KERNEL_PMD) {
374 375
		struct page *page;

376
		list_for_each_entry(page, &pgd_list, lru) {
377 378 379 380 381 382 383 384 385
			pgd_t *pgd;
			pud_t *pud;
			pmd_t *pmd;

			pgd = (pgd_t *)page_address(page) + pgd_index(address);
			pud = pud_offset(pgd, address);
			pmd = pmd_offset(pud, address);
			set_pte_atomic((pte_t *)pmd, pte);
		}
L
Linus Torvalds 已提交
386
	}
387
#endif
L
Linus Torvalds 已提交
388 389
}

I
Ingo Molnar 已提交
390 391 392
static int
try_preserve_large_page(pte_t *kpte, unsigned long address,
			struct cpa_data *cpa)
393
{
T
Thomas Gleixner 已提交
394
	unsigned long nextpage_addr, numpages, pmask, psize, flags, addr, pfn;
395 396
	pte_t new_pte, old_pte, *tmp;
	pgprot_t old_prot, new_prot;
397
	int i, do_split = 1;
398
	unsigned int level;
399

400 401 402
	if (cpa->force_split)
		return 1;

403 404 405 406 407 408 409 410 411 412 413
	spin_lock_irqsave(&pgd_lock, flags);
	/*
	 * Check for races, another CPU might have split this page
	 * up already:
	 */
	tmp = lookup_address(address, &level);
	if (tmp != kpte)
		goto out_unlock;

	switch (level) {
	case PG_LEVEL_2M:
414 415
		psize = PMD_PAGE_SIZE;
		pmask = PMD_PAGE_MASK;
416
		break;
417
#ifdef CONFIG_X86_64
418
	case PG_LEVEL_1G:
419 420
		psize = PUD_PAGE_SIZE;
		pmask = PUD_PAGE_MASK;
421 422
		break;
#endif
423
	default:
I
Ingo Molnar 已提交
424
		do_split = -EINVAL;
425 426 427 428 429 430 431 432 433
		goto out_unlock;
	}

	/*
	 * Calculate the number of pages, which fit into this large
	 * page starting at address:
	 */
	nextpage_addr = (address + psize) & pmask;
	numpages = (nextpage_addr - address) >> PAGE_SHIFT;
434 435
	if (numpages < cpa->numpages)
		cpa->numpages = numpages;
436 437 438 439 440 441 442 443 444

	/*
	 * We are safe now. Check whether the new pgprot is the same:
	 */
	old_pte = *kpte;
	old_prot = new_prot = pte_pgprot(old_pte);

	pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
	pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
T
Thomas Gleixner 已提交
445 446 447 448 449 450 451 452 453

	/*
	 * old_pte points to the large page base address. So we need
	 * to add the offset of the virtual address:
	 */
	pfn = pte_pfn(old_pte) + ((address & (psize - 1)) >> PAGE_SHIFT);
	cpa->pfn = pfn;

	new_prot = static_protections(new_prot, address, pfn);
454

455 456 457 458 459 460
	/*
	 * We need to check the full range, whether
	 * static_protection() requires a different pgprot for one of
	 * the pages in the range we try to preserve:
	 */
	addr = address + PAGE_SIZE;
T
Thomas Gleixner 已提交
461
	pfn++;
462
	for (i = 1; i < cpa->numpages; i++, addr += PAGE_SIZE, pfn++) {
T
Thomas Gleixner 已提交
463
		pgprot_t chk_prot = static_protections(new_prot, addr, pfn);
464 465 466 467 468

		if (pgprot_val(chk_prot) != pgprot_val(new_prot))
			goto out_unlock;
	}

469 470 471 472 473
	/*
	 * If there are no changes, return. maxpages has been updated
	 * above:
	 */
	if (pgprot_val(new_prot) == pgprot_val(old_prot)) {
I
Ingo Molnar 已提交
474
		do_split = 0;
475 476 477 478 479 480 481 482 483 484 485
		goto out_unlock;
	}

	/*
	 * We need to change the attributes. Check, whether we can
	 * change the large page in one go. We request a split, when
	 * the address is not aligned and the number of pages is
	 * smaller than the number of pages in the large page. Note
	 * that we limited the number of possible pages already to
	 * the number of pages in the large page.
	 */
486
	if (address == (nextpage_addr - psize) && cpa->numpages == numpages) {
487 488 489 490 491 492
		/*
		 * The address is aligned and the number of pages
		 * covers the full page.
		 */
		new_pte = pfn_pte(pte_pfn(old_pte), canon_pgprot(new_prot));
		__set_pmd_pte(kpte, address, new_pte);
493
		cpa->flags |= CPA_FLUSHTLB;
I
Ingo Molnar 已提交
494
		do_split = 0;
495 496 497 498
	}

out_unlock:
	spin_unlock_irqrestore(&pgd_lock, flags);
I
Ingo Molnar 已提交
499

I
Ingo Molnar 已提交
500
	return do_split;
501 502
}

503
static int split_large_page(pte_t *kpte, unsigned long address)
504
{
T
Thomas Gleixner 已提交
505
	unsigned long flags, pfn, pfninc = 1;
I
Ingo Molnar 已提交
506
	unsigned int i, level;
507
	pte_t *pbase, *tmp;
I
Ingo Molnar 已提交
508
	pgprot_t ref_prot;
509 510 511 512
	struct page *base;

	if (!debug_pagealloc)
		spin_unlock(&cpa_lock);
513
	base = alloc_pages(GFP_KERNEL | __GFP_NOTRACK, 0);
514 515
	if (!debug_pagealloc)
		spin_lock(&cpa_lock);
S
Suresh Siddha 已提交
516 517
	if (!base)
		return -ENOMEM;
518

T
Thomas Gleixner 已提交
519
	spin_lock_irqsave(&pgd_lock, flags);
520 521 522 523 524
	/*
	 * Check for races, another CPU might have split this page
	 * up for us already:
	 */
	tmp = lookup_address(address, &level);
I
Ingo Molnar 已提交
525
	if (tmp != kpte)
526 527 528
		goto out_unlock;

	pbase = (pte_t *)page_address(base);
529
	paravirt_alloc_pte(&init_mm, page_to_pfn(base));
T
Thomas Gleixner 已提交
530
	ref_prot = pte_pgprot(pte_clrhuge(*kpte));
531 532 533 534 535 536 537
	/*
	 * If we ever want to utilize the PAT bit, we need to
	 * update this function to make sure it's converted from
	 * bit 12 to bit 7 when we cross from the 2MB level to
	 * the 4K level:
	 */
	WARN_ON_ONCE(pgprot_val(ref_prot) & _PAGE_PAT_LARGE);
538

539 540 541 542 543 544 545
#ifdef CONFIG_X86_64
	if (level == PG_LEVEL_1G) {
		pfninc = PMD_PAGE_SIZE >> PAGE_SHIFT;
		pgprot_val(ref_prot) |= _PAGE_PSE;
	}
#endif

546 547 548 549
	/*
	 * Get the target pfn from the original entry:
	 */
	pfn = pte_pfn(*kpte);
550
	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
551
		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
552

553
	if (address >= (unsigned long)__va(0) &&
554 555 556 557 558
		address < (unsigned long)__va(max_low_pfn_mapped << PAGE_SHIFT))
		split_page_count(level);

#ifdef CONFIG_X86_64
	if (address >= (unsigned long)__va(1UL<<32) &&
559 560
		address < (unsigned long)__va(max_pfn_mapped << PAGE_SHIFT))
		split_page_count(level);
561
#endif
562

563
	/*
564
	 * Install the new, split up pagetable.
565
	 *
566 567 568
	 * We use the standard kernel pagetable protections for the new
	 * pagetable protections, the actual ptes set above control the
	 * primary protection behavior:
569
	 */
570
	__set_pmd_pte(kpte, address, mk_pte(base, __pgprot(_KERNPG_TABLE)));
571 572 573 574 575 576 577 578 579 580 581

	/*
	 * Intel Atom errata AAH41 workaround.
	 *
	 * The real fix should be in hw or in a microcode update, but
	 * we also probabilistically try to reduce the window of having
	 * a large TLB mixed with 4K TLBs while instruction fetches are
	 * going on.
	 */
	__flush_tlb_all();

582 583 584
	base = NULL;

out_unlock:
T
Thomas Gleixner 已提交
585 586 587 588
	/*
	 * If we dropped out via the lookup_address check under
	 * pgd_lock then stick the page back into the pool:
	 */
S
Suresh Siddha 已提交
589 590
	if (base)
		__free_page(base);
I
Ingo Molnar 已提交
591
	spin_unlock_irqrestore(&pgd_lock, flags);
592 593 594 595

	return 0;
}

596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625
static int __cpa_process_fault(struct cpa_data *cpa, unsigned long vaddr,
			       int primary)
{
	/*
	 * Ignore all non primary paths.
	 */
	if (!primary)
		return 0;

	/*
	 * Ignore the NULL PTE for kernel identity mapping, as it is expected
	 * to have holes.
	 * Also set numpages to '1' indicating that we processed cpa req for
	 * one virtual address page and its pfn. TBD: numpages can be set based
	 * on the initial value and the level returned by lookup_address().
	 */
	if (within(vaddr, PAGE_OFFSET,
		   PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT))) {
		cpa->numpages = 1;
		cpa->pfn = __pa(vaddr) >> PAGE_SHIFT;
		return 0;
	} else {
		WARN(1, KERN_WARNING "CPA: called for zero pte. "
			"vaddr = %lx cpa->vaddr = %lx\n", vaddr,
			*cpa->vaddr);

		return -EFAULT;
	}
}

T
Thomas Gleixner 已提交
626
static int __change_page_attr(struct cpa_data *cpa, int primary)
627
{
628
	unsigned long address;
629 630
	int do_split, err;
	unsigned int level;
T
Thomas Gleixner 已提交
631
	pte_t *kpte, old_pte;
L
Linus Torvalds 已提交
632

633 634 635 636 637 638
	if (cpa->flags & CPA_PAGES_ARRAY) {
		struct page *page = cpa->pages[cpa->curpage];
		if (unlikely(PageHighMem(page)))
			return 0;
		address = (unsigned long)page_address(page);
	} else if (cpa->flags & CPA_ARRAY)
639 640 641
		address = cpa->vaddr[cpa->curpage];
	else
		address = *cpa->vaddr;
642
repeat:
643
	kpte = lookup_address(address, &level);
L
Linus Torvalds 已提交
644
	if (!kpte)
645
		return __cpa_process_fault(cpa, address, primary);
T
Thomas Gleixner 已提交
646 647

	old_pte = *kpte;
648 649
	if (!pte_val(old_pte))
		return __cpa_process_fault(cpa, address, primary);
650

T
Thomas Gleixner 已提交
651
	if (level == PG_LEVEL_4K) {
T
Thomas Gleixner 已提交
652
		pte_t new_pte;
653
		pgprot_t new_prot = pte_pgprot(old_pte);
T
Thomas Gleixner 已提交
654
		unsigned long pfn = pte_pfn(old_pte);
I
Ingo Molnar 已提交
655

T
Thomas Gleixner 已提交
656 657
		pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
		pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
I
Ingo Molnar 已提交
658

T
Thomas Gleixner 已提交
659
		new_prot = static_protections(new_prot, address, pfn);
I
Ingo Molnar 已提交
660

661 662 663 664 665
		/*
		 * We need to keep the pfn from the existing PTE,
		 * after all we're only going to change it's attributes
		 * not the memory it points to
		 */
T
Thomas Gleixner 已提交
666 667
		new_pte = pfn_pte(pfn, canon_pgprot(new_prot));
		cpa->pfn = pfn;
668 669 670 671 672
		/*
		 * Do we really change anything ?
		 */
		if (pte_val(old_pte) != pte_val(new_pte)) {
			set_pte_atomic(kpte, new_pte);
673
			cpa->flags |= CPA_FLUSHTLB;
674
		}
675
		cpa->numpages = 1;
676
		return 0;
L
Linus Torvalds 已提交
677
	}
678 679 680 681 682

	/*
	 * Check, whether we can keep the large page intact
	 * and just change the pte:
	 */
I
Ingo Molnar 已提交
683
	do_split = try_preserve_large_page(kpte, address, cpa);
684 685
	/*
	 * When the range fits into the existing large page,
686
	 * return. cp->numpages and cpa->tlbflush have been updated in
687 688
	 * try_large_page:
	 */
I
Ingo Molnar 已提交
689 690
	if (do_split <= 0)
		return do_split;
691 692 693 694

	/*
	 * We have to split the large page:
	 */
I
Ingo Molnar 已提交
695 696
	err = split_large_page(kpte, address);
	if (!err) {
697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715
		/*
	 	 * Do a global flush tlb after splitting the large page
	 	 * and before we do the actual change page attribute in the PTE.
	 	 *
	 	 * With out this, we violate the TLB application note, that says
	 	 * "The TLBs may contain both ordinary and large-page
		 *  translations for a 4-KByte range of linear addresses. This
		 *  may occur if software modifies the paging structures so that
		 *  the page size used for the address range changes. If the two
		 *  translations differ with respect to page frame or attributes
		 *  (e.g., permissions), processor behavior is undefined and may
		 *  be implementation-specific."
	 	 *
	 	 * We do this global tlb flush inside the cpa_lock, so that we
		 * don't allow any other cpu, with stale tlb entries change the
		 * page attribute in parallel, that also falls into the
		 * just split large page entry.
	 	 */
		flush_tlb_all();
I
Ingo Molnar 已提交
716 717
		goto repeat;
	}
I
Ingo Molnar 已提交
718

I
Ingo Molnar 已提交
719
	return err;
720
}
L
Linus Torvalds 已提交
721

T
Thomas Gleixner 已提交
722 723 724
static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias);

static int cpa_process_alias(struct cpa_data *cpa)
L
Linus Torvalds 已提交
725
{
T
Thomas Gleixner 已提交
726
	struct cpa_data alias_cpa;
T
Tejun Heo 已提交
727
	unsigned long laddr = (unsigned long)__va(cpa->pfn << PAGE_SHIFT);
728
	unsigned long vaddr;
T
Tejun Heo 已提交
729
	int ret;
730

Y
Yinghai Lu 已提交
731
	if (cpa->pfn >= max_pfn_mapped)
T
Thomas Gleixner 已提交
732
		return 0;
733

734
#ifdef CONFIG_X86_64
Y
Yinghai Lu 已提交
735
	if (cpa->pfn >= max_low_pfn_mapped && cpa->pfn < (1UL<<(32-PAGE_SHIFT)))
736 737
		return 0;
#endif
738 739 740 741
	/*
	 * No need to redo, when the primary call touched the direct
	 * mapping already:
	 */
742 743 744 745 746 747
	if (cpa->flags & CPA_PAGES_ARRAY) {
		struct page *page = cpa->pages[cpa->curpage];
		if (unlikely(PageHighMem(page)))
			return 0;
		vaddr = (unsigned long)page_address(page);
	} else if (cpa->flags & CPA_ARRAY)
748 749 750 751 752
		vaddr = cpa->vaddr[cpa->curpage];
	else
		vaddr = *cpa->vaddr;

	if (!(within(vaddr, PAGE_OFFSET,
753
		    PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) {
754

755
		alias_cpa = *cpa;
T
Tejun Heo 已提交
756
		alias_cpa.vaddr = &laddr;
757
		alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
758

759
		ret = __change_page_attr_set_clr(&alias_cpa, 0);
T
Tejun Heo 已提交
760 761
		if (ret)
			return ret;
762
	}
763 764

#ifdef CONFIG_X86_64
A
Arjan van de Ven 已提交
765
	/*
T
Tejun Heo 已提交
766 767
	 * If the primary call didn't touch the high mapping already
	 * and the physical address is inside the kernel map, we need
768
	 * to touch the high mapped kernel as well:
A
Arjan van de Ven 已提交
769
	 */
T
Tejun Heo 已提交
770 771 772 773 774 775 776
	if (!within(vaddr, (unsigned long)_text, _brk_end) &&
	    within(cpa->pfn, highmap_start_pfn(), highmap_end_pfn())) {
		unsigned long temp_cpa_vaddr = (cpa->pfn << PAGE_SHIFT) +
					       __START_KERNEL_map - phys_base;
		alias_cpa = *cpa;
		alias_cpa.vaddr = &temp_cpa_vaddr;
		alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
T
Thomas Gleixner 已提交
777

T
Tejun Heo 已提交
778 779 780 781 782 783
		/*
		 * The high mapping range is imprecise, so ignore the
		 * return value.
		 */
		__change_page_attr_set_clr(&alias_cpa, 0);
	}
A
Arjan van de Ven 已提交
784
#endif
T
Tejun Heo 已提交
785 786

	return 0;
L
Linus Torvalds 已提交
787 788
}

T
Thomas Gleixner 已提交
789
static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias)
790
{
791
	int ret, numpages = cpa->numpages;
792

793 794 795 796 797
	while (numpages) {
		/*
		 * Store the remaining nr of pages for the large page
		 * preservation check.
		 */
798
		cpa->numpages = numpages;
799
		/* for array changes, we can't use large page */
800
		if (cpa->flags & (CPA_ARRAY | CPA_PAGES_ARRAY))
801
			cpa->numpages = 1;
T
Thomas Gleixner 已提交
802

803 804
		if (!debug_pagealloc)
			spin_lock(&cpa_lock);
T
Thomas Gleixner 已提交
805
		ret = __change_page_attr(cpa, checkalias);
806 807
		if (!debug_pagealloc)
			spin_unlock(&cpa_lock);
808 809 810
		if (ret)
			return ret;

T
Thomas Gleixner 已提交
811 812 813 814 815 816
		if (checkalias) {
			ret = cpa_process_alias(cpa);
			if (ret)
				return ret;
		}

817 818 819 820 821
		/*
		 * Adjust the number of pages with the result of the
		 * CPA operation. Either a large page has been
		 * preserved or a single page update happened.
		 */
822 823
		BUG_ON(cpa->numpages > numpages);
		numpages -= cpa->numpages;
824
		if (cpa->flags & (CPA_PAGES_ARRAY | CPA_ARRAY))
825 826 827 828
			cpa->curpage++;
		else
			*cpa->vaddr += cpa->numpages * PAGE_SIZE;

829
	}
830 831 832
	return 0;
}

833 834 835 836 837 838
static inline int cache_attr(pgprot_t attr)
{
	return pgprot_val(attr) &
		(_PAGE_PAT | _PAGE_PAT_LARGE | _PAGE_PWT | _PAGE_PCD);
}

839
static int change_page_attr_set_clr(unsigned long *addr, int numpages,
840
				    pgprot_t mask_set, pgprot_t mask_clr,
841 842
				    int force_split, int in_flag,
				    struct page **pages)
843
{
T
Thomas Gleixner 已提交
844
	struct cpa_data cpa;
845
	int ret, cache, checkalias;
846
	unsigned long baddr = 0;
847 848 849 850 851 852 853

	/*
	 * Check, if we are requested to change a not supported
	 * feature:
	 */
	mask_set = canon_pgprot(mask_set);
	mask_clr = canon_pgprot(mask_clr);
854
	if (!pgprot_val(mask_set) && !pgprot_val(mask_clr) && !force_split)
855 856
		return 0;

857
	/* Ensure we are PAGE_SIZE aligned */
858
	if (in_flag & CPA_ARRAY) {
859 860 861 862 863 864 865
		int i;
		for (i = 0; i < numpages; i++) {
			if (addr[i] & ~PAGE_MASK) {
				addr[i] &= PAGE_MASK;
				WARN_ON_ONCE(1);
			}
		}
866 867 868 869 870 871 872 873 874 875 876 877
	} else if (!(in_flag & CPA_PAGES_ARRAY)) {
		/*
		 * in_flag of CPA_PAGES_ARRAY implies it is aligned.
		 * No need to cehck in that case
		 */
		if (*addr & ~PAGE_MASK) {
			*addr &= PAGE_MASK;
			/*
			 * People should not be passing in unaligned addresses:
			 */
			WARN_ON_ONCE(1);
		}
878 879 880 881 882
		/*
		 * Save address for cache flush. *addr is modified in the call
		 * to __change_page_attr_set_clr() below.
		 */
		baddr = *addr;
883 884
	}

885 886 887
	/* Must avoid aliasing mappings in the highmem code */
	kmap_flush_unused();

N
Nick Piggin 已提交
888 889
	vm_unmap_aliases();

T
Thomas Gleixner 已提交
890
	cpa.vaddr = addr;
891
	cpa.pages = pages;
T
Thomas Gleixner 已提交
892 893 894
	cpa.numpages = numpages;
	cpa.mask_set = mask_set;
	cpa.mask_clr = mask_clr;
895 896
	cpa.flags = 0;
	cpa.curpage = 0;
897
	cpa.force_split = force_split;
T
Thomas Gleixner 已提交
898

899 900
	if (in_flag & (CPA_ARRAY | CPA_PAGES_ARRAY))
		cpa.flags |= in_flag;
901

902 903 904 905
	/* No alias checking for _NX bit modifications */
	checkalias = (pgprot_val(mask_set) | pgprot_val(mask_clr)) != _PAGE_NX;

	ret = __change_page_attr_set_clr(&cpa, checkalias);
906

907 908 909
	/*
	 * Check whether we really changed something:
	 */
910
	if (!(cpa.flags & CPA_FLUSHTLB))
911
		goto out;
912

913 914 915 916 917 918
	/*
	 * No need to flush, when we did not set any of the caching
	 * attributes:
	 */
	cache = cache_attr(mask_set);

919 920 921
	/*
	 * On success we use clflush, when the CPU supports it to
	 * avoid the wbindv. If the CPU does not support it and in the
922
	 * error case we fall back to cpa_flush_all (which uses
923 924
	 * wbindv):
	 */
925
	if (!ret && cpu_has_clflush) {
926 927 928 929
		if (cpa.flags & (CPA_PAGES_ARRAY | CPA_ARRAY)) {
			cpa_flush_array(addr, numpages, cache,
					cpa.flags, pages);
		} else
930
			cpa_flush_range(baddr, numpages, cache);
931
	} else
932
		cpa_flush_all(cache);
933

934
out:
935 936 937
	return ret;
}

938 939
static inline int change_page_attr_set(unsigned long *addr, int numpages,
				       pgprot_t mask, int array)
940
{
941
	return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0), 0,
942
		(array ? CPA_ARRAY : 0), NULL);
943 944
}

945 946
static inline int change_page_attr_clear(unsigned long *addr, int numpages,
					 pgprot_t mask, int array)
947
{
948
	return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask, 0,
949
		(array ? CPA_ARRAY : 0), NULL);
950 951
}

952 953 954 955 956 957 958 959 960 961 962 963 964 965
static inline int cpa_set_pages_array(struct page **pages, int numpages,
				       pgprot_t mask)
{
	return change_page_attr_set_clr(NULL, numpages, mask, __pgprot(0), 0,
		CPA_PAGES_ARRAY, pages);
}

static inline int cpa_clear_pages_array(struct page **pages, int numpages,
					 pgprot_t mask)
{
	return change_page_attr_set_clr(NULL, numpages, __pgprot(0), mask, 0,
		CPA_PAGES_ARRAY, pages);
}

966
int _set_memory_uc(unsigned long addr, int numpages)
967
{
968 969 970
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
971 972
	return change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
973
}
974 975 976

int set_memory_uc(unsigned long addr, int numpages)
{
977 978
	int ret;

979 980 981
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
982 983 984 985 986 987 988 989 990 991
	ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
			    _PAGE_CACHE_UC_MINUS, NULL);
	if (ret)
		goto out_err;

	ret = _set_memory_uc(addr, numpages);
	if (ret)
		goto out_free;

	return 0;
992

993 994 995 996
out_free:
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
out_err:
	return ret;
997
}
998 999
EXPORT_SYMBOL(set_memory_uc);

1000 1001
int _set_memory_array(unsigned long *addr, int addrinarray,
		unsigned long new_type)
1002
{
1003 1004 1005
	int i, j;
	int ret;

1006 1007 1008 1009
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
	for (i = 0; i < addrinarray; i++) {
1010
		ret = reserve_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE,
1011
					new_type, NULL);
1012 1013
		if (ret)
			goto out_free;
1014 1015
	}

1016
	ret = change_page_attr_set(addr, addrinarray,
1017
				    __pgprot(_PAGE_CACHE_UC_MINUS), 1);
1018 1019 1020 1021 1022 1023

	if (!ret && new_type == _PAGE_CACHE_WC)
		ret = change_page_attr_set_clr(addr, addrinarray,
					       __pgprot(_PAGE_CACHE_WC),
					       __pgprot(_PAGE_CACHE_MASK),
					       0, CPA_ARRAY, NULL);
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
	if (ret)
		goto out_free;

	return 0;

out_free:
	for (j = 0; j < i; j++)
		free_memtype(__pa(addr[j]), __pa(addr[j]) + PAGE_SIZE);

	return ret;
1034
}
1035 1036 1037 1038 1039

int set_memory_array_uc(unsigned long *addr, int addrinarray)
{
	return _set_memory_array(addr, addrinarray, _PAGE_CACHE_UC_MINUS);
}
1040 1041
EXPORT_SYMBOL(set_memory_array_uc);

1042 1043 1044 1045 1046 1047
int set_memory_array_wc(unsigned long *addr, int addrinarray)
{
	return _set_memory_array(addr, addrinarray, _PAGE_CACHE_WC);
}
EXPORT_SYMBOL(set_memory_array_wc);

1048 1049
int _set_memory_wc(unsigned long addr, int numpages)
{
1050
	int ret;
1051 1052
	unsigned long addr_copy = addr;

1053 1054 1055
	ret = change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
	if (!ret) {
1056 1057 1058 1059
		ret = change_page_attr_set_clr(&addr_copy, numpages,
					       __pgprot(_PAGE_CACHE_WC),
					       __pgprot(_PAGE_CACHE_MASK),
					       0, 0, NULL);
1060 1061
	}
	return ret;
1062 1063 1064 1065
}

int set_memory_wc(unsigned long addr, int numpages)
{
1066 1067
	int ret;

1068
	if (!pat_enabled)
1069 1070
		return set_memory_uc(addr, numpages);

1071 1072 1073 1074
	ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
		_PAGE_CACHE_WC, NULL);
	if (ret)
		goto out_err;
1075

1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
	ret = _set_memory_wc(addr, numpages);
	if (ret)
		goto out_free;

	return 0;

out_free:
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
out_err:
	return ret;
1086 1087 1088
}
EXPORT_SYMBOL(set_memory_wc);

1089
int _set_memory_wb(unsigned long addr, int numpages)
1090
{
1091 1092
	return change_page_attr_clear(&addr, numpages,
				      __pgprot(_PAGE_CACHE_MASK), 0);
1093
}
1094 1095 1096

int set_memory_wb(unsigned long addr, int numpages)
{
1097 1098 1099 1100 1101 1102
	int ret;

	ret = _set_memory_wb(addr, numpages);
	if (ret)
		return ret;

1103
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
1104
	return 0;
1105
}
1106 1107
EXPORT_SYMBOL(set_memory_wb);

1108 1109 1110
int set_memory_array_wb(unsigned long *addr, int addrinarray)
{
	int i;
1111 1112 1113 1114
	int ret;

	ret = change_page_attr_clear(addr, addrinarray,
				      __pgprot(_PAGE_CACHE_MASK), 1);
1115 1116
	if (ret)
		return ret;
1117

1118 1119
	for (i = 0; i < addrinarray; i++)
		free_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE);
1120

1121
	return 0;
1122 1123 1124
}
EXPORT_SYMBOL(set_memory_array_wb);

1125 1126
int set_memory_x(unsigned long addr, int numpages)
{
1127 1128 1129
	if (!(__supported_pte_mask & _PAGE_NX))
		return 0;

1130
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_NX), 0);
1131 1132 1133 1134 1135
}
EXPORT_SYMBOL(set_memory_x);

int set_memory_nx(unsigned long addr, int numpages)
{
1136 1137 1138
	if (!(__supported_pte_mask & _PAGE_NX))
		return 0;

1139
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_NX), 0);
1140 1141 1142 1143 1144
}
EXPORT_SYMBOL(set_memory_nx);

int set_memory_ro(unsigned long addr, int numpages)
{
1145
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_RW), 0);
1146
}
1147
EXPORT_SYMBOL_GPL(set_memory_ro);
1148 1149 1150

int set_memory_rw(unsigned long addr, int numpages)
{
1151
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_RW), 0);
1152
}
1153
EXPORT_SYMBOL_GPL(set_memory_rw);
I
Ingo Molnar 已提交
1154 1155 1156

int set_memory_np(unsigned long addr, int numpages)
{
1157
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_PRESENT), 0);
I
Ingo Molnar 已提交
1158
}
1159

1160 1161
int set_memory_4k(unsigned long addr, int numpages)
{
1162
	return change_page_attr_set_clr(&addr, numpages, __pgprot(0),
1163
					__pgprot(0), 1, 0, NULL);
1164 1165
}

1166 1167 1168 1169
int set_pages_uc(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1170
	return set_memory_uc(addr, numpages);
1171 1172 1173
}
EXPORT_SYMBOL(set_pages_uc);

1174 1175
static int _set_pages_array(struct page **pages, int addrinarray,
		unsigned long new_type)
1176 1177 1178 1179 1180
{
	unsigned long start;
	unsigned long end;
	int i;
	int free_idx;
1181
	int ret;
1182 1183

	for (i = 0; i < addrinarray; i++) {
1184 1185 1186
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1187
		end = start + PAGE_SIZE;
1188
		if (reserve_memtype(start, end, new_type, NULL))
1189 1190 1191
			goto err_out;
	}

1192 1193 1194 1195 1196 1197 1198 1199 1200 1201
	ret = cpa_set_pages_array(pages, addrinarray,
			__pgprot(_PAGE_CACHE_UC_MINUS));
	if (!ret && new_type == _PAGE_CACHE_WC)
		ret = change_page_attr_set_clr(NULL, addrinarray,
					       __pgprot(_PAGE_CACHE_WC),
					       __pgprot(_PAGE_CACHE_MASK),
					       0, CPA_PAGES_ARRAY, pages);
	if (ret)
		goto err_out;
	return 0; /* Success */
1202 1203 1204
err_out:
	free_idx = i;
	for (i = 0; i < free_idx; i++) {
1205 1206 1207
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1208 1209 1210 1211 1212
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}
	return -EINVAL;
}
1213 1214 1215 1216 1217

int set_pages_array_uc(struct page **pages, int addrinarray)
{
	return _set_pages_array(pages, addrinarray, _PAGE_CACHE_UC_MINUS);
}
1218 1219
EXPORT_SYMBOL(set_pages_array_uc);

1220 1221 1222 1223 1224 1225
int set_pages_array_wc(struct page **pages, int addrinarray)
{
	return _set_pages_array(pages, addrinarray, _PAGE_CACHE_WC);
}
EXPORT_SYMBOL(set_pages_array_wc);

1226 1227 1228 1229
int set_pages_wb(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1230
	return set_memory_wb(addr, numpages);
1231 1232 1233
}
EXPORT_SYMBOL(set_pages_wb);

1234 1235 1236 1237 1238 1239 1240 1241 1242
int set_pages_array_wb(struct page **pages, int addrinarray)
{
	int retval;
	unsigned long start;
	unsigned long end;
	int i;

	retval = cpa_clear_pages_array(pages, addrinarray,
			__pgprot(_PAGE_CACHE_MASK));
1243 1244
	if (retval)
		return retval;
1245 1246

	for (i = 0; i < addrinarray; i++) {
1247 1248 1249
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1250 1251 1252 1253
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}

1254
	return 0;
1255 1256 1257
}
EXPORT_SYMBOL(set_pages_array_wb);

1258 1259 1260 1261
int set_pages_x(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1262
	return set_memory_x(addr, numpages);
1263 1264 1265 1266 1267 1268 1269
}
EXPORT_SYMBOL(set_pages_x);

int set_pages_nx(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1270
	return set_memory_nx(addr, numpages);
1271 1272 1273 1274 1275 1276 1277
}
EXPORT_SYMBOL(set_pages_nx);

int set_pages_ro(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1278
	return set_memory_ro(addr, numpages);
1279 1280 1281 1282 1283
}

int set_pages_rw(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);
1284

T
Thomas Gleixner 已提交
1285
	return set_memory_rw(addr, numpages);
I
Ingo Molnar 已提交
1286 1287
}

L
Linus Torvalds 已提交
1288
#ifdef CONFIG_DEBUG_PAGEALLOC
I
Ingo Molnar 已提交
1289 1290 1291

static int __set_pages_p(struct page *page, int numpages)
{
1292 1293
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1294 1295
				.numpages = numpages,
				.mask_set = __pgprot(_PAGE_PRESENT | _PAGE_RW),
1296 1297
				.mask_clr = __pgprot(0),
				.flags = 0};
1298

1299 1300 1301 1302 1303 1304 1305
	/*
	 * No alias checking needed for setting present flag. otherwise,
	 * we may need to break large pages for 64-bit kernel text
	 * mappings (this adds to complexity if we want to do this from
	 * atomic context especially). Let's keep it simple!
	 */
	return __change_page_attr_set_clr(&cpa, 0);
I
Ingo Molnar 已提交
1306 1307 1308 1309
}

static int __set_pages_np(struct page *page, int numpages)
{
1310 1311
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1312 1313
				.numpages = numpages,
				.mask_set = __pgprot(0),
1314 1315
				.mask_clr = __pgprot(_PAGE_PRESENT | _PAGE_RW),
				.flags = 0};
1316

1317 1318 1319 1320 1321 1322 1323
	/*
	 * No alias checking needed for setting not present flag. otherwise,
	 * we may need to break large pages for 64-bit kernel text
	 * mappings (this adds to complexity if we want to do this from
	 * atomic context especially). Let's keep it simple!
	 */
	return __change_page_attr_set_clr(&cpa, 0);
I
Ingo Molnar 已提交
1324 1325
}

L
Linus Torvalds 已提交
1326 1327 1328 1329
void kernel_map_pages(struct page *page, int numpages, int enable)
{
	if (PageHighMem(page))
		return;
1330
	if (!enable) {
1331 1332
		debug_check_no_locks_freed(page_address(page),
					   numpages * PAGE_SIZE);
1333
	}
1334

1335 1336 1337 1338 1339 1340
	/*
	 * If page allocator is not up yet then do not call c_p_a():
	 */
	if (!debug_pagealloc_enabled)
		return;

1341
	/*
I
Ingo Molnar 已提交
1342
	 * The return value is ignored as the calls cannot fail.
1343 1344
	 * Large pages for identity mappings are not used at boot time
	 * and hence no memory allocations during large page split.
L
Linus Torvalds 已提交
1345
	 */
I
Ingo Molnar 已提交
1346 1347 1348 1349
	if (enable)
		__set_pages_p(page, numpages);
	else
		__set_pages_np(page, numpages);
1350 1351

	/*
1352 1353
	 * We should perform an IPI and flush all tlbs,
	 * but that can deadlock->flush only current cpu:
L
Linus Torvalds 已提交
1354 1355
	 */
	__flush_tlb_all();
1356 1357
}

1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
#ifdef CONFIG_HIBERNATION

bool kernel_page_present(struct page *page)
{
	unsigned int level;
	pte_t *pte;

	if (PageHighMem(page))
		return false;

	pte = lookup_address((unsigned long)page_address(page), &level);
	return (pte_val(*pte) & _PAGE_PRESENT);
}

#endif /* CONFIG_HIBERNATION */

#endif /* CONFIG_DEBUG_PAGEALLOC */
1375 1376 1377 1378 1379 1380 1381 1382

/*
 * The testcases use internal knowledge of the implementation that shouldn't
 * be exposed to the rest of the kernel. Include these directly here.
 */
#ifdef CONFIG_CPA_DEBUG
#include "pageattr-test.c"
#endif