1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_MM_H
3#define _LINUX_MM_H
4
5#include <linux/errno.h>
6#include <linux/mmdebug.h>
7#include <linux/gfp.h>
8#include <linux/bug.h>
9#include <linux/list.h>
10#include <linux/mmzone.h>
11#include <linux/rbtree.h>
12#include <linux/atomic.h>
13#include <linux/debug_locks.h>
14#include <linux/mm_types.h>
15#include <linux/mmap_lock.h>
16#include <linux/range.h>
17#include <linux/pfn.h>
18#include <linux/percpu-refcount.h>
19#include <linux/bit_spinlock.h>
20#include <linux/shrinker.h>
21#include <linux/resource.h>
22#include <linux/page_ext.h>
23#include <linux/err.h>
24#include <linux/page-flags.h>
25#include <linux/page_ref.h>
26#include <linux/overflow.h>
27#include <linux/sizes.h>
28#include <linux/sched.h>
29#include <linux/pgtable.h>
30#include <linux/kasan.h>
31#include <linux/memremap.h>
32
33struct mempolicy;
34struct anon_vma;
35struct anon_vma_chain;
36struct user_struct;
37struct pt_regs;
38
39extern int sysctl_page_lock_unfairness;
40
41void init_mm_internals(void);
42
43#ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
44extern unsigned long max_mapnr;
45
46static inline void set_max_mapnr(unsigned long limit)
47{
48 max_mapnr = limit;
49}
50#else
51static inline void set_max_mapnr(unsigned long limit) { }
52#endif
53
54extern atomic_long_t _totalram_pages;
55static inline unsigned long totalram_pages(void)
56{
57 return (unsigned long)atomic_long_read(&_totalram_pages);
58}
59
60static inline void totalram_pages_inc(void)
61{
62 atomic_long_inc(&_totalram_pages);
63}
64
65static inline void totalram_pages_dec(void)
66{
67 atomic_long_dec(&_totalram_pages);
68}
69
70static inline void totalram_pages_add(long count)
71{
72 atomic_long_add(count, &_totalram_pages);
73}
74
75extern void * high_memory;
76extern int page_cluster;
77
78#ifdef CONFIG_SYSCTL
79extern int sysctl_legacy_va_layout;
80#else
81#define sysctl_legacy_va_layout 0
82#endif
83
84#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
85extern const int mmap_rnd_bits_min;
86extern const int mmap_rnd_bits_max;
87extern int mmap_rnd_bits __read_mostly;
88#endif
89#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
90extern const int mmap_rnd_compat_bits_min;
91extern const int mmap_rnd_compat_bits_max;
92extern int mmap_rnd_compat_bits __read_mostly;
93#endif
94
95#include <asm/page.h>
96#include <asm/processor.h>
97
98/*
99 * Architectures that support memory tagging (assigning tags to memory regions,
100 * embedding these tags into addresses that point to these memory regions, and
101 * checking that the memory and the pointer tags match on memory accesses)
102 * redefine this macro to strip tags from pointers.
103 * It's defined as noop for architectures that don't support memory tagging.
104 */
105#ifndef untagged_addr
106#define untagged_addr(addr) (addr)
107#endif
108
109#ifndef __pa_symbol
110#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
111#endif
112
113#ifndef page_to_virt
114#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
115#endif
116
117#ifndef lm_alias
118#define lm_alias(x) __va(__pa_symbol(x))
119#endif
120
121/*
122 * To prevent common memory management code establishing
123 * a zero page mapping on a read fault.
124 * This macro should be defined within <asm/pgtable.h>.
125 * s390 does this to prevent multiplexing of hardware bits
126 * related to the physical page in case of virtualization.
127 */
128#ifndef mm_forbids_zeropage
129#define mm_forbids_zeropage(X) (0)
130#endif
131
132/*
133 * On some architectures it is expensive to call memset() for small sizes.
134 * If an architecture decides to implement their own version of
135 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
136 * define their own version of this macro in <asm/pgtable.h>
137 */
138#if BITS_PER_LONG == 64
139/* This function must be updated when the size of struct page grows above 80
140 * or reduces below 56. The idea that compiler optimizes out switch()
141 * statement, and only leaves move/store instructions. Also the compiler can
142 * combine write statements if they are both assignments and can be reordered,
143 * this can result in several of the writes here being dropped.
144 */
145#define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
146static inline void __mm_zero_struct_page(struct page *page)
147{
148 unsigned long *_pp = (void *)page;
149
150 /* Check that struct page is either 56, 64, 72, or 80 bytes */
151 BUILD_BUG_ON(sizeof(struct page) & 7);
152 BUILD_BUG_ON(sizeof(struct page) < 56);
153 BUILD_BUG_ON(sizeof(struct page) > 80);
154
155 switch (sizeof(struct page)) {
156 case 80:
157 _pp[9] = 0;
158 fallthrough;
159 case 72:
160 _pp[8] = 0;
161 fallthrough;
162 case 64:
163 _pp[7] = 0;
164 fallthrough;
165 case 56:
166 _pp[6] = 0;
167 _pp[5] = 0;
168 _pp[4] = 0;
169 _pp[3] = 0;
170 _pp[2] = 0;
171 _pp[1] = 0;
172 _pp[0] = 0;
173 }
174}
175#else
176#define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
177#endif
178
179/*
180 * Default maximum number of active map areas, this limits the number of vmas
181 * per mm struct. Users can overwrite this number by sysctl but there is a
182 * problem.
183 *
184 * When a program's coredump is generated as ELF format, a section is created
185 * per a vma. In ELF, the number of sections is represented in unsigned short.
186 * This means the number of sections should be smaller than 65535 at coredump.
187 * Because the kernel adds some informative sections to a image of program at
188 * generating coredump, we need some margin. The number of extra sections is
189 * 1-3 now and depends on arch. We use "5" as safe margin, here.
190 *
191 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
192 * not a hard limit any more. Although some userspace tools can be surprised by
193 * that.
194 */
195#define MAPCOUNT_ELF_CORE_MARGIN (5)
196#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
197
198extern int sysctl_max_map_count;
199
200extern unsigned long sysctl_user_reserve_kbytes;
201extern unsigned long sysctl_admin_reserve_kbytes;
202
203extern int sysctl_overcommit_memory;
204extern int sysctl_overcommit_ratio;
205extern unsigned long sysctl_overcommit_kbytes;
206
207int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
208 loff_t *);
209int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
210 loff_t *);
211int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
212 loff_t *);
213
214#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
215#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
216#define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
217#else
218#define nth_page(page,n) ((page) + (n))
219#define folio_page_idx(folio, p) ((p) - &(folio)->page)
220#endif
221
222/* to align the pointer to the (next) page boundary */
223#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
224
225/* to align the pointer to the (prev) page boundary */
226#define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
227
228/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
229#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
230
231#define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
232static inline struct folio *lru_to_folio(struct list_head *head)
233{
234 return list_entry((head)->prev, struct folio, lru);
235}
236
237void setup_initial_init_mm(void *start_code, void *end_code,
238 void *end_data, void *brk);
239
240/*
241 * Linux kernel virtual memory manager primitives.
242 * The idea being to have a "virtual" mm in the same way
243 * we have a virtual fs - giving a cleaner interface to the
244 * mm details, and allowing different kinds of memory mappings
245 * (from shared memory to executable loading to arbitrary
246 * mmap() functions).
247 */
248
249struct vm_area_struct *vm_area_alloc(struct mm_struct *);
250struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
251void vm_area_free(struct vm_area_struct *);
252
253#ifndef CONFIG_MMU
254extern struct rb_root nommu_region_tree;
255extern struct rw_semaphore nommu_region_sem;
256
257extern unsigned int kobjsize(const void *objp);
258#endif
259
260/*
261 * vm_flags in vm_area_struct, see mm_types.h.
262 * When changing, update also include/trace/events/mmflags.h
263 */
264#define VM_NONE 0x00000000
265
266#define VM_READ 0x00000001 /* currently active flags */
267#define VM_WRITE 0x00000002
268#define VM_EXEC 0x00000004
269#define VM_SHARED 0x00000008
270
271/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
272#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
273#define VM_MAYWRITE 0x00000020
274#define VM_MAYEXEC 0x00000040
275#define VM_MAYSHARE 0x00000080
276
277#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
278#define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
279#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
280#define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
281
282#define VM_LOCKED 0x00002000
283#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
284
285 /* Used by sys_madvise() */
286#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
287#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
288
289#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
290#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
291#define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
292#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
293#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
294#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
295#define VM_SYNC 0x00800000 /* Synchronous page faults */
296#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
297#define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
298#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
299
300#ifdef CONFIG_MEM_SOFT_DIRTY
301# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
302#else
303# define VM_SOFTDIRTY 0
304#endif
305
306#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
307#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
308#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
309#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
310
311#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
312#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
313#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
314#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
315#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
316#define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
317#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
318#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
319#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
320#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
321#define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
322#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
323
324#ifdef CONFIG_ARCH_HAS_PKEYS
325# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
326# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
327# define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
328# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
329# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
330#ifdef CONFIG_PPC
331# define VM_PKEY_BIT4 VM_HIGH_ARCH_4
332#else
333# define VM_PKEY_BIT4 0
334#endif
335#endif /* CONFIG_ARCH_HAS_PKEYS */
336
337#if defined(CONFIG_X86)
338# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
339#elif defined(CONFIG_PPC)
340# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
341#elif defined(CONFIG_PARISC)
342# define VM_GROWSUP VM_ARCH_1
343#elif defined(CONFIG_IA64)
344# define VM_GROWSUP VM_ARCH_1
345#elif defined(CONFIG_SPARC64)
346# define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
347# define VM_ARCH_CLEAR VM_SPARC_ADI
348#elif defined(CONFIG_ARM64)
349# define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
350# define VM_ARCH_CLEAR VM_ARM64_BTI
351#elif !defined(CONFIG_MMU)
352# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
353#endif
354
355#if defined(CONFIG_ARM64_MTE)
356# define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
357# define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
358#else
359# define VM_MTE VM_NONE
360# define VM_MTE_ALLOWED VM_NONE
361#endif
362
363#ifndef VM_GROWSUP
364# define VM_GROWSUP VM_NONE
365#endif
366
367#ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
368# define VM_UFFD_MINOR_BIT 37
369# define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
370#else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
371# define VM_UFFD_MINOR VM_NONE
372#endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
373
374/* Bits set in the VMA until the stack is in its final location */
375#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
376
377#define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
378
379/* Common data flag combinations */
380#define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
381 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
382#define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
383 VM_MAYWRITE | VM_MAYEXEC)
384#define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
385 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
386
387#ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
388#define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
389#endif
390
391#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
392#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
393#endif
394
395#ifdef CONFIG_STACK_GROWSUP
396#define VM_STACK VM_GROWSUP
397#else
398#define VM_STACK VM_GROWSDOWN
399#endif
400
401#define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
402
403/* VMA basic access permission flags */
404#define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
405
406
407/*
408 * Special vmas that are non-mergable, non-mlock()able.
409 */
410#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
411
412/* This mask prevents VMA from being scanned with khugepaged */
413#define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
414
415/* This mask defines which mm->def_flags a process can inherit its parent */
416#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
417
418/* This mask is used to clear all the VMA flags used by mlock */
419#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
420
421/* Arch-specific flags to clear when updating VM flags on protection change */
422#ifndef VM_ARCH_CLEAR
423# define VM_ARCH_CLEAR VM_NONE
424#endif
425#define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
426
427/*
428 * mapping from the currently active vm_flags protection bits (the
429 * low four bits) to a page protection mask..
430 */
431
432/*
433 * The default fault flags that should be used by most of the
434 * arch-specific page fault handlers.
435 */
436#define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
437 FAULT_FLAG_KILLABLE | \
438 FAULT_FLAG_INTERRUPTIBLE)
439
440/**
441 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
442 * @flags: Fault flags.
443 *
444 * This is mostly used for places where we want to try to avoid taking
445 * the mmap_lock for too long a time when waiting for another condition
446 * to change, in which case we can try to be polite to release the
447 * mmap_lock in the first round to avoid potential starvation of other
448 * processes that would also want the mmap_lock.
449 *
450 * Return: true if the page fault allows retry and this is the first
451 * attempt of the fault handling; false otherwise.
452 */
453static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
454{
455 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
456 (!(flags & FAULT_FLAG_TRIED));
457}
458
459#define FAULT_FLAG_TRACE \
460 { FAULT_FLAG_WRITE, "WRITE" }, \
461 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
462 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
463 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
464 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
465 { FAULT_FLAG_TRIED, "TRIED" }, \
466 { FAULT_FLAG_USER, "USER" }, \
467 { FAULT_FLAG_REMOTE, "REMOTE" }, \
468 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
469 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
470
471/*
472 * vm_fault is filled by the pagefault handler and passed to the vma's
473 * ->fault function. The vma's ->fault is responsible for returning a bitmask
474 * of VM_FAULT_xxx flags that give details about how the fault was handled.
475 *
476 * MM layer fills up gfp_mask for page allocations but fault handler might
477 * alter it if its implementation requires a different allocation context.
478 *
479 * pgoff should be used in favour of virtual_address, if possible.
480 */
481struct vm_fault {
482 const struct {
483 struct vm_area_struct *vma; /* Target VMA */
484 gfp_t gfp_mask; /* gfp mask to be used for allocations */
485 pgoff_t pgoff; /* Logical page offset based on vma */
486 unsigned long address; /* Faulting virtual address - masked */
487 unsigned long real_address; /* Faulting virtual address - unmasked */
488 };
489 enum fault_flag flags; /* FAULT_FLAG_xxx flags
490 * XXX: should really be 'const' */
491 pmd_t *pmd; /* Pointer to pmd entry matching
492 * the 'address' */
493 pud_t *pud; /* Pointer to pud entry matching
494 * the 'address'
495 */
496 union {
497 pte_t orig_pte; /* Value of PTE at the time of fault */
498 pmd_t orig_pmd; /* Value of PMD at the time of fault,
499 * used by PMD fault only.
500 */
501 };
502
503 struct page *cow_page; /* Page handler may use for COW fault */
504 struct page *page; /* ->fault handlers should return a
505 * page here, unless VM_FAULT_NOPAGE
506 * is set (which is also implied by
507 * VM_FAULT_ERROR).
508 */
509 /* These three entries are valid only while holding ptl lock */
510 pte_t *pte; /* Pointer to pte entry matching
511 * the 'address'. NULL if the page
512 * table hasn't been allocated.
513 */
514 spinlock_t *ptl; /* Page table lock.
515 * Protects pte page table if 'pte'
516 * is not NULL, otherwise pmd.
517 */
518 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
519 * vm_ops->map_pages() sets up a page
520 * table from atomic context.
521 * do_fault_around() pre-allocates
522 * page table to avoid allocation from
523 * atomic context.
524 */
525};
526
527/* page entry size for vm->huge_fault() */
528enum page_entry_size {
529 PE_SIZE_PTE = 0,
530 PE_SIZE_PMD,
531 PE_SIZE_PUD,
532};
533
534/*
535 * These are the virtual MM functions - opening of an area, closing and
536 * unmapping it (needed to keep files on disk up-to-date etc), pointer
537 * to the functions called when a no-page or a wp-page exception occurs.
538 */
539struct vm_operations_struct {
540 void (*open)(struct vm_area_struct * area);
541 /**
542 * @close: Called when the VMA is being removed from the MM.
543 * Context: User context. May sleep. Caller holds mmap_lock.
544 */
545 void (*close)(struct vm_area_struct * area);
546 /* Called any time before splitting to check if it's allowed */
547 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
548 int (*mremap)(struct vm_area_struct *area);
549 /*
550 * Called by mprotect() to make driver-specific permission
551 * checks before mprotect() is finalised. The VMA must not
552 * be modified. Returns 0 if eprotect() can proceed.
553 */
554 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
555 unsigned long end, unsigned long newflags);
556 vm_fault_t (*fault)(struct vm_fault *vmf);
557 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
558 enum page_entry_size pe_size);
559 vm_fault_t (*map_pages)(struct vm_fault *vmf,
560 pgoff_t start_pgoff, pgoff_t end_pgoff);
561 unsigned long (*pagesize)(struct vm_area_struct * area);
562
563 /* notification that a previously read-only page is about to become
564 * writable, if an error is returned it will cause a SIGBUS */
565 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
566
567 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
568 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
569
570 /* called by access_process_vm when get_user_pages() fails, typically
571 * for use by special VMAs. See also generic_access_phys() for a generic
572 * implementation useful for any iomem mapping.
573 */
574 int (*access)(struct vm_area_struct *vma, unsigned long addr,
575 void *buf, int len, int write);
576
577 /* Called by the /proc/PID/maps code to ask the vma whether it
578 * has a special name. Returning non-NULL will also cause this
579 * vma to be dumped unconditionally. */
580 const char *(*name)(struct vm_area_struct *vma);
581
582#ifdef CONFIG_NUMA
583 /*
584 * set_policy() op must add a reference to any non-NULL @new mempolicy
585 * to hold the policy upon return. Caller should pass NULL @new to
586 * remove a policy and fall back to surrounding context--i.e. do not
587 * install a MPOL_DEFAULT policy, nor the task or system default
588 * mempolicy.
589 */
590 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
591
592 /*
593 * get_policy() op must add reference [mpol_get()] to any policy at
594 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
595 * in mm/mempolicy.c will do this automatically.
596 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
597 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
598 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
599 * must return NULL--i.e., do not "fallback" to task or system default
600 * policy.
601 */
602 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
603 unsigned long addr);
604#endif
605 /*
606 * Called by vm_normal_page() for special PTEs to find the
607 * page for @addr. This is useful if the default behavior
608 * (using pte_page()) would not find the correct page.
609 */
610 struct page *(*find_special_page)(struct vm_area_struct *vma,
611 unsigned long addr);
612};
613
614static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
615{
616 static const struct vm_operations_struct dummy_vm_ops = {};
617
618 memset(vma, 0, sizeof(*vma));
619 vma->vm_mm = mm;
620 vma->vm_ops = &dummy_vm_ops;
621 INIT_LIST_HEAD(&vma->anon_vma_chain);
622}
623
624static inline void vma_set_anonymous(struct vm_area_struct *vma)
625{
626 vma->vm_ops = NULL;
627}
628
629static inline bool vma_is_anonymous(struct vm_area_struct *vma)
630{
631 return !vma->vm_ops;
632}
633
634static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
635{
636 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
637
638 if (!maybe_stack)
639 return false;
640
641 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
642 VM_STACK_INCOMPLETE_SETUP)
643 return true;
644
645 return false;
646}
647
648static inline bool vma_is_foreign(struct vm_area_struct *vma)
649{
650 if (!current->mm)
651 return true;
652
653 if (current->mm != vma->vm_mm)
654 return true;
655
656 return false;
657}
658
659static inline bool vma_is_accessible(struct vm_area_struct *vma)
660{
661 return vma->vm_flags & VM_ACCESS_FLAGS;
662}
663
664#ifdef CONFIG_SHMEM
665/*
666 * The vma_is_shmem is not inline because it is used only by slow
667 * paths in userfault.
668 */
669bool vma_is_shmem(struct vm_area_struct *vma);
670#else
671static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
672#endif
673
674int vma_is_stack_for_current(struct vm_area_struct *vma);
675
676/* flush_tlb_range() takes a vma, not a mm, and can care about flags */
677#define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
678
679struct mmu_gather;
680struct inode;
681
682static inline unsigned int compound_order(struct page *page)
683{
684 if (!PageHead(page))
685 return 0;
686 return page[1].compound_order;
687}
688
689/**
690 * folio_order - The allocation order of a folio.
691 * @folio: The folio.
692 *
693 * A folio is composed of 2^order pages. See get_order() for the definition
694 * of order.
695 *
696 * Return: The order of the folio.
697 */
698static inline unsigned int folio_order(struct folio *folio)
699{
700 return compound_order(&folio->page);
701}
702
703#include <linux/huge_mm.h>
704
705/*
706 * Methods to modify the page usage count.
707 *
708 * What counts for a page usage:
709 * - cache mapping (page->mapping)
710 * - private data (page->private)
711 * - page mapped in a task's page tables, each mapping
712 * is counted separately
713 *
714 * Also, many kernel routines increase the page count before a critical
715 * routine so they can be sure the page doesn't go away from under them.
716 */
717
718/*
719 * Drop a ref, return true if the refcount fell to zero (the page has no users)
720 */
721static inline int put_page_testzero(struct page *page)
722{
723 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
724 return page_ref_dec_and_test(page);
725}
726
727static inline int folio_put_testzero(struct folio *folio)
728{
729 return put_page_testzero(&folio->page);
730}
731
732/*
733 * Try to grab a ref unless the page has a refcount of zero, return false if
734 * that is the case.
735 * This can be called when MMU is off so it must not access
736 * any of the virtual mappings.
737 */
738static inline bool get_page_unless_zero(struct page *page)
739{
740 return page_ref_add_unless(page, 1, 0);
741}
742
743extern int page_is_ram(unsigned long pfn);
744
745enum {
746 REGION_INTERSECTS,
747 REGION_DISJOINT,
748 REGION_MIXED,
749};
750
751int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
752 unsigned long desc);
753
754/* Support for virtually mapped pages */
755struct page *vmalloc_to_page(const void *addr);
756unsigned long vmalloc_to_pfn(const void *addr);
757
758/*
759 * Determine if an address is within the vmalloc range
760 *
761 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
762 * is no special casing required.
763 */
764
765#ifndef is_ioremap_addr
766#define is_ioremap_addr(x) is_vmalloc_addr(x)
767#endif
768
769#ifdef CONFIG_MMU
770extern bool is_vmalloc_addr(const void *x);
771extern int is_vmalloc_or_module_addr(const void *x);
772#else
773static inline bool is_vmalloc_addr(const void *x)
774{
775 return false;
776}
777static inline int is_vmalloc_or_module_addr(const void *x)
778{
779 return 0;
780}
781#endif
782
783/*
784 * How many times the entire folio is mapped as a single unit (eg by a
785 * PMD or PUD entry). This is probably not what you want, except for
786 * debugging purposes; look at folio_mapcount() or page_mapcount()
787 * instead.
788 */
789static inline int folio_entire_mapcount(struct folio *folio)
790{
791 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
792 return atomic_read(folio_mapcount_ptr(folio)) + 1;
793}
794
795/*
796 * Mapcount of compound page as a whole, does not include mapped sub-pages.
797 *
798 * Must be called only for compound pages.
799 */
800static inline int compound_mapcount(struct page *page)
801{
802 return folio_entire_mapcount(page_folio(page));
803}
804
805/*
806 * The atomic page->_mapcount, starts from -1: so that transitions
807 * both from it and to it can be tracked, using atomic_inc_and_test
808 * and atomic_add_negative(-1).
809 */
810static inline void page_mapcount_reset(struct page *page)
811{
812 atomic_set(&(page)->_mapcount, -1);
813}
814
815int __page_mapcount(struct page *page);
816
817/*
818 * Mapcount of 0-order page; when compound sub-page, includes
819 * compound_mapcount().
820 *
821 * Result is undefined for pages which cannot be mapped into userspace.
822 * For example SLAB or special types of pages. See function page_has_type().
823 * They use this place in struct page differently.
824 */
825static inline int page_mapcount(struct page *page)
826{
827 if (unlikely(PageCompound(page)))
828 return __page_mapcount(page);
829 return atomic_read(&page->_mapcount) + 1;
830}
831
832int folio_mapcount(struct folio *folio);
833
834#ifdef CONFIG_TRANSPARENT_HUGEPAGE
835static inline int total_mapcount(struct page *page)
836{
837 return folio_mapcount(page_folio(page));
838}
839
840#else
841static inline int total_mapcount(struct page *page)
842{
843 return page_mapcount(page);
844}
845#endif
846
847static inline struct page *virt_to_head_page(const void *x)
848{
849 struct page *page = virt_to_page(x);
850
851 return compound_head(page);
852}
853
854static inline struct folio *virt_to_folio(const void *x)
855{
856 struct page *page = virt_to_page(x);
857
858 return page_folio(page);
859}
860
861void __folio_put(struct folio *folio);
862
863void put_pages_list(struct list_head *pages);
864
865void split_page(struct page *page, unsigned int order);
866void folio_copy(struct folio *dst, struct folio *src);
867
868unsigned long nr_free_buffer_pages(void);
869
870/*
871 * Compound pages have a destructor function. Provide a
872 * prototype for that function and accessor functions.
873 * These are _only_ valid on the head of a compound page.
874 */
875typedef void compound_page_dtor(struct page *);
876
877/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
878enum compound_dtor_id {
879 NULL_COMPOUND_DTOR,
880 COMPOUND_PAGE_DTOR,
881#ifdef CONFIG_HUGETLB_PAGE
882 HUGETLB_PAGE_DTOR,
883#endif
884#ifdef CONFIG_TRANSPARENT_HUGEPAGE
885 TRANSHUGE_PAGE_DTOR,
886#endif
887 NR_COMPOUND_DTORS,
888};
889extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
890
891static inline void set_compound_page_dtor(struct page *page,
892 enum compound_dtor_id compound_dtor)
893{
894 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
895 page[1].compound_dtor = compound_dtor;
896}
897
898void destroy_large_folio(struct folio *folio);
899
900static inline int head_compound_pincount(struct page *head)
901{
902 return atomic_read(compound_pincount_ptr(head));
903}
904
905static inline void set_compound_order(struct page *page, unsigned int order)
906{
907 page[1].compound_order = order;
908#ifdef CONFIG_64BIT
909 page[1].compound_nr = 1U << order;
910#endif
911}
912
913/* Returns the number of pages in this potentially compound page. */
914static inline unsigned long compound_nr(struct page *page)
915{
916 if (!PageHead(page))
917 return 1;
918#ifdef CONFIG_64BIT
919 return page[1].compound_nr;
920#else
921 return 1UL << compound_order(page);
922#endif
923}
924
925/* Returns the number of bytes in this potentially compound page. */
926static inline unsigned long page_size(struct page *page)
927{
928 return PAGE_SIZE << compound_order(page);
929}
930
931/* Returns the number of bits needed for the number of bytes in a page */
932static inline unsigned int page_shift(struct page *page)
933{
934 return PAGE_SHIFT + compound_order(page);
935}
936
937/**
938 * thp_order - Order of a transparent huge page.
939 * @page: Head page of a transparent huge page.
940 */
941static inline unsigned int thp_order(struct page *page)
942{
943 VM_BUG_ON_PGFLAGS(PageTail(page), page);
944 return compound_order(page);
945}
946
947/**
948 * thp_nr_pages - The number of regular pages in this huge page.
949 * @page: The head page of a huge page.
950 */
951static inline int thp_nr_pages(struct page *page)
952{
953 VM_BUG_ON_PGFLAGS(PageTail(page), page);
954 return compound_nr(page);
955}
956
957/**
958 * thp_size - Size of a transparent huge page.
959 * @page: Head page of a transparent huge page.
960 *
961 * Return: Number of bytes in this page.
962 */
963static inline unsigned long thp_size(struct page *page)
964{
965 return PAGE_SIZE << thp_order(page);
966}
967
968void free_compound_page(struct page *page);
969
970#ifdef CONFIG_MMU
971/*
972 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
973 * servicing faults for write access. In the normal case, do always want
974 * pte_mkwrite. But get_user_pages can cause write faults for mappings
975 * that do not have writing enabled, when used by access_process_vm.
976 */
977static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
978{
979 if (likely(vma->vm_flags & VM_WRITE))
980 pte = pte_mkwrite(pte);
981 return pte;
982}
983
984vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
985void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
986
987vm_fault_t finish_fault(struct vm_fault *vmf);
988vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
989#endif
990
991/*
992 * Multiple processes may "see" the same page. E.g. for untouched
993 * mappings of /dev/null, all processes see the same page full of
994 * zeroes, and text pages of executables and shared libraries have
995 * only one copy in memory, at most, normally.
996 *
997 * For the non-reserved pages, page_count(page) denotes a reference count.
998 * page_count() == 0 means the page is free. page->lru is then used for
999 * freelist management in the buddy allocator.
1000 * page_count() > 0 means the page has been allocated.
1001 *
1002 * Pages are allocated by the slab allocator in order to provide memory
1003 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1004 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1005 * unless a particular usage is carefully commented. (the responsibility of
1006 * freeing the kmalloc memory is the caller's, of course).
1007 *
1008 * A page may be used by anyone else who does a __get_free_page().
1009 * In this case, page_count still tracks the references, and should only
1010 * be used through the normal accessor functions. The top bits of page->flags
1011 * and page->virtual store page management information, but all other fields
1012 * are unused and could be used privately, carefully. The management of this
1013 * page is the responsibility of the one who allocated it, and those who have
1014 * subsequently been given references to it.
1015 *
1016 * The other pages (we may call them "pagecache pages") are completely
1017 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1018 * The following discussion applies only to them.
1019 *
1020 * A pagecache page contains an opaque `private' member, which belongs to the
1021 * page's address_space. Usually, this is the address of a circular list of
1022 * the page's disk buffers. PG_private must be set to tell the VM to call
1023 * into the filesystem to release these pages.
1024 *
1025 * A page may belong to an inode's memory mapping. In this case, page->mapping
1026 * is the pointer to the inode, and page->index is the file offset of the page,
1027 * in units of PAGE_SIZE.
1028 *
1029 * If pagecache pages are not associated with an inode, they are said to be
1030 * anonymous pages. These may become associated with the swapcache, and in that
1031 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1032 *
1033 * In either case (swapcache or inode backed), the pagecache itself holds one
1034 * reference to the page. Setting PG_private should also increment the
1035 * refcount. The each user mapping also has a reference to the page.
1036 *
1037 * The pagecache pages are stored in a per-mapping radix tree, which is
1038 * rooted at mapping->i_pages, and indexed by offset.
1039 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1040 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1041 *
1042 * All pagecache pages may be subject to I/O:
1043 * - inode pages may need to be read from disk,
1044 * - inode pages which have been modified and are MAP_SHARED may need
1045 * to be written back to the inode on disk,
1046 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1047 * modified may need to be swapped out to swap space and (later) to be read
1048 * back into memory.
1049 */
1050
1051#if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1052DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1053
1054bool __put_devmap_managed_page_refs(struct page *page, int refs);
1055static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1056{
1057 if (!static_branch_unlikely(&devmap_managed_key))
1058 return false;
1059 if (!is_zone_device_page(page))
1060 return false;
1061 return __put_devmap_managed_page_refs(page, refs);
1062}
1063#else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1064static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1065{
1066 return false;
1067}
1068#endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1069
1070static inline bool put_devmap_managed_page(struct page *page)
1071{
1072 return put_devmap_managed_page_refs(page, 1);
1073}
1074
1075/* 127: arbitrary random number, small enough to assemble well */
1076#define folio_ref_zero_or_close_to_overflow(folio) \
1077 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1078
1079/**
1080 * folio_get - Increment the reference count on a folio.
1081 * @folio: The folio.
1082 *
1083 * Context: May be called in any context, as long as you know that
1084 * you have a refcount on the folio. If you do not already have one,
1085 * folio_try_get() may be the right interface for you to use.
1086 */
1087static inline void folio_get(struct folio *folio)
1088{
1089 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1090 folio_ref_inc(folio);
1091}
1092
1093static inline void get_page(struct page *page)
1094{
1095 folio_get(page_folio(page));
1096}
1097
1098bool __must_check try_grab_page(struct page *page, unsigned int flags);
1099
1100static inline __must_check bool try_get_page(struct page *page)
1101{
1102 page = compound_head(page);
1103 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1104 return false;
1105 page_ref_inc(page);
1106 return true;
1107}
1108
1109/**
1110 * folio_put - Decrement the reference count on a folio.
1111 * @folio: The folio.
1112 *
1113 * If the folio's reference count reaches zero, the memory will be
1114 * released back to the page allocator and may be used by another
1115 * allocation immediately. Do not access the memory or the struct folio
1116 * after calling folio_put() unless you can be sure that it wasn't the
1117 * last reference.
1118 *
1119 * Context: May be called in process or interrupt context, but not in NMI
1120 * context. May be called while holding a spinlock.
1121 */
1122static inline void folio_put(struct folio *folio)
1123{
1124 if (folio_put_testzero(folio))
1125 __folio_put(folio);
1126}
1127
1128/**
1129 * folio_put_refs - Reduce the reference count on a folio.
1130 * @folio: The folio.
1131 * @refs: The amount to subtract from the folio's reference count.
1132 *
1133 * If the folio's reference count reaches zero, the memory will be
1134 * released back to the page allocator and may be used by another
1135 * allocation immediately. Do not access the memory or the struct folio
1136 * after calling folio_put_refs() unless you can be sure that these weren't
1137 * the last references.
1138 *
1139 * Context: May be called in process or interrupt context, but not in NMI
1140 * context. May be called while holding a spinlock.
1141 */
1142static inline void folio_put_refs(struct folio *folio, int refs)
1143{
1144 if (folio_ref_sub_and_test(folio, refs))
1145 __folio_put(folio);
1146}
1147
1148void release_pages(struct page **pages, int nr);
1149
1150/**
1151 * folios_put - Decrement the reference count on an array of folios.
1152 * @folios: The folios.
1153 * @nr: How many folios there are.
1154 *
1155 * Like folio_put(), but for an array of folios. This is more efficient
1156 * than writing the loop yourself as it will optimise the locks which
1157 * need to be taken if the folios are freed.
1158 *
1159 * Context: May be called in process or interrupt context, but not in NMI
1160 * context. May be called while holding a spinlock.
1161 */
1162static inline void folios_put(struct folio **folios, unsigned int nr)
1163{
1164 release_pages((struct page **)folios, nr);
1165}
1166
1167static inline void put_page(struct page *page)
1168{
1169 struct folio *folio = page_folio(page);
1170
1171 /*
1172 * For some devmap managed pages we need to catch refcount transition
1173 * from 2 to 1:
1174 */
1175 if (put_devmap_managed_page(&folio->page))
1176 return;
1177 folio_put(folio);
1178}
1179
1180/*
1181 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1182 * the page's refcount so that two separate items are tracked: the original page
1183 * reference count, and also a new count of how many pin_user_pages() calls were
1184 * made against the page. ("gup-pinned" is another term for the latter).
1185 *
1186 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1187 * distinct from normal pages. As such, the unpin_user_page() call (and its
1188 * variants) must be used in order to release gup-pinned pages.
1189 *
1190 * Choice of value:
1191 *
1192 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1193 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1194 * simpler, due to the fact that adding an even power of two to the page
1195 * refcount has the effect of using only the upper N bits, for the code that
1196 * counts up using the bias value. This means that the lower bits are left for
1197 * the exclusive use of the original code that increments and decrements by one
1198 * (or at least, by much smaller values than the bias value).
1199 *
1200 * Of course, once the lower bits overflow into the upper bits (and this is
1201 * OK, because subtraction recovers the original values), then visual inspection
1202 * no longer suffices to directly view the separate counts. However, for normal
1203 * applications that don't have huge page reference counts, this won't be an
1204 * issue.
1205 *
1206 * Locking: the lockless algorithm described in folio_try_get_rcu()
1207 * provides safe operation for get_user_pages(), page_mkclean() and
1208 * other calls that race to set up page table entries.
1209 */
1210#define GUP_PIN_COUNTING_BIAS (1U << 10)
1211
1212void unpin_user_page(struct page *page);
1213void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1214 bool make_dirty);
1215void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1216 bool make_dirty);
1217void unpin_user_pages(struct page **pages, unsigned long npages);
1218
1219static inline bool is_cow_mapping(vm_flags_t flags)
1220{
1221 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1222}
1223
1224#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1225#define SECTION_IN_PAGE_FLAGS
1226#endif
1227
1228/*
1229 * The identification function is mainly used by the buddy allocator for
1230 * determining if two pages could be buddies. We are not really identifying
1231 * the zone since we could be using the section number id if we do not have
1232 * node id available in page flags.
1233 * We only guarantee that it will return the same value for two combinable
1234 * pages in a zone.
1235 */
1236static inline int page_zone_id(struct page *page)
1237{
1238 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1239}
1240
1241#ifdef NODE_NOT_IN_PAGE_FLAGS
1242extern int page_to_nid(const struct page *page);
1243#else
1244static inline int page_to_nid(const struct page *page)
1245{
1246 struct page *p = (struct page *)page;
1247
1248 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1249}
1250#endif
1251
1252static inline int folio_nid(const struct folio *folio)
1253{
1254 return page_to_nid(&folio->page);
1255}
1256
1257#ifdef CONFIG_NUMA_BALANCING
1258static inline int cpu_pid_to_cpupid(int cpu, int pid)
1259{
1260 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1261}
1262
1263static inline int cpupid_to_pid(int cpupid)
1264{
1265 return cpupid & LAST__PID_MASK;
1266}
1267
1268static inline int cpupid_to_cpu(int cpupid)
1269{
1270 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1271}
1272
1273static inline int cpupid_to_nid(int cpupid)
1274{
1275 return cpu_to_node(cpupid_to_cpu(cpupid));
1276}
1277
1278static inline bool cpupid_pid_unset(int cpupid)
1279{
1280 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1281}
1282
1283static inline bool cpupid_cpu_unset(int cpupid)
1284{
1285 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1286}
1287
1288static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1289{
1290 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1291}
1292
1293#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1294#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1295static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1296{
1297 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1298}
1299
1300static inline int page_cpupid_last(struct page *page)
1301{
1302 return page->_last_cpupid;
1303}
1304static inline void page_cpupid_reset_last(struct page *page)
1305{
1306 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1307}
1308#else
1309static inline int page_cpupid_last(struct page *page)
1310{
1311 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1312}
1313
1314extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1315
1316static inline void page_cpupid_reset_last(struct page *page)
1317{
1318 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1319}
1320#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1321#else /* !CONFIG_NUMA_BALANCING */
1322static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1323{
1324 return page_to_nid(page); /* XXX */
1325}
1326
1327static inline int page_cpupid_last(struct page *page)
1328{
1329 return page_to_nid(page); /* XXX */
1330}
1331
1332static inline int cpupid_to_nid(int cpupid)
1333{
1334 return -1;
1335}
1336
1337static inline int cpupid_to_pid(int cpupid)
1338{
1339 return -1;
1340}
1341
1342static inline int cpupid_to_cpu(int cpupid)
1343{
1344 return -1;
1345}
1346
1347static inline int cpu_pid_to_cpupid(int nid, int pid)
1348{
1349 return -1;
1350}
1351
1352static inline bool cpupid_pid_unset(int cpupid)
1353{
1354 return true;
1355}
1356
1357static inline void page_cpupid_reset_last(struct page *page)
1358{
1359}
1360
1361static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1362{
1363 return false;
1364}
1365#endif /* CONFIG_NUMA_BALANCING */
1366
1367#if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1368
1369/*
1370 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1371 * setting tags for all pages to native kernel tag value 0xff, as the default
1372 * value 0x00 maps to 0xff.
1373 */
1374
1375static inline u8 page_kasan_tag(const struct page *page)
1376{
1377 u8 tag = 0xff;
1378
1379 if (kasan_enabled()) {
1380 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1381 tag ^= 0xff;
1382 }
1383
1384 return tag;
1385}
1386
1387static inline void page_kasan_tag_set(struct page *page, u8 tag)
1388{
1389 unsigned long old_flags, flags;
1390
1391 if (!kasan_enabled())
1392 return;
1393
1394 tag ^= 0xff;
1395 old_flags = READ_ONCE(page->flags);
1396 do {
1397 flags = old_flags;
1398 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1399 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1400 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1401}
1402
1403static inline void page_kasan_tag_reset(struct page *page)
1404{
1405 if (kasan_enabled())
1406 page_kasan_tag_set(page, 0xff);
1407}
1408
1409#else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1410
1411static inline u8 page_kasan_tag(const struct page *page)
1412{
1413 return 0xff;
1414}
1415
1416static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1417static inline void page_kasan_tag_reset(struct page *page) { }
1418
1419#endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1420
1421static inline struct zone *page_zone(const struct page *page)
1422{
1423 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1424}
1425
1426static inline pg_data_t *page_pgdat(const struct page *page)
1427{
1428 return NODE_DATA(page_to_nid(page));
1429}
1430
1431static inline struct zone *folio_zone(const struct folio *folio)
1432{
1433 return page_zone(&folio->page);
1434}
1435
1436static inline pg_data_t *folio_pgdat(const struct folio *folio)
1437{
1438 return page_pgdat(&folio->page);
1439}
1440
1441#ifdef SECTION_IN_PAGE_FLAGS
1442static inline void set_page_section(struct page *page, unsigned long section)
1443{
1444 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1445 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1446}
1447
1448static inline unsigned long page_to_section(const struct page *page)
1449{
1450 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1451}
1452#endif
1453
1454/**
1455 * folio_pfn - Return the Page Frame Number of a folio.
1456 * @folio: The folio.
1457 *
1458 * A folio may contain multiple pages. The pages have consecutive
1459 * Page Frame Numbers.
1460 *
1461 * Return: The Page Frame Number of the first page in the folio.
1462 */
1463static inline unsigned long folio_pfn(struct folio *folio)
1464{
1465 return page_to_pfn(&folio->page);
1466}
1467
1468static inline atomic_t *folio_pincount_ptr(struct folio *folio)
1469{
1470 return &folio_page(folio, 1)->compound_pincount;
1471}
1472
1473/**
1474 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1475 * @folio: The folio.
1476 *
1477 * This function checks if a folio has been pinned via a call to
1478 * a function in the pin_user_pages() family.
1479 *
1480 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1481 * because it means "definitely not pinned for DMA", but true means "probably
1482 * pinned for DMA, but possibly a false positive due to having at least
1483 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1484 *
1485 * False positives are OK, because: a) it's unlikely for a folio to
1486 * get that many refcounts, and b) all the callers of this routine are
1487 * expected to be able to deal gracefully with a false positive.
1488 *
1489 * For large folios, the result will be exactly correct. That's because
1490 * we have more tracking data available: the compound_pincount is used
1491 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1492 *
1493 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1494 *
1495 * Return: True, if it is likely that the page has been "dma-pinned".
1496 * False, if the page is definitely not dma-pinned.
1497 */
1498static inline bool folio_maybe_dma_pinned(struct folio *folio)
1499{
1500 if (folio_test_large(folio))
1501 return atomic_read(folio_pincount_ptr(folio)) > 0;
1502
1503 /*
1504 * folio_ref_count() is signed. If that refcount overflows, then
1505 * folio_ref_count() returns a negative value, and callers will avoid
1506 * further incrementing the refcount.
1507 *
1508 * Here, for that overflow case, use the sign bit to count a little
1509 * bit higher via unsigned math, and thus still get an accurate result.
1510 */
1511 return ((unsigned int)folio_ref_count(folio)) >=
1512 GUP_PIN_COUNTING_BIAS;
1513}
1514
1515static inline bool page_maybe_dma_pinned(struct page *page)
1516{
1517 return folio_maybe_dma_pinned(page_folio(page));
1518}
1519
1520/*
1521 * This should most likely only be called during fork() to see whether we
1522 * should break the cow immediately for an anon page on the src mm.
1523 *
1524 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1525 */
1526static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1527 struct page *page)
1528{
1529 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1530
1531 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1532 return false;
1533
1534 return page_maybe_dma_pinned(page);
1535}
1536
1537/* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1538#ifdef CONFIG_MIGRATION
1539static inline bool is_longterm_pinnable_page(struct page *page)
1540{
1541#ifdef CONFIG_CMA
1542 int mt = get_pageblock_migratetype(page);
1543
1544 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1545 return false;
1546#endif
1547 return !(is_device_coherent_page(page) ||
1548 is_zone_movable_page(page) ||
1549 is_zero_pfn(page_to_pfn(page)));
1550}
1551#else
1552static inline bool is_longterm_pinnable_page(struct page *page)
1553{
1554 return true;
1555}
1556#endif
1557
1558static inline bool folio_is_longterm_pinnable(struct folio *folio)
1559{
1560 return is_longterm_pinnable_page(&folio->page);
1561}
1562
1563static inline void set_page_zone(struct page *page, enum zone_type zone)
1564{
1565 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1566 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1567}
1568
1569static inline void set_page_node(struct page *page, unsigned long node)
1570{
1571 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1572 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1573}
1574
1575static inline void set_page_links(struct page *page, enum zone_type zone,
1576 unsigned long node, unsigned long pfn)
1577{
1578 set_page_zone(page, zone);
1579 set_page_node(page, node);
1580#ifdef SECTION_IN_PAGE_FLAGS
1581 set_page_section(page, pfn_to_section_nr(pfn));
1582#endif
1583}
1584
1585/**
1586 * folio_nr_pages - The number of pages in the folio.
1587 * @folio: The folio.
1588 *
1589 * Return: A positive power of two.
1590 */
1591static inline long folio_nr_pages(struct folio *folio)
1592{
1593 return compound_nr(&folio->page);
1594}
1595
1596/**
1597 * folio_next - Move to the next physical folio.
1598 * @folio: The folio we're currently operating on.
1599 *
1600 * If you have physically contiguous memory which may span more than
1601 * one folio (eg a &struct bio_vec), use this function to move from one
1602 * folio to the next. Do not use it if the memory is only virtually
1603 * contiguous as the folios are almost certainly not adjacent to each
1604 * other. This is the folio equivalent to writing ``page++``.
1605 *
1606 * Context: We assume that the folios are refcounted and/or locked at a
1607 * higher level and do not adjust the reference counts.
1608 * Return: The next struct folio.
1609 */
1610static inline struct folio *folio_next(struct folio *folio)
1611{
1612 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1613}
1614
1615/**
1616 * folio_shift - The size of the memory described by this folio.
1617 * @folio: The folio.
1618 *
1619 * A folio represents a number of bytes which is a power-of-two in size.
1620 * This function tells you which power-of-two the folio is. See also
1621 * folio_size() and folio_order().
1622 *
1623 * Context: The caller should have a reference on the folio to prevent
1624 * it from being split. It is not necessary for the folio to be locked.
1625 * Return: The base-2 logarithm of the size of this folio.
1626 */
1627static inline unsigned int folio_shift(struct folio *folio)
1628{
1629 return PAGE_SHIFT + folio_order(folio);
1630}
1631
1632/**
1633 * folio_size - The number of bytes in a folio.
1634 * @folio: The folio.
1635 *
1636 * Context: The caller should have a reference on the folio to prevent
1637 * it from being split. It is not necessary for the folio to be locked.
1638 * Return: The number of bytes in this folio.
1639 */
1640static inline size_t folio_size(struct folio *folio)
1641{
1642 return PAGE_SIZE << folio_order(folio);
1643}
1644
1645#ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1646static inline int arch_make_page_accessible(struct page *page)
1647{
1648 return 0;
1649}
1650#endif
1651
1652#ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1653static inline int arch_make_folio_accessible(struct folio *folio)
1654{
1655 int ret;
1656 long i, nr = folio_nr_pages(folio);
1657
1658 for (i = 0; i < nr; i++) {
1659 ret = arch_make_page_accessible(folio_page(folio, i));
1660 if (ret)
1661 break;
1662 }
1663
1664 return ret;
1665}
1666#endif
1667
1668/*
1669 * Some inline functions in vmstat.h depend on page_zone()
1670 */
1671#include <linux/vmstat.h>
1672
1673static __always_inline void *lowmem_page_address(const struct page *page)
1674{
1675 return page_to_virt(page);
1676}
1677
1678#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1679#define HASHED_PAGE_VIRTUAL
1680#endif
1681
1682#if defined(WANT_PAGE_VIRTUAL)
1683static inline void *page_address(const struct page *page)
1684{
1685 return page->virtual;
1686}
1687static inline void set_page_address(struct page *page, void *address)
1688{
1689 page->virtual = address;
1690}
1691#define page_address_init() do { } while(0)
1692#endif
1693
1694#if defined(HASHED_PAGE_VIRTUAL)
1695void *page_address(const struct page *page);
1696void set_page_address(struct page *page, void *virtual);
1697void page_address_init(void);
1698#endif
1699
1700#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1701#define page_address(page) lowmem_page_address(page)
1702#define set_page_address(page, address) do { } while(0)
1703#define page_address_init() do { } while(0)
1704#endif
1705
1706static inline void *folio_address(const struct folio *folio)
1707{
1708 return page_address(&folio->page);
1709}
1710
1711extern void *page_rmapping(struct page *page);
1712extern pgoff_t __page_file_index(struct page *page);
1713
1714/*
1715 * Return the pagecache index of the passed page. Regular pagecache pages
1716 * use ->index whereas swapcache pages use swp_offset(->private)
1717 */
1718static inline pgoff_t page_index(struct page *page)
1719{
1720 if (unlikely(PageSwapCache(page)))
1721 return __page_file_index(page);
1722 return page->index;
1723}
1724
1725bool page_mapped(struct page *page);
1726bool folio_mapped(struct folio *folio);
1727
1728/*
1729 * Return true only if the page has been allocated with
1730 * ALLOC_NO_WATERMARKS and the low watermark was not
1731 * met implying that the system is under some pressure.
1732 */
1733static inline bool page_is_pfmemalloc(const struct page *page)
1734{
1735 /*
1736 * lru.next has bit 1 set if the page is allocated from the
1737 * pfmemalloc reserves. Callers may simply overwrite it if
1738 * they do not need to preserve that information.
1739 */
1740 return (uintptr_t)page->lru.next & BIT(1);
1741}
1742
1743/*
1744 * Only to be called by the page allocator on a freshly allocated
1745 * page.
1746 */
1747static inline void set_page_pfmemalloc(struct page *page)
1748{
1749 page->lru.next = (void *)BIT(1);
1750}
1751
1752static inline void clear_page_pfmemalloc(struct page *page)
1753{
1754 page->lru.next = NULL;
1755}
1756
1757/*
1758 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1759 */
1760extern void pagefault_out_of_memory(void);
1761
1762#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1763#define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1764#define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1765
1766/*
1767 * Flags passed to show_mem() and show_free_areas() to suppress output in
1768 * various contexts.
1769 */
1770#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1771
1772extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1773
1774#ifdef CONFIG_MMU
1775extern bool can_do_mlock(void);
1776#else
1777static inline bool can_do_mlock(void) { return false; }
1778#endif
1779extern int user_shm_lock(size_t, struct ucounts *);
1780extern void user_shm_unlock(size_t, struct ucounts *);
1781
1782struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1783 pte_t pte);
1784struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1785 pmd_t pmd);
1786
1787void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1788 unsigned long size);
1789void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1790 unsigned long size);
1791void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1792 unsigned long start, unsigned long end);
1793
1794struct mmu_notifier_range;
1795
1796void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1797 unsigned long end, unsigned long floor, unsigned long ceiling);
1798int
1799copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1800int follow_pte(struct mm_struct *mm, unsigned long address,
1801 pte_t **ptepp, spinlock_t **ptlp);
1802int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1803 unsigned long *pfn);
1804int follow_phys(struct vm_area_struct *vma, unsigned long address,
1805 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1806int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1807 void *buf, int len, int write);
1808
1809extern void truncate_pagecache(struct inode *inode, loff_t new);
1810extern void truncate_setsize(struct inode *inode, loff_t newsize);
1811void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1812void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1813int generic_error_remove_page(struct address_space *mapping, struct page *page);
1814
1815#ifdef CONFIG_MMU
1816extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1817 unsigned long address, unsigned int flags,
1818 struct pt_regs *regs);
1819extern int fixup_user_fault(struct mm_struct *mm,
1820 unsigned long address, unsigned int fault_flags,
1821 bool *unlocked);
1822void unmap_mapping_pages(struct address_space *mapping,
1823 pgoff_t start, pgoff_t nr, bool even_cows);
1824void unmap_mapping_range(struct address_space *mapping,
1825 loff_t const holebegin, loff_t const holelen, int even_cows);
1826#else
1827static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1828 unsigned long address, unsigned int flags,
1829 struct pt_regs *regs)
1830{
1831 /* should never happen if there's no MMU */
1832 BUG();
1833 return VM_FAULT_SIGBUS;
1834}
1835static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1836 unsigned int fault_flags, bool *unlocked)
1837{
1838 /* should never happen if there's no MMU */
1839 BUG();
1840 return -EFAULT;
1841}
1842static inline void unmap_mapping_pages(struct address_space *mapping,
1843 pgoff_t start, pgoff_t nr, bool even_cows) { }
1844static inline void unmap_mapping_range(struct address_space *mapping,
1845 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1846#endif
1847
1848static inline void unmap_shared_mapping_range(struct address_space *mapping,
1849 loff_t const holebegin, loff_t const holelen)
1850{
1851 unmap_mapping_range(mapping, holebegin, holelen, 0);
1852}
1853
1854extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1855 void *buf, int len, unsigned int gup_flags);
1856extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1857 void *buf, int len, unsigned int gup_flags);
1858extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1859 void *buf, int len, unsigned int gup_flags);
1860
1861long get_user_pages_remote(struct mm_struct *mm,
1862 unsigned long start, unsigned long nr_pages,
1863 unsigned int gup_flags, struct page **pages,
1864 struct vm_area_struct **vmas, int *locked);
1865long pin_user_pages_remote(struct mm_struct *mm,
1866 unsigned long start, unsigned long nr_pages,
1867 unsigned int gup_flags, struct page **pages,
1868 struct vm_area_struct **vmas, int *locked);
1869long get_user_pages(unsigned long start, unsigned long nr_pages,
1870 unsigned int gup_flags, struct page **pages,
1871 struct vm_area_struct **vmas);
1872long pin_user_pages(unsigned long start, unsigned long nr_pages,
1873 unsigned int gup_flags, struct page **pages,
1874 struct vm_area_struct **vmas);
1875long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1876 struct page **pages, unsigned int gup_flags);
1877long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1878 struct page **pages, unsigned int gup_flags);
1879
1880int get_user_pages_fast(unsigned long start, int nr_pages,
1881 unsigned int gup_flags, struct page **pages);
1882int pin_user_pages_fast(unsigned long start, int nr_pages,
1883 unsigned int gup_flags, struct page **pages);
1884
1885int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1886int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1887 struct task_struct *task, bool bypass_rlim);
1888
1889struct kvec;
1890int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1891 struct page **pages);
1892struct page *get_dump_page(unsigned long addr);
1893
1894bool folio_mark_dirty(struct folio *folio);
1895bool set_page_dirty(struct page *page);
1896int set_page_dirty_lock(struct page *page);
1897
1898int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1899
1900extern unsigned long move_page_tables(struct vm_area_struct *vma,
1901 unsigned long old_addr, struct vm_area_struct *new_vma,
1902 unsigned long new_addr, unsigned long len,
1903 bool need_rmap_locks);
1904
1905/*
1906 * Flags used by change_protection(). For now we make it a bitmap so
1907 * that we can pass in multiple flags just like parameters. However
1908 * for now all the callers are only use one of the flags at the same
1909 * time.
1910 */
1911/*
1912 * Whether we should manually check if we can map individual PTEs writable,
1913 * because something (e.g., COW, uffd-wp) blocks that from happening for all
1914 * PTEs automatically in a writable mapping.
1915 */
1916#define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
1917/* Whether this protection change is for NUMA hints */
1918#define MM_CP_PROT_NUMA (1UL << 1)
1919/* Whether this change is for write protecting */
1920#define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1921#define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1922#define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1923 MM_CP_UFFD_WP_RESOLVE)
1924
1925extern unsigned long change_protection(struct mmu_gather *tlb,
1926 struct vm_area_struct *vma, unsigned long start,
1927 unsigned long end, pgprot_t newprot,
1928 unsigned long cp_flags);
1929extern int mprotect_fixup(struct mmu_gather *tlb, struct vm_area_struct *vma,
1930 struct vm_area_struct **pprev, unsigned long start,
1931 unsigned long end, unsigned long newflags);
1932
1933/*
1934 * doesn't attempt to fault and will return short.
1935 */
1936int get_user_pages_fast_only(unsigned long start, int nr_pages,
1937 unsigned int gup_flags, struct page **pages);
1938int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1939 unsigned int gup_flags, struct page **pages);
1940
1941static inline bool get_user_page_fast_only(unsigned long addr,
1942 unsigned int gup_flags, struct page **pagep)
1943{
1944 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1945}
1946/*
1947 * per-process(per-mm_struct) statistics.
1948 */
1949static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1950{
1951 long val = atomic_long_read(&mm->rss_stat.count[member]);
1952
1953#ifdef SPLIT_RSS_COUNTING
1954 /*
1955 * counter is updated in asynchronous manner and may go to minus.
1956 * But it's never be expected number for users.
1957 */
1958 if (val < 0)
1959 val = 0;
1960#endif
1961 return (unsigned long)val;
1962}
1963
1964void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1965
1966static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1967{
1968 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1969
1970 mm_trace_rss_stat(mm, member, count);
1971}
1972
1973static inline void inc_mm_counter(struct mm_struct *mm, int member)
1974{
1975 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1976
1977 mm_trace_rss_stat(mm, member, count);
1978}
1979
1980static inline void dec_mm_counter(struct mm_struct *mm, int member)
1981{
1982 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1983
1984 mm_trace_rss_stat(mm, member, count);
1985}
1986
1987/* Optimized variant when page is already known not to be PageAnon */
1988static inline int mm_counter_file(struct page *page)
1989{
1990 if (PageSwapBacked(page))
1991 return MM_SHMEMPAGES;
1992 return MM_FILEPAGES;
1993}
1994
1995static inline int mm_counter(struct page *page)
1996{
1997 if (PageAnon(page))
1998 return MM_ANONPAGES;
1999 return mm_counter_file(page);
2000}
2001
2002static inline unsigned long get_mm_rss(struct mm_struct *mm)
2003{
2004 return get_mm_counter(mm, MM_FILEPAGES) +
2005 get_mm_counter(mm, MM_ANONPAGES) +
2006 get_mm_counter(mm, MM_SHMEMPAGES);
2007}
2008
2009static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2010{
2011 return max(mm->hiwater_rss, get_mm_rss(mm));
2012}
2013
2014static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2015{
2016 return max(mm->hiwater_vm, mm->total_vm);
2017}
2018
2019static inline void update_hiwater_rss(struct mm_struct *mm)
2020{
2021 unsigned long _rss = get_mm_rss(mm);
2022
2023 if ((mm)->hiwater_rss < _rss)
2024 (mm)->hiwater_rss = _rss;
2025}
2026
2027static inline void update_hiwater_vm(struct mm_struct *mm)
2028{
2029 if (mm->hiwater_vm < mm->total_vm)
2030 mm->hiwater_vm = mm->total_vm;
2031}
2032
2033static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2034{
2035 mm->hiwater_rss = get_mm_rss(mm);
2036}
2037
2038static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2039 struct mm_struct *mm)
2040{
2041 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2042
2043 if (*maxrss < hiwater_rss)
2044 *maxrss = hiwater_rss;
2045}
2046
2047#if defined(SPLIT_RSS_COUNTING)
2048void sync_mm_rss(struct mm_struct *mm);
2049#else
2050static inline void sync_mm_rss(struct mm_struct *mm)
2051{
2052}
2053#endif
2054
2055#ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2056static inline int pte_special(pte_t pte)
2057{
2058 return 0;
2059}
2060
2061static inline pte_t pte_mkspecial(pte_t pte)
2062{
2063 return pte;
2064}
2065#endif
2066
2067#ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2068static inline int pte_devmap(pte_t pte)
2069{
2070 return 0;
2071}
2072#endif
2073
2074int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2075
2076extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2077 spinlock_t **ptl);
2078static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2079 spinlock_t **ptl)
2080{
2081 pte_t *ptep;
2082 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2083 return ptep;
2084}
2085
2086#ifdef __PAGETABLE_P4D_FOLDED
2087static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2088 unsigned long address)
2089{
2090 return 0;
2091}
2092#else
2093int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2094#endif
2095
2096#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2097static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2098 unsigned long address)
2099{
2100 return 0;
2101}
2102static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2103static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2104
2105#else
2106int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2107
2108static inline void mm_inc_nr_puds(struct mm_struct *mm)
2109{
2110 if (mm_pud_folded(mm))
2111 return;
2112 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2113}
2114
2115static inline void mm_dec_nr_puds(struct mm_struct *mm)
2116{
2117 if (mm_pud_folded(mm))
2118 return;
2119 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2120}
2121#endif
2122
2123#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2124static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2125 unsigned long address)
2126{
2127 return 0;
2128}
2129
2130static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2131static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2132
2133#else
2134int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2135
2136static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2137{
2138 if (mm_pmd_folded(mm))
2139 return;
2140 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2141}
2142
2143static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2144{
2145 if (mm_pmd_folded(mm))
2146 return;
2147 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2148}
2149#endif
2150
2151#ifdef CONFIG_MMU
2152static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2153{
2154 atomic_long_set(&mm->pgtables_bytes, 0);
2155}
2156
2157static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2158{
2159 return atomic_long_read(&mm->pgtables_bytes);
2160}
2161
2162static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2163{
2164 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2165}
2166
2167static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2168{
2169 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2170}
2171#else
2172
2173static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2174static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2175{
2176 return 0;
2177}
2178
2179static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2180static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2181#endif
2182
2183int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2184int __pte_alloc_kernel(pmd_t *pmd);
2185
2186#if defined(CONFIG_MMU)
2187
2188static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2189 unsigned long address)
2190{
2191 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2192 NULL : p4d_offset(pgd, address);
2193}
2194
2195static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2196 unsigned long address)
2197{
2198 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2199 NULL : pud_offset(p4d, address);
2200}
2201
2202static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2203{
2204 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2205 NULL: pmd_offset(pud, address);
2206}
2207#endif /* CONFIG_MMU */
2208
2209#if USE_SPLIT_PTE_PTLOCKS
2210#if ALLOC_SPLIT_PTLOCKS
2211void __init ptlock_cache_init(void);
2212extern bool ptlock_alloc(struct page *page);
2213extern void ptlock_free(struct page *page);
2214
2215static inline spinlock_t *ptlock_ptr(struct page *page)
2216{
2217 return page->ptl;
2218}
2219#else /* ALLOC_SPLIT_PTLOCKS */
2220static inline void ptlock_cache_init(void)
2221{
2222}
2223
2224static inline bool ptlock_alloc(struct page *page)
2225{
2226 return true;
2227}
2228
2229static inline void ptlock_free(struct page *page)
2230{
2231}
2232
2233static inline spinlock_t *ptlock_ptr(struct page *page)
2234{
2235 return &page->ptl;
2236}
2237#endif /* ALLOC_SPLIT_PTLOCKS */
2238
2239static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2240{
2241 return ptlock_ptr(pmd_page(*pmd));
2242}
2243
2244static inline bool ptlock_init(struct page *page)
2245{
2246 /*
2247 * prep_new_page() initialize page->private (and therefore page->ptl)
2248 * with 0. Make sure nobody took it in use in between.
2249 *
2250 * It can happen if arch try to use slab for page table allocation:
2251 * slab code uses page->slab_cache, which share storage with page->ptl.
2252 */
2253 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2254 if (!ptlock_alloc(page))
2255 return false;
2256 spin_lock_init(ptlock_ptr(page));
2257 return true;
2258}
2259
2260#else /* !USE_SPLIT_PTE_PTLOCKS */
2261/*
2262 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2263 */
2264static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2265{
2266 return &mm->page_table_lock;
2267}
2268static inline void ptlock_cache_init(void) {}
2269static inline bool ptlock_init(struct page *page) { return true; }
2270static inline void ptlock_free(struct page *page) {}
2271#endif /* USE_SPLIT_PTE_PTLOCKS */
2272
2273static inline void pgtable_init(void)
2274{
2275 ptlock_cache_init();
2276 pgtable_cache_init();
2277}
2278
2279static inline bool pgtable_pte_page_ctor(struct page *page)
2280{
2281 if (!ptlock_init(page))
2282 return false;
2283 __SetPageTable(page);
2284 inc_lruvec_page_state(page, NR_PAGETABLE);
2285 return true;
2286}
2287
2288static inline void pgtable_pte_page_dtor(struct page *page)
2289{
2290 ptlock_free(page);
2291 __ClearPageTable(page);
2292 dec_lruvec_page_state(page, NR_PAGETABLE);
2293}
2294
2295#define pte_offset_map_lock(mm, pmd, address, ptlp) \
2296({ \
2297 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2298 pte_t *__pte = pte_offset_map(pmd, address); \
2299 *(ptlp) = __ptl; \
2300 spin_lock(__ptl); \
2301 __pte; \
2302})
2303
2304#define pte_unmap_unlock(pte, ptl) do { \
2305 spin_unlock(ptl); \
2306 pte_unmap(pte); \
2307} while (0)
2308
2309#define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2310
2311#define pte_alloc_map(mm, pmd, address) \
2312 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2313
2314#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2315 (pte_alloc(mm, pmd) ? \
2316 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2317
2318#define pte_alloc_kernel(pmd, address) \
2319 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2320 NULL: pte_offset_kernel(pmd, address))
2321
2322#if USE_SPLIT_PMD_PTLOCKS
2323
2324static struct page *pmd_to_page(pmd_t *pmd)
2325{
2326 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2327 return virt_to_page((void *)((unsigned long) pmd & mask));
2328}
2329
2330static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2331{
2332 return ptlock_ptr(pmd_to_page(pmd));
2333}
2334
2335static inline bool pmd_ptlock_init(struct page *page)
2336{
2337#ifdef CONFIG_TRANSPARENT_HUGEPAGE
2338 page->pmd_huge_pte = NULL;
2339#endif
2340 return ptlock_init(page);
2341}
2342
2343static inline void pmd_ptlock_free(struct page *page)
2344{
2345#ifdef CONFIG_TRANSPARENT_HUGEPAGE
2346 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2347#endif
2348 ptlock_free(page);
2349}
2350
2351#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2352
2353#else
2354
2355static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2356{
2357 return &mm->page_table_lock;
2358}
2359
2360static inline bool pmd_ptlock_init(struct page *page) { return true; }
2361static inline void pmd_ptlock_free(struct page *page) {}
2362
2363#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2364
2365#endif
2366
2367static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2368{
2369 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2370 spin_lock(ptl);
2371 return ptl;
2372}
2373
2374static inline bool pgtable_pmd_page_ctor(struct page *page)
2375{
2376 if (!pmd_ptlock_init(page))
2377 return false;
2378 __SetPageTable(page);
2379 inc_lruvec_page_state(page, NR_PAGETABLE);
2380 return true;
2381}
2382
2383static inline void pgtable_pmd_page_dtor(struct page *page)
2384{
2385 pmd_ptlock_free(page);
2386 __ClearPageTable(page);
2387 dec_lruvec_page_state(page, NR_PAGETABLE);
2388}
2389
2390/*
2391 * No scalability reason to split PUD locks yet, but follow the same pattern
2392 * as the PMD locks to make it easier if we decide to. The VM should not be
2393 * considered ready to switch to split PUD locks yet; there may be places
2394 * which need to be converted from page_table_lock.
2395 */
2396static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2397{
2398 return &mm->page_table_lock;
2399}
2400
2401static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2402{
2403 spinlock_t *ptl = pud_lockptr(mm, pud);
2404
2405 spin_lock(ptl);
2406 return ptl;
2407}
2408
2409extern void __init pagecache_init(void);
2410extern void free_initmem(void);
2411
2412/*
2413 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2414 * into the buddy system. The freed pages will be poisoned with pattern
2415 * "poison" if it's within range [0, UCHAR_MAX].
2416 * Return pages freed into the buddy system.
2417 */
2418extern unsigned long free_reserved_area(void *start, void *end,
2419 int poison, const char *s);
2420
2421extern void adjust_managed_page_count(struct page *page, long count);
2422extern void mem_init_print_info(void);
2423
2424extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2425
2426/* Free the reserved page into the buddy system, so it gets managed. */
2427static inline void free_reserved_page(struct page *page)
2428{
2429 ClearPageReserved(page);
2430 init_page_count(page);
2431 __free_page(page);
2432 adjust_managed_page_count(page, 1);
2433}
2434#define free_highmem_page(page) free_reserved_page(page)
2435
2436static inline void mark_page_reserved(struct page *page)
2437{
2438 SetPageReserved(page);
2439 adjust_managed_page_count(page, -1);
2440}
2441
2442/*
2443 * Default method to free all the __init memory into the buddy system.
2444 * The freed pages will be poisoned with pattern "poison" if it's within
2445 * range [0, UCHAR_MAX].
2446 * Return pages freed into the buddy system.
2447 */
2448static inline unsigned long free_initmem_default(int poison)
2449{
2450 extern char __init_begin[], __init_end[];
2451
2452 return free_reserved_area(&__init_begin, &__init_end,
2453 poison, "unused kernel image (initmem)");
2454}
2455
2456static inline unsigned long get_num_physpages(void)
2457{
2458 int nid;
2459 unsigned long phys_pages = 0;
2460
2461 for_each_online_node(nid)
2462 phys_pages += node_present_pages(nid);
2463
2464 return phys_pages;
2465}
2466
2467/*
2468 * Using memblock node mappings, an architecture may initialise its
2469 * zones, allocate the backing mem_map and account for memory holes in an
2470 * architecture independent manner.
2471 *
2472 * An architecture is expected to register range of page frames backed by
2473 * physical memory with memblock_add[_node]() before calling
2474 * free_area_init() passing in the PFN each zone ends at. At a basic
2475 * usage, an architecture is expected to do something like
2476 *
2477 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2478 * max_highmem_pfn};
2479 * for_each_valid_physical_page_range()
2480 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2481 * free_area_init(max_zone_pfns);
2482 */
2483void free_area_init(unsigned long *max_zone_pfn);
2484unsigned long node_map_pfn_alignment(void);
2485unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2486 unsigned long end_pfn);
2487extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2488 unsigned long end_pfn);
2489extern void get_pfn_range_for_nid(unsigned int nid,
2490 unsigned long *start_pfn, unsigned long *end_pfn);
2491extern unsigned long find_min_pfn_with_active_regions(void);
2492
2493#ifndef CONFIG_NUMA
2494static inline int early_pfn_to_nid(unsigned long pfn)
2495{
2496 return 0;
2497}
2498#else
2499/* please see mm/page_alloc.c */
2500extern int __meminit early_pfn_to_nid(unsigned long pfn);
2501#endif
2502
2503extern void set_dma_reserve(unsigned long new_dma_reserve);
2504extern void memmap_init_range(unsigned long, int, unsigned long,
2505 unsigned long, unsigned long, enum meminit_context,
2506 struct vmem_altmap *, int migratetype);
2507extern void setup_per_zone_wmarks(void);
2508extern void calculate_min_free_kbytes(void);
2509extern int __meminit init_per_zone_wmark_min(void);
2510extern void mem_init(void);
2511extern void __init mmap_init(void);
2512extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2513extern long si_mem_available(void);
2514extern void si_meminfo(struct sysinfo * val);
2515extern void si_meminfo_node(struct sysinfo *val, int nid);
2516#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2517extern unsigned long arch_reserved_kernel_pages(void);
2518#endif
2519
2520extern __printf(3, 4)
2521void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2522
2523extern void setup_per_cpu_pageset(void);
2524
2525/* page_alloc.c */
2526extern int min_free_kbytes;
2527extern int watermark_boost_factor;
2528extern int watermark_scale_factor;
2529extern bool arch_has_descending_max_zone_pfns(void);
2530
2531/* nommu.c */
2532extern atomic_long_t mmap_pages_allocated;
2533extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2534
2535/* interval_tree.c */
2536void vma_interval_tree_insert(struct vm_area_struct *node,
2537 struct rb_root_cached *root);
2538void vma_interval_tree_insert_after(struct vm_area_struct *node,
2539 struct vm_area_struct *prev,
2540 struct rb_root_cached *root);
2541void vma_interval_tree_remove(struct vm_area_struct *node,
2542 struct rb_root_cached *root);
2543struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2544 unsigned long start, unsigned long last);
2545struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2546 unsigned long start, unsigned long last);
2547
2548#define vma_interval_tree_foreach(vma, root, start, last) \
2549 for (vma = vma_interval_tree_iter_first(root, start, last); \
2550 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2551
2552void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2553 struct rb_root_cached *root);
2554void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2555 struct rb_root_cached *root);
2556struct anon_vma_chain *
2557anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2558 unsigned long start, unsigned long last);
2559struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2560 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2561#ifdef CONFIG_DEBUG_VM_RB
2562void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2563#endif
2564
2565#define anon_vma_interval_tree_foreach(avc, root, start, last) \
2566 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2567 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2568
2569/* mmap.c */
2570extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2571extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2572 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2573 struct vm_area_struct *expand);
2574static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2575 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2576{
2577 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2578}
2579extern struct vm_area_struct *vma_merge(struct mm_struct *,
2580 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2581 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2582 struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2583extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2584extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2585 unsigned long addr, int new_below);
2586extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2587 unsigned long addr, int new_below);
2588extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2589extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2590 struct rb_node **, struct rb_node *);
2591extern void unlink_file_vma(struct vm_area_struct *);
2592extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2593 unsigned long addr, unsigned long len, pgoff_t pgoff,
2594 bool *need_rmap_locks);
2595extern void exit_mmap(struct mm_struct *);
2596
2597static inline int check_data_rlimit(unsigned long rlim,
2598 unsigned long new,
2599 unsigned long start,
2600 unsigned long end_data,
2601 unsigned long start_data)
2602{
2603 if (rlim < RLIM_INFINITY) {
2604 if (((new - start) + (end_data - start_data)) > rlim)
2605 return -ENOSPC;
2606 }
2607
2608 return 0;
2609}
2610
2611extern int mm_take_all_locks(struct mm_struct *mm);
2612extern void mm_drop_all_locks(struct mm_struct *mm);
2613
2614extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2615extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2616extern struct file *get_mm_exe_file(struct mm_struct *mm);
2617extern struct file *get_task_exe_file(struct task_struct *task);
2618
2619extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2620extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2621
2622extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2623 const struct vm_special_mapping *sm);
2624extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2625 unsigned long addr, unsigned long len,
2626 unsigned long flags,
2627 const struct vm_special_mapping *spec);
2628/* This is an obsolete alternative to _install_special_mapping. */
2629extern int install_special_mapping(struct mm_struct *mm,
2630 unsigned long addr, unsigned long len,
2631 unsigned long flags, struct page **pages);
2632
2633unsigned long randomize_stack_top(unsigned long stack_top);
2634unsigned long randomize_page(unsigned long start, unsigned long range);
2635
2636extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2637
2638extern unsigned long mmap_region(struct file *file, unsigned long addr,
2639 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2640 struct list_head *uf);
2641extern unsigned long do_mmap(struct file *file, unsigned long addr,
2642 unsigned long len, unsigned long prot, unsigned long flags,
2643 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2644extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2645 struct list_head *uf, bool downgrade);
2646extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2647 struct list_head *uf);
2648extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2649
2650#ifdef CONFIG_MMU
2651extern int __mm_populate(unsigned long addr, unsigned long len,
2652 int ignore_errors);
2653static inline void mm_populate(unsigned long addr, unsigned long len)
2654{
2655 /* Ignore errors */
2656 (void) __mm_populate(addr, len, 1);
2657}
2658#else
2659static inline void mm_populate(unsigned long addr, unsigned long len) {}
2660#endif
2661
2662/* These take the mm semaphore themselves */
2663extern int __must_check vm_brk(unsigned long, unsigned long);
2664extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2665extern int vm_munmap(unsigned long, size_t);
2666extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2667 unsigned long, unsigned long,
2668 unsigned long, unsigned long);
2669
2670struct vm_unmapped_area_info {
2671#define VM_UNMAPPED_AREA_TOPDOWN 1
2672 unsigned long flags;
2673 unsigned long length;
2674 unsigned long low_limit;
2675 unsigned long high_limit;
2676 unsigned long align_mask;
2677 unsigned long align_offset;
2678};
2679
2680extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2681
2682/* truncate.c */
2683extern void truncate_inode_pages(struct address_space *, loff_t);
2684extern void truncate_inode_pages_range(struct address_space *,
2685 loff_t lstart, loff_t lend);
2686extern void truncate_inode_pages_final(struct address_space *);
2687
2688/* generic vm_area_ops exported for stackable file systems */
2689extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2690extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2691 pgoff_t start_pgoff, pgoff_t end_pgoff);
2692extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2693
2694extern unsigned long stack_guard_gap;
2695/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2696extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2697
2698/* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2699extern int expand_downwards(struct vm_area_struct *vma,
2700 unsigned long address);
2701#if VM_GROWSUP
2702extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2703#else
2704 #define expand_upwards(vma, address) (0)
2705#endif
2706
2707/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2708extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2709extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2710 struct vm_area_struct **pprev);
2711
2712/**
2713 * find_vma_intersection() - Look up the first VMA which intersects the interval
2714 * @mm: The process address space.
2715 * @start_addr: The inclusive start user address.
2716 * @end_addr: The exclusive end user address.
2717 *
2718 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2719 * start_addr < end_addr.
2720 */
2721static inline
2722struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2723 unsigned long start_addr,
2724 unsigned long end_addr)
2725{
2726 struct vm_area_struct *vma = find_vma(mm, start_addr);
2727
2728 if (vma && end_addr <= vma->vm_start)
2729 vma = NULL;
2730 return vma;
2731}
2732
2733/**
2734 * vma_lookup() - Find a VMA at a specific address
2735 * @mm: The process address space.
2736 * @addr: The user address.
2737 *
2738 * Return: The vm_area_struct at the given address, %NULL otherwise.
2739 */
2740static inline
2741struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2742{
2743 struct vm_area_struct *vma = find_vma(mm, addr);
2744
2745 if (vma && addr < vma->vm_start)
2746 vma = NULL;
2747
2748 return vma;
2749}
2750
2751static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2752{
2753 unsigned long vm_start = vma->vm_start;
2754
2755 if (vma->vm_flags & VM_GROWSDOWN) {
2756 vm_start -= stack_guard_gap;
2757 if (vm_start > vma->vm_start)
2758 vm_start = 0;
2759 }
2760 return vm_start;
2761}
2762
2763static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2764{
2765 unsigned long vm_end = vma->vm_end;
2766
2767 if (vma->vm_flags & VM_GROWSUP) {
2768 vm_end += stack_guard_gap;
2769 if (vm_end < vma->vm_end)
2770 vm_end = -PAGE_SIZE;
2771 }
2772 return vm_end;
2773}
2774
2775static inline unsigned long vma_pages(struct vm_area_struct *vma)
2776{
2777 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2778}
2779
2780/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2781static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2782 unsigned long vm_start, unsigned long vm_end)
2783{
2784 struct vm_area_struct *vma = find_vma(mm, vm_start);
2785
2786 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2787 vma = NULL;
2788
2789 return vma;
2790}
2791
2792static inline bool range_in_vma(struct vm_area_struct *vma,
2793 unsigned long start, unsigned long end)
2794{
2795 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2796}
2797
2798#ifdef CONFIG_MMU
2799pgprot_t vm_get_page_prot(unsigned long vm_flags);
2800void vma_set_page_prot(struct vm_area_struct *vma);
2801#else
2802static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2803{
2804 return __pgprot(0);
2805}
2806static inline void vma_set_page_prot(struct vm_area_struct *vma)
2807{
2808 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2809}
2810#endif
2811
2812void vma_set_file(struct vm_area_struct *vma, struct file *file);
2813
2814#ifdef CONFIG_NUMA_BALANCING
2815unsigned long change_prot_numa(struct vm_area_struct *vma,
2816 unsigned long start, unsigned long end);
2817#endif
2818
2819struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2820int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2821 unsigned long pfn, unsigned long size, pgprot_t);
2822int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2823 unsigned long pfn, unsigned long size, pgprot_t prot);
2824int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2825int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2826 struct page **pages, unsigned long *num);
2827int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2828 unsigned long num);
2829int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2830 unsigned long num);
2831vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2832 unsigned long pfn);
2833vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2834 unsigned long pfn, pgprot_t pgprot);
2835vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2836 pfn_t pfn);
2837vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2838 pfn_t pfn, pgprot_t pgprot);
2839vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2840 unsigned long addr, pfn_t pfn);
2841int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2842
2843static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2844 unsigned long addr, struct page *page)
2845{
2846 int err = vm_insert_page(vma, addr, page);
2847
2848 if (err == -ENOMEM)
2849 return VM_FAULT_OOM;
2850 if (err < 0 && err != -EBUSY)
2851 return VM_FAULT_SIGBUS;
2852
2853 return VM_FAULT_NOPAGE;
2854}
2855
2856#ifndef io_remap_pfn_range
2857static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2858 unsigned long addr, unsigned long pfn,
2859 unsigned long size, pgprot_t prot)
2860{
2861 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2862}
2863#endif
2864
2865static inline vm_fault_t vmf_error(int err)
2866{
2867 if (err == -ENOMEM)
2868 return VM_FAULT_OOM;
2869 return VM_FAULT_SIGBUS;
2870}
2871
2872struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2873 unsigned int foll_flags);
2874
2875#define FOLL_WRITE 0x01 /* check pte is writable */
2876#define FOLL_TOUCH 0x02 /* mark page accessed */
2877#define FOLL_GET 0x04 /* do get_page on page */
2878#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2879#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2880#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2881 * and return without waiting upon it */
2882#define FOLL_NOFAULT 0x80 /* do not fault in pages */
2883#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2884#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2885#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2886#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2887#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2888#define FOLL_COW 0x4000 /* internal GUP flag */
2889#define FOLL_ANON 0x8000 /* don't do file mappings */
2890#define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2891#define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2892#define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2893#define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2894
2895/*
2896 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2897 * other. Here is what they mean, and how to use them:
2898 *
2899 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2900 * period _often_ under userspace control. This is in contrast to
2901 * iov_iter_get_pages(), whose usages are transient.
2902 *
2903 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2904 * lifetime enforced by the filesystem and we need guarantees that longterm
2905 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2906 * the filesystem. Ideas for this coordination include revoking the longterm
2907 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2908 * added after the problem with filesystems was found FS DAX VMAs are
2909 * specifically failed. Filesystem pages are still subject to bugs and use of
2910 * FOLL_LONGTERM should be avoided on those pages.
2911 *
2912 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2913 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2914 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2915 * is due to an incompatibility with the FS DAX check and
2916 * FAULT_FLAG_ALLOW_RETRY.
2917 *
2918 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2919 * that region. And so, CMA attempts to migrate the page before pinning, when
2920 * FOLL_LONGTERM is specified.
2921 *
2922 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2923 * but an additional pin counting system) will be invoked. This is intended for
2924 * anything that gets a page reference and then touches page data (for example,
2925 * Direct IO). This lets the filesystem know that some non-file-system entity is
2926 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2927 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2928 * a call to unpin_user_page().
2929 *
2930 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2931 * and separate refcounting mechanisms, however, and that means that each has
2932 * its own acquire and release mechanisms:
2933 *
2934 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2935 *
2936 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2937 *
2938 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2939 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2940 * calls applied to them, and that's perfectly OK. This is a constraint on the
2941 * callers, not on the pages.)
2942 *
2943 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2944 * directly by the caller. That's in order to help avoid mismatches when
2945 * releasing pages: get_user_pages*() pages must be released via put_page(),
2946 * while pin_user_pages*() pages must be released via unpin_user_page().
2947 *
2948 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2949 */
2950
2951static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2952{
2953 if (vm_fault & VM_FAULT_OOM)
2954 return -ENOMEM;
2955 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2956 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2957 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2958 return -EFAULT;
2959 return 0;
2960}
2961
2962/*
2963 * Indicates for which pages that are write-protected in the page table,
2964 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
2965 * GUP pin will remain consistent with the pages mapped into the page tables
2966 * of the MM.
2967 *
2968 * Temporary unmapping of PageAnonExclusive() pages or clearing of
2969 * PageAnonExclusive() has to protect against concurrent GUP:
2970 * * Ordinary GUP: Using the PT lock
2971 * * GUP-fast and fork(): mm->write_protect_seq
2972 * * GUP-fast and KSM or temporary unmapping (swap, migration):
2973 * clear/invalidate+flush of the page table entry
2974 *
2975 * Must be called with the (sub)page that's actually referenced via the
2976 * page table entry, which might not necessarily be the head page for a
2977 * PTE-mapped THP.
2978 */
2979static inline bool gup_must_unshare(unsigned int flags, struct page *page)
2980{
2981 /*
2982 * FOLL_WRITE is implicitly handled correctly as the page table entry
2983 * has to be writable -- and if it references (part of) an anonymous
2984 * folio, that part is required to be marked exclusive.
2985 */
2986 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
2987 return false;
2988 /*
2989 * Note: PageAnon(page) is stable until the page is actually getting
2990 * freed.
2991 */
2992 if (!PageAnon(page))
2993 return false;
2994 /*
2995 * Note that PageKsm() pages cannot be exclusive, and consequently,
2996 * cannot get pinned.
2997 */
2998 return !PageAnonExclusive(page);
2999}
3000
3001typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3002extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3003 unsigned long size, pte_fn_t fn, void *data);
3004extern int apply_to_existing_page_range(struct mm_struct *mm,
3005 unsigned long address, unsigned long size,
3006 pte_fn_t fn, void *data);
3007
3008extern void init_mem_debugging_and_hardening(void);
3009#ifdef CONFIG_PAGE_POISONING
3010extern void __kernel_poison_pages(struct page *page, int numpages);
3011extern void __kernel_unpoison_pages(struct page *page, int numpages);
3012extern bool _page_poisoning_enabled_early;
3013DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3014static inline bool page_poisoning_enabled(void)
3015{
3016 return _page_poisoning_enabled_early;
3017}
3018/*
3019 * For use in fast paths after init_mem_debugging() has run, or when a
3020 * false negative result is not harmful when called too early.
3021 */
3022static inline bool page_poisoning_enabled_static(void)
3023{
3024 return static_branch_unlikely(&_page_poisoning_enabled);
3025}
3026static inline void kernel_poison_pages(struct page *page, int numpages)
3027{
3028 if (page_poisoning_enabled_static())
3029 __kernel_poison_pages(page, numpages);
3030}
3031static inline void kernel_unpoison_pages(struct page *page, int numpages)
3032{
3033 if (page_poisoning_enabled_static())
3034 __kernel_unpoison_pages(page, numpages);
3035}
3036#else
3037static inline bool page_poisoning_enabled(void) { return false; }
3038static inline bool page_poisoning_enabled_static(void) { return false; }
3039static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3040static inline void kernel_poison_pages(struct page *page, int numpages) { }
3041static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3042#endif
3043
3044DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3045static inline bool want_init_on_alloc(gfp_t flags)
3046{
3047 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3048 &init_on_alloc))
3049 return true;
3050 return flags & __GFP_ZERO;
3051}
3052
3053DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3054static inline bool want_init_on_free(void)
3055{
3056 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3057 &init_on_free);
3058}
3059
3060extern bool _debug_pagealloc_enabled_early;
3061DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3062
3063static inline bool debug_pagealloc_enabled(void)
3064{
3065 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3066 _debug_pagealloc_enabled_early;
3067}
3068
3069/*
3070 * For use in fast paths after init_debug_pagealloc() has run, or when a
3071 * false negative result is not harmful when called too early.
3072 */
3073static inline bool debug_pagealloc_enabled_static(void)
3074{
3075 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3076 return false;
3077
3078 return static_branch_unlikely(&_debug_pagealloc_enabled);
3079}
3080
3081#ifdef CONFIG_DEBUG_PAGEALLOC
3082/*
3083 * To support DEBUG_PAGEALLOC architecture must ensure that
3084 * __kernel_map_pages() never fails
3085 */
3086extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3087
3088static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3089{
3090 if (debug_pagealloc_enabled_static())
3091 __kernel_map_pages(page, numpages, 1);
3092}
3093
3094static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3095{
3096 if (debug_pagealloc_enabled_static())
3097 __kernel_map_pages(page, numpages, 0);
3098}
3099#else /* CONFIG_DEBUG_PAGEALLOC */
3100static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3101static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3102#endif /* CONFIG_DEBUG_PAGEALLOC */
3103
3104#ifdef __HAVE_ARCH_GATE_AREA
3105extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3106extern int in_gate_area_no_mm(unsigned long addr);
3107extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3108#else
3109static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3110{
3111 return NULL;
3112}
3113static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3114static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3115{
3116 return 0;
3117}
3118#endif /* __HAVE_ARCH_GATE_AREA */
3119
3120extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3121
3122#ifdef CONFIG_SYSCTL
3123extern int sysctl_drop_caches;
3124int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3125 loff_t *);
3126#endif
3127
3128void drop_slab(void);
3129
3130#ifndef CONFIG_MMU
3131#define randomize_va_space 0
3132#else
3133extern int randomize_va_space;
3134#endif
3135
3136const char * arch_vma_name(struct vm_area_struct *vma);
3137#ifdef CONFIG_MMU
3138void print_vma_addr(char *prefix, unsigned long rip);
3139#else
3140static inline void print_vma_addr(char *prefix, unsigned long rip)
3141{
3142}
3143#endif
3144
3145void *sparse_buffer_alloc(unsigned long size);
3146struct page * __populate_section_memmap(unsigned long pfn,
3147 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3148 struct dev_pagemap *pgmap);
3149pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3150p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3151pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3152pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3153pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3154 struct vmem_altmap *altmap, struct page *reuse);
3155void *vmemmap_alloc_block(unsigned long size, int node);
3156struct vmem_altmap;
3157void *vmemmap_alloc_block_buf(unsigned long size, int node,
3158 struct vmem_altmap *altmap);
3159void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3160int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3161 int node, struct vmem_altmap *altmap);
3162int vmemmap_populate(unsigned long start, unsigned long end, int node,
3163 struct vmem_altmap *altmap);
3164void vmemmap_populate_print_last(void);
3165#ifdef CONFIG_MEMORY_HOTPLUG
3166void vmemmap_free(unsigned long start, unsigned long end,
3167 struct vmem_altmap *altmap);
3168#endif
3169void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3170 unsigned long nr_pages);
3171
3172enum mf_flags {
3173 MF_COUNT_INCREASED = 1 << 0,
3174 MF_ACTION_REQUIRED = 1 << 1,
3175 MF_MUST_KILL = 1 << 2,
3176 MF_SOFT_OFFLINE = 1 << 3,
3177 MF_UNPOISON = 1 << 4,
3178 MF_SW_SIMULATED = 1 << 5,
3179 MF_NO_RETRY = 1 << 6,
3180};
3181int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3182 unsigned long count, int mf_flags);
3183extern int memory_failure(unsigned long pfn, int flags);
3184extern void memory_failure_queue(unsigned long pfn, int flags);
3185extern void memory_failure_queue_kick(int cpu);
3186extern int unpoison_memory(unsigned long pfn);
3187extern int sysctl_memory_failure_early_kill;
3188extern int sysctl_memory_failure_recovery;
3189extern void shake_page(struct page *p);
3190extern atomic_long_t num_poisoned_pages __read_mostly;
3191extern int soft_offline_page(unsigned long pfn, int flags);
3192#ifdef CONFIG_MEMORY_FAILURE
3193extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags);
3194#else
3195static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags)
3196{
3197 return 0;
3198}
3199#endif
3200
3201#ifndef arch_memory_failure
3202static inline int arch_memory_failure(unsigned long pfn, int flags)
3203{
3204 return -ENXIO;
3205}
3206#endif
3207
3208#ifndef arch_is_platform_page
3209static inline bool arch_is_platform_page(u64 paddr)
3210{
3211 return false;
3212}
3213#endif
3214
3215/*
3216 * Error handlers for various types of pages.
3217 */
3218enum mf_result {
3219 MF_IGNORED, /* Error: cannot be handled */
3220 MF_FAILED, /* Error: handling failed */
3221 MF_DELAYED, /* Will be handled later */
3222 MF_RECOVERED, /* Successfully recovered */
3223};
3224
3225enum mf_action_page_type {
3226 MF_MSG_KERNEL,
3227 MF_MSG_KERNEL_HIGH_ORDER,
3228 MF_MSG_SLAB,
3229 MF_MSG_DIFFERENT_COMPOUND,
3230 MF_MSG_HUGE,
3231 MF_MSG_FREE_HUGE,
3232 MF_MSG_UNMAP_FAILED,
3233 MF_MSG_DIRTY_SWAPCACHE,
3234 MF_MSG_CLEAN_SWAPCACHE,
3235 MF_MSG_DIRTY_MLOCKED_LRU,
3236 MF_MSG_CLEAN_MLOCKED_LRU,
3237 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3238 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3239 MF_MSG_DIRTY_LRU,
3240 MF_MSG_CLEAN_LRU,
3241 MF_MSG_TRUNCATED_LRU,
3242 MF_MSG_BUDDY,
3243 MF_MSG_DAX,
3244 MF_MSG_UNSPLIT_THP,
3245 MF_MSG_UNKNOWN,
3246};
3247
3248#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3249extern void clear_huge_page(struct page *page,
3250 unsigned long addr_hint,
3251 unsigned int pages_per_huge_page);
3252extern void copy_user_huge_page(struct page *dst, struct page *src,
3253 unsigned long addr_hint,
3254 struct vm_area_struct *vma,
3255 unsigned int pages_per_huge_page);
3256extern long copy_huge_page_from_user(struct page *dst_page,
3257 const void __user *usr_src,
3258 unsigned int pages_per_huge_page,
3259 bool allow_pagefault);
3260
3261/**
3262 * vma_is_special_huge - Are transhuge page-table entries considered special?
3263 * @vma: Pointer to the struct vm_area_struct to consider
3264 *
3265 * Whether transhuge page-table entries are considered "special" following
3266 * the definition in vm_normal_page().
3267 *
3268 * Return: true if transhuge page-table entries should be considered special,
3269 * false otherwise.
3270 */
3271static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3272{
3273 return vma_is_dax(vma) || (vma->vm_file &&
3274 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3275}
3276
3277#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3278
3279#ifdef CONFIG_DEBUG_PAGEALLOC
3280extern unsigned int _debug_guardpage_minorder;
3281DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3282
3283static inline unsigned int debug_guardpage_minorder(void)
3284{
3285 return _debug_guardpage_minorder;
3286}
3287
3288static inline bool debug_guardpage_enabled(void)
3289{
3290 return static_branch_unlikely(&_debug_guardpage_enabled);
3291}
3292
3293static inline bool page_is_guard(struct page *page)
3294{
3295 if (!debug_guardpage_enabled())
3296 return false;
3297
3298 return PageGuard(page);
3299}
3300#else
3301static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3302static inline bool debug_guardpage_enabled(void) { return false; }
3303static inline bool page_is_guard(struct page *page) { return false; }
3304#endif /* CONFIG_DEBUG_PAGEALLOC */
3305
3306#if MAX_NUMNODES > 1
3307void __init setup_nr_node_ids(void);
3308#else
3309static inline void setup_nr_node_ids(void) {}
3310#endif
3311
3312extern int memcmp_pages(struct page *page1, struct page *page2);
3313
3314static inline int pages_identical(struct page *page1, struct page *page2)
3315{
3316 return !memcmp_pages(page1, page2);
3317}
3318
3319#ifdef CONFIG_MAPPING_DIRTY_HELPERS
3320unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3321 pgoff_t first_index, pgoff_t nr,
3322 pgoff_t bitmap_pgoff,
3323 unsigned long *bitmap,
3324 pgoff_t *start,
3325 pgoff_t *end);
3326
3327unsigned long wp_shared_mapping_range(struct address_space *mapping,
3328 pgoff_t first_index, pgoff_t nr);
3329#endif
3330
3331extern int sysctl_nr_trim_pages;
3332
3333#ifdef CONFIG_PRINTK
3334void mem_dump_obj(void *object);
3335#else
3336static inline void mem_dump_obj(void *object) {}
3337#endif
3338
3339/**
3340 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3341 * @seals: the seals to check
3342 * @vma: the vma to operate on
3343 *
3344 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3345 * the vma flags. Return 0 if check pass, or <0 for errors.
3346 */
3347static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3348{
3349 if (seals & F_SEAL_FUTURE_WRITE) {
3350 /*
3351 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3352 * "future write" seal active.
3353 */
3354 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3355 return -EPERM;
3356
3357 /*
3358 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3359 * MAP_SHARED and read-only, take care to not allow mprotect to
3360 * revert protections on such mappings. Do this only for shared
3361 * mappings. For private mappings, don't need to mask
3362 * VM_MAYWRITE as we still want them to be COW-writable.
3363 */
3364 if (vma->vm_flags & VM_SHARED)
3365 vma->vm_flags &= ~(VM_MAYWRITE);
3366 }
3367
3368 return 0;
3369}
3370
3371#ifdef CONFIG_ANON_VMA_NAME
3372int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3373 unsigned long len_in,
3374 struct anon_vma_name *anon_name);
3375#else
3376static inline int
3377madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3378 unsigned long len_in, struct anon_vma_name *anon_name) {
3379 return 0;
3380}
3381#endif
3382
3383/*
3384 * Whether to drop the pte markers, for example, the uffd-wp information for
3385 * file-backed memory. This should only be specified when we will completely
3386 * drop the page in the mm, either by truncation or unmapping of the vma. By
3387 * default, the flag is not set.
3388 */
3389#define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
3390
3391#endif /* _LINUX_MM_H */
3392

source code of linux/include/linux/mm.h