1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_MM_H
3#define _LINUX_MM_H
4
5#include <linux/errno.h>
6
7#ifdef __KERNEL__
8
9#include <linux/mmdebug.h>
10#include <linux/gfp.h>
11#include <linux/bug.h>
12#include <linux/list.h>
13#include <linux/mmzone.h>
14#include <linux/rbtree.h>
15#include <linux/atomic.h>
16#include <linux/debug_locks.h>
17#include <linux/mm_types.h>
18#include <linux/range.h>
19#include <linux/pfn.h>
20#include <linux/percpu-refcount.h>
21#include <linux/bit_spinlock.h>
22#include <linux/shrinker.h>
23#include <linux/resource.h>
24#include <linux/page_ext.h>
25#include <linux/err.h>
26#include <linux/page_ref.h>
27#include <linux/memremap.h>
28#include <linux/overflow.h>
29
30struct mempolicy;
31struct anon_vma;
32struct anon_vma_chain;
33struct file_ra_state;
34struct user_struct;
35struct writeback_control;
36struct bdi_writeback;
37
38void init_mm_internals(void);
39
40#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
41extern unsigned long max_mapnr;
42
43static inline void set_max_mapnr(unsigned long limit)
44{
45 max_mapnr = limit;
46}
47#else
48static inline void set_max_mapnr(unsigned long limit) { }
49#endif
50
51extern unsigned long totalram_pages;
52extern void * high_memory;
53extern int page_cluster;
54
55#ifdef CONFIG_SYSCTL
56extern int sysctl_legacy_va_layout;
57#else
58#define sysctl_legacy_va_layout 0
59#endif
60
61#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
62extern const int mmap_rnd_bits_min;
63extern const int mmap_rnd_bits_max;
64extern int mmap_rnd_bits __read_mostly;
65#endif
66#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
67extern const int mmap_rnd_compat_bits_min;
68extern const int mmap_rnd_compat_bits_max;
69extern int mmap_rnd_compat_bits __read_mostly;
70#endif
71
72#include <asm/page.h>
73#include <asm/pgtable.h>
74#include <asm/processor.h>
75
76#ifndef __pa_symbol
77#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
78#endif
79
80#ifndef page_to_virt
81#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
82#endif
83
84#ifndef lm_alias
85#define lm_alias(x) __va(__pa_symbol(x))
86#endif
87
88/*
89 * To prevent common memory management code establishing
90 * a zero page mapping on a read fault.
91 * This macro should be defined within <asm/pgtable.h>.
92 * s390 does this to prevent multiplexing of hardware bits
93 * related to the physical page in case of virtualization.
94 */
95#ifndef mm_forbids_zeropage
96#define mm_forbids_zeropage(X) (0)
97#endif
98
99/*
100 * On some architectures it is expensive to call memset() for small sizes.
101 * Those architectures should provide their own implementation of "struct page"
102 * zeroing by defining this macro in <asm/pgtable.h>.
103 */
104#ifndef mm_zero_struct_page
105#define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
106#endif
107
108/*
109 * Default maximum number of active map areas, this limits the number of vmas
110 * per mm struct. Users can overwrite this number by sysctl but there is a
111 * problem.
112 *
113 * When a program's coredump is generated as ELF format, a section is created
114 * per a vma. In ELF, the number of sections is represented in unsigned short.
115 * This means the number of sections should be smaller than 65535 at coredump.
116 * Because the kernel adds some informative sections to a image of program at
117 * generating coredump, we need some margin. The number of extra sections is
118 * 1-3 now and depends on arch. We use "5" as safe margin, here.
119 *
120 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
121 * not a hard limit any more. Although some userspace tools can be surprised by
122 * that.
123 */
124#define MAPCOUNT_ELF_CORE_MARGIN (5)
125#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
126
127extern int sysctl_max_map_count;
128
129extern unsigned long sysctl_user_reserve_kbytes;
130extern unsigned long sysctl_admin_reserve_kbytes;
131
132extern int sysctl_overcommit_memory;
133extern int sysctl_overcommit_ratio;
134extern unsigned long sysctl_overcommit_kbytes;
135
136extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
137 size_t *, loff_t *);
138extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
139 size_t *, loff_t *);
140
141#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
142
143/* to align the pointer to the (next) page boundary */
144#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
145
146/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
147#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
148
149/*
150 * Linux kernel virtual memory manager primitives.
151 * The idea being to have a "virtual" mm in the same way
152 * we have a virtual fs - giving a cleaner interface to the
153 * mm details, and allowing different kinds of memory mappings
154 * (from shared memory to executable loading to arbitrary
155 * mmap() functions).
156 */
157
158struct vm_area_struct *vm_area_alloc(struct mm_struct *);
159struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
160void vm_area_free(struct vm_area_struct *);
161
162#ifndef CONFIG_MMU
163extern struct rb_root nommu_region_tree;
164extern struct rw_semaphore nommu_region_sem;
165
166extern unsigned int kobjsize(const void *objp);
167#endif
168
169/*
170 * vm_flags in vm_area_struct, see mm_types.h.
171 * When changing, update also include/trace/events/mmflags.h
172 */
173#define VM_NONE 0x00000000
174
175#define VM_READ 0x00000001 /* currently active flags */
176#define VM_WRITE 0x00000002
177#define VM_EXEC 0x00000004
178#define VM_SHARED 0x00000008
179
180/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
181#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
182#define VM_MAYWRITE 0x00000020
183#define VM_MAYEXEC 0x00000040
184#define VM_MAYSHARE 0x00000080
185
186#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
187#define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
188#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
189#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
190#define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
191
192#define VM_LOCKED 0x00002000
193#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
194
195 /* Used by sys_madvise() */
196#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
197#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
198
199#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
200#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
201#define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
202#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
203#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
204#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
205#define VM_SYNC 0x00800000 /* Synchronous page faults */
206#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
207#define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
208#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
209
210#ifdef CONFIG_MEM_SOFT_DIRTY
211# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
212#else
213# define VM_SOFTDIRTY 0
214#endif
215
216#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
217#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
218#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
219#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
220
221#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
222#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
223#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
224#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
225#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
226#define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
227#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
228#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
229#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
230#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
231#define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
232#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
233
234#ifdef CONFIG_ARCH_HAS_PKEYS
235# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
236# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
237# define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
238# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
239# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
240#ifdef CONFIG_PPC
241# define VM_PKEY_BIT4 VM_HIGH_ARCH_4
242#else
243# define VM_PKEY_BIT4 0
244#endif
245#endif /* CONFIG_ARCH_HAS_PKEYS */
246
247#if defined(CONFIG_X86)
248# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
249#elif defined(CONFIG_PPC)
250# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
251#elif defined(CONFIG_PARISC)
252# define VM_GROWSUP VM_ARCH_1
253#elif defined(CONFIG_IA64)
254# define VM_GROWSUP VM_ARCH_1
255#elif defined(CONFIG_SPARC64)
256# define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
257# define VM_ARCH_CLEAR VM_SPARC_ADI
258#elif !defined(CONFIG_MMU)
259# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
260#endif
261
262#if defined(CONFIG_X86_INTEL_MPX)
263/* MPX specific bounds table or bounds directory */
264# define VM_MPX VM_HIGH_ARCH_4
265#else
266# define VM_MPX VM_NONE
267#endif
268
269#ifndef VM_GROWSUP
270# define VM_GROWSUP VM_NONE
271#endif
272
273/* Bits set in the VMA until the stack is in its final location */
274#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
275
276#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
277#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
278#endif
279
280#ifdef CONFIG_STACK_GROWSUP
281#define VM_STACK VM_GROWSUP
282#else
283#define VM_STACK VM_GROWSDOWN
284#endif
285
286#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
287
288/*
289 * Special vmas that are non-mergable, non-mlock()able.
290 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
291 */
292#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
293
294/* This mask defines which mm->def_flags a process can inherit its parent */
295#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
296
297/* This mask is used to clear all the VMA flags used by mlock */
298#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
299
300/* Arch-specific flags to clear when updating VM flags on protection change */
301#ifndef VM_ARCH_CLEAR
302# define VM_ARCH_CLEAR VM_NONE
303#endif
304#define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
305
306/*
307 * mapping from the currently active vm_flags protection bits (the
308 * low four bits) to a page protection mask..
309 */
310extern pgprot_t protection_map[16];
311
312#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
313#define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
314#define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
315#define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
316#define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
317#define FAULT_FLAG_TRIED 0x20 /* Second try */
318#define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
319#define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
320#define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
321
322#define FAULT_FLAG_TRACE \
323 { FAULT_FLAG_WRITE, "WRITE" }, \
324 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
325 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
326 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
327 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
328 { FAULT_FLAG_TRIED, "TRIED" }, \
329 { FAULT_FLAG_USER, "USER" }, \
330 { FAULT_FLAG_REMOTE, "REMOTE" }, \
331 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
332
333/*
334 * vm_fault is filled by the the pagefault handler and passed to the vma's
335 * ->fault function. The vma's ->fault is responsible for returning a bitmask
336 * of VM_FAULT_xxx flags that give details about how the fault was handled.
337 *
338 * MM layer fills up gfp_mask for page allocations but fault handler might
339 * alter it if its implementation requires a different allocation context.
340 *
341 * pgoff should be used in favour of virtual_address, if possible.
342 */
343struct vm_fault {
344 struct vm_area_struct *vma; /* Target VMA */
345 unsigned int flags; /* FAULT_FLAG_xxx flags */
346 gfp_t gfp_mask; /* gfp mask to be used for allocations */
347 pgoff_t pgoff; /* Logical page offset based on vma */
348 unsigned long address; /* Faulting virtual address */
349 pmd_t *pmd; /* Pointer to pmd entry matching
350 * the 'address' */
351 pud_t *pud; /* Pointer to pud entry matching
352 * the 'address'
353 */
354 pte_t orig_pte; /* Value of PTE at the time of fault */
355
356 struct page *cow_page; /* Page handler may use for COW fault */
357 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
358 struct page *page; /* ->fault handlers should return a
359 * page here, unless VM_FAULT_NOPAGE
360 * is set (which is also implied by
361 * VM_FAULT_ERROR).
362 */
363 /* These three entries are valid only while holding ptl lock */
364 pte_t *pte; /* Pointer to pte entry matching
365 * the 'address'. NULL if the page
366 * table hasn't been allocated.
367 */
368 spinlock_t *ptl; /* Page table lock.
369 * Protects pte page table if 'pte'
370 * is not NULL, otherwise pmd.
371 */
372 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
373 * vm_ops->map_pages() calls
374 * alloc_set_pte() from atomic context.
375 * do_fault_around() pre-allocates
376 * page table to avoid allocation from
377 * atomic context.
378 */
379};
380
381/* page entry size for vm->huge_fault() */
382enum page_entry_size {
383 PE_SIZE_PTE = 0,
384 PE_SIZE_PMD,
385 PE_SIZE_PUD,
386};
387
388/*
389 * These are the virtual MM functions - opening of an area, closing and
390 * unmapping it (needed to keep files on disk up-to-date etc), pointer
391 * to the functions called when a no-page or a wp-page exception occurs.
392 */
393struct vm_operations_struct {
394 void (*open)(struct vm_area_struct * area);
395 void (*close)(struct vm_area_struct * area);
396 int (*split)(struct vm_area_struct * area, unsigned long addr);
397 int (*mremap)(struct vm_area_struct * area);
398 vm_fault_t (*fault)(struct vm_fault *vmf);
399 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
400 enum page_entry_size pe_size);
401 void (*map_pages)(struct vm_fault *vmf,
402 pgoff_t start_pgoff, pgoff_t end_pgoff);
403 unsigned long (*pagesize)(struct vm_area_struct * area);
404
405 /* notification that a previously read-only page is about to become
406 * writable, if an error is returned it will cause a SIGBUS */
407 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
408
409 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
410 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
411
412 /* called by access_process_vm when get_user_pages() fails, typically
413 * for use by special VMAs that can switch between memory and hardware
414 */
415 int (*access)(struct vm_area_struct *vma, unsigned long addr,
416 void *buf, int len, int write);
417
418 /* Called by the /proc/PID/maps code to ask the vma whether it
419 * has a special name. Returning non-NULL will also cause this
420 * vma to be dumped unconditionally. */
421 const char *(*name)(struct vm_area_struct *vma);
422
423#ifdef CONFIG_NUMA
424 /*
425 * set_policy() op must add a reference to any non-NULL @new mempolicy
426 * to hold the policy upon return. Caller should pass NULL @new to
427 * remove a policy and fall back to surrounding context--i.e. do not
428 * install a MPOL_DEFAULT policy, nor the task or system default
429 * mempolicy.
430 */
431 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
432
433 /*
434 * get_policy() op must add reference [mpol_get()] to any policy at
435 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
436 * in mm/mempolicy.c will do this automatically.
437 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
438 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
439 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
440 * must return NULL--i.e., do not "fallback" to task or system default
441 * policy.
442 */
443 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
444 unsigned long addr);
445#endif
446 /*
447 * Called by vm_normal_page() for special PTEs to find the
448 * page for @addr. This is useful if the default behavior
449 * (using pte_page()) would not find the correct page.
450 */
451 struct page *(*find_special_page)(struct vm_area_struct *vma,
452 unsigned long addr);
453};
454
455static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
456{
457 static const struct vm_operations_struct dummy_vm_ops = {};
458
459 vma->vm_mm = mm;
460 vma->vm_ops = &dummy_vm_ops;
461 INIT_LIST_HEAD(&vma->anon_vma_chain);
462}
463
464static inline void vma_set_anonymous(struct vm_area_struct *vma)
465{
466 vma->vm_ops = NULL;
467}
468
469/* flush_tlb_range() takes a vma, not a mm, and can care about flags */
470#define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
471
472struct mmu_gather;
473struct inode;
474
475#define page_private(page) ((page)->private)
476#define set_page_private(page, v) ((page)->private = (v))
477
478#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
479static inline int pmd_devmap(pmd_t pmd)
480{
481 return 0;
482}
483static inline int pud_devmap(pud_t pud)
484{
485 return 0;
486}
487static inline int pgd_devmap(pgd_t pgd)
488{
489 return 0;
490}
491#endif
492
493/*
494 * FIXME: take this include out, include page-flags.h in
495 * files which need it (119 of them)
496 */
497#include <linux/page-flags.h>
498#include <linux/huge_mm.h>
499
500/*
501 * Methods to modify the page usage count.
502 *
503 * What counts for a page usage:
504 * - cache mapping (page->mapping)
505 * - private data (page->private)
506 * - page mapped in a task's page tables, each mapping
507 * is counted separately
508 *
509 * Also, many kernel routines increase the page count before a critical
510 * routine so they can be sure the page doesn't go away from under them.
511 */
512
513/*
514 * Drop a ref, return true if the refcount fell to zero (the page has no users)
515 */
516static inline int put_page_testzero(struct page *page)
517{
518 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
519 return page_ref_dec_and_test(page);
520}
521
522/*
523 * Try to grab a ref unless the page has a refcount of zero, return false if
524 * that is the case.
525 * This can be called when MMU is off so it must not access
526 * any of the virtual mappings.
527 */
528static inline int get_page_unless_zero(struct page *page)
529{
530 return page_ref_add_unless(page, 1, 0);
531}
532
533extern int page_is_ram(unsigned long pfn);
534
535enum {
536 REGION_INTERSECTS,
537 REGION_DISJOINT,
538 REGION_MIXED,
539};
540
541int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
542 unsigned long desc);
543
544/* Support for virtually mapped pages */
545struct page *vmalloc_to_page(const void *addr);
546unsigned long vmalloc_to_pfn(const void *addr);
547
548/*
549 * Determine if an address is within the vmalloc range
550 *
551 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
552 * is no special casing required.
553 */
554static inline bool is_vmalloc_addr(const void *x)
555{
556#ifdef CONFIG_MMU
557 unsigned long addr = (unsigned long)x;
558
559 return addr >= VMALLOC_START && addr < VMALLOC_END;
560#else
561 return false;
562#endif
563}
564#ifdef CONFIG_MMU
565extern int is_vmalloc_or_module_addr(const void *x);
566#else
567static inline int is_vmalloc_or_module_addr(const void *x)
568{
569 return 0;
570}
571#endif
572
573extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
574static inline void *kvmalloc(size_t size, gfp_t flags)
575{
576 return kvmalloc_node(size, flags, NUMA_NO_NODE);
577}
578static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
579{
580 return kvmalloc_node(size, flags | __GFP_ZERO, node);
581}
582static inline void *kvzalloc(size_t size, gfp_t flags)
583{
584 return kvmalloc(size, flags | __GFP_ZERO);
585}
586
587static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
588{
589 size_t bytes;
590
591 if (unlikely(check_mul_overflow(n, size, &bytes)))
592 return NULL;
593
594 return kvmalloc(bytes, flags);
595}
596
597static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
598{
599 return kvmalloc_array(n, size, flags | __GFP_ZERO);
600}
601
602extern void kvfree(const void *addr);
603
604static inline atomic_t *compound_mapcount_ptr(struct page *page)
605{
606 return &page[1].compound_mapcount;
607}
608
609static inline int compound_mapcount(struct page *page)
610{
611 VM_BUG_ON_PAGE(!PageCompound(page), page);
612 page = compound_head(page);
613 return atomic_read(compound_mapcount_ptr(page)) + 1;
614}
615
616/*
617 * The atomic page->_mapcount, starts from -1: so that transitions
618 * both from it and to it can be tracked, using atomic_inc_and_test
619 * and atomic_add_negative(-1).
620 */
621static inline void page_mapcount_reset(struct page *page)
622{
623 atomic_set(&(page)->_mapcount, -1);
624}
625
626int __page_mapcount(struct page *page);
627
628static inline int page_mapcount(struct page *page)
629{
630 VM_BUG_ON_PAGE(PageSlab(page), page);
631
632 if (unlikely(PageCompound(page)))
633 return __page_mapcount(page);
634 return atomic_read(&page->_mapcount) + 1;
635}
636
637#ifdef CONFIG_TRANSPARENT_HUGEPAGE
638int total_mapcount(struct page *page);
639int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
640#else
641static inline int total_mapcount(struct page *page)
642{
643 return page_mapcount(page);
644}
645static inline int page_trans_huge_mapcount(struct page *page,
646 int *total_mapcount)
647{
648 int mapcount = page_mapcount(page);
649 if (total_mapcount)
650 *total_mapcount = mapcount;
651 return mapcount;
652}
653#endif
654
655static inline struct page *virt_to_head_page(const void *x)
656{
657 struct page *page = virt_to_page(x);
658
659 return compound_head(page);
660}
661
662void __put_page(struct page *page);
663
664void put_pages_list(struct list_head *pages);
665
666void split_page(struct page *page, unsigned int order);
667
668/*
669 * Compound pages have a destructor function. Provide a
670 * prototype for that function and accessor functions.
671 * These are _only_ valid on the head of a compound page.
672 */
673typedef void compound_page_dtor(struct page *);
674
675/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
676enum compound_dtor_id {
677 NULL_COMPOUND_DTOR,
678 COMPOUND_PAGE_DTOR,
679#ifdef CONFIG_HUGETLB_PAGE
680 HUGETLB_PAGE_DTOR,
681#endif
682#ifdef CONFIG_TRANSPARENT_HUGEPAGE
683 TRANSHUGE_PAGE_DTOR,
684#endif
685 NR_COMPOUND_DTORS,
686};
687extern compound_page_dtor * const compound_page_dtors[];
688
689static inline void set_compound_page_dtor(struct page *page,
690 enum compound_dtor_id compound_dtor)
691{
692 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
693 page[1].compound_dtor = compound_dtor;
694}
695
696static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
697{
698 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
699 return compound_page_dtors[page[1].compound_dtor];
700}
701
702static inline unsigned int compound_order(struct page *page)
703{
704 if (!PageHead(page))
705 return 0;
706 return page[1].compound_order;
707}
708
709static inline void set_compound_order(struct page *page, unsigned int order)
710{
711 page[1].compound_order = order;
712}
713
714void free_compound_page(struct page *page);
715
716#ifdef CONFIG_MMU
717/*
718 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
719 * servicing faults for write access. In the normal case, do always want
720 * pte_mkwrite. But get_user_pages can cause write faults for mappings
721 * that do not have writing enabled, when used by access_process_vm.
722 */
723static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
724{
725 if (likely(vma->vm_flags & VM_WRITE))
726 pte = pte_mkwrite(pte);
727 return pte;
728}
729
730int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
731 struct page *page);
732int finish_fault(struct vm_fault *vmf);
733int finish_mkwrite_fault(struct vm_fault *vmf);
734#endif
735
736/*
737 * Multiple processes may "see" the same page. E.g. for untouched
738 * mappings of /dev/null, all processes see the same page full of
739 * zeroes, and text pages of executables and shared libraries have
740 * only one copy in memory, at most, normally.
741 *
742 * For the non-reserved pages, page_count(page) denotes a reference count.
743 * page_count() == 0 means the page is free. page->lru is then used for
744 * freelist management in the buddy allocator.
745 * page_count() > 0 means the page has been allocated.
746 *
747 * Pages are allocated by the slab allocator in order to provide memory
748 * to kmalloc and kmem_cache_alloc. In this case, the management of the
749 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
750 * unless a particular usage is carefully commented. (the responsibility of
751 * freeing the kmalloc memory is the caller's, of course).
752 *
753 * A page may be used by anyone else who does a __get_free_page().
754 * In this case, page_count still tracks the references, and should only
755 * be used through the normal accessor functions. The top bits of page->flags
756 * and page->virtual store page management information, but all other fields
757 * are unused and could be used privately, carefully. The management of this
758 * page is the responsibility of the one who allocated it, and those who have
759 * subsequently been given references to it.
760 *
761 * The other pages (we may call them "pagecache pages") are completely
762 * managed by the Linux memory manager: I/O, buffers, swapping etc.
763 * The following discussion applies only to them.
764 *
765 * A pagecache page contains an opaque `private' member, which belongs to the
766 * page's address_space. Usually, this is the address of a circular list of
767 * the page's disk buffers. PG_private must be set to tell the VM to call
768 * into the filesystem to release these pages.
769 *
770 * A page may belong to an inode's memory mapping. In this case, page->mapping
771 * is the pointer to the inode, and page->index is the file offset of the page,
772 * in units of PAGE_SIZE.
773 *
774 * If pagecache pages are not associated with an inode, they are said to be
775 * anonymous pages. These may become associated with the swapcache, and in that
776 * case PG_swapcache is set, and page->private is an offset into the swapcache.
777 *
778 * In either case (swapcache or inode backed), the pagecache itself holds one
779 * reference to the page. Setting PG_private should also increment the
780 * refcount. The each user mapping also has a reference to the page.
781 *
782 * The pagecache pages are stored in a per-mapping radix tree, which is
783 * rooted at mapping->i_pages, and indexed by offset.
784 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
785 * lists, we instead now tag pages as dirty/writeback in the radix tree.
786 *
787 * All pagecache pages may be subject to I/O:
788 * - inode pages may need to be read from disk,
789 * - inode pages which have been modified and are MAP_SHARED may need
790 * to be written back to the inode on disk,
791 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
792 * modified may need to be swapped out to swap space and (later) to be read
793 * back into memory.
794 */
795
796/*
797 * The zone field is never updated after free_area_init_core()
798 * sets it, so none of the operations on it need to be atomic.
799 */
800
801/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
802#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
803#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
804#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
805#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
806
807/*
808 * Define the bit shifts to access each section. For non-existent
809 * sections we define the shift as 0; that plus a 0 mask ensures
810 * the compiler will optimise away reference to them.
811 */
812#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
813#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
814#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
815#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
816
817/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
818#ifdef NODE_NOT_IN_PAGE_FLAGS
819#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
820#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
821 SECTIONS_PGOFF : ZONES_PGOFF)
822#else
823#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
824#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
825 NODES_PGOFF : ZONES_PGOFF)
826#endif
827
828#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
829
830#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
831#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
832#endif
833
834#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
835#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
836#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
837#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
838#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
839
840static inline enum zone_type page_zonenum(const struct page *page)
841{
842 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
843}
844
845#ifdef CONFIG_ZONE_DEVICE
846static inline bool is_zone_device_page(const struct page *page)
847{
848 return page_zonenum(page) == ZONE_DEVICE;
849}
850#else
851static inline bool is_zone_device_page(const struct page *page)
852{
853 return false;
854}
855#endif
856
857#ifdef CONFIG_DEV_PAGEMAP_OPS
858void dev_pagemap_get_ops(void);
859void dev_pagemap_put_ops(void);
860void __put_devmap_managed_page(struct page *page);
861DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
862static inline bool put_devmap_managed_page(struct page *page)
863{
864 if (!static_branch_unlikely(&devmap_managed_key))
865 return false;
866 if (!is_zone_device_page(page))
867 return false;
868 switch (page->pgmap->type) {
869 case MEMORY_DEVICE_PRIVATE:
870 case MEMORY_DEVICE_PUBLIC:
871 case MEMORY_DEVICE_FS_DAX:
872 __put_devmap_managed_page(page);
873 return true;
874 default:
875 break;
876 }
877 return false;
878}
879
880static inline bool is_device_private_page(const struct page *page)
881{
882 return is_zone_device_page(page) &&
883 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
884}
885
886static inline bool is_device_public_page(const struct page *page)
887{
888 return is_zone_device_page(page) &&
889 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
890}
891
892#else /* CONFIG_DEV_PAGEMAP_OPS */
893static inline void dev_pagemap_get_ops(void)
894{
895}
896
897static inline void dev_pagemap_put_ops(void)
898{
899}
900
901static inline bool put_devmap_managed_page(struct page *page)
902{
903 return false;
904}
905
906static inline bool is_device_private_page(const struct page *page)
907{
908 return false;
909}
910
911static inline bool is_device_public_page(const struct page *page)
912{
913 return false;
914}
915#endif /* CONFIG_DEV_PAGEMAP_OPS */
916
917static inline void get_page(struct page *page)
918{
919 page = compound_head(page);
920 /*
921 * Getting a normal page or the head of a compound page
922 * requires to already have an elevated page->_refcount.
923 */
924 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
925 page_ref_inc(page);
926}
927
928static inline void put_page(struct page *page)
929{
930 page = compound_head(page);
931
932 /*
933 * For devmap managed pages we need to catch refcount transition from
934 * 2 to 1, when refcount reach one it means the page is free and we
935 * need to inform the device driver through callback. See
936 * include/linux/memremap.h and HMM for details.
937 */
938 if (put_devmap_managed_page(page))
939 return;
940
941 if (put_page_testzero(page))
942 __put_page(page);
943}
944
945#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
946#define SECTION_IN_PAGE_FLAGS
947#endif
948
949/*
950 * The identification function is mainly used by the buddy allocator for
951 * determining if two pages could be buddies. We are not really identifying
952 * the zone since we could be using the section number id if we do not have
953 * node id available in page flags.
954 * We only guarantee that it will return the same value for two combinable
955 * pages in a zone.
956 */
957static inline int page_zone_id(struct page *page)
958{
959 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
960}
961
962static inline int zone_to_nid(struct zone *zone)
963{
964#ifdef CONFIG_NUMA
965 return zone->node;
966#else
967 return 0;
968#endif
969}
970
971#ifdef NODE_NOT_IN_PAGE_FLAGS
972extern int page_to_nid(const struct page *page);
973#else
974static inline int page_to_nid(const struct page *page)
975{
976 struct page *p = (struct page *)page;
977
978 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
979}
980#endif
981
982#ifdef CONFIG_NUMA_BALANCING
983static inline int cpu_pid_to_cpupid(int cpu, int pid)
984{
985 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
986}
987
988static inline int cpupid_to_pid(int cpupid)
989{
990 return cpupid & LAST__PID_MASK;
991}
992
993static inline int cpupid_to_cpu(int cpupid)
994{
995 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
996}
997
998static inline int cpupid_to_nid(int cpupid)
999{
1000 return cpu_to_node(cpupid_to_cpu(cpupid));
1001}
1002
1003static inline bool cpupid_pid_unset(int cpupid)
1004{
1005 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1006}
1007
1008static inline bool cpupid_cpu_unset(int cpupid)
1009{
1010 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1011}
1012
1013static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1014{
1015 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1016}
1017
1018#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1019#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1020static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1021{
1022 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1023}
1024
1025static inline int page_cpupid_last(struct page *page)
1026{
1027 return page->_last_cpupid;
1028}
1029static inline void page_cpupid_reset_last(struct page *page)
1030{
1031 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1032}
1033#else
1034static inline int page_cpupid_last(struct page *page)
1035{
1036 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1037}
1038
1039extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1040
1041static inline void page_cpupid_reset_last(struct page *page)
1042{
1043 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1044}
1045#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1046#else /* !CONFIG_NUMA_BALANCING */
1047static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1048{
1049 return page_to_nid(page); /* XXX */
1050}
1051
1052static inline int page_cpupid_last(struct page *page)
1053{
1054 return page_to_nid(page); /* XXX */
1055}
1056
1057static inline int cpupid_to_nid(int cpupid)
1058{
1059 return -1;
1060}
1061
1062static inline int cpupid_to_pid(int cpupid)
1063{
1064 return -1;
1065}
1066
1067static inline int cpupid_to_cpu(int cpupid)
1068{
1069 return -1;
1070}
1071
1072static inline int cpu_pid_to_cpupid(int nid, int pid)
1073{
1074 return -1;
1075}
1076
1077static inline bool cpupid_pid_unset(int cpupid)
1078{
1079 return 1;
1080}
1081
1082static inline void page_cpupid_reset_last(struct page *page)
1083{
1084}
1085
1086static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1087{
1088 return false;
1089}
1090#endif /* CONFIG_NUMA_BALANCING */
1091
1092static inline struct zone *page_zone(const struct page *page)
1093{
1094 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1095}
1096
1097static inline pg_data_t *page_pgdat(const struct page *page)
1098{
1099 return NODE_DATA(page_to_nid(page));
1100}
1101
1102#ifdef SECTION_IN_PAGE_FLAGS
1103static inline void set_page_section(struct page *page, unsigned long section)
1104{
1105 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1106 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1107}
1108
1109static inline unsigned long page_to_section(const struct page *page)
1110{
1111 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1112}
1113#endif
1114
1115static inline void set_page_zone(struct page *page, enum zone_type zone)
1116{
1117 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1118 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1119}
1120
1121static inline void set_page_node(struct page *page, unsigned long node)
1122{
1123 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1124 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1125}
1126
1127static inline void set_page_links(struct page *page, enum zone_type zone,
1128 unsigned long node, unsigned long pfn)
1129{
1130 set_page_zone(page, zone);
1131 set_page_node(page, node);
1132#ifdef SECTION_IN_PAGE_FLAGS
1133 set_page_section(page, pfn_to_section_nr(pfn));
1134#endif
1135}
1136
1137#ifdef CONFIG_MEMCG
1138static inline struct mem_cgroup *page_memcg(struct page *page)
1139{
1140 return page->mem_cgroup;
1141}
1142static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1143{
1144 WARN_ON_ONCE(!rcu_read_lock_held());
1145 return READ_ONCE(page->mem_cgroup);
1146}
1147#else
1148static inline struct mem_cgroup *page_memcg(struct page *page)
1149{
1150 return NULL;
1151}
1152static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1153{
1154 WARN_ON_ONCE(!rcu_read_lock_held());
1155 return NULL;
1156}
1157#endif
1158
1159/*
1160 * Some inline functions in vmstat.h depend on page_zone()
1161 */
1162#include <linux/vmstat.h>
1163
1164static __always_inline void *lowmem_page_address(const struct page *page)
1165{
1166 return page_to_virt(page);
1167}
1168
1169#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1170#define HASHED_PAGE_VIRTUAL
1171#endif
1172
1173#if defined(WANT_PAGE_VIRTUAL)
1174static inline void *page_address(const struct page *page)
1175{
1176 return page->virtual;
1177}
1178static inline void set_page_address(struct page *page, void *address)
1179{
1180 page->virtual = address;
1181}
1182#define page_address_init() do { } while(0)
1183#endif
1184
1185#if defined(HASHED_PAGE_VIRTUAL)
1186void *page_address(const struct page *page);
1187void set_page_address(struct page *page, void *virtual);
1188void page_address_init(void);
1189#endif
1190
1191#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1192#define page_address(page) lowmem_page_address(page)
1193#define set_page_address(page, address) do { } while(0)
1194#define page_address_init() do { } while(0)
1195#endif
1196
1197extern void *page_rmapping(struct page *page);
1198extern struct anon_vma *page_anon_vma(struct page *page);
1199extern struct address_space *page_mapping(struct page *page);
1200
1201extern struct address_space *__page_file_mapping(struct page *);
1202
1203static inline
1204struct address_space *page_file_mapping(struct page *page)
1205{
1206 if (unlikely(PageSwapCache(page)))
1207 return __page_file_mapping(page);
1208
1209 return page->mapping;
1210}
1211
1212extern pgoff_t __page_file_index(struct page *page);
1213
1214/*
1215 * Return the pagecache index of the passed page. Regular pagecache pages
1216 * use ->index whereas swapcache pages use swp_offset(->private)
1217 */
1218static inline pgoff_t page_index(struct page *page)
1219{
1220 if (unlikely(PageSwapCache(page)))
1221 return __page_file_index(page);
1222 return page->index;
1223}
1224
1225bool page_mapped(struct page *page);
1226struct address_space *page_mapping(struct page *page);
1227struct address_space *page_mapping_file(struct page *page);
1228
1229/*
1230 * Return true only if the page has been allocated with
1231 * ALLOC_NO_WATERMARKS and the low watermark was not
1232 * met implying that the system is under some pressure.
1233 */
1234static inline bool page_is_pfmemalloc(struct page *page)
1235{
1236 /*
1237 * Page index cannot be this large so this must be
1238 * a pfmemalloc page.
1239 */
1240 return page->index == -1UL;
1241}
1242
1243/*
1244 * Only to be called by the page allocator on a freshly allocated
1245 * page.
1246 */
1247static inline void set_page_pfmemalloc(struct page *page)
1248{
1249 page->index = -1UL;
1250}
1251
1252static inline void clear_page_pfmemalloc(struct page *page)
1253{
1254 page->index = 0;
1255}
1256
1257/*
1258 * Different kinds of faults, as returned by handle_mm_fault().
1259 * Used to decide whether a process gets delivered SIGBUS or
1260 * just gets major/minor fault counters bumped up.
1261 */
1262
1263#define VM_FAULT_OOM 0x0001
1264#define VM_FAULT_SIGBUS 0x0002
1265#define VM_FAULT_MAJOR 0x0004
1266#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1267#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1268#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1269#define VM_FAULT_SIGSEGV 0x0040
1270
1271#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1272#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1273#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1274#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1275#define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1276#define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1277 * and needs fsync() to complete (for
1278 * synchronous page faults in DAX) */
1279
1280#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1281 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1282 VM_FAULT_FALLBACK)
1283
1284#define VM_FAULT_RESULT_TRACE \
1285 { VM_FAULT_OOM, "OOM" }, \
1286 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1287 { VM_FAULT_MAJOR, "MAJOR" }, \
1288 { VM_FAULT_WRITE, "WRITE" }, \
1289 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1290 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1291 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1292 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1293 { VM_FAULT_LOCKED, "LOCKED" }, \
1294 { VM_FAULT_RETRY, "RETRY" }, \
1295 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1296 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1297 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1298
1299/* Encode hstate index for a hwpoisoned large page */
1300#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1301#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1302
1303/*
1304 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1305 */
1306extern void pagefault_out_of_memory(void);
1307
1308#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1309
1310/*
1311 * Flags passed to show_mem() and show_free_areas() to suppress output in
1312 * various contexts.
1313 */
1314#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1315
1316extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1317
1318extern bool can_do_mlock(void);
1319extern int user_shm_lock(size_t, struct user_struct *);
1320extern void user_shm_unlock(size_t, struct user_struct *);
1321
1322/*
1323 * Parameter block passed down to zap_pte_range in exceptional cases.
1324 */
1325struct zap_details {
1326 struct address_space *check_mapping; /* Check page->mapping if set */
1327 pgoff_t first_index; /* Lowest page->index to unmap */
1328 pgoff_t last_index; /* Highest page->index to unmap */
1329};
1330
1331struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1332 pte_t pte, bool with_public_device);
1333#define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1334
1335struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1336 pmd_t pmd);
1337
1338void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1339 unsigned long size);
1340void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1341 unsigned long size);
1342void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1343 unsigned long start, unsigned long end);
1344
1345/**
1346 * mm_walk - callbacks for walk_page_range
1347 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1348 * this handler should only handle pud_trans_huge() puds.
1349 * the pmd_entry or pte_entry callbacks will be used for
1350 * regular PUDs.
1351 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1352 * this handler is required to be able to handle
1353 * pmd_trans_huge() pmds. They may simply choose to
1354 * split_huge_page() instead of handling it explicitly.
1355 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1356 * @pte_hole: if set, called for each hole at all levels
1357 * @hugetlb_entry: if set, called for each hugetlb entry
1358 * @test_walk: caller specific callback function to determine whether
1359 * we walk over the current vma or not. Returning 0
1360 * value means "do page table walk over the current vma,"
1361 * and a negative one means "abort current page table walk
1362 * right now." 1 means "skip the current vma."
1363 * @mm: mm_struct representing the target process of page table walk
1364 * @vma: vma currently walked (NULL if walking outside vmas)
1365 * @private: private data for callbacks' usage
1366 *
1367 * (see the comment on walk_page_range() for more details)
1368 */
1369struct mm_walk {
1370 int (*pud_entry)(pud_t *pud, unsigned long addr,
1371 unsigned long next, struct mm_walk *walk);
1372 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1373 unsigned long next, struct mm_walk *walk);
1374 int (*pte_entry)(pte_t *pte, unsigned long addr,
1375 unsigned long next, struct mm_walk *walk);
1376 int (*pte_hole)(unsigned long addr, unsigned long next,
1377 struct mm_walk *walk);
1378 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1379 unsigned long addr, unsigned long next,
1380 struct mm_walk *walk);
1381 int (*test_walk)(unsigned long addr, unsigned long next,
1382 struct mm_walk *walk);
1383 struct mm_struct *mm;
1384 struct vm_area_struct *vma;
1385 void *private;
1386};
1387
1388int walk_page_range(unsigned long addr, unsigned long end,
1389 struct mm_walk *walk);
1390int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1391void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1392 unsigned long end, unsigned long floor, unsigned long ceiling);
1393int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1394 struct vm_area_struct *vma);
1395int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1396 unsigned long *start, unsigned long *end,
1397 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1398int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1399 unsigned long *pfn);
1400int follow_phys(struct vm_area_struct *vma, unsigned long address,
1401 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1402int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1403 void *buf, int len, int write);
1404
1405extern void truncate_pagecache(struct inode *inode, loff_t new);
1406extern void truncate_setsize(struct inode *inode, loff_t newsize);
1407void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1408void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1409int truncate_inode_page(struct address_space *mapping, struct page *page);
1410int generic_error_remove_page(struct address_space *mapping, struct page *page);
1411int invalidate_inode_page(struct page *page);
1412
1413#ifdef CONFIG_MMU
1414extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1415 unsigned int flags);
1416extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1417 unsigned long address, unsigned int fault_flags,
1418 bool *unlocked);
1419void unmap_mapping_pages(struct address_space *mapping,
1420 pgoff_t start, pgoff_t nr, bool even_cows);
1421void unmap_mapping_range(struct address_space *mapping,
1422 loff_t const holebegin, loff_t const holelen, int even_cows);
1423#else
1424static inline int handle_mm_fault(struct vm_area_struct *vma,
1425 unsigned long address, unsigned int flags)
1426{
1427 /* should never happen if there's no MMU */
1428 BUG();
1429 return VM_FAULT_SIGBUS;
1430}
1431static inline int fixup_user_fault(struct task_struct *tsk,
1432 struct mm_struct *mm, unsigned long address,
1433 unsigned int fault_flags, bool *unlocked)
1434{
1435 /* should never happen if there's no MMU */
1436 BUG();
1437 return -EFAULT;
1438}
1439static inline void unmap_mapping_pages(struct address_space *mapping,
1440 pgoff_t start, pgoff_t nr, bool even_cows) { }
1441static inline void unmap_mapping_range(struct address_space *mapping,
1442 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1443#endif
1444
1445static inline void unmap_shared_mapping_range(struct address_space *mapping,
1446 loff_t const holebegin, loff_t const holelen)
1447{
1448 unmap_mapping_range(mapping, holebegin, holelen, 0);
1449}
1450
1451extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1452 void *buf, int len, unsigned int gup_flags);
1453extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1454 void *buf, int len, unsigned int gup_flags);
1455extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1456 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1457
1458long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1459 unsigned long start, unsigned long nr_pages,
1460 unsigned int gup_flags, struct page **pages,
1461 struct vm_area_struct **vmas, int *locked);
1462long get_user_pages(unsigned long start, unsigned long nr_pages,
1463 unsigned int gup_flags, struct page **pages,
1464 struct vm_area_struct **vmas);
1465long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1466 unsigned int gup_flags, struct page **pages, int *locked);
1467long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1468 struct page **pages, unsigned int gup_flags);
1469#ifdef CONFIG_FS_DAX
1470long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1471 unsigned int gup_flags, struct page **pages,
1472 struct vm_area_struct **vmas);
1473#else
1474static inline long get_user_pages_longterm(unsigned long start,
1475 unsigned long nr_pages, unsigned int gup_flags,
1476 struct page **pages, struct vm_area_struct **vmas)
1477{
1478 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1479}
1480#endif /* CONFIG_FS_DAX */
1481
1482int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1483 struct page **pages);
1484
1485/* Container for pinned pfns / pages */
1486struct frame_vector {
1487 unsigned int nr_allocated; /* Number of frames we have space for */
1488 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1489 bool got_ref; /* Did we pin pages by getting page ref? */
1490 bool is_pfns; /* Does array contain pages or pfns? */
1491 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1492 * pfns_vector_pages() or pfns_vector_pfns()
1493 * for access */
1494};
1495
1496struct frame_vector *frame_vector_create(unsigned int nr_frames);
1497void frame_vector_destroy(struct frame_vector *vec);
1498int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1499 unsigned int gup_flags, struct frame_vector *vec);
1500void put_vaddr_frames(struct frame_vector *vec);
1501int frame_vector_to_pages(struct frame_vector *vec);
1502void frame_vector_to_pfns(struct frame_vector *vec);
1503
1504static inline unsigned int frame_vector_count(struct frame_vector *vec)
1505{
1506 return vec->nr_frames;
1507}
1508
1509static inline struct page **frame_vector_pages(struct frame_vector *vec)
1510{
1511 if (vec->is_pfns) {
1512 int err = frame_vector_to_pages(vec);
1513
1514 if (err)
1515 return ERR_PTR(err);
1516 }
1517 return (struct page **)(vec->ptrs);
1518}
1519
1520static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1521{
1522 if (!vec->is_pfns)
1523 frame_vector_to_pfns(vec);
1524 return (unsigned long *)(vec->ptrs);
1525}
1526
1527struct kvec;
1528int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1529 struct page **pages);
1530int get_kernel_page(unsigned long start, int write, struct page **pages);
1531struct page *get_dump_page(unsigned long addr);
1532
1533extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1534extern void do_invalidatepage(struct page *page, unsigned int offset,
1535 unsigned int length);
1536
1537void __set_page_dirty(struct page *, struct address_space *, int warn);
1538int __set_page_dirty_nobuffers(struct page *page);
1539int __set_page_dirty_no_writeback(struct page *page);
1540int redirty_page_for_writepage(struct writeback_control *wbc,
1541 struct page *page);
1542void account_page_dirtied(struct page *page, struct address_space *mapping);
1543void account_page_cleaned(struct page *page, struct address_space *mapping,
1544 struct bdi_writeback *wb);
1545int set_page_dirty(struct page *page);
1546int set_page_dirty_lock(struct page *page);
1547void __cancel_dirty_page(struct page *page);
1548static inline void cancel_dirty_page(struct page *page)
1549{
1550 /* Avoid atomic ops, locking, etc. when not actually needed. */
1551 if (PageDirty(page))
1552 __cancel_dirty_page(page);
1553}
1554int clear_page_dirty_for_io(struct page *page);
1555
1556int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1557
1558static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1559{
1560 return !vma->vm_ops;
1561}
1562
1563#ifdef CONFIG_SHMEM
1564/*
1565 * The vma_is_shmem is not inline because it is used only by slow
1566 * paths in userfault.
1567 */
1568bool vma_is_shmem(struct vm_area_struct *vma);
1569#else
1570static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1571#endif
1572
1573int vma_is_stack_for_current(struct vm_area_struct *vma);
1574
1575extern unsigned long move_page_tables(struct vm_area_struct *vma,
1576 unsigned long old_addr, struct vm_area_struct *new_vma,
1577 unsigned long new_addr, unsigned long len,
1578 bool need_rmap_locks);
1579extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1580 unsigned long end, pgprot_t newprot,
1581 int dirty_accountable, int prot_numa);
1582extern int mprotect_fixup(struct vm_area_struct *vma,
1583 struct vm_area_struct **pprev, unsigned long start,
1584 unsigned long end, unsigned long newflags);
1585
1586/*
1587 * doesn't attempt to fault and will return short.
1588 */
1589int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1590 struct page **pages);
1591/*
1592 * per-process(per-mm_struct) statistics.
1593 */
1594static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1595{
1596 long val = atomic_long_read(&mm->rss_stat.count[member]);
1597
1598#ifdef SPLIT_RSS_COUNTING
1599 /*
1600 * counter is updated in asynchronous manner and may go to minus.
1601 * But it's never be expected number for users.
1602 */
1603 if (val < 0)
1604 val = 0;
1605#endif
1606 return (unsigned long)val;
1607}
1608
1609static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1610{
1611 atomic_long_add(value, &mm->rss_stat.count[member]);
1612}
1613
1614static inline void inc_mm_counter(struct mm_struct *mm, int member)
1615{
1616 atomic_long_inc(&mm->rss_stat.count[member]);
1617}
1618
1619static inline void dec_mm_counter(struct mm_struct *mm, int member)
1620{
1621 atomic_long_dec(&mm->rss_stat.count[member]);
1622}
1623
1624/* Optimized variant when page is already known not to be PageAnon */
1625static inline int mm_counter_file(struct page *page)
1626{
1627 if (PageSwapBacked(page))
1628 return MM_SHMEMPAGES;
1629 return MM_FILEPAGES;
1630}
1631
1632static inline int mm_counter(struct page *page)
1633{
1634 if (PageAnon(page))
1635 return MM_ANONPAGES;
1636 return mm_counter_file(page);
1637}
1638
1639static inline unsigned long get_mm_rss(struct mm_struct *mm)
1640{
1641 return get_mm_counter(mm, MM_FILEPAGES) +
1642 get_mm_counter(mm, MM_ANONPAGES) +
1643 get_mm_counter(mm, MM_SHMEMPAGES);
1644}
1645
1646static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1647{
1648 return max(mm->hiwater_rss, get_mm_rss(mm));
1649}
1650
1651static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1652{
1653 return max(mm->hiwater_vm, mm->total_vm);
1654}
1655
1656static inline void update_hiwater_rss(struct mm_struct *mm)
1657{
1658 unsigned long _rss = get_mm_rss(mm);
1659
1660 if ((mm)->hiwater_rss < _rss)
1661 (mm)->hiwater_rss = _rss;
1662}
1663
1664static inline void update_hiwater_vm(struct mm_struct *mm)
1665{
1666 if (mm->hiwater_vm < mm->total_vm)
1667 mm->hiwater_vm = mm->total_vm;
1668}
1669
1670static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1671{
1672 mm->hiwater_rss = get_mm_rss(mm);
1673}
1674
1675static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1676 struct mm_struct *mm)
1677{
1678 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1679
1680 if (*maxrss < hiwater_rss)
1681 *maxrss = hiwater_rss;
1682}
1683
1684#if defined(SPLIT_RSS_COUNTING)
1685void sync_mm_rss(struct mm_struct *mm);
1686#else
1687static inline void sync_mm_rss(struct mm_struct *mm)
1688{
1689}
1690#endif
1691
1692#ifndef __HAVE_ARCH_PTE_DEVMAP
1693static inline int pte_devmap(pte_t pte)
1694{
1695 return 0;
1696}
1697#endif
1698
1699int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1700
1701extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1702 spinlock_t **ptl);
1703static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1704 spinlock_t **ptl)
1705{
1706 pte_t *ptep;
1707 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1708 return ptep;
1709}
1710
1711#ifdef __PAGETABLE_P4D_FOLDED
1712static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1713 unsigned long address)
1714{
1715 return 0;
1716}
1717#else
1718int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1719#endif
1720
1721#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1722static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1723 unsigned long address)
1724{
1725 return 0;
1726}
1727static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1728static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1729
1730#else
1731int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1732
1733static inline void mm_inc_nr_puds(struct mm_struct *mm)
1734{
1735 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1736}
1737
1738static inline void mm_dec_nr_puds(struct mm_struct *mm)
1739{
1740 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1741}
1742#endif
1743
1744#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1745static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1746 unsigned long address)
1747{
1748 return 0;
1749}
1750
1751static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1752static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1753
1754#else
1755int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1756
1757static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1758{
1759 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1760}
1761
1762static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1763{
1764 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1765}
1766#endif
1767
1768#ifdef CONFIG_MMU
1769static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1770{
1771 atomic_long_set(&mm->pgtables_bytes, 0);
1772}
1773
1774static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1775{
1776 return atomic_long_read(&mm->pgtables_bytes);
1777}
1778
1779static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1780{
1781 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1782}
1783
1784static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1785{
1786 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1787}
1788#else
1789
1790static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1791static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1792{
1793 return 0;
1794}
1795
1796static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1797static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1798#endif
1799
1800int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1801int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1802
1803/*
1804 * The following ifdef needed to get the 4level-fixup.h header to work.
1805 * Remove it when 4level-fixup.h has been removed.
1806 */
1807#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1808
1809#ifndef __ARCH_HAS_5LEVEL_HACK
1810static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1811 unsigned long address)
1812{
1813 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1814 NULL : p4d_offset(pgd, address);
1815}
1816
1817static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1818 unsigned long address)
1819{
1820 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1821 NULL : pud_offset(p4d, address);
1822}
1823#endif /* !__ARCH_HAS_5LEVEL_HACK */
1824
1825static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1826{
1827 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1828 NULL: pmd_offset(pud, address);
1829}
1830#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1831
1832#if USE_SPLIT_PTE_PTLOCKS
1833#if ALLOC_SPLIT_PTLOCKS
1834void __init ptlock_cache_init(void);
1835extern bool ptlock_alloc(struct page *page);
1836extern void ptlock_free(struct page *page);
1837
1838static inline spinlock_t *ptlock_ptr(struct page *page)
1839{
1840 return page->ptl;
1841}
1842#else /* ALLOC_SPLIT_PTLOCKS */
1843static inline void ptlock_cache_init(void)
1844{
1845}
1846
1847static inline bool ptlock_alloc(struct page *page)
1848{
1849 return true;
1850}
1851
1852static inline void ptlock_free(struct page *page)
1853{
1854}
1855
1856static inline spinlock_t *ptlock_ptr(struct page *page)
1857{
1858 return &page->ptl;
1859}
1860#endif /* ALLOC_SPLIT_PTLOCKS */
1861
1862static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1863{
1864 return ptlock_ptr(pmd_page(*pmd));
1865}
1866
1867static inline bool ptlock_init(struct page *page)
1868{
1869 /*
1870 * prep_new_page() initialize page->private (and therefore page->ptl)
1871 * with 0. Make sure nobody took it in use in between.
1872 *
1873 * It can happen if arch try to use slab for page table allocation:
1874 * slab code uses page->slab_cache, which share storage with page->ptl.
1875 */
1876 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1877 if (!ptlock_alloc(page))
1878 return false;
1879 spin_lock_init(ptlock_ptr(page));
1880 return true;
1881}
1882
1883/* Reset page->mapping so free_pages_check won't complain. */
1884static inline void pte_lock_deinit(struct page *page)
1885{
1886 page->mapping = NULL;
1887 ptlock_free(page);
1888}
1889
1890#else /* !USE_SPLIT_PTE_PTLOCKS */
1891/*
1892 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1893 */
1894static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1895{
1896 return &mm->page_table_lock;
1897}
1898static inline void ptlock_cache_init(void) {}
1899static inline bool ptlock_init(struct page *page) { return true; }
1900static inline void pte_lock_deinit(struct page *page) {}
1901#endif /* USE_SPLIT_PTE_PTLOCKS */
1902
1903static inline void pgtable_init(void)
1904{
1905 ptlock_cache_init();
1906 pgtable_cache_init();
1907}
1908
1909static inline bool pgtable_page_ctor(struct page *page)
1910{
1911 if (!ptlock_init(page))
1912 return false;
1913 __SetPageTable(page);
1914 inc_zone_page_state(page, NR_PAGETABLE);
1915 return true;
1916}
1917
1918static inline void pgtable_page_dtor(struct page *page)
1919{
1920 pte_lock_deinit(page);
1921 __ClearPageTable(page);
1922 dec_zone_page_state(page, NR_PAGETABLE);
1923}
1924
1925#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1926({ \
1927 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1928 pte_t *__pte = pte_offset_map(pmd, address); \
1929 *(ptlp) = __ptl; \
1930 spin_lock(__ptl); \
1931 __pte; \
1932})
1933
1934#define pte_unmap_unlock(pte, ptl) do { \
1935 spin_unlock(ptl); \
1936 pte_unmap(pte); \
1937} while (0)
1938
1939#define pte_alloc(mm, pmd, address) \
1940 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1941
1942#define pte_alloc_map(mm, pmd, address) \
1943 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1944
1945#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1946 (pte_alloc(mm, pmd, address) ? \
1947 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1948
1949#define pte_alloc_kernel(pmd, address) \
1950 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1951 NULL: pte_offset_kernel(pmd, address))
1952
1953#if USE_SPLIT_PMD_PTLOCKS
1954
1955static struct page *pmd_to_page(pmd_t *pmd)
1956{
1957 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1958 return virt_to_page((void *)((unsigned long) pmd & mask));
1959}
1960
1961static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1962{
1963 return ptlock_ptr(pmd_to_page(pmd));
1964}
1965
1966static inline bool pgtable_pmd_page_ctor(struct page *page)
1967{
1968#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1969 page->pmd_huge_pte = NULL;
1970#endif
1971 return ptlock_init(page);
1972}
1973
1974static inline void pgtable_pmd_page_dtor(struct page *page)
1975{
1976#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1977 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1978#endif
1979 ptlock_free(page);
1980}
1981
1982#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1983
1984#else
1985
1986static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1987{
1988 return &mm->page_table_lock;
1989}
1990
1991static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1992static inline void pgtable_pmd_page_dtor(struct page *page) {}
1993
1994#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1995
1996#endif
1997
1998static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1999{
2000 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2001 spin_lock(ptl);
2002 return ptl;
2003}
2004
2005/*
2006 * No scalability reason to split PUD locks yet, but follow the same pattern
2007 * as the PMD locks to make it easier if we decide to. The VM should not be
2008 * considered ready to switch to split PUD locks yet; there may be places
2009 * which need to be converted from page_table_lock.
2010 */
2011static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2012{
2013 return &mm->page_table_lock;
2014}
2015
2016static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2017{
2018 spinlock_t *ptl = pud_lockptr(mm, pud);
2019
2020 spin_lock(ptl);
2021 return ptl;
2022}
2023
2024extern void __init pagecache_init(void);
2025extern void free_area_init(unsigned long * zones_size);
2026extern void free_area_init_node(int nid, unsigned long * zones_size,
2027 unsigned long zone_start_pfn, unsigned long *zholes_size);
2028extern void free_initmem(void);
2029
2030/*
2031 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2032 * into the buddy system. The freed pages will be poisoned with pattern
2033 * "poison" if it's within range [0, UCHAR_MAX].
2034 * Return pages freed into the buddy system.
2035 */
2036extern unsigned long free_reserved_area(void *start, void *end,
2037 int poison, char *s);
2038
2039#ifdef CONFIG_HIGHMEM
2040/*
2041 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2042 * and totalram_pages.
2043 */
2044extern void free_highmem_page(struct page *page);
2045#endif
2046
2047extern void adjust_managed_page_count(struct page *page, long count);
2048extern void mem_init_print_info(const char *str);
2049
2050extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2051
2052/* Free the reserved page into the buddy system, so it gets managed. */
2053static inline void __free_reserved_page(struct page *page)
2054{
2055 ClearPageReserved(page);
2056 init_page_count(page);
2057 __free_page(page);
2058}
2059
2060static inline void free_reserved_page(struct page *page)
2061{
2062 __free_reserved_page(page);
2063 adjust_managed_page_count(page, 1);
2064}
2065
2066static inline void mark_page_reserved(struct page *page)
2067{
2068 SetPageReserved(page);
2069 adjust_managed_page_count(page, -1);
2070}
2071
2072/*
2073 * Default method to free all the __init memory into the buddy system.
2074 * The freed pages will be poisoned with pattern "poison" if it's within
2075 * range [0, UCHAR_MAX].
2076 * Return pages freed into the buddy system.
2077 */
2078static inline unsigned long free_initmem_default(int poison)
2079{
2080 extern char __init_begin[], __init_end[];
2081
2082 return free_reserved_area(&__init_begin, &__init_end,
2083 poison, "unused kernel");
2084}
2085
2086static inline unsigned long get_num_physpages(void)
2087{
2088 int nid;
2089 unsigned long phys_pages = 0;
2090
2091 for_each_online_node(nid)
2092 phys_pages += node_present_pages(nid);
2093
2094 return phys_pages;
2095}
2096
2097#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2098/*
2099 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2100 * zones, allocate the backing mem_map and account for memory holes in a more
2101 * architecture independent manner. This is a substitute for creating the
2102 * zone_sizes[] and zholes_size[] arrays and passing them to
2103 * free_area_init_node()
2104 *
2105 * An architecture is expected to register range of page frames backed by
2106 * physical memory with memblock_add[_node]() before calling
2107 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2108 * usage, an architecture is expected to do something like
2109 *
2110 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2111 * max_highmem_pfn};
2112 * for_each_valid_physical_page_range()
2113 * memblock_add_node(base, size, nid)
2114 * free_area_init_nodes(max_zone_pfns);
2115 *
2116 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2117 * registered physical page range. Similarly
2118 * sparse_memory_present_with_active_regions() calls memory_present() for
2119 * each range when SPARSEMEM is enabled.
2120 *
2121 * See mm/page_alloc.c for more information on each function exposed by
2122 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2123 */
2124extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2125unsigned long node_map_pfn_alignment(void);
2126unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2127 unsigned long end_pfn);
2128extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2129 unsigned long end_pfn);
2130extern void get_pfn_range_for_nid(unsigned int nid,
2131 unsigned long *start_pfn, unsigned long *end_pfn);
2132extern unsigned long find_min_pfn_with_active_regions(void);
2133extern void free_bootmem_with_active_regions(int nid,
2134 unsigned long max_low_pfn);
2135extern void sparse_memory_present_with_active_regions(int nid);
2136
2137#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2138
2139#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2140 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2141static inline int __early_pfn_to_nid(unsigned long pfn,
2142 struct mminit_pfnnid_cache *state)
2143{
2144 return 0;
2145}
2146#else
2147/* please see mm/page_alloc.c */
2148extern int __meminit early_pfn_to_nid(unsigned long pfn);
2149/* there is a per-arch backend function. */
2150extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2151 struct mminit_pfnnid_cache *state);
2152#endif
2153
2154#if defined(CONFIG_HAVE_MEMBLOCK) && !defined(CONFIG_FLAT_NODE_MEM_MAP)
2155void zero_resv_unavail(void);
2156#else
2157static inline void zero_resv_unavail(void) {}
2158#endif
2159
2160extern void set_dma_reserve(unsigned long new_dma_reserve);
2161extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2162 enum memmap_context, struct vmem_altmap *);
2163extern void setup_per_zone_wmarks(void);
2164extern int __meminit init_per_zone_wmark_min(void);
2165extern void mem_init(void);
2166extern void __init mmap_init(void);
2167extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2168extern long si_mem_available(void);
2169extern void si_meminfo(struct sysinfo * val);
2170extern void si_meminfo_node(struct sysinfo *val, int nid);
2171#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2172extern unsigned long arch_reserved_kernel_pages(void);
2173#endif
2174
2175extern __printf(3, 4)
2176void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2177
2178extern void setup_per_cpu_pageset(void);
2179
2180extern void zone_pcp_update(struct zone *zone);
2181extern void zone_pcp_reset(struct zone *zone);
2182
2183/* page_alloc.c */
2184extern int min_free_kbytes;
2185extern int watermark_scale_factor;
2186
2187/* nommu.c */
2188extern atomic_long_t mmap_pages_allocated;
2189extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2190
2191/* interval_tree.c */
2192void vma_interval_tree_insert(struct vm_area_struct *node,
2193 struct rb_root_cached *root);
2194void vma_interval_tree_insert_after(struct vm_area_struct *node,
2195 struct vm_area_struct *prev,
2196 struct rb_root_cached *root);
2197void vma_interval_tree_remove(struct vm_area_struct *node,
2198 struct rb_root_cached *root);
2199struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2200 unsigned long start, unsigned long last);
2201struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2202 unsigned long start, unsigned long last);
2203
2204#define vma_interval_tree_foreach(vma, root, start, last) \
2205 for (vma = vma_interval_tree_iter_first(root, start, last); \
2206 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2207
2208void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2209 struct rb_root_cached *root);
2210void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2211 struct rb_root_cached *root);
2212struct anon_vma_chain *
2213anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2214 unsigned long start, unsigned long last);
2215struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2216 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2217#ifdef CONFIG_DEBUG_VM_RB
2218void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2219#endif
2220
2221#define anon_vma_interval_tree_foreach(avc, root, start, last) \
2222 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2223 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2224
2225/* mmap.c */
2226extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2227extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2228 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2229 struct vm_area_struct *expand);
2230static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2231 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2232{
2233 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2234}
2235extern struct vm_area_struct *vma_merge(struct mm_struct *,
2236 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2237 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2238 struct mempolicy *, struct vm_userfaultfd_ctx);
2239extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2240extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2241 unsigned long addr, int new_below);
2242extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2243 unsigned long addr, int new_below);
2244extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2245extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2246 struct rb_node **, struct rb_node *);
2247extern void unlink_file_vma(struct vm_area_struct *);
2248extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2249 unsigned long addr, unsigned long len, pgoff_t pgoff,
2250 bool *need_rmap_locks);
2251extern void exit_mmap(struct mm_struct *);
2252
2253static inline int check_data_rlimit(unsigned long rlim,
2254 unsigned long new,
2255 unsigned long start,
2256 unsigned long end_data,
2257 unsigned long start_data)
2258{
2259 if (rlim < RLIM_INFINITY) {
2260 if (((new - start) + (end_data - start_data)) > rlim)
2261 return -ENOSPC;
2262 }
2263
2264 return 0;
2265}
2266
2267extern int mm_take_all_locks(struct mm_struct *mm);
2268extern void mm_drop_all_locks(struct mm_struct *mm);
2269
2270extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2271extern struct file *get_mm_exe_file(struct mm_struct *mm);
2272extern struct file *get_task_exe_file(struct task_struct *task);
2273
2274extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2275extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2276
2277extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2278 const struct vm_special_mapping *sm);
2279extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2280 unsigned long addr, unsigned long len,
2281 unsigned long flags,
2282 const struct vm_special_mapping *spec);
2283/* This is an obsolete alternative to _install_special_mapping. */
2284extern int install_special_mapping(struct mm_struct *mm,
2285 unsigned long addr, unsigned long len,
2286 unsigned long flags, struct page **pages);
2287
2288extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2289
2290extern unsigned long mmap_region(struct file *file, unsigned long addr,
2291 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2292 struct list_head *uf);
2293extern unsigned long do_mmap(struct file *file, unsigned long addr,
2294 unsigned long len, unsigned long prot, unsigned long flags,
2295 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2296 struct list_head *uf);
2297extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2298 struct list_head *uf);
2299
2300static inline unsigned long
2301do_mmap_pgoff(struct file *file, unsigned long addr,
2302 unsigned long len, unsigned long prot, unsigned long flags,
2303 unsigned long pgoff, unsigned long *populate,
2304 struct list_head *uf)
2305{
2306 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2307}
2308
2309#ifdef CONFIG_MMU
2310extern int __mm_populate(unsigned long addr, unsigned long len,
2311 int ignore_errors);
2312static inline void mm_populate(unsigned long addr, unsigned long len)
2313{
2314 /* Ignore errors */
2315 (void) __mm_populate(addr, len, 1);
2316}
2317#else
2318static inline void mm_populate(unsigned long addr, unsigned long len) {}
2319#endif
2320
2321/* These take the mm semaphore themselves */
2322extern int __must_check vm_brk(unsigned long, unsigned long);
2323extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2324extern int vm_munmap(unsigned long, size_t);
2325extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2326 unsigned long, unsigned long,
2327 unsigned long, unsigned long);
2328
2329struct vm_unmapped_area_info {
2330#define VM_UNMAPPED_AREA_TOPDOWN 1
2331 unsigned long flags;
2332 unsigned long length;
2333 unsigned long low_limit;
2334 unsigned long high_limit;
2335 unsigned long align_mask;
2336 unsigned long align_offset;
2337};
2338
2339extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2340extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2341
2342/*
2343 * Search for an unmapped address range.
2344 *
2345 * We are looking for a range that:
2346 * - does not intersect with any VMA;
2347 * - is contained within the [low_limit, high_limit) interval;
2348 * - is at least the desired size.
2349 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2350 */
2351static inline unsigned long
2352vm_unmapped_area(struct vm_unmapped_area_info *info)
2353{
2354 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2355 return unmapped_area_topdown(info);
2356 else
2357 return unmapped_area(info);
2358}
2359
2360/* truncate.c */
2361extern void truncate_inode_pages(struct address_space *, loff_t);
2362extern void truncate_inode_pages_range(struct address_space *,
2363 loff_t lstart, loff_t lend);
2364extern void truncate_inode_pages_final(struct address_space *);
2365
2366/* generic vm_area_ops exported for stackable file systems */
2367extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2368extern void filemap_map_pages(struct vm_fault *vmf,
2369 pgoff_t start_pgoff, pgoff_t end_pgoff);
2370extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2371
2372/* mm/page-writeback.c */
2373int __must_check write_one_page(struct page *page);
2374void task_dirty_inc(struct task_struct *tsk);
2375
2376/* readahead.c */
2377#define VM_MAX_READAHEAD 128 /* kbytes */
2378#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2379
2380int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2381 pgoff_t offset, unsigned long nr_to_read);
2382
2383void page_cache_sync_readahead(struct address_space *mapping,
2384 struct file_ra_state *ra,
2385 struct file *filp,
2386 pgoff_t offset,
2387 unsigned long size);
2388
2389void page_cache_async_readahead(struct address_space *mapping,
2390 struct file_ra_state *ra,
2391 struct file *filp,
2392 struct page *pg,
2393 pgoff_t offset,
2394 unsigned long size);
2395
2396extern unsigned long stack_guard_gap;
2397/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2398extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2399
2400/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2401extern int expand_downwards(struct vm_area_struct *vma,
2402 unsigned long address);
2403#if VM_GROWSUP
2404extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2405#else
2406 #define expand_upwards(vma, address) (0)
2407#endif
2408
2409/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2410extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2411extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2412 struct vm_area_struct **pprev);
2413
2414/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2415 NULL if none. Assume start_addr < end_addr. */
2416static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2417{
2418 struct vm_area_struct * vma = find_vma(mm,start_addr);
2419
2420 if (vma && end_addr <= vma->vm_start)
2421 vma = NULL;
2422 return vma;
2423}
2424
2425static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2426{
2427 unsigned long vm_start = vma->vm_start;
2428
2429 if (vma->vm_flags & VM_GROWSDOWN) {
2430 vm_start -= stack_guard_gap;
2431 if (vm_start > vma->vm_start)
2432 vm_start = 0;
2433 }
2434 return vm_start;
2435}
2436
2437static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2438{
2439 unsigned long vm_end = vma->vm_end;
2440
2441 if (vma->vm_flags & VM_GROWSUP) {
2442 vm_end += stack_guard_gap;
2443 if (vm_end < vma->vm_end)
2444 vm_end = -PAGE_SIZE;
2445 }
2446 return vm_end;
2447}
2448
2449static inline unsigned long vma_pages(struct vm_area_struct *vma)
2450{
2451 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2452}
2453
2454/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2455static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2456 unsigned long vm_start, unsigned long vm_end)
2457{
2458 struct vm_area_struct *vma = find_vma(mm, vm_start);
2459
2460 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2461 vma = NULL;
2462
2463 return vma;
2464}
2465
2466#ifdef CONFIG_MMU
2467pgprot_t vm_get_page_prot(unsigned long vm_flags);
2468void vma_set_page_prot(struct vm_area_struct *vma);
2469#else
2470static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2471{
2472 return __pgprot(0);
2473}
2474static inline void vma_set_page_prot(struct vm_area_struct *vma)
2475{
2476 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2477}
2478#endif
2479
2480#ifdef CONFIG_NUMA_BALANCING
2481unsigned long change_prot_numa(struct vm_area_struct *vma,
2482 unsigned long start, unsigned long end);
2483#endif
2484
2485struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2486int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2487 unsigned long pfn, unsigned long size, pgprot_t);
2488int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2489int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2490 unsigned long pfn);
2491int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2492 unsigned long pfn, pgprot_t pgprot);
2493int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2494 pfn_t pfn);
2495vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2496 unsigned long addr, pfn_t pfn);
2497int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2498
2499static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2500 unsigned long addr, struct page *page)
2501{
2502 int err = vm_insert_page(vma, addr, page);
2503
2504 if (err == -ENOMEM)
2505 return VM_FAULT_OOM;
2506 if (err < 0 && err != -EBUSY)
2507 return VM_FAULT_SIGBUS;
2508
2509 return VM_FAULT_NOPAGE;
2510}
2511
2512static inline vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma,
2513 unsigned long addr, pfn_t pfn)
2514{
2515 int err = vm_insert_mixed(vma, addr, pfn);
2516
2517 if (err == -ENOMEM)
2518 return VM_FAULT_OOM;
2519 if (err < 0 && err != -EBUSY)
2520 return VM_FAULT_SIGBUS;
2521
2522 return VM_FAULT_NOPAGE;
2523}
2524
2525static inline vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma,
2526 unsigned long addr, unsigned long pfn)
2527{
2528 int err = vm_insert_pfn(vma, addr, pfn);
2529
2530 if (err == -ENOMEM)
2531 return VM_FAULT_OOM;
2532 if (err < 0 && err != -EBUSY)
2533 return VM_FAULT_SIGBUS;
2534
2535 return VM_FAULT_NOPAGE;
2536}
2537
2538static inline vm_fault_t vmf_error(int err)
2539{
2540 if (err == -ENOMEM)
2541 return VM_FAULT_OOM;
2542 return VM_FAULT_SIGBUS;
2543}
2544
2545struct page *follow_page_mask(struct vm_area_struct *vma,
2546 unsigned long address, unsigned int foll_flags,
2547 unsigned int *page_mask);
2548
2549static inline struct page *follow_page(struct vm_area_struct *vma,
2550 unsigned long address, unsigned int foll_flags)
2551{
2552 unsigned int unused_page_mask;
2553 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2554}
2555
2556#define FOLL_WRITE 0x01 /* check pte is writable */
2557#define FOLL_TOUCH 0x02 /* mark page accessed */
2558#define FOLL_GET 0x04 /* do get_page on page */
2559#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2560#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2561#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2562 * and return without waiting upon it */
2563#define FOLL_POPULATE 0x40 /* fault in page */
2564#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2565#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2566#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2567#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2568#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2569#define FOLL_MLOCK 0x1000 /* lock present pages */
2570#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2571#define FOLL_COW 0x4000 /* internal GUP flag */
2572#define FOLL_ANON 0x8000 /* don't do file mappings */
2573
2574static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2575{
2576 if (vm_fault & VM_FAULT_OOM)
2577 return -ENOMEM;
2578 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2579 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2580 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2581 return -EFAULT;
2582 return 0;
2583}
2584
2585typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2586 void *data);
2587extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2588 unsigned long size, pte_fn_t fn, void *data);
2589
2590
2591#ifdef CONFIG_PAGE_POISONING
2592extern bool page_poisoning_enabled(void);
2593extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2594#else
2595static inline bool page_poisoning_enabled(void) { return false; }
2596static inline void kernel_poison_pages(struct page *page, int numpages,
2597 int enable) { }
2598#endif
2599
2600#ifdef CONFIG_DEBUG_PAGEALLOC
2601extern bool _debug_pagealloc_enabled;
2602extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2603
2604static inline bool debug_pagealloc_enabled(void)
2605{
2606 return _debug_pagealloc_enabled;
2607}
2608
2609static inline void
2610kernel_map_pages(struct page *page, int numpages, int enable)
2611{
2612 if (!debug_pagealloc_enabled())
2613 return;
2614
2615 __kernel_map_pages(page, numpages, enable);
2616}
2617#ifdef CONFIG_HIBERNATION
2618extern bool kernel_page_present(struct page *page);
2619#endif /* CONFIG_HIBERNATION */
2620#else /* CONFIG_DEBUG_PAGEALLOC */
2621static inline void
2622kernel_map_pages(struct page *page, int numpages, int enable) {}
2623#ifdef CONFIG_HIBERNATION
2624static inline bool kernel_page_present(struct page *page) { return true; }
2625#endif /* CONFIG_HIBERNATION */
2626static inline bool debug_pagealloc_enabled(void)
2627{
2628 return false;
2629}
2630#endif /* CONFIG_DEBUG_PAGEALLOC */
2631
2632#ifdef __HAVE_ARCH_GATE_AREA
2633extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2634extern int in_gate_area_no_mm(unsigned long addr);
2635extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2636#else
2637static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2638{
2639 return NULL;
2640}
2641static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2642static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2643{
2644 return 0;
2645}
2646#endif /* __HAVE_ARCH_GATE_AREA */
2647
2648extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2649
2650#ifdef CONFIG_SYSCTL
2651extern int sysctl_drop_caches;
2652int drop_caches_sysctl_handler(struct ctl_table *, int,
2653 void __user *, size_t *, loff_t *);
2654#endif
2655
2656void drop_slab(void);
2657void drop_slab_node(int nid);
2658
2659#ifndef CONFIG_MMU
2660#define randomize_va_space 0
2661#else
2662extern int randomize_va_space;
2663#endif
2664
2665const char * arch_vma_name(struct vm_area_struct *vma);
2666void print_vma_addr(char *prefix, unsigned long rip);
2667
2668void *sparse_buffer_alloc(unsigned long size);
2669struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2670 struct vmem_altmap *altmap);
2671pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2672p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2673pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2674pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2675pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2676void *vmemmap_alloc_block(unsigned long size, int node);
2677struct vmem_altmap;
2678void *vmemmap_alloc_block_buf(unsigned long size, int node);
2679void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2680void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2681int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2682 int node);
2683int vmemmap_populate(unsigned long start, unsigned long end, int node,
2684 struct vmem_altmap *altmap);
2685void vmemmap_populate_print_last(void);
2686#ifdef CONFIG_MEMORY_HOTPLUG
2687void vmemmap_free(unsigned long start, unsigned long end,
2688 struct vmem_altmap *altmap);
2689#endif
2690void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2691 unsigned long nr_pages);
2692
2693enum mf_flags {
2694 MF_COUNT_INCREASED = 1 << 0,
2695 MF_ACTION_REQUIRED = 1 << 1,
2696 MF_MUST_KILL = 1 << 2,
2697 MF_SOFT_OFFLINE = 1 << 3,
2698};
2699extern int memory_failure(unsigned long pfn, int flags);
2700extern void memory_failure_queue(unsigned long pfn, int flags);
2701extern int unpoison_memory(unsigned long pfn);
2702extern int get_hwpoison_page(struct page *page);
2703#define put_hwpoison_page(page) put_page(page)
2704extern int sysctl_memory_failure_early_kill;
2705extern int sysctl_memory_failure_recovery;
2706extern void shake_page(struct page *p, int access);
2707extern atomic_long_t num_poisoned_pages __read_mostly;
2708extern int soft_offline_page(struct page *page, int flags);
2709
2710
2711/*
2712 * Error handlers for various types of pages.
2713 */
2714enum mf_result {
2715 MF_IGNORED, /* Error: cannot be handled */
2716 MF_FAILED, /* Error: handling failed */
2717 MF_DELAYED, /* Will be handled later */
2718 MF_RECOVERED, /* Successfully recovered */
2719};
2720
2721enum mf_action_page_type {
2722 MF_MSG_KERNEL,
2723 MF_MSG_KERNEL_HIGH_ORDER,
2724 MF_MSG_SLAB,
2725 MF_MSG_DIFFERENT_COMPOUND,
2726 MF_MSG_POISONED_HUGE,
2727 MF_MSG_HUGE,
2728 MF_MSG_FREE_HUGE,
2729 MF_MSG_NON_PMD_HUGE,
2730 MF_MSG_UNMAP_FAILED,
2731 MF_MSG_DIRTY_SWAPCACHE,
2732 MF_MSG_CLEAN_SWAPCACHE,
2733 MF_MSG_DIRTY_MLOCKED_LRU,
2734 MF_MSG_CLEAN_MLOCKED_LRU,
2735 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2736 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2737 MF_MSG_DIRTY_LRU,
2738 MF_MSG_CLEAN_LRU,
2739 MF_MSG_TRUNCATED_LRU,
2740 MF_MSG_BUDDY,
2741 MF_MSG_BUDDY_2ND,
2742 MF_MSG_UNKNOWN,
2743};
2744
2745#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2746extern void clear_huge_page(struct page *page,
2747 unsigned long addr_hint,
2748 unsigned int pages_per_huge_page);
2749extern void copy_user_huge_page(struct page *dst, struct page *src,
2750 unsigned long addr_hint,
2751 struct vm_area_struct *vma,
2752 unsigned int pages_per_huge_page);
2753extern long copy_huge_page_from_user(struct page *dst_page,
2754 const void __user *usr_src,
2755 unsigned int pages_per_huge_page,
2756 bool allow_pagefault);
2757#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2758
2759extern struct page_ext_operations debug_guardpage_ops;
2760
2761#ifdef CONFIG_DEBUG_PAGEALLOC
2762extern unsigned int _debug_guardpage_minorder;
2763extern bool _debug_guardpage_enabled;
2764
2765static inline unsigned int debug_guardpage_minorder(void)
2766{
2767 return _debug_guardpage_minorder;
2768}
2769
2770static inline bool debug_guardpage_enabled(void)
2771{
2772 return _debug_guardpage_enabled;
2773}
2774
2775static inline bool page_is_guard(struct page *page)
2776{
2777 struct page_ext *page_ext;
2778
2779 if (!debug_guardpage_enabled())
2780 return false;
2781
2782 page_ext = lookup_page_ext(page);
2783 if (unlikely(!page_ext))
2784 return false;
2785
2786 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2787}
2788#else
2789static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2790static inline bool debug_guardpage_enabled(void) { return false; }
2791static inline bool page_is_guard(struct page *page) { return false; }
2792#endif /* CONFIG_DEBUG_PAGEALLOC */
2793
2794#if MAX_NUMNODES > 1
2795void __init setup_nr_node_ids(void);
2796#else
2797static inline void setup_nr_node_ids(void) {}
2798#endif
2799
2800#endif /* __KERNEL__ */
2801#endif /* _LINUX_MM_H */
2802