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 init_mm_internals(void);
43 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
44 extern unsigned long max_mapnr;
46 static inline void set_max_mapnr(unsigned long limit)
51 static inline void set_max_mapnr(unsigned long limit) { }
54 extern atomic_long_t _totalram_pages;
55 static inline unsigned long totalram_pages(void)
57 return (unsigned long)atomic_long_read(&_totalram_pages);
60 static inline void totalram_pages_inc(void)
62 atomic_long_inc(&_totalram_pages);
65 static inline void totalram_pages_dec(void)
67 atomic_long_dec(&_totalram_pages);
70 static inline void totalram_pages_add(long count)
72 atomic_long_add(count, &_totalram_pages);
75 extern void * high_memory;
76 extern int page_cluster;
77 extern const int page_cluster_max;
80 extern int sysctl_legacy_va_layout;
82 #define sysctl_legacy_va_layout 0
85 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
86 extern const int mmap_rnd_bits_min;
87 extern const int mmap_rnd_bits_max;
88 extern int mmap_rnd_bits __read_mostly;
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
91 extern const int mmap_rnd_compat_bits_min;
92 extern const int mmap_rnd_compat_bits_max;
93 extern int mmap_rnd_compat_bits __read_mostly;
97 #include <asm/processor.h>
100 * Architectures that support memory tagging (assigning tags to memory regions,
101 * embedding these tags into addresses that point to these memory regions, and
102 * checking that the memory and the pointer tags match on memory accesses)
103 * redefine this macro to strip tags from pointers.
104 * It's defined as noop for architectures that don't support memory tagging.
106 #ifndef untagged_addr
107 #define untagged_addr(addr) (addr)
111 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
115 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 #define lm_alias(x) __va(__pa_symbol(x))
123 * To prevent common memory management code establishing
124 * a zero page mapping on a read fault.
125 * This macro should be defined within <asm/pgtable.h>.
126 * s390 does this to prevent multiplexing of hardware bits
127 * related to the physical page in case of virtualization.
129 #ifndef mm_forbids_zeropage
130 #define mm_forbids_zeropage(X) (0)
134 * On some architectures it is expensive to call memset() for small sizes.
135 * If an architecture decides to implement their own version of
136 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
137 * define their own version of this macro in <asm/pgtable.h>
139 #if BITS_PER_LONG == 64
140 /* This function must be updated when the size of struct page grows above 80
141 * or reduces below 56. The idea that compiler optimizes out switch()
142 * statement, and only leaves move/store instructions. Also the compiler can
143 * combine write statements if they are both assignments and can be reordered,
144 * this can result in several of the writes here being dropped.
146 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
147 static inline void __mm_zero_struct_page(struct page *page)
149 unsigned long *_pp = (void *)page;
151 /* Check that struct page is either 56, 64, 72, or 80 bytes */
152 BUILD_BUG_ON(sizeof(struct page) & 7);
153 BUILD_BUG_ON(sizeof(struct page) < 56);
154 BUILD_BUG_ON(sizeof(struct page) > 80);
156 switch (sizeof(struct page)) {
177 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
181 * Default maximum number of active map areas, this limits the number of vmas
182 * per mm struct. Users can overwrite this number by sysctl but there is a
185 * When a program's coredump is generated as ELF format, a section is created
186 * per a vma. In ELF, the number of sections is represented in unsigned short.
187 * This means the number of sections should be smaller than 65535 at coredump.
188 * Because the kernel adds some informative sections to a image of program at
189 * generating coredump, we need some margin. The number of extra sections is
190 * 1-3 now and depends on arch. We use "5" as safe margin, here.
192 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
193 * not a hard limit any more. Although some userspace tools can be surprised by
196 #define MAPCOUNT_ELF_CORE_MARGIN (5)
197 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
199 extern int sysctl_max_map_count;
201 extern unsigned long sysctl_user_reserve_kbytes;
202 extern unsigned long sysctl_admin_reserve_kbytes;
204 extern int sysctl_overcommit_memory;
205 extern int sysctl_overcommit_ratio;
206 extern unsigned long sysctl_overcommit_kbytes;
208 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
210 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
212 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
215 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
216 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
217 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
219 #define nth_page(page,n) ((page) + (n))
220 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
223 /* to align the pointer to the (next) page boundary */
224 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
226 /* to align the pointer to the (prev) page boundary */
227 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
229 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
230 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
232 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
233 static inline struct folio *lru_to_folio(struct list_head *head)
235 return list_entry((head)->prev, struct folio, lru);
238 void setup_initial_init_mm(void *start_code, void *end_code,
239 void *end_data, void *brk);
242 * Linux kernel virtual memory manager primitives.
243 * The idea being to have a "virtual" mm in the same way
244 * we have a virtual fs - giving a cleaner interface to the
245 * mm details, and allowing different kinds of memory mappings
246 * (from shared memory to executable loading to arbitrary
250 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
251 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
252 void vm_area_free(struct vm_area_struct *);
255 extern struct rb_root nommu_region_tree;
256 extern struct rw_semaphore nommu_region_sem;
258 extern unsigned int kobjsize(const void *objp);
262 * vm_flags in vm_area_struct, see mm_types.h.
263 * When changing, update also include/trace/events/mmflags.h
265 #define VM_NONE 0x00000000
267 #define VM_READ 0x00000001 /* currently active flags */
268 #define VM_WRITE 0x00000002
269 #define VM_EXEC 0x00000004
270 #define VM_SHARED 0x00000008
272 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
273 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
274 #define VM_MAYWRITE 0x00000020
275 #define VM_MAYEXEC 0x00000040
276 #define VM_MAYSHARE 0x00000080
278 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
279 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
280 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
281 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
283 #define VM_LOCKED 0x00002000
284 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
286 /* Used by sys_madvise() */
287 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
288 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
290 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
291 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
292 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
293 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
294 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
295 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
296 #define VM_SYNC 0x00800000 /* Synchronous page faults */
297 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
298 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
299 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
301 #ifdef CONFIG_MEM_SOFT_DIRTY
302 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
304 # define VM_SOFTDIRTY 0
307 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
308 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
309 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
310 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
312 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
313 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
314 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
315 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
316 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
317 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
319 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
320 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
321 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
322 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
323 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
325 #ifdef CONFIG_ARCH_HAS_PKEYS
326 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
327 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
328 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
329 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
330 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
332 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
334 # define VM_PKEY_BIT4 0
336 #endif /* CONFIG_ARCH_HAS_PKEYS */
338 #if defined(CONFIG_X86)
339 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
340 #elif defined(CONFIG_PPC)
341 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
342 #elif defined(CONFIG_PARISC)
343 # define VM_GROWSUP VM_ARCH_1
344 #elif defined(CONFIG_IA64)
345 # define VM_GROWSUP VM_ARCH_1
346 #elif defined(CONFIG_SPARC64)
347 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
348 # define VM_ARCH_CLEAR VM_SPARC_ADI
349 #elif defined(CONFIG_ARM64)
350 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
351 # define VM_ARCH_CLEAR VM_ARM64_BTI
352 #elif !defined(CONFIG_MMU)
353 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
356 #if defined(CONFIG_ARM64_MTE)
357 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
358 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
360 # define VM_MTE VM_NONE
361 # define VM_MTE_ALLOWED VM_NONE
365 # define VM_GROWSUP VM_NONE
368 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
369 # define VM_UFFD_MINOR_BIT 37
370 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
371 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
372 # define VM_UFFD_MINOR VM_NONE
373 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
375 /* Bits set in the VMA until the stack is in its final location */
376 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
378 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
380 /* Common data flag combinations */
381 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
382 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
383 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
384 VM_MAYWRITE | VM_MAYEXEC)
385 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
386 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
388 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
389 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
392 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
393 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
396 #ifdef CONFIG_STACK_GROWSUP
397 #define VM_STACK VM_GROWSUP
399 #define VM_STACK VM_GROWSDOWN
402 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
404 /* VMA basic access permission flags */
405 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
409 * Special vmas that are non-mergable, non-mlock()able.
411 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
413 /* This mask prevents VMA from being scanned with khugepaged */
414 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
416 /* This mask defines which mm->def_flags a process can inherit its parent */
417 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
419 /* This mask is used to clear all the VMA flags used by mlock */
420 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
422 /* Arch-specific flags to clear when updating VM flags on protection change */
423 #ifndef VM_ARCH_CLEAR
424 # define VM_ARCH_CLEAR VM_NONE
426 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
429 * mapping from the currently active vm_flags protection bits (the
430 * low four bits) to a page protection mask..
434 * The default fault flags that should be used by most of the
435 * arch-specific page fault handlers.
437 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
438 FAULT_FLAG_KILLABLE | \
439 FAULT_FLAG_INTERRUPTIBLE)
442 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
443 * @flags: Fault flags.
445 * This is mostly used for places where we want to try to avoid taking
446 * the mmap_lock for too long a time when waiting for another condition
447 * to change, in which case we can try to be polite to release the
448 * mmap_lock in the first round to avoid potential starvation of other
449 * processes that would also want the mmap_lock.
451 * Return: true if the page fault allows retry and this is the first
452 * attempt of the fault handling; false otherwise.
454 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
456 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
457 (!(flags & FAULT_FLAG_TRIED));
460 #define FAULT_FLAG_TRACE \
461 { FAULT_FLAG_WRITE, "WRITE" }, \
462 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
463 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
464 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
465 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
466 { FAULT_FLAG_TRIED, "TRIED" }, \
467 { FAULT_FLAG_USER, "USER" }, \
468 { FAULT_FLAG_REMOTE, "REMOTE" }, \
469 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
470 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
473 * vm_fault is filled by the pagefault handler and passed to the vma's
474 * ->fault function. The vma's ->fault is responsible for returning a bitmask
475 * of VM_FAULT_xxx flags that give details about how the fault was handled.
477 * MM layer fills up gfp_mask for page allocations but fault handler might
478 * alter it if its implementation requires a different allocation context.
480 * pgoff should be used in favour of virtual_address, if possible.
484 struct vm_area_struct *vma; /* Target VMA */
485 gfp_t gfp_mask; /* gfp mask to be used for allocations */
486 pgoff_t pgoff; /* Logical page offset based on vma */
487 unsigned long address; /* Faulting virtual address - masked */
488 unsigned long real_address; /* Faulting virtual address - unmasked */
490 enum fault_flag flags; /* FAULT_FLAG_xxx flags
491 * XXX: should really be 'const' */
492 pmd_t *pmd; /* Pointer to pmd entry matching
494 pud_t *pud; /* Pointer to pud entry matching
498 pte_t orig_pte; /* Value of PTE at the time of fault */
499 pmd_t orig_pmd; /* Value of PMD at the time of fault,
500 * used by PMD fault only.
504 struct page *cow_page; /* Page handler may use for COW fault */
505 struct page *page; /* ->fault handlers should return a
506 * page here, unless VM_FAULT_NOPAGE
507 * is set (which is also implied by
510 /* These three entries are valid only while holding ptl lock */
511 pte_t *pte; /* Pointer to pte entry matching
512 * the 'address'. NULL if the page
513 * table hasn't been allocated.
515 spinlock_t *ptl; /* Page table lock.
516 * Protects pte page table if 'pte'
517 * is not NULL, otherwise pmd.
519 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
520 * vm_ops->map_pages() sets up a page
521 * table from atomic context.
522 * do_fault_around() pre-allocates
523 * page table to avoid allocation from
528 /* page entry size for vm->huge_fault() */
529 enum page_entry_size {
536 * These are the virtual MM functions - opening of an area, closing and
537 * unmapping it (needed to keep files on disk up-to-date etc), pointer
538 * to the functions called when a no-page or a wp-page exception occurs.
540 struct vm_operations_struct {
541 void (*open)(struct vm_area_struct * area);
543 * @close: Called when the VMA is being removed from the MM.
544 * Context: User context. May sleep. Caller holds mmap_lock.
546 void (*close)(struct vm_area_struct * area);
547 /* Called any time before splitting to check if it's allowed */
548 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
549 int (*mremap)(struct vm_area_struct *area);
551 * Called by mprotect() to make driver-specific permission
552 * checks before mprotect() is finalised. The VMA must not
553 * be modified. Returns 0 if mprotect() can proceed.
555 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
556 unsigned long end, unsigned long newflags);
557 vm_fault_t (*fault)(struct vm_fault *vmf);
558 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
559 enum page_entry_size pe_size);
560 vm_fault_t (*map_pages)(struct vm_fault *vmf,
561 pgoff_t start_pgoff, pgoff_t end_pgoff);
562 unsigned long (*pagesize)(struct vm_area_struct * area);
564 /* notification that a previously read-only page is about to become
565 * writable, if an error is returned it will cause a SIGBUS */
566 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
568 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
569 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
571 /* called by access_process_vm when get_user_pages() fails, typically
572 * for use by special VMAs. See also generic_access_phys() for a generic
573 * implementation useful for any iomem mapping.
575 int (*access)(struct vm_area_struct *vma, unsigned long addr,
576 void *buf, int len, int write);
578 /* Called by the /proc/PID/maps code to ask the vma whether it
579 * has a special name. Returning non-NULL will also cause this
580 * vma to be dumped unconditionally. */
581 const char *(*name)(struct vm_area_struct *vma);
585 * set_policy() op must add a reference to any non-NULL @new mempolicy
586 * to hold the policy upon return. Caller should pass NULL @new to
587 * remove a policy and fall back to surrounding context--i.e. do not
588 * install a MPOL_DEFAULT policy, nor the task or system default
591 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
594 * get_policy() op must add reference [mpol_get()] to any policy at
595 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
596 * in mm/mempolicy.c will do this automatically.
597 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
598 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
599 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
600 * must return NULL--i.e., do not "fallback" to task or system default
603 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
607 * Called by vm_normal_page() for special PTEs to find the
608 * page for @addr. This is useful if the default behavior
609 * (using pte_page()) would not find the correct page.
611 struct page *(*find_special_page)(struct vm_area_struct *vma,
615 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
617 static const struct vm_operations_struct dummy_vm_ops = {};
619 memset(vma, 0, sizeof(*vma));
621 vma->vm_ops = &dummy_vm_ops;
622 INIT_LIST_HEAD(&vma->anon_vma_chain);
625 static inline void vma_set_anonymous(struct vm_area_struct *vma)
630 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
635 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
637 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
642 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
643 VM_STACK_INCOMPLETE_SETUP)
649 static inline bool vma_is_foreign(struct vm_area_struct *vma)
654 if (current->mm != vma->vm_mm)
660 static inline bool vma_is_accessible(struct vm_area_struct *vma)
662 return vma->vm_flags & VM_ACCESS_FLAGS;
666 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
668 return mas_find(&vmi->mas, max);
671 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
674 * Uses vma_find() to get the first VMA when the iterator starts.
675 * Calling mas_next() could skip the first entry.
677 return vma_find(vmi, ULONG_MAX);
680 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
682 return mas_prev(&vmi->mas, 0);
685 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
687 return vmi->mas.index;
690 #define for_each_vma(__vmi, __vma) \
691 while (((__vma) = vma_next(&(__vmi))) != NULL)
693 /* The MM code likes to work with exclusive end addresses */
694 #define for_each_vma_range(__vmi, __vma, __end) \
695 while (((__vma) = vma_find(&(__vmi), (__end) - 1)) != NULL)
699 * The vma_is_shmem is not inline because it is used only by slow
700 * paths in userfault.
702 bool vma_is_shmem(struct vm_area_struct *vma);
703 bool vma_is_anon_shmem(struct vm_area_struct *vma);
705 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
706 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
709 int vma_is_stack_for_current(struct vm_area_struct *vma);
711 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
712 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
717 static inline unsigned int compound_order(struct page *page)
721 return page[1].compound_order;
725 * folio_order - The allocation order of a folio.
728 * A folio is composed of 2^order pages. See get_order() for the definition
731 * Return: The order of the folio.
733 static inline unsigned int folio_order(struct folio *folio)
735 if (!folio_test_large(folio))
737 return folio->_folio_order;
740 #include <linux/huge_mm.h>
743 * Methods to modify the page usage count.
745 * What counts for a page usage:
746 * - cache mapping (page->mapping)
747 * - private data (page->private)
748 * - page mapped in a task's page tables, each mapping
749 * is counted separately
751 * Also, many kernel routines increase the page count before a critical
752 * routine so they can be sure the page doesn't go away from under them.
756 * Drop a ref, return true if the refcount fell to zero (the page has no users)
758 static inline int put_page_testzero(struct page *page)
760 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
761 return page_ref_dec_and_test(page);
764 static inline int folio_put_testzero(struct folio *folio)
766 return put_page_testzero(&folio->page);
770 * Try to grab a ref unless the page has a refcount of zero, return false if
772 * This can be called when MMU is off so it must not access
773 * any of the virtual mappings.
775 static inline bool get_page_unless_zero(struct page *page)
777 return page_ref_add_unless(page, 1, 0);
780 extern int page_is_ram(unsigned long pfn);
788 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
791 /* Support for virtually mapped pages */
792 struct page *vmalloc_to_page(const void *addr);
793 unsigned long vmalloc_to_pfn(const void *addr);
796 * Determine if an address is within the vmalloc range
798 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
799 * is no special casing required.
802 #ifndef is_ioremap_addr
803 #define is_ioremap_addr(x) is_vmalloc_addr(x)
807 extern bool is_vmalloc_addr(const void *x);
808 extern int is_vmalloc_or_module_addr(const void *x);
810 static inline bool is_vmalloc_addr(const void *x)
814 static inline int is_vmalloc_or_module_addr(const void *x)
821 * How many times the entire folio is mapped as a single unit (eg by a
822 * PMD or PUD entry). This is probably not what you want, except for
823 * debugging purposes - it does not include PTE-mapped sub-pages; look
824 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
826 static inline int folio_entire_mapcount(struct folio *folio)
828 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
829 return atomic_read(folio_mapcount_ptr(folio)) + 1;
833 * Mapcount of compound page as a whole, does not include mapped sub-pages.
834 * Must be called only on head of compound page.
836 static inline int head_compound_mapcount(struct page *head)
838 return atomic_read(compound_mapcount_ptr(head)) + 1;
842 * If a 16GB hugetlb page were mapped by PTEs of all of its 4kB sub-pages,
843 * its subpages_mapcount would be 0x400000: choose the COMPOUND_MAPPED bit
844 * above that range, instead of 2*(PMD_SIZE/PAGE_SIZE). Hugetlb currently
845 * leaves subpages_mapcount at 0, but avoid surprise if it participates later.
847 #define COMPOUND_MAPPED 0x800000
848 #define SUBPAGES_MAPPED (COMPOUND_MAPPED - 1)
851 * Number of sub-pages mapped by PTE, does not include compound mapcount.
852 * Must be called only on head of compound page.
854 static inline int head_subpages_mapcount(struct page *head)
856 return atomic_read(subpages_mapcount_ptr(head)) & SUBPAGES_MAPPED;
860 * The atomic page->_mapcount, starts from -1: so that transitions
861 * both from it and to it can be tracked, using atomic_inc_and_test
862 * and atomic_add_negative(-1).
864 static inline void page_mapcount_reset(struct page *page)
866 atomic_set(&(page)->_mapcount, -1);
870 * Mapcount of 0-order page; when compound sub-page, includes
871 * compound_mapcount of compound_head of page.
873 * Result is undefined for pages which cannot be mapped into userspace.
874 * For example SLAB or special types of pages. See function page_has_type().
875 * They use this place in struct page differently.
877 static inline int page_mapcount(struct page *page)
879 int mapcount = atomic_read(&page->_mapcount) + 1;
881 if (likely(!PageCompound(page)))
883 page = compound_head(page);
884 return head_compound_mapcount(page) + mapcount;
887 int total_compound_mapcount(struct page *head);
890 * folio_mapcount() - Calculate the number of mappings of this folio.
893 * A large folio tracks both how many times the entire folio is mapped,
894 * and how many times each individual page in the folio is mapped.
895 * This function calculates the total number of times the folio is
898 * Return: The number of times this folio is mapped.
900 static inline int folio_mapcount(struct folio *folio)
902 if (likely(!folio_test_large(folio)))
903 return atomic_read(&folio->_mapcount) + 1;
904 return total_compound_mapcount(&folio->page);
907 static inline int total_mapcount(struct page *page)
909 if (likely(!PageCompound(page)))
910 return atomic_read(&page->_mapcount) + 1;
911 return total_compound_mapcount(compound_head(page));
914 static inline bool folio_large_is_mapped(struct folio *folio)
917 * Reading folio_mapcount_ptr() below could be omitted if hugetlb
918 * participated in incrementing subpages_mapcount when compound mapped.
920 return atomic_read(folio_subpages_mapcount_ptr(folio)) > 0 ||
921 atomic_read(folio_mapcount_ptr(folio)) >= 0;
925 * folio_mapped - Is this folio mapped into userspace?
928 * Return: True if any page in this folio is referenced by user page tables.
930 static inline bool folio_mapped(struct folio *folio)
932 if (likely(!folio_test_large(folio)))
933 return atomic_read(&folio->_mapcount) >= 0;
934 return folio_large_is_mapped(folio);
938 * Return true if this page is mapped into pagetables.
939 * For compound page it returns true if any sub-page of compound page is mapped,
940 * even if this particular sub-page is not itself mapped by any PTE or PMD.
942 static inline bool page_mapped(struct page *page)
944 if (likely(!PageCompound(page)))
945 return atomic_read(&page->_mapcount) >= 0;
946 return folio_large_is_mapped(page_folio(page));
949 static inline struct page *virt_to_head_page(const void *x)
951 struct page *page = virt_to_page(x);
953 return compound_head(page);
956 static inline struct folio *virt_to_folio(const void *x)
958 struct page *page = virt_to_page(x);
960 return page_folio(page);
963 void __folio_put(struct folio *folio);
965 void put_pages_list(struct list_head *pages);
967 void split_page(struct page *page, unsigned int order);
968 void folio_copy(struct folio *dst, struct folio *src);
970 unsigned long nr_free_buffer_pages(void);
973 * Compound pages have a destructor function. Provide a
974 * prototype for that function and accessor functions.
975 * These are _only_ valid on the head of a compound page.
977 typedef void compound_page_dtor(struct page *);
979 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
980 enum compound_dtor_id {
983 #ifdef CONFIG_HUGETLB_PAGE
986 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
991 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
993 static inline void set_compound_page_dtor(struct page *page,
994 enum compound_dtor_id compound_dtor)
996 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
997 page[1].compound_dtor = compound_dtor;
1000 void destroy_large_folio(struct folio *folio);
1002 static inline int head_compound_pincount(struct page *head)
1004 return atomic_read(compound_pincount_ptr(head));
1007 static inline void set_compound_order(struct page *page, unsigned int order)
1009 page[1].compound_order = order;
1011 page[1].compound_nr = 1U << order;
1015 /* Returns the number of pages in this potentially compound page. */
1016 static inline unsigned long compound_nr(struct page *page)
1018 if (!PageHead(page))
1021 return page[1].compound_nr;
1023 return 1UL << compound_order(page);
1027 /* Returns the number of bytes in this potentially compound page. */
1028 static inline unsigned long page_size(struct page *page)
1030 return PAGE_SIZE << compound_order(page);
1033 /* Returns the number of bits needed for the number of bytes in a page */
1034 static inline unsigned int page_shift(struct page *page)
1036 return PAGE_SHIFT + compound_order(page);
1040 * thp_order - Order of a transparent huge page.
1041 * @page: Head page of a transparent huge page.
1043 static inline unsigned int thp_order(struct page *page)
1045 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1046 return compound_order(page);
1050 * thp_nr_pages - The number of regular pages in this huge page.
1051 * @page: The head page of a huge page.
1053 static inline int thp_nr_pages(struct page *page)
1055 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1056 return compound_nr(page);
1060 * thp_size - Size of a transparent huge page.
1061 * @page: Head page of a transparent huge page.
1063 * Return: Number of bytes in this page.
1065 static inline unsigned long thp_size(struct page *page)
1067 return PAGE_SIZE << thp_order(page);
1070 void free_compound_page(struct page *page);
1074 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1075 * servicing faults for write access. In the normal case, do always want
1076 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1077 * that do not have writing enabled, when used by access_process_vm.
1079 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1081 if (likely(vma->vm_flags & VM_WRITE))
1082 pte = pte_mkwrite(pte);
1086 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1087 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1089 vm_fault_t finish_fault(struct vm_fault *vmf);
1090 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1094 * Multiple processes may "see" the same page. E.g. for untouched
1095 * mappings of /dev/null, all processes see the same page full of
1096 * zeroes, and text pages of executables and shared libraries have
1097 * only one copy in memory, at most, normally.
1099 * For the non-reserved pages, page_count(page) denotes a reference count.
1100 * page_count() == 0 means the page is free. page->lru is then used for
1101 * freelist management in the buddy allocator.
1102 * page_count() > 0 means the page has been allocated.
1104 * Pages are allocated by the slab allocator in order to provide memory
1105 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1106 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1107 * unless a particular usage is carefully commented. (the responsibility of
1108 * freeing the kmalloc memory is the caller's, of course).
1110 * A page may be used by anyone else who does a __get_free_page().
1111 * In this case, page_count still tracks the references, and should only
1112 * be used through the normal accessor functions. The top bits of page->flags
1113 * and page->virtual store page management information, but all other fields
1114 * are unused and could be used privately, carefully. The management of this
1115 * page is the responsibility of the one who allocated it, and those who have
1116 * subsequently been given references to it.
1118 * The other pages (we may call them "pagecache pages") are completely
1119 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1120 * The following discussion applies only to them.
1122 * A pagecache page contains an opaque `private' member, which belongs to the
1123 * page's address_space. Usually, this is the address of a circular list of
1124 * the page's disk buffers. PG_private must be set to tell the VM to call
1125 * into the filesystem to release these pages.
1127 * A page may belong to an inode's memory mapping. In this case, page->mapping
1128 * is the pointer to the inode, and page->index is the file offset of the page,
1129 * in units of PAGE_SIZE.
1131 * If pagecache pages are not associated with an inode, they are said to be
1132 * anonymous pages. These may become associated with the swapcache, and in that
1133 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1135 * In either case (swapcache or inode backed), the pagecache itself holds one
1136 * reference to the page. Setting PG_private should also increment the
1137 * refcount. The each user mapping also has a reference to the page.
1139 * The pagecache pages are stored in a per-mapping radix tree, which is
1140 * rooted at mapping->i_pages, and indexed by offset.
1141 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1142 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1144 * All pagecache pages may be subject to I/O:
1145 * - inode pages may need to be read from disk,
1146 * - inode pages which have been modified and are MAP_SHARED may need
1147 * to be written back to the inode on disk,
1148 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1149 * modified may need to be swapped out to swap space and (later) to be read
1153 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1154 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1156 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1157 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1159 if (!static_branch_unlikely(&devmap_managed_key))
1161 if (!is_zone_device_page(page))
1163 return __put_devmap_managed_page_refs(page, refs);
1165 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1166 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1170 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1172 static inline bool put_devmap_managed_page(struct page *page)
1174 return put_devmap_managed_page_refs(page, 1);
1177 /* 127: arbitrary random number, small enough to assemble well */
1178 #define folio_ref_zero_or_close_to_overflow(folio) \
1179 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1182 * folio_get - Increment the reference count on a folio.
1183 * @folio: The folio.
1185 * Context: May be called in any context, as long as you know that
1186 * you have a refcount on the folio. If you do not already have one,
1187 * folio_try_get() may be the right interface for you to use.
1189 static inline void folio_get(struct folio *folio)
1191 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1192 folio_ref_inc(folio);
1195 static inline void get_page(struct page *page)
1197 folio_get(page_folio(page));
1200 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1202 static inline __must_check bool try_get_page(struct page *page)
1204 page = compound_head(page);
1205 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1212 * folio_put - Decrement the reference count on a folio.
1213 * @folio: The folio.
1215 * If the folio's reference count reaches zero, the memory will be
1216 * released back to the page allocator and may be used by another
1217 * allocation immediately. Do not access the memory or the struct folio
1218 * after calling folio_put() unless you can be sure that it wasn't the
1221 * Context: May be called in process or interrupt context, but not in NMI
1222 * context. May be called while holding a spinlock.
1224 static inline void folio_put(struct folio *folio)
1226 if (folio_put_testzero(folio))
1231 * folio_put_refs - Reduce the reference count on a folio.
1232 * @folio: The folio.
1233 * @refs: The amount to subtract from the folio's reference count.
1235 * If the folio's reference count reaches zero, the memory will be
1236 * released back to the page allocator and may be used by another
1237 * allocation immediately. Do not access the memory or the struct folio
1238 * after calling folio_put_refs() unless you can be sure that these weren't
1239 * the last references.
1241 * Context: May be called in process or interrupt context, but not in NMI
1242 * context. May be called while holding a spinlock.
1244 static inline void folio_put_refs(struct folio *folio, int refs)
1246 if (folio_ref_sub_and_test(folio, refs))
1251 * release_pages - release an array of pages or folios
1253 * This just releases a simple array of multiple pages, and
1254 * accepts various different forms of said page array: either
1255 * a regular old boring array of pages, an array of folios, or
1256 * an array of encoded page pointers.
1258 * The transparent union syntax for this kind of "any of these
1259 * argument types" is all kinds of ugly, so look away.
1262 struct page **pages;
1263 struct folio **folios;
1264 struct encoded_page **encoded_pages;
1265 } release_pages_arg __attribute__ ((__transparent_union__));
1267 void release_pages(release_pages_arg, int nr);
1270 * folios_put - Decrement the reference count on an array of folios.
1271 * @folios: The folios.
1272 * @nr: How many folios there are.
1274 * Like folio_put(), but for an array of folios. This is more efficient
1275 * than writing the loop yourself as it will optimise the locks which
1276 * need to be taken if the folios are freed.
1278 * Context: May be called in process or interrupt context, but not in NMI
1279 * context. May be called while holding a spinlock.
1281 static inline void folios_put(struct folio **folios, unsigned int nr)
1283 release_pages(folios, nr);
1286 static inline void put_page(struct page *page)
1288 struct folio *folio = page_folio(page);
1291 * For some devmap managed pages we need to catch refcount transition
1294 if (put_devmap_managed_page(&folio->page))
1300 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1301 * the page's refcount so that two separate items are tracked: the original page
1302 * reference count, and also a new count of how many pin_user_pages() calls were
1303 * made against the page. ("gup-pinned" is another term for the latter).
1305 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1306 * distinct from normal pages. As such, the unpin_user_page() call (and its
1307 * variants) must be used in order to release gup-pinned pages.
1311 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1312 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1313 * simpler, due to the fact that adding an even power of two to the page
1314 * refcount has the effect of using only the upper N bits, for the code that
1315 * counts up using the bias value. This means that the lower bits are left for
1316 * the exclusive use of the original code that increments and decrements by one
1317 * (or at least, by much smaller values than the bias value).
1319 * Of course, once the lower bits overflow into the upper bits (and this is
1320 * OK, because subtraction recovers the original values), then visual inspection
1321 * no longer suffices to directly view the separate counts. However, for normal
1322 * applications that don't have huge page reference counts, this won't be an
1325 * Locking: the lockless algorithm described in folio_try_get_rcu()
1326 * provides safe operation for get_user_pages(), page_mkclean() and
1327 * other calls that race to set up page table entries.
1329 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1331 void unpin_user_page(struct page *page);
1332 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1334 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1336 void unpin_user_pages(struct page **pages, unsigned long npages);
1338 static inline bool is_cow_mapping(vm_flags_t flags)
1340 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1343 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1344 #define SECTION_IN_PAGE_FLAGS
1348 * The identification function is mainly used by the buddy allocator for
1349 * determining if two pages could be buddies. We are not really identifying
1350 * the zone since we could be using the section number id if we do not have
1351 * node id available in page flags.
1352 * We only guarantee that it will return the same value for two combinable
1355 static inline int page_zone_id(struct page *page)
1357 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1360 #ifdef NODE_NOT_IN_PAGE_FLAGS
1361 extern int page_to_nid(const struct page *page);
1363 static inline int page_to_nid(const struct page *page)
1365 struct page *p = (struct page *)page;
1367 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1371 static inline int folio_nid(const struct folio *folio)
1373 return page_to_nid(&folio->page);
1376 #ifdef CONFIG_NUMA_BALANCING
1377 /* page access time bits needs to hold at least 4 seconds */
1378 #define PAGE_ACCESS_TIME_MIN_BITS 12
1379 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1380 #define PAGE_ACCESS_TIME_BUCKETS \
1381 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1383 #define PAGE_ACCESS_TIME_BUCKETS 0
1386 #define PAGE_ACCESS_TIME_MASK \
1387 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1389 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1391 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1394 static inline int cpupid_to_pid(int cpupid)
1396 return cpupid & LAST__PID_MASK;
1399 static inline int cpupid_to_cpu(int cpupid)
1401 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1404 static inline int cpupid_to_nid(int cpupid)
1406 return cpu_to_node(cpupid_to_cpu(cpupid));
1409 static inline bool cpupid_pid_unset(int cpupid)
1411 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1414 static inline bool cpupid_cpu_unset(int cpupid)
1416 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1419 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1421 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1424 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1425 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1426 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1428 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1431 static inline int page_cpupid_last(struct page *page)
1433 return page->_last_cpupid;
1435 static inline void page_cpupid_reset_last(struct page *page)
1437 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1440 static inline int page_cpupid_last(struct page *page)
1442 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1445 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1447 static inline void page_cpupid_reset_last(struct page *page)
1449 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1451 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1453 static inline int xchg_page_access_time(struct page *page, int time)
1457 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1458 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1460 #else /* !CONFIG_NUMA_BALANCING */
1461 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1463 return page_to_nid(page); /* XXX */
1466 static inline int xchg_page_access_time(struct page *page, int time)
1471 static inline int page_cpupid_last(struct page *page)
1473 return page_to_nid(page); /* XXX */
1476 static inline int cpupid_to_nid(int cpupid)
1481 static inline int cpupid_to_pid(int cpupid)
1486 static inline int cpupid_to_cpu(int cpupid)
1491 static inline int cpu_pid_to_cpupid(int nid, int pid)
1496 static inline bool cpupid_pid_unset(int cpupid)
1501 static inline void page_cpupid_reset_last(struct page *page)
1505 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1509 #endif /* CONFIG_NUMA_BALANCING */
1511 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1514 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1515 * setting tags for all pages to native kernel tag value 0xff, as the default
1516 * value 0x00 maps to 0xff.
1519 static inline u8 page_kasan_tag(const struct page *page)
1523 if (kasan_enabled()) {
1524 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1531 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1533 unsigned long old_flags, flags;
1535 if (!kasan_enabled())
1539 old_flags = READ_ONCE(page->flags);
1542 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1543 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1544 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1547 static inline void page_kasan_tag_reset(struct page *page)
1549 if (kasan_enabled())
1550 page_kasan_tag_set(page, 0xff);
1553 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1555 static inline u8 page_kasan_tag(const struct page *page)
1560 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1561 static inline void page_kasan_tag_reset(struct page *page) { }
1563 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1565 static inline struct zone *page_zone(const struct page *page)
1567 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1570 static inline pg_data_t *page_pgdat(const struct page *page)
1572 return NODE_DATA(page_to_nid(page));
1575 static inline struct zone *folio_zone(const struct folio *folio)
1577 return page_zone(&folio->page);
1580 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1582 return page_pgdat(&folio->page);
1585 #ifdef SECTION_IN_PAGE_FLAGS
1586 static inline void set_page_section(struct page *page, unsigned long section)
1588 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1589 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1592 static inline unsigned long page_to_section(const struct page *page)
1594 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1599 * folio_pfn - Return the Page Frame Number of a folio.
1600 * @folio: The folio.
1602 * A folio may contain multiple pages. The pages have consecutive
1603 * Page Frame Numbers.
1605 * Return: The Page Frame Number of the first page in the folio.
1607 static inline unsigned long folio_pfn(struct folio *folio)
1609 return page_to_pfn(&folio->page);
1612 static inline struct folio *pfn_folio(unsigned long pfn)
1614 return page_folio(pfn_to_page(pfn));
1617 static inline atomic_t *folio_pincount_ptr(struct folio *folio)
1619 return &folio_page(folio, 1)->compound_pincount;
1623 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1624 * @folio: The folio.
1626 * This function checks if a folio has been pinned via a call to
1627 * a function in the pin_user_pages() family.
1629 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1630 * because it means "definitely not pinned for DMA", but true means "probably
1631 * pinned for DMA, but possibly a false positive due to having at least
1632 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1634 * False positives are OK, because: a) it's unlikely for a folio to
1635 * get that many refcounts, and b) all the callers of this routine are
1636 * expected to be able to deal gracefully with a false positive.
1638 * For large folios, the result will be exactly correct. That's because
1639 * we have more tracking data available: the compound_pincount is used
1640 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1642 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1644 * Return: True, if it is likely that the page has been "dma-pinned".
1645 * False, if the page is definitely not dma-pinned.
1647 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1649 if (folio_test_large(folio))
1650 return atomic_read(folio_pincount_ptr(folio)) > 0;
1653 * folio_ref_count() is signed. If that refcount overflows, then
1654 * folio_ref_count() returns a negative value, and callers will avoid
1655 * further incrementing the refcount.
1657 * Here, for that overflow case, use the sign bit to count a little
1658 * bit higher via unsigned math, and thus still get an accurate result.
1660 return ((unsigned int)folio_ref_count(folio)) >=
1661 GUP_PIN_COUNTING_BIAS;
1664 static inline bool page_maybe_dma_pinned(struct page *page)
1666 return folio_maybe_dma_pinned(page_folio(page));
1670 * This should most likely only be called during fork() to see whether we
1671 * should break the cow immediately for an anon page on the src mm.
1673 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1675 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1678 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1680 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1683 return page_maybe_dma_pinned(page);
1686 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1687 #ifdef CONFIG_MIGRATION
1688 static inline bool is_longterm_pinnable_page(struct page *page)
1691 int mt = get_pageblock_migratetype(page);
1693 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1696 /* The zero page may always be pinned */
1697 if (is_zero_pfn(page_to_pfn(page)))
1700 /* Coherent device memory must always allow eviction. */
1701 if (is_device_coherent_page(page))
1704 /* Otherwise, non-movable zone pages can be pinned. */
1705 return !is_zone_movable_page(page);
1708 static inline bool is_longterm_pinnable_page(struct page *page)
1714 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1716 return is_longterm_pinnable_page(&folio->page);
1719 static inline void set_page_zone(struct page *page, enum zone_type zone)
1721 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1722 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1725 static inline void set_page_node(struct page *page, unsigned long node)
1727 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1728 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1731 static inline void set_page_links(struct page *page, enum zone_type zone,
1732 unsigned long node, unsigned long pfn)
1734 set_page_zone(page, zone);
1735 set_page_node(page, node);
1736 #ifdef SECTION_IN_PAGE_FLAGS
1737 set_page_section(page, pfn_to_section_nr(pfn));
1742 * folio_nr_pages - The number of pages in the folio.
1743 * @folio: The folio.
1745 * Return: A positive power of two.
1747 static inline long folio_nr_pages(struct folio *folio)
1749 if (!folio_test_large(folio))
1752 return folio->_folio_nr_pages;
1754 return 1L << folio->_folio_order;
1759 * folio_next - Move to the next physical folio.
1760 * @folio: The folio we're currently operating on.
1762 * If you have physically contiguous memory which may span more than
1763 * one folio (eg a &struct bio_vec), use this function to move from one
1764 * folio to the next. Do not use it if the memory is only virtually
1765 * contiguous as the folios are almost certainly not adjacent to each
1766 * other. This is the folio equivalent to writing ``page++``.
1768 * Context: We assume that the folios are refcounted and/or locked at a
1769 * higher level and do not adjust the reference counts.
1770 * Return: The next struct folio.
1772 static inline struct folio *folio_next(struct folio *folio)
1774 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1778 * folio_shift - The size of the memory described by this folio.
1779 * @folio: The folio.
1781 * A folio represents a number of bytes which is a power-of-two in size.
1782 * This function tells you which power-of-two the folio is. See also
1783 * folio_size() and folio_order().
1785 * Context: The caller should have a reference on the folio to prevent
1786 * it from being split. It is not necessary for the folio to be locked.
1787 * Return: The base-2 logarithm of the size of this folio.
1789 static inline unsigned int folio_shift(struct folio *folio)
1791 return PAGE_SHIFT + folio_order(folio);
1795 * folio_size - The number of bytes in a folio.
1796 * @folio: The folio.
1798 * Context: The caller should have a reference on the folio to prevent
1799 * it from being split. It is not necessary for the folio to be locked.
1800 * Return: The number of bytes in this folio.
1802 static inline size_t folio_size(struct folio *folio)
1804 return PAGE_SIZE << folio_order(folio);
1807 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1808 static inline int arch_make_page_accessible(struct page *page)
1814 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1815 static inline int arch_make_folio_accessible(struct folio *folio)
1818 long i, nr = folio_nr_pages(folio);
1820 for (i = 0; i < nr; i++) {
1821 ret = arch_make_page_accessible(folio_page(folio, i));
1831 * Some inline functions in vmstat.h depend on page_zone()
1833 #include <linux/vmstat.h>
1835 static __always_inline void *lowmem_page_address(const struct page *page)
1837 return page_to_virt(page);
1840 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1841 #define HASHED_PAGE_VIRTUAL
1844 #if defined(WANT_PAGE_VIRTUAL)
1845 static inline void *page_address(const struct page *page)
1847 return page->virtual;
1849 static inline void set_page_address(struct page *page, void *address)
1851 page->virtual = address;
1853 #define page_address_init() do { } while(0)
1856 #if defined(HASHED_PAGE_VIRTUAL)
1857 void *page_address(const struct page *page);
1858 void set_page_address(struct page *page, void *virtual);
1859 void page_address_init(void);
1862 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1863 #define page_address(page) lowmem_page_address(page)
1864 #define set_page_address(page, address) do { } while(0)
1865 #define page_address_init() do { } while(0)
1868 static inline void *folio_address(const struct folio *folio)
1870 return page_address(&folio->page);
1873 extern void *page_rmapping(struct page *page);
1874 extern pgoff_t __page_file_index(struct page *page);
1877 * Return the pagecache index of the passed page. Regular pagecache pages
1878 * use ->index whereas swapcache pages use swp_offset(->private)
1880 static inline pgoff_t page_index(struct page *page)
1882 if (unlikely(PageSwapCache(page)))
1883 return __page_file_index(page);
1888 * Return true only if the page has been allocated with
1889 * ALLOC_NO_WATERMARKS and the low watermark was not
1890 * met implying that the system is under some pressure.
1892 static inline bool page_is_pfmemalloc(const struct page *page)
1895 * lru.next has bit 1 set if the page is allocated from the
1896 * pfmemalloc reserves. Callers may simply overwrite it if
1897 * they do not need to preserve that information.
1899 return (uintptr_t)page->lru.next & BIT(1);
1903 * Only to be called by the page allocator on a freshly allocated
1906 static inline void set_page_pfmemalloc(struct page *page)
1908 page->lru.next = (void *)BIT(1);
1911 static inline void clear_page_pfmemalloc(struct page *page)
1913 page->lru.next = NULL;
1917 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1919 extern void pagefault_out_of_memory(void);
1921 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1922 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1923 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1926 * Flags passed to show_mem() and show_free_areas() to suppress output in
1929 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1931 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
1932 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
1934 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
1938 * Parameter block passed down to zap_pte_range in exceptional cases.
1940 struct zap_details {
1941 struct folio *single_folio; /* Locked folio to be unmapped */
1942 bool even_cows; /* Zap COWed private pages too? */
1943 zap_flags_t zap_flags; /* Extra flags for zapping */
1947 * Whether to drop the pte markers, for example, the uffd-wp information for
1948 * file-backed memory. This should only be specified when we will completely
1949 * drop the page in the mm, either by truncation or unmapping of the vma. By
1950 * default, the flag is not set.
1952 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
1953 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
1954 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
1957 extern bool can_do_mlock(void);
1959 static inline bool can_do_mlock(void) { return false; }
1961 extern int user_shm_lock(size_t, struct ucounts *);
1962 extern void user_shm_unlock(size_t, struct ucounts *);
1964 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1966 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1969 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1970 unsigned long size);
1971 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1972 unsigned long size);
1973 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
1974 unsigned long size, struct zap_details *details);
1975 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
1976 struct vm_area_struct *start_vma, unsigned long start,
1979 struct mmu_notifier_range;
1981 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1982 unsigned long end, unsigned long floor, unsigned long ceiling);
1984 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1985 int follow_pte(struct mm_struct *mm, unsigned long address,
1986 pte_t **ptepp, spinlock_t **ptlp);
1987 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1988 unsigned long *pfn);
1989 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1990 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1991 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1992 void *buf, int len, int write);
1994 extern void truncate_pagecache(struct inode *inode, loff_t new);
1995 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1996 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1997 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1998 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2001 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2002 unsigned long address, unsigned int flags,
2003 struct pt_regs *regs);
2004 extern int fixup_user_fault(struct mm_struct *mm,
2005 unsigned long address, unsigned int fault_flags,
2007 void unmap_mapping_pages(struct address_space *mapping,
2008 pgoff_t start, pgoff_t nr, bool even_cows);
2009 void unmap_mapping_range(struct address_space *mapping,
2010 loff_t const holebegin, loff_t const holelen, int even_cows);
2012 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2013 unsigned long address, unsigned int flags,
2014 struct pt_regs *regs)
2016 /* should never happen if there's no MMU */
2018 return VM_FAULT_SIGBUS;
2020 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2021 unsigned int fault_flags, bool *unlocked)
2023 /* should never happen if there's no MMU */
2027 static inline void unmap_mapping_pages(struct address_space *mapping,
2028 pgoff_t start, pgoff_t nr, bool even_cows) { }
2029 static inline void unmap_mapping_range(struct address_space *mapping,
2030 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2033 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2034 loff_t const holebegin, loff_t const holelen)
2036 unmap_mapping_range(mapping, holebegin, holelen, 0);
2039 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2040 void *buf, int len, unsigned int gup_flags);
2041 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2042 void *buf, int len, unsigned int gup_flags);
2043 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2044 void *buf, int len, unsigned int gup_flags);
2046 long get_user_pages_remote(struct mm_struct *mm,
2047 unsigned long start, unsigned long nr_pages,
2048 unsigned int gup_flags, struct page **pages,
2049 struct vm_area_struct **vmas, int *locked);
2050 long pin_user_pages_remote(struct mm_struct *mm,
2051 unsigned long start, unsigned long nr_pages,
2052 unsigned int gup_flags, struct page **pages,
2053 struct vm_area_struct **vmas, int *locked);
2054 long get_user_pages(unsigned long start, unsigned long nr_pages,
2055 unsigned int gup_flags, struct page **pages,
2056 struct vm_area_struct **vmas);
2057 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2058 unsigned int gup_flags, struct page **pages,
2059 struct vm_area_struct **vmas);
2060 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2061 struct page **pages, unsigned int gup_flags);
2062 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2063 struct page **pages, unsigned int gup_flags);
2065 int get_user_pages_fast(unsigned long start, int nr_pages,
2066 unsigned int gup_flags, struct page **pages);
2067 int pin_user_pages_fast(unsigned long start, int nr_pages,
2068 unsigned int gup_flags, struct page **pages);
2070 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2071 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2072 struct task_struct *task, bool bypass_rlim);
2075 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
2076 struct page **pages);
2077 struct page *get_dump_page(unsigned long addr);
2079 bool folio_mark_dirty(struct folio *folio);
2080 bool set_page_dirty(struct page *page);
2081 int set_page_dirty_lock(struct page *page);
2083 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2085 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2086 unsigned long old_addr, struct vm_area_struct *new_vma,
2087 unsigned long new_addr, unsigned long len,
2088 bool need_rmap_locks);
2091 * Flags used by change_protection(). For now we make it a bitmap so
2092 * that we can pass in multiple flags just like parameters. However
2093 * for now all the callers are only use one of the flags at the same
2097 * Whether we should manually check if we can map individual PTEs writable,
2098 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2099 * PTEs automatically in a writable mapping.
2101 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2102 /* Whether this protection change is for NUMA hints */
2103 #define MM_CP_PROT_NUMA (1UL << 1)
2104 /* Whether this change is for write protecting */
2105 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2106 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2107 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2108 MM_CP_UFFD_WP_RESOLVE)
2110 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2111 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2114 * We want to check manually if we can change individual PTEs writable
2115 * if we can't do that automatically for all PTEs in a mapping. For
2116 * private mappings, that's always the case when we have write
2117 * permissions as we properly have to handle COW.
2119 if (vma->vm_flags & VM_SHARED)
2120 return vma_wants_writenotify(vma, vma->vm_page_prot);
2121 return !!(vma->vm_flags & VM_WRITE);
2124 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2126 extern unsigned long change_protection(struct mmu_gather *tlb,
2127 struct vm_area_struct *vma, unsigned long start,
2128 unsigned long end, pgprot_t newprot,
2129 unsigned long cp_flags);
2130 extern int mprotect_fixup(struct mmu_gather *tlb, struct vm_area_struct *vma,
2131 struct vm_area_struct **pprev, unsigned long start,
2132 unsigned long end, unsigned long newflags);
2135 * doesn't attempt to fault and will return short.
2137 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2138 unsigned int gup_flags, struct page **pages);
2139 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
2140 unsigned int gup_flags, struct page **pages);
2142 static inline bool get_user_page_fast_only(unsigned long addr,
2143 unsigned int gup_flags, struct page **pagep)
2145 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2148 * per-process(per-mm_struct) statistics.
2150 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2152 return percpu_counter_read_positive(&mm->rss_stat[member]);
2155 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2157 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2159 percpu_counter_add(&mm->rss_stat[member], value);
2161 mm_trace_rss_stat(mm, member);
2164 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2166 percpu_counter_inc(&mm->rss_stat[member]);
2168 mm_trace_rss_stat(mm, member);
2171 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2173 percpu_counter_dec(&mm->rss_stat[member]);
2175 mm_trace_rss_stat(mm, member);
2178 /* Optimized variant when page is already known not to be PageAnon */
2179 static inline int mm_counter_file(struct page *page)
2181 if (PageSwapBacked(page))
2182 return MM_SHMEMPAGES;
2183 return MM_FILEPAGES;
2186 static inline int mm_counter(struct page *page)
2189 return MM_ANONPAGES;
2190 return mm_counter_file(page);
2193 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2195 return get_mm_counter(mm, MM_FILEPAGES) +
2196 get_mm_counter(mm, MM_ANONPAGES) +
2197 get_mm_counter(mm, MM_SHMEMPAGES);
2200 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2202 return max(mm->hiwater_rss, get_mm_rss(mm));
2205 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2207 return max(mm->hiwater_vm, mm->total_vm);
2210 static inline void update_hiwater_rss(struct mm_struct *mm)
2212 unsigned long _rss = get_mm_rss(mm);
2214 if ((mm)->hiwater_rss < _rss)
2215 (mm)->hiwater_rss = _rss;
2218 static inline void update_hiwater_vm(struct mm_struct *mm)
2220 if (mm->hiwater_vm < mm->total_vm)
2221 mm->hiwater_vm = mm->total_vm;
2224 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2226 mm->hiwater_rss = get_mm_rss(mm);
2229 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2230 struct mm_struct *mm)
2232 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2234 if (*maxrss < hiwater_rss)
2235 *maxrss = hiwater_rss;
2238 #if defined(SPLIT_RSS_COUNTING)
2239 void sync_mm_rss(struct mm_struct *mm);
2241 static inline void sync_mm_rss(struct mm_struct *mm)
2246 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2247 static inline int pte_special(pte_t pte)
2252 static inline pte_t pte_mkspecial(pte_t pte)
2258 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2259 static inline int pte_devmap(pte_t pte)
2265 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2267 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2271 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2275 #ifdef __PAGETABLE_P4D_FOLDED
2276 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2277 unsigned long address)
2282 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2285 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2286 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2287 unsigned long address)
2291 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2292 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2295 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2297 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2299 if (mm_pud_folded(mm))
2301 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2304 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2306 if (mm_pud_folded(mm))
2308 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2312 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2313 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2314 unsigned long address)
2319 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2320 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2323 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2325 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2327 if (mm_pmd_folded(mm))
2329 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2332 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2334 if (mm_pmd_folded(mm))
2336 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2341 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2343 atomic_long_set(&mm->pgtables_bytes, 0);
2346 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2348 return atomic_long_read(&mm->pgtables_bytes);
2351 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2353 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2356 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2358 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2362 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2363 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2368 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2369 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2372 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2373 int __pte_alloc_kernel(pmd_t *pmd);
2375 #if defined(CONFIG_MMU)
2377 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2378 unsigned long address)
2380 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2381 NULL : p4d_offset(pgd, address);
2384 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2385 unsigned long address)
2387 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2388 NULL : pud_offset(p4d, address);
2391 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2393 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2394 NULL: pmd_offset(pud, address);
2396 #endif /* CONFIG_MMU */
2398 #if USE_SPLIT_PTE_PTLOCKS
2399 #if ALLOC_SPLIT_PTLOCKS
2400 void __init ptlock_cache_init(void);
2401 extern bool ptlock_alloc(struct page *page);
2402 extern void ptlock_free(struct page *page);
2404 static inline spinlock_t *ptlock_ptr(struct page *page)
2408 #else /* ALLOC_SPLIT_PTLOCKS */
2409 static inline void ptlock_cache_init(void)
2413 static inline bool ptlock_alloc(struct page *page)
2418 static inline void ptlock_free(struct page *page)
2422 static inline spinlock_t *ptlock_ptr(struct page *page)
2426 #endif /* ALLOC_SPLIT_PTLOCKS */
2428 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2430 return ptlock_ptr(pmd_page(*pmd));
2433 static inline bool ptlock_init(struct page *page)
2436 * prep_new_page() initialize page->private (and therefore page->ptl)
2437 * with 0. Make sure nobody took it in use in between.
2439 * It can happen if arch try to use slab for page table allocation:
2440 * slab code uses page->slab_cache, which share storage with page->ptl.
2442 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2443 if (!ptlock_alloc(page))
2445 spin_lock_init(ptlock_ptr(page));
2449 #else /* !USE_SPLIT_PTE_PTLOCKS */
2451 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2453 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2455 return &mm->page_table_lock;
2457 static inline void ptlock_cache_init(void) {}
2458 static inline bool ptlock_init(struct page *page) { return true; }
2459 static inline void ptlock_free(struct page *page) {}
2460 #endif /* USE_SPLIT_PTE_PTLOCKS */
2462 static inline void pgtable_init(void)
2464 ptlock_cache_init();
2465 pgtable_cache_init();
2468 static inline bool pgtable_pte_page_ctor(struct page *page)
2470 if (!ptlock_init(page))
2472 __SetPageTable(page);
2473 inc_lruvec_page_state(page, NR_PAGETABLE);
2477 static inline void pgtable_pte_page_dtor(struct page *page)
2480 __ClearPageTable(page);
2481 dec_lruvec_page_state(page, NR_PAGETABLE);
2484 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2486 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2487 pte_t *__pte = pte_offset_map(pmd, address); \
2493 #define pte_unmap_unlock(pte, ptl) do { \
2498 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2500 #define pte_alloc_map(mm, pmd, address) \
2501 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2503 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2504 (pte_alloc(mm, pmd) ? \
2505 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2507 #define pte_alloc_kernel(pmd, address) \
2508 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2509 NULL: pte_offset_kernel(pmd, address))
2511 #if USE_SPLIT_PMD_PTLOCKS
2513 static struct page *pmd_to_page(pmd_t *pmd)
2515 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2516 return virt_to_page((void *)((unsigned long) pmd & mask));
2519 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2521 return ptlock_ptr(pmd_to_page(pmd));
2524 static inline bool pmd_ptlock_init(struct page *page)
2526 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2527 page->pmd_huge_pte = NULL;
2529 return ptlock_init(page);
2532 static inline void pmd_ptlock_free(struct page *page)
2534 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2535 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2540 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2544 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2546 return &mm->page_table_lock;
2549 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2550 static inline void pmd_ptlock_free(struct page *page) {}
2552 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2556 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2558 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2563 static inline bool pgtable_pmd_page_ctor(struct page *page)
2565 if (!pmd_ptlock_init(page))
2567 __SetPageTable(page);
2568 inc_lruvec_page_state(page, NR_PAGETABLE);
2572 static inline void pgtable_pmd_page_dtor(struct page *page)
2574 pmd_ptlock_free(page);
2575 __ClearPageTable(page);
2576 dec_lruvec_page_state(page, NR_PAGETABLE);
2580 * No scalability reason to split PUD locks yet, but follow the same pattern
2581 * as the PMD locks to make it easier if we decide to. The VM should not be
2582 * considered ready to switch to split PUD locks yet; there may be places
2583 * which need to be converted from page_table_lock.
2585 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2587 return &mm->page_table_lock;
2590 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2592 spinlock_t *ptl = pud_lockptr(mm, pud);
2598 extern void __init pagecache_init(void);
2599 extern void free_initmem(void);
2602 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2603 * into the buddy system. The freed pages will be poisoned with pattern
2604 * "poison" if it's within range [0, UCHAR_MAX].
2605 * Return pages freed into the buddy system.
2607 extern unsigned long free_reserved_area(void *start, void *end,
2608 int poison, const char *s);
2610 extern void adjust_managed_page_count(struct page *page, long count);
2611 extern void mem_init_print_info(void);
2613 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2615 /* Free the reserved page into the buddy system, so it gets managed. */
2616 static inline void free_reserved_page(struct page *page)
2618 ClearPageReserved(page);
2619 init_page_count(page);
2621 adjust_managed_page_count(page, 1);
2623 #define free_highmem_page(page) free_reserved_page(page)
2625 static inline void mark_page_reserved(struct page *page)
2627 SetPageReserved(page);
2628 adjust_managed_page_count(page, -1);
2632 * Default method to free all the __init memory into the buddy system.
2633 * The freed pages will be poisoned with pattern "poison" if it's within
2634 * range [0, UCHAR_MAX].
2635 * Return pages freed into the buddy system.
2637 static inline unsigned long free_initmem_default(int poison)
2639 extern char __init_begin[], __init_end[];
2641 return free_reserved_area(&__init_begin, &__init_end,
2642 poison, "unused kernel image (initmem)");
2645 static inline unsigned long get_num_physpages(void)
2648 unsigned long phys_pages = 0;
2650 for_each_online_node(nid)
2651 phys_pages += node_present_pages(nid);
2657 * Using memblock node mappings, an architecture may initialise its
2658 * zones, allocate the backing mem_map and account for memory holes in an
2659 * architecture independent manner.
2661 * An architecture is expected to register range of page frames backed by
2662 * physical memory with memblock_add[_node]() before calling
2663 * free_area_init() passing in the PFN each zone ends at. At a basic
2664 * usage, an architecture is expected to do something like
2666 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2668 * for_each_valid_physical_page_range()
2669 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2670 * free_area_init(max_zone_pfns);
2672 void free_area_init(unsigned long *max_zone_pfn);
2673 unsigned long node_map_pfn_alignment(void);
2674 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2675 unsigned long end_pfn);
2676 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2677 unsigned long end_pfn);
2678 extern void get_pfn_range_for_nid(unsigned int nid,
2679 unsigned long *start_pfn, unsigned long *end_pfn);
2682 static inline int early_pfn_to_nid(unsigned long pfn)
2687 /* please see mm/page_alloc.c */
2688 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2691 extern void set_dma_reserve(unsigned long new_dma_reserve);
2692 extern void memmap_init_range(unsigned long, int, unsigned long,
2693 unsigned long, unsigned long, enum meminit_context,
2694 struct vmem_altmap *, int migratetype);
2695 extern void setup_per_zone_wmarks(void);
2696 extern void calculate_min_free_kbytes(void);
2697 extern int __meminit init_per_zone_wmark_min(void);
2698 extern void mem_init(void);
2699 extern void __init mmap_init(void);
2701 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2702 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2704 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2706 extern long si_mem_available(void);
2707 extern void si_meminfo(struct sysinfo * val);
2708 extern void si_meminfo_node(struct sysinfo *val, int nid);
2709 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2710 extern unsigned long arch_reserved_kernel_pages(void);
2713 extern __printf(3, 4)
2714 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2716 extern void setup_per_cpu_pageset(void);
2719 extern int min_free_kbytes;
2720 extern int watermark_boost_factor;
2721 extern int watermark_scale_factor;
2722 extern bool arch_has_descending_max_zone_pfns(void);
2725 extern atomic_long_t mmap_pages_allocated;
2726 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2728 /* interval_tree.c */
2729 void vma_interval_tree_insert(struct vm_area_struct *node,
2730 struct rb_root_cached *root);
2731 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2732 struct vm_area_struct *prev,
2733 struct rb_root_cached *root);
2734 void vma_interval_tree_remove(struct vm_area_struct *node,
2735 struct rb_root_cached *root);
2736 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2737 unsigned long start, unsigned long last);
2738 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2739 unsigned long start, unsigned long last);
2741 #define vma_interval_tree_foreach(vma, root, start, last) \
2742 for (vma = vma_interval_tree_iter_first(root, start, last); \
2743 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2745 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2746 struct rb_root_cached *root);
2747 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2748 struct rb_root_cached *root);
2749 struct anon_vma_chain *
2750 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2751 unsigned long start, unsigned long last);
2752 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2753 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2754 #ifdef CONFIG_DEBUG_VM_RB
2755 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2758 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2759 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2760 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2763 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2764 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2765 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2766 struct vm_area_struct *expand);
2767 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2768 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2770 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2772 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2773 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2774 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2775 struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2776 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2777 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2778 unsigned long addr, int new_below);
2779 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2780 unsigned long addr, int new_below);
2781 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2782 extern void unlink_file_vma(struct vm_area_struct *);
2783 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2784 unsigned long addr, unsigned long len, pgoff_t pgoff,
2785 bool *need_rmap_locks);
2786 extern void exit_mmap(struct mm_struct *);
2788 void vma_mas_store(struct vm_area_struct *vma, struct ma_state *mas);
2789 void vma_mas_remove(struct vm_area_struct *vma, struct ma_state *mas);
2791 static inline int check_data_rlimit(unsigned long rlim,
2793 unsigned long start,
2794 unsigned long end_data,
2795 unsigned long start_data)
2797 if (rlim < RLIM_INFINITY) {
2798 if (((new - start) + (end_data - start_data)) > rlim)
2805 extern int mm_take_all_locks(struct mm_struct *mm);
2806 extern void mm_drop_all_locks(struct mm_struct *mm);
2808 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2809 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2810 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2811 extern struct file *get_task_exe_file(struct task_struct *task);
2813 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2814 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2816 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2817 const struct vm_special_mapping *sm);
2818 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2819 unsigned long addr, unsigned long len,
2820 unsigned long flags,
2821 const struct vm_special_mapping *spec);
2822 /* This is an obsolete alternative to _install_special_mapping. */
2823 extern int install_special_mapping(struct mm_struct *mm,
2824 unsigned long addr, unsigned long len,
2825 unsigned long flags, struct page **pages);
2827 unsigned long randomize_stack_top(unsigned long stack_top);
2828 unsigned long randomize_page(unsigned long start, unsigned long range);
2830 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2832 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2833 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2834 struct list_head *uf);
2835 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2836 unsigned long len, unsigned long prot, unsigned long flags,
2837 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2838 extern int do_mas_munmap(struct ma_state *mas, struct mm_struct *mm,
2839 unsigned long start, size_t len, struct list_head *uf,
2841 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2842 struct list_head *uf);
2843 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2846 extern int __mm_populate(unsigned long addr, unsigned long len,
2848 static inline void mm_populate(unsigned long addr, unsigned long len)
2851 (void) __mm_populate(addr, len, 1);
2854 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2857 /* These take the mm semaphore themselves */
2858 extern int __must_check vm_brk(unsigned long, unsigned long);
2859 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2860 extern int vm_munmap(unsigned long, size_t);
2861 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2862 unsigned long, unsigned long,
2863 unsigned long, unsigned long);
2865 struct vm_unmapped_area_info {
2866 #define VM_UNMAPPED_AREA_TOPDOWN 1
2867 unsigned long flags;
2868 unsigned long length;
2869 unsigned long low_limit;
2870 unsigned long high_limit;
2871 unsigned long align_mask;
2872 unsigned long align_offset;
2875 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2878 extern void truncate_inode_pages(struct address_space *, loff_t);
2879 extern void truncate_inode_pages_range(struct address_space *,
2880 loff_t lstart, loff_t lend);
2881 extern void truncate_inode_pages_final(struct address_space *);
2883 /* generic vm_area_ops exported for stackable file systems */
2884 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2885 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2886 pgoff_t start_pgoff, pgoff_t end_pgoff);
2887 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2889 extern unsigned long stack_guard_gap;
2890 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2891 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2893 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2894 extern int expand_downwards(struct vm_area_struct *vma,
2895 unsigned long address);
2897 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2899 #define expand_upwards(vma, address) (0)
2902 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2903 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2904 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2905 struct vm_area_struct **pprev);
2908 * Look up the first VMA which intersects the interval [start_addr, end_addr)
2909 * NULL if none. Assume start_addr < end_addr.
2911 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2912 unsigned long start_addr, unsigned long end_addr);
2915 * vma_lookup() - Find a VMA at a specific address
2916 * @mm: The process address space.
2917 * @addr: The user address.
2919 * Return: The vm_area_struct at the given address, %NULL otherwise.
2922 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2924 return mtree_load(&mm->mm_mt, addr);
2927 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2929 unsigned long vm_start = vma->vm_start;
2931 if (vma->vm_flags & VM_GROWSDOWN) {
2932 vm_start -= stack_guard_gap;
2933 if (vm_start > vma->vm_start)
2939 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2941 unsigned long vm_end = vma->vm_end;
2943 if (vma->vm_flags & VM_GROWSUP) {
2944 vm_end += stack_guard_gap;
2945 if (vm_end < vma->vm_end)
2946 vm_end = -PAGE_SIZE;
2951 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2953 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2956 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2957 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2958 unsigned long vm_start, unsigned long vm_end)
2960 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
2962 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2968 static inline bool range_in_vma(struct vm_area_struct *vma,
2969 unsigned long start, unsigned long end)
2971 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2975 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2976 void vma_set_page_prot(struct vm_area_struct *vma);
2978 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2982 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2984 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2988 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2990 #ifdef CONFIG_NUMA_BALANCING
2991 unsigned long change_prot_numa(struct vm_area_struct *vma,
2992 unsigned long start, unsigned long end);
2995 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2996 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2997 unsigned long pfn, unsigned long size, pgprot_t);
2998 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2999 unsigned long pfn, unsigned long size, pgprot_t prot);
3000 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3001 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3002 struct page **pages, unsigned long *num);
3003 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3005 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3007 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3009 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3010 unsigned long pfn, pgprot_t pgprot);
3011 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3013 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
3014 pfn_t pfn, pgprot_t pgprot);
3015 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3016 unsigned long addr, pfn_t pfn);
3017 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3019 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3020 unsigned long addr, struct page *page)
3022 int err = vm_insert_page(vma, addr, page);
3025 return VM_FAULT_OOM;
3026 if (err < 0 && err != -EBUSY)
3027 return VM_FAULT_SIGBUS;
3029 return VM_FAULT_NOPAGE;
3032 #ifndef io_remap_pfn_range
3033 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3034 unsigned long addr, unsigned long pfn,
3035 unsigned long size, pgprot_t prot)
3037 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3041 static inline vm_fault_t vmf_error(int err)
3044 return VM_FAULT_OOM;
3045 return VM_FAULT_SIGBUS;
3048 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3049 unsigned int foll_flags);
3051 #define FOLL_WRITE 0x01 /* check pte is writable */
3052 #define FOLL_TOUCH 0x02 /* mark page accessed */
3053 #define FOLL_GET 0x04 /* do get_page on page */
3054 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
3055 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
3056 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
3057 * and return without waiting upon it */
3058 #define FOLL_NOFAULT 0x80 /* do not fault in pages */
3059 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
3060 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
3061 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
3062 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
3063 #define FOLL_ANON 0x8000 /* don't do file mappings */
3064 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
3065 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
3066 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
3067 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
3070 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
3071 * other. Here is what they mean, and how to use them:
3073 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
3074 * period _often_ under userspace control. This is in contrast to
3075 * iov_iter_get_pages(), whose usages are transient.
3077 * FIXME: For pages which are part of a filesystem, mappings are subject to the
3078 * lifetime enforced by the filesystem and we need guarantees that longterm
3079 * users like RDMA and V4L2 only establish mappings which coordinate usage with
3080 * the filesystem. Ideas for this coordination include revoking the longterm
3081 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
3082 * added after the problem with filesystems was found FS DAX VMAs are
3083 * specifically failed. Filesystem pages are still subject to bugs and use of
3084 * FOLL_LONGTERM should be avoided on those pages.
3086 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
3087 * Currently only get_user_pages() and get_user_pages_fast() support this flag
3088 * and calls to get_user_pages_[un]locked are specifically not allowed. This
3089 * is due to an incompatibility with the FS DAX check and
3090 * FAULT_FLAG_ALLOW_RETRY.
3092 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
3093 * that region. And so, CMA attempts to migrate the page before pinning, when
3094 * FOLL_LONGTERM is specified.
3096 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
3097 * but an additional pin counting system) will be invoked. This is intended for
3098 * anything that gets a page reference and then touches page data (for example,
3099 * Direct IO). This lets the filesystem know that some non-file-system entity is
3100 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
3101 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
3102 * a call to unpin_user_page().
3104 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
3105 * and separate refcounting mechanisms, however, and that means that each has
3106 * its own acquire and release mechanisms:
3108 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
3110 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
3112 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
3113 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
3114 * calls applied to them, and that's perfectly OK. This is a constraint on the
3115 * callers, not on the pages.)
3117 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
3118 * directly by the caller. That's in order to help avoid mismatches when
3119 * releasing pages: get_user_pages*() pages must be released via put_page(),
3120 * while pin_user_pages*() pages must be released via unpin_user_page().
3122 * Please see Documentation/core-api/pin_user_pages.rst for more information.
3125 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3127 if (vm_fault & VM_FAULT_OOM)
3129 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3130 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3131 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3137 * Indicates for which pages that are write-protected in the page table,
3138 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
3139 * GUP pin will remain consistent with the pages mapped into the page tables
3142 * Temporary unmapping of PageAnonExclusive() pages or clearing of
3143 * PageAnonExclusive() has to protect against concurrent GUP:
3144 * * Ordinary GUP: Using the PT lock
3145 * * GUP-fast and fork(): mm->write_protect_seq
3146 * * GUP-fast and KSM or temporary unmapping (swap, migration): see
3147 * page_try_share_anon_rmap()
3149 * Must be called with the (sub)page that's actually referenced via the
3150 * page table entry, which might not necessarily be the head page for a
3153 static inline bool gup_must_unshare(unsigned int flags, struct page *page)
3156 * FOLL_WRITE is implicitly handled correctly as the page table entry
3157 * has to be writable -- and if it references (part of) an anonymous
3158 * folio, that part is required to be marked exclusive.
3160 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
3163 * Note: PageAnon(page) is stable until the page is actually getting
3166 if (!PageAnon(page))
3169 /* Paired with a memory barrier in page_try_share_anon_rmap(). */
3170 if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
3174 * Note that PageKsm() pages cannot be exclusive, and consequently,
3175 * cannot get pinned.
3177 return !PageAnonExclusive(page);
3181 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3182 * a (NUMA hinting) fault is required.
3184 static inline bool gup_can_follow_protnone(unsigned int flags)
3187 * FOLL_FORCE has to be able to make progress even if the VMA is
3188 * inaccessible. Further, FOLL_FORCE access usually does not represent
3189 * application behaviour and we should avoid triggering NUMA hinting
3192 return flags & FOLL_FORCE;
3195 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3196 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3197 unsigned long size, pte_fn_t fn, void *data);
3198 extern int apply_to_existing_page_range(struct mm_struct *mm,
3199 unsigned long address, unsigned long size,
3200 pte_fn_t fn, void *data);
3202 extern void __init init_mem_debugging_and_hardening(void);
3203 #ifdef CONFIG_PAGE_POISONING
3204 extern void __kernel_poison_pages(struct page *page, int numpages);
3205 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3206 extern bool _page_poisoning_enabled_early;
3207 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3208 static inline bool page_poisoning_enabled(void)
3210 return _page_poisoning_enabled_early;
3213 * For use in fast paths after init_mem_debugging() has run, or when a
3214 * false negative result is not harmful when called too early.
3216 static inline bool page_poisoning_enabled_static(void)
3218 return static_branch_unlikely(&_page_poisoning_enabled);
3220 static inline void kernel_poison_pages(struct page *page, int numpages)
3222 if (page_poisoning_enabled_static())
3223 __kernel_poison_pages(page, numpages);
3225 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3227 if (page_poisoning_enabled_static())
3228 __kernel_unpoison_pages(page, numpages);
3231 static inline bool page_poisoning_enabled(void) { return false; }
3232 static inline bool page_poisoning_enabled_static(void) { return false; }
3233 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3234 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3235 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3238 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3239 static inline bool want_init_on_alloc(gfp_t flags)
3241 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3244 return flags & __GFP_ZERO;
3247 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3248 static inline bool want_init_on_free(void)
3250 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3254 extern bool _debug_pagealloc_enabled_early;
3255 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3257 static inline bool debug_pagealloc_enabled(void)
3259 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3260 _debug_pagealloc_enabled_early;
3264 * For use in fast paths after init_debug_pagealloc() has run, or when a
3265 * false negative result is not harmful when called too early.
3267 static inline bool debug_pagealloc_enabled_static(void)
3269 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3272 return static_branch_unlikely(&_debug_pagealloc_enabled);
3275 #ifdef CONFIG_DEBUG_PAGEALLOC
3277 * To support DEBUG_PAGEALLOC architecture must ensure that
3278 * __kernel_map_pages() never fails
3280 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3282 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3284 if (debug_pagealloc_enabled_static())
3285 __kernel_map_pages(page, numpages, 1);
3288 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3290 if (debug_pagealloc_enabled_static())
3291 __kernel_map_pages(page, numpages, 0);
3293 #else /* CONFIG_DEBUG_PAGEALLOC */
3294 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3295 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3296 #endif /* CONFIG_DEBUG_PAGEALLOC */
3298 #ifdef __HAVE_ARCH_GATE_AREA
3299 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3300 extern int in_gate_area_no_mm(unsigned long addr);
3301 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3303 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3307 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3308 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3312 #endif /* __HAVE_ARCH_GATE_AREA */
3314 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3316 #ifdef CONFIG_SYSCTL
3317 extern int sysctl_drop_caches;
3318 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3322 void drop_slab(void);
3325 #define randomize_va_space 0
3327 extern int randomize_va_space;
3330 const char * arch_vma_name(struct vm_area_struct *vma);
3332 void print_vma_addr(char *prefix, unsigned long rip);
3334 static inline void print_vma_addr(char *prefix, unsigned long rip)
3339 void *sparse_buffer_alloc(unsigned long size);
3340 struct page * __populate_section_memmap(unsigned long pfn,
3341 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3342 struct dev_pagemap *pgmap);
3343 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3344 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3345 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3346 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3347 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3348 struct vmem_altmap *altmap, struct page *reuse);
3349 void *vmemmap_alloc_block(unsigned long size, int node);
3351 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3352 struct vmem_altmap *altmap);
3353 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3354 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3355 int node, struct vmem_altmap *altmap);
3356 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3357 struct vmem_altmap *altmap);
3358 void vmemmap_populate_print_last(void);
3359 #ifdef CONFIG_MEMORY_HOTPLUG
3360 void vmemmap_free(unsigned long start, unsigned long end,
3361 struct vmem_altmap *altmap);
3363 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3364 unsigned long nr_pages);
3367 MF_COUNT_INCREASED = 1 << 0,
3368 MF_ACTION_REQUIRED = 1 << 1,
3369 MF_MUST_KILL = 1 << 2,
3370 MF_SOFT_OFFLINE = 1 << 3,
3371 MF_UNPOISON = 1 << 4,
3372 MF_SW_SIMULATED = 1 << 5,
3373 MF_NO_RETRY = 1 << 6,
3375 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3376 unsigned long count, int mf_flags);
3377 extern int memory_failure(unsigned long pfn, int flags);
3378 extern void memory_failure_queue_kick(int cpu);
3379 extern int unpoison_memory(unsigned long pfn);
3380 extern int sysctl_memory_failure_early_kill;
3381 extern int sysctl_memory_failure_recovery;
3382 extern void shake_page(struct page *p);
3383 extern atomic_long_t num_poisoned_pages __read_mostly;
3384 extern int soft_offline_page(unsigned long pfn, int flags);
3385 #ifdef CONFIG_MEMORY_FAILURE
3386 extern void memory_failure_queue(unsigned long pfn, int flags);
3387 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3388 bool *migratable_cleared);
3389 void num_poisoned_pages_inc(unsigned long pfn);
3390 void num_poisoned_pages_sub(unsigned long pfn, long i);
3392 static inline void memory_failure_queue(unsigned long pfn, int flags)
3396 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3397 bool *migratable_cleared)
3402 static inline void num_poisoned_pages_inc(unsigned long pfn)
3406 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3411 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3412 extern void memblk_nr_poison_inc(unsigned long pfn);
3413 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3415 static inline void memblk_nr_poison_inc(unsigned long pfn)
3419 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3424 #ifndef arch_memory_failure
3425 static inline int arch_memory_failure(unsigned long pfn, int flags)
3431 #ifndef arch_is_platform_page
3432 static inline bool arch_is_platform_page(u64 paddr)
3439 * Error handlers for various types of pages.
3442 MF_IGNORED, /* Error: cannot be handled */
3443 MF_FAILED, /* Error: handling failed */
3444 MF_DELAYED, /* Will be handled later */
3445 MF_RECOVERED, /* Successfully recovered */
3448 enum mf_action_page_type {
3450 MF_MSG_KERNEL_HIGH_ORDER,
3452 MF_MSG_DIFFERENT_COMPOUND,
3455 MF_MSG_UNMAP_FAILED,
3456 MF_MSG_DIRTY_SWAPCACHE,
3457 MF_MSG_CLEAN_SWAPCACHE,
3458 MF_MSG_DIRTY_MLOCKED_LRU,
3459 MF_MSG_CLEAN_MLOCKED_LRU,
3460 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3461 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3464 MF_MSG_TRUNCATED_LRU,
3471 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3472 extern void clear_huge_page(struct page *page,
3473 unsigned long addr_hint,
3474 unsigned int pages_per_huge_page);
3475 extern void copy_user_huge_page(struct page *dst, struct page *src,
3476 unsigned long addr_hint,
3477 struct vm_area_struct *vma,
3478 unsigned int pages_per_huge_page);
3479 extern long copy_huge_page_from_user(struct page *dst_page,
3480 const void __user *usr_src,
3481 unsigned int pages_per_huge_page,
3482 bool allow_pagefault);
3485 * vma_is_special_huge - Are transhuge page-table entries considered special?
3486 * @vma: Pointer to the struct vm_area_struct to consider
3488 * Whether transhuge page-table entries are considered "special" following
3489 * the definition in vm_normal_page().
3491 * Return: true if transhuge page-table entries should be considered special,
3494 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3496 return vma_is_dax(vma) || (vma->vm_file &&
3497 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3500 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3502 #ifdef CONFIG_DEBUG_PAGEALLOC
3503 extern unsigned int _debug_guardpage_minorder;
3504 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3506 static inline unsigned int debug_guardpage_minorder(void)
3508 return _debug_guardpage_minorder;
3511 static inline bool debug_guardpage_enabled(void)
3513 return static_branch_unlikely(&_debug_guardpage_enabled);
3516 static inline bool page_is_guard(struct page *page)
3518 if (!debug_guardpage_enabled())
3521 return PageGuard(page);
3524 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3525 static inline bool debug_guardpage_enabled(void) { return false; }
3526 static inline bool page_is_guard(struct page *page) { return false; }
3527 #endif /* CONFIG_DEBUG_PAGEALLOC */
3529 #if MAX_NUMNODES > 1
3530 void __init setup_nr_node_ids(void);
3532 static inline void setup_nr_node_ids(void) {}
3535 extern int memcmp_pages(struct page *page1, struct page *page2);
3537 static inline int pages_identical(struct page *page1, struct page *page2)
3539 return !memcmp_pages(page1, page2);
3542 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3543 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3544 pgoff_t first_index, pgoff_t nr,
3545 pgoff_t bitmap_pgoff,
3546 unsigned long *bitmap,
3550 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3551 pgoff_t first_index, pgoff_t nr);
3554 extern int sysctl_nr_trim_pages;
3556 #ifdef CONFIG_PRINTK
3557 void mem_dump_obj(void *object);
3559 static inline void mem_dump_obj(void *object) {}
3563 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3564 * @seals: the seals to check
3565 * @vma: the vma to operate on
3567 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3568 * the vma flags. Return 0 if check pass, or <0 for errors.
3570 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3572 if (seals & F_SEAL_FUTURE_WRITE) {
3574 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3575 * "future write" seal active.
3577 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3581 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3582 * MAP_SHARED and read-only, take care to not allow mprotect to
3583 * revert protections on such mappings. Do this only for shared
3584 * mappings. For private mappings, don't need to mask
3585 * VM_MAYWRITE as we still want them to be COW-writable.
3587 if (vma->vm_flags & VM_SHARED)
3588 vma->vm_flags &= ~(VM_MAYWRITE);
3594 #ifdef CONFIG_ANON_VMA_NAME
3595 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3596 unsigned long len_in,
3597 struct anon_vma_name *anon_name);
3600 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3601 unsigned long len_in, struct anon_vma_name *anon_name) {
3606 #endif /* _LINUX_MM_H */