1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
6 #include <linux/mmdebug.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>
35 struct anon_vma_chain;
39 extern int sysctl_page_lock_unfairness;
41 void mm_core_init(void);
42 void init_mm_internals(void);
44 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
45 extern unsigned long max_mapnr;
47 static inline void set_max_mapnr(unsigned long limit)
52 static inline void set_max_mapnr(unsigned long limit) { }
55 extern atomic_long_t _totalram_pages;
56 static inline unsigned long totalram_pages(void)
58 return (unsigned long)atomic_long_read(&_totalram_pages);
61 static inline void totalram_pages_inc(void)
63 atomic_long_inc(&_totalram_pages);
66 static inline void totalram_pages_dec(void)
68 atomic_long_dec(&_totalram_pages);
71 static inline void totalram_pages_add(long count)
73 atomic_long_add(count, &_totalram_pages);
76 extern void * high_memory;
77 extern int page_cluster;
78 extern const int page_cluster_max;
81 extern int sysctl_legacy_va_layout;
83 #define sysctl_legacy_va_layout 0
86 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
87 extern const int mmap_rnd_bits_min;
88 extern const int mmap_rnd_bits_max;
89 extern int mmap_rnd_bits __read_mostly;
91 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
92 extern const int mmap_rnd_compat_bits_min;
93 extern const int mmap_rnd_compat_bits_max;
94 extern int mmap_rnd_compat_bits __read_mostly;
98 #include <asm/processor.h>
101 * Architectures that support memory tagging (assigning tags to memory regions,
102 * embedding these tags into addresses that point to these memory regions, and
103 * checking that the memory and the pointer tags match on memory accesses)
104 * redefine this macro to strip tags from pointers.
105 * It's defined as noop for architectures that don't support memory tagging.
107 #ifndef untagged_addr
108 #define untagged_addr(addr) (addr)
112 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
116 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
120 #define lm_alias(x) __va(__pa_symbol(x))
124 * To prevent common memory management code establishing
125 * a zero page mapping on a read fault.
126 * This macro should be defined within <asm/pgtable.h>.
127 * s390 does this to prevent multiplexing of hardware bits
128 * related to the physical page in case of virtualization.
130 #ifndef mm_forbids_zeropage
131 #define mm_forbids_zeropage(X) (0)
135 * On some architectures it is expensive to call memset() for small sizes.
136 * If an architecture decides to implement their own version of
137 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
138 * define their own version of this macro in <asm/pgtable.h>
140 #if BITS_PER_LONG == 64
141 /* This function must be updated when the size of struct page grows above 96
142 * or reduces below 56. The idea that compiler optimizes out switch()
143 * statement, and only leaves move/store instructions. Also the compiler can
144 * combine write statements if they are both assignments and can be reordered,
145 * this can result in several of the writes here being dropped.
147 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
148 static inline void __mm_zero_struct_page(struct page *page)
150 unsigned long *_pp = (void *)page;
152 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
153 BUILD_BUG_ON(sizeof(struct page) & 7);
154 BUILD_BUG_ON(sizeof(struct page) < 56);
155 BUILD_BUG_ON(sizeof(struct page) > 96);
157 switch (sizeof(struct page)) {
184 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
188 * Default maximum number of active map areas, this limits the number of vmas
189 * per mm struct. Users can overwrite this number by sysctl but there is a
192 * When a program's coredump is generated as ELF format, a section is created
193 * per a vma. In ELF, the number of sections is represented in unsigned short.
194 * This means the number of sections should be smaller than 65535 at coredump.
195 * Because the kernel adds some informative sections to a image of program at
196 * generating coredump, we need some margin. The number of extra sections is
197 * 1-3 now and depends on arch. We use "5" as safe margin, here.
199 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
200 * not a hard limit any more. Although some userspace tools can be surprised by
203 #define MAPCOUNT_ELF_CORE_MARGIN (5)
204 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
206 extern int sysctl_max_map_count;
208 extern unsigned long sysctl_user_reserve_kbytes;
209 extern unsigned long sysctl_admin_reserve_kbytes;
211 extern int sysctl_overcommit_memory;
212 extern int sysctl_overcommit_ratio;
213 extern unsigned long sysctl_overcommit_kbytes;
215 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
217 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
219 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
222 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
223 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
224 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
226 #define nth_page(page,n) ((page) + (n))
227 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
230 /* to align the pointer to the (next) page boundary */
231 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
233 /* to align the pointer to the (prev) page boundary */
234 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
236 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
237 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
239 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
240 static inline struct folio *lru_to_folio(struct list_head *head)
242 return list_entry((head)->prev, struct folio, lru);
245 void setup_initial_init_mm(void *start_code, void *end_code,
246 void *end_data, void *brk);
249 * Linux kernel virtual memory manager primitives.
250 * The idea being to have a "virtual" mm in the same way
251 * we have a virtual fs - giving a cleaner interface to the
252 * mm details, and allowing different kinds of memory mappings
253 * (from shared memory to executable loading to arbitrary
257 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
258 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
259 void vm_area_free(struct vm_area_struct *);
262 extern struct rb_root nommu_region_tree;
263 extern struct rw_semaphore nommu_region_sem;
265 extern unsigned int kobjsize(const void *objp);
269 * vm_flags in vm_area_struct, see mm_types.h.
270 * When changing, update also include/trace/events/mmflags.h
272 #define VM_NONE 0x00000000
274 #define VM_READ 0x00000001 /* currently active flags */
275 #define VM_WRITE 0x00000002
276 #define VM_EXEC 0x00000004
277 #define VM_SHARED 0x00000008
279 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
280 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
281 #define VM_MAYWRITE 0x00000020
282 #define VM_MAYEXEC 0x00000040
283 #define VM_MAYSHARE 0x00000080
285 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
287 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
288 #else /* CONFIG_MMU */
289 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
290 #define VM_UFFD_MISSING 0
291 #endif /* CONFIG_MMU */
292 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
293 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
295 #define VM_LOCKED 0x00002000
296 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
298 /* Used by sys_madvise() */
299 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
300 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
302 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
303 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
304 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
305 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
306 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
307 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
308 #define VM_SYNC 0x00800000 /* Synchronous page faults */
309 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
310 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
311 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
313 #ifdef CONFIG_MEM_SOFT_DIRTY
314 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
316 # define VM_SOFTDIRTY 0
319 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
320 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
321 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
322 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
324 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
325 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
326 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
327 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
328 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
329 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
330 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
331 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
332 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
333 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
334 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
335 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
337 #ifdef CONFIG_ARCH_HAS_PKEYS
338 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
339 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
340 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
341 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
342 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
344 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
346 # define VM_PKEY_BIT4 0
348 #endif /* CONFIG_ARCH_HAS_PKEYS */
350 #if defined(CONFIG_X86)
351 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
352 #elif defined(CONFIG_PPC)
353 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
354 #elif defined(CONFIG_PARISC)
355 # define VM_GROWSUP VM_ARCH_1
356 #elif defined(CONFIG_IA64)
357 # define VM_GROWSUP VM_ARCH_1
358 #elif defined(CONFIG_SPARC64)
359 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
360 # define VM_ARCH_CLEAR VM_SPARC_ADI
361 #elif defined(CONFIG_ARM64)
362 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
363 # define VM_ARCH_CLEAR VM_ARM64_BTI
364 #elif !defined(CONFIG_MMU)
365 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
368 #if defined(CONFIG_ARM64_MTE)
369 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
370 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
372 # define VM_MTE VM_NONE
373 # define VM_MTE_ALLOWED VM_NONE
377 # define VM_GROWSUP VM_NONE
380 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
381 # define VM_UFFD_MINOR_BIT 37
382 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
383 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
384 # define VM_UFFD_MINOR VM_NONE
385 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
387 /* Bits set in the VMA until the stack is in its final location */
388 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
390 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
392 /* Common data flag combinations */
393 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
394 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
395 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
396 VM_MAYWRITE | VM_MAYEXEC)
397 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
398 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
400 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
401 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
404 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
405 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
408 #ifdef CONFIG_STACK_GROWSUP
409 #define VM_STACK VM_GROWSUP
411 #define VM_STACK VM_GROWSDOWN
414 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
416 /* VMA basic access permission flags */
417 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
421 * Special vmas that are non-mergable, non-mlock()able.
423 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
425 /* This mask prevents VMA from being scanned with khugepaged */
426 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
428 /* This mask defines which mm->def_flags a process can inherit its parent */
429 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
431 /* This mask represents all the VMA flag bits used by mlock */
432 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
434 /* Arch-specific flags to clear when updating VM flags on protection change */
435 #ifndef VM_ARCH_CLEAR
436 # define VM_ARCH_CLEAR VM_NONE
438 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
441 * mapping from the currently active vm_flags protection bits (the
442 * low four bits) to a page protection mask..
446 * The default fault flags that should be used by most of the
447 * arch-specific page fault handlers.
449 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
450 FAULT_FLAG_KILLABLE | \
451 FAULT_FLAG_INTERRUPTIBLE)
454 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
455 * @flags: Fault flags.
457 * This is mostly used for places where we want to try to avoid taking
458 * the mmap_lock for too long a time when waiting for another condition
459 * to change, in which case we can try to be polite to release the
460 * mmap_lock in the first round to avoid potential starvation of other
461 * processes that would also want the mmap_lock.
463 * Return: true if the page fault allows retry and this is the first
464 * attempt of the fault handling; false otherwise.
466 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
468 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
469 (!(flags & FAULT_FLAG_TRIED));
472 #define FAULT_FLAG_TRACE \
473 { FAULT_FLAG_WRITE, "WRITE" }, \
474 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
475 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
476 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
477 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
478 { FAULT_FLAG_TRIED, "TRIED" }, \
479 { FAULT_FLAG_USER, "USER" }, \
480 { FAULT_FLAG_REMOTE, "REMOTE" }, \
481 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
482 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
485 * vm_fault is filled by the pagefault handler and passed to the vma's
486 * ->fault function. The vma's ->fault is responsible for returning a bitmask
487 * of VM_FAULT_xxx flags that give details about how the fault was handled.
489 * MM layer fills up gfp_mask for page allocations but fault handler might
490 * alter it if its implementation requires a different allocation context.
492 * pgoff should be used in favour of virtual_address, if possible.
496 struct vm_area_struct *vma; /* Target VMA */
497 gfp_t gfp_mask; /* gfp mask to be used for allocations */
498 pgoff_t pgoff; /* Logical page offset based on vma */
499 unsigned long address; /* Faulting virtual address - masked */
500 unsigned long real_address; /* Faulting virtual address - unmasked */
502 enum fault_flag flags; /* FAULT_FLAG_xxx flags
503 * XXX: should really be 'const' */
504 pmd_t *pmd; /* Pointer to pmd entry matching
506 pud_t *pud; /* Pointer to pud entry matching
510 pte_t orig_pte; /* Value of PTE at the time of fault */
511 pmd_t orig_pmd; /* Value of PMD at the time of fault,
512 * used by PMD fault only.
516 struct page *cow_page; /* Page handler may use for COW fault */
517 struct page *page; /* ->fault handlers should return a
518 * page here, unless VM_FAULT_NOPAGE
519 * is set (which is also implied by
522 /* These three entries are valid only while holding ptl lock */
523 pte_t *pte; /* Pointer to pte entry matching
524 * the 'address'. NULL if the page
525 * table hasn't been allocated.
527 spinlock_t *ptl; /* Page table lock.
528 * Protects pte page table if 'pte'
529 * is not NULL, otherwise pmd.
531 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
532 * vm_ops->map_pages() sets up a page
533 * table from atomic context.
534 * do_fault_around() pre-allocates
535 * page table to avoid allocation from
540 /* page entry size for vm->huge_fault() */
541 enum page_entry_size {
548 * These are the virtual MM functions - opening of an area, closing and
549 * unmapping it (needed to keep files on disk up-to-date etc), pointer
550 * to the functions called when a no-page or a wp-page exception occurs.
552 struct vm_operations_struct {
553 void (*open)(struct vm_area_struct * area);
555 * @close: Called when the VMA is being removed from the MM.
556 * Context: User context. May sleep. Caller holds mmap_lock.
558 void (*close)(struct vm_area_struct * area);
559 /* Called any time before splitting to check if it's allowed */
560 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
561 int (*mremap)(struct vm_area_struct *area);
563 * Called by mprotect() to make driver-specific permission
564 * checks before mprotect() is finalised. The VMA must not
565 * be modified. Returns 0 if mprotect() can proceed.
567 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
568 unsigned long end, unsigned long newflags);
569 vm_fault_t (*fault)(struct vm_fault *vmf);
570 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
571 enum page_entry_size pe_size);
572 vm_fault_t (*map_pages)(struct vm_fault *vmf,
573 pgoff_t start_pgoff, pgoff_t end_pgoff);
574 unsigned long (*pagesize)(struct vm_area_struct * area);
576 /* notification that a previously read-only page is about to become
577 * writable, if an error is returned it will cause a SIGBUS */
578 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
580 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
581 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
583 /* called by access_process_vm when get_user_pages() fails, typically
584 * for use by special VMAs. See also generic_access_phys() for a generic
585 * implementation useful for any iomem mapping.
587 int (*access)(struct vm_area_struct *vma, unsigned long addr,
588 void *buf, int len, int write);
590 /* Called by the /proc/PID/maps code to ask the vma whether it
591 * has a special name. Returning non-NULL will also cause this
592 * vma to be dumped unconditionally. */
593 const char *(*name)(struct vm_area_struct *vma);
597 * set_policy() op must add a reference to any non-NULL @new mempolicy
598 * to hold the policy upon return. Caller should pass NULL @new to
599 * remove a policy and fall back to surrounding context--i.e. do not
600 * install a MPOL_DEFAULT policy, nor the task or system default
603 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
606 * get_policy() op must add reference [mpol_get()] to any policy at
607 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
608 * in mm/mempolicy.c will do this automatically.
609 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
610 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
611 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
612 * must return NULL--i.e., do not "fallback" to task or system default
615 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
619 * Called by vm_normal_page() for special PTEs to find the
620 * page for @addr. This is useful if the default behavior
621 * (using pte_page()) would not find the correct page.
623 struct page *(*find_special_page)(struct vm_area_struct *vma,
627 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
629 static const struct vm_operations_struct dummy_vm_ops = {};
631 memset(vma, 0, sizeof(*vma));
633 vma->vm_ops = &dummy_vm_ops;
634 INIT_LIST_HEAD(&vma->anon_vma_chain);
637 /* Use when VMA is not part of the VMA tree and needs no locking */
638 static inline void vm_flags_init(struct vm_area_struct *vma,
641 ACCESS_PRIVATE(vma, __vm_flags) = flags;
644 /* Use when VMA is part of the VMA tree and modifications need coordination */
645 static inline void vm_flags_reset(struct vm_area_struct *vma,
648 mmap_assert_write_locked(vma->vm_mm);
649 vm_flags_init(vma, flags);
652 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
655 mmap_assert_write_locked(vma->vm_mm);
656 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
659 static inline void vm_flags_set(struct vm_area_struct *vma,
662 mmap_assert_write_locked(vma->vm_mm);
663 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
666 static inline void vm_flags_clear(struct vm_area_struct *vma,
669 mmap_assert_write_locked(vma->vm_mm);
670 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
674 * Use only if VMA is not part of the VMA tree or has no other users and
675 * therefore needs no locking.
677 static inline void __vm_flags_mod(struct vm_area_struct *vma,
678 vm_flags_t set, vm_flags_t clear)
680 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
684 * Use only when the order of set/clear operations is unimportant, otherwise
685 * use vm_flags_{set|clear} explicitly.
687 static inline void vm_flags_mod(struct vm_area_struct *vma,
688 vm_flags_t set, vm_flags_t clear)
690 mmap_assert_write_locked(vma->vm_mm);
691 __vm_flags_mod(vma, set, clear);
694 static inline void vma_set_anonymous(struct vm_area_struct *vma)
699 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
704 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
706 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
711 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
712 VM_STACK_INCOMPLETE_SETUP)
718 static inline bool vma_is_foreign(struct vm_area_struct *vma)
723 if (current->mm != vma->vm_mm)
729 static inline bool vma_is_accessible(struct vm_area_struct *vma)
731 return vma->vm_flags & VM_ACCESS_FLAGS;
735 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
737 return mas_find(&vmi->mas, max - 1);
740 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
743 * Uses mas_find() to get the first VMA when the iterator starts.
744 * Calling mas_next() could skip the first entry.
746 return mas_find(&vmi->mas, ULONG_MAX);
749 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
751 return mas_prev(&vmi->mas, 0);
754 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
756 return vmi->mas.index;
759 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
761 return vmi->mas.last + 1;
763 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
766 return mas_expected_entries(&vmi->mas, count);
769 /* Free any unused preallocations */
770 static inline void vma_iter_free(struct vma_iterator *vmi)
772 mas_destroy(&vmi->mas);
775 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
776 struct vm_area_struct *vma)
778 vmi->mas.index = vma->vm_start;
779 vmi->mas.last = vma->vm_end - 1;
780 mas_store(&vmi->mas, vma);
781 if (unlikely(mas_is_err(&vmi->mas)))
787 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
789 mas_pause(&vmi->mas);
792 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
794 mas_set(&vmi->mas, addr);
797 #define for_each_vma(__vmi, __vma) \
798 while (((__vma) = vma_next(&(__vmi))) != NULL)
800 /* The MM code likes to work with exclusive end addresses */
801 #define for_each_vma_range(__vmi, __vma, __end) \
802 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
806 * The vma_is_shmem is not inline because it is used only by slow
807 * paths in userfault.
809 bool vma_is_shmem(struct vm_area_struct *vma);
810 bool vma_is_anon_shmem(struct vm_area_struct *vma);
812 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
813 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
816 int vma_is_stack_for_current(struct vm_area_struct *vma);
818 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
819 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
825 * compound_order() can be called without holding a reference, which means
826 * that niceties like page_folio() don't work. These callers should be
827 * prepared to handle wild return values. For example, PG_head may be
828 * set before _folio_order is initialised, or this may be a tail page.
829 * See compaction.c for some good examples.
831 static inline unsigned int compound_order(struct page *page)
833 struct folio *folio = (struct folio *)page;
835 if (!test_bit(PG_head, &folio->flags))
837 return folio->_folio_order;
841 * folio_order - The allocation order of a folio.
844 * A folio is composed of 2^order pages. See get_order() for the definition
847 * Return: The order of the folio.
849 static inline unsigned int folio_order(struct folio *folio)
851 if (!folio_test_large(folio))
853 return folio->_folio_order;
856 #include <linux/huge_mm.h>
859 * Methods to modify the page usage count.
861 * What counts for a page usage:
862 * - cache mapping (page->mapping)
863 * - private data (page->private)
864 * - page mapped in a task's page tables, each mapping
865 * is counted separately
867 * Also, many kernel routines increase the page count before a critical
868 * routine so they can be sure the page doesn't go away from under them.
872 * Drop a ref, return true if the refcount fell to zero (the page has no users)
874 static inline int put_page_testzero(struct page *page)
876 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
877 return page_ref_dec_and_test(page);
880 static inline int folio_put_testzero(struct folio *folio)
882 return put_page_testzero(&folio->page);
886 * Try to grab a ref unless the page has a refcount of zero, return false if
888 * This can be called when MMU is off so it must not access
889 * any of the virtual mappings.
891 static inline bool get_page_unless_zero(struct page *page)
893 return page_ref_add_unless(page, 1, 0);
896 static inline struct folio *folio_get_nontail_page(struct page *page)
898 if (unlikely(!get_page_unless_zero(page)))
900 return (struct folio *)page;
903 extern int page_is_ram(unsigned long pfn);
911 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
914 /* Support for virtually mapped pages */
915 struct page *vmalloc_to_page(const void *addr);
916 unsigned long vmalloc_to_pfn(const void *addr);
919 * Determine if an address is within the vmalloc range
921 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
922 * is no special casing required.
925 #ifndef is_ioremap_addr
926 #define is_ioremap_addr(x) is_vmalloc_addr(x)
930 extern bool is_vmalloc_addr(const void *x);
931 extern int is_vmalloc_or_module_addr(const void *x);
933 static inline bool is_vmalloc_addr(const void *x)
937 static inline int is_vmalloc_or_module_addr(const void *x)
944 * How many times the entire folio is mapped as a single unit (eg by a
945 * PMD or PUD entry). This is probably not what you want, except for
946 * debugging purposes - it does not include PTE-mapped sub-pages; look
947 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
949 static inline int folio_entire_mapcount(struct folio *folio)
951 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
952 return atomic_read(&folio->_entire_mapcount) + 1;
956 * The atomic page->_mapcount, starts from -1: so that transitions
957 * both from it and to it can be tracked, using atomic_inc_and_test
958 * and atomic_add_negative(-1).
960 static inline void page_mapcount_reset(struct page *page)
962 atomic_set(&(page)->_mapcount, -1);
966 * page_mapcount() - Number of times this precise page is mapped.
969 * The number of times this page is mapped. If this page is part of
970 * a large folio, it includes the number of times this page is mapped
971 * as part of that folio.
973 * The result is undefined for pages which cannot be mapped into userspace.
974 * For example SLAB or special types of pages. See function page_has_type().
975 * They use this field in struct page differently.
977 static inline int page_mapcount(struct page *page)
979 int mapcount = atomic_read(&page->_mapcount) + 1;
981 if (unlikely(PageCompound(page)))
982 mapcount += folio_entire_mapcount(page_folio(page));
987 int folio_total_mapcount(struct folio *folio);
990 * folio_mapcount() - Calculate the number of mappings of this folio.
993 * A large folio tracks both how many times the entire folio is mapped,
994 * and how many times each individual page in the folio is mapped.
995 * This function calculates the total number of times the folio is
998 * Return: The number of times this folio is mapped.
1000 static inline int folio_mapcount(struct folio *folio)
1002 if (likely(!folio_test_large(folio)))
1003 return atomic_read(&folio->_mapcount) + 1;
1004 return folio_total_mapcount(folio);
1007 static inline int total_mapcount(struct page *page)
1009 if (likely(!PageCompound(page)))
1010 return atomic_read(&page->_mapcount) + 1;
1011 return folio_total_mapcount(page_folio(page));
1014 static inline bool folio_large_is_mapped(struct folio *folio)
1017 * Reading _entire_mapcount below could be omitted if hugetlb
1018 * participated in incrementing nr_pages_mapped when compound mapped.
1020 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1021 atomic_read(&folio->_entire_mapcount) >= 0;
1025 * folio_mapped - Is this folio mapped into userspace?
1026 * @folio: The folio.
1028 * Return: True if any page in this folio is referenced by user page tables.
1030 static inline bool folio_mapped(struct folio *folio)
1032 if (likely(!folio_test_large(folio)))
1033 return atomic_read(&folio->_mapcount) >= 0;
1034 return folio_large_is_mapped(folio);
1038 * Return true if this page is mapped into pagetables.
1039 * For compound page it returns true if any sub-page of compound page is mapped,
1040 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1042 static inline bool page_mapped(struct page *page)
1044 if (likely(!PageCompound(page)))
1045 return atomic_read(&page->_mapcount) >= 0;
1046 return folio_large_is_mapped(page_folio(page));
1049 static inline struct page *virt_to_head_page(const void *x)
1051 struct page *page = virt_to_page(x);
1053 return compound_head(page);
1056 static inline struct folio *virt_to_folio(const void *x)
1058 struct page *page = virt_to_page(x);
1060 return page_folio(page);
1063 void __folio_put(struct folio *folio);
1065 void put_pages_list(struct list_head *pages);
1067 void split_page(struct page *page, unsigned int order);
1068 void folio_copy(struct folio *dst, struct folio *src);
1070 unsigned long nr_free_buffer_pages(void);
1073 * Compound pages have a destructor function. Provide a
1074 * prototype for that function and accessor functions.
1075 * These are _only_ valid on the head of a compound page.
1077 typedef void compound_page_dtor(struct page *);
1079 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1080 enum compound_dtor_id {
1083 #ifdef CONFIG_HUGETLB_PAGE
1086 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1087 TRANSHUGE_PAGE_DTOR,
1091 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1093 static inline void set_compound_page_dtor(struct page *page,
1094 enum compound_dtor_id compound_dtor)
1096 struct folio *folio = (struct folio *)page;
1098 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1099 VM_BUG_ON_PAGE(!PageHead(page), page);
1100 folio->_folio_dtor = compound_dtor;
1103 static inline void folio_set_compound_dtor(struct folio *folio,
1104 enum compound_dtor_id compound_dtor)
1106 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1107 folio->_folio_dtor = compound_dtor;
1110 void destroy_large_folio(struct folio *folio);
1112 static inline void set_compound_order(struct page *page, unsigned int order)
1114 struct folio *folio = (struct folio *)page;
1116 folio->_folio_order = order;
1118 folio->_folio_nr_pages = 1U << order;
1122 /* Returns the number of bytes in this potentially compound page. */
1123 static inline unsigned long page_size(struct page *page)
1125 return PAGE_SIZE << compound_order(page);
1128 /* Returns the number of bits needed for the number of bytes in a page */
1129 static inline unsigned int page_shift(struct page *page)
1131 return PAGE_SHIFT + compound_order(page);
1135 * thp_order - Order of a transparent huge page.
1136 * @page: Head page of a transparent huge page.
1138 static inline unsigned int thp_order(struct page *page)
1140 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1141 return compound_order(page);
1145 * thp_size - Size of a transparent huge page.
1146 * @page: Head page of a transparent huge page.
1148 * Return: Number of bytes in this page.
1150 static inline unsigned long thp_size(struct page *page)
1152 return PAGE_SIZE << thp_order(page);
1155 void free_compound_page(struct page *page);
1159 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1160 * servicing faults for write access. In the normal case, do always want
1161 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1162 * that do not have writing enabled, when used by access_process_vm.
1164 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1166 if (likely(vma->vm_flags & VM_WRITE))
1167 pte = pte_mkwrite(pte);
1171 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1172 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1174 vm_fault_t finish_fault(struct vm_fault *vmf);
1175 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1179 * Multiple processes may "see" the same page. E.g. for untouched
1180 * mappings of /dev/null, all processes see the same page full of
1181 * zeroes, and text pages of executables and shared libraries have
1182 * only one copy in memory, at most, normally.
1184 * For the non-reserved pages, page_count(page) denotes a reference count.
1185 * page_count() == 0 means the page is free. page->lru is then used for
1186 * freelist management in the buddy allocator.
1187 * page_count() > 0 means the page has been allocated.
1189 * Pages are allocated by the slab allocator in order to provide memory
1190 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1191 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1192 * unless a particular usage is carefully commented. (the responsibility of
1193 * freeing the kmalloc memory is the caller's, of course).
1195 * A page may be used by anyone else who does a __get_free_page().
1196 * In this case, page_count still tracks the references, and should only
1197 * be used through the normal accessor functions. The top bits of page->flags
1198 * and page->virtual store page management information, but all other fields
1199 * are unused and could be used privately, carefully. The management of this
1200 * page is the responsibility of the one who allocated it, and those who have
1201 * subsequently been given references to it.
1203 * The other pages (we may call them "pagecache pages") are completely
1204 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1205 * The following discussion applies only to them.
1207 * A pagecache page contains an opaque `private' member, which belongs to the
1208 * page's address_space. Usually, this is the address of a circular list of
1209 * the page's disk buffers. PG_private must be set to tell the VM to call
1210 * into the filesystem to release these pages.
1212 * A page may belong to an inode's memory mapping. In this case, page->mapping
1213 * is the pointer to the inode, and page->index is the file offset of the page,
1214 * in units of PAGE_SIZE.
1216 * If pagecache pages are not associated with an inode, they are said to be
1217 * anonymous pages. These may become associated with the swapcache, and in that
1218 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1220 * In either case (swapcache or inode backed), the pagecache itself holds one
1221 * reference to the page. Setting PG_private should also increment the
1222 * refcount. The each user mapping also has a reference to the page.
1224 * The pagecache pages are stored in a per-mapping radix tree, which is
1225 * rooted at mapping->i_pages, and indexed by offset.
1226 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1227 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1229 * All pagecache pages may be subject to I/O:
1230 * - inode pages may need to be read from disk,
1231 * - inode pages which have been modified and are MAP_SHARED may need
1232 * to be written back to the inode on disk,
1233 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1234 * modified may need to be swapped out to swap space and (later) to be read
1238 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1239 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1241 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1242 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1244 if (!static_branch_unlikely(&devmap_managed_key))
1246 if (!is_zone_device_page(page))
1248 return __put_devmap_managed_page_refs(page, refs);
1250 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1251 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1255 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1257 static inline bool put_devmap_managed_page(struct page *page)
1259 return put_devmap_managed_page_refs(page, 1);
1262 /* 127: arbitrary random number, small enough to assemble well */
1263 #define folio_ref_zero_or_close_to_overflow(folio) \
1264 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1267 * folio_get - Increment the reference count on a folio.
1268 * @folio: The folio.
1270 * Context: May be called in any context, as long as you know that
1271 * you have a refcount on the folio. If you do not already have one,
1272 * folio_try_get() may be the right interface for you to use.
1274 static inline void folio_get(struct folio *folio)
1276 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1277 folio_ref_inc(folio);
1280 static inline void get_page(struct page *page)
1282 folio_get(page_folio(page));
1285 static inline __must_check bool try_get_page(struct page *page)
1287 page = compound_head(page);
1288 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1295 * folio_put - Decrement the reference count on a folio.
1296 * @folio: The folio.
1298 * If the folio's reference count reaches zero, the memory will be
1299 * released back to the page allocator and may be used by another
1300 * allocation immediately. Do not access the memory or the struct folio
1301 * after calling folio_put() unless you can be sure that it wasn't the
1304 * Context: May be called in process or interrupt context, but not in NMI
1305 * context. May be called while holding a spinlock.
1307 static inline void folio_put(struct folio *folio)
1309 if (folio_put_testzero(folio))
1314 * folio_put_refs - Reduce the reference count on a folio.
1315 * @folio: The folio.
1316 * @refs: The amount to subtract from the folio's reference count.
1318 * If the folio's reference count reaches zero, the memory will be
1319 * released back to the page allocator and may be used by another
1320 * allocation immediately. Do not access the memory or the struct folio
1321 * after calling folio_put_refs() unless you can be sure that these weren't
1322 * the last references.
1324 * Context: May be called in process or interrupt context, but not in NMI
1325 * context. May be called while holding a spinlock.
1327 static inline void folio_put_refs(struct folio *folio, int refs)
1329 if (folio_ref_sub_and_test(folio, refs))
1334 * union release_pages_arg - an array of pages or folios
1336 * release_pages() releases a simple array of multiple pages, and
1337 * accepts various different forms of said page array: either
1338 * a regular old boring array of pages, an array of folios, or
1339 * an array of encoded page pointers.
1341 * The transparent union syntax for this kind of "any of these
1342 * argument types" is all kinds of ugly, so look away.
1345 struct page **pages;
1346 struct folio **folios;
1347 struct encoded_page **encoded_pages;
1348 } release_pages_arg __attribute__ ((__transparent_union__));
1350 void release_pages(release_pages_arg, int nr);
1353 * folios_put - Decrement the reference count on an array of folios.
1354 * @folios: The folios.
1355 * @nr: How many folios there are.
1357 * Like folio_put(), but for an array of folios. This is more efficient
1358 * than writing the loop yourself as it will optimise the locks which
1359 * need to be taken if the folios are freed.
1361 * Context: May be called in process or interrupt context, but not in NMI
1362 * context. May be called while holding a spinlock.
1364 static inline void folios_put(struct folio **folios, unsigned int nr)
1366 release_pages(folios, nr);
1369 static inline void put_page(struct page *page)
1371 struct folio *folio = page_folio(page);
1374 * For some devmap managed pages we need to catch refcount transition
1377 if (put_devmap_managed_page(&folio->page))
1383 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1384 * the page's refcount so that two separate items are tracked: the original page
1385 * reference count, and also a new count of how many pin_user_pages() calls were
1386 * made against the page. ("gup-pinned" is another term for the latter).
1388 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1389 * distinct from normal pages. As such, the unpin_user_page() call (and its
1390 * variants) must be used in order to release gup-pinned pages.
1394 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1395 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1396 * simpler, due to the fact that adding an even power of two to the page
1397 * refcount has the effect of using only the upper N bits, for the code that
1398 * counts up using the bias value. This means that the lower bits are left for
1399 * the exclusive use of the original code that increments and decrements by one
1400 * (or at least, by much smaller values than the bias value).
1402 * Of course, once the lower bits overflow into the upper bits (and this is
1403 * OK, because subtraction recovers the original values), then visual inspection
1404 * no longer suffices to directly view the separate counts. However, for normal
1405 * applications that don't have huge page reference counts, this won't be an
1408 * Locking: the lockless algorithm described in folio_try_get_rcu()
1409 * provides safe operation for get_user_pages(), page_mkclean() and
1410 * other calls that race to set up page table entries.
1412 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1414 void unpin_user_page(struct page *page);
1415 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1417 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1419 void unpin_user_pages(struct page **pages, unsigned long npages);
1421 static inline bool is_cow_mapping(vm_flags_t flags)
1423 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1427 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1430 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1431 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1432 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1433 * underlying memory if ptrace is active, so this is only possible if
1434 * ptrace does not apply. Note that there is no mprotect() to upgrade
1435 * write permissions later.
1437 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1441 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1442 #define SECTION_IN_PAGE_FLAGS
1446 * The identification function is mainly used by the buddy allocator for
1447 * determining if two pages could be buddies. We are not really identifying
1448 * the zone since we could be using the section number id if we do not have
1449 * node id available in page flags.
1450 * We only guarantee that it will return the same value for two combinable
1453 static inline int page_zone_id(struct page *page)
1455 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1458 #ifdef NODE_NOT_IN_PAGE_FLAGS
1459 extern int page_to_nid(const struct page *page);
1461 static inline int page_to_nid(const struct page *page)
1463 struct page *p = (struct page *)page;
1465 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1469 static inline int folio_nid(const struct folio *folio)
1471 return page_to_nid(&folio->page);
1474 #ifdef CONFIG_NUMA_BALANCING
1475 /* page access time bits needs to hold at least 4 seconds */
1476 #define PAGE_ACCESS_TIME_MIN_BITS 12
1477 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1478 #define PAGE_ACCESS_TIME_BUCKETS \
1479 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1481 #define PAGE_ACCESS_TIME_BUCKETS 0
1484 #define PAGE_ACCESS_TIME_MASK \
1485 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1487 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1489 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1492 static inline int cpupid_to_pid(int cpupid)
1494 return cpupid & LAST__PID_MASK;
1497 static inline int cpupid_to_cpu(int cpupid)
1499 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1502 static inline int cpupid_to_nid(int cpupid)
1504 return cpu_to_node(cpupid_to_cpu(cpupid));
1507 static inline bool cpupid_pid_unset(int cpupid)
1509 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1512 static inline bool cpupid_cpu_unset(int cpupid)
1514 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1517 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1519 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1522 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1523 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1524 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1526 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1529 static inline int page_cpupid_last(struct page *page)
1531 return page->_last_cpupid;
1533 static inline void page_cpupid_reset_last(struct page *page)
1535 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1538 static inline int page_cpupid_last(struct page *page)
1540 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1543 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1545 static inline void page_cpupid_reset_last(struct page *page)
1547 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1549 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1551 static inline int xchg_page_access_time(struct page *page, int time)
1555 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1556 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1558 #else /* !CONFIG_NUMA_BALANCING */
1559 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1561 return page_to_nid(page); /* XXX */
1564 static inline int xchg_page_access_time(struct page *page, int time)
1569 static inline int page_cpupid_last(struct page *page)
1571 return page_to_nid(page); /* XXX */
1574 static inline int cpupid_to_nid(int cpupid)
1579 static inline int cpupid_to_pid(int cpupid)
1584 static inline int cpupid_to_cpu(int cpupid)
1589 static inline int cpu_pid_to_cpupid(int nid, int pid)
1594 static inline bool cpupid_pid_unset(int cpupid)
1599 static inline void page_cpupid_reset_last(struct page *page)
1603 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1607 #endif /* CONFIG_NUMA_BALANCING */
1609 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1612 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1613 * setting tags for all pages to native kernel tag value 0xff, as the default
1614 * value 0x00 maps to 0xff.
1617 static inline u8 page_kasan_tag(const struct page *page)
1621 if (kasan_enabled()) {
1622 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1629 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1631 unsigned long old_flags, flags;
1633 if (!kasan_enabled())
1637 old_flags = READ_ONCE(page->flags);
1640 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1641 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1642 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1645 static inline void page_kasan_tag_reset(struct page *page)
1647 if (kasan_enabled())
1648 page_kasan_tag_set(page, 0xff);
1651 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1653 static inline u8 page_kasan_tag(const struct page *page)
1658 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1659 static inline void page_kasan_tag_reset(struct page *page) { }
1661 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1663 static inline struct zone *page_zone(const struct page *page)
1665 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1668 static inline pg_data_t *page_pgdat(const struct page *page)
1670 return NODE_DATA(page_to_nid(page));
1673 static inline struct zone *folio_zone(const struct folio *folio)
1675 return page_zone(&folio->page);
1678 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1680 return page_pgdat(&folio->page);
1683 #ifdef SECTION_IN_PAGE_FLAGS
1684 static inline void set_page_section(struct page *page, unsigned long section)
1686 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1687 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1690 static inline unsigned long page_to_section(const struct page *page)
1692 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1697 * folio_pfn - Return the Page Frame Number of a folio.
1698 * @folio: The folio.
1700 * A folio may contain multiple pages. The pages have consecutive
1701 * Page Frame Numbers.
1703 * Return: The Page Frame Number of the first page in the folio.
1705 static inline unsigned long folio_pfn(struct folio *folio)
1707 return page_to_pfn(&folio->page);
1710 static inline struct folio *pfn_folio(unsigned long pfn)
1712 return page_folio(pfn_to_page(pfn));
1716 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1717 * @folio: The folio.
1719 * This function checks if a folio has been pinned via a call to
1720 * a function in the pin_user_pages() family.
1722 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1723 * because it means "definitely not pinned for DMA", but true means "probably
1724 * pinned for DMA, but possibly a false positive due to having at least
1725 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1727 * False positives are OK, because: a) it's unlikely for a folio to
1728 * get that many refcounts, and b) all the callers of this routine are
1729 * expected to be able to deal gracefully with a false positive.
1731 * For large folios, the result will be exactly correct. That's because
1732 * we have more tracking data available: the _pincount field is used
1733 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1735 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1737 * Return: True, if it is likely that the page has been "dma-pinned".
1738 * False, if the page is definitely not dma-pinned.
1740 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1742 if (folio_test_large(folio))
1743 return atomic_read(&folio->_pincount) > 0;
1746 * folio_ref_count() is signed. If that refcount overflows, then
1747 * folio_ref_count() returns a negative value, and callers will avoid
1748 * further incrementing the refcount.
1750 * Here, for that overflow case, use the sign bit to count a little
1751 * bit higher via unsigned math, and thus still get an accurate result.
1753 return ((unsigned int)folio_ref_count(folio)) >=
1754 GUP_PIN_COUNTING_BIAS;
1757 static inline bool page_maybe_dma_pinned(struct page *page)
1759 return folio_maybe_dma_pinned(page_folio(page));
1763 * This should most likely only be called during fork() to see whether we
1764 * should break the cow immediately for an anon page on the src mm.
1766 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1768 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1771 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1773 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1776 return page_maybe_dma_pinned(page);
1779 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1780 #ifdef CONFIG_MIGRATION
1781 static inline bool is_longterm_pinnable_page(struct page *page)
1784 int mt = get_pageblock_migratetype(page);
1786 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1789 /* The zero page may always be pinned */
1790 if (is_zero_pfn(page_to_pfn(page)))
1793 /* Coherent device memory must always allow eviction. */
1794 if (is_device_coherent_page(page))
1797 /* Otherwise, non-movable zone pages can be pinned. */
1798 return !is_zone_movable_page(page);
1801 static inline bool is_longterm_pinnable_page(struct page *page)
1807 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1809 return is_longterm_pinnable_page(&folio->page);
1812 static inline void set_page_zone(struct page *page, enum zone_type zone)
1814 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1815 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1818 static inline void set_page_node(struct page *page, unsigned long node)
1820 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1821 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1824 static inline void set_page_links(struct page *page, enum zone_type zone,
1825 unsigned long node, unsigned long pfn)
1827 set_page_zone(page, zone);
1828 set_page_node(page, node);
1829 #ifdef SECTION_IN_PAGE_FLAGS
1830 set_page_section(page, pfn_to_section_nr(pfn));
1835 * folio_nr_pages - The number of pages in the folio.
1836 * @folio: The folio.
1838 * Return: A positive power of two.
1840 static inline long folio_nr_pages(struct folio *folio)
1842 if (!folio_test_large(folio))
1845 return folio->_folio_nr_pages;
1847 return 1L << folio->_folio_order;
1852 * compound_nr() returns the number of pages in this potentially compound
1853 * page. compound_nr() can be called on a tail page, and is defined to
1854 * return 1 in that case.
1856 static inline unsigned long compound_nr(struct page *page)
1858 struct folio *folio = (struct folio *)page;
1860 if (!test_bit(PG_head, &folio->flags))
1863 return folio->_folio_nr_pages;
1865 return 1L << folio->_folio_order;
1870 * thp_nr_pages - The number of regular pages in this huge page.
1871 * @page: The head page of a huge page.
1873 static inline int thp_nr_pages(struct page *page)
1875 return folio_nr_pages((struct folio *)page);
1879 * folio_next - Move to the next physical folio.
1880 * @folio: The folio we're currently operating on.
1882 * If you have physically contiguous memory which may span more than
1883 * one folio (eg a &struct bio_vec), use this function to move from one
1884 * folio to the next. Do not use it if the memory is only virtually
1885 * contiguous as the folios are almost certainly not adjacent to each
1886 * other. This is the folio equivalent to writing ``page++``.
1888 * Context: We assume that the folios are refcounted and/or locked at a
1889 * higher level and do not adjust the reference counts.
1890 * Return: The next struct folio.
1892 static inline struct folio *folio_next(struct folio *folio)
1894 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1898 * folio_shift - The size of the memory described by this folio.
1899 * @folio: The folio.
1901 * A folio represents a number of bytes which is a power-of-two in size.
1902 * This function tells you which power-of-two the folio is. See also
1903 * folio_size() and folio_order().
1905 * Context: The caller should have a reference on the folio to prevent
1906 * it from being split. It is not necessary for the folio to be locked.
1907 * Return: The base-2 logarithm of the size of this folio.
1909 static inline unsigned int folio_shift(struct folio *folio)
1911 return PAGE_SHIFT + folio_order(folio);
1915 * folio_size - The number of bytes in a folio.
1916 * @folio: The folio.
1918 * Context: The caller should have a reference on the folio to prevent
1919 * it from being split. It is not necessary for the folio to be locked.
1920 * Return: The number of bytes in this folio.
1922 static inline size_t folio_size(struct folio *folio)
1924 return PAGE_SIZE << folio_order(folio);
1928 * folio_estimated_sharers - Estimate the number of sharers of a folio.
1929 * @folio: The folio.
1931 * folio_estimated_sharers() aims to serve as a function to efficiently
1932 * estimate the number of processes sharing a folio. This is done by
1933 * looking at the precise mapcount of the first subpage in the folio, and
1934 * assuming the other subpages are the same. This may not be true for large
1935 * folios. If you want exact mapcounts for exact calculations, look at
1936 * page_mapcount() or folio_total_mapcount().
1938 * Return: The estimated number of processes sharing a folio.
1940 static inline int folio_estimated_sharers(struct folio *folio)
1942 return page_mapcount(folio_page(folio, 0));
1945 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1946 static inline int arch_make_page_accessible(struct page *page)
1952 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1953 static inline int arch_make_folio_accessible(struct folio *folio)
1956 long i, nr = folio_nr_pages(folio);
1958 for (i = 0; i < nr; i++) {
1959 ret = arch_make_page_accessible(folio_page(folio, i));
1969 * Some inline functions in vmstat.h depend on page_zone()
1971 #include <linux/vmstat.h>
1973 static __always_inline void *lowmem_page_address(const struct page *page)
1975 return page_to_virt(page);
1978 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1979 #define HASHED_PAGE_VIRTUAL
1982 #if defined(WANT_PAGE_VIRTUAL)
1983 static inline void *page_address(const struct page *page)
1985 return page->virtual;
1987 static inline void set_page_address(struct page *page, void *address)
1989 page->virtual = address;
1991 #define page_address_init() do { } while(0)
1994 #if defined(HASHED_PAGE_VIRTUAL)
1995 void *page_address(const struct page *page);
1996 void set_page_address(struct page *page, void *virtual);
1997 void page_address_init(void);
2000 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2001 #define page_address(page) lowmem_page_address(page)
2002 #define set_page_address(page, address) do { } while(0)
2003 #define page_address_init() do { } while(0)
2006 static inline void *folio_address(const struct folio *folio)
2008 return page_address(&folio->page);
2011 extern void *page_rmapping(struct page *page);
2012 extern pgoff_t __page_file_index(struct page *page);
2015 * Return the pagecache index of the passed page. Regular pagecache pages
2016 * use ->index whereas swapcache pages use swp_offset(->private)
2018 static inline pgoff_t page_index(struct page *page)
2020 if (unlikely(PageSwapCache(page)))
2021 return __page_file_index(page);
2026 * Return true only if the page has been allocated with
2027 * ALLOC_NO_WATERMARKS and the low watermark was not
2028 * met implying that the system is under some pressure.
2030 static inline bool page_is_pfmemalloc(const struct page *page)
2033 * lru.next has bit 1 set if the page is allocated from the
2034 * pfmemalloc reserves. Callers may simply overwrite it if
2035 * they do not need to preserve that information.
2037 return (uintptr_t)page->lru.next & BIT(1);
2041 * Return true only if the folio has been allocated with
2042 * ALLOC_NO_WATERMARKS and the low watermark was not
2043 * met implying that the system is under some pressure.
2045 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2048 * lru.next has bit 1 set if the page is allocated from the
2049 * pfmemalloc reserves. Callers may simply overwrite it if
2050 * they do not need to preserve that information.
2052 return (uintptr_t)folio->lru.next & BIT(1);
2056 * Only to be called by the page allocator on a freshly allocated
2059 static inline void set_page_pfmemalloc(struct page *page)
2061 page->lru.next = (void *)BIT(1);
2064 static inline void clear_page_pfmemalloc(struct page *page)
2066 page->lru.next = NULL;
2070 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2072 extern void pagefault_out_of_memory(void);
2074 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2075 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2076 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2079 * Flags passed to show_mem() and show_free_areas() to suppress output in
2082 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2084 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2085 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2087 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2091 * Parameter block passed down to zap_pte_range in exceptional cases.
2093 struct zap_details {
2094 struct folio *single_folio; /* Locked folio to be unmapped */
2095 bool even_cows; /* Zap COWed private pages too? */
2096 zap_flags_t zap_flags; /* Extra flags for zapping */
2100 * Whether to drop the pte markers, for example, the uffd-wp information for
2101 * file-backed memory. This should only be specified when we will completely
2102 * drop the page in the mm, either by truncation or unmapping of the vma. By
2103 * default, the flag is not set.
2105 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2106 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2107 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2109 #ifdef CONFIG_SCHED_MM_CID
2110 void sched_mm_cid_before_execve(struct task_struct *t);
2111 void sched_mm_cid_after_execve(struct task_struct *t);
2112 void sched_mm_cid_fork(struct task_struct *t);
2113 void sched_mm_cid_exit_signals(struct task_struct *t);
2114 static inline int task_mm_cid(struct task_struct *t)
2119 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2120 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2121 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2122 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2123 static inline int task_mm_cid(struct task_struct *t)
2126 * Use the processor id as a fall-back when the mm cid feature is
2127 * disabled. This provides functional per-cpu data structure accesses
2128 * in user-space, althrough it won't provide the memory usage benefits.
2130 return raw_smp_processor_id();
2135 extern bool can_do_mlock(void);
2137 static inline bool can_do_mlock(void) { return false; }
2139 extern int user_shm_lock(size_t, struct ucounts *);
2140 extern void user_shm_unlock(size_t, struct ucounts *);
2142 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2144 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2146 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2149 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2150 unsigned long size);
2151 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2152 unsigned long size, struct zap_details *details);
2153 static inline void zap_vma_pages(struct vm_area_struct *vma)
2155 zap_page_range_single(vma, vma->vm_start,
2156 vma->vm_end - vma->vm_start, NULL);
2158 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2159 struct vm_area_struct *start_vma, unsigned long start,
2160 unsigned long end, bool mm_wr_locked);
2162 struct mmu_notifier_range;
2164 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2165 unsigned long end, unsigned long floor, unsigned long ceiling);
2167 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2168 int follow_pte(struct mm_struct *mm, unsigned long address,
2169 pte_t **ptepp, spinlock_t **ptlp);
2170 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2171 unsigned long *pfn);
2172 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2173 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2174 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2175 void *buf, int len, int write);
2177 extern void truncate_pagecache(struct inode *inode, loff_t new);
2178 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2179 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2180 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2181 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2184 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2185 unsigned long address, unsigned int flags,
2186 struct pt_regs *regs);
2187 extern int fixup_user_fault(struct mm_struct *mm,
2188 unsigned long address, unsigned int fault_flags,
2190 void unmap_mapping_pages(struct address_space *mapping,
2191 pgoff_t start, pgoff_t nr, bool even_cows);
2192 void unmap_mapping_range(struct address_space *mapping,
2193 loff_t const holebegin, loff_t const holelen, int even_cows);
2195 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2196 unsigned long address, unsigned int flags,
2197 struct pt_regs *regs)
2199 /* should never happen if there's no MMU */
2201 return VM_FAULT_SIGBUS;
2203 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2204 unsigned int fault_flags, bool *unlocked)
2206 /* should never happen if there's no MMU */
2210 static inline void unmap_mapping_pages(struct address_space *mapping,
2211 pgoff_t start, pgoff_t nr, bool even_cows) { }
2212 static inline void unmap_mapping_range(struct address_space *mapping,
2213 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2216 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2217 loff_t const holebegin, loff_t const holelen)
2219 unmap_mapping_range(mapping, holebegin, holelen, 0);
2222 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2223 void *buf, int len, unsigned int gup_flags);
2224 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2225 void *buf, int len, unsigned int gup_flags);
2226 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2227 void *buf, int len, unsigned int gup_flags);
2229 long get_user_pages_remote(struct mm_struct *mm,
2230 unsigned long start, unsigned long nr_pages,
2231 unsigned int gup_flags, struct page **pages,
2232 struct vm_area_struct **vmas, int *locked);
2233 long pin_user_pages_remote(struct mm_struct *mm,
2234 unsigned long start, unsigned long nr_pages,
2235 unsigned int gup_flags, struct page **pages,
2236 struct vm_area_struct **vmas, int *locked);
2237 long get_user_pages(unsigned long start, unsigned long nr_pages,
2238 unsigned int gup_flags, struct page **pages,
2239 struct vm_area_struct **vmas);
2240 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2241 unsigned int gup_flags, struct page **pages,
2242 struct vm_area_struct **vmas);
2243 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2244 struct page **pages, unsigned int gup_flags);
2245 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2246 struct page **pages, unsigned int gup_flags);
2248 int get_user_pages_fast(unsigned long start, int nr_pages,
2249 unsigned int gup_flags, struct page **pages);
2250 int pin_user_pages_fast(unsigned long start, int nr_pages,
2251 unsigned int gup_flags, struct page **pages);
2253 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2254 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2255 struct task_struct *task, bool bypass_rlim);
2258 struct page *get_dump_page(unsigned long addr);
2260 bool folio_mark_dirty(struct folio *folio);
2261 bool set_page_dirty(struct page *page);
2262 int set_page_dirty_lock(struct page *page);
2264 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2266 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2267 unsigned long old_addr, struct vm_area_struct *new_vma,
2268 unsigned long new_addr, unsigned long len,
2269 bool need_rmap_locks);
2272 * Flags used by change_protection(). For now we make it a bitmap so
2273 * that we can pass in multiple flags just like parameters. However
2274 * for now all the callers are only use one of the flags at the same
2278 * Whether we should manually check if we can map individual PTEs writable,
2279 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2280 * PTEs automatically in a writable mapping.
2282 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2283 /* Whether this protection change is for NUMA hints */
2284 #define MM_CP_PROT_NUMA (1UL << 1)
2285 /* Whether this change is for write protecting */
2286 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2287 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2288 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2289 MM_CP_UFFD_WP_RESOLVE)
2291 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2292 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2295 * We want to check manually if we can change individual PTEs writable
2296 * if we can't do that automatically for all PTEs in a mapping. For
2297 * private mappings, that's always the case when we have write
2298 * permissions as we properly have to handle COW.
2300 if (vma->vm_flags & VM_SHARED)
2301 return vma_wants_writenotify(vma, vma->vm_page_prot);
2302 return !!(vma->vm_flags & VM_WRITE);
2305 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2307 extern long change_protection(struct mmu_gather *tlb,
2308 struct vm_area_struct *vma, unsigned long start,
2309 unsigned long end, unsigned long cp_flags);
2310 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2311 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2312 unsigned long start, unsigned long end, unsigned long newflags);
2315 * doesn't attempt to fault and will return short.
2317 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2318 unsigned int gup_flags, struct page **pages);
2320 static inline bool get_user_page_fast_only(unsigned long addr,
2321 unsigned int gup_flags, struct page **pagep)
2323 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2326 * per-process(per-mm_struct) statistics.
2328 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2330 return percpu_counter_read_positive(&mm->rss_stat[member]);
2333 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2335 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2337 percpu_counter_add(&mm->rss_stat[member], value);
2339 mm_trace_rss_stat(mm, member);
2342 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2344 percpu_counter_inc(&mm->rss_stat[member]);
2346 mm_trace_rss_stat(mm, member);
2349 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2351 percpu_counter_dec(&mm->rss_stat[member]);
2353 mm_trace_rss_stat(mm, member);
2356 /* Optimized variant when page is already known not to be PageAnon */
2357 static inline int mm_counter_file(struct page *page)
2359 if (PageSwapBacked(page))
2360 return MM_SHMEMPAGES;
2361 return MM_FILEPAGES;
2364 static inline int mm_counter(struct page *page)
2367 return MM_ANONPAGES;
2368 return mm_counter_file(page);
2371 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2373 return get_mm_counter(mm, MM_FILEPAGES) +
2374 get_mm_counter(mm, MM_ANONPAGES) +
2375 get_mm_counter(mm, MM_SHMEMPAGES);
2378 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2380 return max(mm->hiwater_rss, get_mm_rss(mm));
2383 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2385 return max(mm->hiwater_vm, mm->total_vm);
2388 static inline void update_hiwater_rss(struct mm_struct *mm)
2390 unsigned long _rss = get_mm_rss(mm);
2392 if ((mm)->hiwater_rss < _rss)
2393 (mm)->hiwater_rss = _rss;
2396 static inline void update_hiwater_vm(struct mm_struct *mm)
2398 if (mm->hiwater_vm < mm->total_vm)
2399 mm->hiwater_vm = mm->total_vm;
2402 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2404 mm->hiwater_rss = get_mm_rss(mm);
2407 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2408 struct mm_struct *mm)
2410 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2412 if (*maxrss < hiwater_rss)
2413 *maxrss = hiwater_rss;
2416 #if defined(SPLIT_RSS_COUNTING)
2417 void sync_mm_rss(struct mm_struct *mm);
2419 static inline void sync_mm_rss(struct mm_struct *mm)
2424 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2425 static inline int pte_special(pte_t pte)
2430 static inline pte_t pte_mkspecial(pte_t pte)
2436 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2437 static inline int pte_devmap(pte_t pte)
2443 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2445 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2449 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2453 #ifdef __PAGETABLE_P4D_FOLDED
2454 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2455 unsigned long address)
2460 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2463 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2464 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2465 unsigned long address)
2469 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2470 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2473 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2475 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2477 if (mm_pud_folded(mm))
2479 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2482 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2484 if (mm_pud_folded(mm))
2486 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2490 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2491 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2492 unsigned long address)
2497 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2498 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2501 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2503 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2505 if (mm_pmd_folded(mm))
2507 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2510 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2512 if (mm_pmd_folded(mm))
2514 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2519 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2521 atomic_long_set(&mm->pgtables_bytes, 0);
2524 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2526 return atomic_long_read(&mm->pgtables_bytes);
2529 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2531 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2534 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2536 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2540 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2541 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2546 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2547 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2550 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2551 int __pte_alloc_kernel(pmd_t *pmd);
2553 #if defined(CONFIG_MMU)
2555 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2556 unsigned long address)
2558 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2559 NULL : p4d_offset(pgd, address);
2562 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2563 unsigned long address)
2565 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2566 NULL : pud_offset(p4d, address);
2569 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2571 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2572 NULL: pmd_offset(pud, address);
2574 #endif /* CONFIG_MMU */
2576 #if USE_SPLIT_PTE_PTLOCKS
2577 #if ALLOC_SPLIT_PTLOCKS
2578 void __init ptlock_cache_init(void);
2579 extern bool ptlock_alloc(struct page *page);
2580 extern void ptlock_free(struct page *page);
2582 static inline spinlock_t *ptlock_ptr(struct page *page)
2586 #else /* ALLOC_SPLIT_PTLOCKS */
2587 static inline void ptlock_cache_init(void)
2591 static inline bool ptlock_alloc(struct page *page)
2596 static inline void ptlock_free(struct page *page)
2600 static inline spinlock_t *ptlock_ptr(struct page *page)
2604 #endif /* ALLOC_SPLIT_PTLOCKS */
2606 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2608 return ptlock_ptr(pmd_page(*pmd));
2611 static inline bool ptlock_init(struct page *page)
2614 * prep_new_page() initialize page->private (and therefore page->ptl)
2615 * with 0. Make sure nobody took it in use in between.
2617 * It can happen if arch try to use slab for page table allocation:
2618 * slab code uses page->slab_cache, which share storage with page->ptl.
2620 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2621 if (!ptlock_alloc(page))
2623 spin_lock_init(ptlock_ptr(page));
2627 #else /* !USE_SPLIT_PTE_PTLOCKS */
2629 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2631 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2633 return &mm->page_table_lock;
2635 static inline void ptlock_cache_init(void) {}
2636 static inline bool ptlock_init(struct page *page) { return true; }
2637 static inline void ptlock_free(struct page *page) {}
2638 #endif /* USE_SPLIT_PTE_PTLOCKS */
2640 static inline bool pgtable_pte_page_ctor(struct page *page)
2642 if (!ptlock_init(page))
2644 __SetPageTable(page);
2645 inc_lruvec_page_state(page, NR_PAGETABLE);
2649 static inline void pgtable_pte_page_dtor(struct page *page)
2652 __ClearPageTable(page);
2653 dec_lruvec_page_state(page, NR_PAGETABLE);
2656 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2658 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2659 pte_t *__pte = pte_offset_map(pmd, address); \
2665 #define pte_unmap_unlock(pte, ptl) do { \
2670 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2672 #define pte_alloc_map(mm, pmd, address) \
2673 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2675 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2676 (pte_alloc(mm, pmd) ? \
2677 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2679 #define pte_alloc_kernel(pmd, address) \
2680 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2681 NULL: pte_offset_kernel(pmd, address))
2683 #if USE_SPLIT_PMD_PTLOCKS
2685 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2687 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2688 return virt_to_page((void *)((unsigned long) pmd & mask));
2691 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2693 return ptlock_ptr(pmd_pgtable_page(pmd));
2696 static inline bool pmd_ptlock_init(struct page *page)
2698 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2699 page->pmd_huge_pte = NULL;
2701 return ptlock_init(page);
2704 static inline void pmd_ptlock_free(struct page *page)
2706 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2707 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2712 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2716 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2718 return &mm->page_table_lock;
2721 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2722 static inline void pmd_ptlock_free(struct page *page) {}
2724 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2728 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2730 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2735 static inline bool pgtable_pmd_page_ctor(struct page *page)
2737 if (!pmd_ptlock_init(page))
2739 __SetPageTable(page);
2740 inc_lruvec_page_state(page, NR_PAGETABLE);
2744 static inline void pgtable_pmd_page_dtor(struct page *page)
2746 pmd_ptlock_free(page);
2747 __ClearPageTable(page);
2748 dec_lruvec_page_state(page, NR_PAGETABLE);
2752 * No scalability reason to split PUD locks yet, but follow the same pattern
2753 * as the PMD locks to make it easier if we decide to. The VM should not be
2754 * considered ready to switch to split PUD locks yet; there may be places
2755 * which need to be converted from page_table_lock.
2757 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2759 return &mm->page_table_lock;
2762 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2764 spinlock_t *ptl = pud_lockptr(mm, pud);
2770 extern void __init pagecache_init(void);
2771 extern void free_initmem(void);
2774 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2775 * into the buddy system. The freed pages will be poisoned with pattern
2776 * "poison" if it's within range [0, UCHAR_MAX].
2777 * Return pages freed into the buddy system.
2779 extern unsigned long free_reserved_area(void *start, void *end,
2780 int poison, const char *s);
2782 extern void adjust_managed_page_count(struct page *page, long count);
2784 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2786 /* Free the reserved page into the buddy system, so it gets managed. */
2787 static inline void free_reserved_page(struct page *page)
2789 ClearPageReserved(page);
2790 init_page_count(page);
2792 adjust_managed_page_count(page, 1);
2794 #define free_highmem_page(page) free_reserved_page(page)
2796 static inline void mark_page_reserved(struct page *page)
2798 SetPageReserved(page);
2799 adjust_managed_page_count(page, -1);
2803 * Default method to free all the __init memory into the buddy system.
2804 * The freed pages will be poisoned with pattern "poison" if it's within
2805 * range [0, UCHAR_MAX].
2806 * Return pages freed into the buddy system.
2808 static inline unsigned long free_initmem_default(int poison)
2810 extern char __init_begin[], __init_end[];
2812 return free_reserved_area(&__init_begin, &__init_end,
2813 poison, "unused kernel image (initmem)");
2816 static inline unsigned long get_num_physpages(void)
2819 unsigned long phys_pages = 0;
2821 for_each_online_node(nid)
2822 phys_pages += node_present_pages(nid);
2828 * Using memblock node mappings, an architecture may initialise its
2829 * zones, allocate the backing mem_map and account for memory holes in an
2830 * architecture independent manner.
2832 * An architecture is expected to register range of page frames backed by
2833 * physical memory with memblock_add[_node]() before calling
2834 * free_area_init() passing in the PFN each zone ends at. At a basic
2835 * usage, an architecture is expected to do something like
2837 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2839 * for_each_valid_physical_page_range()
2840 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2841 * free_area_init(max_zone_pfns);
2843 void free_area_init(unsigned long *max_zone_pfn);
2844 unsigned long node_map_pfn_alignment(void);
2845 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2846 unsigned long end_pfn);
2847 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2848 unsigned long end_pfn);
2849 extern void get_pfn_range_for_nid(unsigned int nid,
2850 unsigned long *start_pfn, unsigned long *end_pfn);
2853 static inline int early_pfn_to_nid(unsigned long pfn)
2858 /* please see mm/page_alloc.c */
2859 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2862 extern void set_dma_reserve(unsigned long new_dma_reserve);
2863 extern void memmap_init_range(unsigned long, int, unsigned long,
2864 unsigned long, unsigned long, enum meminit_context,
2865 struct vmem_altmap *, int migratetype);
2866 extern void setup_per_zone_wmarks(void);
2867 extern void calculate_min_free_kbytes(void);
2868 extern int __meminit init_per_zone_wmark_min(void);
2869 extern void mem_init(void);
2870 extern void __init mmap_init(void);
2872 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2873 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2875 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2877 extern long si_mem_available(void);
2878 extern void si_meminfo(struct sysinfo * val);
2879 extern void si_meminfo_node(struct sysinfo *val, int nid);
2880 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2881 extern unsigned long arch_reserved_kernel_pages(void);
2884 extern __printf(3, 4)
2885 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2887 extern void setup_per_cpu_pageset(void);
2890 extern int min_free_kbytes;
2891 extern int watermark_boost_factor;
2892 extern int watermark_scale_factor;
2893 extern bool arch_has_descending_max_zone_pfns(void);
2896 extern atomic_long_t mmap_pages_allocated;
2897 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2899 /* interval_tree.c */
2900 void vma_interval_tree_insert(struct vm_area_struct *node,
2901 struct rb_root_cached *root);
2902 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2903 struct vm_area_struct *prev,
2904 struct rb_root_cached *root);
2905 void vma_interval_tree_remove(struct vm_area_struct *node,
2906 struct rb_root_cached *root);
2907 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2908 unsigned long start, unsigned long last);
2909 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2910 unsigned long start, unsigned long last);
2912 #define vma_interval_tree_foreach(vma, root, start, last) \
2913 for (vma = vma_interval_tree_iter_first(root, start, last); \
2914 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2916 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2917 struct rb_root_cached *root);
2918 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2919 struct rb_root_cached *root);
2920 struct anon_vma_chain *
2921 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2922 unsigned long start, unsigned long last);
2923 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2924 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2925 #ifdef CONFIG_DEBUG_VM_RB
2926 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2929 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2930 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2931 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2934 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2935 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
2936 unsigned long start, unsigned long end, pgoff_t pgoff,
2937 struct vm_area_struct *next);
2938 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
2939 unsigned long start, unsigned long end, pgoff_t pgoff);
2940 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
2941 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
2942 unsigned long end, unsigned long vm_flags, struct anon_vma *,
2943 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
2944 struct anon_vma_name *);
2945 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2946 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
2947 unsigned long addr, int new_below);
2948 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
2949 unsigned long addr, int new_below);
2950 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2951 extern void unlink_file_vma(struct vm_area_struct *);
2952 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2953 unsigned long addr, unsigned long len, pgoff_t pgoff,
2954 bool *need_rmap_locks);
2955 extern void exit_mmap(struct mm_struct *);
2957 static inline int check_data_rlimit(unsigned long rlim,
2959 unsigned long start,
2960 unsigned long end_data,
2961 unsigned long start_data)
2963 if (rlim < RLIM_INFINITY) {
2964 if (((new - start) + (end_data - start_data)) > rlim)
2971 extern int mm_take_all_locks(struct mm_struct *mm);
2972 extern void mm_drop_all_locks(struct mm_struct *mm);
2974 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2975 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2976 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2977 extern struct file *get_task_exe_file(struct task_struct *task);
2979 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2980 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2982 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2983 const struct vm_special_mapping *sm);
2984 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2985 unsigned long addr, unsigned long len,
2986 unsigned long flags,
2987 const struct vm_special_mapping *spec);
2988 /* This is an obsolete alternative to _install_special_mapping. */
2989 extern int install_special_mapping(struct mm_struct *mm,
2990 unsigned long addr, unsigned long len,
2991 unsigned long flags, struct page **pages);
2993 unsigned long randomize_stack_top(unsigned long stack_top);
2994 unsigned long randomize_page(unsigned long start, unsigned long range);
2996 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2998 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2999 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3000 struct list_head *uf);
3001 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3002 unsigned long len, unsigned long prot, unsigned long flags,
3003 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3004 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3005 unsigned long start, size_t len, struct list_head *uf,
3007 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3008 struct list_head *uf);
3009 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3012 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3013 unsigned long start, unsigned long end,
3014 struct list_head *uf, bool downgrade);
3015 extern int __mm_populate(unsigned long addr, unsigned long len,
3017 static inline void mm_populate(unsigned long addr, unsigned long len)
3020 (void) __mm_populate(addr, len, 1);
3023 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3026 /* These take the mm semaphore themselves */
3027 extern int __must_check vm_brk(unsigned long, unsigned long);
3028 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3029 extern int vm_munmap(unsigned long, size_t);
3030 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3031 unsigned long, unsigned long,
3032 unsigned long, unsigned long);
3034 struct vm_unmapped_area_info {
3035 #define VM_UNMAPPED_AREA_TOPDOWN 1
3036 unsigned long flags;
3037 unsigned long length;
3038 unsigned long low_limit;
3039 unsigned long high_limit;
3040 unsigned long align_mask;
3041 unsigned long align_offset;
3044 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3047 extern void truncate_inode_pages(struct address_space *, loff_t);
3048 extern void truncate_inode_pages_range(struct address_space *,
3049 loff_t lstart, loff_t lend);
3050 extern void truncate_inode_pages_final(struct address_space *);
3052 /* generic vm_area_ops exported for stackable file systems */
3053 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3054 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3055 pgoff_t start_pgoff, pgoff_t end_pgoff);
3056 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3058 extern unsigned long stack_guard_gap;
3059 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3060 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
3062 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3063 extern int expand_downwards(struct vm_area_struct *vma,
3064 unsigned long address);
3066 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
3068 #define expand_upwards(vma, address) (0)
3071 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3072 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3073 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3074 struct vm_area_struct **pprev);
3077 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3078 * NULL if none. Assume start_addr < end_addr.
3080 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3081 unsigned long start_addr, unsigned long end_addr);
3084 * vma_lookup() - Find a VMA at a specific address
3085 * @mm: The process address space.
3086 * @addr: The user address.
3088 * Return: The vm_area_struct at the given address, %NULL otherwise.
3091 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3093 return mtree_load(&mm->mm_mt, addr);
3096 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3098 unsigned long vm_start = vma->vm_start;
3100 if (vma->vm_flags & VM_GROWSDOWN) {
3101 vm_start -= stack_guard_gap;
3102 if (vm_start > vma->vm_start)
3108 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3110 unsigned long vm_end = vma->vm_end;
3112 if (vma->vm_flags & VM_GROWSUP) {
3113 vm_end += stack_guard_gap;
3114 if (vm_end < vma->vm_end)
3115 vm_end = -PAGE_SIZE;
3120 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3122 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3125 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3126 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3127 unsigned long vm_start, unsigned long vm_end)
3129 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3131 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3137 static inline bool range_in_vma(struct vm_area_struct *vma,
3138 unsigned long start, unsigned long end)
3140 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3144 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3145 void vma_set_page_prot(struct vm_area_struct *vma);
3147 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3151 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3153 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3157 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3159 #ifdef CONFIG_NUMA_BALANCING
3160 unsigned long change_prot_numa(struct vm_area_struct *vma,
3161 unsigned long start, unsigned long end);
3164 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3165 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3166 unsigned long pfn, unsigned long size, pgprot_t);
3167 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3168 unsigned long pfn, unsigned long size, pgprot_t prot);
3169 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3170 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3171 struct page **pages, unsigned long *num);
3172 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3174 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3176 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3178 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3179 unsigned long pfn, pgprot_t pgprot);
3180 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3182 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3183 unsigned long addr, pfn_t pfn);
3184 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3186 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3187 unsigned long addr, struct page *page)
3189 int err = vm_insert_page(vma, addr, page);
3192 return VM_FAULT_OOM;
3193 if (err < 0 && err != -EBUSY)
3194 return VM_FAULT_SIGBUS;
3196 return VM_FAULT_NOPAGE;
3199 #ifndef io_remap_pfn_range
3200 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3201 unsigned long addr, unsigned long pfn,
3202 unsigned long size, pgprot_t prot)
3204 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3208 static inline vm_fault_t vmf_error(int err)
3211 return VM_FAULT_OOM;
3212 return VM_FAULT_SIGBUS;
3215 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3216 unsigned int foll_flags);
3218 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3220 if (vm_fault & VM_FAULT_OOM)
3222 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3223 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3224 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3230 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3231 * a (NUMA hinting) fault is required.
3233 static inline bool gup_can_follow_protnone(unsigned int flags)
3236 * FOLL_FORCE has to be able to make progress even if the VMA is
3237 * inaccessible. Further, FOLL_FORCE access usually does not represent
3238 * application behaviour and we should avoid triggering NUMA hinting
3241 return flags & FOLL_FORCE;
3244 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3245 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3246 unsigned long size, pte_fn_t fn, void *data);
3247 extern int apply_to_existing_page_range(struct mm_struct *mm,
3248 unsigned long address, unsigned long size,
3249 pte_fn_t fn, void *data);
3251 #ifdef CONFIG_PAGE_POISONING
3252 extern void __kernel_poison_pages(struct page *page, int numpages);
3253 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3254 extern bool _page_poisoning_enabled_early;
3255 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3256 static inline bool page_poisoning_enabled(void)
3258 return _page_poisoning_enabled_early;
3261 * For use in fast paths after init_mem_debugging() has run, or when a
3262 * false negative result is not harmful when called too early.
3264 static inline bool page_poisoning_enabled_static(void)
3266 return static_branch_unlikely(&_page_poisoning_enabled);
3268 static inline void kernel_poison_pages(struct page *page, int numpages)
3270 if (page_poisoning_enabled_static())
3271 __kernel_poison_pages(page, numpages);
3273 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3275 if (page_poisoning_enabled_static())
3276 __kernel_unpoison_pages(page, numpages);
3279 static inline bool page_poisoning_enabled(void) { return false; }
3280 static inline bool page_poisoning_enabled_static(void) { return false; }
3281 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3282 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3283 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3286 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3287 static inline bool want_init_on_alloc(gfp_t flags)
3289 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3292 return flags & __GFP_ZERO;
3295 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3296 static inline bool want_init_on_free(void)
3298 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3302 extern bool _debug_pagealloc_enabled_early;
3303 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3305 static inline bool debug_pagealloc_enabled(void)
3307 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3308 _debug_pagealloc_enabled_early;
3312 * For use in fast paths after init_debug_pagealloc() has run, or when a
3313 * false negative result is not harmful when called too early.
3315 static inline bool debug_pagealloc_enabled_static(void)
3317 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3320 return static_branch_unlikely(&_debug_pagealloc_enabled);
3323 #ifdef CONFIG_DEBUG_PAGEALLOC
3325 * To support DEBUG_PAGEALLOC architecture must ensure that
3326 * __kernel_map_pages() never fails
3328 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3330 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3332 if (debug_pagealloc_enabled_static())
3333 __kernel_map_pages(page, numpages, 1);
3336 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3338 if (debug_pagealloc_enabled_static())
3339 __kernel_map_pages(page, numpages, 0);
3341 #else /* CONFIG_DEBUG_PAGEALLOC */
3342 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3343 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3344 #endif /* CONFIG_DEBUG_PAGEALLOC */
3346 #ifdef __HAVE_ARCH_GATE_AREA
3347 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3348 extern int in_gate_area_no_mm(unsigned long addr);
3349 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3351 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3355 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3356 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3360 #endif /* __HAVE_ARCH_GATE_AREA */
3362 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3364 #ifdef CONFIG_SYSCTL
3365 extern int sysctl_drop_caches;
3366 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3370 void drop_slab(void);
3373 #define randomize_va_space 0
3375 extern int randomize_va_space;
3378 const char * arch_vma_name(struct vm_area_struct *vma);
3380 void print_vma_addr(char *prefix, unsigned long rip);
3382 static inline void print_vma_addr(char *prefix, unsigned long rip)
3387 void *sparse_buffer_alloc(unsigned long size);
3388 struct page * __populate_section_memmap(unsigned long pfn,
3389 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3390 struct dev_pagemap *pgmap);
3391 void pmd_init(void *addr);
3392 void pud_init(void *addr);
3393 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3394 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3395 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3396 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3397 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3398 struct vmem_altmap *altmap, struct page *reuse);
3399 void *vmemmap_alloc_block(unsigned long size, int node);
3401 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3402 struct vmem_altmap *altmap);
3403 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3404 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3405 unsigned long addr, unsigned long next);
3406 int vmemmap_check_pmd(pmd_t *pmd, int node,
3407 unsigned long addr, unsigned long next);
3408 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3409 int node, struct vmem_altmap *altmap);
3410 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3411 int node, struct vmem_altmap *altmap);
3412 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3413 struct vmem_altmap *altmap);
3414 void vmemmap_populate_print_last(void);
3415 #ifdef CONFIG_MEMORY_HOTPLUG
3416 void vmemmap_free(unsigned long start, unsigned long end,
3417 struct vmem_altmap *altmap);
3419 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3420 unsigned long nr_pages);
3423 MF_COUNT_INCREASED = 1 << 0,
3424 MF_ACTION_REQUIRED = 1 << 1,
3425 MF_MUST_KILL = 1 << 2,
3426 MF_SOFT_OFFLINE = 1 << 3,
3427 MF_UNPOISON = 1 << 4,
3428 MF_SW_SIMULATED = 1 << 5,
3429 MF_NO_RETRY = 1 << 6,
3431 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3432 unsigned long count, int mf_flags);
3433 extern int memory_failure(unsigned long pfn, int flags);
3434 extern void memory_failure_queue_kick(int cpu);
3435 extern int unpoison_memory(unsigned long pfn);
3436 extern int sysctl_memory_failure_early_kill;
3437 extern int sysctl_memory_failure_recovery;
3438 extern void shake_page(struct page *p);
3439 extern atomic_long_t num_poisoned_pages __read_mostly;
3440 extern int soft_offline_page(unsigned long pfn, int flags);
3441 #ifdef CONFIG_MEMORY_FAILURE
3442 extern void memory_failure_queue(unsigned long pfn, int flags);
3443 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3444 bool *migratable_cleared);
3445 void num_poisoned_pages_inc(unsigned long pfn);
3446 void num_poisoned_pages_sub(unsigned long pfn, long i);
3448 static inline void memory_failure_queue(unsigned long pfn, int flags)
3452 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3453 bool *migratable_cleared)
3458 static inline void num_poisoned_pages_inc(unsigned long pfn)
3462 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3467 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3468 extern void memblk_nr_poison_inc(unsigned long pfn);
3469 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3471 static inline void memblk_nr_poison_inc(unsigned long pfn)
3475 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3480 #ifndef arch_memory_failure
3481 static inline int arch_memory_failure(unsigned long pfn, int flags)
3487 #ifndef arch_is_platform_page
3488 static inline bool arch_is_platform_page(u64 paddr)
3495 * Error handlers for various types of pages.
3498 MF_IGNORED, /* Error: cannot be handled */
3499 MF_FAILED, /* Error: handling failed */
3500 MF_DELAYED, /* Will be handled later */
3501 MF_RECOVERED, /* Successfully recovered */
3504 enum mf_action_page_type {
3506 MF_MSG_KERNEL_HIGH_ORDER,
3508 MF_MSG_DIFFERENT_COMPOUND,
3511 MF_MSG_UNMAP_FAILED,
3512 MF_MSG_DIRTY_SWAPCACHE,
3513 MF_MSG_CLEAN_SWAPCACHE,
3514 MF_MSG_DIRTY_MLOCKED_LRU,
3515 MF_MSG_CLEAN_MLOCKED_LRU,
3516 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3517 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3520 MF_MSG_TRUNCATED_LRU,
3528 * Sysfs entries for memory failure handling statistics.
3530 extern const struct attribute_group memory_failure_attr_group;
3532 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3533 extern void clear_huge_page(struct page *page,
3534 unsigned long addr_hint,
3535 unsigned int pages_per_huge_page);
3536 extern void copy_user_huge_page(struct page *dst, struct page *src,
3537 unsigned long addr_hint,
3538 struct vm_area_struct *vma,
3539 unsigned int pages_per_huge_page);
3540 extern long copy_huge_page_from_user(struct page *dst_page,
3541 const void __user *usr_src,
3542 unsigned int pages_per_huge_page,
3543 bool allow_pagefault);
3546 * vma_is_special_huge - Are transhuge page-table entries considered special?
3547 * @vma: Pointer to the struct vm_area_struct to consider
3549 * Whether transhuge page-table entries are considered "special" following
3550 * the definition in vm_normal_page().
3552 * Return: true if transhuge page-table entries should be considered special,
3555 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3557 return vma_is_dax(vma) || (vma->vm_file &&
3558 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3561 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3563 #ifdef CONFIG_DEBUG_PAGEALLOC
3564 extern unsigned int _debug_guardpage_minorder;
3565 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3567 static inline unsigned int debug_guardpage_minorder(void)
3569 return _debug_guardpage_minorder;
3572 static inline bool debug_guardpage_enabled(void)
3574 return static_branch_unlikely(&_debug_guardpage_enabled);
3577 static inline bool page_is_guard(struct page *page)
3579 if (!debug_guardpage_enabled())
3582 return PageGuard(page);
3585 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3586 static inline bool debug_guardpage_enabled(void) { return false; }
3587 static inline bool page_is_guard(struct page *page) { return false; }
3588 #endif /* CONFIG_DEBUG_PAGEALLOC */
3590 #if MAX_NUMNODES > 1
3591 void __init setup_nr_node_ids(void);
3593 static inline void setup_nr_node_ids(void) {}
3596 extern int memcmp_pages(struct page *page1, struct page *page2);
3598 static inline int pages_identical(struct page *page1, struct page *page2)
3600 return !memcmp_pages(page1, page2);
3603 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3604 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3605 pgoff_t first_index, pgoff_t nr,
3606 pgoff_t bitmap_pgoff,
3607 unsigned long *bitmap,
3611 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3612 pgoff_t first_index, pgoff_t nr);
3615 extern int sysctl_nr_trim_pages;
3617 #ifdef CONFIG_PRINTK
3618 void mem_dump_obj(void *object);
3620 static inline void mem_dump_obj(void *object) {}
3624 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3625 * @seals: the seals to check
3626 * @vma: the vma to operate on
3628 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3629 * the vma flags. Return 0 if check pass, or <0 for errors.
3631 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3633 if (seals & F_SEAL_FUTURE_WRITE) {
3635 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3636 * "future write" seal active.
3638 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3642 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3643 * MAP_SHARED and read-only, take care to not allow mprotect to
3644 * revert protections on such mappings. Do this only for shared
3645 * mappings. For private mappings, don't need to mask
3646 * VM_MAYWRITE as we still want them to be COW-writable.
3648 if (vma->vm_flags & VM_SHARED)
3649 vm_flags_clear(vma, VM_MAYWRITE);
3655 #ifdef CONFIG_ANON_VMA_NAME
3656 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3657 unsigned long len_in,
3658 struct anon_vma_name *anon_name);
3661 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3662 unsigned long len_in, struct anon_vma_name *anon_name) {
3667 #endif /* _LINUX_MM_H */