pageattr.c 33.8 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
{
	pgprot_t forbidden = __pgprot(0);
258
	pgprot_t required = __pgprot(0);
259

I
Ingo Molnar 已提交
260
	/*
261 262
	 * 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
263
	 */
T
Thomas Gleixner 已提交
264
	if (within(pfn, BIOS_BEGIN >> PAGE_SHIFT, BIOS_END >> PAGE_SHIFT))
265 266 267 268
		pgprot_val(forbidden) |= _PAGE_NX;

	/*
	 * The kernel text needs to be executable for obvious reasons
T
Thomas Gleixner 已提交
269 270
	 * Does not cover __inittext since that is gone later on. On
	 * 64bit we do not enforce !NX on the low mapping
271 272 273
	 */
	if (within(address, (unsigned long)_text, (unsigned long)_etext))
		pgprot_val(forbidden) |= _PAGE_NX;
274 275

	/*
T
Thomas Gleixner 已提交
276 277
	 * The .rodata section needs to be read-only. Using the pfn
	 * catches all aliases.
278
	 */
T
Thomas Gleixner 已提交
279 280
	if (within(pfn, __pa((unsigned long)__start_rodata) >> PAGE_SHIFT,
		   __pa((unsigned long)__end_rodata) >> PAGE_SHIFT))
281
		pgprot_val(forbidden) |= _PAGE_RW;
282 283 284 285 286 287
	/*
	 * .data and .bss should always be writable.
	 */
	if (within(address, (unsigned long)_sdata, (unsigned long)_edata) ||
	    within(address, (unsigned long)__bss_start, (unsigned long)__bss_stop))
		pgprot_val(required) |= _PAGE_RW;
288

289
#if defined(CONFIG_X86_64) && defined(CONFIG_DEBUG_RODATA)
290
	/*
291 292 293 294
	 * 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.
295 296 297 298
	 *
	 * This will preserve the large page mappings for kernel text/data
	 * at no extra cost.
	 */
299 300
	if (kernel_set_to_readonly &&
	    within(address, (unsigned long)_text,
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
		   (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;
	}
324 325
#endif

326
	prot = __pgprot(pgprot_val(prot) & ~pgprot_val(forbidden));
327
	prot = __pgprot(pgprot_val(prot) | pgprot_val(required));
I
Ingo Molnar 已提交
328 329 330 331

	return prot;
}

T
Thomas Gleixner 已提交
332 333 334 335 336 337 338 339
/*
 * 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.
 */
340
pte_t *lookup_address(unsigned long address, unsigned int *level)
341
{
L
Linus Torvalds 已提交
342 343 344
	pgd_t *pgd = pgd_offset_k(address);
	pud_t *pud;
	pmd_t *pmd;
345

T
Thomas Gleixner 已提交
346 347
	*level = PG_LEVEL_NONE;

L
Linus Torvalds 已提交
348 349
	if (pgd_none(*pgd))
		return NULL;
I
Ingo Molnar 已提交
350

L
Linus Torvalds 已提交
351 352 353
	pud = pud_offset(pgd, address);
	if (pud_none(*pud))
		return NULL;
354 355 356 357 358

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

L
Linus Torvalds 已提交
359 360 361
	pmd = pmd_offset(pud, address);
	if (pmd_none(*pmd))
		return NULL;
T
Thomas Gleixner 已提交
362 363

	*level = PG_LEVEL_2M;
T
Thomas Gleixner 已提交
364
	if (pmd_large(*pmd) || !pmd_present(*pmd))
L
Linus Torvalds 已提交
365 366
		return (pte_t *)pmd;

T
Thomas Gleixner 已提交
367
	*level = PG_LEVEL_4K;
I
Ingo Molnar 已提交
368

369 370
	return pte_offset_kernel(pmd, address);
}
371
EXPORT_SYMBOL_GPL(lookup_address);
372

I
Ingo Molnar 已提交
373 374 375
/*
 * Set the new pmd in all the pgds we know about:
 */
I
Ingo Molnar 已提交
376
static void __set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)
377 378 379
{
	/* change init_mm */
	set_pte_atomic(kpte, pte);
380
#ifdef CONFIG_X86_32
381
	if (!SHARED_KERNEL_PMD) {
382 383
		struct page *page;

384
		list_for_each_entry(page, &pgd_list, lru) {
385 386 387 388 389 390 391 392 393
			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
394
	}
395
#endif
L
Linus Torvalds 已提交
396 397
}

I
Ingo Molnar 已提交
398 399 400
static int
try_preserve_large_page(pte_t *kpte, unsigned long address,
			struct cpa_data *cpa)
401
{
T
Thomas Gleixner 已提交
402
	unsigned long nextpage_addr, numpages, pmask, psize, flags, addr, pfn;
403
	pte_t new_pte, old_pte, *tmp;
404
	pgprot_t old_prot, new_prot, req_prot;
405
	int i, do_split = 1;
406
	unsigned int level;
407

408 409 410
	if (cpa->force_split)
		return 1;

411 412 413 414 415 416 417 418 419 420 421
	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:
422 423
		psize = PMD_PAGE_SIZE;
		pmask = PMD_PAGE_MASK;
424
		break;
425
#ifdef CONFIG_X86_64
426
	case PG_LEVEL_1G:
427 428
		psize = PUD_PAGE_SIZE;
		pmask = PUD_PAGE_MASK;
429 430
		break;
#endif
431
	default:
I
Ingo Molnar 已提交
432
		do_split = -EINVAL;
433 434 435 436 437 438 439 440 441
		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;
442 443
	if (numpages < cpa->numpages)
		cpa->numpages = numpages;
444 445 446 447 448

	/*
	 * We are safe now. Check whether the new pgprot is the same:
	 */
	old_pte = *kpte;
449
	old_prot = new_prot = req_prot = pte_pgprot(old_pte);
450

451 452
	pgprot_val(req_prot) &= ~pgprot_val(cpa->mask_clr);
	pgprot_val(req_prot) |= pgprot_val(cpa->mask_set);
T
Thomas Gleixner 已提交
453 454 455 456 457 458 459 460

	/*
	 * 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;

461
	new_prot = static_protections(req_prot, address, pfn);
462

463 464 465 466 467
	/*
	 * 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:
	 */
468 469 470 471
	addr = address & pmask;
	pfn = pte_pfn(old_pte);
	for (i = 0; i < (psize >> PAGE_SHIFT); i++, addr += PAGE_SIZE, pfn++) {
		pgprot_t chk_prot = static_protections(req_prot, addr, pfn);
472 473 474 475 476

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

477 478 479 480 481
	/*
	 * 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
482
		do_split = 0;
483 484 485 486 487 488 489 490 491 492 493
		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.
	 */
494
	if (address == (address & pmask) && cpa->numpages == (psize >> PAGE_SHIFT)) {
495 496 497 498 499 500
		/*
		 * 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);
501
		cpa->flags |= CPA_FLUSHTLB;
I
Ingo Molnar 已提交
502
		do_split = 0;
503 504 505 506
	}

out_unlock:
	spin_unlock_irqrestore(&pgd_lock, flags);
I
Ingo Molnar 已提交
507

I
Ingo Molnar 已提交
508
	return do_split;
509 510
}

511
static int split_large_page(pte_t *kpte, unsigned long address)
512
{
T
Thomas Gleixner 已提交
513
	unsigned long flags, pfn, pfninc = 1;
I
Ingo Molnar 已提交
514
	unsigned int i, level;
515
	pte_t *pbase, *tmp;
I
Ingo Molnar 已提交
516
	pgprot_t ref_prot;
517 518 519 520
	struct page *base;

	if (!debug_pagealloc)
		spin_unlock(&cpa_lock);
521
	base = alloc_pages(GFP_KERNEL | __GFP_NOTRACK, 0);
522 523
	if (!debug_pagealloc)
		spin_lock(&cpa_lock);
S
Suresh Siddha 已提交
524 525
	if (!base)
		return -ENOMEM;
526

T
Thomas Gleixner 已提交
527
	spin_lock_irqsave(&pgd_lock, flags);
528 529 530 531 532
	/*
	 * 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
533
	if (tmp != kpte)
534 535 536
		goto out_unlock;

	pbase = (pte_t *)page_address(base);
537
	paravirt_alloc_pte(&init_mm, page_to_pfn(base));
T
Thomas Gleixner 已提交
538
	ref_prot = pte_pgprot(pte_clrhuge(*kpte));
539 540 541 542 543 544 545
	/*
	 * 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);
546

547 548 549 550 551 552 553
#ifdef CONFIG_X86_64
	if (level == PG_LEVEL_1G) {
		pfninc = PMD_PAGE_SIZE >> PAGE_SHIFT;
		pgprot_val(ref_prot) |= _PAGE_PSE;
	}
#endif

554 555 556 557
	/*
	 * Get the target pfn from the original entry:
	 */
	pfn = pte_pfn(*kpte);
558
	for (i = 0; i < PTRS_PER_PTE; i++, pfn += pfninc)
559
		set_pte(&pbase[i], pfn_pte(pfn, ref_prot));
560

561
	if (address >= (unsigned long)__va(0) &&
562 563 564 565 566
		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) &&
567 568
		address < (unsigned long)__va(max_pfn_mapped << PAGE_SHIFT))
		split_page_count(level);
569
#endif
570

571
	/*
572
	 * Install the new, split up pagetable.
573
	 *
574 575 576
	 * We use the standard kernel pagetable protections for the new
	 * pagetable protections, the actual ptes set above control the
	 * primary protection behavior:
577
	 */
578
	__set_pmd_pte(kpte, address, mk_pte(base, __pgprot(_KERNPG_TABLE)));
579 580 581 582 583 584 585 586 587 588 589

	/*
	 * 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();

590 591 592
	base = NULL;

out_unlock:
T
Thomas Gleixner 已提交
593 594 595 596
	/*
	 * 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
597 598
	if (base)
		__free_page(base);
I
Ingo Molnar 已提交
599
	spin_unlock_irqrestore(&pgd_lock, flags);
600 601 602 603

	return 0;
}

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
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
634
static int __change_page_attr(struct cpa_data *cpa, int primary)
635
{
636
	unsigned long address;
637 638
	int do_split, err;
	unsigned int level;
T
Thomas Gleixner 已提交
639
	pte_t *kpte, old_pte;
L
Linus Torvalds 已提交
640

641 642 643 644 645 646
	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)
647 648 649
		address = cpa->vaddr[cpa->curpage];
	else
		address = *cpa->vaddr;
650
repeat:
651
	kpte = lookup_address(address, &level);
L
Linus Torvalds 已提交
652
	if (!kpte)
653
		return __cpa_process_fault(cpa, address, primary);
T
Thomas Gleixner 已提交
654 655

	old_pte = *kpte;
656 657
	if (!pte_val(old_pte))
		return __cpa_process_fault(cpa, address, primary);
658

T
Thomas Gleixner 已提交
659
	if (level == PG_LEVEL_4K) {
T
Thomas Gleixner 已提交
660
		pte_t new_pte;
661
		pgprot_t new_prot = pte_pgprot(old_pte);
T
Thomas Gleixner 已提交
662
		unsigned long pfn = pte_pfn(old_pte);
I
Ingo Molnar 已提交
663

T
Thomas Gleixner 已提交
664 665
		pgprot_val(new_prot) &= ~pgprot_val(cpa->mask_clr);
		pgprot_val(new_prot) |= pgprot_val(cpa->mask_set);
I
Ingo Molnar 已提交
666

T
Thomas Gleixner 已提交
667
		new_prot = static_protections(new_prot, address, pfn);
I
Ingo Molnar 已提交
668

669 670 671 672 673
		/*
		 * 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
674 675
		new_pte = pfn_pte(pfn, canon_pgprot(new_prot));
		cpa->pfn = pfn;
676 677 678 679 680
		/*
		 * Do we really change anything ?
		 */
		if (pte_val(old_pte) != pte_val(new_pte)) {
			set_pte_atomic(kpte, new_pte);
681
			cpa->flags |= CPA_FLUSHTLB;
682
		}
683
		cpa->numpages = 1;
684
		return 0;
L
Linus Torvalds 已提交
685
	}
686 687 688 689 690

	/*
	 * Check, whether we can keep the large page intact
	 * and just change the pte:
	 */
I
Ingo Molnar 已提交
691
	do_split = try_preserve_large_page(kpte, address, cpa);
692 693
	/*
	 * When the range fits into the existing large page,
694
	 * return. cp->numpages and cpa->tlbflush have been updated in
695 696
	 * try_large_page:
	 */
I
Ingo Molnar 已提交
697 698
	if (do_split <= 0)
		return do_split;
699 700 701 702

	/*
	 * We have to split the large page:
	 */
I
Ingo Molnar 已提交
703 704
	err = split_large_page(kpte, address);
	if (!err) {
705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
		/*
	 	 * 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
724 725
		goto repeat;
	}
I
Ingo Molnar 已提交
726

I
Ingo Molnar 已提交
727
	return err;
728
}
L
Linus Torvalds 已提交
729

T
Thomas Gleixner 已提交
730 731 732
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
733
{
T
Thomas Gleixner 已提交
734
	struct cpa_data alias_cpa;
T
Tejun Heo 已提交
735
	unsigned long laddr = (unsigned long)__va(cpa->pfn << PAGE_SHIFT);
736
	unsigned long vaddr;
T
Tejun Heo 已提交
737
	int ret;
738

Y
Yinghai Lu 已提交
739
	if (cpa->pfn >= max_pfn_mapped)
T
Thomas Gleixner 已提交
740
		return 0;
741

742
#ifdef CONFIG_X86_64
Y
Yinghai Lu 已提交
743
	if (cpa->pfn >= max_low_pfn_mapped && cpa->pfn < (1UL<<(32-PAGE_SHIFT)))
744 745
		return 0;
#endif
746 747 748 749
	/*
	 * No need to redo, when the primary call touched the direct
	 * mapping already:
	 */
750 751 752 753 754 755
	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)
756 757 758 759 760
		vaddr = cpa->vaddr[cpa->curpage];
	else
		vaddr = *cpa->vaddr;

	if (!(within(vaddr, PAGE_OFFSET,
761
		    PAGE_OFFSET + (max_pfn_mapped << PAGE_SHIFT)))) {
762

763
		alias_cpa = *cpa;
T
Tejun Heo 已提交
764
		alias_cpa.vaddr = &laddr;
765
		alias_cpa.flags &= ~(CPA_PAGES_ARRAY | CPA_ARRAY);
766

767
		ret = __change_page_attr_set_clr(&alias_cpa, 0);
T
Tejun Heo 已提交
768 769
		if (ret)
			return ret;
770
	}
771 772

#ifdef CONFIG_X86_64
A
Arjan van de Ven 已提交
773
	/*
T
Tejun Heo 已提交
774 775
	 * If the primary call didn't touch the high mapping already
	 * and the physical address is inside the kernel map, we need
776
	 * to touch the high mapped kernel as well:
A
Arjan van de Ven 已提交
777
	 */
T
Tejun Heo 已提交
778 779 780 781 782 783 784
	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
785

T
Tejun Heo 已提交
786 787 788 789 790 791
		/*
		 * 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
792
#endif
T
Tejun Heo 已提交
793 794

	return 0;
L
Linus Torvalds 已提交
795 796
}

T
Thomas Gleixner 已提交
797
static int __change_page_attr_set_clr(struct cpa_data *cpa, int checkalias)
798
{
799
	int ret, numpages = cpa->numpages;
800

801 802 803 804 805
	while (numpages) {
		/*
		 * Store the remaining nr of pages for the large page
		 * preservation check.
		 */
806
		cpa->numpages = numpages;
807
		/* for array changes, we can't use large page */
808
		if (cpa->flags & (CPA_ARRAY | CPA_PAGES_ARRAY))
809
			cpa->numpages = 1;
T
Thomas Gleixner 已提交
810

811 812
		if (!debug_pagealloc)
			spin_lock(&cpa_lock);
T
Thomas Gleixner 已提交
813
		ret = __change_page_attr(cpa, checkalias);
814 815
		if (!debug_pagealloc)
			spin_unlock(&cpa_lock);
816 817 818
		if (ret)
			return ret;

T
Thomas Gleixner 已提交
819 820 821 822 823 824
		if (checkalias) {
			ret = cpa_process_alias(cpa);
			if (ret)
				return ret;
		}

825 826 827 828 829
		/*
		 * 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.
		 */
830 831
		BUG_ON(cpa->numpages > numpages);
		numpages -= cpa->numpages;
832
		if (cpa->flags & (CPA_PAGES_ARRAY | CPA_ARRAY))
833 834 835 836
			cpa->curpage++;
		else
			*cpa->vaddr += cpa->numpages * PAGE_SIZE;

837
	}
838 839 840
	return 0;
}

841 842 843 844 845 846
static inline int cache_attr(pgprot_t attr)
{
	return pgprot_val(attr) &
		(_PAGE_PAT | _PAGE_PAT_LARGE | _PAGE_PWT | _PAGE_PCD);
}

847
static int change_page_attr_set_clr(unsigned long *addr, int numpages,
848
				    pgprot_t mask_set, pgprot_t mask_clr,
849 850
				    int force_split, int in_flag,
				    struct page **pages)
851
{
T
Thomas Gleixner 已提交
852
	struct cpa_data cpa;
853
	int ret, cache, checkalias;
854
	unsigned long baddr = 0;
855 856 857 858 859 860 861

	/*
	 * Check, if we are requested to change a not supported
	 * feature:
	 */
	mask_set = canon_pgprot(mask_set);
	mask_clr = canon_pgprot(mask_clr);
862
	if (!pgprot_val(mask_set) && !pgprot_val(mask_clr) && !force_split)
863 864
		return 0;

865
	/* Ensure we are PAGE_SIZE aligned */
866
	if (in_flag & CPA_ARRAY) {
867 868 869 870 871 872 873
		int i;
		for (i = 0; i < numpages; i++) {
			if (addr[i] & ~PAGE_MASK) {
				addr[i] &= PAGE_MASK;
				WARN_ON_ONCE(1);
			}
		}
874 875 876 877 878 879 880 881 882 883 884 885
	} 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);
		}
886 887 888 889 890
		/*
		 * Save address for cache flush. *addr is modified in the call
		 * to __change_page_attr_set_clr() below.
		 */
		baddr = *addr;
891 892
	}

893 894 895
	/* Must avoid aliasing mappings in the highmem code */
	kmap_flush_unused();

N
Nick Piggin 已提交
896 897
	vm_unmap_aliases();

T
Thomas Gleixner 已提交
898
	cpa.vaddr = addr;
899
	cpa.pages = pages;
T
Thomas Gleixner 已提交
900 901 902
	cpa.numpages = numpages;
	cpa.mask_set = mask_set;
	cpa.mask_clr = mask_clr;
903 904
	cpa.flags = 0;
	cpa.curpage = 0;
905
	cpa.force_split = force_split;
T
Thomas Gleixner 已提交
906

907 908
	if (in_flag & (CPA_ARRAY | CPA_PAGES_ARRAY))
		cpa.flags |= in_flag;
909

910 911 912 913
	/* 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);
914

915 916 917
	/*
	 * Check whether we really changed something:
	 */
918
	if (!(cpa.flags & CPA_FLUSHTLB))
919
		goto out;
920

921 922 923 924 925 926
	/*
	 * No need to flush, when we did not set any of the caching
	 * attributes:
	 */
	cache = cache_attr(mask_set);

927 928 929
	/*
	 * On success we use clflush, when the CPU supports it to
	 * avoid the wbindv. If the CPU does not support it and in the
930
	 * error case we fall back to cpa_flush_all (which uses
931 932
	 * wbindv):
	 */
933
	if (!ret && cpu_has_clflush) {
934 935 936 937
		if (cpa.flags & (CPA_PAGES_ARRAY | CPA_ARRAY)) {
			cpa_flush_array(addr, numpages, cache,
					cpa.flags, pages);
		} else
938
			cpa_flush_range(baddr, numpages, cache);
939
	} else
940
		cpa_flush_all(cache);
941

942
out:
943 944 945
	return ret;
}

946 947
static inline int change_page_attr_set(unsigned long *addr, int numpages,
				       pgprot_t mask, int array)
948
{
949
	return change_page_attr_set_clr(addr, numpages, mask, __pgprot(0), 0,
950
		(array ? CPA_ARRAY : 0), NULL);
951 952
}

953 954
static inline int change_page_attr_clear(unsigned long *addr, int numpages,
					 pgprot_t mask, int array)
955
{
956
	return change_page_attr_set_clr(addr, numpages, __pgprot(0), mask, 0,
957
		(array ? CPA_ARRAY : 0), NULL);
958 959
}

960 961 962 963 964 965 966 967 968 969 970 971 972 973
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);
}

974
int _set_memory_uc(unsigned long addr, int numpages)
975
{
976 977 978
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
979 980
	return change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
981
}
982 983 984

int set_memory_uc(unsigned long addr, int numpages)
{
985 986
	int ret;

987 988 989
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
990 991 992 993 994 995 996 997 998 999
	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;
1000

1001 1002 1003 1004
out_free:
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
out_err:
	return ret;
1005
}
1006 1007
EXPORT_SYMBOL(set_memory_uc);

1008 1009
int _set_memory_array(unsigned long *addr, int addrinarray,
		unsigned long new_type)
1010
{
1011 1012 1013
	int i, j;
	int ret;

1014 1015 1016 1017
	/*
	 * for now UC MINUS. see comments in ioremap_nocache()
	 */
	for (i = 0; i < addrinarray; i++) {
1018
		ret = reserve_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE,
1019
					new_type, NULL);
1020 1021
		if (ret)
			goto out_free;
1022 1023
	}

1024
	ret = change_page_attr_set(addr, addrinarray,
1025
				    __pgprot(_PAGE_CACHE_UC_MINUS), 1);
1026 1027 1028 1029 1030 1031

	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);
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
	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;
1042
}
1043 1044 1045 1046 1047

int set_memory_array_uc(unsigned long *addr, int addrinarray)
{
	return _set_memory_array(addr, addrinarray, _PAGE_CACHE_UC_MINUS);
}
1048 1049
EXPORT_SYMBOL(set_memory_array_uc);

1050 1051 1052 1053 1054 1055
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);

1056 1057
int _set_memory_wc(unsigned long addr, int numpages)
{
1058
	int ret;
1059 1060
	unsigned long addr_copy = addr;

1061 1062 1063
	ret = change_page_attr_set(&addr, numpages,
				    __pgprot(_PAGE_CACHE_UC_MINUS), 0);
	if (!ret) {
1064 1065 1066 1067
		ret = change_page_attr_set_clr(&addr_copy, numpages,
					       __pgprot(_PAGE_CACHE_WC),
					       __pgprot(_PAGE_CACHE_MASK),
					       0, 0, NULL);
1068 1069
	}
	return ret;
1070 1071 1072 1073
}

int set_memory_wc(unsigned long addr, int numpages)
{
1074 1075
	int ret;

1076
	if (!pat_enabled)
1077 1078
		return set_memory_uc(addr, numpages);

1079 1080 1081 1082
	ret = reserve_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE,
		_PAGE_CACHE_WC, NULL);
	if (ret)
		goto out_err;
1083

1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
	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;
1094 1095 1096
}
EXPORT_SYMBOL(set_memory_wc);

1097
int _set_memory_wb(unsigned long addr, int numpages)
1098
{
1099 1100
	return change_page_attr_clear(&addr, numpages,
				      __pgprot(_PAGE_CACHE_MASK), 0);
1101
}
1102 1103 1104

int set_memory_wb(unsigned long addr, int numpages)
{
1105 1106 1107 1108 1109 1110
	int ret;

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

1111
	free_memtype(__pa(addr), __pa(addr) + numpages * PAGE_SIZE);
1112
	return 0;
1113
}
1114 1115
EXPORT_SYMBOL(set_memory_wb);

1116 1117 1118
int set_memory_array_wb(unsigned long *addr, int addrinarray)
{
	int i;
1119 1120 1121 1122
	int ret;

	ret = change_page_attr_clear(addr, addrinarray,
				      __pgprot(_PAGE_CACHE_MASK), 1);
1123 1124
	if (ret)
		return ret;
1125

1126 1127
	for (i = 0; i < addrinarray; i++)
		free_memtype(__pa(addr[i]), __pa(addr[i]) + PAGE_SIZE);
1128

1129
	return 0;
1130 1131 1132
}
EXPORT_SYMBOL(set_memory_array_wb);

1133 1134
int set_memory_x(unsigned long addr, int numpages)
{
1135 1136 1137
	if (!(__supported_pte_mask & _PAGE_NX))
		return 0;

1138
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_NX), 0);
1139 1140 1141 1142 1143
}
EXPORT_SYMBOL(set_memory_x);

int set_memory_nx(unsigned long addr, int numpages)
{
1144 1145 1146
	if (!(__supported_pte_mask & _PAGE_NX))
		return 0;

1147
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_NX), 0);
1148 1149 1150 1151 1152
}
EXPORT_SYMBOL(set_memory_nx);

int set_memory_ro(unsigned long addr, int numpages)
{
1153
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_RW), 0);
1154
}
1155
EXPORT_SYMBOL_GPL(set_memory_ro);
1156 1157 1158

int set_memory_rw(unsigned long addr, int numpages)
{
1159
	return change_page_attr_set(&addr, numpages, __pgprot(_PAGE_RW), 0);
1160
}
1161
EXPORT_SYMBOL_GPL(set_memory_rw);
I
Ingo Molnar 已提交
1162 1163 1164

int set_memory_np(unsigned long addr, int numpages)
{
1165
	return change_page_attr_clear(&addr, numpages, __pgprot(_PAGE_PRESENT), 0);
I
Ingo Molnar 已提交
1166
}
1167

1168 1169
int set_memory_4k(unsigned long addr, int numpages)
{
1170
	return change_page_attr_set_clr(&addr, numpages, __pgprot(0),
1171
					__pgprot(0), 1, 0, NULL);
1172 1173
}

1174 1175 1176 1177
int set_pages_uc(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1178
	return set_memory_uc(addr, numpages);
1179 1180 1181
}
EXPORT_SYMBOL(set_pages_uc);

1182 1183
static int _set_pages_array(struct page **pages, int addrinarray,
		unsigned long new_type)
1184 1185 1186 1187 1188
{
	unsigned long start;
	unsigned long end;
	int i;
	int free_idx;
1189
	int ret;
1190 1191

	for (i = 0; i < addrinarray; i++) {
1192 1193 1194
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1195
		end = start + PAGE_SIZE;
1196
		if (reserve_memtype(start, end, new_type, NULL))
1197 1198 1199
			goto err_out;
	}

1200 1201 1202 1203 1204 1205 1206 1207 1208 1209
	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 */
1210 1211 1212
err_out:
	free_idx = i;
	for (i = 0; i < free_idx; i++) {
1213 1214 1215
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1216 1217 1218 1219 1220
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}
	return -EINVAL;
}
1221 1222 1223 1224 1225

int set_pages_array_uc(struct page **pages, int addrinarray)
{
	return _set_pages_array(pages, addrinarray, _PAGE_CACHE_UC_MINUS);
}
1226 1227
EXPORT_SYMBOL(set_pages_array_uc);

1228 1229 1230 1231 1232 1233
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);

1234 1235 1236 1237
int set_pages_wb(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1238
	return set_memory_wb(addr, numpages);
1239 1240 1241
}
EXPORT_SYMBOL(set_pages_wb);

1242 1243 1244 1245 1246 1247 1248 1249 1250
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));
1251 1252
	if (retval)
		return retval;
1253 1254

	for (i = 0; i < addrinarray; i++) {
1255 1256 1257
		if (PageHighMem(pages[i]))
			continue;
		start = page_to_pfn(pages[i]) << PAGE_SHIFT;
1258 1259 1260 1261
		end = start + PAGE_SIZE;
		free_memtype(start, end);
	}

1262
	return 0;
1263 1264 1265
}
EXPORT_SYMBOL(set_pages_array_wb);

1266 1267 1268 1269
int set_pages_x(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);

T
Thomas Gleixner 已提交
1270
	return set_memory_x(addr, numpages);
1271 1272 1273 1274 1275 1276 1277
}
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
1278
	return set_memory_nx(addr, numpages);
1279 1280 1281 1282 1283 1284 1285
}
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
1286
	return set_memory_ro(addr, numpages);
1287 1288 1289 1290 1291
}

int set_pages_rw(struct page *page, int numpages)
{
	unsigned long addr = (unsigned long)page_address(page);
1292

T
Thomas Gleixner 已提交
1293
	return set_memory_rw(addr, numpages);
I
Ingo Molnar 已提交
1294 1295
}

L
Linus Torvalds 已提交
1296
#ifdef CONFIG_DEBUG_PAGEALLOC
I
Ingo Molnar 已提交
1297 1298 1299

static int __set_pages_p(struct page *page, int numpages)
{
1300 1301
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1302 1303
				.numpages = numpages,
				.mask_set = __pgprot(_PAGE_PRESENT | _PAGE_RW),
1304 1305
				.mask_clr = __pgprot(0),
				.flags = 0};
1306

1307 1308 1309 1310 1311 1312 1313
	/*
	 * 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
1314 1315 1316 1317
}

static int __set_pages_np(struct page *page, int numpages)
{
1318 1319
	unsigned long tempaddr = (unsigned long) page_address(page);
	struct cpa_data cpa = { .vaddr = &tempaddr,
T
Thomas Gleixner 已提交
1320 1321
				.numpages = numpages,
				.mask_set = __pgprot(0),
1322 1323
				.mask_clr = __pgprot(_PAGE_PRESENT | _PAGE_RW),
				.flags = 0};
1324

1325 1326 1327 1328 1329 1330 1331
	/*
	 * 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
1332 1333
}

L
Linus Torvalds 已提交
1334 1335 1336 1337
void kernel_map_pages(struct page *page, int numpages, int enable)
{
	if (PageHighMem(page))
		return;
1338
	if (!enable) {
1339 1340
		debug_check_no_locks_freed(page_address(page),
					   numpages * PAGE_SIZE);
1341
	}
1342

1343 1344 1345 1346 1347 1348
	/*
	 * If page allocator is not up yet then do not call c_p_a():
	 */
	if (!debug_pagealloc_enabled)
		return;

1349
	/*
I
Ingo Molnar 已提交
1350
	 * The return value is ignored as the calls cannot fail.
1351 1352
	 * 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
1353
	 */
I
Ingo Molnar 已提交
1354 1355 1356 1357
	if (enable)
		__set_pages_p(page, numpages);
	else
		__set_pages_np(page, numpages);
1358 1359

	/*
1360 1361
	 * We should perform an IPI and flush all tlbs,
	 * but that can deadlock->flush only current cpu:
L
Linus Torvalds 已提交
1362 1363
	 */
	__flush_tlb_all();
1364 1365
}

1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382
#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 */
1383 1384 1385 1386 1387 1388 1389 1390

/*
 * 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