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>
32 #include <linux/slab.h>
36 struct anon_vma_chain;
40 extern int sysctl_page_lock_unfairness;
42 void mm_core_init(void);
43 void init_mm_internals(void);
45 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
46 extern unsigned long max_mapnr;
48 static inline void set_max_mapnr(unsigned long limit)
53 static inline void set_max_mapnr(unsigned long limit) { }
56 extern atomic_long_t _totalram_pages;
57 static inline unsigned long totalram_pages(void)
59 return (unsigned long)atomic_long_read(&_totalram_pages);
62 static inline void totalram_pages_inc(void)
64 atomic_long_inc(&_totalram_pages);
67 static inline void totalram_pages_dec(void)
69 atomic_long_dec(&_totalram_pages);
72 static inline void totalram_pages_add(long count)
74 atomic_long_add(count, &_totalram_pages);
77 extern void * high_memory;
78 extern int page_cluster;
79 extern const int page_cluster_max;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/processor.h>
102 * Architectures that support memory tagging (assigning tags to memory regions,
103 * embedding these tags into addresses that point to these memory regions, and
104 * checking that the memory and the pointer tags match on memory accesses)
105 * redefine this macro to strip tags from pointers.
106 * It's defined as noop for architectures that don't support memory tagging.
108 #ifndef untagged_addr
109 #define untagged_addr(addr) (addr)
113 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 #define lm_alias(x) __va(__pa_symbol(x))
125 * To prevent common memory management code establishing
126 * a zero page mapping on a read fault.
127 * This macro should be defined within <asm/pgtable.h>.
128 * s390 does this to prevent multiplexing of hardware bits
129 * related to the physical page in case of virtualization.
131 #ifndef mm_forbids_zeropage
132 #define mm_forbids_zeropage(X) (0)
136 * On some architectures it is expensive to call memset() for small sizes.
137 * If an architecture decides to implement their own version of
138 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
139 * define their own version of this macro in <asm/pgtable.h>
141 #if BITS_PER_LONG == 64
142 /* This function must be updated when the size of struct page grows above 96
143 * or reduces below 56. The idea that compiler optimizes out switch()
144 * statement, and only leaves move/store instructions. Also the compiler can
145 * combine write statements if they are both assignments and can be reordered,
146 * this can result in several of the writes here being dropped.
148 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
149 static inline void __mm_zero_struct_page(struct page *page)
151 unsigned long *_pp = (void *)page;
153 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
154 BUILD_BUG_ON(sizeof(struct page) & 7);
155 BUILD_BUG_ON(sizeof(struct page) < 56);
156 BUILD_BUG_ON(sizeof(struct page) > 96);
158 switch (sizeof(struct page)) {
185 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
189 * Default maximum number of active map areas, this limits the number of vmas
190 * per mm struct. Users can overwrite this number by sysctl but there is a
193 * When a program's coredump is generated as ELF format, a section is created
194 * per a vma. In ELF, the number of sections is represented in unsigned short.
195 * This means the number of sections should be smaller than 65535 at coredump.
196 * Because the kernel adds some informative sections to a image of program at
197 * generating coredump, we need some margin. The number of extra sections is
198 * 1-3 now and depends on arch. We use "5" as safe margin, here.
200 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
201 * not a hard limit any more. Although some userspace tools can be surprised by
204 #define MAPCOUNT_ELF_CORE_MARGIN (5)
205 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
207 extern int sysctl_max_map_count;
209 extern unsigned long sysctl_user_reserve_kbytes;
210 extern unsigned long sysctl_admin_reserve_kbytes;
212 extern int sysctl_overcommit_memory;
213 extern int sysctl_overcommit_ratio;
214 extern unsigned long sysctl_overcommit_kbytes;
216 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
218 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
220 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
223 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
224 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
225 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
227 #define nth_page(page,n) ((page) + (n))
228 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
231 /* to align the pointer to the (next) page boundary */
232 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
234 /* to align the pointer to the (prev) page boundary */
235 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
237 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
238 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
240 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
241 static inline struct folio *lru_to_folio(struct list_head *head)
243 return list_entry((head)->prev, struct folio, lru);
246 void setup_initial_init_mm(void *start_code, void *end_code,
247 void *end_data, void *brk);
250 * Linux kernel virtual memory manager primitives.
251 * The idea being to have a "virtual" mm in the same way
252 * we have a virtual fs - giving a cleaner interface to the
253 * mm details, and allowing different kinds of memory mappings
254 * (from shared memory to executable loading to arbitrary
258 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
259 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
260 void vm_area_free(struct vm_area_struct *);
261 /* Use only if VMA has no other users */
262 void __vm_area_free(struct vm_area_struct *vma);
265 extern struct rb_root nommu_region_tree;
266 extern struct rw_semaphore nommu_region_sem;
268 extern unsigned int kobjsize(const void *objp);
272 * vm_flags in vm_area_struct, see mm_types.h.
273 * When changing, update also include/trace/events/mmflags.h
275 #define VM_NONE 0x00000000
277 #define VM_READ 0x00000001 /* currently active flags */
278 #define VM_WRITE 0x00000002
279 #define VM_EXEC 0x00000004
280 #define VM_SHARED 0x00000008
282 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
283 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
284 #define VM_MAYWRITE 0x00000020
285 #define VM_MAYEXEC 0x00000040
286 #define VM_MAYSHARE 0x00000080
288 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
290 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
291 #else /* CONFIG_MMU */
292 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
293 #define VM_UFFD_MISSING 0
294 #endif /* CONFIG_MMU */
295 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
296 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
298 #define VM_LOCKED 0x00002000
299 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
301 /* Used by sys_madvise() */
302 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
303 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
305 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
306 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
307 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
308 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
309 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
310 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
311 #define VM_SYNC 0x00800000 /* Synchronous page faults */
312 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
313 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
314 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
316 #ifdef CONFIG_MEM_SOFT_DIRTY
317 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
319 # define VM_SOFTDIRTY 0
322 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
323 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
324 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
325 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
327 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
328 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
329 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
330 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
331 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
332 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
333 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
334 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
335 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
336 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
337 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
338 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
340 #ifdef CONFIG_ARCH_HAS_PKEYS
341 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
342 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
343 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
344 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
345 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
347 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
349 # define VM_PKEY_BIT4 0
351 #endif /* CONFIG_ARCH_HAS_PKEYS */
353 #if defined(CONFIG_X86)
354 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
355 #elif defined(CONFIG_PPC)
356 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
357 #elif defined(CONFIG_PARISC)
358 # define VM_GROWSUP VM_ARCH_1
359 #elif defined(CONFIG_IA64)
360 # define VM_GROWSUP VM_ARCH_1
361 #elif defined(CONFIG_SPARC64)
362 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
363 # define VM_ARCH_CLEAR VM_SPARC_ADI
364 #elif defined(CONFIG_ARM64)
365 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
366 # define VM_ARCH_CLEAR VM_ARM64_BTI
367 #elif !defined(CONFIG_MMU)
368 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
371 #if defined(CONFIG_ARM64_MTE)
372 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
373 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
375 # define VM_MTE VM_NONE
376 # define VM_MTE_ALLOWED VM_NONE
380 # define VM_GROWSUP VM_NONE
383 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
384 # define VM_UFFD_MINOR_BIT 37
385 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
386 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
387 # define VM_UFFD_MINOR VM_NONE
388 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
390 /* Bits set in the VMA until the stack is in its final location */
391 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
393 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
395 /* Common data flag combinations */
396 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
397 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
398 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
399 VM_MAYWRITE | VM_MAYEXEC)
400 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
401 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
403 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
404 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
407 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
408 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
411 #ifdef CONFIG_STACK_GROWSUP
412 #define VM_STACK VM_GROWSUP
414 #define VM_STACK VM_GROWSDOWN
417 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
419 /* VMA basic access permission flags */
420 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
424 * Special vmas that are non-mergable, non-mlock()able.
426 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
428 /* This mask prevents VMA from being scanned with khugepaged */
429 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
431 /* This mask defines which mm->def_flags a process can inherit its parent */
432 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
434 /* This mask represents all the VMA flag bits used by mlock */
435 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
437 /* Arch-specific flags to clear when updating VM flags on protection change */
438 #ifndef VM_ARCH_CLEAR
439 # define VM_ARCH_CLEAR VM_NONE
441 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
444 * mapping from the currently active vm_flags protection bits (the
445 * low four bits) to a page protection mask..
449 * The default fault flags that should be used by most of the
450 * arch-specific page fault handlers.
452 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
453 FAULT_FLAG_KILLABLE | \
454 FAULT_FLAG_INTERRUPTIBLE)
457 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
458 * @flags: Fault flags.
460 * This is mostly used for places where we want to try to avoid taking
461 * the mmap_lock for too long a time when waiting for another condition
462 * to change, in which case we can try to be polite to release the
463 * mmap_lock in the first round to avoid potential starvation of other
464 * processes that would also want the mmap_lock.
466 * Return: true if the page fault allows retry and this is the first
467 * attempt of the fault handling; false otherwise.
469 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
471 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
472 (!(flags & FAULT_FLAG_TRIED));
475 #define FAULT_FLAG_TRACE \
476 { FAULT_FLAG_WRITE, "WRITE" }, \
477 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
478 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
479 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
480 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
481 { FAULT_FLAG_TRIED, "TRIED" }, \
482 { FAULT_FLAG_USER, "USER" }, \
483 { FAULT_FLAG_REMOTE, "REMOTE" }, \
484 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
485 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
486 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
489 * vm_fault is filled by the pagefault handler and passed to the vma's
490 * ->fault function. The vma's ->fault is responsible for returning a bitmask
491 * of VM_FAULT_xxx flags that give details about how the fault was handled.
493 * MM layer fills up gfp_mask for page allocations but fault handler might
494 * alter it if its implementation requires a different allocation context.
496 * pgoff should be used in favour of virtual_address, if possible.
500 struct vm_area_struct *vma; /* Target VMA */
501 gfp_t gfp_mask; /* gfp mask to be used for allocations */
502 pgoff_t pgoff; /* Logical page offset based on vma */
503 unsigned long address; /* Faulting virtual address - masked */
504 unsigned long real_address; /* Faulting virtual address - unmasked */
506 enum fault_flag flags; /* FAULT_FLAG_xxx flags
507 * XXX: should really be 'const' */
508 pmd_t *pmd; /* Pointer to pmd entry matching
510 pud_t *pud; /* Pointer to pud entry matching
514 pte_t orig_pte; /* Value of PTE at the time of fault */
515 pmd_t orig_pmd; /* Value of PMD at the time of fault,
516 * used by PMD fault only.
520 struct page *cow_page; /* Page handler may use for COW fault */
521 struct page *page; /* ->fault handlers should return a
522 * page here, unless VM_FAULT_NOPAGE
523 * is set (which is also implied by
526 /* These three entries are valid only while holding ptl lock */
527 pte_t *pte; /* Pointer to pte entry matching
528 * the 'address'. NULL if the page
529 * table hasn't been allocated.
531 spinlock_t *ptl; /* Page table lock.
532 * Protects pte page table if 'pte'
533 * is not NULL, otherwise pmd.
535 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
536 * vm_ops->map_pages() sets up a page
537 * table from atomic context.
538 * do_fault_around() pre-allocates
539 * page table to avoid allocation from
544 /* page entry size for vm->huge_fault() */
545 enum page_entry_size {
552 * These are the virtual MM functions - opening of an area, closing and
553 * unmapping it (needed to keep files on disk up-to-date etc), pointer
554 * to the functions called when a no-page or a wp-page exception occurs.
556 struct vm_operations_struct {
557 void (*open)(struct vm_area_struct * area);
559 * @close: Called when the VMA is being removed from the MM.
560 * Context: User context. May sleep. Caller holds mmap_lock.
562 void (*close)(struct vm_area_struct * area);
563 /* Called any time before splitting to check if it's allowed */
564 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
565 int (*mremap)(struct vm_area_struct *area);
567 * Called by mprotect() to make driver-specific permission
568 * checks before mprotect() is finalised. The VMA must not
569 * be modified. Returns 0 if mprotect() can proceed.
571 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
572 unsigned long end, unsigned long newflags);
573 vm_fault_t (*fault)(struct vm_fault *vmf);
574 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
575 enum page_entry_size pe_size);
576 vm_fault_t (*map_pages)(struct vm_fault *vmf,
577 pgoff_t start_pgoff, pgoff_t end_pgoff);
578 unsigned long (*pagesize)(struct vm_area_struct * area);
580 /* notification that a previously read-only page is about to become
581 * writable, if an error is returned it will cause a SIGBUS */
582 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
584 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
585 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
587 /* called by access_process_vm when get_user_pages() fails, typically
588 * for use by special VMAs. See also generic_access_phys() for a generic
589 * implementation useful for any iomem mapping.
591 int (*access)(struct vm_area_struct *vma, unsigned long addr,
592 void *buf, int len, int write);
594 /* Called by the /proc/PID/maps code to ask the vma whether it
595 * has a special name. Returning non-NULL will also cause this
596 * vma to be dumped unconditionally. */
597 const char *(*name)(struct vm_area_struct *vma);
601 * set_policy() op must add a reference to any non-NULL @new mempolicy
602 * to hold the policy upon return. Caller should pass NULL @new to
603 * remove a policy and fall back to surrounding context--i.e. do not
604 * install a MPOL_DEFAULT policy, nor the task or system default
607 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
610 * get_policy() op must add reference [mpol_get()] to any policy at
611 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
612 * in mm/mempolicy.c will do this automatically.
613 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
614 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
615 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
616 * must return NULL--i.e., do not "fallback" to task or system default
619 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
623 * Called by vm_normal_page() for special PTEs to find the
624 * page for @addr. This is useful if the default behavior
625 * (using pte_page()) would not find the correct page.
627 struct page *(*find_special_page)(struct vm_area_struct *vma,
631 #ifdef CONFIG_NUMA_BALANCING
632 static inline void vma_numab_state_init(struct vm_area_struct *vma)
634 vma->numab_state = NULL;
636 static inline void vma_numab_state_free(struct vm_area_struct *vma)
638 kfree(vma->numab_state);
641 static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
642 static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
643 #endif /* CONFIG_NUMA_BALANCING */
645 #ifdef CONFIG_PER_VMA_LOCK
647 * Try to read-lock a vma. The function is allowed to occasionally yield false
648 * locked result to avoid performance overhead, in which case we fall back to
649 * using mmap_lock. The function should never yield false unlocked result.
651 static inline bool vma_start_read(struct vm_area_struct *vma)
653 /* Check before locking. A race might cause false locked result. */
654 if (vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))
657 if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
661 * Overflow might produce false locked result.
662 * False unlocked result is impossible because we modify and check
663 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
664 * modification invalidates all existing locks.
666 if (unlikely(vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))) {
667 up_read(&vma->vm_lock->lock);
673 static inline void vma_end_read(struct vm_area_struct *vma)
675 rcu_read_lock(); /* keeps vma alive till the end of up_read */
676 up_read(&vma->vm_lock->lock);
680 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
682 mmap_assert_write_locked(vma->vm_mm);
685 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
686 * mm->mm_lock_seq can't be concurrently modified.
688 *mm_lock_seq = READ_ONCE(vma->vm_mm->mm_lock_seq);
689 return (vma->vm_lock_seq == *mm_lock_seq);
692 static inline void vma_start_write(struct vm_area_struct *vma)
696 if (__is_vma_write_locked(vma, &mm_lock_seq))
699 down_write(&vma->vm_lock->lock);
700 vma->vm_lock_seq = mm_lock_seq;
701 up_write(&vma->vm_lock->lock);
704 static inline bool vma_try_start_write(struct vm_area_struct *vma)
708 if (__is_vma_write_locked(vma, &mm_lock_seq))
711 if (!down_write_trylock(&vma->vm_lock->lock))
714 vma->vm_lock_seq = mm_lock_seq;
715 up_write(&vma->vm_lock->lock);
719 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
723 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
726 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
728 /* When detaching vma should be write-locked */
730 vma_assert_write_locked(vma);
731 vma->detached = detached;
734 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
735 unsigned long address);
737 #else /* CONFIG_PER_VMA_LOCK */
739 static inline void vma_init_lock(struct vm_area_struct *vma) {}
740 static inline bool vma_start_read(struct vm_area_struct *vma)
742 static inline void vma_end_read(struct vm_area_struct *vma) {}
743 static inline void vma_start_write(struct vm_area_struct *vma) {}
744 static inline bool vma_try_start_write(struct vm_area_struct *vma)
746 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
747 static inline void vma_mark_detached(struct vm_area_struct *vma,
750 #endif /* CONFIG_PER_VMA_LOCK */
753 * WARNING: vma_init does not initialize vma->vm_lock.
754 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
756 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
758 static const struct vm_operations_struct dummy_vm_ops = {};
760 memset(vma, 0, sizeof(*vma));
762 vma->vm_ops = &dummy_vm_ops;
763 INIT_LIST_HEAD(&vma->anon_vma_chain);
764 vma_mark_detached(vma, false);
765 vma_numab_state_init(vma);
768 /* Use when VMA is not part of the VMA tree and needs no locking */
769 static inline void vm_flags_init(struct vm_area_struct *vma,
772 ACCESS_PRIVATE(vma, __vm_flags) = flags;
775 /* Use when VMA is part of the VMA tree and modifications need coordination */
776 static inline void vm_flags_reset(struct vm_area_struct *vma,
779 vma_start_write(vma);
780 vm_flags_init(vma, flags);
783 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
786 vma_start_write(vma);
787 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
790 static inline void vm_flags_set(struct vm_area_struct *vma,
793 vma_start_write(vma);
794 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
797 static inline void vm_flags_clear(struct vm_area_struct *vma,
800 vma_start_write(vma);
801 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
805 * Use only if VMA is not part of the VMA tree or has no other users and
806 * therefore needs no locking.
808 static inline void __vm_flags_mod(struct vm_area_struct *vma,
809 vm_flags_t set, vm_flags_t clear)
811 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
815 * Use only when the order of set/clear operations is unimportant, otherwise
816 * use vm_flags_{set|clear} explicitly.
818 static inline void vm_flags_mod(struct vm_area_struct *vma,
819 vm_flags_t set, vm_flags_t clear)
821 vma_start_write(vma);
822 __vm_flags_mod(vma, set, clear);
825 static inline void vma_set_anonymous(struct vm_area_struct *vma)
830 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
835 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
837 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
842 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
843 VM_STACK_INCOMPLETE_SETUP)
849 static inline bool vma_is_foreign(struct vm_area_struct *vma)
854 if (current->mm != vma->vm_mm)
860 static inline bool vma_is_accessible(struct vm_area_struct *vma)
862 return vma->vm_flags & VM_ACCESS_FLAGS;
866 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
868 return mas_find(&vmi->mas, max - 1);
871 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
874 * Uses mas_find() to get the first VMA when the iterator starts.
875 * Calling mas_next() could skip the first entry.
877 return mas_find(&vmi->mas, ULONG_MAX);
880 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
882 return mas_prev(&vmi->mas, 0);
885 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
887 return vmi->mas.index;
890 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
892 return vmi->mas.last + 1;
894 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
897 return mas_expected_entries(&vmi->mas, count);
900 /* Free any unused preallocations */
901 static inline void vma_iter_free(struct vma_iterator *vmi)
903 mas_destroy(&vmi->mas);
906 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
907 struct vm_area_struct *vma)
909 vmi->mas.index = vma->vm_start;
910 vmi->mas.last = vma->vm_end - 1;
911 mas_store(&vmi->mas, vma);
912 if (unlikely(mas_is_err(&vmi->mas)))
918 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
920 mas_pause(&vmi->mas);
923 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
925 mas_set(&vmi->mas, addr);
928 #define for_each_vma(__vmi, __vma) \
929 while (((__vma) = vma_next(&(__vmi))) != NULL)
931 /* The MM code likes to work with exclusive end addresses */
932 #define for_each_vma_range(__vmi, __vma, __end) \
933 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
937 * The vma_is_shmem is not inline because it is used only by slow
938 * paths in userfault.
940 bool vma_is_shmem(struct vm_area_struct *vma);
941 bool vma_is_anon_shmem(struct vm_area_struct *vma);
943 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
944 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
947 int vma_is_stack_for_current(struct vm_area_struct *vma);
949 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
950 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
956 * compound_order() can be called without holding a reference, which means
957 * that niceties like page_folio() don't work. These callers should be
958 * prepared to handle wild return values. For example, PG_head may be
959 * set before _folio_order is initialised, or this may be a tail page.
960 * See compaction.c for some good examples.
962 static inline unsigned int compound_order(struct page *page)
964 struct folio *folio = (struct folio *)page;
966 if (!test_bit(PG_head, &folio->flags))
968 return folio->_folio_order;
972 * folio_order - The allocation order of a folio.
975 * A folio is composed of 2^order pages. See get_order() for the definition
978 * Return: The order of the folio.
980 static inline unsigned int folio_order(struct folio *folio)
982 if (!folio_test_large(folio))
984 return folio->_folio_order;
987 #include <linux/huge_mm.h>
990 * Methods to modify the page usage count.
992 * What counts for a page usage:
993 * - cache mapping (page->mapping)
994 * - private data (page->private)
995 * - page mapped in a task's page tables, each mapping
996 * is counted separately
998 * Also, many kernel routines increase the page count before a critical
999 * routine so they can be sure the page doesn't go away from under them.
1003 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1005 static inline int put_page_testzero(struct page *page)
1007 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1008 return page_ref_dec_and_test(page);
1011 static inline int folio_put_testzero(struct folio *folio)
1013 return put_page_testzero(&folio->page);
1017 * Try to grab a ref unless the page has a refcount of zero, return false if
1019 * This can be called when MMU is off so it must not access
1020 * any of the virtual mappings.
1022 static inline bool get_page_unless_zero(struct page *page)
1024 return page_ref_add_unless(page, 1, 0);
1027 static inline struct folio *folio_get_nontail_page(struct page *page)
1029 if (unlikely(!get_page_unless_zero(page)))
1031 return (struct folio *)page;
1034 extern int page_is_ram(unsigned long pfn);
1042 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1043 unsigned long desc);
1045 /* Support for virtually mapped pages */
1046 struct page *vmalloc_to_page(const void *addr);
1047 unsigned long vmalloc_to_pfn(const void *addr);
1050 * Determine if an address is within the vmalloc range
1052 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1053 * is no special casing required.
1056 #ifndef is_ioremap_addr
1057 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1061 extern bool is_vmalloc_addr(const void *x);
1062 extern int is_vmalloc_or_module_addr(const void *x);
1064 static inline bool is_vmalloc_addr(const void *x)
1068 static inline int is_vmalloc_or_module_addr(const void *x)
1075 * How many times the entire folio is mapped as a single unit (eg by a
1076 * PMD or PUD entry). This is probably not what you want, except for
1077 * debugging purposes - it does not include PTE-mapped sub-pages; look
1078 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1080 static inline int folio_entire_mapcount(struct folio *folio)
1082 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1083 return atomic_read(&folio->_entire_mapcount) + 1;
1087 * The atomic page->_mapcount, starts from -1: so that transitions
1088 * both from it and to it can be tracked, using atomic_inc_and_test
1089 * and atomic_add_negative(-1).
1091 static inline void page_mapcount_reset(struct page *page)
1093 atomic_set(&(page)->_mapcount, -1);
1097 * page_mapcount() - Number of times this precise page is mapped.
1100 * The number of times this page is mapped. If this page is part of
1101 * a large folio, it includes the number of times this page is mapped
1102 * as part of that folio.
1104 * The result is undefined for pages which cannot be mapped into userspace.
1105 * For example SLAB or special types of pages. See function page_has_type().
1106 * They use this field in struct page differently.
1108 static inline int page_mapcount(struct page *page)
1110 int mapcount = atomic_read(&page->_mapcount) + 1;
1112 if (unlikely(PageCompound(page)))
1113 mapcount += folio_entire_mapcount(page_folio(page));
1118 int folio_total_mapcount(struct folio *folio);
1121 * folio_mapcount() - Calculate the number of mappings of this folio.
1122 * @folio: The folio.
1124 * A large folio tracks both how many times the entire folio is mapped,
1125 * and how many times each individual page in the folio is mapped.
1126 * This function calculates the total number of times the folio is
1129 * Return: The number of times this folio is mapped.
1131 static inline int folio_mapcount(struct folio *folio)
1133 if (likely(!folio_test_large(folio)))
1134 return atomic_read(&folio->_mapcount) + 1;
1135 return folio_total_mapcount(folio);
1138 static inline int total_mapcount(struct page *page)
1140 if (likely(!PageCompound(page)))
1141 return atomic_read(&page->_mapcount) + 1;
1142 return folio_total_mapcount(page_folio(page));
1145 static inline bool folio_large_is_mapped(struct folio *folio)
1148 * Reading _entire_mapcount below could be omitted if hugetlb
1149 * participated in incrementing nr_pages_mapped when compound mapped.
1151 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1152 atomic_read(&folio->_entire_mapcount) >= 0;
1156 * folio_mapped - Is this folio mapped into userspace?
1157 * @folio: The folio.
1159 * Return: True if any page in this folio is referenced by user page tables.
1161 static inline bool folio_mapped(struct folio *folio)
1163 if (likely(!folio_test_large(folio)))
1164 return atomic_read(&folio->_mapcount) >= 0;
1165 return folio_large_is_mapped(folio);
1169 * Return true if this page is mapped into pagetables.
1170 * For compound page it returns true if any sub-page of compound page is mapped,
1171 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1173 static inline bool page_mapped(struct page *page)
1175 if (likely(!PageCompound(page)))
1176 return atomic_read(&page->_mapcount) >= 0;
1177 return folio_large_is_mapped(page_folio(page));
1180 static inline struct page *virt_to_head_page(const void *x)
1182 struct page *page = virt_to_page(x);
1184 return compound_head(page);
1187 static inline struct folio *virt_to_folio(const void *x)
1189 struct page *page = virt_to_page(x);
1191 return page_folio(page);
1194 void __folio_put(struct folio *folio);
1196 void put_pages_list(struct list_head *pages);
1198 void split_page(struct page *page, unsigned int order);
1199 void folio_copy(struct folio *dst, struct folio *src);
1201 unsigned long nr_free_buffer_pages(void);
1204 * Compound pages have a destructor function. Provide a
1205 * prototype for that function and accessor functions.
1206 * These are _only_ valid on the head of a compound page.
1208 typedef void compound_page_dtor(struct page *);
1210 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1211 enum compound_dtor_id {
1214 #ifdef CONFIG_HUGETLB_PAGE
1217 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1218 TRANSHUGE_PAGE_DTOR,
1222 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1224 static inline void set_compound_page_dtor(struct page *page,
1225 enum compound_dtor_id compound_dtor)
1227 struct folio *folio = (struct folio *)page;
1229 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1230 VM_BUG_ON_PAGE(!PageHead(page), page);
1231 folio->_folio_dtor = compound_dtor;
1234 static inline void folio_set_compound_dtor(struct folio *folio,
1235 enum compound_dtor_id compound_dtor)
1237 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1238 folio->_folio_dtor = compound_dtor;
1241 void destroy_large_folio(struct folio *folio);
1243 static inline void set_compound_order(struct page *page, unsigned int order)
1245 struct folio *folio = (struct folio *)page;
1247 folio->_folio_order = order;
1249 folio->_folio_nr_pages = 1U << order;
1253 /* Returns the number of bytes in this potentially compound page. */
1254 static inline unsigned long page_size(struct page *page)
1256 return PAGE_SIZE << compound_order(page);
1259 /* Returns the number of bits needed for the number of bytes in a page */
1260 static inline unsigned int page_shift(struct page *page)
1262 return PAGE_SHIFT + compound_order(page);
1266 * thp_order - Order of a transparent huge page.
1267 * @page: Head page of a transparent huge page.
1269 static inline unsigned int thp_order(struct page *page)
1271 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1272 return compound_order(page);
1276 * thp_size - Size of a transparent huge page.
1277 * @page: Head page of a transparent huge page.
1279 * Return: Number of bytes in this page.
1281 static inline unsigned long thp_size(struct page *page)
1283 return PAGE_SIZE << thp_order(page);
1286 void free_compound_page(struct page *page);
1290 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1291 * servicing faults for write access. In the normal case, do always want
1292 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1293 * that do not have writing enabled, when used by access_process_vm.
1295 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1297 if (likely(vma->vm_flags & VM_WRITE))
1298 pte = pte_mkwrite(pte);
1302 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1303 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1305 vm_fault_t finish_fault(struct vm_fault *vmf);
1306 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1310 * Multiple processes may "see" the same page. E.g. for untouched
1311 * mappings of /dev/null, all processes see the same page full of
1312 * zeroes, and text pages of executables and shared libraries have
1313 * only one copy in memory, at most, normally.
1315 * For the non-reserved pages, page_count(page) denotes a reference count.
1316 * page_count() == 0 means the page is free. page->lru is then used for
1317 * freelist management in the buddy allocator.
1318 * page_count() > 0 means the page has been allocated.
1320 * Pages are allocated by the slab allocator in order to provide memory
1321 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1322 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1323 * unless a particular usage is carefully commented. (the responsibility of
1324 * freeing the kmalloc memory is the caller's, of course).
1326 * A page may be used by anyone else who does a __get_free_page().
1327 * In this case, page_count still tracks the references, and should only
1328 * be used through the normal accessor functions. The top bits of page->flags
1329 * and page->virtual store page management information, but all other fields
1330 * are unused and could be used privately, carefully. The management of this
1331 * page is the responsibility of the one who allocated it, and those who have
1332 * subsequently been given references to it.
1334 * The other pages (we may call them "pagecache pages") are completely
1335 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1336 * The following discussion applies only to them.
1338 * A pagecache page contains an opaque `private' member, which belongs to the
1339 * page's address_space. Usually, this is the address of a circular list of
1340 * the page's disk buffers. PG_private must be set to tell the VM to call
1341 * into the filesystem to release these pages.
1343 * A page may belong to an inode's memory mapping. In this case, page->mapping
1344 * is the pointer to the inode, and page->index is the file offset of the page,
1345 * in units of PAGE_SIZE.
1347 * If pagecache pages are not associated with an inode, they are said to be
1348 * anonymous pages. These may become associated with the swapcache, and in that
1349 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1351 * In either case (swapcache or inode backed), the pagecache itself holds one
1352 * reference to the page. Setting PG_private should also increment the
1353 * refcount. The each user mapping also has a reference to the page.
1355 * The pagecache pages are stored in a per-mapping radix tree, which is
1356 * rooted at mapping->i_pages, and indexed by offset.
1357 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1358 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1360 * All pagecache pages may be subject to I/O:
1361 * - inode pages may need to be read from disk,
1362 * - inode pages which have been modified and are MAP_SHARED may need
1363 * to be written back to the inode on disk,
1364 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1365 * modified may need to be swapped out to swap space and (later) to be read
1369 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1370 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1372 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1373 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1375 if (!static_branch_unlikely(&devmap_managed_key))
1377 if (!is_zone_device_page(page))
1379 return __put_devmap_managed_page_refs(page, refs);
1381 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1382 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1386 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1388 static inline bool put_devmap_managed_page(struct page *page)
1390 return put_devmap_managed_page_refs(page, 1);
1393 /* 127: arbitrary random number, small enough to assemble well */
1394 #define folio_ref_zero_or_close_to_overflow(folio) \
1395 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1398 * folio_get - Increment the reference count on a folio.
1399 * @folio: The folio.
1401 * Context: May be called in any context, as long as you know that
1402 * you have a refcount on the folio. If you do not already have one,
1403 * folio_try_get() may be the right interface for you to use.
1405 static inline void folio_get(struct folio *folio)
1407 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1408 folio_ref_inc(folio);
1411 static inline void get_page(struct page *page)
1413 folio_get(page_folio(page));
1416 static inline __must_check bool try_get_page(struct page *page)
1418 page = compound_head(page);
1419 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1426 * folio_put - Decrement the reference count on a folio.
1427 * @folio: The folio.
1429 * If the folio's reference count reaches zero, the memory will be
1430 * released back to the page allocator and may be used by another
1431 * allocation immediately. Do not access the memory or the struct folio
1432 * after calling folio_put() unless you can be sure that it wasn't the
1435 * Context: May be called in process or interrupt context, but not in NMI
1436 * context. May be called while holding a spinlock.
1438 static inline void folio_put(struct folio *folio)
1440 if (folio_put_testzero(folio))
1445 * folio_put_refs - Reduce the reference count on a folio.
1446 * @folio: The folio.
1447 * @refs: The amount to subtract from the folio's reference count.
1449 * If the folio's reference count reaches zero, the memory will be
1450 * released back to the page allocator and may be used by another
1451 * allocation immediately. Do not access the memory or the struct folio
1452 * after calling folio_put_refs() unless you can be sure that these weren't
1453 * the last references.
1455 * Context: May be called in process or interrupt context, but not in NMI
1456 * context. May be called while holding a spinlock.
1458 static inline void folio_put_refs(struct folio *folio, int refs)
1460 if (folio_ref_sub_and_test(folio, refs))
1465 * union release_pages_arg - an array of pages or folios
1467 * release_pages() releases a simple array of multiple pages, and
1468 * accepts various different forms of said page array: either
1469 * a regular old boring array of pages, an array of folios, or
1470 * an array of encoded page pointers.
1472 * The transparent union syntax for this kind of "any of these
1473 * argument types" is all kinds of ugly, so look away.
1476 struct page **pages;
1477 struct folio **folios;
1478 struct encoded_page **encoded_pages;
1479 } release_pages_arg __attribute__ ((__transparent_union__));
1481 void release_pages(release_pages_arg, int nr);
1484 * folios_put - Decrement the reference count on an array of folios.
1485 * @folios: The folios.
1486 * @nr: How many folios there are.
1488 * Like folio_put(), but for an array of folios. This is more efficient
1489 * than writing the loop yourself as it will optimise the locks which
1490 * need to be taken if the folios are freed.
1492 * Context: May be called in process or interrupt context, but not in NMI
1493 * context. May be called while holding a spinlock.
1495 static inline void folios_put(struct folio **folios, unsigned int nr)
1497 release_pages(folios, nr);
1500 static inline void put_page(struct page *page)
1502 struct folio *folio = page_folio(page);
1505 * For some devmap managed pages we need to catch refcount transition
1508 if (put_devmap_managed_page(&folio->page))
1514 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1515 * the page's refcount so that two separate items are tracked: the original page
1516 * reference count, and also a new count of how many pin_user_pages() calls were
1517 * made against the page. ("gup-pinned" is another term for the latter).
1519 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1520 * distinct from normal pages. As such, the unpin_user_page() call (and its
1521 * variants) must be used in order to release gup-pinned pages.
1525 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1526 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1527 * simpler, due to the fact that adding an even power of two to the page
1528 * refcount has the effect of using only the upper N bits, for the code that
1529 * counts up using the bias value. This means that the lower bits are left for
1530 * the exclusive use of the original code that increments and decrements by one
1531 * (or at least, by much smaller values than the bias value).
1533 * Of course, once the lower bits overflow into the upper bits (and this is
1534 * OK, because subtraction recovers the original values), then visual inspection
1535 * no longer suffices to directly view the separate counts. However, for normal
1536 * applications that don't have huge page reference counts, this won't be an
1539 * Locking: the lockless algorithm described in folio_try_get_rcu()
1540 * provides safe operation for get_user_pages(), page_mkclean() and
1541 * other calls that race to set up page table entries.
1543 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1545 void unpin_user_page(struct page *page);
1546 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1548 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1550 void unpin_user_pages(struct page **pages, unsigned long npages);
1552 static inline bool is_cow_mapping(vm_flags_t flags)
1554 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1558 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1561 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1562 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1563 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1564 * underlying memory if ptrace is active, so this is only possible if
1565 * ptrace does not apply. Note that there is no mprotect() to upgrade
1566 * write permissions later.
1568 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1572 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1573 #define SECTION_IN_PAGE_FLAGS
1577 * The identification function is mainly used by the buddy allocator for
1578 * determining if two pages could be buddies. We are not really identifying
1579 * the zone since we could be using the section number id if we do not have
1580 * node id available in page flags.
1581 * We only guarantee that it will return the same value for two combinable
1584 static inline int page_zone_id(struct page *page)
1586 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1589 #ifdef NODE_NOT_IN_PAGE_FLAGS
1590 extern int page_to_nid(const struct page *page);
1592 static inline int page_to_nid(const struct page *page)
1594 struct page *p = (struct page *)page;
1596 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1600 static inline int folio_nid(const struct folio *folio)
1602 return page_to_nid(&folio->page);
1605 #ifdef CONFIG_NUMA_BALANCING
1606 /* page access time bits needs to hold at least 4 seconds */
1607 #define PAGE_ACCESS_TIME_MIN_BITS 12
1608 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1609 #define PAGE_ACCESS_TIME_BUCKETS \
1610 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1612 #define PAGE_ACCESS_TIME_BUCKETS 0
1615 #define PAGE_ACCESS_TIME_MASK \
1616 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1618 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1620 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1623 static inline int cpupid_to_pid(int cpupid)
1625 return cpupid & LAST__PID_MASK;
1628 static inline int cpupid_to_cpu(int cpupid)
1630 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1633 static inline int cpupid_to_nid(int cpupid)
1635 return cpu_to_node(cpupid_to_cpu(cpupid));
1638 static inline bool cpupid_pid_unset(int cpupid)
1640 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1643 static inline bool cpupid_cpu_unset(int cpupid)
1645 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1648 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1650 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1653 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1654 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1655 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1657 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1660 static inline int page_cpupid_last(struct page *page)
1662 return page->_last_cpupid;
1664 static inline void page_cpupid_reset_last(struct page *page)
1666 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1669 static inline int page_cpupid_last(struct page *page)
1671 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1674 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1676 static inline void page_cpupid_reset_last(struct page *page)
1678 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1680 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1682 static inline int xchg_page_access_time(struct page *page, int time)
1686 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1687 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1690 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1692 unsigned int pid_bit;
1694 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1695 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) {
1696 __set_bit(pid_bit, &vma->numab_state->access_pids[1]);
1699 #else /* !CONFIG_NUMA_BALANCING */
1700 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1702 return page_to_nid(page); /* XXX */
1705 static inline int xchg_page_access_time(struct page *page, int time)
1710 static inline int page_cpupid_last(struct page *page)
1712 return page_to_nid(page); /* XXX */
1715 static inline int cpupid_to_nid(int cpupid)
1720 static inline int cpupid_to_pid(int cpupid)
1725 static inline int cpupid_to_cpu(int cpupid)
1730 static inline int cpu_pid_to_cpupid(int nid, int pid)
1735 static inline bool cpupid_pid_unset(int cpupid)
1740 static inline void page_cpupid_reset_last(struct page *page)
1744 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1749 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1752 #endif /* CONFIG_NUMA_BALANCING */
1754 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1757 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1758 * setting tags for all pages to native kernel tag value 0xff, as the default
1759 * value 0x00 maps to 0xff.
1762 static inline u8 page_kasan_tag(const struct page *page)
1766 if (kasan_enabled()) {
1767 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1774 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1776 unsigned long old_flags, flags;
1778 if (!kasan_enabled())
1782 old_flags = READ_ONCE(page->flags);
1785 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1786 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1787 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1790 static inline void page_kasan_tag_reset(struct page *page)
1792 if (kasan_enabled())
1793 page_kasan_tag_set(page, 0xff);
1796 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1798 static inline u8 page_kasan_tag(const struct page *page)
1803 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1804 static inline void page_kasan_tag_reset(struct page *page) { }
1806 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1808 static inline struct zone *page_zone(const struct page *page)
1810 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1813 static inline pg_data_t *page_pgdat(const struct page *page)
1815 return NODE_DATA(page_to_nid(page));
1818 static inline struct zone *folio_zone(const struct folio *folio)
1820 return page_zone(&folio->page);
1823 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1825 return page_pgdat(&folio->page);
1828 #ifdef SECTION_IN_PAGE_FLAGS
1829 static inline void set_page_section(struct page *page, unsigned long section)
1831 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1832 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1835 static inline unsigned long page_to_section(const struct page *page)
1837 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1842 * folio_pfn - Return the Page Frame Number of a folio.
1843 * @folio: The folio.
1845 * A folio may contain multiple pages. The pages have consecutive
1846 * Page Frame Numbers.
1848 * Return: The Page Frame Number of the first page in the folio.
1850 static inline unsigned long folio_pfn(struct folio *folio)
1852 return page_to_pfn(&folio->page);
1855 static inline struct folio *pfn_folio(unsigned long pfn)
1857 return page_folio(pfn_to_page(pfn));
1861 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1862 * @folio: The folio.
1864 * This function checks if a folio has been pinned via a call to
1865 * a function in the pin_user_pages() family.
1867 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1868 * because it means "definitely not pinned for DMA", but true means "probably
1869 * pinned for DMA, but possibly a false positive due to having at least
1870 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1872 * False positives are OK, because: a) it's unlikely for a folio to
1873 * get that many refcounts, and b) all the callers of this routine are
1874 * expected to be able to deal gracefully with a false positive.
1876 * For large folios, the result will be exactly correct. That's because
1877 * we have more tracking data available: the _pincount field is used
1878 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1880 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1882 * Return: True, if it is likely that the page has been "dma-pinned".
1883 * False, if the page is definitely not dma-pinned.
1885 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1887 if (folio_test_large(folio))
1888 return atomic_read(&folio->_pincount) > 0;
1891 * folio_ref_count() is signed. If that refcount overflows, then
1892 * folio_ref_count() returns a negative value, and callers will avoid
1893 * further incrementing the refcount.
1895 * Here, for that overflow case, use the sign bit to count a little
1896 * bit higher via unsigned math, and thus still get an accurate result.
1898 return ((unsigned int)folio_ref_count(folio)) >=
1899 GUP_PIN_COUNTING_BIAS;
1902 static inline bool page_maybe_dma_pinned(struct page *page)
1904 return folio_maybe_dma_pinned(page_folio(page));
1908 * This should most likely only be called during fork() to see whether we
1909 * should break the cow immediately for an anon page on the src mm.
1911 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1913 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1916 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1918 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1921 return page_maybe_dma_pinned(page);
1924 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1925 #ifdef CONFIG_MIGRATION
1926 static inline bool is_longterm_pinnable_page(struct page *page)
1929 int mt = get_pageblock_migratetype(page);
1931 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1934 /* The zero page may always be pinned */
1935 if (is_zero_pfn(page_to_pfn(page)))
1938 /* Coherent device memory must always allow eviction. */
1939 if (is_device_coherent_page(page))
1942 /* Otherwise, non-movable zone pages can be pinned. */
1943 return !is_zone_movable_page(page);
1946 static inline bool is_longterm_pinnable_page(struct page *page)
1952 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1954 return is_longterm_pinnable_page(&folio->page);
1957 static inline void set_page_zone(struct page *page, enum zone_type zone)
1959 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1960 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1963 static inline void set_page_node(struct page *page, unsigned long node)
1965 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1966 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1969 static inline void set_page_links(struct page *page, enum zone_type zone,
1970 unsigned long node, unsigned long pfn)
1972 set_page_zone(page, zone);
1973 set_page_node(page, node);
1974 #ifdef SECTION_IN_PAGE_FLAGS
1975 set_page_section(page, pfn_to_section_nr(pfn));
1980 * folio_nr_pages - The number of pages in the folio.
1981 * @folio: The folio.
1983 * Return: A positive power of two.
1985 static inline long folio_nr_pages(struct folio *folio)
1987 if (!folio_test_large(folio))
1990 return folio->_folio_nr_pages;
1992 return 1L << folio->_folio_order;
1997 * compound_nr() returns the number of pages in this potentially compound
1998 * page. compound_nr() can be called on a tail page, and is defined to
1999 * return 1 in that case.
2001 static inline unsigned long compound_nr(struct page *page)
2003 struct folio *folio = (struct folio *)page;
2005 if (!test_bit(PG_head, &folio->flags))
2008 return folio->_folio_nr_pages;
2010 return 1L << folio->_folio_order;
2015 * thp_nr_pages - The number of regular pages in this huge page.
2016 * @page: The head page of a huge page.
2018 static inline int thp_nr_pages(struct page *page)
2020 return folio_nr_pages((struct folio *)page);
2024 * folio_next - Move to the next physical folio.
2025 * @folio: The folio we're currently operating on.
2027 * If you have physically contiguous memory which may span more than
2028 * one folio (eg a &struct bio_vec), use this function to move from one
2029 * folio to the next. Do not use it if the memory is only virtually
2030 * contiguous as the folios are almost certainly not adjacent to each
2031 * other. This is the folio equivalent to writing ``page++``.
2033 * Context: We assume that the folios are refcounted and/or locked at a
2034 * higher level and do not adjust the reference counts.
2035 * Return: The next struct folio.
2037 static inline struct folio *folio_next(struct folio *folio)
2039 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2043 * folio_shift - The size of the memory described by this folio.
2044 * @folio: The folio.
2046 * A folio represents a number of bytes which is a power-of-two in size.
2047 * This function tells you which power-of-two the folio is. See also
2048 * folio_size() and folio_order().
2050 * Context: The caller should have a reference on the folio to prevent
2051 * it from being split. It is not necessary for the folio to be locked.
2052 * Return: The base-2 logarithm of the size of this folio.
2054 static inline unsigned int folio_shift(struct folio *folio)
2056 return PAGE_SHIFT + folio_order(folio);
2060 * folio_size - The number of bytes in a folio.
2061 * @folio: The folio.
2063 * Context: The caller should have a reference on the folio to prevent
2064 * it from being split. It is not necessary for the folio to be locked.
2065 * Return: The number of bytes in this folio.
2067 static inline size_t folio_size(struct folio *folio)
2069 return PAGE_SIZE << folio_order(folio);
2073 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2074 * @folio: The folio.
2076 * folio_estimated_sharers() aims to serve as a function to efficiently
2077 * estimate the number of processes sharing a folio. This is done by
2078 * looking at the precise mapcount of the first subpage in the folio, and
2079 * assuming the other subpages are the same. This may not be true for large
2080 * folios. If you want exact mapcounts for exact calculations, look at
2081 * page_mapcount() or folio_total_mapcount().
2083 * Return: The estimated number of processes sharing a folio.
2085 static inline int folio_estimated_sharers(struct folio *folio)
2087 return page_mapcount(folio_page(folio, 0));
2090 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2091 static inline int arch_make_page_accessible(struct page *page)
2097 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2098 static inline int arch_make_folio_accessible(struct folio *folio)
2101 long i, nr = folio_nr_pages(folio);
2103 for (i = 0; i < nr; i++) {
2104 ret = arch_make_page_accessible(folio_page(folio, i));
2114 * Some inline functions in vmstat.h depend on page_zone()
2116 #include <linux/vmstat.h>
2118 static __always_inline void *lowmem_page_address(const struct page *page)
2120 return page_to_virt(page);
2123 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2124 #define HASHED_PAGE_VIRTUAL
2127 #if defined(WANT_PAGE_VIRTUAL)
2128 static inline void *page_address(const struct page *page)
2130 return page->virtual;
2132 static inline void set_page_address(struct page *page, void *address)
2134 page->virtual = address;
2136 #define page_address_init() do { } while(0)
2139 #if defined(HASHED_PAGE_VIRTUAL)
2140 void *page_address(const struct page *page);
2141 void set_page_address(struct page *page, void *virtual);
2142 void page_address_init(void);
2145 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2146 #define page_address(page) lowmem_page_address(page)
2147 #define set_page_address(page, address) do { } while(0)
2148 #define page_address_init() do { } while(0)
2151 static inline void *folio_address(const struct folio *folio)
2153 return page_address(&folio->page);
2156 extern void *page_rmapping(struct page *page);
2157 extern pgoff_t __page_file_index(struct page *page);
2160 * Return the pagecache index of the passed page. Regular pagecache pages
2161 * use ->index whereas swapcache pages use swp_offset(->private)
2163 static inline pgoff_t page_index(struct page *page)
2165 if (unlikely(PageSwapCache(page)))
2166 return __page_file_index(page);
2171 * Return true only if the page has been allocated with
2172 * ALLOC_NO_WATERMARKS and the low watermark was not
2173 * met implying that the system is under some pressure.
2175 static inline bool page_is_pfmemalloc(const struct page *page)
2178 * lru.next has bit 1 set if the page is allocated from the
2179 * pfmemalloc reserves. Callers may simply overwrite it if
2180 * they do not need to preserve that information.
2182 return (uintptr_t)page->lru.next & BIT(1);
2186 * Return true only if the folio has been allocated with
2187 * ALLOC_NO_WATERMARKS and the low watermark was not
2188 * met implying that the system is under some pressure.
2190 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2193 * lru.next has bit 1 set if the page is allocated from the
2194 * pfmemalloc reserves. Callers may simply overwrite it if
2195 * they do not need to preserve that information.
2197 return (uintptr_t)folio->lru.next & BIT(1);
2201 * Only to be called by the page allocator on a freshly allocated
2204 static inline void set_page_pfmemalloc(struct page *page)
2206 page->lru.next = (void *)BIT(1);
2209 static inline void clear_page_pfmemalloc(struct page *page)
2211 page->lru.next = NULL;
2215 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2217 extern void pagefault_out_of_memory(void);
2219 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2220 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2221 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2224 * Flags passed to show_mem() and show_free_areas() to suppress output in
2227 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2229 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2230 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2232 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2236 * Parameter block passed down to zap_pte_range in exceptional cases.
2238 struct zap_details {
2239 struct folio *single_folio; /* Locked folio to be unmapped */
2240 bool even_cows; /* Zap COWed private pages too? */
2241 zap_flags_t zap_flags; /* Extra flags for zapping */
2245 * Whether to drop the pte markers, for example, the uffd-wp information for
2246 * file-backed memory. This should only be specified when we will completely
2247 * drop the page in the mm, either by truncation or unmapping of the vma. By
2248 * default, the flag is not set.
2250 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2251 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2252 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2254 #ifdef CONFIG_SCHED_MM_CID
2255 void sched_mm_cid_before_execve(struct task_struct *t);
2256 void sched_mm_cid_after_execve(struct task_struct *t);
2257 void sched_mm_cid_fork(struct task_struct *t);
2258 void sched_mm_cid_exit_signals(struct task_struct *t);
2259 static inline int task_mm_cid(struct task_struct *t)
2264 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2265 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2266 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2267 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2268 static inline int task_mm_cid(struct task_struct *t)
2271 * Use the processor id as a fall-back when the mm cid feature is
2272 * disabled. This provides functional per-cpu data structure accesses
2273 * in user-space, althrough it won't provide the memory usage benefits.
2275 return raw_smp_processor_id();
2280 extern bool can_do_mlock(void);
2282 static inline bool can_do_mlock(void) { return false; }
2284 extern int user_shm_lock(size_t, struct ucounts *);
2285 extern void user_shm_unlock(size_t, struct ucounts *);
2287 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2289 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2291 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2294 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2295 unsigned long size);
2296 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2297 unsigned long size, struct zap_details *details);
2298 static inline void zap_vma_pages(struct vm_area_struct *vma)
2300 zap_page_range_single(vma, vma->vm_start,
2301 vma->vm_end - vma->vm_start, NULL);
2303 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2304 struct vm_area_struct *start_vma, unsigned long start,
2305 unsigned long end, bool mm_wr_locked);
2307 struct mmu_notifier_range;
2309 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2310 unsigned long end, unsigned long floor, unsigned long ceiling);
2312 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2313 int follow_pte(struct mm_struct *mm, unsigned long address,
2314 pte_t **ptepp, spinlock_t **ptlp);
2315 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2316 unsigned long *pfn);
2317 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2318 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2319 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2320 void *buf, int len, int write);
2322 extern void truncate_pagecache(struct inode *inode, loff_t new);
2323 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2324 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2325 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2326 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2329 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2330 unsigned long address, unsigned int flags,
2331 struct pt_regs *regs);
2332 extern int fixup_user_fault(struct mm_struct *mm,
2333 unsigned long address, unsigned int fault_flags,
2335 void unmap_mapping_pages(struct address_space *mapping,
2336 pgoff_t start, pgoff_t nr, bool even_cows);
2337 void unmap_mapping_range(struct address_space *mapping,
2338 loff_t const holebegin, loff_t const holelen, int even_cows);
2340 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2341 unsigned long address, unsigned int flags,
2342 struct pt_regs *regs)
2344 /* should never happen if there's no MMU */
2346 return VM_FAULT_SIGBUS;
2348 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2349 unsigned int fault_flags, bool *unlocked)
2351 /* should never happen if there's no MMU */
2355 static inline void unmap_mapping_pages(struct address_space *mapping,
2356 pgoff_t start, pgoff_t nr, bool even_cows) { }
2357 static inline void unmap_mapping_range(struct address_space *mapping,
2358 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2361 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2362 loff_t const holebegin, loff_t const holelen)
2364 unmap_mapping_range(mapping, holebegin, holelen, 0);
2367 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2368 void *buf, int len, unsigned int gup_flags);
2369 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2370 void *buf, int len, unsigned int gup_flags);
2371 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2372 void *buf, int len, unsigned int gup_flags);
2374 long get_user_pages_remote(struct mm_struct *mm,
2375 unsigned long start, unsigned long nr_pages,
2376 unsigned int gup_flags, struct page **pages,
2377 struct vm_area_struct **vmas, int *locked);
2378 long pin_user_pages_remote(struct mm_struct *mm,
2379 unsigned long start, unsigned long nr_pages,
2380 unsigned int gup_flags, struct page **pages,
2381 struct vm_area_struct **vmas, int *locked);
2382 long get_user_pages(unsigned long start, unsigned long nr_pages,
2383 unsigned int gup_flags, struct page **pages,
2384 struct vm_area_struct **vmas);
2385 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2386 unsigned int gup_flags, struct page **pages,
2387 struct vm_area_struct **vmas);
2388 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2389 struct page **pages, unsigned int gup_flags);
2390 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2391 struct page **pages, unsigned int gup_flags);
2393 int get_user_pages_fast(unsigned long start, int nr_pages,
2394 unsigned int gup_flags, struct page **pages);
2395 int pin_user_pages_fast(unsigned long start, int nr_pages,
2396 unsigned int gup_flags, struct page **pages);
2398 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2399 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2400 struct task_struct *task, bool bypass_rlim);
2403 struct page *get_dump_page(unsigned long addr);
2405 bool folio_mark_dirty(struct folio *folio);
2406 bool set_page_dirty(struct page *page);
2407 int set_page_dirty_lock(struct page *page);
2409 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2411 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2412 unsigned long old_addr, struct vm_area_struct *new_vma,
2413 unsigned long new_addr, unsigned long len,
2414 bool need_rmap_locks);
2417 * Flags used by change_protection(). For now we make it a bitmap so
2418 * that we can pass in multiple flags just like parameters. However
2419 * for now all the callers are only use one of the flags at the same
2423 * Whether we should manually check if we can map individual PTEs writable,
2424 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2425 * PTEs automatically in a writable mapping.
2427 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2428 /* Whether this protection change is for NUMA hints */
2429 #define MM_CP_PROT_NUMA (1UL << 1)
2430 /* Whether this change is for write protecting */
2431 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2432 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2433 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2434 MM_CP_UFFD_WP_RESOLVE)
2436 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2437 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2440 * We want to check manually if we can change individual PTEs writable
2441 * if we can't do that automatically for all PTEs in a mapping. For
2442 * private mappings, that's always the case when we have write
2443 * permissions as we properly have to handle COW.
2445 if (vma->vm_flags & VM_SHARED)
2446 return vma_wants_writenotify(vma, vma->vm_page_prot);
2447 return !!(vma->vm_flags & VM_WRITE);
2450 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2452 extern long change_protection(struct mmu_gather *tlb,
2453 struct vm_area_struct *vma, unsigned long start,
2454 unsigned long end, unsigned long cp_flags);
2455 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2456 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2457 unsigned long start, unsigned long end, unsigned long newflags);
2460 * doesn't attempt to fault and will return short.
2462 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2463 unsigned int gup_flags, struct page **pages);
2465 static inline bool get_user_page_fast_only(unsigned long addr,
2466 unsigned int gup_flags, struct page **pagep)
2468 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2471 * per-process(per-mm_struct) statistics.
2473 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2475 return percpu_counter_read_positive(&mm->rss_stat[member]);
2478 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2480 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2482 percpu_counter_add(&mm->rss_stat[member], value);
2484 mm_trace_rss_stat(mm, member);
2487 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2489 percpu_counter_inc(&mm->rss_stat[member]);
2491 mm_trace_rss_stat(mm, member);
2494 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2496 percpu_counter_dec(&mm->rss_stat[member]);
2498 mm_trace_rss_stat(mm, member);
2501 /* Optimized variant when page is already known not to be PageAnon */
2502 static inline int mm_counter_file(struct page *page)
2504 if (PageSwapBacked(page))
2505 return MM_SHMEMPAGES;
2506 return MM_FILEPAGES;
2509 static inline int mm_counter(struct page *page)
2512 return MM_ANONPAGES;
2513 return mm_counter_file(page);
2516 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2518 return get_mm_counter(mm, MM_FILEPAGES) +
2519 get_mm_counter(mm, MM_ANONPAGES) +
2520 get_mm_counter(mm, MM_SHMEMPAGES);
2523 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2525 return max(mm->hiwater_rss, get_mm_rss(mm));
2528 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2530 return max(mm->hiwater_vm, mm->total_vm);
2533 static inline void update_hiwater_rss(struct mm_struct *mm)
2535 unsigned long _rss = get_mm_rss(mm);
2537 if ((mm)->hiwater_rss < _rss)
2538 (mm)->hiwater_rss = _rss;
2541 static inline void update_hiwater_vm(struct mm_struct *mm)
2543 if (mm->hiwater_vm < mm->total_vm)
2544 mm->hiwater_vm = mm->total_vm;
2547 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2549 mm->hiwater_rss = get_mm_rss(mm);
2552 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2553 struct mm_struct *mm)
2555 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2557 if (*maxrss < hiwater_rss)
2558 *maxrss = hiwater_rss;
2561 #if defined(SPLIT_RSS_COUNTING)
2562 void sync_mm_rss(struct mm_struct *mm);
2564 static inline void sync_mm_rss(struct mm_struct *mm)
2569 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2570 static inline int pte_special(pte_t pte)
2575 static inline pte_t pte_mkspecial(pte_t pte)
2581 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2582 static inline int pte_devmap(pte_t pte)
2588 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2590 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2594 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2598 #ifdef __PAGETABLE_P4D_FOLDED
2599 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2600 unsigned long address)
2605 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2608 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2609 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2610 unsigned long address)
2614 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2615 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2618 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2620 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2622 if (mm_pud_folded(mm))
2624 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2627 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2629 if (mm_pud_folded(mm))
2631 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2635 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2636 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2637 unsigned long address)
2642 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2643 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2646 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2648 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2650 if (mm_pmd_folded(mm))
2652 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2655 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2657 if (mm_pmd_folded(mm))
2659 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2664 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2666 atomic_long_set(&mm->pgtables_bytes, 0);
2669 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2671 return atomic_long_read(&mm->pgtables_bytes);
2674 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2676 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2679 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2681 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2685 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2686 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2691 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2692 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2695 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2696 int __pte_alloc_kernel(pmd_t *pmd);
2698 #if defined(CONFIG_MMU)
2700 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2701 unsigned long address)
2703 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2704 NULL : p4d_offset(pgd, address);
2707 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2708 unsigned long address)
2710 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2711 NULL : pud_offset(p4d, address);
2714 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2716 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2717 NULL: pmd_offset(pud, address);
2719 #endif /* CONFIG_MMU */
2721 #if USE_SPLIT_PTE_PTLOCKS
2722 #if ALLOC_SPLIT_PTLOCKS
2723 void __init ptlock_cache_init(void);
2724 extern bool ptlock_alloc(struct page *page);
2725 extern void ptlock_free(struct page *page);
2727 static inline spinlock_t *ptlock_ptr(struct page *page)
2731 #else /* ALLOC_SPLIT_PTLOCKS */
2732 static inline void ptlock_cache_init(void)
2736 static inline bool ptlock_alloc(struct page *page)
2741 static inline void ptlock_free(struct page *page)
2745 static inline spinlock_t *ptlock_ptr(struct page *page)
2749 #endif /* ALLOC_SPLIT_PTLOCKS */
2751 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2753 return ptlock_ptr(pmd_page(*pmd));
2756 static inline bool ptlock_init(struct page *page)
2759 * prep_new_page() initialize page->private (and therefore page->ptl)
2760 * with 0. Make sure nobody took it in use in between.
2762 * It can happen if arch try to use slab for page table allocation:
2763 * slab code uses page->slab_cache, which share storage with page->ptl.
2765 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2766 if (!ptlock_alloc(page))
2768 spin_lock_init(ptlock_ptr(page));
2772 #else /* !USE_SPLIT_PTE_PTLOCKS */
2774 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2776 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2778 return &mm->page_table_lock;
2780 static inline void ptlock_cache_init(void) {}
2781 static inline bool ptlock_init(struct page *page) { return true; }
2782 static inline void ptlock_free(struct page *page) {}
2783 #endif /* USE_SPLIT_PTE_PTLOCKS */
2785 static inline bool pgtable_pte_page_ctor(struct page *page)
2787 if (!ptlock_init(page))
2789 __SetPageTable(page);
2790 inc_lruvec_page_state(page, NR_PAGETABLE);
2794 static inline void pgtable_pte_page_dtor(struct page *page)
2797 __ClearPageTable(page);
2798 dec_lruvec_page_state(page, NR_PAGETABLE);
2801 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2803 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2804 pte_t *__pte = pte_offset_map(pmd, address); \
2810 #define pte_unmap_unlock(pte, ptl) do { \
2815 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2817 #define pte_alloc_map(mm, pmd, address) \
2818 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2820 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2821 (pte_alloc(mm, pmd) ? \
2822 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2824 #define pte_alloc_kernel(pmd, address) \
2825 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2826 NULL: pte_offset_kernel(pmd, address))
2828 #if USE_SPLIT_PMD_PTLOCKS
2830 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2832 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2833 return virt_to_page((void *)((unsigned long) pmd & mask));
2836 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2838 return ptlock_ptr(pmd_pgtable_page(pmd));
2841 static inline bool pmd_ptlock_init(struct page *page)
2843 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2844 page->pmd_huge_pte = NULL;
2846 return ptlock_init(page);
2849 static inline void pmd_ptlock_free(struct page *page)
2851 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2852 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2857 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2861 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2863 return &mm->page_table_lock;
2866 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2867 static inline void pmd_ptlock_free(struct page *page) {}
2869 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2873 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2875 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2880 static inline bool pgtable_pmd_page_ctor(struct page *page)
2882 if (!pmd_ptlock_init(page))
2884 __SetPageTable(page);
2885 inc_lruvec_page_state(page, NR_PAGETABLE);
2889 static inline void pgtable_pmd_page_dtor(struct page *page)
2891 pmd_ptlock_free(page);
2892 __ClearPageTable(page);
2893 dec_lruvec_page_state(page, NR_PAGETABLE);
2897 * No scalability reason to split PUD locks yet, but follow the same pattern
2898 * as the PMD locks to make it easier if we decide to. The VM should not be
2899 * considered ready to switch to split PUD locks yet; there may be places
2900 * which need to be converted from page_table_lock.
2902 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2904 return &mm->page_table_lock;
2907 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2909 spinlock_t *ptl = pud_lockptr(mm, pud);
2915 extern void __init pagecache_init(void);
2916 extern void free_initmem(void);
2919 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2920 * into the buddy system. The freed pages will be poisoned with pattern
2921 * "poison" if it's within range [0, UCHAR_MAX].
2922 * Return pages freed into the buddy system.
2924 extern unsigned long free_reserved_area(void *start, void *end,
2925 int poison, const char *s);
2927 extern void adjust_managed_page_count(struct page *page, long count);
2929 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2931 /* Free the reserved page into the buddy system, so it gets managed. */
2932 static inline void free_reserved_page(struct page *page)
2934 ClearPageReserved(page);
2935 init_page_count(page);
2937 adjust_managed_page_count(page, 1);
2939 #define free_highmem_page(page) free_reserved_page(page)
2941 static inline void mark_page_reserved(struct page *page)
2943 SetPageReserved(page);
2944 adjust_managed_page_count(page, -1);
2948 * Default method to free all the __init memory into the buddy system.
2949 * The freed pages will be poisoned with pattern "poison" if it's within
2950 * range [0, UCHAR_MAX].
2951 * Return pages freed into the buddy system.
2953 static inline unsigned long free_initmem_default(int poison)
2955 extern char __init_begin[], __init_end[];
2957 return free_reserved_area(&__init_begin, &__init_end,
2958 poison, "unused kernel image (initmem)");
2961 static inline unsigned long get_num_physpages(void)
2964 unsigned long phys_pages = 0;
2966 for_each_online_node(nid)
2967 phys_pages += node_present_pages(nid);
2973 * Using memblock node mappings, an architecture may initialise its
2974 * zones, allocate the backing mem_map and account for memory holes in an
2975 * architecture independent manner.
2977 * An architecture is expected to register range of page frames backed by
2978 * physical memory with memblock_add[_node]() before calling
2979 * free_area_init() passing in the PFN each zone ends at. At a basic
2980 * usage, an architecture is expected to do something like
2982 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2984 * for_each_valid_physical_page_range()
2985 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2986 * free_area_init(max_zone_pfns);
2988 void free_area_init(unsigned long *max_zone_pfn);
2989 unsigned long node_map_pfn_alignment(void);
2990 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2991 unsigned long end_pfn);
2992 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2993 unsigned long end_pfn);
2994 extern void get_pfn_range_for_nid(unsigned int nid,
2995 unsigned long *start_pfn, unsigned long *end_pfn);
2998 static inline int early_pfn_to_nid(unsigned long pfn)
3003 /* please see mm/page_alloc.c */
3004 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3007 extern void set_dma_reserve(unsigned long new_dma_reserve);
3008 extern void memmap_init_range(unsigned long, int, unsigned long,
3009 unsigned long, unsigned long, enum meminit_context,
3010 struct vmem_altmap *, int migratetype);
3011 extern void setup_per_zone_wmarks(void);
3012 extern void calculate_min_free_kbytes(void);
3013 extern int __meminit init_per_zone_wmark_min(void);
3014 extern void mem_init(void);
3015 extern void __init mmap_init(void);
3017 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
3018 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
3020 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
3022 extern long si_mem_available(void);
3023 extern void si_meminfo(struct sysinfo * val);
3024 extern void si_meminfo_node(struct sysinfo *val, int nid);
3025 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3026 extern unsigned long arch_reserved_kernel_pages(void);
3029 extern __printf(3, 4)
3030 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3032 extern void setup_per_cpu_pageset(void);
3035 extern int min_free_kbytes;
3036 extern int watermark_boost_factor;
3037 extern int watermark_scale_factor;
3040 extern atomic_long_t mmap_pages_allocated;
3041 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3043 /* interval_tree.c */
3044 void vma_interval_tree_insert(struct vm_area_struct *node,
3045 struct rb_root_cached *root);
3046 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3047 struct vm_area_struct *prev,
3048 struct rb_root_cached *root);
3049 void vma_interval_tree_remove(struct vm_area_struct *node,
3050 struct rb_root_cached *root);
3051 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3052 unsigned long start, unsigned long last);
3053 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3054 unsigned long start, unsigned long last);
3056 #define vma_interval_tree_foreach(vma, root, start, last) \
3057 for (vma = vma_interval_tree_iter_first(root, start, last); \
3058 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3060 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3061 struct rb_root_cached *root);
3062 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3063 struct rb_root_cached *root);
3064 struct anon_vma_chain *
3065 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3066 unsigned long start, unsigned long last);
3067 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3068 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3069 #ifdef CONFIG_DEBUG_VM_RB
3070 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3073 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3074 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3075 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3078 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3079 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3080 unsigned long start, unsigned long end, pgoff_t pgoff,
3081 struct vm_area_struct *next);
3082 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3083 unsigned long start, unsigned long end, pgoff_t pgoff);
3084 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3085 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3086 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3087 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3088 struct anon_vma_name *);
3089 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3090 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3091 unsigned long addr, int new_below);
3092 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3093 unsigned long addr, int new_below);
3094 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3095 extern void unlink_file_vma(struct vm_area_struct *);
3096 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3097 unsigned long addr, unsigned long len, pgoff_t pgoff,
3098 bool *need_rmap_locks);
3099 extern void exit_mmap(struct mm_struct *);
3101 static inline int check_data_rlimit(unsigned long rlim,
3103 unsigned long start,
3104 unsigned long end_data,
3105 unsigned long start_data)
3107 if (rlim < RLIM_INFINITY) {
3108 if (((new - start) + (end_data - start_data)) > rlim)
3115 extern int mm_take_all_locks(struct mm_struct *mm);
3116 extern void mm_drop_all_locks(struct mm_struct *mm);
3118 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3119 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3120 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3121 extern struct file *get_task_exe_file(struct task_struct *task);
3123 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3124 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3126 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3127 const struct vm_special_mapping *sm);
3128 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3129 unsigned long addr, unsigned long len,
3130 unsigned long flags,
3131 const struct vm_special_mapping *spec);
3132 /* This is an obsolete alternative to _install_special_mapping. */
3133 extern int install_special_mapping(struct mm_struct *mm,
3134 unsigned long addr, unsigned long len,
3135 unsigned long flags, struct page **pages);
3137 unsigned long randomize_stack_top(unsigned long stack_top);
3138 unsigned long randomize_page(unsigned long start, unsigned long range);
3140 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3142 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3143 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3144 struct list_head *uf);
3145 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3146 unsigned long len, unsigned long prot, unsigned long flags,
3147 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3148 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3149 unsigned long start, size_t len, struct list_head *uf,
3151 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3152 struct list_head *uf);
3153 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3156 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3157 unsigned long start, unsigned long end,
3158 struct list_head *uf, bool downgrade);
3159 extern int __mm_populate(unsigned long addr, unsigned long len,
3161 static inline void mm_populate(unsigned long addr, unsigned long len)
3164 (void) __mm_populate(addr, len, 1);
3167 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3170 /* These take the mm semaphore themselves */
3171 extern int __must_check vm_brk(unsigned long, unsigned long);
3172 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3173 extern int vm_munmap(unsigned long, size_t);
3174 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3175 unsigned long, unsigned long,
3176 unsigned long, unsigned long);
3178 struct vm_unmapped_area_info {
3179 #define VM_UNMAPPED_AREA_TOPDOWN 1
3180 unsigned long flags;
3181 unsigned long length;
3182 unsigned long low_limit;
3183 unsigned long high_limit;
3184 unsigned long align_mask;
3185 unsigned long align_offset;
3188 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3191 extern void truncate_inode_pages(struct address_space *, loff_t);
3192 extern void truncate_inode_pages_range(struct address_space *,
3193 loff_t lstart, loff_t lend);
3194 extern void truncate_inode_pages_final(struct address_space *);
3196 /* generic vm_area_ops exported for stackable file systems */
3197 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3198 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3199 pgoff_t start_pgoff, pgoff_t end_pgoff);
3200 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3202 extern unsigned long stack_guard_gap;
3203 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3204 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
3206 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3207 extern int expand_downwards(struct vm_area_struct *vma,
3208 unsigned long address);
3210 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
3212 #define expand_upwards(vma, address) (0)
3215 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3216 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3217 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3218 struct vm_area_struct **pprev);
3221 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3222 * NULL if none. Assume start_addr < end_addr.
3224 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3225 unsigned long start_addr, unsigned long end_addr);
3228 * vma_lookup() - Find a VMA at a specific address
3229 * @mm: The process address space.
3230 * @addr: The user address.
3232 * Return: The vm_area_struct at the given address, %NULL otherwise.
3235 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3237 return mtree_load(&mm->mm_mt, addr);
3240 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3242 unsigned long vm_start = vma->vm_start;
3244 if (vma->vm_flags & VM_GROWSDOWN) {
3245 vm_start -= stack_guard_gap;
3246 if (vm_start > vma->vm_start)
3252 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3254 unsigned long vm_end = vma->vm_end;
3256 if (vma->vm_flags & VM_GROWSUP) {
3257 vm_end += stack_guard_gap;
3258 if (vm_end < vma->vm_end)
3259 vm_end = -PAGE_SIZE;
3264 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3266 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3269 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3270 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3271 unsigned long vm_start, unsigned long vm_end)
3273 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3275 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3281 static inline bool range_in_vma(struct vm_area_struct *vma,
3282 unsigned long start, unsigned long end)
3284 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3288 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3289 void vma_set_page_prot(struct vm_area_struct *vma);
3291 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3295 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3297 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3301 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3303 #ifdef CONFIG_NUMA_BALANCING
3304 unsigned long change_prot_numa(struct vm_area_struct *vma,
3305 unsigned long start, unsigned long end);
3308 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3309 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3310 unsigned long pfn, unsigned long size, pgprot_t);
3311 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3312 unsigned long pfn, unsigned long size, pgprot_t prot);
3313 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3314 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3315 struct page **pages, unsigned long *num);
3316 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3318 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3320 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3322 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3323 unsigned long pfn, pgprot_t pgprot);
3324 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3326 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3327 unsigned long addr, pfn_t pfn);
3328 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3330 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3331 unsigned long addr, struct page *page)
3333 int err = vm_insert_page(vma, addr, page);
3336 return VM_FAULT_OOM;
3337 if (err < 0 && err != -EBUSY)
3338 return VM_FAULT_SIGBUS;
3340 return VM_FAULT_NOPAGE;
3343 #ifndef io_remap_pfn_range
3344 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3345 unsigned long addr, unsigned long pfn,
3346 unsigned long size, pgprot_t prot)
3348 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3352 static inline vm_fault_t vmf_error(int err)
3355 return VM_FAULT_OOM;
3356 return VM_FAULT_SIGBUS;
3359 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3360 unsigned int foll_flags);
3362 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3364 if (vm_fault & VM_FAULT_OOM)
3366 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3367 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3368 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3374 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3375 * a (NUMA hinting) fault is required.
3377 static inline bool gup_can_follow_protnone(unsigned int flags)
3380 * FOLL_FORCE has to be able to make progress even if the VMA is
3381 * inaccessible. Further, FOLL_FORCE access usually does not represent
3382 * application behaviour and we should avoid triggering NUMA hinting
3385 return flags & FOLL_FORCE;
3388 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3389 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3390 unsigned long size, pte_fn_t fn, void *data);
3391 extern int apply_to_existing_page_range(struct mm_struct *mm,
3392 unsigned long address, unsigned long size,
3393 pte_fn_t fn, void *data);
3395 #ifdef CONFIG_PAGE_POISONING
3396 extern void __kernel_poison_pages(struct page *page, int numpages);
3397 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3398 extern bool _page_poisoning_enabled_early;
3399 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3400 static inline bool page_poisoning_enabled(void)
3402 return _page_poisoning_enabled_early;
3405 * For use in fast paths after init_mem_debugging() has run, or when a
3406 * false negative result is not harmful when called too early.
3408 static inline bool page_poisoning_enabled_static(void)
3410 return static_branch_unlikely(&_page_poisoning_enabled);
3412 static inline void kernel_poison_pages(struct page *page, int numpages)
3414 if (page_poisoning_enabled_static())
3415 __kernel_poison_pages(page, numpages);
3417 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3419 if (page_poisoning_enabled_static())
3420 __kernel_unpoison_pages(page, numpages);
3423 static inline bool page_poisoning_enabled(void) { return false; }
3424 static inline bool page_poisoning_enabled_static(void) { return false; }
3425 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3426 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3427 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3430 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3431 static inline bool want_init_on_alloc(gfp_t flags)
3433 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3436 return flags & __GFP_ZERO;
3439 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3440 static inline bool want_init_on_free(void)
3442 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3446 extern bool _debug_pagealloc_enabled_early;
3447 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3449 static inline bool debug_pagealloc_enabled(void)
3451 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3452 _debug_pagealloc_enabled_early;
3456 * For use in fast paths after init_debug_pagealloc() has run, or when a
3457 * false negative result is not harmful when called too early.
3459 static inline bool debug_pagealloc_enabled_static(void)
3461 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3464 return static_branch_unlikely(&_debug_pagealloc_enabled);
3467 #ifdef CONFIG_DEBUG_PAGEALLOC
3469 * To support DEBUG_PAGEALLOC architecture must ensure that
3470 * __kernel_map_pages() never fails
3472 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3474 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3476 if (debug_pagealloc_enabled_static())
3477 __kernel_map_pages(page, numpages, 1);
3480 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3482 if (debug_pagealloc_enabled_static())
3483 __kernel_map_pages(page, numpages, 0);
3485 #else /* CONFIG_DEBUG_PAGEALLOC */
3486 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3487 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3488 #endif /* CONFIG_DEBUG_PAGEALLOC */
3490 #ifdef __HAVE_ARCH_GATE_AREA
3491 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3492 extern int in_gate_area_no_mm(unsigned long addr);
3493 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3495 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3499 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3500 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3504 #endif /* __HAVE_ARCH_GATE_AREA */
3506 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3508 #ifdef CONFIG_SYSCTL
3509 extern int sysctl_drop_caches;
3510 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3514 void drop_slab(void);
3517 #define randomize_va_space 0
3519 extern int randomize_va_space;
3522 const char * arch_vma_name(struct vm_area_struct *vma);
3524 void print_vma_addr(char *prefix, unsigned long rip);
3526 static inline void print_vma_addr(char *prefix, unsigned long rip)
3531 void *sparse_buffer_alloc(unsigned long size);
3532 struct page * __populate_section_memmap(unsigned long pfn,
3533 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3534 struct dev_pagemap *pgmap);
3535 void pmd_init(void *addr);
3536 void pud_init(void *addr);
3537 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3538 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3539 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3540 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3541 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3542 struct vmem_altmap *altmap, struct page *reuse);
3543 void *vmemmap_alloc_block(unsigned long size, int node);
3545 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3546 struct vmem_altmap *altmap);
3547 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3548 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3549 unsigned long addr, unsigned long next);
3550 int vmemmap_check_pmd(pmd_t *pmd, int node,
3551 unsigned long addr, unsigned long next);
3552 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3553 int node, struct vmem_altmap *altmap);
3554 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3555 int node, struct vmem_altmap *altmap);
3556 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3557 struct vmem_altmap *altmap);
3558 void vmemmap_populate_print_last(void);
3559 #ifdef CONFIG_MEMORY_HOTPLUG
3560 void vmemmap_free(unsigned long start, unsigned long end,
3561 struct vmem_altmap *altmap);
3564 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP
3565 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3566 struct dev_pagemap *pgmap)
3568 return is_power_of_2(sizeof(struct page)) &&
3569 pgmap && (pgmap_vmemmap_nr(pgmap) > 1) && !altmap;
3572 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3573 struct dev_pagemap *pgmap)
3579 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3580 unsigned long nr_pages);
3583 MF_COUNT_INCREASED = 1 << 0,
3584 MF_ACTION_REQUIRED = 1 << 1,
3585 MF_MUST_KILL = 1 << 2,
3586 MF_SOFT_OFFLINE = 1 << 3,
3587 MF_UNPOISON = 1 << 4,
3588 MF_SW_SIMULATED = 1 << 5,
3589 MF_NO_RETRY = 1 << 6,
3591 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3592 unsigned long count, int mf_flags);
3593 extern int memory_failure(unsigned long pfn, int flags);
3594 extern void memory_failure_queue_kick(int cpu);
3595 extern int unpoison_memory(unsigned long pfn);
3596 extern void shake_page(struct page *p);
3597 extern atomic_long_t num_poisoned_pages __read_mostly;
3598 extern int soft_offline_page(unsigned long pfn, int flags);
3599 #ifdef CONFIG_MEMORY_FAILURE
3600 extern void memory_failure_queue(unsigned long pfn, int flags);
3601 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3602 bool *migratable_cleared);
3603 void num_poisoned_pages_inc(unsigned long pfn);
3604 void num_poisoned_pages_sub(unsigned long pfn, long i);
3605 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3607 static inline void memory_failure_queue(unsigned long pfn, int flags)
3611 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3612 bool *migratable_cleared)
3617 static inline void num_poisoned_pages_inc(unsigned long pfn)
3621 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3626 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3627 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3628 struct vm_area_struct *vma, struct list_head *to_kill,
3629 unsigned long ksm_addr);
3632 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3633 extern void memblk_nr_poison_inc(unsigned long pfn);
3634 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3636 static inline void memblk_nr_poison_inc(unsigned long pfn)
3640 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3645 #ifndef arch_memory_failure
3646 static inline int arch_memory_failure(unsigned long pfn, int flags)
3652 #ifndef arch_is_platform_page
3653 static inline bool arch_is_platform_page(u64 paddr)
3660 * Error handlers for various types of pages.
3663 MF_IGNORED, /* Error: cannot be handled */
3664 MF_FAILED, /* Error: handling failed */
3665 MF_DELAYED, /* Will be handled later */
3666 MF_RECOVERED, /* Successfully recovered */
3669 enum mf_action_page_type {
3671 MF_MSG_KERNEL_HIGH_ORDER,
3673 MF_MSG_DIFFERENT_COMPOUND,
3676 MF_MSG_UNMAP_FAILED,
3677 MF_MSG_DIRTY_SWAPCACHE,
3678 MF_MSG_CLEAN_SWAPCACHE,
3679 MF_MSG_DIRTY_MLOCKED_LRU,
3680 MF_MSG_CLEAN_MLOCKED_LRU,
3681 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3682 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3685 MF_MSG_TRUNCATED_LRU,
3693 * Sysfs entries for memory failure handling statistics.
3695 extern const struct attribute_group memory_failure_attr_group;
3697 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3698 extern void clear_huge_page(struct page *page,
3699 unsigned long addr_hint,
3700 unsigned int pages_per_huge_page);
3701 int copy_user_large_folio(struct folio *dst, struct folio *src,
3702 unsigned long addr_hint,
3703 struct vm_area_struct *vma);
3704 long copy_folio_from_user(struct folio *dst_folio,
3705 const void __user *usr_src,
3706 bool allow_pagefault);
3709 * vma_is_special_huge - Are transhuge page-table entries considered special?
3710 * @vma: Pointer to the struct vm_area_struct to consider
3712 * Whether transhuge page-table entries are considered "special" following
3713 * the definition in vm_normal_page().
3715 * Return: true if transhuge page-table entries should be considered special,
3718 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3720 return vma_is_dax(vma) || (vma->vm_file &&
3721 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3724 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3726 #ifdef CONFIG_DEBUG_PAGEALLOC
3727 extern unsigned int _debug_guardpage_minorder;
3728 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3730 static inline unsigned int debug_guardpage_minorder(void)
3732 return _debug_guardpage_minorder;
3735 static inline bool debug_guardpage_enabled(void)
3737 return static_branch_unlikely(&_debug_guardpage_enabled);
3740 static inline bool page_is_guard(struct page *page)
3742 if (!debug_guardpage_enabled())
3745 return PageGuard(page);
3748 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3749 static inline bool debug_guardpage_enabled(void) { return false; }
3750 static inline bool page_is_guard(struct page *page) { return false; }
3751 #endif /* CONFIG_DEBUG_PAGEALLOC */
3753 #if MAX_NUMNODES > 1
3754 void __init setup_nr_node_ids(void);
3756 static inline void setup_nr_node_ids(void) {}
3759 extern int memcmp_pages(struct page *page1, struct page *page2);
3761 static inline int pages_identical(struct page *page1, struct page *page2)
3763 return !memcmp_pages(page1, page2);
3766 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3767 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3768 pgoff_t first_index, pgoff_t nr,
3769 pgoff_t bitmap_pgoff,
3770 unsigned long *bitmap,
3774 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3775 pgoff_t first_index, pgoff_t nr);
3778 extern int sysctl_nr_trim_pages;
3780 #ifdef CONFIG_PRINTK
3781 void mem_dump_obj(void *object);
3783 static inline void mem_dump_obj(void *object) {}
3787 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3788 * @seals: the seals to check
3789 * @vma: the vma to operate on
3791 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3792 * the vma flags. Return 0 if check pass, or <0 for errors.
3794 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3796 if (seals & F_SEAL_FUTURE_WRITE) {
3798 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3799 * "future write" seal active.
3801 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3805 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3806 * MAP_SHARED and read-only, take care to not allow mprotect to
3807 * revert protections on such mappings. Do this only for shared
3808 * mappings. For private mappings, don't need to mask
3809 * VM_MAYWRITE as we still want them to be COW-writable.
3811 if (vma->vm_flags & VM_SHARED)
3812 vm_flags_clear(vma, VM_MAYWRITE);
3818 #ifdef CONFIG_ANON_VMA_NAME
3819 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3820 unsigned long len_in,
3821 struct anon_vma_name *anon_name);
3824 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3825 unsigned long len_in, struct anon_vma_name *anon_name) {
3830 #endif /* _LINUX_MM_H */