[mirror_ubuntu-focal-kernel.git] / arch / x86 / mm / gup.c

/*
 * Lockless get_user_pages_fast for x86
 *
 * Copyright (C) 2008 Nick Piggin
 * Copyright (C) 2008 Novell Inc.
 */
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/vmstat.h>
#include <linux/highmem.h>
#include <linux/swap.h>
#include <linux/memremap.h>

#include <asm/mmu_context.h>
#include <asm/pgtable.h>

static inline pte_t gup_get_pte(pte_t *ptep)
{
#ifndef CONFIG_X86_PAE
	return READ_ONCE(*ptep);
#else
	/*
	 * With get_user_pages_fast, we walk down the pagetables without taking
	 * any locks.  For this we would like to load the pointers atomically,
	 * but that is not possible (without expensive cmpxchg8b) on PAE.  What
	 * we do have is the guarantee that a pte will only either go from not
	 * present to present, or present to not present or both -- it will not
	 * switch to a completely different present page without a TLB flush in
	 * between; something that we are blocking by holding interrupts off.
	 *
	 * Setting ptes from not present to present goes:
	 * ptep->pte_high = h;
	 * smp_wmb();
	 * ptep->pte_low = l;
	 *
	 * And present to not present goes:
	 * ptep->pte_low = 0;
	 * smp_wmb();
	 * ptep->pte_high = 0;
	 *
	 * We must ensure here that the load of pte_low sees l iff pte_high
	 * sees h. We load pte_high *after* loading pte_low, which ensures we
	 * don't see an older value of pte_high.  *Then* we recheck pte_low,
	 * which ensures that we haven't picked up a changed pte high. We might
	 * have got rubbish values from pte_low and pte_high, but we are
	 * guaranteed that pte_low will not have the present bit set *unless*
	 * it is 'l'. And get_user_pages_fast only operates on present ptes, so
	 * we're safe.
	 *
	 * gup_get_pte should not be used or copied outside gup.c without being
	 * very careful -- it does not atomically load the pte or anything that
	 * is likely to be useful for you.
	 */
	pte_t pte;

retry:
	pte.pte_low = ptep->pte_low;
	smp_rmb();
	pte.pte_high = ptep->pte_high;
	smp_rmb();
	if (unlikely(pte.pte_low != ptep->pte_low))
		goto retry;

	return pte;
#endif
}

static void undo_dev_pagemap(int *nr, int nr_start, struct page **pages)
{
	while ((*nr) - nr_start) {
		struct page *page = pages[--(*nr)];

		ClearPageReferenced(page);
		put_page(page);
	}
}

/*
 * 'pteval' can come from a pte, pmd or pud.  We only check
 * _PAGE_PRESENT, _PAGE_USER, and _PAGE_RW in here which are the
 * same value on all 3 types.
 */
static inline int pte_allows_gup(unsigned long pteval, int write)
{
	unsigned long need_pte_bits = _PAGE_PRESENT|_PAGE_USER;

	if (write)
		need_pte_bits |= _PAGE_RW;

	if ((pteval & need_pte_bits) != need_pte_bits)
		return 0;

	/* Check memory protection keys permissions. */
	if (!__pkru_allows_pkey(pte_flags_pkey(pteval), write))
		return 0;

	return 1;
}

/*
 * The performance critical leaf functions are made noinline otherwise gcc
 * inlines everything into a single function which results in too much
 * register pressure.
 */
static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
		unsigned long end, int write, struct page **pages, int *nr)
{
	struct dev_pagemap *pgmap = NULL;
	int nr_start = *nr;
	pte_t *ptep;

	ptep = pte_offset_map(&pmd, addr);
	do {
		pte_t pte = gup_get_pte(ptep);
		struct page *page;

		/* Similar to the PMD case, NUMA hinting must take slow path */
		if (pte_protnone(pte)) {
			pte_unmap(ptep);
			return 0;
		}

		if (pte_devmap(pte)) {
			pgmap = get_dev_pagemap(pte_pfn(pte), pgmap);
			if (unlikely(!pgmap)) {
				undo_dev_pagemap(nr, nr_start, pages);
				pte_unmap(ptep);
				return 0;
			}
		} else if (!pte_allows_gup(pte_val(pte), write) ||
			   pte_special(pte)) {
			pte_unmap(ptep);
			return 0;
		}
		VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
		page = pte_page(pte);
		get_page(page);
		put_dev_pagemap(pgmap);
		SetPageReferenced(page);
		pages[*nr] = page;
		(*nr)++;

	} while (ptep++, addr += PAGE_SIZE, addr != end);
	pte_unmap(ptep - 1);

	return 1;
}

static inline void get_head_page_multiple(struct page *page, int nr)
{
	VM_BUG_ON_PAGE(page != compound_head(page), page);
	VM_BUG_ON_PAGE(page_count(page) == 0, page);
	page_ref_add(page, nr);
	SetPageReferenced(page);
}

static int __gup_device_huge(unsigned long pfn, unsigned long addr,
		unsigned long end, struct page **pages, int *nr)
{
	int nr_start = *nr;
	struct dev_pagemap *pgmap = NULL;

	do {
		struct page *page = pfn_to_page(pfn);

		pgmap = get_dev_pagemap(pfn, pgmap);
		if (unlikely(!pgmap)) {
			undo_dev_pagemap(nr, nr_start, pages);
			return 0;
		}
		SetPageReferenced(page);
		pages[*nr] = page;
		get_page(page);
		put_dev_pagemap(pgmap);
		(*nr)++;
		pfn++;
	} while (addr += PAGE_SIZE, addr != end);
	return 1;
}

static int __gup_device_huge_pmd(pmd_t pmd, unsigned long addr,
		unsigned long end, struct page **pages, int *nr)
{
	unsigned long fault_pfn;

	fault_pfn = pmd_pfn(pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
	return __gup_device_huge(fault_pfn, addr, end, pages, nr);
}

static int __gup_device_huge_pud(pud_t pud, unsigned long addr,
		unsigned long end, struct page **pages, int *nr)
{
	unsigned long fault_pfn;

	fault_pfn = pud_pfn(pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
	return __gup_device_huge(fault_pfn, addr, end, pages, nr);
}

static noinline int gup_huge_pmd(pmd_t pmd, unsigned long addr,
		unsigned long end, int write, struct page **pages, int *nr)
{
	struct page *head, *page;
	int refs;

	if (!pte_allows_gup(pmd_val(pmd), write))
		return 0;

	VM_BUG_ON(!pfn_valid(pmd_pfn(pmd)));
	if (pmd_devmap(pmd))
		return __gup_device_huge_pmd(pmd, addr, end, pages, nr);

	/* hugepages are never "special" */
	VM_BUG_ON(pmd_flags(pmd) & _PAGE_SPECIAL);

	refs = 0;
	head = pmd_page(pmd);
	page = head + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
	do {
		VM_BUG_ON_PAGE(compound_head(page) != head, page);
		pages[*nr] = page;
		(*nr)++;
		page++;
		refs++;
	} while (addr += PAGE_SIZE, addr != end);
	get_head_page_multiple(head, refs);

	return 1;
}

static int gup_pmd_range(pud_t pud, unsigned long addr, unsigned long end,
		int write, struct page **pages, int *nr)
{
	unsigned long next;
	pmd_t *pmdp;

	pmdp = pmd_offset(&pud, addr);
	do {
		pmd_t pmd = *pmdp;

		next = pmd_addr_end(addr, end);
		if (pmd_none(pmd))
			return 0;
		if (unlikely(pmd_large(pmd) || !pmd_present(pmd))) {
			/*
			 * NUMA hinting faults need to be handled in the GUP
			 * slowpath for accounting purposes and so that they
			 * can be serialised against THP migration.
			 */
			if (pmd_protnone(pmd))
				return 0;
			if (!gup_huge_pmd(pmd, addr, next, write, pages, nr))
				return 0;
		} else {
			if (!gup_pte_range(pmd, addr, next, write, pages, nr))
				return 0;
		}
	} while (pmdp++, addr = next, addr != end);

	return 1;
}

static noinline int gup_huge_pud(pud_t pud, unsigned long addr,
		unsigned long end, int write, struct page **pages, int *nr)
{
	struct page *head, *page;
	int refs;

	if (!pte_allows_gup(pud_val(pud), write))
		return 0;

	VM_BUG_ON(!pfn_valid(pud_pfn(pud)));
	if (pud_devmap(pud))
		return __gup_device_huge_pud(pud, addr, end, pages, nr);

	/* hugepages are never "special" */
	VM_BUG_ON(pud_flags(pud) & _PAGE_SPECIAL);

	refs = 0;
	head = pud_page(pud);
	page = head + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
	do {
		VM_BUG_ON_PAGE(compound_head(page) != head, page);
		pages[*nr] = page;
		(*nr)++;
		page++;
		refs++;
	} while (addr += PAGE_SIZE, addr != end);
	get_head_page_multiple(head, refs);

	return 1;
}

static int gup_pud_range(pgd_t pgd, unsigned long addr, unsigned long end,
			int write, struct page **pages, int *nr)
{
	unsigned long next;
	pud_t *pudp;

	pudp = pud_offset(&pgd, addr);
	do {
		pud_t pud = *pudp;

		next = pud_addr_end(addr, end);
		if (pud_none(pud))
			return 0;
		if (unlikely(pud_large(pud))) {
			if (!gup_huge_pud(pud, addr, next, write, pages, nr))
				return 0;
		} else {
			if (!gup_pmd_range(pud, addr, next, write, pages, nr))
				return 0;
		}
	} while (pudp++, addr = next, addr != end);

	return 1;
}

/*
 * Like get_user_pages_fast() except its IRQ-safe in that it won't fall
 * back to the regular GUP.
 */
int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
			  struct page **pages)
{
	struct mm_struct *mm = current->mm;
	unsigned long addr, len, end;
	unsigned long next;
	unsigned long flags;
	pgd_t *pgdp;
	int nr = 0;

	start &= PAGE_MASK;
	addr = start;
	len = (unsigned long) nr_pages << PAGE_SHIFT;
	end = start + len;
	if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ,
					(void __user *)start, len)))
		return 0;

	/*
	 * XXX: batch / limit 'nr', to avoid large irq off latency
	 * needs some instrumenting to determine the common sizes used by
	 * important workloads (eg. DB2), and whether limiting the batch size
	 * will decrease performance.
	 *
	 * It seems like we're in the clear for the moment. Direct-IO is
	 * the main guy that batches up lots of get_user_pages, and even
	 * they are limited to 64-at-a-time which is not so many.
	 */
	/*
	 * This doesn't prevent pagetable teardown, but does prevent
	 * the pagetables and pages from being freed on x86.
	 *
	 * So long as we atomically load page table pointers versus teardown
	 * (which we do on x86, with the above PAE exception), we can follow the
	 * address down to the the page and take a ref on it.
	 */
	local_irq_save(flags);
	pgdp = pgd_offset(mm, addr);
	do {
		pgd_t pgd = *pgdp;

		next = pgd_addr_end(addr, end);
		if (pgd_none(pgd))
			break;
		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
			break;
	} while (pgdp++, addr = next, addr != end);
	local_irq_restore(flags);

	return nr;
}

/**
 * get_user_pages_fast() - pin user pages in memory
 * @start:	starting user address
 * @nr_pages:	number of pages from start to pin
 * @write:	whether pages will be written to
 * @pages:	array that receives pointers to the pages pinned.
 * 		Should be at least nr_pages long.
 *
 * Attempt to pin user pages in memory without taking mm->mmap_sem.
 * If not successful, it will fall back to taking the lock and
 * calling get_user_pages().
 *
 * Returns number of pages pinned. This may be fewer than the number
 * requested. If nr_pages is 0 or negative, returns 0. If no pages
 * were pinned, returns -errno.
 */
int get_user_pages_fast(unsigned long start, int nr_pages, int write,
			struct page **pages)
{
	struct mm_struct *mm = current->mm;
	unsigned long addr, len, end;
	unsigned long next;
	pgd_t *pgdp;
	int nr = 0;

	start &= PAGE_MASK;
	addr = start;
	len = (unsigned long) nr_pages << PAGE_SHIFT;

	end = start + len;
	if (end < start)
		goto slow_irqon;

#ifdef CONFIG_X86_64
	if (end >> __VIRTUAL_MASK_SHIFT)
		goto slow_irqon;
#endif

	/*
	 * XXX: batch / limit 'nr', to avoid large irq off latency
	 * needs some instrumenting to determine the common sizes used by
	 * important workloads (eg. DB2), and whether limiting the batch size
	 * will decrease performance.
	 *
	 * It seems like we're in the clear for the moment. Direct-IO is
	 * the main guy that batches up lots of get_user_pages, and even
	 * they are limited to 64-at-a-time which is not so many.
	 */
	/*
	 * This doesn't prevent pagetable teardown, but does prevent
	 * the pagetables and pages from being freed on x86.
	 *
	 * So long as we atomically load page table pointers versus teardown
	 * (which we do on x86, with the above PAE exception), we can follow the
	 * address down to the the page and take a ref on it.
	 */
	local_irq_disable();
	pgdp = pgd_offset(mm, addr);
	do {
		pgd_t pgd = *pgdp;

		next = pgd_addr_end(addr, end);
		if (pgd_none(pgd))
			goto slow;
		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
			goto slow;
	} while (pgdp++, addr = next, addr != end);
	local_irq_enable();

	VM_BUG_ON(nr != (end - start) >> PAGE_SHIFT);
	return nr;

	{
		int ret;

slow:
		local_irq_enable();
slow_irqon:
		/* Try to get the remaining pages with get_user_pages */
		start += nr << PAGE_SHIFT;
		pages += nr;

		ret = get_user_pages_unlocked(start,
					      (end - start) >> PAGE_SHIFT,
					      pages, write ? FOLL_WRITE : 0);

		/* Have to be a bit careful with return values */
		if (nr > 0) {
			if (ret < 0)
				ret = nr;
			else
				ret += nr;
		}

		return ret;
	}
}
Commit	Line	Data
8174c430 NP	1	/*
	2	* Lockless get_user_pages_fast for x86
	3	*
	4	* Copyright (C) 2008 Nick Piggin
	5	* Copyright (C) 2008 Novell Inc.
	6	*/
	7	#include <linux/sched.h>
	8	#include <linux/mm.h>
	9	#include <linux/vmstat.h>
	10	#include <linux/highmem.h>
8ee53820	11	#include <linux/swap.h>
3565fce3	12	#include <linux/memremap.h>
8174c430	13
33a709b2	14	#include <asm/mmu_context.h>
8174c430 NP	15	#include <asm/pgtable.h>
	16
	17	static inline pte_t gup_get_pte(pte_t *ptep)
	18	{
	19	#ifndef CONFIG_X86_PAE
14cf3d97	20	return READ_ONCE(*ptep);
8174c430 NP	21	#else
	22	/*
	23	* With get_user_pages_fast, we walk down the pagetables without taking
ab09809f	24	* any locks. For this we would like to load the pointers atomically,
8174c430 NP	25	* but that is not possible (without expensive cmpxchg8b) on PAE. What
	26	* we do have is the guarantee that a pte will only either go from not
	27	* present to present, or present to not present or both -- it will not
	28	* switch to a completely different present page without a TLB flush in
	29	* between; something that we are blocking by holding interrupts off.
	30	*
	31	* Setting ptes from not present to present goes:
	32	* ptep->pte_high = h;
	33	* smp_wmb();
	34	* ptep->pte_low = l;
	35	*
	36	* And present to not present goes:
	37	* ptep->pte_low = 0;
	38	* smp_wmb();
	39	* ptep->pte_high = 0;
	40	*
	41	* We must ensure here that the load of pte_low sees l iff pte_high
	42	* sees h. We load pte_high after loading pte_low, which ensures we
	43	* don't see an older value of pte_high. Then we recheck pte_low,
	44	* which ensures that we haven't picked up a changed pte high. We might
	45	* have got rubbish values from pte_low and pte_high, but we are
	46	* guaranteed that pte_low will not have the present bit set unless
	47	* it is 'l'. And get_user_pages_fast only operates on present ptes, so
	48	* we're safe.
	49	*
	50	* gup_get_pte should not be used or copied outside gup.c without being
	51	* very careful -- it does not atomically load the pte or anything that
	52	* is likely to be useful for you.
	53	*/
	54	pte_t pte;
	55
	56	retry:
	57	pte.pte_low = ptep->pte_low;
	58	smp_rmb();
	59	pte.pte_high = ptep->pte_high;
	60	smp_rmb();
	61	if (unlikely(pte.pte_low != ptep->pte_low))
	62	goto retry;
	63
	64	return pte;
	65	#endif
	66	}
	67
3565fce3 DW	68	static void undo_dev_pagemap(int nr, int nr_start, struct page *pages)
	69	{
	70	while ((*nr) - nr_start) {
	71	struct page page = pages[--(nr)];
	72
	73	ClearPageReferenced(page);
	74	put_page(page);
	75	}
	76	}
	77
1874f689 DH	78	/*
	79	* 'pteval' can come from a pte, pmd or pud. We only check
	80	* _PAGE_PRESENT, _PAGE_USER, and _PAGE_RW in here which are the
	81	* same value on all 3 types.
	82	*/
	83	static inline int pte_allows_gup(unsigned long pteval, int write)
	84	{
	85	unsigned long need_pte_bits = _PAGE_PRESENT\|_PAGE_USER;
	86
	87	if (write)
	88	need_pte_bits \|= _PAGE_RW;
	89
	90	if ((pteval & need_pte_bits) != need_pte_bits)
	91	return 0;
	92
33a709b2 DH	93	/* Check memory protection keys permissions. */
	94	if (!__pkru_allows_pkey(pte_flags_pkey(pteval), write))
	95	return 0;
	96
1874f689 DH	97	return 1;
	98	}
	99
8174c430 NP	100	/*
	101	* The performance critical leaf functions are made noinline otherwise gcc
	102	* inlines everything into a single function which results in too much
	103	* register pressure.
	104	*/
	105	static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
	106	unsigned long end, int write, struct page *pages, int nr)
	107	{
3565fce3	108	struct dev_pagemap *pgmap = NULL;
3565fce3	109	int nr_start = *nr;
8174c430 NP	110	pte_t *ptep;
8174c430 NP	111
8174c430 NP	112	ptep = pte_offset_map(&pmd, addr);
	113	do {
	114	pte_t pte = gup_get_pte(ptep);
	115	struct page *page;
	116
2b4847e7	117	/* Similar to the PMD case, NUMA hinting must take slow path */
8a0516ed	118	if (pte_protnone(pte)) {
2b4847e7 MG	119	pte_unmap(ptep);
	120	return 0;
	121	}
	122
3565fce3 DW	123	if (pte_devmap(pte)) {
	124	pgmap = get_dev_pagemap(pte_pfn(pte), pgmap);
	125	if (unlikely(!pgmap)) {
	126	undo_dev_pagemap(nr, nr_start, pages);
	127	pte_unmap(ptep);
	128	return 0;
	129	}
1874f689 DH	130	} else if (!pte_allows_gup(pte_val(pte), write) \|\|
1874f689 DH	131	pte_special(pte)) {
8174c430 NP	132	pte_unmap(ptep);
	133	return 0;
	134	}
	135	VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
457a98b0	136	page = pte_page(pte);
8174c430	137	get_page(page);
3565fce3	138	put_dev_pagemap(pgmap);
8ee53820	139	SetPageReferenced(page);
8174c430 NP	140	pages[*nr] = page;
	141	(*nr)++;
	142
	143	} while (ptep++, addr += PAGE_SIZE, addr != end);
	144	pte_unmap(ptep - 1);
	145
	146	return 1;
	147	}
	148
	149	static inline void get_head_page_multiple(struct page *page, int nr)
	150	{
309381fe SL	151	VM_BUG_ON_PAGE(page != compound_head(page), page);
309381fe SL	152	VM_BUG_ON_PAGE(page_count(page) == 0, page);
fe896d18	153	page_ref_add(page, nr);
8ee53820	154	SetPageReferenced(page);
8174c430 NP	155	}
8174c430 NP	156
220ced16	157	static int __gup_device_huge(unsigned long pfn, unsigned long addr,
3565fce3 DW	158	unsigned long end, struct page *pages, int nr)
	159	{
	160	int nr_start = *nr;
3565fce3 DW	161	struct dev_pagemap *pgmap = NULL;
3565fce3 DW	162
3565fce3 DW	163	do {
	164	struct page *page = pfn_to_page(pfn);
	165
	166	pgmap = get_dev_pagemap(pfn, pgmap);
	167	if (unlikely(!pgmap)) {
	168	undo_dev_pagemap(nr, nr_start, pages);
	169	return 0;
	170	}
	171	SetPageReferenced(page);
	172	pages[*nr] = page;
	173	get_page(page);
	174	put_dev_pagemap(pgmap);
	175	(*nr)++;
	176	pfn++;
	177	} while (addr += PAGE_SIZE, addr != end);
	178	return 1;
	179	}
	180
220ced16 DW	181	static int __gup_device_huge_pmd(pmd_t pmd, unsigned long addr,
	182	unsigned long end, struct page *pages, int nr)
	183	{
	184	unsigned long fault_pfn;
	185
	186	fault_pfn = pmd_pfn(pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
	187	return __gup_device_huge(fault_pfn, addr, end, pages, nr);
	188	}
	189
	190	static int __gup_device_huge_pud(pud_t pud, unsigned long addr,
	191	unsigned long end, struct page *pages, int nr)
	192	{
	193	unsigned long fault_pfn;
	194
	195	fault_pfn = pud_pfn(pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
	196	return __gup_device_huge(fault_pfn, addr, end, pages, nr);
	197	}
	198
8174c430 NP	199	static noinline int gup_huge_pmd(pmd_t pmd, unsigned long addr,
	200	unsigned long end, int write, struct page *pages, int nr)
	201	{
8174c430 NP	202	struct page head, page;
	203	int refs;
	204
1874f689	205	if (!pte_allows_gup(pmd_val(pmd), write))
8174c430	206	return 0;
3565fce3 DW	207
	208	VM_BUG_ON(!pfn_valid(pmd_pfn(pmd)));
	209	if (pmd_devmap(pmd))
	210	return __gup_device_huge_pmd(pmd, addr, end, pages, nr);
	211
8174c430	212	/* hugepages are never "special" */
daf3e35c	213	VM_BUG_ON(pmd_flags(pmd) & _PAGE_SPECIAL);
8174c430 NP	214
8174c430 NP	215	refs = 0;
daf3e35c	216	head = pmd_page(pmd);
652ea695	217	page = head + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
8174c430	218	do {
309381fe	219	VM_BUG_ON_PAGE(compound_head(page) != head, page);
8174c430 NP	220	pages[*nr] = page;
	221	(*nr)++;
	222	page++;
	223	refs++;
	224	} while (addr += PAGE_SIZE, addr != end);
	225	get_head_page_multiple(head, refs);
	226
	227	return 1;
	228	}
	229
	230	static int gup_pmd_range(pud_t pud, unsigned long addr, unsigned long end,
	231	int write, struct page *pages, int nr)
	232	{
	233	unsigned long next;
	234	pmd_t *pmdp;
	235
	236	pmdp = pmd_offset(&pud, addr);
	237	do {
	238	pmd_t pmd = *pmdp;
	239
	240	next = pmd_addr_end(addr, end);
1f19617d	241	if (pmd_none(pmd))
8174c430	242	return 0;
cbef8478	243	if (unlikely(pmd_large(pmd) \|\| !pmd_present(pmd))) {
2b4847e7 MG	244	/*
	245	* NUMA hinting faults need to be handled in the GUP
	246	* slowpath for accounting purposes and so that they
	247	* can be serialised against THP migration.
	248	*/
8a0516ed	249	if (pmd_protnone(pmd))
2b4847e7	250	return 0;
8174c430 NP	251	if (!gup_huge_pmd(pmd, addr, next, write, pages, nr))
	252	return 0;
	253	} else {
	254	if (!gup_pte_range(pmd, addr, next, write, pages, nr))
	255	return 0;
	256	}
	257	} while (pmdp++, addr = next, addr != end);
	258
	259	return 1;
	260	}
	261
652ea695 NP	262	static noinline int gup_huge_pud(pud_t pud, unsigned long addr,
	263	unsigned long end, int write, struct page *pages, int nr)
	264	{
652ea695 NP	265	struct page head, page;
	266	int refs;
	267
1874f689	268	if (!pte_allows_gup(pud_val(pud), write))
652ea695	269	return 0;
220ced16 DW	270
	271	VM_BUG_ON(!pfn_valid(pud_pfn(pud)));
	272	if (pud_devmap(pud))
	273	return __gup_device_huge_pud(pud, addr, end, pages, nr);
	274
652ea695	275	/* hugepages are never "special" */
daf3e35c	276	VM_BUG_ON(pud_flags(pud) & _PAGE_SPECIAL);
652ea695 NP	277
652ea695 NP	278	refs = 0;
daf3e35c	279	head = pud_page(pud);
652ea695 NP	280	page = head + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
652ea695 NP	281	do {
309381fe	282	VM_BUG_ON_PAGE(compound_head(page) != head, page);
652ea695 NP	283	pages[*nr] = page;
	284	(*nr)++;
	285	page++;
	286	refs++;
	287	} while (addr += PAGE_SIZE, addr != end);
	288	get_head_page_multiple(head, refs);
	289
	290	return 1;
	291	}
	292
8174c430 NP	293	static int gup_pud_range(pgd_t pgd, unsigned long addr, unsigned long end,
	294	int write, struct page *pages, int nr)
	295	{
	296	unsigned long next;
	297	pud_t *pudp;
	298
	299	pudp = pud_offset(&pgd, addr);
	300	do {
	301	pud_t pud = *pudp;
	302
	303	next = pud_addr_end(addr, end);
	304	if (pud_none(pud))
	305	return 0;
652ea695 NP	306	if (unlikely(pud_large(pud))) {
	307	if (!gup_huge_pud(pud, addr, next, write, pages, nr))
	308	return 0;
	309	} else {
	310	if (!gup_pmd_range(pud, addr, next, write, pages, nr))
	311	return 0;
	312	}
8174c430 NP	313	} while (pudp++, addr = next, addr != end);
	314
	315	return 1;
	316	}
	317
465a454f PZ	318	/*
	319	* Like get_user_pages_fast() except its IRQ-safe in that it won't fall
	320	* back to the regular GUP.
	321	*/
	322	int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
	323	struct page **pages)
	324	{
	325	struct mm_struct *mm = current->mm;
	326	unsigned long addr, len, end;
	327	unsigned long next;
	328	unsigned long flags;
	329	pgd_t *pgdp;
	330	int nr = 0;
	331
	332	start &= PAGE_MASK;
	333	addr = start;
	334	len = (unsigned long) nr_pages << PAGE_SHIFT;
	335	end = start + len;
	336	if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ,
	337	(void __user *)start, len)))
	338	return 0;
	339
	340	/*
	341	* XXX: batch / limit 'nr', to avoid large irq off latency
	342	* needs some instrumenting to determine the common sizes used by
	343	* important workloads (eg. DB2), and whether limiting the batch size
	344	* will decrease performance.
	345	*
	346	* It seems like we're in the clear for the moment. Direct-IO is
	347	* the main guy that batches up lots of get_user_pages, and even
	348	* they are limited to 64-at-a-time which is not so many.
	349	*/
	350	/*
	351	* This doesn't prevent pagetable teardown, but does prevent
	352	* the pagetables and pages from being freed on x86.
	353	*
	354	* So long as we atomically load page table pointers versus teardown
	355	* (which we do on x86, with the above PAE exception), we can follow the
	356	* address down to the the page and take a ref on it.
	357	*/
	358	local_irq_save(flags);
	359	pgdp = pgd_offset(mm, addr);
	360	do {
	361	pgd_t pgd = *pgdp;
	362
	363	next = pgd_addr_end(addr, end);
	364	if (pgd_none(pgd))
	365	break;
	366	if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
	367	break;
	368	} while (pgdp++, addr = next, addr != end);
	369	local_irq_restore(flags);
	370
	371	return nr;
	372	}
	373
a0d22f48 AG	374	/**
	375	* get_user_pages_fast() - pin user pages in memory
	376	* @start: starting user address
	377	* @nr_pages: number of pages from start to pin
	378	* @write: whether pages will be written to
	379	* @pages: array that receives pointers to the pages pinned.
	380	* Should be at least nr_pages long.
	381	*
	382	* Attempt to pin user pages in memory without taking mm->mmap_sem.
	383	* If not successful, it will fall back to taking the lock and
	384	* calling get_user_pages().
	385	*
	386	* Returns number of pages pinned. This may be fewer than the number
	387	* requested. If nr_pages is 0 or negative, returns 0. If no pages
	388	* were pinned, returns -errno.
	389	*/
8174c430 NP	390	int get_user_pages_fast(unsigned long start, int nr_pages, int write,
	391	struct page **pages)
	392	{
	393	struct mm_struct *mm = current->mm;
9b79022c	394	unsigned long addr, len, end;
8174c430 NP	395	unsigned long next;
	396	pgd_t *pgdp;
	397	int nr = 0;
	398
9b79022c LT	399	start &= PAGE_MASK;
	400	addr = start;
	401	len = (unsigned long) nr_pages << PAGE_SHIFT;
7f818906	402
9b79022c	403	end = start + len;
7f818906 LT	404	if (end < start)
	405	goto slow_irqon;
	406
	407	#ifdef CONFIG_X86_64
	408	if (end >> __VIRTUAL_MASK_SHIFT)
8174c430	409	goto slow_irqon;
7f818906	410	#endif
8174c430 NP	411
	412	/*
	413	* XXX: batch / limit 'nr', to avoid large irq off latency
	414	* needs some instrumenting to determine the common sizes used by
	415	* important workloads (eg. DB2), and whether limiting the batch size
	416	* will decrease performance.
	417	*
	418	* It seems like we're in the clear for the moment. Direct-IO is
	419	* the main guy that batches up lots of get_user_pages, and even
	420	* they are limited to 64-at-a-time which is not so many.
	421	*/
	422	/*
	423	* This doesn't prevent pagetable teardown, but does prevent
	424	* the pagetables and pages from being freed on x86.
	425	*
	426	* So long as we atomically load page table pointers versus teardown
	427	* (which we do on x86, with the above PAE exception), we can follow the
	428	* address down to the the page and take a ref on it.
	429	*/
	430	local_irq_disable();
	431	pgdp = pgd_offset(mm, addr);
	432	do {
	433	pgd_t pgd = *pgdp;
	434
	435	next = pgd_addr_end(addr, end);
	436	if (pgd_none(pgd))
	437	goto slow;
	438	if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
	439	goto slow;
	440	} while (pgdp++, addr = next, addr != end);
	441	local_irq_enable();
	442
	443	VM_BUG_ON(nr != (end - start) >> PAGE_SHIFT);
	444	return nr;
	445
	446	{
	447	int ret;
	448
	449	slow:
	450	local_irq_enable();
	451	slow_irqon:
	452	/* Try to get the remaining pages with get_user_pages */
	453	start += nr << PAGE_SHIFT;
	454	pages += nr;
	455
d4edcf0d	456	ret = get_user_pages_unlocked(start,
a7b78075	457	(end - start) >> PAGE_SHIFT,
c164154f	458	pages, write ? FOLL_WRITE : 0);
8174c430 NP	459
	460	/* Have to be a bit careful with return values */
	461	if (nr > 0) {
	462	if (ret < 0)
	463	ret = nr;
	464	else
	465	ret += nr;
	466	}
	467
	468	return ret;
	469	}
	470	}