signal/ptrace: Add force_sig_ptrace_errno_trap and use it where needed
[linux-block.git] / include / linux / mm.h
... / ...
CommitLineData
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
29struct mempolicy;
30struct anon_vma;
31struct anon_vma_chain;
32struct file_ra_state;
33struct user_struct;
34struct writeback_control;
35struct bdi_writeback;
36
37void init_mm_internals(void);
38
39#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40extern unsigned long max_mapnr;
41
42static inline void set_max_mapnr(unsigned long limit)
43{
44 max_mapnr = limit;
45}
46#else
47static inline void set_max_mapnr(unsigned long limit) { }
48#endif
49
50extern unsigned long totalram_pages;
51extern void * high_memory;
52extern int page_cluster;
53
54#ifdef CONFIG_SYSCTL
55extern int sysctl_legacy_va_layout;
56#else
57#define sysctl_legacy_va_layout 0
58#endif
59
60#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61extern const int mmap_rnd_bits_min;
62extern const int mmap_rnd_bits_max;
63extern int mmap_rnd_bits __read_mostly;
64#endif
65#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66extern const int mmap_rnd_compat_bits_min;
67extern const int mmap_rnd_compat_bits_max;
68extern int mmap_rnd_compat_bits __read_mostly;
69#endif
70
71#include <asm/page.h>
72#include <asm/pgtable.h>
73#include <asm/processor.h>
74
75#ifndef __pa_symbol
76#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
77#endif
78
79#ifndef page_to_virt
80#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
81#endif
82
83#ifndef lm_alias
84#define lm_alias(x) __va(__pa_symbol(x))
85#endif
86
87/*
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
93 */
94#ifndef mm_forbids_zeropage
95#define mm_forbids_zeropage(X) (0)
96#endif
97
98/*
99 * On some architectures it is expensive to call memset() for small sizes.
100 * Those architectures should provide their own implementation of "struct page"
101 * zeroing by defining this macro in <asm/pgtable.h>.
102 */
103#ifndef mm_zero_struct_page
104#define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
105#endif
106
107/*
108 * Default maximum number of active map areas, this limits the number of vmas
109 * per mm struct. Users can overwrite this number by sysctl but there is a
110 * problem.
111 *
112 * When a program's coredump is generated as ELF format, a section is created
113 * per a vma. In ELF, the number of sections is represented in unsigned short.
114 * This means the number of sections should be smaller than 65535 at coredump.
115 * Because the kernel adds some informative sections to a image of program at
116 * generating coredump, we need some margin. The number of extra sections is
117 * 1-3 now and depends on arch. We use "5" as safe margin, here.
118 *
119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120 * not a hard limit any more. Although some userspace tools can be surprised by
121 * that.
122 */
123#define MAPCOUNT_ELF_CORE_MARGIN (5)
124#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
125
126extern int sysctl_max_map_count;
127
128extern unsigned long sysctl_user_reserve_kbytes;
129extern unsigned long sysctl_admin_reserve_kbytes;
130
131extern int sysctl_overcommit_memory;
132extern int sysctl_overcommit_ratio;
133extern unsigned long sysctl_overcommit_kbytes;
134
135extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
136 size_t *, loff_t *);
137extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
138 size_t *, loff_t *);
139
140#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
141
142/* to align the pointer to the (next) page boundary */
143#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
144
145/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
146#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
147
148/*
149 * Linux kernel virtual memory manager primitives.
150 * The idea being to have a "virtual" mm in the same way
151 * we have a virtual fs - giving a cleaner interface to the
152 * mm details, and allowing different kinds of memory mappings
153 * (from shared memory to executable loading to arbitrary
154 * mmap() functions).
155 */
156
157extern struct kmem_cache *vm_area_cachep;
158
159#ifndef CONFIG_MMU
160extern struct rb_root nommu_region_tree;
161extern struct rw_semaphore nommu_region_sem;
162
163extern unsigned int kobjsize(const void *objp);
164#endif
165
166/*
167 * vm_flags in vm_area_struct, see mm_types.h.
168 * When changing, update also include/trace/events/mmflags.h
169 */
170#define VM_NONE 0x00000000
171
172#define VM_READ 0x00000001 /* currently active flags */
173#define VM_WRITE 0x00000002
174#define VM_EXEC 0x00000004
175#define VM_SHARED 0x00000008
176
177/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
178#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
179#define VM_MAYWRITE 0x00000020
180#define VM_MAYEXEC 0x00000040
181#define VM_MAYSHARE 0x00000080
182
183#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
184#define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
185#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
186#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
187#define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
188
189#define VM_LOCKED 0x00002000
190#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
191
192 /* Used by sys_madvise() */
193#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
194#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
195
196#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
197#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
198#define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
199#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
200#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
201#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
202#define VM_SYNC 0x00800000 /* Synchronous page faults */
203#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
204#define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
205#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
206
207#ifdef CONFIG_MEM_SOFT_DIRTY
208# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
209#else
210# define VM_SOFTDIRTY 0
211#endif
212
213#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
214#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
215#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
216#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
217
218#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
219#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
220#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
221#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
222#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
223#define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
224#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
225#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
226#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
227#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
228#define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
229#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
230
231#if defined(CONFIG_X86)
232# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
233#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
234# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
235# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
236# define VM_PKEY_BIT1 VM_HIGH_ARCH_1
237# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
238# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
239#endif
240#elif defined(CONFIG_PPC)
241# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
242#elif defined(CONFIG_PARISC)
243# define VM_GROWSUP VM_ARCH_1
244#elif defined(CONFIG_METAG)
245# define VM_GROWSUP VM_ARCH_1
246#elif defined(CONFIG_IA64)
247# define VM_GROWSUP VM_ARCH_1
248#elif !defined(CONFIG_MMU)
249# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
250#endif
251
252#if defined(CONFIG_X86_INTEL_MPX)
253/* MPX specific bounds table or bounds directory */
254# define VM_MPX VM_HIGH_ARCH_4
255#else
256# define VM_MPX VM_NONE
257#endif
258
259#ifndef VM_GROWSUP
260# define VM_GROWSUP VM_NONE
261#endif
262
263/* Bits set in the VMA until the stack is in its final location */
264#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
265
266#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
267#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
268#endif
269
270#ifdef CONFIG_STACK_GROWSUP
271#define VM_STACK VM_GROWSUP
272#else
273#define VM_STACK VM_GROWSDOWN
274#endif
275
276#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
277
278/*
279 * Special vmas that are non-mergable, non-mlock()able.
280 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
281 */
282#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
283
284/* This mask defines which mm->def_flags a process can inherit its parent */
285#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
286
287/* This mask is used to clear all the VMA flags used by mlock */
288#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
289
290/*
291 * mapping from the currently active vm_flags protection bits (the
292 * low four bits) to a page protection mask..
293 */
294extern pgprot_t protection_map[16];
295
296#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
297#define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
298#define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
299#define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
300#define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
301#define FAULT_FLAG_TRIED 0x20 /* Second try */
302#define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
303#define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
304#define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
305
306#define FAULT_FLAG_TRACE \
307 { FAULT_FLAG_WRITE, "WRITE" }, \
308 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
309 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
310 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
311 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
312 { FAULT_FLAG_TRIED, "TRIED" }, \
313 { FAULT_FLAG_USER, "USER" }, \
314 { FAULT_FLAG_REMOTE, "REMOTE" }, \
315 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
316
317/*
318 * vm_fault is filled by the the pagefault handler and passed to the vma's
319 * ->fault function. The vma's ->fault is responsible for returning a bitmask
320 * of VM_FAULT_xxx flags that give details about how the fault was handled.
321 *
322 * MM layer fills up gfp_mask for page allocations but fault handler might
323 * alter it if its implementation requires a different allocation context.
324 *
325 * pgoff should be used in favour of virtual_address, if possible.
326 */
327struct vm_fault {
328 struct vm_area_struct *vma; /* Target VMA */
329 unsigned int flags; /* FAULT_FLAG_xxx flags */
330 gfp_t gfp_mask; /* gfp mask to be used for allocations */
331 pgoff_t pgoff; /* Logical page offset based on vma */
332 unsigned long address; /* Faulting virtual address */
333 pmd_t *pmd; /* Pointer to pmd entry matching
334 * the 'address' */
335 pud_t *pud; /* Pointer to pud entry matching
336 * the 'address'
337 */
338 pte_t orig_pte; /* Value of PTE at the time of fault */
339
340 struct page *cow_page; /* Page handler may use for COW fault */
341 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
342 struct page *page; /* ->fault handlers should return a
343 * page here, unless VM_FAULT_NOPAGE
344 * is set (which is also implied by
345 * VM_FAULT_ERROR).
346 */
347 /* These three entries are valid only while holding ptl lock */
348 pte_t *pte; /* Pointer to pte entry matching
349 * the 'address'. NULL if the page
350 * table hasn't been allocated.
351 */
352 spinlock_t *ptl; /* Page table lock.
353 * Protects pte page table if 'pte'
354 * is not NULL, otherwise pmd.
355 */
356 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
357 * vm_ops->map_pages() calls
358 * alloc_set_pte() from atomic context.
359 * do_fault_around() pre-allocates
360 * page table to avoid allocation from
361 * atomic context.
362 */
363};
364
365/* page entry size for vm->huge_fault() */
366enum page_entry_size {
367 PE_SIZE_PTE = 0,
368 PE_SIZE_PMD,
369 PE_SIZE_PUD,
370};
371
372/*
373 * These are the virtual MM functions - opening of an area, closing and
374 * unmapping it (needed to keep files on disk up-to-date etc), pointer
375 * to the functions called when a no-page or a wp-page exception occurs.
376 */
377struct vm_operations_struct {
378 void (*open)(struct vm_area_struct * area);
379 void (*close)(struct vm_area_struct * area);
380 int (*split)(struct vm_area_struct * area, unsigned long addr);
381 int (*mremap)(struct vm_area_struct * area);
382 int (*fault)(struct vm_fault *vmf);
383 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
384 void (*map_pages)(struct vm_fault *vmf,
385 pgoff_t start_pgoff, pgoff_t end_pgoff);
386
387 /* notification that a previously read-only page is about to become
388 * writable, if an error is returned it will cause a SIGBUS */
389 int (*page_mkwrite)(struct vm_fault *vmf);
390
391 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
392 int (*pfn_mkwrite)(struct vm_fault *vmf);
393
394 /* called by access_process_vm when get_user_pages() fails, typically
395 * for use by special VMAs that can switch between memory and hardware
396 */
397 int (*access)(struct vm_area_struct *vma, unsigned long addr,
398 void *buf, int len, int write);
399
400 /* Called by the /proc/PID/maps code to ask the vma whether it
401 * has a special name. Returning non-NULL will also cause this
402 * vma to be dumped unconditionally. */
403 const char *(*name)(struct vm_area_struct *vma);
404
405#ifdef CONFIG_NUMA
406 /*
407 * set_policy() op must add a reference to any non-NULL @new mempolicy
408 * to hold the policy upon return. Caller should pass NULL @new to
409 * remove a policy and fall back to surrounding context--i.e. do not
410 * install a MPOL_DEFAULT policy, nor the task or system default
411 * mempolicy.
412 */
413 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
414
415 /*
416 * get_policy() op must add reference [mpol_get()] to any policy at
417 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
418 * in mm/mempolicy.c will do this automatically.
419 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
420 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
421 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
422 * must return NULL--i.e., do not "fallback" to task or system default
423 * policy.
424 */
425 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
426 unsigned long addr);
427#endif
428 /*
429 * Called by vm_normal_page() for special PTEs to find the
430 * page for @addr. This is useful if the default behavior
431 * (using pte_page()) would not find the correct page.
432 */
433 struct page *(*find_special_page)(struct vm_area_struct *vma,
434 unsigned long addr);
435};
436
437struct mmu_gather;
438struct inode;
439
440#define page_private(page) ((page)->private)
441#define set_page_private(page, v) ((page)->private = (v))
442
443#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
444static inline int pmd_devmap(pmd_t pmd)
445{
446 return 0;
447}
448static inline int pud_devmap(pud_t pud)
449{
450 return 0;
451}
452static inline int pgd_devmap(pgd_t pgd)
453{
454 return 0;
455}
456#endif
457
458/*
459 * FIXME: take this include out, include page-flags.h in
460 * files which need it (119 of them)
461 */
462#include <linux/page-flags.h>
463#include <linux/huge_mm.h>
464
465/*
466 * Methods to modify the page usage count.
467 *
468 * What counts for a page usage:
469 * - cache mapping (page->mapping)
470 * - private data (page->private)
471 * - page mapped in a task's page tables, each mapping
472 * is counted separately
473 *
474 * Also, many kernel routines increase the page count before a critical
475 * routine so they can be sure the page doesn't go away from under them.
476 */
477
478/*
479 * Drop a ref, return true if the refcount fell to zero (the page has no users)
480 */
481static inline int put_page_testzero(struct page *page)
482{
483 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
484 return page_ref_dec_and_test(page);
485}
486
487/*
488 * Try to grab a ref unless the page has a refcount of zero, return false if
489 * that is the case.
490 * This can be called when MMU is off so it must not access
491 * any of the virtual mappings.
492 */
493static inline int get_page_unless_zero(struct page *page)
494{
495 return page_ref_add_unless(page, 1, 0);
496}
497
498extern int page_is_ram(unsigned long pfn);
499
500enum {
501 REGION_INTERSECTS,
502 REGION_DISJOINT,
503 REGION_MIXED,
504};
505
506int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
507 unsigned long desc);
508
509/* Support for virtually mapped pages */
510struct page *vmalloc_to_page(const void *addr);
511unsigned long vmalloc_to_pfn(const void *addr);
512
513/*
514 * Determine if an address is within the vmalloc range
515 *
516 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
517 * is no special casing required.
518 */
519static inline bool is_vmalloc_addr(const void *x)
520{
521#ifdef CONFIG_MMU
522 unsigned long addr = (unsigned long)x;
523
524 return addr >= VMALLOC_START && addr < VMALLOC_END;
525#else
526 return false;
527#endif
528}
529#ifdef CONFIG_MMU
530extern int is_vmalloc_or_module_addr(const void *x);
531#else
532static inline int is_vmalloc_or_module_addr(const void *x)
533{
534 return 0;
535}
536#endif
537
538extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
539static inline void *kvmalloc(size_t size, gfp_t flags)
540{
541 return kvmalloc_node(size, flags, NUMA_NO_NODE);
542}
543static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
544{
545 return kvmalloc_node(size, flags | __GFP_ZERO, node);
546}
547static inline void *kvzalloc(size_t size, gfp_t flags)
548{
549 return kvmalloc(size, flags | __GFP_ZERO);
550}
551
552static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
553{
554 if (size != 0 && n > SIZE_MAX / size)
555 return NULL;
556
557 return kvmalloc(n * size, flags);
558}
559
560extern void kvfree(const void *addr);
561
562static inline atomic_t *compound_mapcount_ptr(struct page *page)
563{
564 return &page[1].compound_mapcount;
565}
566
567static inline int compound_mapcount(struct page *page)
568{
569 VM_BUG_ON_PAGE(!PageCompound(page), page);
570 page = compound_head(page);
571 return atomic_read(compound_mapcount_ptr(page)) + 1;
572}
573
574/*
575 * The atomic page->_mapcount, starts from -1: so that transitions
576 * both from it and to it can be tracked, using atomic_inc_and_test
577 * and atomic_add_negative(-1).
578 */
579static inline void page_mapcount_reset(struct page *page)
580{
581 atomic_set(&(page)->_mapcount, -1);
582}
583
584int __page_mapcount(struct page *page);
585
586static inline int page_mapcount(struct page *page)
587{
588 VM_BUG_ON_PAGE(PageSlab(page), page);
589
590 if (unlikely(PageCompound(page)))
591 return __page_mapcount(page);
592 return atomic_read(&page->_mapcount) + 1;
593}
594
595#ifdef CONFIG_TRANSPARENT_HUGEPAGE
596int total_mapcount(struct page *page);
597int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
598#else
599static inline int total_mapcount(struct page *page)
600{
601 return page_mapcount(page);
602}
603static inline int page_trans_huge_mapcount(struct page *page,
604 int *total_mapcount)
605{
606 int mapcount = page_mapcount(page);
607 if (total_mapcount)
608 *total_mapcount = mapcount;
609 return mapcount;
610}
611#endif
612
613static inline struct page *virt_to_head_page(const void *x)
614{
615 struct page *page = virt_to_page(x);
616
617 return compound_head(page);
618}
619
620void __put_page(struct page *page);
621
622void put_pages_list(struct list_head *pages);
623
624void split_page(struct page *page, unsigned int order);
625
626/*
627 * Compound pages have a destructor function. Provide a
628 * prototype for that function and accessor functions.
629 * These are _only_ valid on the head of a compound page.
630 */
631typedef void compound_page_dtor(struct page *);
632
633/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
634enum compound_dtor_id {
635 NULL_COMPOUND_DTOR,
636 COMPOUND_PAGE_DTOR,
637#ifdef CONFIG_HUGETLB_PAGE
638 HUGETLB_PAGE_DTOR,
639#endif
640#ifdef CONFIG_TRANSPARENT_HUGEPAGE
641 TRANSHUGE_PAGE_DTOR,
642#endif
643 NR_COMPOUND_DTORS,
644};
645extern compound_page_dtor * const compound_page_dtors[];
646
647static inline void set_compound_page_dtor(struct page *page,
648 enum compound_dtor_id compound_dtor)
649{
650 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
651 page[1].compound_dtor = compound_dtor;
652}
653
654static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
655{
656 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
657 return compound_page_dtors[page[1].compound_dtor];
658}
659
660static inline unsigned int compound_order(struct page *page)
661{
662 if (!PageHead(page))
663 return 0;
664 return page[1].compound_order;
665}
666
667static inline void set_compound_order(struct page *page, unsigned int order)
668{
669 page[1].compound_order = order;
670}
671
672void free_compound_page(struct page *page);
673
674#ifdef CONFIG_MMU
675/*
676 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
677 * servicing faults for write access. In the normal case, do always want
678 * pte_mkwrite. But get_user_pages can cause write faults for mappings
679 * that do not have writing enabled, when used by access_process_vm.
680 */
681static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
682{
683 if (likely(vma->vm_flags & VM_WRITE))
684 pte = pte_mkwrite(pte);
685 return pte;
686}
687
688int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
689 struct page *page);
690int finish_fault(struct vm_fault *vmf);
691int finish_mkwrite_fault(struct vm_fault *vmf);
692#endif
693
694/*
695 * Multiple processes may "see" the same page. E.g. for untouched
696 * mappings of /dev/null, all processes see the same page full of
697 * zeroes, and text pages of executables and shared libraries have
698 * only one copy in memory, at most, normally.
699 *
700 * For the non-reserved pages, page_count(page) denotes a reference count.
701 * page_count() == 0 means the page is free. page->lru is then used for
702 * freelist management in the buddy allocator.
703 * page_count() > 0 means the page has been allocated.
704 *
705 * Pages are allocated by the slab allocator in order to provide memory
706 * to kmalloc and kmem_cache_alloc. In this case, the management of the
707 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
708 * unless a particular usage is carefully commented. (the responsibility of
709 * freeing the kmalloc memory is the caller's, of course).
710 *
711 * A page may be used by anyone else who does a __get_free_page().
712 * In this case, page_count still tracks the references, and should only
713 * be used through the normal accessor functions. The top bits of page->flags
714 * and page->virtual store page management information, but all other fields
715 * are unused and could be used privately, carefully. The management of this
716 * page is the responsibility of the one who allocated it, and those who have
717 * subsequently been given references to it.
718 *
719 * The other pages (we may call them "pagecache pages") are completely
720 * managed by the Linux memory manager: I/O, buffers, swapping etc.
721 * The following discussion applies only to them.
722 *
723 * A pagecache page contains an opaque `private' member, which belongs to the
724 * page's address_space. Usually, this is the address of a circular list of
725 * the page's disk buffers. PG_private must be set to tell the VM to call
726 * into the filesystem to release these pages.
727 *
728 * A page may belong to an inode's memory mapping. In this case, page->mapping
729 * is the pointer to the inode, and page->index is the file offset of the page,
730 * in units of PAGE_SIZE.
731 *
732 * If pagecache pages are not associated with an inode, they are said to be
733 * anonymous pages. These may become associated with the swapcache, and in that
734 * case PG_swapcache is set, and page->private is an offset into the swapcache.
735 *
736 * In either case (swapcache or inode backed), the pagecache itself holds one
737 * reference to the page. Setting PG_private should also increment the
738 * refcount. The each user mapping also has a reference to the page.
739 *
740 * The pagecache pages are stored in a per-mapping radix tree, which is
741 * rooted at mapping->page_tree, and indexed by offset.
742 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
743 * lists, we instead now tag pages as dirty/writeback in the radix tree.
744 *
745 * All pagecache pages may be subject to I/O:
746 * - inode pages may need to be read from disk,
747 * - inode pages which have been modified and are MAP_SHARED may need
748 * to be written back to the inode on disk,
749 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
750 * modified may need to be swapped out to swap space and (later) to be read
751 * back into memory.
752 */
753
754/*
755 * The zone field is never updated after free_area_init_core()
756 * sets it, so none of the operations on it need to be atomic.
757 */
758
759/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
760#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
761#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
762#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
763#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
764
765/*
766 * Define the bit shifts to access each section. For non-existent
767 * sections we define the shift as 0; that plus a 0 mask ensures
768 * the compiler will optimise away reference to them.
769 */
770#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
771#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
772#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
773#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
774
775/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
776#ifdef NODE_NOT_IN_PAGE_FLAGS
777#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
778#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
779 SECTIONS_PGOFF : ZONES_PGOFF)
780#else
781#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
782#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
783 NODES_PGOFF : ZONES_PGOFF)
784#endif
785
786#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
787
788#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
790#endif
791
792#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
793#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
794#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
795#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
796#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
797
798static inline enum zone_type page_zonenum(const struct page *page)
799{
800 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
801}
802
803#ifdef CONFIG_ZONE_DEVICE
804static inline bool is_zone_device_page(const struct page *page)
805{
806 return page_zonenum(page) == ZONE_DEVICE;
807}
808#else
809static inline bool is_zone_device_page(const struct page *page)
810{
811 return false;
812}
813#endif
814
815#if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
816void put_zone_device_private_or_public_page(struct page *page);
817DECLARE_STATIC_KEY_FALSE(device_private_key);
818#define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
819static inline bool is_device_private_page(const struct page *page);
820static inline bool is_device_public_page(const struct page *page);
821#else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
822static inline void put_zone_device_private_or_public_page(struct page *page)
823{
824}
825#define IS_HMM_ENABLED 0
826static inline bool is_device_private_page(const struct page *page)
827{
828 return false;
829}
830static inline bool is_device_public_page(const struct page *page)
831{
832 return false;
833}
834#endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
835
836
837static inline void get_page(struct page *page)
838{
839 page = compound_head(page);
840 /*
841 * Getting a normal page or the head of a compound page
842 * requires to already have an elevated page->_refcount.
843 */
844 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
845 page_ref_inc(page);
846}
847
848static inline void put_page(struct page *page)
849{
850 page = compound_head(page);
851
852 /*
853 * For private device pages we need to catch refcount transition from
854 * 2 to 1, when refcount reach one it means the private device page is
855 * free and we need to inform the device driver through callback. See
856 * include/linux/memremap.h and HMM for details.
857 */
858 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
859 unlikely(is_device_public_page(page)))) {
860 put_zone_device_private_or_public_page(page);
861 return;
862 }
863
864 if (put_page_testzero(page))
865 __put_page(page);
866}
867
868#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
869#define SECTION_IN_PAGE_FLAGS
870#endif
871
872/*
873 * The identification function is mainly used by the buddy allocator for
874 * determining if two pages could be buddies. We are not really identifying
875 * the zone since we could be using the section number id if we do not have
876 * node id available in page flags.
877 * We only guarantee that it will return the same value for two combinable
878 * pages in a zone.
879 */
880static inline int page_zone_id(struct page *page)
881{
882 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
883}
884
885static inline int zone_to_nid(struct zone *zone)
886{
887#ifdef CONFIG_NUMA
888 return zone->node;
889#else
890 return 0;
891#endif
892}
893
894#ifdef NODE_NOT_IN_PAGE_FLAGS
895extern int page_to_nid(const struct page *page);
896#else
897static inline int page_to_nid(const struct page *page)
898{
899 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
900}
901#endif
902
903#ifdef CONFIG_NUMA_BALANCING
904static inline int cpu_pid_to_cpupid(int cpu, int pid)
905{
906 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
907}
908
909static inline int cpupid_to_pid(int cpupid)
910{
911 return cpupid & LAST__PID_MASK;
912}
913
914static inline int cpupid_to_cpu(int cpupid)
915{
916 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
917}
918
919static inline int cpupid_to_nid(int cpupid)
920{
921 return cpu_to_node(cpupid_to_cpu(cpupid));
922}
923
924static inline bool cpupid_pid_unset(int cpupid)
925{
926 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
927}
928
929static inline bool cpupid_cpu_unset(int cpupid)
930{
931 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
932}
933
934static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
935{
936 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
937}
938
939#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
940#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
941static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
942{
943 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
944}
945
946static inline int page_cpupid_last(struct page *page)
947{
948 return page->_last_cpupid;
949}
950static inline void page_cpupid_reset_last(struct page *page)
951{
952 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
953}
954#else
955static inline int page_cpupid_last(struct page *page)
956{
957 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
958}
959
960extern int page_cpupid_xchg_last(struct page *page, int cpupid);
961
962static inline void page_cpupid_reset_last(struct page *page)
963{
964 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
965}
966#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
967#else /* !CONFIG_NUMA_BALANCING */
968static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
969{
970 return page_to_nid(page); /* XXX */
971}
972
973static inline int page_cpupid_last(struct page *page)
974{
975 return page_to_nid(page); /* XXX */
976}
977
978static inline int cpupid_to_nid(int cpupid)
979{
980 return -1;
981}
982
983static inline int cpupid_to_pid(int cpupid)
984{
985 return -1;
986}
987
988static inline int cpupid_to_cpu(int cpupid)
989{
990 return -1;
991}
992
993static inline int cpu_pid_to_cpupid(int nid, int pid)
994{
995 return -1;
996}
997
998static inline bool cpupid_pid_unset(int cpupid)
999{
1000 return 1;
1001}
1002
1003static inline void page_cpupid_reset_last(struct page *page)
1004{
1005}
1006
1007static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1008{
1009 return false;
1010}
1011#endif /* CONFIG_NUMA_BALANCING */
1012
1013static inline struct zone *page_zone(const struct page *page)
1014{
1015 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1016}
1017
1018static inline pg_data_t *page_pgdat(const struct page *page)
1019{
1020 return NODE_DATA(page_to_nid(page));
1021}
1022
1023#ifdef SECTION_IN_PAGE_FLAGS
1024static inline void set_page_section(struct page *page, unsigned long section)
1025{
1026 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1027 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1028}
1029
1030static inline unsigned long page_to_section(const struct page *page)
1031{
1032 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1033}
1034#endif
1035
1036static inline void set_page_zone(struct page *page, enum zone_type zone)
1037{
1038 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1039 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1040}
1041
1042static inline void set_page_node(struct page *page, unsigned long node)
1043{
1044 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1045 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1046}
1047
1048static inline void set_page_links(struct page *page, enum zone_type zone,
1049 unsigned long node, unsigned long pfn)
1050{
1051 set_page_zone(page, zone);
1052 set_page_node(page, node);
1053#ifdef SECTION_IN_PAGE_FLAGS
1054 set_page_section(page, pfn_to_section_nr(pfn));
1055#endif
1056}
1057
1058#ifdef CONFIG_MEMCG
1059static inline struct mem_cgroup *page_memcg(struct page *page)
1060{
1061 return page->mem_cgroup;
1062}
1063static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1064{
1065 WARN_ON_ONCE(!rcu_read_lock_held());
1066 return READ_ONCE(page->mem_cgroup);
1067}
1068#else
1069static inline struct mem_cgroup *page_memcg(struct page *page)
1070{
1071 return NULL;
1072}
1073static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1074{
1075 WARN_ON_ONCE(!rcu_read_lock_held());
1076 return NULL;
1077}
1078#endif
1079
1080/*
1081 * Some inline functions in vmstat.h depend on page_zone()
1082 */
1083#include <linux/vmstat.h>
1084
1085static __always_inline void *lowmem_page_address(const struct page *page)
1086{
1087 return page_to_virt(page);
1088}
1089
1090#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1091#define HASHED_PAGE_VIRTUAL
1092#endif
1093
1094#if defined(WANT_PAGE_VIRTUAL)
1095static inline void *page_address(const struct page *page)
1096{
1097 return page->virtual;
1098}
1099static inline void set_page_address(struct page *page, void *address)
1100{
1101 page->virtual = address;
1102}
1103#define page_address_init() do { } while(0)
1104#endif
1105
1106#if defined(HASHED_PAGE_VIRTUAL)
1107void *page_address(const struct page *page);
1108void set_page_address(struct page *page, void *virtual);
1109void page_address_init(void);
1110#endif
1111
1112#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1113#define page_address(page) lowmem_page_address(page)
1114#define set_page_address(page, address) do { } while(0)
1115#define page_address_init() do { } while(0)
1116#endif
1117
1118extern void *page_rmapping(struct page *page);
1119extern struct anon_vma *page_anon_vma(struct page *page);
1120extern struct address_space *page_mapping(struct page *page);
1121
1122extern struct address_space *__page_file_mapping(struct page *);
1123
1124static inline
1125struct address_space *page_file_mapping(struct page *page)
1126{
1127 if (unlikely(PageSwapCache(page)))
1128 return __page_file_mapping(page);
1129
1130 return page->mapping;
1131}
1132
1133extern pgoff_t __page_file_index(struct page *page);
1134
1135/*
1136 * Return the pagecache index of the passed page. Regular pagecache pages
1137 * use ->index whereas swapcache pages use swp_offset(->private)
1138 */
1139static inline pgoff_t page_index(struct page *page)
1140{
1141 if (unlikely(PageSwapCache(page)))
1142 return __page_file_index(page);
1143 return page->index;
1144}
1145
1146bool page_mapped(struct page *page);
1147struct address_space *page_mapping(struct page *page);
1148
1149/*
1150 * Return true only if the page has been allocated with
1151 * ALLOC_NO_WATERMARKS and the low watermark was not
1152 * met implying that the system is under some pressure.
1153 */
1154static inline bool page_is_pfmemalloc(struct page *page)
1155{
1156 /*
1157 * Page index cannot be this large so this must be
1158 * a pfmemalloc page.
1159 */
1160 return page->index == -1UL;
1161}
1162
1163/*
1164 * Only to be called by the page allocator on a freshly allocated
1165 * page.
1166 */
1167static inline void set_page_pfmemalloc(struct page *page)
1168{
1169 page->index = -1UL;
1170}
1171
1172static inline void clear_page_pfmemalloc(struct page *page)
1173{
1174 page->index = 0;
1175}
1176
1177/*
1178 * Different kinds of faults, as returned by handle_mm_fault().
1179 * Used to decide whether a process gets delivered SIGBUS or
1180 * just gets major/minor fault counters bumped up.
1181 */
1182
1183#define VM_FAULT_OOM 0x0001
1184#define VM_FAULT_SIGBUS 0x0002
1185#define VM_FAULT_MAJOR 0x0004
1186#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1187#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1188#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1189#define VM_FAULT_SIGSEGV 0x0040
1190
1191#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1192#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1193#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1194#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1195#define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1196#define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1197 * and needs fsync() to complete (for
1198 * synchronous page faults in DAX) */
1199
1200#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1201 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1202 VM_FAULT_FALLBACK)
1203
1204#define VM_FAULT_RESULT_TRACE \
1205 { VM_FAULT_OOM, "OOM" }, \
1206 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1207 { VM_FAULT_MAJOR, "MAJOR" }, \
1208 { VM_FAULT_WRITE, "WRITE" }, \
1209 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1210 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1211 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1212 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1213 { VM_FAULT_LOCKED, "LOCKED" }, \
1214 { VM_FAULT_RETRY, "RETRY" }, \
1215 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1216 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1217 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1218
1219/* Encode hstate index for a hwpoisoned large page */
1220#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1221#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1222
1223/*
1224 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1225 */
1226extern void pagefault_out_of_memory(void);
1227
1228#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1229
1230/*
1231 * Flags passed to show_mem() and show_free_areas() to suppress output in
1232 * various contexts.
1233 */
1234#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1235
1236extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1237
1238extern bool can_do_mlock(void);
1239extern int user_shm_lock(size_t, struct user_struct *);
1240extern void user_shm_unlock(size_t, struct user_struct *);
1241
1242/*
1243 * Parameter block passed down to zap_pte_range in exceptional cases.
1244 */
1245struct zap_details {
1246 struct address_space *check_mapping; /* Check page->mapping if set */
1247 pgoff_t first_index; /* Lowest page->index to unmap */
1248 pgoff_t last_index; /* Highest page->index to unmap */
1249};
1250
1251struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1252 pte_t pte, bool with_public_device);
1253#define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1254
1255struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1256 pmd_t pmd);
1257
1258int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1259 unsigned long size);
1260void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1261 unsigned long size);
1262void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1263 unsigned long start, unsigned long end);
1264
1265/**
1266 * mm_walk - callbacks for walk_page_range
1267 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1268 * this handler should only handle pud_trans_huge() puds.
1269 * the pmd_entry or pte_entry callbacks will be used for
1270 * regular PUDs.
1271 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1272 * this handler is required to be able to handle
1273 * pmd_trans_huge() pmds. They may simply choose to
1274 * split_huge_page() instead of handling it explicitly.
1275 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1276 * @pte_hole: if set, called for each hole at all levels
1277 * @hugetlb_entry: if set, called for each hugetlb entry
1278 * @test_walk: caller specific callback function to determine whether
1279 * we walk over the current vma or not. Returning 0
1280 * value means "do page table walk over the current vma,"
1281 * and a negative one means "abort current page table walk
1282 * right now." 1 means "skip the current vma."
1283 * @mm: mm_struct representing the target process of page table walk
1284 * @vma: vma currently walked (NULL if walking outside vmas)
1285 * @private: private data for callbacks' usage
1286 *
1287 * (see the comment on walk_page_range() for more details)
1288 */
1289struct mm_walk {
1290 int (*pud_entry)(pud_t *pud, unsigned long addr,
1291 unsigned long next, struct mm_walk *walk);
1292 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1293 unsigned long next, struct mm_walk *walk);
1294 int (*pte_entry)(pte_t *pte, unsigned long addr,
1295 unsigned long next, struct mm_walk *walk);
1296 int (*pte_hole)(unsigned long addr, unsigned long next,
1297 struct mm_walk *walk);
1298 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1299 unsigned long addr, unsigned long next,
1300 struct mm_walk *walk);
1301 int (*test_walk)(unsigned long addr, unsigned long next,
1302 struct mm_walk *walk);
1303 struct mm_struct *mm;
1304 struct vm_area_struct *vma;
1305 void *private;
1306};
1307
1308int walk_page_range(unsigned long addr, unsigned long end,
1309 struct mm_walk *walk);
1310int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1311void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1312 unsigned long end, unsigned long floor, unsigned long ceiling);
1313int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1314 struct vm_area_struct *vma);
1315void unmap_mapping_range(struct address_space *mapping,
1316 loff_t const holebegin, loff_t const holelen, int even_cows);
1317int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1318 unsigned long *start, unsigned long *end,
1319 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1320int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1321 unsigned long *pfn);
1322int follow_phys(struct vm_area_struct *vma, unsigned long address,
1323 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1324int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1325 void *buf, int len, int write);
1326
1327static inline void unmap_shared_mapping_range(struct address_space *mapping,
1328 loff_t const holebegin, loff_t const holelen)
1329{
1330 unmap_mapping_range(mapping, holebegin, holelen, 0);
1331}
1332
1333extern void truncate_pagecache(struct inode *inode, loff_t new);
1334extern void truncate_setsize(struct inode *inode, loff_t newsize);
1335void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1336void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1337int truncate_inode_page(struct address_space *mapping, struct page *page);
1338int generic_error_remove_page(struct address_space *mapping, struct page *page);
1339int invalidate_inode_page(struct page *page);
1340
1341#ifdef CONFIG_MMU
1342extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1343 unsigned int flags);
1344extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1345 unsigned long address, unsigned int fault_flags,
1346 bool *unlocked);
1347#else
1348static inline int handle_mm_fault(struct vm_area_struct *vma,
1349 unsigned long address, unsigned int flags)
1350{
1351 /* should never happen if there's no MMU */
1352 BUG();
1353 return VM_FAULT_SIGBUS;
1354}
1355static inline int fixup_user_fault(struct task_struct *tsk,
1356 struct mm_struct *mm, unsigned long address,
1357 unsigned int fault_flags, bool *unlocked)
1358{
1359 /* should never happen if there's no MMU */
1360 BUG();
1361 return -EFAULT;
1362}
1363#endif
1364
1365extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1366 unsigned int gup_flags);
1367extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1368 void *buf, int len, unsigned int gup_flags);
1369extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1370 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1371
1372long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1373 unsigned long start, unsigned long nr_pages,
1374 unsigned int gup_flags, struct page **pages,
1375 struct vm_area_struct **vmas, int *locked);
1376long get_user_pages(unsigned long start, unsigned long nr_pages,
1377 unsigned int gup_flags, struct page **pages,
1378 struct vm_area_struct **vmas);
1379long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1380 unsigned int gup_flags, struct page **pages, int *locked);
1381long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1382 struct page **pages, unsigned int gup_flags);
1383#ifdef CONFIG_FS_DAX
1384long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1385 unsigned int gup_flags, struct page **pages,
1386 struct vm_area_struct **vmas);
1387#else
1388static inline long get_user_pages_longterm(unsigned long start,
1389 unsigned long nr_pages, unsigned int gup_flags,
1390 struct page **pages, struct vm_area_struct **vmas)
1391{
1392 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1393}
1394#endif /* CONFIG_FS_DAX */
1395
1396int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1397 struct page **pages);
1398
1399/* Container for pinned pfns / pages */
1400struct frame_vector {
1401 unsigned int nr_allocated; /* Number of frames we have space for */
1402 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1403 bool got_ref; /* Did we pin pages by getting page ref? */
1404 bool is_pfns; /* Does array contain pages or pfns? */
1405 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1406 * pfns_vector_pages() or pfns_vector_pfns()
1407 * for access */
1408};
1409
1410struct frame_vector *frame_vector_create(unsigned int nr_frames);
1411void frame_vector_destroy(struct frame_vector *vec);
1412int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1413 unsigned int gup_flags, struct frame_vector *vec);
1414void put_vaddr_frames(struct frame_vector *vec);
1415int frame_vector_to_pages(struct frame_vector *vec);
1416void frame_vector_to_pfns(struct frame_vector *vec);
1417
1418static inline unsigned int frame_vector_count(struct frame_vector *vec)
1419{
1420 return vec->nr_frames;
1421}
1422
1423static inline struct page **frame_vector_pages(struct frame_vector *vec)
1424{
1425 if (vec->is_pfns) {
1426 int err = frame_vector_to_pages(vec);
1427
1428 if (err)
1429 return ERR_PTR(err);
1430 }
1431 return (struct page **)(vec->ptrs);
1432}
1433
1434static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1435{
1436 if (!vec->is_pfns)
1437 frame_vector_to_pfns(vec);
1438 return (unsigned long *)(vec->ptrs);
1439}
1440
1441struct kvec;
1442int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1443 struct page **pages);
1444int get_kernel_page(unsigned long start, int write, struct page **pages);
1445struct page *get_dump_page(unsigned long addr);
1446
1447extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1448extern void do_invalidatepage(struct page *page, unsigned int offset,
1449 unsigned int length);
1450
1451int __set_page_dirty_nobuffers(struct page *page);
1452int __set_page_dirty_no_writeback(struct page *page);
1453int redirty_page_for_writepage(struct writeback_control *wbc,
1454 struct page *page);
1455void account_page_dirtied(struct page *page, struct address_space *mapping);
1456void account_page_cleaned(struct page *page, struct address_space *mapping,
1457 struct bdi_writeback *wb);
1458int set_page_dirty(struct page *page);
1459int set_page_dirty_lock(struct page *page);
1460void __cancel_dirty_page(struct page *page);
1461static inline void cancel_dirty_page(struct page *page)
1462{
1463 /* Avoid atomic ops, locking, etc. when not actually needed. */
1464 if (PageDirty(page))
1465 __cancel_dirty_page(page);
1466}
1467int clear_page_dirty_for_io(struct page *page);
1468
1469int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1470
1471static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1472{
1473 return !vma->vm_ops;
1474}
1475
1476#ifdef CONFIG_SHMEM
1477/*
1478 * The vma_is_shmem is not inline because it is used only by slow
1479 * paths in userfault.
1480 */
1481bool vma_is_shmem(struct vm_area_struct *vma);
1482#else
1483static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1484#endif
1485
1486int vma_is_stack_for_current(struct vm_area_struct *vma);
1487
1488extern unsigned long move_page_tables(struct vm_area_struct *vma,
1489 unsigned long old_addr, struct vm_area_struct *new_vma,
1490 unsigned long new_addr, unsigned long len,
1491 bool need_rmap_locks);
1492extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1493 unsigned long end, pgprot_t newprot,
1494 int dirty_accountable, int prot_numa);
1495extern int mprotect_fixup(struct vm_area_struct *vma,
1496 struct vm_area_struct **pprev, unsigned long start,
1497 unsigned long end, unsigned long newflags);
1498
1499/*
1500 * doesn't attempt to fault and will return short.
1501 */
1502int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1503 struct page **pages);
1504/*
1505 * per-process(per-mm_struct) statistics.
1506 */
1507static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1508{
1509 long val = atomic_long_read(&mm->rss_stat.count[member]);
1510
1511#ifdef SPLIT_RSS_COUNTING
1512 /*
1513 * counter is updated in asynchronous manner and may go to minus.
1514 * But it's never be expected number for users.
1515 */
1516 if (val < 0)
1517 val = 0;
1518#endif
1519 return (unsigned long)val;
1520}
1521
1522static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1523{
1524 atomic_long_add(value, &mm->rss_stat.count[member]);
1525}
1526
1527static inline void inc_mm_counter(struct mm_struct *mm, int member)
1528{
1529 atomic_long_inc(&mm->rss_stat.count[member]);
1530}
1531
1532static inline void dec_mm_counter(struct mm_struct *mm, int member)
1533{
1534 atomic_long_dec(&mm->rss_stat.count[member]);
1535}
1536
1537/* Optimized variant when page is already known not to be PageAnon */
1538static inline int mm_counter_file(struct page *page)
1539{
1540 if (PageSwapBacked(page))
1541 return MM_SHMEMPAGES;
1542 return MM_FILEPAGES;
1543}
1544
1545static inline int mm_counter(struct page *page)
1546{
1547 if (PageAnon(page))
1548 return MM_ANONPAGES;
1549 return mm_counter_file(page);
1550}
1551
1552static inline unsigned long get_mm_rss(struct mm_struct *mm)
1553{
1554 return get_mm_counter(mm, MM_FILEPAGES) +
1555 get_mm_counter(mm, MM_ANONPAGES) +
1556 get_mm_counter(mm, MM_SHMEMPAGES);
1557}
1558
1559static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1560{
1561 return max(mm->hiwater_rss, get_mm_rss(mm));
1562}
1563
1564static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1565{
1566 return max(mm->hiwater_vm, mm->total_vm);
1567}
1568
1569static inline void update_hiwater_rss(struct mm_struct *mm)
1570{
1571 unsigned long _rss = get_mm_rss(mm);
1572
1573 if ((mm)->hiwater_rss < _rss)
1574 (mm)->hiwater_rss = _rss;
1575}
1576
1577static inline void update_hiwater_vm(struct mm_struct *mm)
1578{
1579 if (mm->hiwater_vm < mm->total_vm)
1580 mm->hiwater_vm = mm->total_vm;
1581}
1582
1583static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1584{
1585 mm->hiwater_rss = get_mm_rss(mm);
1586}
1587
1588static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1589 struct mm_struct *mm)
1590{
1591 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1592
1593 if (*maxrss < hiwater_rss)
1594 *maxrss = hiwater_rss;
1595}
1596
1597#if defined(SPLIT_RSS_COUNTING)
1598void sync_mm_rss(struct mm_struct *mm);
1599#else
1600static inline void sync_mm_rss(struct mm_struct *mm)
1601{
1602}
1603#endif
1604
1605#ifndef __HAVE_ARCH_PTE_DEVMAP
1606static inline int pte_devmap(pte_t pte)
1607{
1608 return 0;
1609}
1610#endif
1611
1612int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1613
1614extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1615 spinlock_t **ptl);
1616static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1617 spinlock_t **ptl)
1618{
1619 pte_t *ptep;
1620 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1621 return ptep;
1622}
1623
1624#ifdef __PAGETABLE_P4D_FOLDED
1625static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1626 unsigned long address)
1627{
1628 return 0;
1629}
1630#else
1631int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1632#endif
1633
1634#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1635static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1636 unsigned long address)
1637{
1638 return 0;
1639}
1640static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1641static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1642
1643#else
1644int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1645
1646static inline void mm_inc_nr_puds(struct mm_struct *mm)
1647{
1648 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1649}
1650
1651static inline void mm_dec_nr_puds(struct mm_struct *mm)
1652{
1653 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1654}
1655#endif
1656
1657#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1658static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1659 unsigned long address)
1660{
1661 return 0;
1662}
1663
1664static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1665static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1666
1667#else
1668int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1669
1670static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1671{
1672 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1673}
1674
1675static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1676{
1677 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1678}
1679#endif
1680
1681#ifdef CONFIG_MMU
1682static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1683{
1684 atomic_long_set(&mm->pgtables_bytes, 0);
1685}
1686
1687static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1688{
1689 return atomic_long_read(&mm->pgtables_bytes);
1690}
1691
1692static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1693{
1694 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1695}
1696
1697static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1698{
1699 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1700}
1701#else
1702
1703static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1704static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1705{
1706 return 0;
1707}
1708
1709static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1710static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1711#endif
1712
1713int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1714int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1715
1716/*
1717 * The following ifdef needed to get the 4level-fixup.h header to work.
1718 * Remove it when 4level-fixup.h has been removed.
1719 */
1720#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1721
1722#ifndef __ARCH_HAS_5LEVEL_HACK
1723static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1724 unsigned long address)
1725{
1726 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1727 NULL : p4d_offset(pgd, address);
1728}
1729
1730static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1731 unsigned long address)
1732{
1733 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1734 NULL : pud_offset(p4d, address);
1735}
1736#endif /* !__ARCH_HAS_5LEVEL_HACK */
1737
1738static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1739{
1740 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1741 NULL: pmd_offset(pud, address);
1742}
1743#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1744
1745#if USE_SPLIT_PTE_PTLOCKS
1746#if ALLOC_SPLIT_PTLOCKS
1747void __init ptlock_cache_init(void);
1748extern bool ptlock_alloc(struct page *page);
1749extern void ptlock_free(struct page *page);
1750
1751static inline spinlock_t *ptlock_ptr(struct page *page)
1752{
1753 return page->ptl;
1754}
1755#else /* ALLOC_SPLIT_PTLOCKS */
1756static inline void ptlock_cache_init(void)
1757{
1758}
1759
1760static inline bool ptlock_alloc(struct page *page)
1761{
1762 return true;
1763}
1764
1765static inline void ptlock_free(struct page *page)
1766{
1767}
1768
1769static inline spinlock_t *ptlock_ptr(struct page *page)
1770{
1771 return &page->ptl;
1772}
1773#endif /* ALLOC_SPLIT_PTLOCKS */
1774
1775static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1776{
1777 return ptlock_ptr(pmd_page(*pmd));
1778}
1779
1780static inline bool ptlock_init(struct page *page)
1781{
1782 /*
1783 * prep_new_page() initialize page->private (and therefore page->ptl)
1784 * with 0. Make sure nobody took it in use in between.
1785 *
1786 * It can happen if arch try to use slab for page table allocation:
1787 * slab code uses page->slab_cache, which share storage with page->ptl.
1788 */
1789 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1790 if (!ptlock_alloc(page))
1791 return false;
1792 spin_lock_init(ptlock_ptr(page));
1793 return true;
1794}
1795
1796/* Reset page->mapping so free_pages_check won't complain. */
1797static inline void pte_lock_deinit(struct page *page)
1798{
1799 page->mapping = NULL;
1800 ptlock_free(page);
1801}
1802
1803#else /* !USE_SPLIT_PTE_PTLOCKS */
1804/*
1805 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1806 */
1807static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1808{
1809 return &mm->page_table_lock;
1810}
1811static inline void ptlock_cache_init(void) {}
1812static inline bool ptlock_init(struct page *page) { return true; }
1813static inline void pte_lock_deinit(struct page *page) {}
1814#endif /* USE_SPLIT_PTE_PTLOCKS */
1815
1816static inline void pgtable_init(void)
1817{
1818 ptlock_cache_init();
1819 pgtable_cache_init();
1820}
1821
1822static inline bool pgtable_page_ctor(struct page *page)
1823{
1824 if (!ptlock_init(page))
1825 return false;
1826 inc_zone_page_state(page, NR_PAGETABLE);
1827 return true;
1828}
1829
1830static inline void pgtable_page_dtor(struct page *page)
1831{
1832 pte_lock_deinit(page);
1833 dec_zone_page_state(page, NR_PAGETABLE);
1834}
1835
1836#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1837({ \
1838 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1839 pte_t *__pte = pte_offset_map(pmd, address); \
1840 *(ptlp) = __ptl; \
1841 spin_lock(__ptl); \
1842 __pte; \
1843})
1844
1845#define pte_unmap_unlock(pte, ptl) do { \
1846 spin_unlock(ptl); \
1847 pte_unmap(pte); \
1848} while (0)
1849
1850#define pte_alloc(mm, pmd, address) \
1851 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1852
1853#define pte_alloc_map(mm, pmd, address) \
1854 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1855
1856#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1857 (pte_alloc(mm, pmd, address) ? \
1858 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1859
1860#define pte_alloc_kernel(pmd, address) \
1861 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1862 NULL: pte_offset_kernel(pmd, address))
1863
1864#if USE_SPLIT_PMD_PTLOCKS
1865
1866static struct page *pmd_to_page(pmd_t *pmd)
1867{
1868 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1869 return virt_to_page((void *)((unsigned long) pmd & mask));
1870}
1871
1872static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1873{
1874 return ptlock_ptr(pmd_to_page(pmd));
1875}
1876
1877static inline bool pgtable_pmd_page_ctor(struct page *page)
1878{
1879#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1880 page->pmd_huge_pte = NULL;
1881#endif
1882 return ptlock_init(page);
1883}
1884
1885static inline void pgtable_pmd_page_dtor(struct page *page)
1886{
1887#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1888 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1889#endif
1890 ptlock_free(page);
1891}
1892
1893#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1894
1895#else
1896
1897static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1898{
1899 return &mm->page_table_lock;
1900}
1901
1902static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1903static inline void pgtable_pmd_page_dtor(struct page *page) {}
1904
1905#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1906
1907#endif
1908
1909static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1910{
1911 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1912 spin_lock(ptl);
1913 return ptl;
1914}
1915
1916/*
1917 * No scalability reason to split PUD locks yet, but follow the same pattern
1918 * as the PMD locks to make it easier if we decide to. The VM should not be
1919 * considered ready to switch to split PUD locks yet; there may be places
1920 * which need to be converted from page_table_lock.
1921 */
1922static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1923{
1924 return &mm->page_table_lock;
1925}
1926
1927static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1928{
1929 spinlock_t *ptl = pud_lockptr(mm, pud);
1930
1931 spin_lock(ptl);
1932 return ptl;
1933}
1934
1935extern void __init pagecache_init(void);
1936extern void free_area_init(unsigned long * zones_size);
1937extern void free_area_init_node(int nid, unsigned long * zones_size,
1938 unsigned long zone_start_pfn, unsigned long *zholes_size);
1939extern void free_initmem(void);
1940
1941/*
1942 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1943 * into the buddy system. The freed pages will be poisoned with pattern
1944 * "poison" if it's within range [0, UCHAR_MAX].
1945 * Return pages freed into the buddy system.
1946 */
1947extern unsigned long free_reserved_area(void *start, void *end,
1948 int poison, char *s);
1949
1950#ifdef CONFIG_HIGHMEM
1951/*
1952 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1953 * and totalram_pages.
1954 */
1955extern void free_highmem_page(struct page *page);
1956#endif
1957
1958extern void adjust_managed_page_count(struct page *page, long count);
1959extern void mem_init_print_info(const char *str);
1960
1961extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1962
1963/* Free the reserved page into the buddy system, so it gets managed. */
1964static inline void __free_reserved_page(struct page *page)
1965{
1966 ClearPageReserved(page);
1967 init_page_count(page);
1968 __free_page(page);
1969}
1970
1971static inline void free_reserved_page(struct page *page)
1972{
1973 __free_reserved_page(page);
1974 adjust_managed_page_count(page, 1);
1975}
1976
1977static inline void mark_page_reserved(struct page *page)
1978{
1979 SetPageReserved(page);
1980 adjust_managed_page_count(page, -1);
1981}
1982
1983/*
1984 * Default method to free all the __init memory into the buddy system.
1985 * The freed pages will be poisoned with pattern "poison" if it's within
1986 * range [0, UCHAR_MAX].
1987 * Return pages freed into the buddy system.
1988 */
1989static inline unsigned long free_initmem_default(int poison)
1990{
1991 extern char __init_begin[], __init_end[];
1992
1993 return free_reserved_area(&__init_begin, &__init_end,
1994 poison, "unused kernel");
1995}
1996
1997static inline unsigned long get_num_physpages(void)
1998{
1999 int nid;
2000 unsigned long phys_pages = 0;
2001
2002 for_each_online_node(nid)
2003 phys_pages += node_present_pages(nid);
2004
2005 return phys_pages;
2006}
2007
2008#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2009/*
2010 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2011 * zones, allocate the backing mem_map and account for memory holes in a more
2012 * architecture independent manner. This is a substitute for creating the
2013 * zone_sizes[] and zholes_size[] arrays and passing them to
2014 * free_area_init_node()
2015 *
2016 * An architecture is expected to register range of page frames backed by
2017 * physical memory with memblock_add[_node]() before calling
2018 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2019 * usage, an architecture is expected to do something like
2020 *
2021 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2022 * max_highmem_pfn};
2023 * for_each_valid_physical_page_range()
2024 * memblock_add_node(base, size, nid)
2025 * free_area_init_nodes(max_zone_pfns);
2026 *
2027 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2028 * registered physical page range. Similarly
2029 * sparse_memory_present_with_active_regions() calls memory_present() for
2030 * each range when SPARSEMEM is enabled.
2031 *
2032 * See mm/page_alloc.c for more information on each function exposed by
2033 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2034 */
2035extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2036unsigned long node_map_pfn_alignment(void);
2037unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2038 unsigned long end_pfn);
2039extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2040 unsigned long end_pfn);
2041extern void get_pfn_range_for_nid(unsigned int nid,
2042 unsigned long *start_pfn, unsigned long *end_pfn);
2043extern unsigned long find_min_pfn_with_active_regions(void);
2044extern void free_bootmem_with_active_regions(int nid,
2045 unsigned long max_low_pfn);
2046extern void sparse_memory_present_with_active_regions(int nid);
2047
2048#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2049
2050#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2051 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2052static inline int __early_pfn_to_nid(unsigned long pfn,
2053 struct mminit_pfnnid_cache *state)
2054{
2055 return 0;
2056}
2057#else
2058/* please see mm/page_alloc.c */
2059extern int __meminit early_pfn_to_nid(unsigned long pfn);
2060/* there is a per-arch backend function. */
2061extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2062 struct mminit_pfnnid_cache *state);
2063#endif
2064
2065#ifdef CONFIG_HAVE_MEMBLOCK
2066void zero_resv_unavail(void);
2067#else
2068static inline void zero_resv_unavail(void) {}
2069#endif
2070
2071extern void set_dma_reserve(unsigned long new_dma_reserve);
2072extern void memmap_init_zone(unsigned long, int, unsigned long,
2073 unsigned long, enum memmap_context);
2074extern void setup_per_zone_wmarks(void);
2075extern int __meminit init_per_zone_wmark_min(void);
2076extern void mem_init(void);
2077extern void __init mmap_init(void);
2078extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2079extern long si_mem_available(void);
2080extern void si_meminfo(struct sysinfo * val);
2081extern void si_meminfo_node(struct sysinfo *val, int nid);
2082#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2083extern unsigned long arch_reserved_kernel_pages(void);
2084#endif
2085
2086extern __printf(3, 4)
2087void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2088
2089extern void setup_per_cpu_pageset(void);
2090
2091extern void zone_pcp_update(struct zone *zone);
2092extern void zone_pcp_reset(struct zone *zone);
2093
2094/* page_alloc.c */
2095extern int min_free_kbytes;
2096extern int watermark_scale_factor;
2097
2098/* nommu.c */
2099extern atomic_long_t mmap_pages_allocated;
2100extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2101
2102/* interval_tree.c */
2103void vma_interval_tree_insert(struct vm_area_struct *node,
2104 struct rb_root_cached *root);
2105void vma_interval_tree_insert_after(struct vm_area_struct *node,
2106 struct vm_area_struct *prev,
2107 struct rb_root_cached *root);
2108void vma_interval_tree_remove(struct vm_area_struct *node,
2109 struct rb_root_cached *root);
2110struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2111 unsigned long start, unsigned long last);
2112struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2113 unsigned long start, unsigned long last);
2114
2115#define vma_interval_tree_foreach(vma, root, start, last) \
2116 for (vma = vma_interval_tree_iter_first(root, start, last); \
2117 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2118
2119void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2120 struct rb_root_cached *root);
2121void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2122 struct rb_root_cached *root);
2123struct anon_vma_chain *
2124anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2125 unsigned long start, unsigned long last);
2126struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2127 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2128#ifdef CONFIG_DEBUG_VM_RB
2129void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2130#endif
2131
2132#define anon_vma_interval_tree_foreach(avc, root, start, last) \
2133 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2134 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2135
2136/* mmap.c */
2137extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2138extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2139 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2140 struct vm_area_struct *expand);
2141static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2142 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2143{
2144 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2145}
2146extern struct vm_area_struct *vma_merge(struct mm_struct *,
2147 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2148 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2149 struct mempolicy *, struct vm_userfaultfd_ctx);
2150extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2151extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2152 unsigned long addr, int new_below);
2153extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2154 unsigned long addr, int new_below);
2155extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2156extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2157 struct rb_node **, struct rb_node *);
2158extern void unlink_file_vma(struct vm_area_struct *);
2159extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2160 unsigned long addr, unsigned long len, pgoff_t pgoff,
2161 bool *need_rmap_locks);
2162extern void exit_mmap(struct mm_struct *);
2163
2164static inline int check_data_rlimit(unsigned long rlim,
2165 unsigned long new,
2166 unsigned long start,
2167 unsigned long end_data,
2168 unsigned long start_data)
2169{
2170 if (rlim < RLIM_INFINITY) {
2171 if (((new - start) + (end_data - start_data)) > rlim)
2172 return -ENOSPC;
2173 }
2174
2175 return 0;
2176}
2177
2178extern int mm_take_all_locks(struct mm_struct *mm);
2179extern void mm_drop_all_locks(struct mm_struct *mm);
2180
2181extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2182extern struct file *get_mm_exe_file(struct mm_struct *mm);
2183extern struct file *get_task_exe_file(struct task_struct *task);
2184
2185extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2186extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2187
2188extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2189 const struct vm_special_mapping *sm);
2190extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2191 unsigned long addr, unsigned long len,
2192 unsigned long flags,
2193 const struct vm_special_mapping *spec);
2194/* This is an obsolete alternative to _install_special_mapping. */
2195extern int install_special_mapping(struct mm_struct *mm,
2196 unsigned long addr, unsigned long len,
2197 unsigned long flags, struct page **pages);
2198
2199extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2200
2201extern unsigned long mmap_region(struct file *file, unsigned long addr,
2202 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2203 struct list_head *uf);
2204extern unsigned long do_mmap(struct file *file, unsigned long addr,
2205 unsigned long len, unsigned long prot, unsigned long flags,
2206 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2207 struct list_head *uf);
2208extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2209 struct list_head *uf);
2210
2211static inline unsigned long
2212do_mmap_pgoff(struct file *file, unsigned long addr,
2213 unsigned long len, unsigned long prot, unsigned long flags,
2214 unsigned long pgoff, unsigned long *populate,
2215 struct list_head *uf)
2216{
2217 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2218}
2219
2220#ifdef CONFIG_MMU
2221extern int __mm_populate(unsigned long addr, unsigned long len,
2222 int ignore_errors);
2223static inline void mm_populate(unsigned long addr, unsigned long len)
2224{
2225 /* Ignore errors */
2226 (void) __mm_populate(addr, len, 1);
2227}
2228#else
2229static inline void mm_populate(unsigned long addr, unsigned long len) {}
2230#endif
2231
2232/* These take the mm semaphore themselves */
2233extern int __must_check vm_brk(unsigned long, unsigned long);
2234extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2235extern int vm_munmap(unsigned long, size_t);
2236extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2237 unsigned long, unsigned long,
2238 unsigned long, unsigned long);
2239
2240struct vm_unmapped_area_info {
2241#define VM_UNMAPPED_AREA_TOPDOWN 1
2242 unsigned long flags;
2243 unsigned long length;
2244 unsigned long low_limit;
2245 unsigned long high_limit;
2246 unsigned long align_mask;
2247 unsigned long align_offset;
2248};
2249
2250extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2251extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2252
2253/*
2254 * Search for an unmapped address range.
2255 *
2256 * We are looking for a range that:
2257 * - does not intersect with any VMA;
2258 * - is contained within the [low_limit, high_limit) interval;
2259 * - is at least the desired size.
2260 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2261 */
2262static inline unsigned long
2263vm_unmapped_area(struct vm_unmapped_area_info *info)
2264{
2265 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2266 return unmapped_area_topdown(info);
2267 else
2268 return unmapped_area(info);
2269}
2270
2271/* truncate.c */
2272extern void truncate_inode_pages(struct address_space *, loff_t);
2273extern void truncate_inode_pages_range(struct address_space *,
2274 loff_t lstart, loff_t lend);
2275extern void truncate_inode_pages_final(struct address_space *);
2276
2277/* generic vm_area_ops exported for stackable file systems */
2278extern int filemap_fault(struct vm_fault *vmf);
2279extern void filemap_map_pages(struct vm_fault *vmf,
2280 pgoff_t start_pgoff, pgoff_t end_pgoff);
2281extern int filemap_page_mkwrite(struct vm_fault *vmf);
2282
2283/* mm/page-writeback.c */
2284int __must_check write_one_page(struct page *page);
2285void task_dirty_inc(struct task_struct *tsk);
2286
2287/* readahead.c */
2288#define VM_MAX_READAHEAD 128 /* kbytes */
2289#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2290
2291int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2292 pgoff_t offset, unsigned long nr_to_read);
2293
2294void page_cache_sync_readahead(struct address_space *mapping,
2295 struct file_ra_state *ra,
2296 struct file *filp,
2297 pgoff_t offset,
2298 unsigned long size);
2299
2300void page_cache_async_readahead(struct address_space *mapping,
2301 struct file_ra_state *ra,
2302 struct file *filp,
2303 struct page *pg,
2304 pgoff_t offset,
2305 unsigned long size);
2306
2307extern unsigned long stack_guard_gap;
2308/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2309extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2310
2311/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2312extern int expand_downwards(struct vm_area_struct *vma,
2313 unsigned long address);
2314#if VM_GROWSUP
2315extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2316#else
2317 #define expand_upwards(vma, address) (0)
2318#endif
2319
2320/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2321extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2322extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2323 struct vm_area_struct **pprev);
2324
2325/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2326 NULL if none. Assume start_addr < end_addr. */
2327static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2328{
2329 struct vm_area_struct * vma = find_vma(mm,start_addr);
2330
2331 if (vma && end_addr <= vma->vm_start)
2332 vma = NULL;
2333 return vma;
2334}
2335
2336static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2337{
2338 unsigned long vm_start = vma->vm_start;
2339
2340 if (vma->vm_flags & VM_GROWSDOWN) {
2341 vm_start -= stack_guard_gap;
2342 if (vm_start > vma->vm_start)
2343 vm_start = 0;
2344 }
2345 return vm_start;
2346}
2347
2348static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2349{
2350 unsigned long vm_end = vma->vm_end;
2351
2352 if (vma->vm_flags & VM_GROWSUP) {
2353 vm_end += stack_guard_gap;
2354 if (vm_end < vma->vm_end)
2355 vm_end = -PAGE_SIZE;
2356 }
2357 return vm_end;
2358}
2359
2360static inline unsigned long vma_pages(struct vm_area_struct *vma)
2361{
2362 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2363}
2364
2365/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2366static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2367 unsigned long vm_start, unsigned long vm_end)
2368{
2369 struct vm_area_struct *vma = find_vma(mm, vm_start);
2370
2371 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2372 vma = NULL;
2373
2374 return vma;
2375}
2376
2377#ifdef CONFIG_MMU
2378pgprot_t vm_get_page_prot(unsigned long vm_flags);
2379void vma_set_page_prot(struct vm_area_struct *vma);
2380#else
2381static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2382{
2383 return __pgprot(0);
2384}
2385static inline void vma_set_page_prot(struct vm_area_struct *vma)
2386{
2387 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2388}
2389#endif
2390
2391#ifdef CONFIG_NUMA_BALANCING
2392unsigned long change_prot_numa(struct vm_area_struct *vma,
2393 unsigned long start, unsigned long end);
2394#endif
2395
2396struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2397int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2398 unsigned long pfn, unsigned long size, pgprot_t);
2399int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2400int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2401 unsigned long pfn);
2402int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2403 unsigned long pfn, pgprot_t pgprot);
2404int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2405 pfn_t pfn);
2406int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2407 pfn_t pfn);
2408int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2409
2410
2411struct page *follow_page_mask(struct vm_area_struct *vma,
2412 unsigned long address, unsigned int foll_flags,
2413 unsigned int *page_mask);
2414
2415static inline struct page *follow_page(struct vm_area_struct *vma,
2416 unsigned long address, unsigned int foll_flags)
2417{
2418 unsigned int unused_page_mask;
2419 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2420}
2421
2422#define FOLL_WRITE 0x01 /* check pte is writable */
2423#define FOLL_TOUCH 0x02 /* mark page accessed */
2424#define FOLL_GET 0x04 /* do get_page on page */
2425#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2426#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2427#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2428 * and return without waiting upon it */
2429#define FOLL_POPULATE 0x40 /* fault in page */
2430#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2431#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2432#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2433#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2434#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2435#define FOLL_MLOCK 0x1000 /* lock present pages */
2436#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2437#define FOLL_COW 0x4000 /* internal GUP flag */
2438
2439static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2440{
2441 if (vm_fault & VM_FAULT_OOM)
2442 return -ENOMEM;
2443 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2444 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2445 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2446 return -EFAULT;
2447 return 0;
2448}
2449
2450typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2451 void *data);
2452extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2453 unsigned long size, pte_fn_t fn, void *data);
2454
2455
2456#ifdef CONFIG_PAGE_POISONING
2457extern bool page_poisoning_enabled(void);
2458extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2459extern bool page_is_poisoned(struct page *page);
2460#else
2461static inline bool page_poisoning_enabled(void) { return false; }
2462static inline void kernel_poison_pages(struct page *page, int numpages,
2463 int enable) { }
2464static inline bool page_is_poisoned(struct page *page) { return false; }
2465#endif
2466
2467#ifdef CONFIG_DEBUG_PAGEALLOC
2468extern bool _debug_pagealloc_enabled;
2469extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2470
2471static inline bool debug_pagealloc_enabled(void)
2472{
2473 return _debug_pagealloc_enabled;
2474}
2475
2476static inline void
2477kernel_map_pages(struct page *page, int numpages, int enable)
2478{
2479 if (!debug_pagealloc_enabled())
2480 return;
2481
2482 __kernel_map_pages(page, numpages, enable);
2483}
2484#ifdef CONFIG_HIBERNATION
2485extern bool kernel_page_present(struct page *page);
2486#endif /* CONFIG_HIBERNATION */
2487#else /* CONFIG_DEBUG_PAGEALLOC */
2488static inline void
2489kernel_map_pages(struct page *page, int numpages, int enable) {}
2490#ifdef CONFIG_HIBERNATION
2491static inline bool kernel_page_present(struct page *page) { return true; }
2492#endif /* CONFIG_HIBERNATION */
2493static inline bool debug_pagealloc_enabled(void)
2494{
2495 return false;
2496}
2497#endif /* CONFIG_DEBUG_PAGEALLOC */
2498
2499#ifdef __HAVE_ARCH_GATE_AREA
2500extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2501extern int in_gate_area_no_mm(unsigned long addr);
2502extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2503#else
2504static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2505{
2506 return NULL;
2507}
2508static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2509static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2510{
2511 return 0;
2512}
2513#endif /* __HAVE_ARCH_GATE_AREA */
2514
2515extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2516
2517#ifdef CONFIG_SYSCTL
2518extern int sysctl_drop_caches;
2519int drop_caches_sysctl_handler(struct ctl_table *, int,
2520 void __user *, size_t *, loff_t *);
2521#endif
2522
2523void drop_slab(void);
2524void drop_slab_node(int nid);
2525
2526#ifndef CONFIG_MMU
2527#define randomize_va_space 0
2528#else
2529extern int randomize_va_space;
2530#endif
2531
2532const char * arch_vma_name(struct vm_area_struct *vma);
2533void print_vma_addr(char *prefix, unsigned long rip);
2534
2535void sparse_mem_maps_populate_node(struct page **map_map,
2536 unsigned long pnum_begin,
2537 unsigned long pnum_end,
2538 unsigned long map_count,
2539 int nodeid);
2540
2541struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2542pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2543p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2544pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2545pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2546pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2547void *vmemmap_alloc_block(unsigned long size, int node);
2548struct vmem_altmap;
2549void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2550 struct vmem_altmap *altmap);
2551static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2552{
2553 return __vmemmap_alloc_block_buf(size, node, NULL);
2554}
2555
2556void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2557int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2558 int node);
2559int vmemmap_populate(unsigned long start, unsigned long end, int node);
2560void vmemmap_populate_print_last(void);
2561#ifdef CONFIG_MEMORY_HOTPLUG
2562void vmemmap_free(unsigned long start, unsigned long end);
2563#endif
2564void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2565 unsigned long nr_pages);
2566
2567enum mf_flags {
2568 MF_COUNT_INCREASED = 1 << 0,
2569 MF_ACTION_REQUIRED = 1 << 1,
2570 MF_MUST_KILL = 1 << 2,
2571 MF_SOFT_OFFLINE = 1 << 3,
2572};
2573extern int memory_failure(unsigned long pfn, int trapno, int flags);
2574extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2575extern int unpoison_memory(unsigned long pfn);
2576extern int get_hwpoison_page(struct page *page);
2577#define put_hwpoison_page(page) put_page(page)
2578extern int sysctl_memory_failure_early_kill;
2579extern int sysctl_memory_failure_recovery;
2580extern void shake_page(struct page *p, int access);
2581extern atomic_long_t num_poisoned_pages;
2582extern int soft_offline_page(struct page *page, int flags);
2583
2584
2585/*
2586 * Error handlers for various types of pages.
2587 */
2588enum mf_result {
2589 MF_IGNORED, /* Error: cannot be handled */
2590 MF_FAILED, /* Error: handling failed */
2591 MF_DELAYED, /* Will be handled later */
2592 MF_RECOVERED, /* Successfully recovered */
2593};
2594
2595enum mf_action_page_type {
2596 MF_MSG_KERNEL,
2597 MF_MSG_KERNEL_HIGH_ORDER,
2598 MF_MSG_SLAB,
2599 MF_MSG_DIFFERENT_COMPOUND,
2600 MF_MSG_POISONED_HUGE,
2601 MF_MSG_HUGE,
2602 MF_MSG_FREE_HUGE,
2603 MF_MSG_UNMAP_FAILED,
2604 MF_MSG_DIRTY_SWAPCACHE,
2605 MF_MSG_CLEAN_SWAPCACHE,
2606 MF_MSG_DIRTY_MLOCKED_LRU,
2607 MF_MSG_CLEAN_MLOCKED_LRU,
2608 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2609 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2610 MF_MSG_DIRTY_LRU,
2611 MF_MSG_CLEAN_LRU,
2612 MF_MSG_TRUNCATED_LRU,
2613 MF_MSG_BUDDY,
2614 MF_MSG_BUDDY_2ND,
2615 MF_MSG_UNKNOWN,
2616};
2617
2618#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2619extern void clear_huge_page(struct page *page,
2620 unsigned long addr_hint,
2621 unsigned int pages_per_huge_page);
2622extern void copy_user_huge_page(struct page *dst, struct page *src,
2623 unsigned long addr, struct vm_area_struct *vma,
2624 unsigned int pages_per_huge_page);
2625extern long copy_huge_page_from_user(struct page *dst_page,
2626 const void __user *usr_src,
2627 unsigned int pages_per_huge_page,
2628 bool allow_pagefault);
2629#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2630
2631extern struct page_ext_operations debug_guardpage_ops;
2632
2633#ifdef CONFIG_DEBUG_PAGEALLOC
2634extern unsigned int _debug_guardpage_minorder;
2635extern bool _debug_guardpage_enabled;
2636
2637static inline unsigned int debug_guardpage_minorder(void)
2638{
2639 return _debug_guardpage_minorder;
2640}
2641
2642static inline bool debug_guardpage_enabled(void)
2643{
2644 return _debug_guardpage_enabled;
2645}
2646
2647static inline bool page_is_guard(struct page *page)
2648{
2649 struct page_ext *page_ext;
2650
2651 if (!debug_guardpage_enabled())
2652 return false;
2653
2654 page_ext = lookup_page_ext(page);
2655 if (unlikely(!page_ext))
2656 return false;
2657
2658 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2659}
2660#else
2661static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2662static inline bool debug_guardpage_enabled(void) { return false; }
2663static inline bool page_is_guard(struct page *page) { return false; }
2664#endif /* CONFIG_DEBUG_PAGEALLOC */
2665
2666#if MAX_NUMNODES > 1
2667void __init setup_nr_node_ids(void);
2668#else
2669static inline void setup_nr_node_ids(void) {}
2670#endif
2671
2672#endif /* __KERNEL__ */
2673#endif /* _LINUX_MM_H */