pageattr.c 31.2 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
#include <linux/sched.h>
L
Linus Torvalds 已提交
9
#include <linux/slab.h>
10
#include <linux/mm.h>
11
#include <linux/interrupt.h>
12 13
#include <linux/seq_file.h>
#include <linux/debugfs.h>
14
#include <linux/pfn.h>
15
#include <linux/percpu.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
}

148
static void __cpa_flush_all(void *arg)
T
Thomas Gleixner 已提交
149
{
150 151
	unsigned long cache = (unsigned long)arg;

T
Thomas Gleixner 已提交
152 153 154 155 156 157
	/*
	 * Flush all to work around Errata in early athlons regarding
	 * large page flushing.
	 */
	__flush_tlb_all();

158
	if (cache && boot_cpu_data.x86 >= 4)
T
Thomas Gleixner 已提交
159 160 161
		wbinvd();
}

162
static void cpa_flush_all(unsigned long cache)
T
Thomas Gleixner 已提交
163 164 165
{
	BUG_ON(irqs_disabled());

166
	on_each_cpu(__cpa_flush_all, (void *) cache, 1);
T
Thomas Gleixner 已提交
167 168
}

169 170 171 172 173 174 175 176 177 178
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();
}

179
static void cpa_flush_range(unsigned long start, int numpages, int cache)
180
{
I
Ingo Molnar 已提交
181 182 183
	unsigned int i, level;
	unsigned long addr;

184
	BUG_ON(irqs_disabled());
I
Ingo Molnar 已提交
185
	WARN_ON(PAGE_ALIGN(start) != start);
186

187
	on_each_cpu(__cpa_flush_range, NULL, 1);
188

189 190 191
	if (!cache)
		return;

T
Thomas Gleixner 已提交
192 193 194 195 196 197
	/*
	 * 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
198 199 200 201 202 203
	for (i = 0, addr = start; i < numpages; i++, addr += PAGE_SIZE) {
		pte_t *pte = lookup_address(addr, &level);

		/*
		 * Only flush present addresses:
		 */
204
		if (pte && (pte_val(*pte) & _PAGE_PRESENT))
I
Ingo Molnar 已提交
205 206
			clflush_cache_range((void *) addr, PAGE_SIZE);
	}
207 208
}

209 210
static void cpa_flush_array(unsigned long *start, int numpages, int cache,
			    int in_flags, struct page **pages)
211 212
{
	unsigned int i, level;
213
	unsigned long do_wbinvd = cache && numpages >= 1024; /* 4M threshold */
214 215 216

	BUG_ON(irqs_disabled());

217
	on_each_cpu(__cpa_flush_all, (void *) do_wbinvd, 1);
218

219
	if (!cache || do_wbinvd)
220 221 222 223 224 225 226 227
		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:
	 */
228 229 230 231 232 233 234 235 236 237
	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);
238 239 240 241 242

		/*
		 * Only flush present addresses:
		 */
		if (pte && (pte_val(*pte) & _PAGE_PRESENT))
243
			clflush_cache_range((void *)addr, PAGE_SIZE);
244 245 246
	}
}

247 248 249 250 251 252
/*
 * 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
253 254
static inline pgprot_t static_protections(pgprot_t prot, unsigned long address,
				   unsigned long pfn)
255 256 257
{
	pgprot_t forbidden = __pgprot(0);

I
Ingo Molnar 已提交
258
	/*
259 260
	 * 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
261
	 */
T
Thomas Gleixner 已提交
262
	if (within(pfn, BIOS_BEGIN >> PAGE_SHIFT, BIOS_END >> PAGE_SHIFT))
263 264 265 266
		pgprot_val(forbidden) |= _PAGE_NX;

	/*
	 * The kernel text needs to be executable for obvious reasons
T
Thomas Gleixner 已提交
267 268
	 * Does not cover __inittext since that is gone later on. On
	 * 64bit we do not enforce !NX on the low mapping
269 270 271
	 */
	if (within(address, (unsigned long)_text, (unsigned long)_etext))
		pgprot_val(forbidden) |= _PAGE_NX;
272 273

	/*
T
Thomas Gleixner 已提交
274 275
	 * The .rodata section needs to be read-only. Using the pfn
	 * catches all aliases.
276
	 */
T
Thomas Gleixner 已提交
277 278
	if (within(pfn, __pa((unsigned long)__start_rodata) >> PAGE_SHIFT,
		   __pa((unsigned long)__end_rodata) >> PAGE_SHIFT))
279
		pgprot_val(forbidden) |= _PAGE_RW;
280 281

	prot = __pgprot(pgprot_val(prot) & ~pgprot_val(forbidden));
I
Ingo Molnar 已提交
282 283 284 285

	return prot;
}

T
Thomas Gleixner 已提交
286 287 288 289 290 291 292 293
/*
 * 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.
 */
294
pte_t *lookup_address(unsigned long address, unsigned int *level)
295
{
L
Linus Torvalds 已提交
296 297 298
	pgd_t *pgd = pgd_offset_k(address);
	pud_t *pud;
	pmd_t *pmd;
299

T
Thomas Gleixner 已提交
300 301
	*level = PG_LEVEL_NONE;

L
Linus Torvalds 已提交
302 303
	if (pgd_none(*pgd))
		return NULL;
I
Ingo Molnar 已提交
304

L
Linus Torvalds 已提交
305 306 307
	pud = pud_offset(pgd, address);
	if (pud_none(*pud))
		return NULL;
308 309 310 311 312

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

L
Linus Torvalds 已提交
313 314 315
	pmd = pmd_offset(pud, address);
	if (pmd_none(*pmd))
		return NULL;
T
Thomas Gleixner 已提交
316 317

	*level = PG_LEVEL_2M;
T
Thomas Gleixner 已提交
318
	if (pmd_large(*pmd) || !pmd_present(*pmd))
L
Linus Torvalds 已提交
319 320
		return (pte_t *)pmd;

T
Thomas Gleixner 已提交
321
	*level = PG_LEVEL_4K;
I
Ingo Molnar 已提交
322

323 324
	return pte_offset_kernel(pmd, address);
}
325
EXPORT_SYMBOL_GPL(lookup_address);
326

I
Ingo Molnar 已提交
327 328 329
/*
 * Set the new pmd in all the pgds we know about:
 */
I
Ingo Molnar 已提交
330
static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
331 332 333
{
	/* change init_mm */
	set_pte_atomic(kpte, pte);
334
#ifdef CONFIG_X86_32
335
	if (!SHARED_KERNEL_PMD) {
336 337
		struct page *page;

338
		list_for_each_entry(page, &pgd_list, lru) {
339 340 341 342 343 344 345 346 347
			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
348
	}
349
#endif
L
Linus Torvalds 已提交
350 351
}

I
Ingo Molnar 已提交
352 353 354
static int
try_preserve_large_page(pte_t *kpte, unsigned long address,
			struct cpa_data *cpa)
355
{
T
Thomas Gleixner 已提交
356
	unsigned long nextpage_addr, numpages, pmask, psize, flags, addr, pfn;
357 358
	pte_t new_pte, old_pte, *tmp;
	pgprot_t old_prot, new_prot;
359
	int i, do_split = 1;
360
	unsigned int level;
361

362 363 364
	if (cpa->force_split)
		return 1;

365 366 367 368 369 370 371 372 373 374 375
	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:
376 377
		psize = PMD_PAGE_SIZE;
		pmask = PMD_PAGE_MASK;
378
		break;
379
#ifdef CONFIG_X86_64
380
	case PG_LEVEL_1G:
381 382
		psize = PUD_PAGE_SIZE;
		pmask = PUD_PAGE_MASK;
383 384
		break;
#endif
385
	default:
I
Ingo Molnar 已提交
386
		do_split = -EINVAL;
387 388 389 390 391 392 393 394 395
		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;
396 397
	if (numpages < cpa->numpages)
		cpa->numpages = numpages;
398 399 400 401 402 403 404 405 406

	/*
	 * 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
407 408 409 410 411 412 413 414 415

	/*
	 * 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);
416

417 418 419 420 421 422
	/*
	 * 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
423
	pfn++;
424
	for (i = 1; i < cpa->numpages; i++, addr += PAGE_SIZE, pfn++) {
T
Thomas Gleixner 已提交
425
		pgprot_t chk_prot = static_protections(new_prot, addr, pfn);
426 427 428 429 430

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

431 432 433 434 435
	/*
	 * 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
436
		do_split = 0;
437 438 439 440 441 442 443 444 445 446 447
		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.
	 */
448
	if (address == (nextpage_addr - psize) && cpa->numpages == numpages) {
449 450 451 452 453 454
		/*
		 * 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);
455
		cpa->flags |= CPA_FLUSHTLB;
I
Ingo Molnar 已提交
456
		do_split = 0;
457 458 459 460
	}

out_unlock:
	spin_unlock_irqrestore(&pgd_lock, flags);
I
Ingo Molnar 已提交
461

I
Ingo Molnar 已提交
462
	return do_split;
463 464
}

465
static int split_large_page(pte_t *kpte, unsigned long address)
466
{
T
Thomas Gleixner 已提交
467
	unsigned long flags, pfn, pfninc = 1;
I
Ingo Molnar 已提交
468
	unsigned int i, level;
469
	pte_t *pbase, *tmp;
I
Ingo Molnar 已提交
470
	pgprot_t ref_prot;
471 472 473 474
	struct page *base;

	if (!debug_pagealloc)
		spin_unlock(&cpa_lock);
475
	base = alloc_pages(GFP_KERNEL | __GFP_NOTRACK, 0);
476 477
	if (!debug_pagealloc)
		spin_lock(&cpa_lock);
S
Suresh Siddha 已提交
478 479
	if (!base)
		return -ENOMEM;
480

T
Thomas Gleixner 已提交
481
	spin_lock_irqsave(&pgd_lock, flags);
482 483 484 485 486
	/*
	 * 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
487
	if (tmp != kpte)
488 489 490
		goto out_unlock;

	pbase = (pte_t *)page_address(base);
491
	paravirt_alloc_pte(&init_mm, page_to_pfn(base));
T
Thomas Gleixner 已提交
492
	ref_prot = pte_pgprot(pte_clrhuge(*kpte));
493 494 495 496 497 498 499
	/*
	 * 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);
500

501 502 503 504 505 506 507
#ifdef CONFIG_X86_64
	if (level == PG_LEVEL_1G) {
		pfninc = PMD_PAGE_SIZE >> PAGE_SHIFT;
		pgprot_val(ref_prot) |= _PAGE_PSE;
	}
#endif

508 509 510 511
	/*
	 * Get the target pfn from the original entry:
	 */
	pfn = pte_pfn(*kpte);
512
	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
513
		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
514

515
	if (address >= (unsigned long)__va(0) &&
516 517 518 519 520
		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) &&
521 522
		address < (unsigned long)__va(max_pfn_mapped << PAGE_SHIFT))
		split_page_count(level);
523
#endif
524

525
	/*
526
	 * Install the new, split up pagetable.
527
	 *
528 529 530
	 * We use the standard kernel pagetable protections for the new
	 * pagetable protections, the actual ptes set above control the
	 * primary protection behavior:
531
	 */
532
	__set_pmd_pte(kpte, address, mk_pte(base, __pgprot(_KERNPG_TABLE)));
533 534 535 536 537 538 539 540 541 542 543

	/*
	 * 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();

544 545 546
	base = NULL;

out_unlock:
T
Thomas Gleixner 已提交
547 548 549 550
	/*
	 * 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
551 552
	if (base)
		__free_page(base);
I
Ingo Molnar 已提交
553
	spin_unlock_irqrestore(&pgd_lock, flags);
554 555 556 557

	return 0;
}

558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
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
588
static int __change_page_attr(struct cpa_data *cpa, int primary)
589
{
590
	unsigned long address;
591 592
	int do_split, err;
	unsigned int level;
T
Thomas Gleixner 已提交
593
	pte_t *kpte, old_pte;
L
Linus Torvalds 已提交
594

595 596 597 598 599 600
	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)
601 602 603
		address = cpa->vaddr[cpa->curpage];
	else
		address = *cpa->vaddr;
604
repeat:
605
	kpte = lookup_address(address, &level);
L
Linus Torvalds 已提交
606
	if (!kpte)
607
		return __cpa_process_fault(cpa, address, primary);
T
Thomas Gleixner 已提交
608 609

	old_pte = *kpte;
610 611
	if (!pte_val(old_pte))
		return __cpa_process_fault(cpa, address, primary);
612

T
Thomas Gleixner 已提交
613
	if (level == PG_LEVEL_4K) {
T
Thomas Gleixner 已提交
614
		pte_t new_pte;
615
		pgprot_t new_prot = pte_pgprot(old_pte);
T
Thomas Gleixner 已提交
616
		unsigned long pfn = pte_pfn(old_pte);
I
Ingo Molnar 已提交
617

T
Thomas Gleixner 已提交
618 619
		pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
		pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
I
Ingo Molnar 已提交
620

T
Thomas Gleixner 已提交
621
		new_prot = static_protections(new_prot, address, pfn);
I
Ingo Molnar 已提交
622

623 624 625 626 627
		/*
		 * 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
628 629
		new_pte = pfn_pte(pfn, canon_pgprot(new_prot));
		cpa->pfn = pfn;
630 631 632 633 634
		/*
		 * Do we really change anything ?
		 */
		if (pte_val(old_pte) != pte_val(new_pte)) {
			set_pte_atomic(kpte, new_pte);
635
			cpa->flags |= CPA_FLUSHTLB;
636
		}
637
		cpa->numpages = 1;
638
		return 0;
L
Linus Torvalds 已提交
639
	}
640 641 642 643 644

	/*
	 * Check, whether we can keep the large page intact
	 * and just change the pte:
	 */
I
Ingo Molnar 已提交
645
	do_split = try_preserve_large_page(kpte, address, cpa);
646 647
	/*
	 * When the range fits into the existing large page,
648
	 * return. cp->numpages and cpa->tlbflush have been updated in
649 650
	 * try_large_page:
	 */
I
Ingo Molnar 已提交
651 652
	if (do_split <= 0)
		return do_split;
653 654 655 656

	/*
	 * We have to split the large page:
	 */
I
Ingo Molnar 已提交
657 658
	err = split_large_page(kpte, address);
	if (!err) {
659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
		/*
	 	 * 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
678 679
		goto repeat;
	}
I
Ingo Molnar 已提交
680

I
Ingo Molnar 已提交
681
	return err;
682
}
L
Linus Torvalds 已提交
683

T
Thomas Gleixner 已提交
684 685 686
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
687
{
T
Thomas Gleixner 已提交
688
	struct cpa_data alias_cpa;
T
Tejun Heo 已提交
689
	unsigned long laddr = (unsigned long)__va(cpa->pfn << PAGE_SHIFT);
690
	unsigned long vaddr, remapped;
T
Tejun Heo 已提交
691
	int ret;
692

Y
Yinghai Lu 已提交
693
	if (cpa->pfn >= max_pfn_mapped)
T
Thomas Gleixner 已提交
694
		return 0;
695

696
#ifdef CONFIG_X86_64
Y
Yinghai Lu 已提交
697
	if (cpa->pfn >= max_low_pfn_mapped && cpa->pfn < (1UL<<(32-PAGE_SHIFT)))
698 699
		return 0;
#endif
700 701 702 703
	/*
	 * No need to redo, when the primary call touched the direct
	 * mapping already:
	 */
704 705 706 707 708 709
	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)
710 711 712 713 714
		vaddr = cpa->vaddr[cpa->curpage];
	else
		vaddr = *cpa->vaddr;

	if (!(within(vaddr, PAGE_OFFSET,
715
		    PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) {
716

717
		alias_cpa = *cpa;
T
Tejun Heo 已提交
718
		alias_cpa.vaddr = &laddr;
719
		alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
720

721
		ret = __change_page_attr_set_clr(&alias_cpa, 0);
T
Tejun Heo 已提交
722 723
		if (ret)
			return ret;
724
	}
725 726

#ifdef CONFIG_X86_64
A
Arjan van de Ven 已提交
727
	/*
T
Tejun Heo 已提交
728 729
	 * If the primary call didn't touch the high mapping already
	 * and the physical address is inside the kernel map, we need
730
	 * to touch the high mapped kernel as well:
A
Arjan van de Ven 已提交
731
	 */
T
Tejun Heo 已提交
732 733 734 735 736 737 738
	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
739

T
Tejun Heo 已提交
740 741 742 743 744 745
		/*
		 * 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
746
#endif
T
Tejun Heo 已提交
747

748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765
	/*
	 * If the PMD page was partially used for per-cpu remapping,
	 * the recycled area needs to be split and modified.  Because
	 * the area is always proper subset of a PMD page
	 * cpa->numpages is guaranteed to be 1 for these areas, so
	 * there's no need to loop over and check for further remaps.
	 */
	remapped = (unsigned long)pcpu_lpage_remapped((void *)laddr);
	if (remapped) {
		WARN_ON(cpa->numpages > 1);
		alias_cpa = *cpa;
		alias_cpa.vaddr = &remapped;
		alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
		ret = __change_page_attr_set_clr(&alias_cpa, 0);
		if (ret)
			return ret;
	}

T
Tejun Heo 已提交
766
	return 0;
L
Linus Torvalds 已提交
767 768
}

T
Thomas Gleixner 已提交
769
static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias)
770
{
771
	int ret, numpages = cpa->numpages;
772

773 774 775 776 777
	while (numpages) {
		/*
		 * Store the remaining nr of pages for the large page
		 * preservation check.
		 */
778
		cpa->numpages = numpages;
779
		/* for array changes, we can't use large page */
780
		if (cpa->flags & (CPA_ARRAY | CPA_PAGES_ARRAY))
781
			cpa->numpages = 1;
T
Thomas Gleixner 已提交
782

783 784
		if (!debug_pagealloc)
			spin_lock(&cpa_lock);
T
Thomas Gleixner 已提交
785
		ret = __change_page_attr(cpa, checkalias);
786 787
		if (!debug_pagealloc)
			spin_unlock(&cpa_lock);
788 789 790
		if (ret)
			return ret;

T
Thomas Gleixner 已提交
791 792 793 794 795 796
		if (checkalias) {
			ret = cpa_process_alias(cpa);
			if (ret)
				return ret;
		}

797 798 799 800 801
		/*
		 * 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.
		 */
802 803
		BUG_ON(cpa->numpages > numpages);
		numpages -= cpa->numpages;
804
		if (cpa->flags & (CPA_PAGES_ARRAY | CPA_ARRAY))
805 806 807 808
			cpa->curpage++;
		else
			*cpa->vaddr += cpa->numpages * PAGE_SIZE;

809
	}
810 811 812
	return 0;
}

813 814 815 816 817 818
static inline int cache_attr(pgprot_t attr)
{
	return pgprot_val(attr) &
		(_PAGE_PAT | _PAGE_PAT_LARGE | _PAGE_PWT | _PAGE_PCD);
}

819
static int change_page_attr_set_clr(unsigned long *addr, int numpages,
820
				    pgprot_t mask_set, pgprot_t mask_clr,
821 822
				    int force_split, int in_flag,
				    struct page **pages)
823
{
T
Thomas Gleixner 已提交
824
	struct cpa_data cpa;
825
	int ret, cache, checkalias;
826 827 828 829 830 831 832

	/*
	 * Check, if we are requested to change a not supported
	 * feature:
	 */
	mask_set = canon_pgprot(mask_set);
	mask_clr = canon_pgprot(mask_clr);
833
	if (!pgprot_val(mask_set) && !pgprot_val(mask_clr) && !force_split)
834 835
		return 0;

836
	/* Ensure we are PAGE_SIZE aligned */
837
	if (in_flag & CPA_ARRAY) {
838 839 840 841 842 843 844
		int i;
		for (i = 0; i < numpages; i++) {
			if (addr[i] & ~PAGE_MASK) {
				addr[i] &= PAGE_MASK;
				WARN_ON_ONCE(1);
			}
		}
845 846 847 848 849 850 851 852 853 854 855 856
	} 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);
		}
857 858
	}

859 860 861
	/* Must avoid aliasing mappings in the highmem code */
	kmap_flush_unused();

N
Nick Piggin 已提交
862 863
	vm_unmap_aliases();

T
Thomas Gleixner 已提交
864
	cpa.vaddr = addr;
865
	cpa.pages = pages;
T
Thomas Gleixner 已提交
866 867 868
	cpa.numpages = numpages;
	cpa.mask_set = mask_set;
	cpa.mask_clr = mask_clr;
869 870
	cpa.flags = 0;
	cpa.curpage = 0;
871
	cpa.force_split = force_split;
T
Thomas Gleixner 已提交
872

873 874
	if (in_flag & (CPA_ARRAY | CPA_PAGES_ARRAY))
		cpa.flags |= in_flag;
875

876 877 878 879
	/* 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);
880

881 882 883
	/*
	 * Check whether we really changed something:
	 */
884
	if (!(cpa.flags & CPA_FLUSHTLB))
885
		goto out;
886

887 888 889 890 891 892
	/*
	 * No need to flush, when we did not set any of the caching
	 * attributes:
	 */
	cache = cache_attr(mask_set);

893 894 895
	/*
	 * On success we use clflush, when the CPU supports it to
	 * avoid the wbindv. If the CPU does not support it and in the
896
	 * error case we fall back to cpa_flush_all (which uses
897 898
	 * wbindv):
	 */
899
	if (!ret && cpu_has_clflush) {
900 901 902 903
		if (cpa.flags & (CPA_PAGES_ARRAY | CPA_ARRAY)) {
			cpa_flush_array(addr, numpages, cache,
					cpa.flags, pages);
		} else
904 905
			cpa_flush_range(*addr, numpages, cache);
	} else
906
		cpa_flush_all(cache);
907

908
out:
909 910 911
	return ret;
}

912 913
static inline int change_page_attr_set(unsigned long *addr, int numpages,
				       pgprot_t mask, int array)
914
{
915
	return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0), 0,
916
		(array ? CPA_ARRAY : 0), NULL);
917 918
}

919 920
static inline int change_page_attr_clear(unsigned long *addr, int numpages,
					 pgprot_t mask, int array)
921
{
922
	return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask, 0,
923
		(array ? CPA_ARRAY : 0), NULL);
924 925
}

926 927 928 929 930 931 932 933 934 935 936 937 938 939
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);
}

940
int _set_memory_uc(unsigned long addr, int numpages)
941
{
942 943 944
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
945 946
	return change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
947
}
948 949 950

int set_memory_uc(unsigned long addr, int numpages)
{
951 952
	int ret;

953 954 955
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
956 957 958 959 960 961 962 963 964 965
	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;
966

967 968 969 970
out_free:
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
out_err:
	return ret;
971
}
972 973
EXPORT_SYMBOL(set_memory_uc);

974 975
int set_memory_array_uc(unsigned long *addr, int addrinarray)
{
976 977 978
	int i, j;
	int ret;

979 980 981 982
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
	for (i = 0; i < addrinarray; i++) {
983 984 985 986
		ret = reserve_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE,
					_PAGE_CACHE_UC_MINUS, NULL);
		if (ret)
			goto out_free;
987 988
	}

989
	ret = change_page_attr_set(addr, addrinarray,
990
				    __pgprot(_PAGE_CACHE_UC_MINUS), 1);
991 992 993 994 995 996 997 998 999 1000
	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;
1001 1002 1003
}
EXPORT_SYMBOL(set_memory_array_uc);

1004 1005
int _set_memory_wc(unsigned long addr, int numpages)
{
1006
	int ret;
1007 1008
	unsigned long addr_copy = addr;

1009 1010 1011
	ret = change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
	if (!ret) {
1012 1013 1014 1015
		ret = change_page_attr_set_clr(&addr_copy, numpages,
					       __pgprot(_PAGE_CACHE_WC),
					       __pgprot(_PAGE_CACHE_MASK),
					       0, 0, NULL);
1016 1017
	}
	return ret;
1018 1019 1020 1021
}

int set_memory_wc(unsigned long addr, int numpages)
{
1022 1023
	int ret;

1024
	if (!pat_enabled)
1025 1026
		return set_memory_uc(addr, numpages);

1027 1028 1029 1030
	ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
		_PAGE_CACHE_WC, NULL);
	if (ret)
		goto out_err;
1031

1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
	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;
1042 1043 1044
}
EXPORT_SYMBOL(set_memory_wc);

1045
int _set_memory_wb(unsigned long addr, int numpages)
1046
{
1047 1048
	return change_page_attr_clear(&addr, numpages,
				      __pgprot(_PAGE_CACHE_MASK), 0);
1049
}
1050 1051 1052

int set_memory_wb(unsigned long addr, int numpages)
{
1053 1054 1055 1056 1057 1058
	int ret;

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

1059
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
1060
	return 0;
1061
}
1062 1063
EXPORT_SYMBOL(set_memory_wb);

1064 1065 1066
int set_memory_array_wb(unsigned long *addr, int addrinarray)
{
	int i;
1067 1068 1069 1070
	int ret;

	ret = change_page_attr_clear(addr, addrinarray,
				      __pgprot(_PAGE_CACHE_MASK), 1);
1071 1072
	if (ret)
		return ret;
1073

1074 1075
	for (i = 0; i < addrinarray; i++)
		free_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE);
1076

1077
	return 0;
1078 1079 1080
}
EXPORT_SYMBOL(set_memory_array_wb);

1081 1082
int set_memory_x(unsigned long addr, int numpages)
{
1083
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_NX), 0);
1084 1085 1086 1087 1088
}
EXPORT_SYMBOL(set_memory_x);

int set_memory_nx(unsigned long addr, int numpages)
{
1089
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_NX), 0);
1090 1091 1092 1093 1094
}
EXPORT_SYMBOL(set_memory_nx);

int set_memory_ro(unsigned long addr, int numpages)
{
1095
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_RW), 0);
1096
}
1097
EXPORT_SYMBOL_GPL(set_memory_ro);
1098 1099 1100

int set_memory_rw(unsigned long addr, int numpages)
{
1101
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_RW), 0);
1102
}
1103
EXPORT_SYMBOL_GPL(set_memory_rw);
I
Ingo Molnar 已提交
1104 1105 1106

int set_memory_np(unsigned long addr, int numpages)
{
1107
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_PRESENT), 0);
I
Ingo Molnar 已提交
1108
}
1109

1110 1111
int set_memory_4k(unsigned long addr, int numpages)
{
1112
	return change_page_attr_set_clr(&addr, numpages, __pgprot(0),
1113
					__pgprot(0), 1, 0, NULL);
1114 1115
}

1116 1117 1118 1119
int set_pages_uc(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1120
	return set_memory_uc(addr, numpages);
1121 1122 1123
}
EXPORT_SYMBOL(set_pages_uc);

1124 1125 1126 1127 1128 1129 1130 1131
int set_pages_array_uc(struct page **pages, int addrinarray)
{
	unsigned long start;
	unsigned long end;
	int i;
	int free_idx;

	for (i = 0; i < addrinarray; i++) {
1132 1133 1134
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
		end = start + PAGE_SIZE;
		if (reserve_memtype(start, end, _PAGE_CACHE_UC_MINUS, NULL))
			goto err_out;
	}

	if (cpa_set_pages_array(pages, addrinarray,
			__pgprot(_PAGE_CACHE_UC_MINUS)) == 0) {
		return 0; /* Success */
	}
err_out:
	free_idx = i;
	for (i = 0; i < free_idx; i++) {
1147 1148 1149
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1150 1151 1152 1153 1154 1155 1156
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}
	return -EINVAL;
}
EXPORT_SYMBOL(set_pages_array_uc);

1157 1158 1159 1160
int set_pages_wb(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1161
	return set_memory_wb(addr, numpages);
1162 1163 1164
}
EXPORT_SYMBOL(set_pages_wb);

1165 1166 1167 1168 1169 1170 1171 1172 1173
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));
1174 1175
	if (retval)
		return retval;
1176 1177

	for (i = 0; i < addrinarray; i++) {
1178 1179 1180
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1181 1182 1183 1184
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}

1185
	return 0;
1186 1187 1188
}
EXPORT_SYMBOL(set_pages_array_wb);

1189 1190 1191 1192
int set_pages_x(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1193
	return set_memory_x(addr, numpages);
1194 1195 1196 1197 1198 1199 1200
}
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
1201
	return set_memory_nx(addr, numpages);
1202 1203 1204 1205 1206 1207 1208
}
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
1209
	return set_memory_ro(addr, numpages);
1210 1211 1212 1213 1214
}

int set_pages_rw(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);
1215

T
Thomas Gleixner 已提交
1216
	return set_memory_rw(addr, numpages);
I
Ingo Molnar 已提交
1217 1218
}

L
Linus Torvalds 已提交
1219
#ifdef CONFIG_DEBUG_PAGEALLOC
I
Ingo Molnar 已提交
1220 1221 1222

static int __set_pages_p(struct page *page, int numpages)
{
1223 1224
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1225 1226
				.numpages = numpages,
				.mask_set = __pgprot(_PAGE_PRESENT | _PAGE_RW),
1227 1228
				.mask_clr = __pgprot(0),
				.flags = 0};
1229

1230 1231 1232 1233 1234 1235 1236
	/*
	 * 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
1237 1238 1239 1240
}

static int __set_pages_np(struct page *page, int numpages)
{
1241 1242
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1243 1244
				.numpages = numpages,
				.mask_set = __pgprot(0),
1245 1246
				.mask_clr = __pgprot(_PAGE_PRESENT | _PAGE_RW),
				.flags = 0};
1247

1248 1249 1250 1251 1252 1253 1254
	/*
	 * 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
1255 1256
}

L
Linus Torvalds 已提交
1257 1258 1259 1260
void kernel_map_pages(struct page *page, int numpages, int enable)
{
	if (PageHighMem(page))
		return;
1261
	if (!enable) {
1262 1263
		debug_check_no_locks_freed(page_address(page),
					   numpages * PAGE_SIZE);
1264
	}
1265

1266 1267 1268 1269 1270 1271
	/*
	 * If page allocator is not up yet then do not call c_p_a():
	 */
	if (!debug_pagealloc_enabled)
		return;

1272
	/*
I
Ingo Molnar 已提交
1273
	 * The return value is ignored as the calls cannot fail.
1274 1275
	 * 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
1276
	 */
I
Ingo Molnar 已提交
1277 1278 1279 1280
	if (enable)
		__set_pages_p(page, numpages);
	else
		__set_pages_np(page, numpages);
1281 1282

	/*
1283 1284
	 * We should perform an IPI and flush all tlbs,
	 * but that can deadlock->flush only current cpu:
L
Linus Torvalds 已提交
1285 1286
	 */
	__flush_tlb_all();
1287 1288
}

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305
#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 */
1306 1307 1308 1309 1310 1311 1312 1313

/*
 * 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