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 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
106 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
110 #define lm_alias(x) __va(__pa_symbol(x))
114 * To prevent common memory management code establishing
115 * a zero page mapping on a read fault.
116 * This macro should be defined within <asm/pgtable.h>.
117 * s390 does this to prevent multiplexing of hardware bits
118 * related to the physical page in case of virtualization.
120 #ifndef mm_forbids_zeropage
121 #define mm_forbids_zeropage(X) (0)
125 * On some architectures it is expensive to call memset() for small sizes.
126 * If an architecture decides to implement their own version of
127 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
128 * define their own version of this macro in <asm/pgtable.h>
130 #if BITS_PER_LONG == 64
131 /* This function must be updated when the size of struct page grows above 96
132 * or reduces below 56. The idea that compiler optimizes out switch()
133 * statement, and only leaves move/store instructions. Also the compiler can
134 * combine write statements if they are both assignments and can be reordered,
135 * this can result in several of the writes here being dropped.
137 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
138 static inline void __mm_zero_struct_page(struct page *page)
140 unsigned long *_pp = (void *)page;
142 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
143 BUILD_BUG_ON(sizeof(struct page) & 7);
144 BUILD_BUG_ON(sizeof(struct page) < 56);
145 BUILD_BUG_ON(sizeof(struct page) > 96);
147 switch (sizeof(struct page)) {
174 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
178 * Default maximum number of active map areas, this limits the number of vmas
179 * per mm struct. Users can overwrite this number by sysctl but there is a
182 * When a program's coredump is generated as ELF format, a section is created
183 * per a vma. In ELF, the number of sections is represented in unsigned short.
184 * This means the number of sections should be smaller than 65535 at coredump.
185 * Because the kernel adds some informative sections to a image of program at
186 * generating coredump, we need some margin. The number of extra sections is
187 * 1-3 now and depends on arch. We use "5" as safe margin, here.
189 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
190 * not a hard limit any more. Although some userspace tools can be surprised by
193 #define MAPCOUNT_ELF_CORE_MARGIN (5)
194 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
196 extern int sysctl_max_map_count;
198 extern unsigned long sysctl_user_reserve_kbytes;
199 extern unsigned long sysctl_admin_reserve_kbytes;
201 extern int sysctl_overcommit_memory;
202 extern int sysctl_overcommit_ratio;
203 extern unsigned long sysctl_overcommit_kbytes;
205 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
207 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
209 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
212 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
213 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
214 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
216 #define nth_page(page,n) ((page) + (n))
217 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
220 /* to align the pointer to the (next) page boundary */
221 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
223 /* to align the pointer to the (prev) page boundary */
224 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
226 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
227 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
229 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
230 static inline struct folio *lru_to_folio(struct list_head *head)
232 return list_entry((head)->prev, struct folio, lru);
235 void setup_initial_init_mm(void *start_code, void *end_code,
236 void *end_data, void *brk);
239 * Linux kernel virtual memory manager primitives.
240 * The idea being to have a "virtual" mm in the same way
241 * we have a virtual fs - giving a cleaner interface to the
242 * mm details, and allowing different kinds of memory mappings
243 * (from shared memory to executable loading to arbitrary
247 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
248 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
249 void vm_area_free(struct vm_area_struct *);
250 /* Use only if VMA has no other users */
251 void __vm_area_free(struct vm_area_struct *vma);
254 extern struct rb_root nommu_region_tree;
255 extern struct rw_semaphore nommu_region_sem;
257 extern unsigned int kobjsize(const void *objp);
261 * vm_flags in vm_area_struct, see mm_types.h.
262 * When changing, update also include/trace/events/mmflags.h
264 #define VM_NONE 0x00000000
266 #define VM_READ 0x00000001 /* currently active flags */
267 #define VM_WRITE 0x00000002
268 #define VM_EXEC 0x00000004
269 #define VM_SHARED 0x00000008
271 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
272 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
273 #define VM_MAYWRITE 0x00000020
274 #define VM_MAYEXEC 0x00000040
275 #define VM_MAYSHARE 0x00000080
277 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
279 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
280 #else /* CONFIG_MMU */
281 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
282 #define VM_UFFD_MISSING 0
283 #endif /* CONFIG_MMU */
284 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
285 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
287 #define VM_LOCKED 0x00002000
288 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
290 /* Used by sys_madvise() */
291 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
292 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
294 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
295 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
296 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
297 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
298 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
299 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
300 #define VM_SYNC 0x00800000 /* Synchronous page faults */
301 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
302 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
303 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
305 #ifdef CONFIG_MEM_SOFT_DIRTY
306 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
308 # define VM_SOFTDIRTY 0
311 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
312 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
313 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
314 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
316 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
317 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
323 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
324 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
325 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
326 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
327 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
329 #ifdef CONFIG_ARCH_HAS_PKEYS
330 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
331 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
332 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
333 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
334 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
336 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
338 # define VM_PKEY_BIT4 0
340 #endif /* CONFIG_ARCH_HAS_PKEYS */
342 #if defined(CONFIG_X86)
343 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
344 #elif defined(CONFIG_PPC)
345 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
346 #elif defined(CONFIG_PARISC)
347 # define VM_GROWSUP VM_ARCH_1
348 #elif defined(CONFIG_IA64)
349 # define VM_GROWSUP VM_ARCH_1
350 #elif defined(CONFIG_SPARC64)
351 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
352 # define VM_ARCH_CLEAR VM_SPARC_ADI
353 #elif defined(CONFIG_ARM64)
354 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
355 # define VM_ARCH_CLEAR VM_ARM64_BTI
356 #elif !defined(CONFIG_MMU)
357 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
360 #if defined(CONFIG_ARM64_MTE)
361 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
362 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
364 # define VM_MTE VM_NONE
365 # define VM_MTE_ALLOWED VM_NONE
369 # define VM_GROWSUP VM_NONE
372 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
373 # define VM_UFFD_MINOR_BIT 37
374 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
375 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
376 # define VM_UFFD_MINOR VM_NONE
377 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
379 /* Bits set in the VMA until the stack is in its final location */
380 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
382 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
384 /* Common data flag combinations */
385 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
386 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
387 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
388 VM_MAYWRITE | VM_MAYEXEC)
389 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
390 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
392 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
393 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
396 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
397 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
400 #ifdef CONFIG_STACK_GROWSUP
401 #define VM_STACK VM_GROWSUP
403 #define VM_STACK VM_GROWSDOWN
406 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
408 /* VMA basic access permission flags */
409 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
413 * Special vmas that are non-mergable, non-mlock()able.
415 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
417 /* This mask prevents VMA from being scanned with khugepaged */
418 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
420 /* This mask defines which mm->def_flags a process can inherit its parent */
421 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
423 /* This mask represents all the VMA flag bits used by mlock */
424 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
426 /* Arch-specific flags to clear when updating VM flags on protection change */
427 #ifndef VM_ARCH_CLEAR
428 # define VM_ARCH_CLEAR VM_NONE
430 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
433 * mapping from the currently active vm_flags protection bits (the
434 * low four bits) to a page protection mask..
438 * The default fault flags that should be used by most of the
439 * arch-specific page fault handlers.
441 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
442 FAULT_FLAG_KILLABLE | \
443 FAULT_FLAG_INTERRUPTIBLE)
446 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
447 * @flags: Fault flags.
449 * This is mostly used for places where we want to try to avoid taking
450 * the mmap_lock for too long a time when waiting for another condition
451 * to change, in which case we can try to be polite to release the
452 * mmap_lock in the first round to avoid potential starvation of other
453 * processes that would also want the mmap_lock.
455 * Return: true if the page fault allows retry and this is the first
456 * attempt of the fault handling; false otherwise.
458 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
460 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
461 (!(flags & FAULT_FLAG_TRIED));
464 #define FAULT_FLAG_TRACE \
465 { FAULT_FLAG_WRITE, "WRITE" }, \
466 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
467 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
468 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
469 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
470 { FAULT_FLAG_TRIED, "TRIED" }, \
471 { FAULT_FLAG_USER, "USER" }, \
472 { FAULT_FLAG_REMOTE, "REMOTE" }, \
473 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
474 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
475 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
478 * vm_fault is filled by the pagefault handler and passed to the vma's
479 * ->fault function. The vma's ->fault is responsible for returning a bitmask
480 * of VM_FAULT_xxx flags that give details about how the fault was handled.
482 * MM layer fills up gfp_mask for page allocations but fault handler might
483 * alter it if its implementation requires a different allocation context.
485 * pgoff should be used in favour of virtual_address, if possible.
489 struct vm_area_struct *vma; /* Target VMA */
490 gfp_t gfp_mask; /* gfp mask to be used for allocations */
491 pgoff_t pgoff; /* Logical page offset based on vma */
492 unsigned long address; /* Faulting virtual address - masked */
493 unsigned long real_address; /* Faulting virtual address - unmasked */
495 enum fault_flag flags; /* FAULT_FLAG_xxx flags
496 * XXX: should really be 'const' */
497 pmd_t *pmd; /* Pointer to pmd entry matching
499 pud_t *pud; /* Pointer to pud entry matching
503 pte_t orig_pte; /* Value of PTE at the time of fault */
504 pmd_t orig_pmd; /* Value of PMD at the time of fault,
505 * used by PMD fault only.
509 struct page *cow_page; /* Page handler may use for COW fault */
510 struct page *page; /* ->fault handlers should return a
511 * page here, unless VM_FAULT_NOPAGE
512 * is set (which is also implied by
515 /* These three entries are valid only while holding ptl lock */
516 pte_t *pte; /* Pointer to pte entry matching
517 * the 'address'. NULL if the page
518 * table hasn't been allocated.
520 spinlock_t *ptl; /* Page table lock.
521 * Protects pte page table if 'pte'
522 * is not NULL, otherwise pmd.
524 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
525 * vm_ops->map_pages() sets up a page
526 * table from atomic context.
527 * do_fault_around() pre-allocates
528 * page table to avoid allocation from
533 /* page entry size for vm->huge_fault() */
534 enum page_entry_size {
541 * These are the virtual MM functions - opening of an area, closing and
542 * unmapping it (needed to keep files on disk up-to-date etc), pointer
543 * to the functions called when a no-page or a wp-page exception occurs.
545 struct vm_operations_struct {
546 void (*open)(struct vm_area_struct * area);
548 * @close: Called when the VMA is being removed from the MM.
549 * Context: User context. May sleep. Caller holds mmap_lock.
551 void (*close)(struct vm_area_struct * area);
552 /* Called any time before splitting to check if it's allowed */
553 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
554 int (*mremap)(struct vm_area_struct *area);
556 * Called by mprotect() to make driver-specific permission
557 * checks before mprotect() is finalised. The VMA must not
558 * be modified. Returns 0 if mprotect() can proceed.
560 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
561 unsigned long end, unsigned long newflags);
562 vm_fault_t (*fault)(struct vm_fault *vmf);
563 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
564 enum page_entry_size pe_size);
565 vm_fault_t (*map_pages)(struct vm_fault *vmf,
566 pgoff_t start_pgoff, pgoff_t end_pgoff);
567 unsigned long (*pagesize)(struct vm_area_struct * area);
569 /* notification that a previously read-only page is about to become
570 * writable, if an error is returned it will cause a SIGBUS */
571 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
573 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
574 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
576 /* called by access_process_vm when get_user_pages() fails, typically
577 * for use by special VMAs. See also generic_access_phys() for a generic
578 * implementation useful for any iomem mapping.
580 int (*access)(struct vm_area_struct *vma, unsigned long addr,
581 void *buf, int len, int write);
583 /* Called by the /proc/PID/maps code to ask the vma whether it
584 * has a special name. Returning non-NULL will also cause this
585 * vma to be dumped unconditionally. */
586 const char *(*name)(struct vm_area_struct *vma);
590 * set_policy() op must add a reference to any non-NULL @new mempolicy
591 * to hold the policy upon return. Caller should pass NULL @new to
592 * remove a policy and fall back to surrounding context--i.e. do not
593 * install a MPOL_DEFAULT policy, nor the task or system default
596 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
599 * get_policy() op must add reference [mpol_get()] to any policy at
600 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
601 * in mm/mempolicy.c will do this automatically.
602 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
603 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
604 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
605 * must return NULL--i.e., do not "fallback" to task or system default
608 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
612 * Called by vm_normal_page() for special PTEs to find the
613 * page for @addr. This is useful if the default behavior
614 * (using pte_page()) would not find the correct page.
616 struct page *(*find_special_page)(struct vm_area_struct *vma,
620 #ifdef CONFIG_NUMA_BALANCING
621 static inline void vma_numab_state_init(struct vm_area_struct *vma)
623 vma->numab_state = NULL;
625 static inline void vma_numab_state_free(struct vm_area_struct *vma)
627 kfree(vma->numab_state);
630 static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
631 static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
632 #endif /* CONFIG_NUMA_BALANCING */
634 #ifdef CONFIG_PER_VMA_LOCK
636 * Try to read-lock a vma. The function is allowed to occasionally yield false
637 * locked result to avoid performance overhead, in which case we fall back to
638 * using mmap_lock. The function should never yield false unlocked result.
640 static inline bool vma_start_read(struct vm_area_struct *vma)
642 /* Check before locking. A race might cause false locked result. */
643 if (vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))
646 if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
650 * Overflow might produce false locked result.
651 * False unlocked result is impossible because we modify and check
652 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
653 * modification invalidates all existing locks.
655 if (unlikely(vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))) {
656 up_read(&vma->vm_lock->lock);
662 static inline void vma_end_read(struct vm_area_struct *vma)
664 rcu_read_lock(); /* keeps vma alive till the end of up_read */
665 up_read(&vma->vm_lock->lock);
669 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
671 mmap_assert_write_locked(vma->vm_mm);
674 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
675 * mm->mm_lock_seq can't be concurrently modified.
677 *mm_lock_seq = READ_ONCE(vma->vm_mm->mm_lock_seq);
678 return (vma->vm_lock_seq == *mm_lock_seq);
681 static inline void vma_start_write(struct vm_area_struct *vma)
685 if (__is_vma_write_locked(vma, &mm_lock_seq))
688 down_write(&vma->vm_lock->lock);
689 vma->vm_lock_seq = mm_lock_seq;
690 up_write(&vma->vm_lock->lock);
693 static inline bool vma_try_start_write(struct vm_area_struct *vma)
697 if (__is_vma_write_locked(vma, &mm_lock_seq))
700 if (!down_write_trylock(&vma->vm_lock->lock))
703 vma->vm_lock_seq = mm_lock_seq;
704 up_write(&vma->vm_lock->lock);
708 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
712 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
715 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
717 /* When detaching vma should be write-locked */
719 vma_assert_write_locked(vma);
720 vma->detached = detached;
723 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
724 unsigned long address);
726 #else /* CONFIG_PER_VMA_LOCK */
728 static inline void vma_init_lock(struct vm_area_struct *vma) {}
729 static inline bool vma_start_read(struct vm_area_struct *vma)
731 static inline void vma_end_read(struct vm_area_struct *vma) {}
732 static inline void vma_start_write(struct vm_area_struct *vma) {}
733 static inline bool vma_try_start_write(struct vm_area_struct *vma)
735 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
736 static inline void vma_mark_detached(struct vm_area_struct *vma,
739 #endif /* CONFIG_PER_VMA_LOCK */
742 * WARNING: vma_init does not initialize vma->vm_lock.
743 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
745 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
747 static const struct vm_operations_struct dummy_vm_ops = {};
749 memset(vma, 0, sizeof(*vma));
751 vma->vm_ops = &dummy_vm_ops;
752 INIT_LIST_HEAD(&vma->anon_vma_chain);
753 vma_mark_detached(vma, false);
754 vma_numab_state_init(vma);
757 /* Use when VMA is not part of the VMA tree and needs no locking */
758 static inline void vm_flags_init(struct vm_area_struct *vma,
761 ACCESS_PRIVATE(vma, __vm_flags) = flags;
764 /* Use when VMA is part of the VMA tree and modifications need coordination */
765 static inline void vm_flags_reset(struct vm_area_struct *vma,
768 vma_start_write(vma);
769 vm_flags_init(vma, flags);
772 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
775 vma_start_write(vma);
776 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
779 static inline void vm_flags_set(struct vm_area_struct *vma,
782 vma_start_write(vma);
783 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
786 static inline void vm_flags_clear(struct vm_area_struct *vma,
789 vma_start_write(vma);
790 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
794 * Use only if VMA is not part of the VMA tree or has no other users and
795 * therefore needs no locking.
797 static inline void __vm_flags_mod(struct vm_area_struct *vma,
798 vm_flags_t set, vm_flags_t clear)
800 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
804 * Use only when the order of set/clear operations is unimportant, otherwise
805 * use vm_flags_{set|clear} explicitly.
807 static inline void vm_flags_mod(struct vm_area_struct *vma,
808 vm_flags_t set, vm_flags_t clear)
810 vma_start_write(vma);
811 __vm_flags_mod(vma, set, clear);
814 static inline void vma_set_anonymous(struct vm_area_struct *vma)
819 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
824 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
826 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
831 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
832 VM_STACK_INCOMPLETE_SETUP)
838 static inline bool vma_is_foreign(struct vm_area_struct *vma)
843 if (current->mm != vma->vm_mm)
849 static inline bool vma_is_accessible(struct vm_area_struct *vma)
851 return vma->vm_flags & VM_ACCESS_FLAGS;
855 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
857 return mas_find(&vmi->mas, max - 1);
860 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
863 * Uses mas_find() to get the first VMA when the iterator starts.
864 * Calling mas_next() could skip the first entry.
866 return mas_find(&vmi->mas, ULONG_MAX);
869 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
871 return mas_prev(&vmi->mas, 0);
874 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
876 return vmi->mas.index;
879 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
881 return vmi->mas.last + 1;
883 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
886 return mas_expected_entries(&vmi->mas, count);
889 /* Free any unused preallocations */
890 static inline void vma_iter_free(struct vma_iterator *vmi)
892 mas_destroy(&vmi->mas);
895 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
896 struct vm_area_struct *vma)
898 vmi->mas.index = vma->vm_start;
899 vmi->mas.last = vma->vm_end - 1;
900 mas_store(&vmi->mas, vma);
901 if (unlikely(mas_is_err(&vmi->mas)))
907 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
909 mas_pause(&vmi->mas);
912 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
914 mas_set(&vmi->mas, addr);
917 #define for_each_vma(__vmi, __vma) \
918 while (((__vma) = vma_next(&(__vmi))) != NULL)
920 /* The MM code likes to work with exclusive end addresses */
921 #define for_each_vma_range(__vmi, __vma, __end) \
922 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
926 * The vma_is_shmem is not inline because it is used only by slow
927 * paths in userfault.
929 bool vma_is_shmem(struct vm_area_struct *vma);
930 bool vma_is_anon_shmem(struct vm_area_struct *vma);
932 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
933 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
936 int vma_is_stack_for_current(struct vm_area_struct *vma);
938 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
939 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
945 * compound_order() can be called without holding a reference, which means
946 * that niceties like page_folio() don't work. These callers should be
947 * prepared to handle wild return values. For example, PG_head may be
948 * set before _folio_order is initialised, or this may be a tail page.
949 * See compaction.c for some good examples.
951 static inline unsigned int compound_order(struct page *page)
953 struct folio *folio = (struct folio *)page;
955 if (!test_bit(PG_head, &folio->flags))
957 return folio->_folio_order;
961 * folio_order - The allocation order of a folio.
964 * A folio is composed of 2^order pages. See get_order() for the definition
967 * Return: The order of the folio.
969 static inline unsigned int folio_order(struct folio *folio)
971 if (!folio_test_large(folio))
973 return folio->_folio_order;
976 #include <linux/huge_mm.h>
979 * Methods to modify the page usage count.
981 * What counts for a page usage:
982 * - cache mapping (page->mapping)
983 * - private data (page->private)
984 * - page mapped in a task's page tables, each mapping
985 * is counted separately
987 * Also, many kernel routines increase the page count before a critical
988 * routine so they can be sure the page doesn't go away from under them.
992 * Drop a ref, return true if the refcount fell to zero (the page has no users)
994 static inline int put_page_testzero(struct page *page)
996 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
997 return page_ref_dec_and_test(page);
1000 static inline int folio_put_testzero(struct folio *folio)
1002 return put_page_testzero(&folio->page);
1006 * Try to grab a ref unless the page has a refcount of zero, return false if
1008 * This can be called when MMU is off so it must not access
1009 * any of the virtual mappings.
1011 static inline bool get_page_unless_zero(struct page *page)
1013 return page_ref_add_unless(page, 1, 0);
1016 static inline struct folio *folio_get_nontail_page(struct page *page)
1018 if (unlikely(!get_page_unless_zero(page)))
1020 return (struct folio *)page;
1023 extern int page_is_ram(unsigned long pfn);
1031 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1032 unsigned long desc);
1034 /* Support for virtually mapped pages */
1035 struct page *vmalloc_to_page(const void *addr);
1036 unsigned long vmalloc_to_pfn(const void *addr);
1039 * Determine if an address is within the vmalloc range
1041 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1042 * is no special casing required.
1045 #ifndef is_ioremap_addr
1046 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1050 extern bool is_vmalloc_addr(const void *x);
1051 extern int is_vmalloc_or_module_addr(const void *x);
1053 static inline bool is_vmalloc_addr(const void *x)
1057 static inline int is_vmalloc_or_module_addr(const void *x)
1064 * How many times the entire folio is mapped as a single unit (eg by a
1065 * PMD or PUD entry). This is probably not what you want, except for
1066 * debugging purposes - it does not include PTE-mapped sub-pages; look
1067 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1069 static inline int folio_entire_mapcount(struct folio *folio)
1071 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1072 return atomic_read(&folio->_entire_mapcount) + 1;
1076 * The atomic page->_mapcount, starts from -1: so that transitions
1077 * both from it and to it can be tracked, using atomic_inc_and_test
1078 * and atomic_add_negative(-1).
1080 static inline void page_mapcount_reset(struct page *page)
1082 atomic_set(&(page)->_mapcount, -1);
1086 * page_mapcount() - Number of times this precise page is mapped.
1089 * The number of times this page is mapped. If this page is part of
1090 * a large folio, it includes the number of times this page is mapped
1091 * as part of that folio.
1093 * The result is undefined for pages which cannot be mapped into userspace.
1094 * For example SLAB or special types of pages. See function page_has_type().
1095 * They use this field in struct page differently.
1097 static inline int page_mapcount(struct page *page)
1099 int mapcount = atomic_read(&page->_mapcount) + 1;
1101 if (unlikely(PageCompound(page)))
1102 mapcount += folio_entire_mapcount(page_folio(page));
1107 int folio_total_mapcount(struct folio *folio);
1110 * folio_mapcount() - Calculate the number of mappings of this folio.
1111 * @folio: The folio.
1113 * A large folio tracks both how many times the entire folio is mapped,
1114 * and how many times each individual page in the folio is mapped.
1115 * This function calculates the total number of times the folio is
1118 * Return: The number of times this folio is mapped.
1120 static inline int folio_mapcount(struct folio *folio)
1122 if (likely(!folio_test_large(folio)))
1123 return atomic_read(&folio->_mapcount) + 1;
1124 return folio_total_mapcount(folio);
1127 static inline int total_mapcount(struct page *page)
1129 if (likely(!PageCompound(page)))
1130 return atomic_read(&page->_mapcount) + 1;
1131 return folio_total_mapcount(page_folio(page));
1134 static inline bool folio_large_is_mapped(struct folio *folio)
1137 * Reading _entire_mapcount below could be omitted if hugetlb
1138 * participated in incrementing nr_pages_mapped when compound mapped.
1140 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1141 atomic_read(&folio->_entire_mapcount) >= 0;
1145 * folio_mapped - Is this folio mapped into userspace?
1146 * @folio: The folio.
1148 * Return: True if any page in this folio is referenced by user page tables.
1150 static inline bool folio_mapped(struct folio *folio)
1152 if (likely(!folio_test_large(folio)))
1153 return atomic_read(&folio->_mapcount) >= 0;
1154 return folio_large_is_mapped(folio);
1158 * Return true if this page is mapped into pagetables.
1159 * For compound page it returns true if any sub-page of compound page is mapped,
1160 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1162 static inline bool page_mapped(struct page *page)
1164 if (likely(!PageCompound(page)))
1165 return atomic_read(&page->_mapcount) >= 0;
1166 return folio_large_is_mapped(page_folio(page));
1169 static inline struct page *virt_to_head_page(const void *x)
1171 struct page *page = virt_to_page(x);
1173 return compound_head(page);
1176 static inline struct folio *virt_to_folio(const void *x)
1178 struct page *page = virt_to_page(x);
1180 return page_folio(page);
1183 void __folio_put(struct folio *folio);
1185 void put_pages_list(struct list_head *pages);
1187 void split_page(struct page *page, unsigned int order);
1188 void folio_copy(struct folio *dst, struct folio *src);
1190 unsigned long nr_free_buffer_pages(void);
1193 * Compound pages have a destructor function. Provide a
1194 * prototype for that function and accessor functions.
1195 * These are _only_ valid on the head of a compound page.
1197 typedef void compound_page_dtor(struct page *);
1199 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1200 enum compound_dtor_id {
1203 #ifdef CONFIG_HUGETLB_PAGE
1206 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1207 TRANSHUGE_PAGE_DTOR,
1211 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1213 static inline void set_compound_page_dtor(struct page *page,
1214 enum compound_dtor_id compound_dtor)
1216 struct folio *folio = (struct folio *)page;
1218 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1219 VM_BUG_ON_PAGE(!PageHead(page), page);
1220 folio->_folio_dtor = compound_dtor;
1223 static inline void folio_set_compound_dtor(struct folio *folio,
1224 enum compound_dtor_id compound_dtor)
1226 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1227 folio->_folio_dtor = compound_dtor;
1230 void destroy_large_folio(struct folio *folio);
1232 static inline void set_compound_order(struct page *page, unsigned int order)
1234 struct folio *folio = (struct folio *)page;
1236 folio->_folio_order = order;
1238 folio->_folio_nr_pages = 1U << order;
1242 /* Returns the number of bytes in this potentially compound page. */
1243 static inline unsigned long page_size(struct page *page)
1245 return PAGE_SIZE << compound_order(page);
1248 /* Returns the number of bits needed for the number of bytes in a page */
1249 static inline unsigned int page_shift(struct page *page)
1251 return PAGE_SHIFT + compound_order(page);
1255 * thp_order - Order of a transparent huge page.
1256 * @page: Head page of a transparent huge page.
1258 static inline unsigned int thp_order(struct page *page)
1260 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1261 return compound_order(page);
1265 * thp_size - Size of a transparent huge page.
1266 * @page: Head page of a transparent huge page.
1268 * Return: Number of bytes in this page.
1270 static inline unsigned long thp_size(struct page *page)
1272 return PAGE_SIZE << thp_order(page);
1275 void free_compound_page(struct page *page);
1279 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1280 * servicing faults for write access. In the normal case, do always want
1281 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1282 * that do not have writing enabled, when used by access_process_vm.
1284 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1286 if (likely(vma->vm_flags & VM_WRITE))
1287 pte = pte_mkwrite(pte);
1291 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1292 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1294 vm_fault_t finish_fault(struct vm_fault *vmf);
1295 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1299 * Multiple processes may "see" the same page. E.g. for untouched
1300 * mappings of /dev/null, all processes see the same page full of
1301 * zeroes, and text pages of executables and shared libraries have
1302 * only one copy in memory, at most, normally.
1304 * For the non-reserved pages, page_count(page) denotes a reference count.
1305 * page_count() == 0 means the page is free. page->lru is then used for
1306 * freelist management in the buddy allocator.
1307 * page_count() > 0 means the page has been allocated.
1309 * Pages are allocated by the slab allocator in order to provide memory
1310 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1311 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1312 * unless a particular usage is carefully commented. (the responsibility of
1313 * freeing the kmalloc memory is the caller's, of course).
1315 * A page may be used by anyone else who does a __get_free_page().
1316 * In this case, page_count still tracks the references, and should only
1317 * be used through the normal accessor functions. The top bits of page->flags
1318 * and page->virtual store page management information, but all other fields
1319 * are unused and could be used privately, carefully. The management of this
1320 * page is the responsibility of the one who allocated it, and those who have
1321 * subsequently been given references to it.
1323 * The other pages (we may call them "pagecache pages") are completely
1324 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1325 * The following discussion applies only to them.
1327 * A pagecache page contains an opaque `private' member, which belongs to the
1328 * page's address_space. Usually, this is the address of a circular list of
1329 * the page's disk buffers. PG_private must be set to tell the VM to call
1330 * into the filesystem to release these pages.
1332 * A page may belong to an inode's memory mapping. In this case, page->mapping
1333 * is the pointer to the inode, and page->index is the file offset of the page,
1334 * in units of PAGE_SIZE.
1336 * If pagecache pages are not associated with an inode, they are said to be
1337 * anonymous pages. These may become associated with the swapcache, and in that
1338 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1340 * In either case (swapcache or inode backed), the pagecache itself holds one
1341 * reference to the page. Setting PG_private should also increment the
1342 * refcount. The each user mapping also has a reference to the page.
1344 * The pagecache pages are stored in a per-mapping radix tree, which is
1345 * rooted at mapping->i_pages, and indexed by offset.
1346 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1347 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1349 * All pagecache pages may be subject to I/O:
1350 * - inode pages may need to be read from disk,
1351 * - inode pages which have been modified and are MAP_SHARED may need
1352 * to be written back to the inode on disk,
1353 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1354 * modified may need to be swapped out to swap space and (later) to be read
1358 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1359 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1361 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1362 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1364 if (!static_branch_unlikely(&devmap_managed_key))
1366 if (!is_zone_device_page(page))
1368 return __put_devmap_managed_page_refs(page, refs);
1370 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1371 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1375 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1377 static inline bool put_devmap_managed_page(struct page *page)
1379 return put_devmap_managed_page_refs(page, 1);
1382 /* 127: arbitrary random number, small enough to assemble well */
1383 #define folio_ref_zero_or_close_to_overflow(folio) \
1384 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1387 * folio_get - Increment the reference count on a folio.
1388 * @folio: The folio.
1390 * Context: May be called in any context, as long as you know that
1391 * you have a refcount on the folio. If you do not already have one,
1392 * folio_try_get() may be the right interface for you to use.
1394 static inline void folio_get(struct folio *folio)
1396 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1397 folio_ref_inc(folio);
1400 static inline void get_page(struct page *page)
1402 folio_get(page_folio(page));
1405 static inline __must_check bool try_get_page(struct page *page)
1407 page = compound_head(page);
1408 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1415 * folio_put - Decrement the reference count on a folio.
1416 * @folio: The folio.
1418 * If the folio's reference count reaches zero, the memory will be
1419 * released back to the page allocator and may be used by another
1420 * allocation immediately. Do not access the memory or the struct folio
1421 * after calling folio_put() unless you can be sure that it wasn't the
1424 * Context: May be called in process or interrupt context, but not in NMI
1425 * context. May be called while holding a spinlock.
1427 static inline void folio_put(struct folio *folio)
1429 if (folio_put_testzero(folio))
1434 * folio_put_refs - Reduce the reference count on a folio.
1435 * @folio: The folio.
1436 * @refs: The amount to subtract from the folio's reference count.
1438 * If the folio's reference count reaches zero, the memory will be
1439 * released back to the page allocator and may be used by another
1440 * allocation immediately. Do not access the memory or the struct folio
1441 * after calling folio_put_refs() unless you can be sure that these weren't
1442 * the last references.
1444 * Context: May be called in process or interrupt context, but not in NMI
1445 * context. May be called while holding a spinlock.
1447 static inline void folio_put_refs(struct folio *folio, int refs)
1449 if (folio_ref_sub_and_test(folio, refs))
1454 * union release_pages_arg - an array of pages or folios
1456 * release_pages() releases a simple array of multiple pages, and
1457 * accepts various different forms of said page array: either
1458 * a regular old boring array of pages, an array of folios, or
1459 * an array of encoded page pointers.
1461 * The transparent union syntax for this kind of "any of these
1462 * argument types" is all kinds of ugly, so look away.
1465 struct page **pages;
1466 struct folio **folios;
1467 struct encoded_page **encoded_pages;
1468 } release_pages_arg __attribute__ ((__transparent_union__));
1470 void release_pages(release_pages_arg, int nr);
1473 * folios_put - Decrement the reference count on an array of folios.
1474 * @folios: The folios.
1475 * @nr: How many folios there are.
1477 * Like folio_put(), but for an array of folios. This is more efficient
1478 * than writing the loop yourself as it will optimise the locks which
1479 * need to be taken if the folios are freed.
1481 * Context: May be called in process or interrupt context, but not in NMI
1482 * context. May be called while holding a spinlock.
1484 static inline void folios_put(struct folio **folios, unsigned int nr)
1486 release_pages(folios, nr);
1489 static inline void put_page(struct page *page)
1491 struct folio *folio = page_folio(page);
1494 * For some devmap managed pages we need to catch refcount transition
1497 if (put_devmap_managed_page(&folio->page))
1503 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1504 * the page's refcount so that two separate items are tracked: the original page
1505 * reference count, and also a new count of how many pin_user_pages() calls were
1506 * made against the page. ("gup-pinned" is another term for the latter).
1508 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1509 * distinct from normal pages. As such, the unpin_user_page() call (and its
1510 * variants) must be used in order to release gup-pinned pages.
1514 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1515 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1516 * simpler, due to the fact that adding an even power of two to the page
1517 * refcount has the effect of using only the upper N bits, for the code that
1518 * counts up using the bias value. This means that the lower bits are left for
1519 * the exclusive use of the original code that increments and decrements by one
1520 * (or at least, by much smaller values than the bias value).
1522 * Of course, once the lower bits overflow into the upper bits (and this is
1523 * OK, because subtraction recovers the original values), then visual inspection
1524 * no longer suffices to directly view the separate counts. However, for normal
1525 * applications that don't have huge page reference counts, this won't be an
1528 * Locking: the lockless algorithm described in folio_try_get_rcu()
1529 * provides safe operation for get_user_pages(), page_mkclean() and
1530 * other calls that race to set up page table entries.
1532 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1534 void unpin_user_page(struct page *page);
1535 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1537 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1539 void unpin_user_pages(struct page **pages, unsigned long npages);
1541 static inline bool is_cow_mapping(vm_flags_t flags)
1543 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1547 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1550 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1551 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1552 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1553 * underlying memory if ptrace is active, so this is only possible if
1554 * ptrace does not apply. Note that there is no mprotect() to upgrade
1555 * write permissions later.
1557 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1561 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1562 #define SECTION_IN_PAGE_FLAGS
1566 * The identification function is mainly used by the buddy allocator for
1567 * determining if two pages could be buddies. We are not really identifying
1568 * the zone since we could be using the section number id if we do not have
1569 * node id available in page flags.
1570 * We only guarantee that it will return the same value for two combinable
1573 static inline int page_zone_id(struct page *page)
1575 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1578 #ifdef NODE_NOT_IN_PAGE_FLAGS
1579 extern int page_to_nid(const struct page *page);
1581 static inline int page_to_nid(const struct page *page)
1583 struct page *p = (struct page *)page;
1585 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1589 static inline int folio_nid(const struct folio *folio)
1591 return page_to_nid(&folio->page);
1594 #ifdef CONFIG_NUMA_BALANCING
1595 /* page access time bits needs to hold at least 4 seconds */
1596 #define PAGE_ACCESS_TIME_MIN_BITS 12
1597 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1598 #define PAGE_ACCESS_TIME_BUCKETS \
1599 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1601 #define PAGE_ACCESS_TIME_BUCKETS 0
1604 #define PAGE_ACCESS_TIME_MASK \
1605 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1607 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1609 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1612 static inline int cpupid_to_pid(int cpupid)
1614 return cpupid & LAST__PID_MASK;
1617 static inline int cpupid_to_cpu(int cpupid)
1619 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1622 static inline int cpupid_to_nid(int cpupid)
1624 return cpu_to_node(cpupid_to_cpu(cpupid));
1627 static inline bool cpupid_pid_unset(int cpupid)
1629 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1632 static inline bool cpupid_cpu_unset(int cpupid)
1634 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1637 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1639 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1642 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1643 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1644 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1646 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1649 static inline int page_cpupid_last(struct page *page)
1651 return page->_last_cpupid;
1653 static inline void page_cpupid_reset_last(struct page *page)
1655 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1658 static inline int page_cpupid_last(struct page *page)
1660 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1663 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1665 static inline void page_cpupid_reset_last(struct page *page)
1667 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1669 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1671 static inline int xchg_page_access_time(struct page *page, int time)
1675 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1676 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1679 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1681 unsigned int pid_bit;
1683 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1684 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) {
1685 __set_bit(pid_bit, &vma->numab_state->access_pids[1]);
1688 #else /* !CONFIG_NUMA_BALANCING */
1689 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1691 return page_to_nid(page); /* XXX */
1694 static inline int xchg_page_access_time(struct page *page, int time)
1699 static inline int page_cpupid_last(struct page *page)
1701 return page_to_nid(page); /* XXX */
1704 static inline int cpupid_to_nid(int cpupid)
1709 static inline int cpupid_to_pid(int cpupid)
1714 static inline int cpupid_to_cpu(int cpupid)
1719 static inline int cpu_pid_to_cpupid(int nid, int pid)
1724 static inline bool cpupid_pid_unset(int cpupid)
1729 static inline void page_cpupid_reset_last(struct page *page)
1733 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1738 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1741 #endif /* CONFIG_NUMA_BALANCING */
1743 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1746 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1747 * setting tags for all pages to native kernel tag value 0xff, as the default
1748 * value 0x00 maps to 0xff.
1751 static inline u8 page_kasan_tag(const struct page *page)
1755 if (kasan_enabled()) {
1756 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1763 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1765 unsigned long old_flags, flags;
1767 if (!kasan_enabled())
1771 old_flags = READ_ONCE(page->flags);
1774 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1775 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1776 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1779 static inline void page_kasan_tag_reset(struct page *page)
1781 if (kasan_enabled())
1782 page_kasan_tag_set(page, 0xff);
1785 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1787 static inline u8 page_kasan_tag(const struct page *page)
1792 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1793 static inline void page_kasan_tag_reset(struct page *page) { }
1795 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1797 static inline struct zone *page_zone(const struct page *page)
1799 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1802 static inline pg_data_t *page_pgdat(const struct page *page)
1804 return NODE_DATA(page_to_nid(page));
1807 static inline struct zone *folio_zone(const struct folio *folio)
1809 return page_zone(&folio->page);
1812 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1814 return page_pgdat(&folio->page);
1817 #ifdef SECTION_IN_PAGE_FLAGS
1818 static inline void set_page_section(struct page *page, unsigned long section)
1820 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1821 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1824 static inline unsigned long page_to_section(const struct page *page)
1826 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1831 * folio_pfn - Return the Page Frame Number of a folio.
1832 * @folio: The folio.
1834 * A folio may contain multiple pages. The pages have consecutive
1835 * Page Frame Numbers.
1837 * Return: The Page Frame Number of the first page in the folio.
1839 static inline unsigned long folio_pfn(struct folio *folio)
1841 return page_to_pfn(&folio->page);
1844 static inline struct folio *pfn_folio(unsigned long pfn)
1846 return page_folio(pfn_to_page(pfn));
1850 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1851 * @folio: The folio.
1853 * This function checks if a folio has been pinned via a call to
1854 * a function in the pin_user_pages() family.
1856 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1857 * because it means "definitely not pinned for DMA", but true means "probably
1858 * pinned for DMA, but possibly a false positive due to having at least
1859 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1861 * False positives are OK, because: a) it's unlikely for a folio to
1862 * get that many refcounts, and b) all the callers of this routine are
1863 * expected to be able to deal gracefully with a false positive.
1865 * For large folios, the result will be exactly correct. That's because
1866 * we have more tracking data available: the _pincount field is used
1867 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1869 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1871 * Return: True, if it is likely that the page has been "dma-pinned".
1872 * False, if the page is definitely not dma-pinned.
1874 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1876 if (folio_test_large(folio))
1877 return atomic_read(&folio->_pincount) > 0;
1880 * folio_ref_count() is signed. If that refcount overflows, then
1881 * folio_ref_count() returns a negative value, and callers will avoid
1882 * further incrementing the refcount.
1884 * Here, for that overflow case, use the sign bit to count a little
1885 * bit higher via unsigned math, and thus still get an accurate result.
1887 return ((unsigned int)folio_ref_count(folio)) >=
1888 GUP_PIN_COUNTING_BIAS;
1891 static inline bool page_maybe_dma_pinned(struct page *page)
1893 return folio_maybe_dma_pinned(page_folio(page));
1897 * This should most likely only be called during fork() to see whether we
1898 * should break the cow immediately for an anon page on the src mm.
1900 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1902 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1905 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1907 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1910 return page_maybe_dma_pinned(page);
1914 * is_zero_page - Query if a page is a zero page
1915 * @page: The page to query
1917 * This returns true if @page is one of the permanent zero pages.
1919 static inline bool is_zero_page(const struct page *page)
1921 return is_zero_pfn(page_to_pfn(page));
1925 * is_zero_folio - Query if a folio is a zero page
1926 * @folio: The folio to query
1928 * This returns true if @folio is one of the permanent zero pages.
1930 static inline bool is_zero_folio(const struct folio *folio)
1932 return is_zero_page(&folio->page);
1935 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1936 #ifdef CONFIG_MIGRATION
1937 static inline bool is_longterm_pinnable_page(struct page *page)
1940 int mt = get_pageblock_migratetype(page);
1942 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1945 /* The zero page can be "pinned" but gets special handling. */
1946 if (is_zero_page(page))
1949 /* Coherent device memory must always allow eviction. */
1950 if (is_device_coherent_page(page))
1953 /* Otherwise, non-movable zone pages can be pinned. */
1954 return !is_zone_movable_page(page);
1957 static inline bool is_longterm_pinnable_page(struct page *page)
1963 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1965 return is_longterm_pinnable_page(&folio->page);
1968 static inline void set_page_zone(struct page *page, enum zone_type zone)
1970 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1971 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1974 static inline void set_page_node(struct page *page, unsigned long node)
1976 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1977 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1980 static inline void set_page_links(struct page *page, enum zone_type zone,
1981 unsigned long node, unsigned long pfn)
1983 set_page_zone(page, zone);
1984 set_page_node(page, node);
1985 #ifdef SECTION_IN_PAGE_FLAGS
1986 set_page_section(page, pfn_to_section_nr(pfn));
1991 * folio_nr_pages - The number of pages in the folio.
1992 * @folio: The folio.
1994 * Return: A positive power of two.
1996 static inline long folio_nr_pages(struct folio *folio)
1998 if (!folio_test_large(folio))
2001 return folio->_folio_nr_pages;
2003 return 1L << folio->_folio_order;
2008 * compound_nr() returns the number of pages in this potentially compound
2009 * page. compound_nr() can be called on a tail page, and is defined to
2010 * return 1 in that case.
2012 static inline unsigned long compound_nr(struct page *page)
2014 struct folio *folio = (struct folio *)page;
2016 if (!test_bit(PG_head, &folio->flags))
2019 return folio->_folio_nr_pages;
2021 return 1L << folio->_folio_order;
2026 * thp_nr_pages - The number of regular pages in this huge page.
2027 * @page: The head page of a huge page.
2029 static inline int thp_nr_pages(struct page *page)
2031 return folio_nr_pages((struct folio *)page);
2035 * folio_next - Move to the next physical folio.
2036 * @folio: The folio we're currently operating on.
2038 * If you have physically contiguous memory which may span more than
2039 * one folio (eg a &struct bio_vec), use this function to move from one
2040 * folio to the next. Do not use it if the memory is only virtually
2041 * contiguous as the folios are almost certainly not adjacent to each
2042 * other. This is the folio equivalent to writing ``page++``.
2044 * Context: We assume that the folios are refcounted and/or locked at a
2045 * higher level and do not adjust the reference counts.
2046 * Return: The next struct folio.
2048 static inline struct folio *folio_next(struct folio *folio)
2050 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2054 * folio_shift - The size of the memory described by this folio.
2055 * @folio: The folio.
2057 * A folio represents a number of bytes which is a power-of-two in size.
2058 * This function tells you which power-of-two the folio is. See also
2059 * folio_size() and folio_order().
2061 * Context: The caller should have a reference on the folio to prevent
2062 * it from being split. It is not necessary for the folio to be locked.
2063 * Return: The base-2 logarithm of the size of this folio.
2065 static inline unsigned int folio_shift(struct folio *folio)
2067 return PAGE_SHIFT + folio_order(folio);
2071 * folio_size - The number of bytes in a folio.
2072 * @folio: The folio.
2074 * Context: The caller should have a reference on the folio to prevent
2075 * it from being split. It is not necessary for the folio to be locked.
2076 * Return: The number of bytes in this folio.
2078 static inline size_t folio_size(struct folio *folio)
2080 return PAGE_SIZE << folio_order(folio);
2084 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2085 * @folio: The folio.
2087 * folio_estimated_sharers() aims to serve as a function to efficiently
2088 * estimate the number of processes sharing a folio. This is done by
2089 * looking at the precise mapcount of the first subpage in the folio, and
2090 * assuming the other subpages are the same. This may not be true for large
2091 * folios. If you want exact mapcounts for exact calculations, look at
2092 * page_mapcount() or folio_total_mapcount().
2094 * Return: The estimated number of processes sharing a folio.
2096 static inline int folio_estimated_sharers(struct folio *folio)
2098 return page_mapcount(folio_page(folio, 0));
2101 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2102 static inline int arch_make_page_accessible(struct page *page)
2108 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2109 static inline int arch_make_folio_accessible(struct folio *folio)
2112 long i, nr = folio_nr_pages(folio);
2114 for (i = 0; i < nr; i++) {
2115 ret = arch_make_page_accessible(folio_page(folio, i));
2125 * Some inline functions in vmstat.h depend on page_zone()
2127 #include <linux/vmstat.h>
2129 static __always_inline void *lowmem_page_address(const struct page *page)
2131 return page_to_virt(page);
2134 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2135 #define HASHED_PAGE_VIRTUAL
2138 #if defined(WANT_PAGE_VIRTUAL)
2139 static inline void *page_address(const struct page *page)
2141 return page->virtual;
2143 static inline void set_page_address(struct page *page, void *address)
2145 page->virtual = address;
2147 #define page_address_init() do { } while(0)
2150 #if defined(HASHED_PAGE_VIRTUAL)
2151 void *page_address(const struct page *page);
2152 void set_page_address(struct page *page, void *virtual);
2153 void page_address_init(void);
2156 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2157 #define page_address(page) lowmem_page_address(page)
2158 #define set_page_address(page, address) do { } while(0)
2159 #define page_address_init() do { } while(0)
2162 static inline void *folio_address(const struct folio *folio)
2164 return page_address(&folio->page);
2167 extern void *page_rmapping(struct page *page);
2168 extern pgoff_t __page_file_index(struct page *page);
2171 * Return the pagecache index of the passed page. Regular pagecache pages
2172 * use ->index whereas swapcache pages use swp_offset(->private)
2174 static inline pgoff_t page_index(struct page *page)
2176 if (unlikely(PageSwapCache(page)))
2177 return __page_file_index(page);
2182 * Return true only if the page has been allocated with
2183 * ALLOC_NO_WATERMARKS and the low watermark was not
2184 * met implying that the system is under some pressure.
2186 static inline bool page_is_pfmemalloc(const struct page *page)
2189 * lru.next has bit 1 set if the page is allocated from the
2190 * pfmemalloc reserves. Callers may simply overwrite it if
2191 * they do not need to preserve that information.
2193 return (uintptr_t)page->lru.next & BIT(1);
2197 * Return true only if the folio has been allocated with
2198 * ALLOC_NO_WATERMARKS and the low watermark was not
2199 * met implying that the system is under some pressure.
2201 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2204 * lru.next has bit 1 set if the page is allocated from the
2205 * pfmemalloc reserves. Callers may simply overwrite it if
2206 * they do not need to preserve that information.
2208 return (uintptr_t)folio->lru.next & BIT(1);
2212 * Only to be called by the page allocator on a freshly allocated
2215 static inline void set_page_pfmemalloc(struct page *page)
2217 page->lru.next = (void *)BIT(1);
2220 static inline void clear_page_pfmemalloc(struct page *page)
2222 page->lru.next = NULL;
2226 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2228 extern void pagefault_out_of_memory(void);
2230 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2231 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2232 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2235 * Flags passed to show_mem() and show_free_areas() to suppress output in
2238 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2240 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2241 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2243 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2247 * Parameter block passed down to zap_pte_range in exceptional cases.
2249 struct zap_details {
2250 struct folio *single_folio; /* Locked folio to be unmapped */
2251 bool even_cows; /* Zap COWed private pages too? */
2252 zap_flags_t zap_flags; /* Extra flags for zapping */
2256 * Whether to drop the pte markers, for example, the uffd-wp information for
2257 * file-backed memory. This should only be specified when we will completely
2258 * drop the page in the mm, either by truncation or unmapping of the vma. By
2259 * default, the flag is not set.
2261 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2262 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2263 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2265 #ifdef CONFIG_SCHED_MM_CID
2266 void sched_mm_cid_before_execve(struct task_struct *t);
2267 void sched_mm_cid_after_execve(struct task_struct *t);
2268 void sched_mm_cid_fork(struct task_struct *t);
2269 void sched_mm_cid_exit_signals(struct task_struct *t);
2270 static inline int task_mm_cid(struct task_struct *t)
2275 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2276 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2277 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2278 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2279 static inline int task_mm_cid(struct task_struct *t)
2282 * Use the processor id as a fall-back when the mm cid feature is
2283 * disabled. This provides functional per-cpu data structure accesses
2284 * in user-space, althrough it won't provide the memory usage benefits.
2286 return raw_smp_processor_id();
2291 extern bool can_do_mlock(void);
2293 static inline bool can_do_mlock(void) { return false; }
2295 extern int user_shm_lock(size_t, struct ucounts *);
2296 extern void user_shm_unlock(size_t, struct ucounts *);
2298 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2300 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2302 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2305 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2306 unsigned long size);
2307 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2308 unsigned long size, struct zap_details *details);
2309 static inline void zap_vma_pages(struct vm_area_struct *vma)
2311 zap_page_range_single(vma, vma->vm_start,
2312 vma->vm_end - vma->vm_start, NULL);
2314 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2315 struct vm_area_struct *start_vma, unsigned long start,
2316 unsigned long end, bool mm_wr_locked);
2318 struct mmu_notifier_range;
2320 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2321 unsigned long end, unsigned long floor, unsigned long ceiling);
2323 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2324 int follow_pte(struct mm_struct *mm, unsigned long address,
2325 pte_t **ptepp, spinlock_t **ptlp);
2326 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2327 unsigned long *pfn);
2328 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2329 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2330 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2331 void *buf, int len, int write);
2333 extern void truncate_pagecache(struct inode *inode, loff_t new);
2334 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2335 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2336 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2337 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2340 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2341 unsigned long address, unsigned int flags,
2342 struct pt_regs *regs);
2343 extern int fixup_user_fault(struct mm_struct *mm,
2344 unsigned long address, unsigned int fault_flags,
2346 void unmap_mapping_pages(struct address_space *mapping,
2347 pgoff_t start, pgoff_t nr, bool even_cows);
2348 void unmap_mapping_range(struct address_space *mapping,
2349 loff_t const holebegin, loff_t const holelen, int even_cows);
2351 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2352 unsigned long address, unsigned int flags,
2353 struct pt_regs *regs)
2355 /* should never happen if there's no MMU */
2357 return VM_FAULT_SIGBUS;
2359 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2360 unsigned int fault_flags, bool *unlocked)
2362 /* should never happen if there's no MMU */
2366 static inline void unmap_mapping_pages(struct address_space *mapping,
2367 pgoff_t start, pgoff_t nr, bool even_cows) { }
2368 static inline void unmap_mapping_range(struct address_space *mapping,
2369 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2372 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2373 loff_t const holebegin, loff_t const holelen)
2375 unmap_mapping_range(mapping, holebegin, holelen, 0);
2378 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2379 void *buf, int len, unsigned int gup_flags);
2380 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2381 void *buf, int len, unsigned int gup_flags);
2382 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2383 void *buf, int len, unsigned int gup_flags);
2385 long get_user_pages_remote(struct mm_struct *mm,
2386 unsigned long start, unsigned long nr_pages,
2387 unsigned int gup_flags, struct page **pages,
2388 struct vm_area_struct **vmas, int *locked);
2389 long pin_user_pages_remote(struct mm_struct *mm,
2390 unsigned long start, unsigned long nr_pages,
2391 unsigned int gup_flags, struct page **pages,
2392 struct vm_area_struct **vmas, int *locked);
2393 long get_user_pages(unsigned long start, unsigned long nr_pages,
2394 unsigned int gup_flags, struct page **pages,
2395 struct vm_area_struct **vmas);
2396 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2397 unsigned int gup_flags, struct page **pages,
2398 struct vm_area_struct **vmas);
2399 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2400 struct page **pages, unsigned int gup_flags);
2401 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2402 struct page **pages, unsigned int gup_flags);
2404 int get_user_pages_fast(unsigned long start, int nr_pages,
2405 unsigned int gup_flags, struct page **pages);
2406 int pin_user_pages_fast(unsigned long start, int nr_pages,
2407 unsigned int gup_flags, struct page **pages);
2409 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2410 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2411 struct task_struct *task, bool bypass_rlim);
2414 struct page *get_dump_page(unsigned long addr);
2416 bool folio_mark_dirty(struct folio *folio);
2417 bool set_page_dirty(struct page *page);
2418 int set_page_dirty_lock(struct page *page);
2420 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2422 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2423 unsigned long old_addr, struct vm_area_struct *new_vma,
2424 unsigned long new_addr, unsigned long len,
2425 bool need_rmap_locks);
2428 * Flags used by change_protection(). For now we make it a bitmap so
2429 * that we can pass in multiple flags just like parameters. However
2430 * for now all the callers are only use one of the flags at the same
2434 * Whether we should manually check if we can map individual PTEs writable,
2435 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2436 * PTEs automatically in a writable mapping.
2438 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2439 /* Whether this protection change is for NUMA hints */
2440 #define MM_CP_PROT_NUMA (1UL << 1)
2441 /* Whether this change is for write protecting */
2442 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2443 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2444 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2445 MM_CP_UFFD_WP_RESOLVE)
2447 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2448 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2451 * We want to check manually if we can change individual PTEs writable
2452 * if we can't do that automatically for all PTEs in a mapping. For
2453 * private mappings, that's always the case when we have write
2454 * permissions as we properly have to handle COW.
2456 if (vma->vm_flags & VM_SHARED)
2457 return vma_wants_writenotify(vma, vma->vm_page_prot);
2458 return !!(vma->vm_flags & VM_WRITE);
2461 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2463 extern long change_protection(struct mmu_gather *tlb,
2464 struct vm_area_struct *vma, unsigned long start,
2465 unsigned long end, unsigned long cp_flags);
2466 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2467 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2468 unsigned long start, unsigned long end, unsigned long newflags);
2471 * doesn't attempt to fault and will return short.
2473 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2474 unsigned int gup_flags, struct page **pages);
2476 static inline bool get_user_page_fast_only(unsigned long addr,
2477 unsigned int gup_flags, struct page **pagep)
2479 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2482 * per-process(per-mm_struct) statistics.
2484 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2486 return percpu_counter_read_positive(&mm->rss_stat[member]);
2489 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2491 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2493 percpu_counter_add(&mm->rss_stat[member], value);
2495 mm_trace_rss_stat(mm, member);
2498 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2500 percpu_counter_inc(&mm->rss_stat[member]);
2502 mm_trace_rss_stat(mm, member);
2505 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2507 percpu_counter_dec(&mm->rss_stat[member]);
2509 mm_trace_rss_stat(mm, member);
2512 /* Optimized variant when page is already known not to be PageAnon */
2513 static inline int mm_counter_file(struct page *page)
2515 if (PageSwapBacked(page))
2516 return MM_SHMEMPAGES;
2517 return MM_FILEPAGES;
2520 static inline int mm_counter(struct page *page)
2523 return MM_ANONPAGES;
2524 return mm_counter_file(page);
2527 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2529 return get_mm_counter(mm, MM_FILEPAGES) +
2530 get_mm_counter(mm, MM_ANONPAGES) +
2531 get_mm_counter(mm, MM_SHMEMPAGES);
2534 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2536 return max(mm->hiwater_rss, get_mm_rss(mm));
2539 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2541 return max(mm->hiwater_vm, mm->total_vm);
2544 static inline void update_hiwater_rss(struct mm_struct *mm)
2546 unsigned long _rss = get_mm_rss(mm);
2548 if ((mm)->hiwater_rss < _rss)
2549 (mm)->hiwater_rss = _rss;
2552 static inline void update_hiwater_vm(struct mm_struct *mm)
2554 if (mm->hiwater_vm < mm->total_vm)
2555 mm->hiwater_vm = mm->total_vm;
2558 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2560 mm->hiwater_rss = get_mm_rss(mm);
2563 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2564 struct mm_struct *mm)
2566 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2568 if (*maxrss < hiwater_rss)
2569 *maxrss = hiwater_rss;
2572 #if defined(SPLIT_RSS_COUNTING)
2573 void sync_mm_rss(struct mm_struct *mm);
2575 static inline void sync_mm_rss(struct mm_struct *mm)
2580 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2581 static inline int pte_special(pte_t pte)
2586 static inline pte_t pte_mkspecial(pte_t pte)
2592 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2593 static inline int pte_devmap(pte_t pte)
2599 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2601 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2605 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2609 #ifdef __PAGETABLE_P4D_FOLDED
2610 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2611 unsigned long address)
2616 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2619 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2620 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2621 unsigned long address)
2625 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2626 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2629 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2631 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2633 if (mm_pud_folded(mm))
2635 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2638 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2640 if (mm_pud_folded(mm))
2642 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2646 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2647 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2648 unsigned long address)
2653 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2654 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2657 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2659 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2661 if (mm_pmd_folded(mm))
2663 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2666 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2668 if (mm_pmd_folded(mm))
2670 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2675 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2677 atomic_long_set(&mm->pgtables_bytes, 0);
2680 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2682 return atomic_long_read(&mm->pgtables_bytes);
2685 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2687 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2690 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2692 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2696 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2697 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2702 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2703 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2706 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2707 int __pte_alloc_kernel(pmd_t *pmd);
2709 #if defined(CONFIG_MMU)
2711 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2712 unsigned long address)
2714 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2715 NULL : p4d_offset(pgd, address);
2718 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2719 unsigned long address)
2721 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2722 NULL : pud_offset(p4d, address);
2725 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2727 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2728 NULL: pmd_offset(pud, address);
2730 #endif /* CONFIG_MMU */
2732 #if USE_SPLIT_PTE_PTLOCKS
2733 #if ALLOC_SPLIT_PTLOCKS
2734 void __init ptlock_cache_init(void);
2735 extern bool ptlock_alloc(struct page *page);
2736 extern void ptlock_free(struct page *page);
2738 static inline spinlock_t *ptlock_ptr(struct page *page)
2742 #else /* ALLOC_SPLIT_PTLOCKS */
2743 static inline void ptlock_cache_init(void)
2747 static inline bool ptlock_alloc(struct page *page)
2752 static inline void ptlock_free(struct page *page)
2756 static inline spinlock_t *ptlock_ptr(struct page *page)
2760 #endif /* ALLOC_SPLIT_PTLOCKS */
2762 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2764 return ptlock_ptr(pmd_page(*pmd));
2767 static inline bool ptlock_init(struct page *page)
2770 * prep_new_page() initialize page->private (and therefore page->ptl)
2771 * with 0. Make sure nobody took it in use in between.
2773 * It can happen if arch try to use slab for page table allocation:
2774 * slab code uses page->slab_cache, which share storage with page->ptl.
2776 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2777 if (!ptlock_alloc(page))
2779 spin_lock_init(ptlock_ptr(page));
2783 #else /* !USE_SPLIT_PTE_PTLOCKS */
2785 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2787 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2789 return &mm->page_table_lock;
2791 static inline void ptlock_cache_init(void) {}
2792 static inline bool ptlock_init(struct page *page) { return true; }
2793 static inline void ptlock_free(struct page *page) {}
2794 #endif /* USE_SPLIT_PTE_PTLOCKS */
2796 static inline bool pgtable_pte_page_ctor(struct page *page)
2798 if (!ptlock_init(page))
2800 __SetPageTable(page);
2801 inc_lruvec_page_state(page, NR_PAGETABLE);
2805 static inline void pgtable_pte_page_dtor(struct page *page)
2808 __ClearPageTable(page);
2809 dec_lruvec_page_state(page, NR_PAGETABLE);
2812 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2814 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2815 pte_t *__pte = pte_offset_map(pmd, address); \
2821 #define pte_unmap_unlock(pte, ptl) do { \
2826 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2828 #define pte_alloc_map(mm, pmd, address) \
2829 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2831 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2832 (pte_alloc(mm, pmd) ? \
2833 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2835 #define pte_alloc_kernel(pmd, address) \
2836 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2837 NULL: pte_offset_kernel(pmd, address))
2839 #if USE_SPLIT_PMD_PTLOCKS
2841 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2843 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2844 return virt_to_page((void *)((unsigned long) pmd & mask));
2847 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2849 return ptlock_ptr(pmd_pgtable_page(pmd));
2852 static inline bool pmd_ptlock_init(struct page *page)
2854 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2855 page->pmd_huge_pte = NULL;
2857 return ptlock_init(page);
2860 static inline void pmd_ptlock_free(struct page *page)
2862 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2863 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2868 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2872 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2874 return &mm->page_table_lock;
2877 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2878 static inline void pmd_ptlock_free(struct page *page) {}
2880 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2884 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2886 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2891 static inline bool pgtable_pmd_page_ctor(struct page *page)
2893 if (!pmd_ptlock_init(page))
2895 __SetPageTable(page);
2896 inc_lruvec_page_state(page, NR_PAGETABLE);
2900 static inline void pgtable_pmd_page_dtor(struct page *page)
2902 pmd_ptlock_free(page);
2903 __ClearPageTable(page);
2904 dec_lruvec_page_state(page, NR_PAGETABLE);
2908 * No scalability reason to split PUD locks yet, but follow the same pattern
2909 * as the PMD locks to make it easier if we decide to. The VM should not be
2910 * considered ready to switch to split PUD locks yet; there may be places
2911 * which need to be converted from page_table_lock.
2913 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2915 return &mm->page_table_lock;
2918 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2920 spinlock_t *ptl = pud_lockptr(mm, pud);
2926 extern void __init pagecache_init(void);
2927 extern void free_initmem(void);
2930 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2931 * into the buddy system. The freed pages will be poisoned with pattern
2932 * "poison" if it's within range [0, UCHAR_MAX].
2933 * Return pages freed into the buddy system.
2935 extern unsigned long free_reserved_area(void *start, void *end,
2936 int poison, const char *s);
2938 extern void adjust_managed_page_count(struct page *page, long count);
2940 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2942 /* Free the reserved page into the buddy system, so it gets managed. */
2943 static inline void free_reserved_page(struct page *page)
2945 ClearPageReserved(page);
2946 init_page_count(page);
2948 adjust_managed_page_count(page, 1);
2950 #define free_highmem_page(page) free_reserved_page(page)
2952 static inline void mark_page_reserved(struct page *page)
2954 SetPageReserved(page);
2955 adjust_managed_page_count(page, -1);
2959 * Default method to free all the __init memory into the buddy system.
2960 * The freed pages will be poisoned with pattern "poison" if it's within
2961 * range [0, UCHAR_MAX].
2962 * Return pages freed into the buddy system.
2964 static inline unsigned long free_initmem_default(int poison)
2966 extern char __init_begin[], __init_end[];
2968 return free_reserved_area(&__init_begin, &__init_end,
2969 poison, "unused kernel image (initmem)");
2972 static inline unsigned long get_num_physpages(void)
2975 unsigned long phys_pages = 0;
2977 for_each_online_node(nid)
2978 phys_pages += node_present_pages(nid);
2984 * Using memblock node mappings, an architecture may initialise its
2985 * zones, allocate the backing mem_map and account for memory holes in an
2986 * architecture independent manner.
2988 * An architecture is expected to register range of page frames backed by
2989 * physical memory with memblock_add[_node]() before calling
2990 * free_area_init() passing in the PFN each zone ends at. At a basic
2991 * usage, an architecture is expected to do something like
2993 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2995 * for_each_valid_physical_page_range()
2996 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2997 * free_area_init(max_zone_pfns);
2999 void free_area_init(unsigned long *max_zone_pfn);
3000 unsigned long node_map_pfn_alignment(void);
3001 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
3002 unsigned long end_pfn);
3003 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3004 unsigned long end_pfn);
3005 extern void get_pfn_range_for_nid(unsigned int nid,
3006 unsigned long *start_pfn, unsigned long *end_pfn);
3009 static inline int early_pfn_to_nid(unsigned long pfn)
3014 /* please see mm/page_alloc.c */
3015 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3018 extern void set_dma_reserve(unsigned long new_dma_reserve);
3019 extern void memmap_init_range(unsigned long, int, unsigned long,
3020 unsigned long, unsigned long, enum meminit_context,
3021 struct vmem_altmap *, int migratetype);
3022 extern void setup_per_zone_wmarks(void);
3023 extern void calculate_min_free_kbytes(void);
3024 extern int __meminit init_per_zone_wmark_min(void);
3025 extern void mem_init(void);
3026 extern void __init mmap_init(void);
3028 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
3029 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
3031 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
3033 extern long si_mem_available(void);
3034 extern void si_meminfo(struct sysinfo * val);
3035 extern void si_meminfo_node(struct sysinfo *val, int nid);
3036 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3037 extern unsigned long arch_reserved_kernel_pages(void);
3040 extern __printf(3, 4)
3041 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3043 extern void setup_per_cpu_pageset(void);
3046 extern int min_free_kbytes;
3047 extern int watermark_boost_factor;
3048 extern int watermark_scale_factor;
3051 extern atomic_long_t mmap_pages_allocated;
3052 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3054 /* interval_tree.c */
3055 void vma_interval_tree_insert(struct vm_area_struct *node,
3056 struct rb_root_cached *root);
3057 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3058 struct vm_area_struct *prev,
3059 struct rb_root_cached *root);
3060 void vma_interval_tree_remove(struct vm_area_struct *node,
3061 struct rb_root_cached *root);
3062 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3063 unsigned long start, unsigned long last);
3064 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3065 unsigned long start, unsigned long last);
3067 #define vma_interval_tree_foreach(vma, root, start, last) \
3068 for (vma = vma_interval_tree_iter_first(root, start, last); \
3069 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3071 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3072 struct rb_root_cached *root);
3073 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3074 struct rb_root_cached *root);
3075 struct anon_vma_chain *
3076 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3077 unsigned long start, unsigned long last);
3078 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3079 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3080 #ifdef CONFIG_DEBUG_VM_RB
3081 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3084 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3085 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3086 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3089 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3090 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3091 unsigned long start, unsigned long end, pgoff_t pgoff,
3092 struct vm_area_struct *next);
3093 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3094 unsigned long start, unsigned long end, pgoff_t pgoff);
3095 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3096 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3097 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3098 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3099 struct anon_vma_name *);
3100 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3101 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3102 unsigned long addr, int new_below);
3103 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3104 unsigned long addr, int new_below);
3105 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3106 extern void unlink_file_vma(struct vm_area_struct *);
3107 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3108 unsigned long addr, unsigned long len, pgoff_t pgoff,
3109 bool *need_rmap_locks);
3110 extern void exit_mmap(struct mm_struct *);
3112 static inline int check_data_rlimit(unsigned long rlim,
3114 unsigned long start,
3115 unsigned long end_data,
3116 unsigned long start_data)
3118 if (rlim < RLIM_INFINITY) {
3119 if (((new - start) + (end_data - start_data)) > rlim)
3126 extern int mm_take_all_locks(struct mm_struct *mm);
3127 extern void mm_drop_all_locks(struct mm_struct *mm);
3129 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3130 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3131 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3132 extern struct file *get_task_exe_file(struct task_struct *task);
3134 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3135 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3137 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3138 const struct vm_special_mapping *sm);
3139 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3140 unsigned long addr, unsigned long len,
3141 unsigned long flags,
3142 const struct vm_special_mapping *spec);
3143 /* This is an obsolete alternative to _install_special_mapping. */
3144 extern int install_special_mapping(struct mm_struct *mm,
3145 unsigned long addr, unsigned long len,
3146 unsigned long flags, struct page **pages);
3148 unsigned long randomize_stack_top(unsigned long stack_top);
3149 unsigned long randomize_page(unsigned long start, unsigned long range);
3151 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3153 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3154 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3155 struct list_head *uf);
3156 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3157 unsigned long len, unsigned long prot, unsigned long flags,
3158 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3159 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3160 unsigned long start, size_t len, struct list_head *uf,
3162 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3163 struct list_head *uf);
3164 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3167 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3168 unsigned long start, unsigned long end,
3169 struct list_head *uf, bool downgrade);
3170 extern int __mm_populate(unsigned long addr, unsigned long len,
3172 static inline void mm_populate(unsigned long addr, unsigned long len)
3175 (void) __mm_populate(addr, len, 1);
3178 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3181 /* These take the mm semaphore themselves */
3182 extern int __must_check vm_brk(unsigned long, unsigned long);
3183 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3184 extern int vm_munmap(unsigned long, size_t);
3185 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3186 unsigned long, unsigned long,
3187 unsigned long, unsigned long);
3189 struct vm_unmapped_area_info {
3190 #define VM_UNMAPPED_AREA_TOPDOWN 1
3191 unsigned long flags;
3192 unsigned long length;
3193 unsigned long low_limit;
3194 unsigned long high_limit;
3195 unsigned long align_mask;
3196 unsigned long align_offset;
3199 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3202 extern void truncate_inode_pages(struct address_space *, loff_t);
3203 extern void truncate_inode_pages_range(struct address_space *,
3204 loff_t lstart, loff_t lend);
3205 extern void truncate_inode_pages_final(struct address_space *);
3207 /* generic vm_area_ops exported for stackable file systems */
3208 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3209 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3210 pgoff_t start_pgoff, pgoff_t end_pgoff);
3211 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3213 extern unsigned long stack_guard_gap;
3214 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3215 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
3217 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3218 extern int expand_downwards(struct vm_area_struct *vma,
3219 unsigned long address);
3221 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
3223 #define expand_upwards(vma, address) (0)
3226 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3227 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3228 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3229 struct vm_area_struct **pprev);
3232 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3233 * NULL if none. Assume start_addr < end_addr.
3235 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3236 unsigned long start_addr, unsigned long end_addr);
3239 * vma_lookup() - Find a VMA at a specific address
3240 * @mm: The process address space.
3241 * @addr: The user address.
3243 * Return: The vm_area_struct at the given address, %NULL otherwise.
3246 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3248 return mtree_load(&mm->mm_mt, addr);
3251 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3253 unsigned long vm_start = vma->vm_start;
3255 if (vma->vm_flags & VM_GROWSDOWN) {
3256 vm_start -= stack_guard_gap;
3257 if (vm_start > vma->vm_start)
3263 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3265 unsigned long vm_end = vma->vm_end;
3267 if (vma->vm_flags & VM_GROWSUP) {
3268 vm_end += stack_guard_gap;
3269 if (vm_end < vma->vm_end)
3270 vm_end = -PAGE_SIZE;
3275 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3277 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3280 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3281 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3282 unsigned long vm_start, unsigned long vm_end)
3284 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3286 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3292 static inline bool range_in_vma(struct vm_area_struct *vma,
3293 unsigned long start, unsigned long end)
3295 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3299 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3300 void vma_set_page_prot(struct vm_area_struct *vma);
3302 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3306 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3308 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3312 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3314 #ifdef CONFIG_NUMA_BALANCING
3315 unsigned long change_prot_numa(struct vm_area_struct *vma,
3316 unsigned long start, unsigned long end);
3319 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3320 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3321 unsigned long pfn, unsigned long size, pgprot_t);
3322 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3323 unsigned long pfn, unsigned long size, pgprot_t prot);
3324 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3325 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3326 struct page **pages, unsigned long *num);
3327 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3329 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3331 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3333 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3334 unsigned long pfn, pgprot_t pgprot);
3335 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3337 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3338 unsigned long addr, pfn_t pfn);
3339 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3341 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3342 unsigned long addr, struct page *page)
3344 int err = vm_insert_page(vma, addr, page);
3347 return VM_FAULT_OOM;
3348 if (err < 0 && err != -EBUSY)
3349 return VM_FAULT_SIGBUS;
3351 return VM_FAULT_NOPAGE;
3354 #ifndef io_remap_pfn_range
3355 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3356 unsigned long addr, unsigned long pfn,
3357 unsigned long size, pgprot_t prot)
3359 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3363 static inline vm_fault_t vmf_error(int err)
3366 return VM_FAULT_OOM;
3367 return VM_FAULT_SIGBUS;
3370 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3371 unsigned int foll_flags);
3373 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3375 if (vm_fault & VM_FAULT_OOM)
3377 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3378 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3379 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3385 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3386 * a (NUMA hinting) fault is required.
3388 static inline bool gup_can_follow_protnone(unsigned int flags)
3391 * FOLL_FORCE has to be able to make progress even if the VMA is
3392 * inaccessible. Further, FOLL_FORCE access usually does not represent
3393 * application behaviour and we should avoid triggering NUMA hinting
3396 return flags & FOLL_FORCE;
3399 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3400 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3401 unsigned long size, pte_fn_t fn, void *data);
3402 extern int apply_to_existing_page_range(struct mm_struct *mm,
3403 unsigned long address, unsigned long size,
3404 pte_fn_t fn, void *data);
3406 #ifdef CONFIG_PAGE_POISONING
3407 extern void __kernel_poison_pages(struct page *page, int numpages);
3408 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3409 extern bool _page_poisoning_enabled_early;
3410 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3411 static inline bool page_poisoning_enabled(void)
3413 return _page_poisoning_enabled_early;
3416 * For use in fast paths after init_mem_debugging() has run, or when a
3417 * false negative result is not harmful when called too early.
3419 static inline bool page_poisoning_enabled_static(void)
3421 return static_branch_unlikely(&_page_poisoning_enabled);
3423 static inline void kernel_poison_pages(struct page *page, int numpages)
3425 if (page_poisoning_enabled_static())
3426 __kernel_poison_pages(page, numpages);
3428 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3430 if (page_poisoning_enabled_static())
3431 __kernel_unpoison_pages(page, numpages);
3434 static inline bool page_poisoning_enabled(void) { return false; }
3435 static inline bool page_poisoning_enabled_static(void) { return false; }
3436 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3437 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3438 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3441 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3442 static inline bool want_init_on_alloc(gfp_t flags)
3444 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3447 return flags & __GFP_ZERO;
3450 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3451 static inline bool want_init_on_free(void)
3453 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3457 extern bool _debug_pagealloc_enabled_early;
3458 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3460 static inline bool debug_pagealloc_enabled(void)
3462 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3463 _debug_pagealloc_enabled_early;
3467 * For use in fast paths after init_debug_pagealloc() has run, or when a
3468 * false negative result is not harmful when called too early.
3470 static inline bool debug_pagealloc_enabled_static(void)
3472 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3475 return static_branch_unlikely(&_debug_pagealloc_enabled);
3478 #ifdef CONFIG_DEBUG_PAGEALLOC
3480 * To support DEBUG_PAGEALLOC architecture must ensure that
3481 * __kernel_map_pages() never fails
3483 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3485 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3487 if (debug_pagealloc_enabled_static())
3488 __kernel_map_pages(page, numpages, 1);
3491 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3493 if (debug_pagealloc_enabled_static())
3494 __kernel_map_pages(page, numpages, 0);
3496 #else /* CONFIG_DEBUG_PAGEALLOC */
3497 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3498 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3499 #endif /* CONFIG_DEBUG_PAGEALLOC */
3501 #ifdef __HAVE_ARCH_GATE_AREA
3502 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3503 extern int in_gate_area_no_mm(unsigned long addr);
3504 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3506 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3510 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3511 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3515 #endif /* __HAVE_ARCH_GATE_AREA */
3517 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3519 #ifdef CONFIG_SYSCTL
3520 extern int sysctl_drop_caches;
3521 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3525 void drop_slab(void);
3528 #define randomize_va_space 0
3530 extern int randomize_va_space;
3533 const char * arch_vma_name(struct vm_area_struct *vma);
3535 void print_vma_addr(char *prefix, unsigned long rip);
3537 static inline void print_vma_addr(char *prefix, unsigned long rip)
3542 void *sparse_buffer_alloc(unsigned long size);
3543 struct page * __populate_section_memmap(unsigned long pfn,
3544 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3545 struct dev_pagemap *pgmap);
3546 void pmd_init(void *addr);
3547 void pud_init(void *addr);
3548 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3549 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3550 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3551 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3552 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3553 struct vmem_altmap *altmap, struct page *reuse);
3554 void *vmemmap_alloc_block(unsigned long size, int node);
3556 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3557 struct vmem_altmap *altmap);
3558 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3559 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3560 unsigned long addr, unsigned long next);
3561 int vmemmap_check_pmd(pmd_t *pmd, int node,
3562 unsigned long addr, unsigned long next);
3563 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3564 int node, struct vmem_altmap *altmap);
3565 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3566 int node, struct vmem_altmap *altmap);
3567 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3568 struct vmem_altmap *altmap);
3569 void vmemmap_populate_print_last(void);
3570 #ifdef CONFIG_MEMORY_HOTPLUG
3571 void vmemmap_free(unsigned long start, unsigned long end,
3572 struct vmem_altmap *altmap);
3575 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP
3576 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3577 struct dev_pagemap *pgmap)
3579 return is_power_of_2(sizeof(struct page)) &&
3580 pgmap && (pgmap_vmemmap_nr(pgmap) > 1) && !altmap;
3583 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3584 struct dev_pagemap *pgmap)
3590 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3591 unsigned long nr_pages);
3594 MF_COUNT_INCREASED = 1 << 0,
3595 MF_ACTION_REQUIRED = 1 << 1,
3596 MF_MUST_KILL = 1 << 2,
3597 MF_SOFT_OFFLINE = 1 << 3,
3598 MF_UNPOISON = 1 << 4,
3599 MF_SW_SIMULATED = 1 << 5,
3600 MF_NO_RETRY = 1 << 6,
3602 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3603 unsigned long count, int mf_flags);
3604 extern int memory_failure(unsigned long pfn, int flags);
3605 extern void memory_failure_queue_kick(int cpu);
3606 extern int unpoison_memory(unsigned long pfn);
3607 extern void shake_page(struct page *p);
3608 extern atomic_long_t num_poisoned_pages __read_mostly;
3609 extern int soft_offline_page(unsigned long pfn, int flags);
3610 #ifdef CONFIG_MEMORY_FAILURE
3611 extern void memory_failure_queue(unsigned long pfn, int flags);
3612 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3613 bool *migratable_cleared);
3614 void num_poisoned_pages_inc(unsigned long pfn);
3615 void num_poisoned_pages_sub(unsigned long pfn, long i);
3616 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3618 static inline void memory_failure_queue(unsigned long pfn, int flags)
3622 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3623 bool *migratable_cleared)
3628 static inline void num_poisoned_pages_inc(unsigned long pfn)
3632 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3637 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3638 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3639 struct vm_area_struct *vma, struct list_head *to_kill,
3640 unsigned long ksm_addr);
3643 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3644 extern void memblk_nr_poison_inc(unsigned long pfn);
3645 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3647 static inline void memblk_nr_poison_inc(unsigned long pfn)
3651 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3656 #ifndef arch_memory_failure
3657 static inline int arch_memory_failure(unsigned long pfn, int flags)
3663 #ifndef arch_is_platform_page
3664 static inline bool arch_is_platform_page(u64 paddr)
3671 * Error handlers for various types of pages.
3674 MF_IGNORED, /* Error: cannot be handled */
3675 MF_FAILED, /* Error: handling failed */
3676 MF_DELAYED, /* Will be handled later */
3677 MF_RECOVERED, /* Successfully recovered */
3680 enum mf_action_page_type {
3682 MF_MSG_KERNEL_HIGH_ORDER,
3684 MF_MSG_DIFFERENT_COMPOUND,
3687 MF_MSG_UNMAP_FAILED,
3688 MF_MSG_DIRTY_SWAPCACHE,
3689 MF_MSG_CLEAN_SWAPCACHE,
3690 MF_MSG_DIRTY_MLOCKED_LRU,
3691 MF_MSG_CLEAN_MLOCKED_LRU,
3692 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3693 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3696 MF_MSG_TRUNCATED_LRU,
3704 * Sysfs entries for memory failure handling statistics.
3706 extern const struct attribute_group memory_failure_attr_group;
3708 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3709 extern void clear_huge_page(struct page *page,
3710 unsigned long addr_hint,
3711 unsigned int pages_per_huge_page);
3712 int copy_user_large_folio(struct folio *dst, struct folio *src,
3713 unsigned long addr_hint,
3714 struct vm_area_struct *vma);
3715 long copy_folio_from_user(struct folio *dst_folio,
3716 const void __user *usr_src,
3717 bool allow_pagefault);
3720 * vma_is_special_huge - Are transhuge page-table entries considered special?
3721 * @vma: Pointer to the struct vm_area_struct to consider
3723 * Whether transhuge page-table entries are considered "special" following
3724 * the definition in vm_normal_page().
3726 * Return: true if transhuge page-table entries should be considered special,
3729 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3731 return vma_is_dax(vma) || (vma->vm_file &&
3732 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3735 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3737 #ifdef CONFIG_DEBUG_PAGEALLOC
3738 extern unsigned int _debug_guardpage_minorder;
3739 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3741 static inline unsigned int debug_guardpage_minorder(void)
3743 return _debug_guardpage_minorder;
3746 static inline bool debug_guardpage_enabled(void)
3748 return static_branch_unlikely(&_debug_guardpage_enabled);
3751 static inline bool page_is_guard(struct page *page)
3753 if (!debug_guardpage_enabled())
3756 return PageGuard(page);
3759 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3760 static inline bool debug_guardpage_enabled(void) { return false; }
3761 static inline bool page_is_guard(struct page *page) { return false; }
3762 #endif /* CONFIG_DEBUG_PAGEALLOC */
3764 #if MAX_NUMNODES > 1
3765 void __init setup_nr_node_ids(void);
3767 static inline void setup_nr_node_ids(void) {}
3770 extern int memcmp_pages(struct page *page1, struct page *page2);
3772 static inline int pages_identical(struct page *page1, struct page *page2)
3774 return !memcmp_pages(page1, page2);
3777 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3778 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3779 pgoff_t first_index, pgoff_t nr,
3780 pgoff_t bitmap_pgoff,
3781 unsigned long *bitmap,
3785 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3786 pgoff_t first_index, pgoff_t nr);
3789 extern int sysctl_nr_trim_pages;
3791 #ifdef CONFIG_PRINTK
3792 void mem_dump_obj(void *object);
3794 static inline void mem_dump_obj(void *object) {}
3798 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3799 * @seals: the seals to check
3800 * @vma: the vma to operate on
3802 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3803 * the vma flags. Return 0 if check pass, or <0 for errors.
3805 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3807 if (seals & F_SEAL_FUTURE_WRITE) {
3809 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3810 * "future write" seal active.
3812 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3816 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3817 * MAP_SHARED and read-only, take care to not allow mprotect to
3818 * revert protections on such mappings. Do this only for shared
3819 * mappings. For private mappings, don't need to mask
3820 * VM_MAYWRITE as we still want them to be COW-writable.
3822 if (vma->vm_flags & VM_SHARED)
3823 vm_flags_clear(vma, VM_MAYWRITE);
3829 #ifdef CONFIG_ANON_VMA_NAME
3830 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3831 unsigned long len_in,
3832 struct anon_vma_name *anon_name);
3835 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3836 unsigned long len_in, struct anon_vma_name *anon_name) {
3841 #endif /* _LINUX_MM_H */