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 | VM_STACK_EARLY)
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
402 #define VM_STACK_EARLY VM_GROWSDOWN
404 #define VM_STACK VM_GROWSDOWN
405 #define VM_STACK_EARLY 0
408 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
410 /* VMA basic access permission flags */
411 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
415 * Special vmas that are non-mergable, non-mlock()able.
417 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
419 /* This mask prevents VMA from being scanned with khugepaged */
420 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
422 /* This mask defines which mm->def_flags a process can inherit its parent */
423 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
425 /* This mask represents all the VMA flag bits used by mlock */
426 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
428 /* Arch-specific flags to clear when updating VM flags on protection change */
429 #ifndef VM_ARCH_CLEAR
430 # define VM_ARCH_CLEAR VM_NONE
432 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
435 * mapping from the currently active vm_flags protection bits (the
436 * low four bits) to a page protection mask..
440 * The default fault flags that should be used by most of the
441 * arch-specific page fault handlers.
443 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
444 FAULT_FLAG_KILLABLE | \
445 FAULT_FLAG_INTERRUPTIBLE)
448 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
449 * @flags: Fault flags.
451 * This is mostly used for places where we want to try to avoid taking
452 * the mmap_lock for too long a time when waiting for another condition
453 * to change, in which case we can try to be polite to release the
454 * mmap_lock in the first round to avoid potential starvation of other
455 * processes that would also want the mmap_lock.
457 * Return: true if the page fault allows retry and this is the first
458 * attempt of the fault handling; false otherwise.
460 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
462 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
463 (!(flags & FAULT_FLAG_TRIED));
466 #define FAULT_FLAG_TRACE \
467 { FAULT_FLAG_WRITE, "WRITE" }, \
468 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
469 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
470 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
471 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
472 { FAULT_FLAG_TRIED, "TRIED" }, \
473 { FAULT_FLAG_USER, "USER" }, \
474 { FAULT_FLAG_REMOTE, "REMOTE" }, \
475 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
476 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
477 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
480 * vm_fault is filled by the pagefault handler and passed to the vma's
481 * ->fault function. The vma's ->fault is responsible for returning a bitmask
482 * of VM_FAULT_xxx flags that give details about how the fault was handled.
484 * MM layer fills up gfp_mask for page allocations but fault handler might
485 * alter it if its implementation requires a different allocation context.
487 * pgoff should be used in favour of virtual_address, if possible.
491 struct vm_area_struct *vma; /* Target VMA */
492 gfp_t gfp_mask; /* gfp mask to be used for allocations */
493 pgoff_t pgoff; /* Logical page offset based on vma */
494 unsigned long address; /* Faulting virtual address - masked */
495 unsigned long real_address; /* Faulting virtual address - unmasked */
497 enum fault_flag flags; /* FAULT_FLAG_xxx flags
498 * XXX: should really be 'const' */
499 pmd_t *pmd; /* Pointer to pmd entry matching
501 pud_t *pud; /* Pointer to pud entry matching
505 pte_t orig_pte; /* Value of PTE at the time of fault */
506 pmd_t orig_pmd; /* Value of PMD at the time of fault,
507 * used by PMD fault only.
511 struct page *cow_page; /* Page handler may use for COW fault */
512 struct page *page; /* ->fault handlers should return a
513 * page here, unless VM_FAULT_NOPAGE
514 * is set (which is also implied by
517 /* These three entries are valid only while holding ptl lock */
518 pte_t *pte; /* Pointer to pte entry matching
519 * the 'address'. NULL if the page
520 * table hasn't been allocated.
522 spinlock_t *ptl; /* Page table lock.
523 * Protects pte page table if 'pte'
524 * is not NULL, otherwise pmd.
526 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
527 * vm_ops->map_pages() sets up a page
528 * table from atomic context.
529 * do_fault_around() pre-allocates
530 * page table to avoid allocation from
535 /* page entry size for vm->huge_fault() */
536 enum page_entry_size {
543 * These are the virtual MM functions - opening of an area, closing and
544 * unmapping it (needed to keep files on disk up-to-date etc), pointer
545 * to the functions called when a no-page or a wp-page exception occurs.
547 struct vm_operations_struct {
548 void (*open)(struct vm_area_struct * area);
550 * @close: Called when the VMA is being removed from the MM.
551 * Context: User context. May sleep. Caller holds mmap_lock.
553 void (*close)(struct vm_area_struct * area);
554 /* Called any time before splitting to check if it's allowed */
555 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
556 int (*mremap)(struct vm_area_struct *area);
558 * Called by mprotect() to make driver-specific permission
559 * checks before mprotect() is finalised. The VMA must not
560 * be modified. Returns 0 if mprotect() can proceed.
562 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
563 unsigned long end, unsigned long newflags);
564 vm_fault_t (*fault)(struct vm_fault *vmf);
565 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
566 enum page_entry_size pe_size);
567 vm_fault_t (*map_pages)(struct vm_fault *vmf,
568 pgoff_t start_pgoff, pgoff_t end_pgoff);
569 unsigned long (*pagesize)(struct vm_area_struct * area);
571 /* notification that a previously read-only page is about to become
572 * writable, if an error is returned it will cause a SIGBUS */
573 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
575 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
576 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
578 /* called by access_process_vm when get_user_pages() fails, typically
579 * for use by special VMAs. See also generic_access_phys() for a generic
580 * implementation useful for any iomem mapping.
582 int (*access)(struct vm_area_struct *vma, unsigned long addr,
583 void *buf, int len, int write);
585 /* Called by the /proc/PID/maps code to ask the vma whether it
586 * has a special name. Returning non-NULL will also cause this
587 * vma to be dumped unconditionally. */
588 const char *(*name)(struct vm_area_struct *vma);
592 * set_policy() op must add a reference to any non-NULL @new mempolicy
593 * to hold the policy upon return. Caller should pass NULL @new to
594 * remove a policy and fall back to surrounding context--i.e. do not
595 * install a MPOL_DEFAULT policy, nor the task or system default
598 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
601 * get_policy() op must add reference [mpol_get()] to any policy at
602 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
603 * in mm/mempolicy.c will do this automatically.
604 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
605 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
606 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
607 * must return NULL--i.e., do not "fallback" to task or system default
610 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
614 * Called by vm_normal_page() for special PTEs to find the
615 * page for @addr. This is useful if the default behavior
616 * (using pte_page()) would not find the correct page.
618 struct page *(*find_special_page)(struct vm_area_struct *vma,
622 #ifdef CONFIG_NUMA_BALANCING
623 static inline void vma_numab_state_init(struct vm_area_struct *vma)
625 vma->numab_state = NULL;
627 static inline void vma_numab_state_free(struct vm_area_struct *vma)
629 kfree(vma->numab_state);
632 static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
633 static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
634 #endif /* CONFIG_NUMA_BALANCING */
636 #ifdef CONFIG_PER_VMA_LOCK
638 * Try to read-lock a vma. The function is allowed to occasionally yield false
639 * locked result to avoid performance overhead, in which case we fall back to
640 * using mmap_lock. The function should never yield false unlocked result.
642 static inline bool vma_start_read(struct vm_area_struct *vma)
645 * Check before locking. A race might cause false locked result.
646 * We can use READ_ONCE() for the mm_lock_seq here, and don't need
647 * ACQUIRE semantics, because this is just a lockless check whose result
648 * we don't rely on for anything - the mm_lock_seq read against which we
649 * need ordering is below.
651 if (READ_ONCE(vma->vm_lock_seq) == READ_ONCE(vma->vm_mm->mm_lock_seq))
654 if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
658 * Overflow might produce false locked result.
659 * False unlocked result is impossible because we modify and check
660 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
661 * modification invalidates all existing locks.
663 * We must use ACQUIRE semantics for the mm_lock_seq so that if we are
664 * racing with vma_end_write_all(), we only start reading from the VMA
665 * after it has been unlocked.
666 * This pairs with RELEASE semantics in vma_end_write_all().
668 if (unlikely(vma->vm_lock_seq == smp_load_acquire(&vma->vm_mm->mm_lock_seq))) {
669 up_read(&vma->vm_lock->lock);
675 static inline void vma_end_read(struct vm_area_struct *vma)
677 rcu_read_lock(); /* keeps vma alive till the end of up_read */
678 up_read(&vma->vm_lock->lock);
682 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
684 mmap_assert_write_locked(vma->vm_mm);
687 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
688 * mm->mm_lock_seq can't be concurrently modified.
690 *mm_lock_seq = vma->vm_mm->mm_lock_seq;
691 return (vma->vm_lock_seq == *mm_lock_seq);
695 * Begin writing to a VMA.
696 * Exclude concurrent readers under the per-VMA lock until the currently
697 * write-locked mmap_lock is dropped or downgraded.
699 static inline void vma_start_write(struct vm_area_struct *vma)
703 if (__is_vma_write_locked(vma, &mm_lock_seq))
706 down_write(&vma->vm_lock->lock);
708 * We should use WRITE_ONCE() here because we can have concurrent reads
709 * from the early lockless pessimistic check in vma_start_read().
710 * We don't really care about the correctness of that early check, but
711 * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy.
713 WRITE_ONCE(vma->vm_lock_seq, mm_lock_seq);
714 up_write(&vma->vm_lock->lock);
717 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
721 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
724 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
726 /* When detaching vma should be write-locked */
728 vma_assert_write_locked(vma);
729 vma->detached = detached;
732 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
733 unsigned long address);
735 #else /* CONFIG_PER_VMA_LOCK */
737 static inline bool vma_start_read(struct vm_area_struct *vma)
739 static inline void vma_end_read(struct vm_area_struct *vma) {}
740 static inline void vma_start_write(struct vm_area_struct *vma) {}
741 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
742 static inline void vma_mark_detached(struct vm_area_struct *vma,
745 static inline struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
746 unsigned long address)
751 #endif /* CONFIG_PER_VMA_LOCK */
754 * WARNING: vma_init does not initialize vma->vm_lock.
755 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
757 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
759 static const struct vm_operations_struct dummy_vm_ops = {};
761 memset(vma, 0, sizeof(*vma));
763 vma->vm_ops = &dummy_vm_ops;
764 INIT_LIST_HEAD(&vma->anon_vma_chain);
765 vma_mark_detached(vma, false);
766 vma_numab_state_init(vma);
769 /* Use when VMA is not part of the VMA tree and needs no locking */
770 static inline void vm_flags_init(struct vm_area_struct *vma,
773 ACCESS_PRIVATE(vma, __vm_flags) = flags;
776 /* Use when VMA is part of the VMA tree and modifications need coordination */
777 static inline void vm_flags_reset(struct vm_area_struct *vma,
780 vma_start_write(vma);
781 vm_flags_init(vma, flags);
784 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
787 vma_start_write(vma);
788 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
791 static inline void vm_flags_set(struct vm_area_struct *vma,
794 vma_start_write(vma);
795 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
798 static inline void vm_flags_clear(struct vm_area_struct *vma,
801 vma_start_write(vma);
802 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
806 * Use only if VMA is not part of the VMA tree or has no other users and
807 * therefore needs no locking.
809 static inline void __vm_flags_mod(struct vm_area_struct *vma,
810 vm_flags_t set, vm_flags_t clear)
812 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
816 * Use only when the order of set/clear operations is unimportant, otherwise
817 * use vm_flags_{set|clear} explicitly.
819 static inline void vm_flags_mod(struct vm_area_struct *vma,
820 vm_flags_t set, vm_flags_t clear)
822 vma_start_write(vma);
823 __vm_flags_mod(vma, set, clear);
826 static inline void vma_set_anonymous(struct vm_area_struct *vma)
831 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
836 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
838 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
843 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
844 VM_STACK_INCOMPLETE_SETUP)
850 static inline bool vma_is_foreign(struct vm_area_struct *vma)
855 if (current->mm != vma->vm_mm)
861 static inline bool vma_is_accessible(struct vm_area_struct *vma)
863 return vma->vm_flags & VM_ACCESS_FLAGS;
867 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
869 return mas_find(&vmi->mas, max - 1);
872 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
875 * Uses mas_find() to get the first VMA when the iterator starts.
876 * Calling mas_next() could skip the first entry.
878 return mas_find(&vmi->mas, ULONG_MAX);
882 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
884 return mas_next_range(&vmi->mas, ULONG_MAX);
888 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
890 return mas_prev(&vmi->mas, 0);
894 struct vm_area_struct *vma_iter_prev_range(struct vma_iterator *vmi)
896 return mas_prev_range(&vmi->mas, 0);
899 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
901 return vmi->mas.index;
904 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
906 return vmi->mas.last + 1;
908 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
911 return mas_expected_entries(&vmi->mas, count);
914 /* Free any unused preallocations */
915 static inline void vma_iter_free(struct vma_iterator *vmi)
917 mas_destroy(&vmi->mas);
920 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
921 struct vm_area_struct *vma)
923 vmi->mas.index = vma->vm_start;
924 vmi->mas.last = vma->vm_end - 1;
925 mas_store(&vmi->mas, vma);
926 if (unlikely(mas_is_err(&vmi->mas)))
932 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
934 mas_pause(&vmi->mas);
937 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
939 mas_set(&vmi->mas, addr);
942 #define for_each_vma(__vmi, __vma) \
943 while (((__vma) = vma_next(&(__vmi))) != NULL)
945 /* The MM code likes to work with exclusive end addresses */
946 #define for_each_vma_range(__vmi, __vma, __end) \
947 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
951 * The vma_is_shmem is not inline because it is used only by slow
952 * paths in userfault.
954 bool vma_is_shmem(struct vm_area_struct *vma);
955 bool vma_is_anon_shmem(struct vm_area_struct *vma);
957 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
958 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
961 int vma_is_stack_for_current(struct vm_area_struct *vma);
963 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
964 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
970 * compound_order() can be called without holding a reference, which means
971 * that niceties like page_folio() don't work. These callers should be
972 * prepared to handle wild return values. For example, PG_head may be
973 * set before _folio_order is initialised, or this may be a tail page.
974 * See compaction.c for some good examples.
976 static inline unsigned int compound_order(struct page *page)
978 struct folio *folio = (struct folio *)page;
980 if (!test_bit(PG_head, &folio->flags))
982 return folio->_folio_order;
986 * folio_order - The allocation order of a folio.
989 * A folio is composed of 2^order pages. See get_order() for the definition
992 * Return: The order of the folio.
994 static inline unsigned int folio_order(struct folio *folio)
996 if (!folio_test_large(folio))
998 return folio->_folio_order;
1001 #include <linux/huge_mm.h>
1004 * Methods to modify the page usage count.
1006 * What counts for a page usage:
1007 * - cache mapping (page->mapping)
1008 * - private data (page->private)
1009 * - page mapped in a task's page tables, each mapping
1010 * is counted separately
1012 * Also, many kernel routines increase the page count before a critical
1013 * routine so they can be sure the page doesn't go away from under them.
1017 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1019 static inline int put_page_testzero(struct page *page)
1021 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1022 return page_ref_dec_and_test(page);
1025 static inline int folio_put_testzero(struct folio *folio)
1027 return put_page_testzero(&folio->page);
1031 * Try to grab a ref unless the page has a refcount of zero, return false if
1033 * This can be called when MMU is off so it must not access
1034 * any of the virtual mappings.
1036 static inline bool get_page_unless_zero(struct page *page)
1038 return page_ref_add_unless(page, 1, 0);
1041 static inline struct folio *folio_get_nontail_page(struct page *page)
1043 if (unlikely(!get_page_unless_zero(page)))
1045 return (struct folio *)page;
1048 extern int page_is_ram(unsigned long pfn);
1056 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1057 unsigned long desc);
1059 /* Support for virtually mapped pages */
1060 struct page *vmalloc_to_page(const void *addr);
1061 unsigned long vmalloc_to_pfn(const void *addr);
1064 * Determine if an address is within the vmalloc range
1066 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1067 * is no special casing required.
1070 extern bool is_vmalloc_addr(const void *x);
1071 extern int is_vmalloc_or_module_addr(const void *x);
1073 static inline bool is_vmalloc_addr(const void *x)
1077 static inline int is_vmalloc_or_module_addr(const void *x)
1084 * How many times the entire folio is mapped as a single unit (eg by a
1085 * PMD or PUD entry). This is probably not what you want, except for
1086 * debugging purposes - it does not include PTE-mapped sub-pages; look
1087 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1089 static inline int folio_entire_mapcount(struct folio *folio)
1091 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1092 return atomic_read(&folio->_entire_mapcount) + 1;
1096 * The atomic page->_mapcount, starts from -1: so that transitions
1097 * both from it and to it can be tracked, using atomic_inc_and_test
1098 * and atomic_add_negative(-1).
1100 static inline void page_mapcount_reset(struct page *page)
1102 atomic_set(&(page)->_mapcount, -1);
1106 * page_mapcount() - Number of times this precise page is mapped.
1109 * The number of times this page is mapped. If this page is part of
1110 * a large folio, it includes the number of times this page is mapped
1111 * as part of that folio.
1113 * The result is undefined for pages which cannot be mapped into userspace.
1114 * For example SLAB or special types of pages. See function page_has_type().
1115 * They use this field in struct page differently.
1117 static inline int page_mapcount(struct page *page)
1119 int mapcount = atomic_read(&page->_mapcount) + 1;
1121 if (unlikely(PageCompound(page)))
1122 mapcount += folio_entire_mapcount(page_folio(page));
1127 int folio_total_mapcount(struct folio *folio);
1130 * folio_mapcount() - Calculate the number of mappings of this folio.
1131 * @folio: The folio.
1133 * A large folio tracks both how many times the entire folio is mapped,
1134 * and how many times each individual page in the folio is mapped.
1135 * This function calculates the total number of times the folio is
1138 * Return: The number of times this folio is mapped.
1140 static inline int folio_mapcount(struct folio *folio)
1142 if (likely(!folio_test_large(folio)))
1143 return atomic_read(&folio->_mapcount) + 1;
1144 return folio_total_mapcount(folio);
1147 static inline int total_mapcount(struct page *page)
1149 if (likely(!PageCompound(page)))
1150 return atomic_read(&page->_mapcount) + 1;
1151 return folio_total_mapcount(page_folio(page));
1154 static inline bool folio_large_is_mapped(struct folio *folio)
1157 * Reading _entire_mapcount below could be omitted if hugetlb
1158 * participated in incrementing nr_pages_mapped when compound mapped.
1160 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1161 atomic_read(&folio->_entire_mapcount) >= 0;
1165 * folio_mapped - Is this folio mapped into userspace?
1166 * @folio: The folio.
1168 * Return: True if any page in this folio is referenced by user page tables.
1170 static inline bool folio_mapped(struct folio *folio)
1172 if (likely(!folio_test_large(folio)))
1173 return atomic_read(&folio->_mapcount) >= 0;
1174 return folio_large_is_mapped(folio);
1178 * Return true if this page is mapped into pagetables.
1179 * For compound page it returns true if any sub-page of compound page is mapped,
1180 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1182 static inline bool page_mapped(struct page *page)
1184 if (likely(!PageCompound(page)))
1185 return atomic_read(&page->_mapcount) >= 0;
1186 return folio_large_is_mapped(page_folio(page));
1189 static inline struct page *virt_to_head_page(const void *x)
1191 struct page *page = virt_to_page(x);
1193 return compound_head(page);
1196 static inline struct folio *virt_to_folio(const void *x)
1198 struct page *page = virt_to_page(x);
1200 return page_folio(page);
1203 void __folio_put(struct folio *folio);
1205 void put_pages_list(struct list_head *pages);
1207 void split_page(struct page *page, unsigned int order);
1208 void folio_copy(struct folio *dst, struct folio *src);
1210 unsigned long nr_free_buffer_pages(void);
1213 * Compound pages have a destructor function. Provide a
1214 * prototype for that function and accessor functions.
1215 * These are _only_ valid on the head of a compound page.
1217 typedef void compound_page_dtor(struct page *);
1219 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1220 enum compound_dtor_id {
1223 #ifdef CONFIG_HUGETLB_PAGE
1226 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1227 TRANSHUGE_PAGE_DTOR,
1232 static inline void folio_set_compound_dtor(struct folio *folio,
1233 enum compound_dtor_id compound_dtor)
1235 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1236 folio->_folio_dtor = compound_dtor;
1239 void destroy_large_folio(struct folio *folio);
1241 /* Returns the number of bytes in this potentially compound page. */
1242 static inline unsigned long page_size(struct page *page)
1244 return PAGE_SIZE << compound_order(page);
1247 /* Returns the number of bits needed for the number of bytes in a page */
1248 static inline unsigned int page_shift(struct page *page)
1250 return PAGE_SHIFT + compound_order(page);
1254 * thp_order - Order of a transparent huge page.
1255 * @page: Head page of a transparent huge page.
1257 static inline unsigned int thp_order(struct page *page)
1259 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1260 return compound_order(page);
1264 * thp_size - Size of a transparent huge page.
1265 * @page: Head page of a transparent huge page.
1267 * Return: Number of bytes in this page.
1269 static inline unsigned long thp_size(struct page *page)
1271 return PAGE_SIZE << thp_order(page);
1274 void free_compound_page(struct page *page);
1278 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1279 * servicing faults for write access. In the normal case, do always want
1280 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1281 * that do not have writing enabled, when used by access_process_vm.
1283 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1285 if (likely(vma->vm_flags & VM_WRITE))
1286 pte = pte_mkwrite(pte);
1290 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1291 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1293 vm_fault_t finish_fault(struct vm_fault *vmf);
1294 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1298 * Multiple processes may "see" the same page. E.g. for untouched
1299 * mappings of /dev/null, all processes see the same page full of
1300 * zeroes, and text pages of executables and shared libraries have
1301 * only one copy in memory, at most, normally.
1303 * For the non-reserved pages, page_count(page) denotes a reference count.
1304 * page_count() == 0 means the page is free. page->lru is then used for
1305 * freelist management in the buddy allocator.
1306 * page_count() > 0 means the page has been allocated.
1308 * Pages are allocated by the slab allocator in order to provide memory
1309 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1310 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1311 * unless a particular usage is carefully commented. (the responsibility of
1312 * freeing the kmalloc memory is the caller's, of course).
1314 * A page may be used by anyone else who does a __get_free_page().
1315 * In this case, page_count still tracks the references, and should only
1316 * be used through the normal accessor functions. The top bits of page->flags
1317 * and page->virtual store page management information, but all other fields
1318 * are unused and could be used privately, carefully. The management of this
1319 * page is the responsibility of the one who allocated it, and those who have
1320 * subsequently been given references to it.
1322 * The other pages (we may call them "pagecache pages") are completely
1323 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1324 * The following discussion applies only to them.
1326 * A pagecache page contains an opaque `private' member, which belongs to the
1327 * page's address_space. Usually, this is the address of a circular list of
1328 * the page's disk buffers. PG_private must be set to tell the VM to call
1329 * into the filesystem to release these pages.
1331 * A page may belong to an inode's memory mapping. In this case, page->mapping
1332 * is the pointer to the inode, and page->index is the file offset of the page,
1333 * in units of PAGE_SIZE.
1335 * If pagecache pages are not associated with an inode, they are said to be
1336 * anonymous pages. These may become associated with the swapcache, and in that
1337 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1339 * In either case (swapcache or inode backed), the pagecache itself holds one
1340 * reference to the page. Setting PG_private should also increment the
1341 * refcount. The each user mapping also has a reference to the page.
1343 * The pagecache pages are stored in a per-mapping radix tree, which is
1344 * rooted at mapping->i_pages, and indexed by offset.
1345 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1346 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1348 * All pagecache pages may be subject to I/O:
1349 * - inode pages may need to be read from disk,
1350 * - inode pages which have been modified and are MAP_SHARED may need
1351 * to be written back to the inode on disk,
1352 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1353 * modified may need to be swapped out to swap space and (later) to be read
1357 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1358 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1360 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1361 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1363 if (!static_branch_unlikely(&devmap_managed_key))
1365 if (!is_zone_device_page(page))
1367 return __put_devmap_managed_page_refs(page, refs);
1369 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1370 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1374 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1376 static inline bool put_devmap_managed_page(struct page *page)
1378 return put_devmap_managed_page_refs(page, 1);
1381 /* 127: arbitrary random number, small enough to assemble well */
1382 #define folio_ref_zero_or_close_to_overflow(folio) \
1383 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1386 * folio_get - Increment the reference count on a folio.
1387 * @folio: The folio.
1389 * Context: May be called in any context, as long as you know that
1390 * you have a refcount on the folio. If you do not already have one,
1391 * folio_try_get() may be the right interface for you to use.
1393 static inline void folio_get(struct folio *folio)
1395 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1396 folio_ref_inc(folio);
1399 static inline void get_page(struct page *page)
1401 folio_get(page_folio(page));
1404 static inline __must_check bool try_get_page(struct page *page)
1406 page = compound_head(page);
1407 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1414 * folio_put - Decrement the reference count on a folio.
1415 * @folio: The folio.
1417 * If the folio's reference count reaches zero, the memory will be
1418 * released back to the page allocator and may be used by another
1419 * allocation immediately. Do not access the memory or the struct folio
1420 * after calling folio_put() unless you can be sure that it wasn't the
1423 * Context: May be called in process or interrupt context, but not in NMI
1424 * context. May be called while holding a spinlock.
1426 static inline void folio_put(struct folio *folio)
1428 if (folio_put_testzero(folio))
1433 * folio_put_refs - Reduce the reference count on a folio.
1434 * @folio: The folio.
1435 * @refs: The amount to subtract from the folio's reference count.
1437 * If the folio's reference count reaches zero, the memory will be
1438 * released back to the page allocator and may be used by another
1439 * allocation immediately. Do not access the memory or the struct folio
1440 * after calling folio_put_refs() unless you can be sure that these weren't
1441 * the last references.
1443 * Context: May be called in process or interrupt context, but not in NMI
1444 * context. May be called while holding a spinlock.
1446 static inline void folio_put_refs(struct folio *folio, int refs)
1448 if (folio_ref_sub_and_test(folio, refs))
1453 * union release_pages_arg - an array of pages or folios
1455 * release_pages() releases a simple array of multiple pages, and
1456 * accepts various different forms of said page array: either
1457 * a regular old boring array of pages, an array of folios, or
1458 * an array of encoded page pointers.
1460 * The transparent union syntax for this kind of "any of these
1461 * argument types" is all kinds of ugly, so look away.
1464 struct page **pages;
1465 struct folio **folios;
1466 struct encoded_page **encoded_pages;
1467 } release_pages_arg __attribute__ ((__transparent_union__));
1469 void release_pages(release_pages_arg, int nr);
1472 * folios_put - Decrement the reference count on an array of folios.
1473 * @folios: The folios.
1474 * @nr: How many folios there are.
1476 * Like folio_put(), but for an array of folios. This is more efficient
1477 * than writing the loop yourself as it will optimise the locks which
1478 * need to be taken if the folios are freed.
1480 * Context: May be called in process or interrupt context, but not in NMI
1481 * context. May be called while holding a spinlock.
1483 static inline void folios_put(struct folio **folios, unsigned int nr)
1485 release_pages(folios, nr);
1488 static inline void put_page(struct page *page)
1490 struct folio *folio = page_folio(page);
1493 * For some devmap managed pages we need to catch refcount transition
1496 if (put_devmap_managed_page(&folio->page))
1502 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1503 * the page's refcount so that two separate items are tracked: the original page
1504 * reference count, and also a new count of how many pin_user_pages() calls were
1505 * made against the page. ("gup-pinned" is another term for the latter).
1507 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1508 * distinct from normal pages. As such, the unpin_user_page() call (and its
1509 * variants) must be used in order to release gup-pinned pages.
1513 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1514 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1515 * simpler, due to the fact that adding an even power of two to the page
1516 * refcount has the effect of using only the upper N bits, for the code that
1517 * counts up using the bias value. This means that the lower bits are left for
1518 * the exclusive use of the original code that increments and decrements by one
1519 * (or at least, by much smaller values than the bias value).
1521 * Of course, once the lower bits overflow into the upper bits (and this is
1522 * OK, because subtraction recovers the original values), then visual inspection
1523 * no longer suffices to directly view the separate counts. However, for normal
1524 * applications that don't have huge page reference counts, this won't be an
1527 * Locking: the lockless algorithm described in folio_try_get_rcu()
1528 * provides safe operation for get_user_pages(), page_mkclean() and
1529 * other calls that race to set up page table entries.
1531 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1533 void unpin_user_page(struct page *page);
1534 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1536 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1538 void unpin_user_pages(struct page **pages, unsigned long npages);
1540 static inline bool is_cow_mapping(vm_flags_t flags)
1542 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1546 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1549 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1550 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1551 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1552 * underlying memory if ptrace is active, so this is only possible if
1553 * ptrace does not apply. Note that there is no mprotect() to upgrade
1554 * write permissions later.
1556 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1560 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1561 #define SECTION_IN_PAGE_FLAGS
1565 * The identification function is mainly used by the buddy allocator for
1566 * determining if two pages could be buddies. We are not really identifying
1567 * the zone since we could be using the section number id if we do not have
1568 * node id available in page flags.
1569 * We only guarantee that it will return the same value for two combinable
1572 static inline int page_zone_id(struct page *page)
1574 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1577 #ifdef NODE_NOT_IN_PAGE_FLAGS
1578 extern int page_to_nid(const struct page *page);
1580 static inline int page_to_nid(const struct page *page)
1582 struct page *p = (struct page *)page;
1584 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1588 static inline int folio_nid(const struct folio *folio)
1590 return page_to_nid(&folio->page);
1593 #ifdef CONFIG_NUMA_BALANCING
1594 /* page access time bits needs to hold at least 4 seconds */
1595 #define PAGE_ACCESS_TIME_MIN_BITS 12
1596 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1597 #define PAGE_ACCESS_TIME_BUCKETS \
1598 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1600 #define PAGE_ACCESS_TIME_BUCKETS 0
1603 #define PAGE_ACCESS_TIME_MASK \
1604 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1606 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1608 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1611 static inline int cpupid_to_pid(int cpupid)
1613 return cpupid & LAST__PID_MASK;
1616 static inline int cpupid_to_cpu(int cpupid)
1618 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1621 static inline int cpupid_to_nid(int cpupid)
1623 return cpu_to_node(cpupid_to_cpu(cpupid));
1626 static inline bool cpupid_pid_unset(int cpupid)
1628 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1631 static inline bool cpupid_cpu_unset(int cpupid)
1633 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1636 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1638 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1641 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1642 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1643 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1645 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1648 static inline int page_cpupid_last(struct page *page)
1650 return page->_last_cpupid;
1652 static inline void page_cpupid_reset_last(struct page *page)
1654 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1657 static inline int page_cpupid_last(struct page *page)
1659 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1662 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1664 static inline void page_cpupid_reset_last(struct page *page)
1666 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1668 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1670 static inline int xchg_page_access_time(struct page *page, int time)
1674 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1675 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1678 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1680 unsigned int pid_bit;
1682 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1683 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) {
1684 __set_bit(pid_bit, &vma->numab_state->access_pids[1]);
1687 #else /* !CONFIG_NUMA_BALANCING */
1688 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1690 return page_to_nid(page); /* XXX */
1693 static inline int xchg_page_access_time(struct page *page, int time)
1698 static inline int page_cpupid_last(struct page *page)
1700 return page_to_nid(page); /* XXX */
1703 static inline int cpupid_to_nid(int cpupid)
1708 static inline int cpupid_to_pid(int cpupid)
1713 static inline int cpupid_to_cpu(int cpupid)
1718 static inline int cpu_pid_to_cpupid(int nid, int pid)
1723 static inline bool cpupid_pid_unset(int cpupid)
1728 static inline void page_cpupid_reset_last(struct page *page)
1732 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1737 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1740 #endif /* CONFIG_NUMA_BALANCING */
1742 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1745 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1746 * setting tags for all pages to native kernel tag value 0xff, as the default
1747 * value 0x00 maps to 0xff.
1750 static inline u8 page_kasan_tag(const struct page *page)
1754 if (kasan_enabled()) {
1755 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1762 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1764 unsigned long old_flags, flags;
1766 if (!kasan_enabled())
1770 old_flags = READ_ONCE(page->flags);
1773 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1774 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1775 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1778 static inline void page_kasan_tag_reset(struct page *page)
1780 if (kasan_enabled())
1781 page_kasan_tag_set(page, 0xff);
1784 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1786 static inline u8 page_kasan_tag(const struct page *page)
1791 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1792 static inline void page_kasan_tag_reset(struct page *page) { }
1794 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1796 static inline struct zone *page_zone(const struct page *page)
1798 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1801 static inline pg_data_t *page_pgdat(const struct page *page)
1803 return NODE_DATA(page_to_nid(page));
1806 static inline struct zone *folio_zone(const struct folio *folio)
1808 return page_zone(&folio->page);
1811 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1813 return page_pgdat(&folio->page);
1816 #ifdef SECTION_IN_PAGE_FLAGS
1817 static inline void set_page_section(struct page *page, unsigned long section)
1819 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1820 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1823 static inline unsigned long page_to_section(const struct page *page)
1825 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1830 * folio_pfn - Return the Page Frame Number of a folio.
1831 * @folio: The folio.
1833 * A folio may contain multiple pages. The pages have consecutive
1834 * Page Frame Numbers.
1836 * Return: The Page Frame Number of the first page in the folio.
1838 static inline unsigned long folio_pfn(struct folio *folio)
1840 return page_to_pfn(&folio->page);
1843 static inline struct folio *pfn_folio(unsigned long pfn)
1845 return page_folio(pfn_to_page(pfn));
1849 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1850 * @folio: The folio.
1852 * This function checks if a folio has been pinned via a call to
1853 * a function in the pin_user_pages() family.
1855 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1856 * because it means "definitely not pinned for DMA", but true means "probably
1857 * pinned for DMA, but possibly a false positive due to having at least
1858 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1860 * False positives are OK, because: a) it's unlikely for a folio to
1861 * get that many refcounts, and b) all the callers of this routine are
1862 * expected to be able to deal gracefully with a false positive.
1864 * For large folios, the result will be exactly correct. That's because
1865 * we have more tracking data available: the _pincount field is used
1866 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1868 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1870 * Return: True, if it is likely that the page has been "dma-pinned".
1871 * False, if the page is definitely not dma-pinned.
1873 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1875 if (folio_test_large(folio))
1876 return atomic_read(&folio->_pincount) > 0;
1879 * folio_ref_count() is signed. If that refcount overflows, then
1880 * folio_ref_count() returns a negative value, and callers will avoid
1881 * further incrementing the refcount.
1883 * Here, for that overflow case, use the sign bit to count a little
1884 * bit higher via unsigned math, and thus still get an accurate result.
1886 return ((unsigned int)folio_ref_count(folio)) >=
1887 GUP_PIN_COUNTING_BIAS;
1890 static inline bool page_maybe_dma_pinned(struct page *page)
1892 return folio_maybe_dma_pinned(page_folio(page));
1896 * This should most likely only be called during fork() to see whether we
1897 * should break the cow immediately for an anon page on the src mm.
1899 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1901 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1904 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1906 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1909 return page_maybe_dma_pinned(page);
1913 * is_zero_page - Query if a page is a zero page
1914 * @page: The page to query
1916 * This returns true if @page is one of the permanent zero pages.
1918 static inline bool is_zero_page(const struct page *page)
1920 return is_zero_pfn(page_to_pfn(page));
1924 * is_zero_folio - Query if a folio is a zero page
1925 * @folio: The folio to query
1927 * This returns true if @folio is one of the permanent zero pages.
1929 static inline bool is_zero_folio(const struct folio *folio)
1931 return is_zero_page(&folio->page);
1934 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1935 #ifdef CONFIG_MIGRATION
1936 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1939 int mt = folio_migratetype(folio);
1941 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1944 /* The zero page can be "pinned" but gets special handling. */
1945 if (is_zero_folio(folio))
1948 /* Coherent device memory must always allow eviction. */
1949 if (folio_is_device_coherent(folio))
1952 /* Otherwise, non-movable zone folios can be pinned. */
1953 return !folio_is_zone_movable(folio);
1957 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1963 static inline void set_page_zone(struct page *page, enum zone_type zone)
1965 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1966 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1969 static inline void set_page_node(struct page *page, unsigned long node)
1971 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1972 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1975 static inline void set_page_links(struct page *page, enum zone_type zone,
1976 unsigned long node, unsigned long pfn)
1978 set_page_zone(page, zone);
1979 set_page_node(page, node);
1980 #ifdef SECTION_IN_PAGE_FLAGS
1981 set_page_section(page, pfn_to_section_nr(pfn));
1986 * folio_nr_pages - The number of pages in the folio.
1987 * @folio: The folio.
1989 * Return: A positive power of two.
1991 static inline long folio_nr_pages(struct folio *folio)
1993 if (!folio_test_large(folio))
1996 return folio->_folio_nr_pages;
1998 return 1L << folio->_folio_order;
2003 * compound_nr() returns the number of pages in this potentially compound
2004 * page. compound_nr() can be called on a tail page, and is defined to
2005 * return 1 in that case.
2007 static inline unsigned long compound_nr(struct page *page)
2009 struct folio *folio = (struct folio *)page;
2011 if (!test_bit(PG_head, &folio->flags))
2014 return folio->_folio_nr_pages;
2016 return 1L << folio->_folio_order;
2021 * thp_nr_pages - The number of regular pages in this huge page.
2022 * @page: The head page of a huge page.
2024 static inline int thp_nr_pages(struct page *page)
2026 return folio_nr_pages((struct folio *)page);
2030 * folio_next - Move to the next physical folio.
2031 * @folio: The folio we're currently operating on.
2033 * If you have physically contiguous memory which may span more than
2034 * one folio (eg a &struct bio_vec), use this function to move from one
2035 * folio to the next. Do not use it if the memory is only virtually
2036 * contiguous as the folios are almost certainly not adjacent to each
2037 * other. This is the folio equivalent to writing ``page++``.
2039 * Context: We assume that the folios are refcounted and/or locked at a
2040 * higher level and do not adjust the reference counts.
2041 * Return: The next struct folio.
2043 static inline struct folio *folio_next(struct folio *folio)
2045 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2049 * folio_shift - The size of the memory described by this folio.
2050 * @folio: The folio.
2052 * A folio represents a number of bytes which is a power-of-two in size.
2053 * This function tells you which power-of-two the folio is. See also
2054 * folio_size() and folio_order().
2056 * Context: The caller should have a reference on the folio to prevent
2057 * it from being split. It is not necessary for the folio to be locked.
2058 * Return: The base-2 logarithm of the size of this folio.
2060 static inline unsigned int folio_shift(struct folio *folio)
2062 return PAGE_SHIFT + folio_order(folio);
2066 * folio_size - The number of bytes in a folio.
2067 * @folio: The folio.
2069 * Context: The caller should have a reference on the folio to prevent
2070 * it from being split. It is not necessary for the folio to be locked.
2071 * Return: The number of bytes in this folio.
2073 static inline size_t folio_size(struct folio *folio)
2075 return PAGE_SIZE << folio_order(folio);
2079 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2080 * @folio: The folio.
2082 * folio_estimated_sharers() aims to serve as a function to efficiently
2083 * estimate the number of processes sharing a folio. This is done by
2084 * looking at the precise mapcount of the first subpage in the folio, and
2085 * assuming the other subpages are the same. This may not be true for large
2086 * folios. If you want exact mapcounts for exact calculations, look at
2087 * page_mapcount() or folio_total_mapcount().
2089 * Return: The estimated number of processes sharing a folio.
2091 static inline int folio_estimated_sharers(struct folio *folio)
2093 return page_mapcount(folio_page(folio, 0));
2096 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2097 static inline int arch_make_page_accessible(struct page *page)
2103 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2104 static inline int arch_make_folio_accessible(struct folio *folio)
2107 long i, nr = folio_nr_pages(folio);
2109 for (i = 0; i < nr; i++) {
2110 ret = arch_make_page_accessible(folio_page(folio, i));
2120 * Some inline functions in vmstat.h depend on page_zone()
2122 #include <linux/vmstat.h>
2124 static __always_inline void *lowmem_page_address(const struct page *page)
2126 return page_to_virt(page);
2129 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2130 #define HASHED_PAGE_VIRTUAL
2133 #if defined(WANT_PAGE_VIRTUAL)
2134 static inline void *page_address(const struct page *page)
2136 return page->virtual;
2138 static inline void set_page_address(struct page *page, void *address)
2140 page->virtual = address;
2142 #define page_address_init() do { } while(0)
2145 #if defined(HASHED_PAGE_VIRTUAL)
2146 void *page_address(const struct page *page);
2147 void set_page_address(struct page *page, void *virtual);
2148 void page_address_init(void);
2151 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2152 #define page_address(page) lowmem_page_address(page)
2153 #define set_page_address(page, address) do { } while(0)
2154 #define page_address_init() do { } while(0)
2157 static inline void *folio_address(const struct folio *folio)
2159 return page_address(&folio->page);
2162 extern pgoff_t __page_file_index(struct page *page);
2165 * Return the pagecache index of the passed page. Regular pagecache pages
2166 * use ->index whereas swapcache pages use swp_offset(->private)
2168 static inline pgoff_t page_index(struct page *page)
2170 if (unlikely(PageSwapCache(page)))
2171 return __page_file_index(page);
2176 * Return true only if the page has been allocated with
2177 * ALLOC_NO_WATERMARKS and the low watermark was not
2178 * met implying that the system is under some pressure.
2180 static inline bool page_is_pfmemalloc(const struct page *page)
2183 * lru.next has bit 1 set if the page is allocated from the
2184 * pfmemalloc reserves. Callers may simply overwrite it if
2185 * they do not need to preserve that information.
2187 return (uintptr_t)page->lru.next & BIT(1);
2191 * Return true only if the folio has been allocated with
2192 * ALLOC_NO_WATERMARKS and the low watermark was not
2193 * met implying that the system is under some pressure.
2195 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2198 * lru.next has bit 1 set if the page is allocated from the
2199 * pfmemalloc reserves. Callers may simply overwrite it if
2200 * they do not need to preserve that information.
2202 return (uintptr_t)folio->lru.next & BIT(1);
2206 * Only to be called by the page allocator on a freshly allocated
2209 static inline void set_page_pfmemalloc(struct page *page)
2211 page->lru.next = (void *)BIT(1);
2214 static inline void clear_page_pfmemalloc(struct page *page)
2216 page->lru.next = NULL;
2220 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2222 extern void pagefault_out_of_memory(void);
2224 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2225 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2226 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2229 * Parameter block passed down to zap_pte_range in exceptional cases.
2231 struct zap_details {
2232 struct folio *single_folio; /* Locked folio to be unmapped */
2233 bool even_cows; /* Zap COWed private pages too? */
2234 zap_flags_t zap_flags; /* Extra flags for zapping */
2238 * Whether to drop the pte markers, for example, the uffd-wp information for
2239 * file-backed memory. This should only be specified when we will completely
2240 * drop the page in the mm, either by truncation or unmapping of the vma. By
2241 * default, the flag is not set.
2243 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2244 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2245 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2247 #ifdef CONFIG_SCHED_MM_CID
2248 void sched_mm_cid_before_execve(struct task_struct *t);
2249 void sched_mm_cid_after_execve(struct task_struct *t);
2250 void sched_mm_cid_fork(struct task_struct *t);
2251 void sched_mm_cid_exit_signals(struct task_struct *t);
2252 static inline int task_mm_cid(struct task_struct *t)
2257 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2258 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2259 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2260 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2261 static inline int task_mm_cid(struct task_struct *t)
2264 * Use the processor id as a fall-back when the mm cid feature is
2265 * disabled. This provides functional per-cpu data structure accesses
2266 * in user-space, althrough it won't provide the memory usage benefits.
2268 return raw_smp_processor_id();
2273 extern bool can_do_mlock(void);
2275 static inline bool can_do_mlock(void) { return false; }
2277 extern int user_shm_lock(size_t, struct ucounts *);
2278 extern void user_shm_unlock(size_t, struct ucounts *);
2280 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2282 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2284 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2287 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2288 unsigned long size);
2289 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2290 unsigned long size, struct zap_details *details);
2291 static inline void zap_vma_pages(struct vm_area_struct *vma)
2293 zap_page_range_single(vma, vma->vm_start,
2294 vma->vm_end - vma->vm_start, NULL);
2296 void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
2297 struct vm_area_struct *start_vma, unsigned long start,
2298 unsigned long end, unsigned long tree_end, bool mm_wr_locked);
2300 struct mmu_notifier_range;
2302 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2303 unsigned long end, unsigned long floor, unsigned long ceiling);
2305 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2306 int follow_pte(struct mm_struct *mm, unsigned long address,
2307 pte_t **ptepp, spinlock_t **ptlp);
2308 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2309 unsigned long *pfn);
2310 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2311 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2312 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2313 void *buf, int len, int write);
2315 extern void truncate_pagecache(struct inode *inode, loff_t new);
2316 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2317 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2318 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2319 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2321 struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
2322 unsigned long address, struct pt_regs *regs);
2325 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2326 unsigned long address, unsigned int flags,
2327 struct pt_regs *regs);
2328 extern int fixup_user_fault(struct mm_struct *mm,
2329 unsigned long address, unsigned int fault_flags,
2331 void unmap_mapping_pages(struct address_space *mapping,
2332 pgoff_t start, pgoff_t nr, bool even_cows);
2333 void unmap_mapping_range(struct address_space *mapping,
2334 loff_t const holebegin, loff_t const holelen, int even_cows);
2336 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2337 unsigned long address, unsigned int flags,
2338 struct pt_regs *regs)
2340 /* should never happen if there's no MMU */
2342 return VM_FAULT_SIGBUS;
2344 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2345 unsigned int fault_flags, bool *unlocked)
2347 /* should never happen if there's no MMU */
2351 static inline void unmap_mapping_pages(struct address_space *mapping,
2352 pgoff_t start, pgoff_t nr, bool even_cows) { }
2353 static inline void unmap_mapping_range(struct address_space *mapping,
2354 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2357 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2358 loff_t const holebegin, loff_t const holelen)
2360 unmap_mapping_range(mapping, holebegin, holelen, 0);
2363 static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
2364 unsigned long addr);
2366 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2367 void *buf, int len, unsigned int gup_flags);
2368 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2369 void *buf, int len, unsigned int gup_flags);
2370 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2371 void *buf, int len, unsigned int gup_flags);
2373 long get_user_pages_remote(struct mm_struct *mm,
2374 unsigned long start, unsigned long nr_pages,
2375 unsigned int gup_flags, struct page **pages,
2377 long pin_user_pages_remote(struct mm_struct *mm,
2378 unsigned long start, unsigned long nr_pages,
2379 unsigned int gup_flags, struct page **pages,
2382 static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
2385 struct vm_area_struct **vmap)
2388 struct vm_area_struct *vma;
2389 int got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
2392 return ERR_PTR(got);
2396 vma = vma_lookup(mm, addr);
2397 if (WARN_ON_ONCE(!vma)) {
2399 return ERR_PTR(-EINVAL);
2406 long get_user_pages(unsigned long start, unsigned long nr_pages,
2407 unsigned int gup_flags, struct page **pages);
2408 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2409 unsigned int gup_flags, struct page **pages);
2410 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2411 struct page **pages, unsigned int gup_flags);
2412 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2413 struct page **pages, unsigned int gup_flags);
2415 int get_user_pages_fast(unsigned long start, int nr_pages,
2416 unsigned int gup_flags, struct page **pages);
2417 int pin_user_pages_fast(unsigned long start, int nr_pages,
2418 unsigned int gup_flags, struct page **pages);
2419 void folio_add_pin(struct folio *folio);
2421 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2422 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2423 struct task_struct *task, bool bypass_rlim);
2426 struct page *get_dump_page(unsigned long addr);
2428 bool folio_mark_dirty(struct folio *folio);
2429 bool set_page_dirty(struct page *page);
2430 int set_page_dirty_lock(struct page *page);
2432 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2434 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2435 unsigned long old_addr, struct vm_area_struct *new_vma,
2436 unsigned long new_addr, unsigned long len,
2437 bool need_rmap_locks);
2440 * Flags used by change_protection(). For now we make it a bitmap so
2441 * that we can pass in multiple flags just like parameters. However
2442 * for now all the callers are only use one of the flags at the same
2446 * Whether we should manually check if we can map individual PTEs writable,
2447 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2448 * PTEs automatically in a writable mapping.
2450 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2451 /* Whether this protection change is for NUMA hints */
2452 #define MM_CP_PROT_NUMA (1UL << 1)
2453 /* Whether this change is for write protecting */
2454 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2455 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2456 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2457 MM_CP_UFFD_WP_RESOLVE)
2459 bool vma_needs_dirty_tracking(struct vm_area_struct *vma);
2460 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2461 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2464 * We want to check manually if we can change individual PTEs writable
2465 * if we can't do that automatically for all PTEs in a mapping. For
2466 * private mappings, that's always the case when we have write
2467 * permissions as we properly have to handle COW.
2469 if (vma->vm_flags & VM_SHARED)
2470 return vma_wants_writenotify(vma, vma->vm_page_prot);
2471 return !!(vma->vm_flags & VM_WRITE);
2474 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2476 extern long change_protection(struct mmu_gather *tlb,
2477 struct vm_area_struct *vma, unsigned long start,
2478 unsigned long end, unsigned long cp_flags);
2479 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2480 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2481 unsigned long start, unsigned long end, unsigned long newflags);
2484 * doesn't attempt to fault and will return short.
2486 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2487 unsigned int gup_flags, struct page **pages);
2489 static inline bool get_user_page_fast_only(unsigned long addr,
2490 unsigned int gup_flags, struct page **pagep)
2492 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2495 * per-process(per-mm_struct) statistics.
2497 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2499 return percpu_counter_read_positive(&mm->rss_stat[member]);
2502 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2504 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2506 percpu_counter_add(&mm->rss_stat[member], value);
2508 mm_trace_rss_stat(mm, member);
2511 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2513 percpu_counter_inc(&mm->rss_stat[member]);
2515 mm_trace_rss_stat(mm, member);
2518 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2520 percpu_counter_dec(&mm->rss_stat[member]);
2522 mm_trace_rss_stat(mm, member);
2525 /* Optimized variant when page is already known not to be PageAnon */
2526 static inline int mm_counter_file(struct page *page)
2528 if (PageSwapBacked(page))
2529 return MM_SHMEMPAGES;
2530 return MM_FILEPAGES;
2533 static inline int mm_counter(struct page *page)
2536 return MM_ANONPAGES;
2537 return mm_counter_file(page);
2540 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2542 return get_mm_counter(mm, MM_FILEPAGES) +
2543 get_mm_counter(mm, MM_ANONPAGES) +
2544 get_mm_counter(mm, MM_SHMEMPAGES);
2547 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2549 return max(mm->hiwater_rss, get_mm_rss(mm));
2552 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2554 return max(mm->hiwater_vm, mm->total_vm);
2557 static inline void update_hiwater_rss(struct mm_struct *mm)
2559 unsigned long _rss = get_mm_rss(mm);
2561 if ((mm)->hiwater_rss < _rss)
2562 (mm)->hiwater_rss = _rss;
2565 static inline void update_hiwater_vm(struct mm_struct *mm)
2567 if (mm->hiwater_vm < mm->total_vm)
2568 mm->hiwater_vm = mm->total_vm;
2571 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2573 mm->hiwater_rss = get_mm_rss(mm);
2576 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2577 struct mm_struct *mm)
2579 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2581 if (*maxrss < hiwater_rss)
2582 *maxrss = hiwater_rss;
2585 #if defined(SPLIT_RSS_COUNTING)
2586 void sync_mm_rss(struct mm_struct *mm);
2588 static inline void sync_mm_rss(struct mm_struct *mm)
2593 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2594 static inline int pte_special(pte_t pte)
2599 static inline pte_t pte_mkspecial(pte_t pte)
2605 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2606 static inline int pte_devmap(pte_t pte)
2612 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2614 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2618 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2622 #ifdef __PAGETABLE_P4D_FOLDED
2623 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2624 unsigned long address)
2629 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2632 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2633 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2634 unsigned long address)
2638 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2639 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2642 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2644 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2646 if (mm_pud_folded(mm))
2648 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2651 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2653 if (mm_pud_folded(mm))
2655 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2659 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2660 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2661 unsigned long address)
2666 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2667 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2670 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2672 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2674 if (mm_pmd_folded(mm))
2676 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2679 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2681 if (mm_pmd_folded(mm))
2683 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2688 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2690 atomic_long_set(&mm->pgtables_bytes, 0);
2693 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2695 return atomic_long_read(&mm->pgtables_bytes);
2698 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2700 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2703 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2705 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2709 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2710 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2715 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2716 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2719 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2720 int __pte_alloc_kernel(pmd_t *pmd);
2722 #if defined(CONFIG_MMU)
2724 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2725 unsigned long address)
2727 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2728 NULL : p4d_offset(pgd, address);
2731 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2732 unsigned long address)
2734 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2735 NULL : pud_offset(p4d, address);
2738 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2740 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2741 NULL: pmd_offset(pud, address);
2743 #endif /* CONFIG_MMU */
2745 #if USE_SPLIT_PTE_PTLOCKS
2746 #if ALLOC_SPLIT_PTLOCKS
2747 void __init ptlock_cache_init(void);
2748 extern bool ptlock_alloc(struct page *page);
2749 extern void ptlock_free(struct page *page);
2751 static inline spinlock_t *ptlock_ptr(struct page *page)
2755 #else /* ALLOC_SPLIT_PTLOCKS */
2756 static inline void ptlock_cache_init(void)
2760 static inline bool ptlock_alloc(struct page *page)
2765 static inline void ptlock_free(struct page *page)
2769 static inline spinlock_t *ptlock_ptr(struct page *page)
2773 #endif /* ALLOC_SPLIT_PTLOCKS */
2775 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2777 return ptlock_ptr(pmd_page(*pmd));
2780 static inline bool ptlock_init(struct page *page)
2783 * prep_new_page() initialize page->private (and therefore page->ptl)
2784 * with 0. Make sure nobody took it in use in between.
2786 * It can happen if arch try to use slab for page table allocation:
2787 * slab code uses page->slab_cache, which share storage with page->ptl.
2789 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2790 if (!ptlock_alloc(page))
2792 spin_lock_init(ptlock_ptr(page));
2796 #else /* !USE_SPLIT_PTE_PTLOCKS */
2798 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2800 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2802 return &mm->page_table_lock;
2804 static inline void ptlock_cache_init(void) {}
2805 static inline bool ptlock_init(struct page *page) { return true; }
2806 static inline void ptlock_free(struct page *page) {}
2807 #endif /* USE_SPLIT_PTE_PTLOCKS */
2809 static inline bool pgtable_pte_page_ctor(struct page *page)
2811 if (!ptlock_init(page))
2813 __SetPageTable(page);
2814 inc_lruvec_page_state(page, NR_PAGETABLE);
2818 static inline void pgtable_pte_page_dtor(struct page *page)
2821 __ClearPageTable(page);
2822 dec_lruvec_page_state(page, NR_PAGETABLE);
2825 pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
2826 static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
2828 return __pte_offset_map(pmd, addr, NULL);
2831 pte_t *__pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2832 unsigned long addr, spinlock_t **ptlp);
2833 static inline pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2834 unsigned long addr, spinlock_t **ptlp)
2838 __cond_lock(*ptlp, pte = __pte_offset_map_lock(mm, pmd, addr, ptlp));
2842 pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
2843 unsigned long addr, spinlock_t **ptlp);
2845 #define pte_unmap_unlock(pte, ptl) do { \
2850 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2852 #define pte_alloc_map(mm, pmd, address) \
2853 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2855 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2856 (pte_alloc(mm, pmd) ? \
2857 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2859 #define pte_alloc_kernel(pmd, address) \
2860 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2861 NULL: pte_offset_kernel(pmd, address))
2863 #if USE_SPLIT_PMD_PTLOCKS
2865 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2867 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2868 return virt_to_page((void *)((unsigned long) pmd & mask));
2871 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2873 return ptlock_ptr(pmd_pgtable_page(pmd));
2876 static inline bool pmd_ptlock_init(struct page *page)
2878 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2879 page->pmd_huge_pte = NULL;
2881 return ptlock_init(page);
2884 static inline void pmd_ptlock_free(struct page *page)
2886 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2887 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2892 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2896 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2898 return &mm->page_table_lock;
2901 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2902 static inline void pmd_ptlock_free(struct page *page) {}
2904 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2908 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2910 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2915 static inline bool pgtable_pmd_page_ctor(struct page *page)
2917 if (!pmd_ptlock_init(page))
2919 __SetPageTable(page);
2920 inc_lruvec_page_state(page, NR_PAGETABLE);
2924 static inline void pgtable_pmd_page_dtor(struct page *page)
2926 pmd_ptlock_free(page);
2927 __ClearPageTable(page);
2928 dec_lruvec_page_state(page, NR_PAGETABLE);
2932 * No scalability reason to split PUD locks yet, but follow the same pattern
2933 * as the PMD locks to make it easier if we decide to. The VM should not be
2934 * considered ready to switch to split PUD locks yet; there may be places
2935 * which need to be converted from page_table_lock.
2937 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2939 return &mm->page_table_lock;
2942 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2944 spinlock_t *ptl = pud_lockptr(mm, pud);
2950 extern void __init pagecache_init(void);
2951 extern void free_initmem(void);
2954 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2955 * into the buddy system. The freed pages will be poisoned with pattern
2956 * "poison" if it's within range [0, UCHAR_MAX].
2957 * Return pages freed into the buddy system.
2959 extern unsigned long free_reserved_area(void *start, void *end,
2960 int poison, const char *s);
2962 extern void adjust_managed_page_count(struct page *page, long count);
2964 extern void reserve_bootmem_region(phys_addr_t start,
2965 phys_addr_t end, int nid);
2967 /* Free the reserved page into the buddy system, so it gets managed. */
2968 static inline void free_reserved_page(struct page *page)
2970 ClearPageReserved(page);
2971 init_page_count(page);
2973 adjust_managed_page_count(page, 1);
2975 #define free_highmem_page(page) free_reserved_page(page)
2977 static inline void mark_page_reserved(struct page *page)
2979 SetPageReserved(page);
2980 adjust_managed_page_count(page, -1);
2984 * Default method to free all the __init memory into the buddy system.
2985 * The freed pages will be poisoned with pattern "poison" if it's within
2986 * range [0, UCHAR_MAX].
2987 * Return pages freed into the buddy system.
2989 static inline unsigned long free_initmem_default(int poison)
2991 extern char __init_begin[], __init_end[];
2993 return free_reserved_area(&__init_begin, &__init_end,
2994 poison, "unused kernel image (initmem)");
2997 static inline unsigned long get_num_physpages(void)
3000 unsigned long phys_pages = 0;
3002 for_each_online_node(nid)
3003 phys_pages += node_present_pages(nid);
3009 * Using memblock node mappings, an architecture may initialise its
3010 * zones, allocate the backing mem_map and account for memory holes in an
3011 * architecture independent manner.
3013 * An architecture is expected to register range of page frames backed by
3014 * physical memory with memblock_add[_node]() before calling
3015 * free_area_init() passing in the PFN each zone ends at. At a basic
3016 * usage, an architecture is expected to do something like
3018 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3020 * for_each_valid_physical_page_range()
3021 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3022 * free_area_init(max_zone_pfns);
3024 void free_area_init(unsigned long *max_zone_pfn);
3025 unsigned long node_map_pfn_alignment(void);
3026 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
3027 unsigned long end_pfn);
3028 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3029 unsigned long end_pfn);
3030 extern void get_pfn_range_for_nid(unsigned int nid,
3031 unsigned long *start_pfn, unsigned long *end_pfn);
3034 static inline int early_pfn_to_nid(unsigned long pfn)
3039 /* please see mm/page_alloc.c */
3040 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3043 extern void set_dma_reserve(unsigned long new_dma_reserve);
3044 extern void mem_init(void);
3045 extern void __init mmap_init(void);
3047 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
3048 static inline void show_mem(void)
3050 __show_mem(0, NULL, MAX_NR_ZONES - 1);
3052 extern long si_mem_available(void);
3053 extern void si_meminfo(struct sysinfo * val);
3054 extern void si_meminfo_node(struct sysinfo *val, int nid);
3055 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3056 extern unsigned long arch_reserved_kernel_pages(void);
3059 extern __printf(3, 4)
3060 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3062 extern void setup_per_cpu_pageset(void);
3065 extern atomic_long_t mmap_pages_allocated;
3066 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3068 /* interval_tree.c */
3069 void vma_interval_tree_insert(struct vm_area_struct *node,
3070 struct rb_root_cached *root);
3071 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3072 struct vm_area_struct *prev,
3073 struct rb_root_cached *root);
3074 void vma_interval_tree_remove(struct vm_area_struct *node,
3075 struct rb_root_cached *root);
3076 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3077 unsigned long start, unsigned long last);
3078 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3079 unsigned long start, unsigned long last);
3081 #define vma_interval_tree_foreach(vma, root, start, last) \
3082 for (vma = vma_interval_tree_iter_first(root, start, last); \
3083 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3085 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3086 struct rb_root_cached *root);
3087 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3088 struct rb_root_cached *root);
3089 struct anon_vma_chain *
3090 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3091 unsigned long start, unsigned long last);
3092 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3093 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3094 #ifdef CONFIG_DEBUG_VM_RB
3095 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3098 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3099 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3100 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3103 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3104 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3105 unsigned long start, unsigned long end, pgoff_t pgoff,
3106 struct vm_area_struct *next);
3107 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3108 unsigned long start, unsigned long end, pgoff_t pgoff);
3109 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3110 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3111 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3112 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3113 struct anon_vma_name *);
3114 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3115 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3116 unsigned long addr, int new_below);
3117 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3118 unsigned long addr, int new_below);
3119 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3120 extern void unlink_file_vma(struct vm_area_struct *);
3121 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3122 unsigned long addr, unsigned long len, pgoff_t pgoff,
3123 bool *need_rmap_locks);
3124 extern void exit_mmap(struct mm_struct *);
3126 static inline int check_data_rlimit(unsigned long rlim,
3128 unsigned long start,
3129 unsigned long end_data,
3130 unsigned long start_data)
3132 if (rlim < RLIM_INFINITY) {
3133 if (((new - start) + (end_data - start_data)) > rlim)
3140 extern int mm_take_all_locks(struct mm_struct *mm);
3141 extern void mm_drop_all_locks(struct mm_struct *mm);
3143 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3144 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3145 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3146 extern struct file *get_task_exe_file(struct task_struct *task);
3148 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3149 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3151 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3152 const struct vm_special_mapping *sm);
3153 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3154 unsigned long addr, unsigned long len,
3155 unsigned long flags,
3156 const struct vm_special_mapping *spec);
3157 /* This is an obsolete alternative to _install_special_mapping. */
3158 extern int install_special_mapping(struct mm_struct *mm,
3159 unsigned long addr, unsigned long len,
3160 unsigned long flags, struct page **pages);
3162 unsigned long randomize_stack_top(unsigned long stack_top);
3163 unsigned long randomize_page(unsigned long start, unsigned long range);
3165 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3167 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3168 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3169 struct list_head *uf);
3170 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3171 unsigned long len, unsigned long prot, unsigned long flags,
3172 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3173 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3174 unsigned long start, size_t len, struct list_head *uf,
3176 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3177 struct list_head *uf);
3178 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3181 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3182 unsigned long start, unsigned long end,
3183 struct list_head *uf, bool unlock);
3184 extern int __mm_populate(unsigned long addr, unsigned long len,
3186 static inline void mm_populate(unsigned long addr, unsigned long len)
3189 (void) __mm_populate(addr, len, 1);
3192 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3195 /* These take the mm semaphore themselves */
3196 extern int __must_check vm_brk(unsigned long, unsigned long);
3197 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3198 extern int vm_munmap(unsigned long, size_t);
3199 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3200 unsigned long, unsigned long,
3201 unsigned long, unsigned long);
3203 struct vm_unmapped_area_info {
3204 #define VM_UNMAPPED_AREA_TOPDOWN 1
3205 unsigned long flags;
3206 unsigned long length;
3207 unsigned long low_limit;
3208 unsigned long high_limit;
3209 unsigned long align_mask;
3210 unsigned long align_offset;
3213 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3216 extern void truncate_inode_pages(struct address_space *, loff_t);
3217 extern void truncate_inode_pages_range(struct address_space *,
3218 loff_t lstart, loff_t lend);
3219 extern void truncate_inode_pages_final(struct address_space *);
3221 /* generic vm_area_ops exported for stackable file systems */
3222 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3223 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3224 pgoff_t start_pgoff, pgoff_t end_pgoff);
3225 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3227 extern unsigned long stack_guard_gap;
3228 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3229 int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
3230 struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
3232 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3233 int expand_downwards(struct vm_area_struct *vma, unsigned long address);
3235 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3236 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3237 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3238 struct vm_area_struct **pprev);
3241 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3242 * NULL if none. Assume start_addr < end_addr.
3244 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3245 unsigned long start_addr, unsigned long end_addr);
3248 * vma_lookup() - Find a VMA at a specific address
3249 * @mm: The process address space.
3250 * @addr: The user address.
3252 * Return: The vm_area_struct at the given address, %NULL otherwise.
3255 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3257 return mtree_load(&mm->mm_mt, addr);
3260 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3262 unsigned long vm_start = vma->vm_start;
3264 if (vma->vm_flags & VM_GROWSDOWN) {
3265 vm_start -= stack_guard_gap;
3266 if (vm_start > vma->vm_start)
3272 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3274 unsigned long vm_end = vma->vm_end;
3276 if (vma->vm_flags & VM_GROWSUP) {
3277 vm_end += stack_guard_gap;
3278 if (vm_end < vma->vm_end)
3279 vm_end = -PAGE_SIZE;
3284 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3286 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3289 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3290 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3291 unsigned long vm_start, unsigned long vm_end)
3293 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3295 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3301 static inline bool range_in_vma(struct vm_area_struct *vma,
3302 unsigned long start, unsigned long end)
3304 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3308 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3309 void vma_set_page_prot(struct vm_area_struct *vma);
3311 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3315 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3317 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3321 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3323 #ifdef CONFIG_NUMA_BALANCING
3324 unsigned long change_prot_numa(struct vm_area_struct *vma,
3325 unsigned long start, unsigned long end);
3328 struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
3329 unsigned long addr);
3330 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3331 unsigned long pfn, unsigned long size, pgprot_t);
3332 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3333 unsigned long pfn, unsigned long size, pgprot_t prot);
3334 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3335 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3336 struct page **pages, unsigned long *num);
3337 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3339 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3341 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3343 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3344 unsigned long pfn, pgprot_t pgprot);
3345 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3347 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3348 unsigned long addr, pfn_t pfn);
3349 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3351 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3352 unsigned long addr, struct page *page)
3354 int err = vm_insert_page(vma, addr, page);
3357 return VM_FAULT_OOM;
3358 if (err < 0 && err != -EBUSY)
3359 return VM_FAULT_SIGBUS;
3361 return VM_FAULT_NOPAGE;
3364 #ifndef io_remap_pfn_range
3365 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3366 unsigned long addr, unsigned long pfn,
3367 unsigned long size, pgprot_t prot)
3369 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3373 static inline vm_fault_t vmf_error(int err)
3376 return VM_FAULT_OOM;
3377 else if (err == -EHWPOISON)
3378 return VM_FAULT_HWPOISON;
3379 return VM_FAULT_SIGBUS;
3382 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3383 unsigned int foll_flags);
3385 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3387 if (vm_fault & VM_FAULT_OOM)
3389 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3390 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3391 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3397 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3398 * a (NUMA hinting) fault is required.
3400 static inline bool gup_can_follow_protnone(unsigned int flags)
3403 * FOLL_FORCE has to be able to make progress even if the VMA is
3404 * inaccessible. Further, FOLL_FORCE access usually does not represent
3405 * application behaviour and we should avoid triggering NUMA hinting
3408 return flags & FOLL_FORCE;
3411 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3412 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3413 unsigned long size, pte_fn_t fn, void *data);
3414 extern int apply_to_existing_page_range(struct mm_struct *mm,
3415 unsigned long address, unsigned long size,
3416 pte_fn_t fn, void *data);
3418 #ifdef CONFIG_PAGE_POISONING
3419 extern void __kernel_poison_pages(struct page *page, int numpages);
3420 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3421 extern bool _page_poisoning_enabled_early;
3422 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3423 static inline bool page_poisoning_enabled(void)
3425 return _page_poisoning_enabled_early;
3428 * For use in fast paths after init_mem_debugging() has run, or when a
3429 * false negative result is not harmful when called too early.
3431 static inline bool page_poisoning_enabled_static(void)
3433 return static_branch_unlikely(&_page_poisoning_enabled);
3435 static inline void kernel_poison_pages(struct page *page, int numpages)
3437 if (page_poisoning_enabled_static())
3438 __kernel_poison_pages(page, numpages);
3440 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3442 if (page_poisoning_enabled_static())
3443 __kernel_unpoison_pages(page, numpages);
3446 static inline bool page_poisoning_enabled(void) { return false; }
3447 static inline bool page_poisoning_enabled_static(void) { return false; }
3448 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3449 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3450 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3453 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3454 static inline bool want_init_on_alloc(gfp_t flags)
3456 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3459 return flags & __GFP_ZERO;
3462 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3463 static inline bool want_init_on_free(void)
3465 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3469 extern bool _debug_pagealloc_enabled_early;
3470 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3472 static inline bool debug_pagealloc_enabled(void)
3474 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3475 _debug_pagealloc_enabled_early;
3479 * For use in fast paths after mem_debugging_and_hardening_init() has run,
3480 * or when a false negative result is not harmful when called too early.
3482 static inline bool debug_pagealloc_enabled_static(void)
3484 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3487 return static_branch_unlikely(&_debug_pagealloc_enabled);
3491 * To support DEBUG_PAGEALLOC architecture must ensure that
3492 * __kernel_map_pages() never fails
3494 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3495 #ifdef CONFIG_DEBUG_PAGEALLOC
3496 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3498 if (debug_pagealloc_enabled_static())
3499 __kernel_map_pages(page, numpages, 1);
3502 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3504 if (debug_pagealloc_enabled_static())
3505 __kernel_map_pages(page, numpages, 0);
3508 extern unsigned int _debug_guardpage_minorder;
3509 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3511 static inline unsigned int debug_guardpage_minorder(void)
3513 return _debug_guardpage_minorder;
3516 static inline bool debug_guardpage_enabled(void)
3518 return static_branch_unlikely(&_debug_guardpage_enabled);
3521 static inline bool page_is_guard(struct page *page)
3523 if (!debug_guardpage_enabled())
3526 return PageGuard(page);
3529 bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order,
3531 static inline bool set_page_guard(struct zone *zone, struct page *page,
3532 unsigned int order, int migratetype)
3534 if (!debug_guardpage_enabled())
3536 return __set_page_guard(zone, page, order, migratetype);
3539 void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order,
3541 static inline void clear_page_guard(struct zone *zone, struct page *page,
3542 unsigned int order, int migratetype)
3544 if (!debug_guardpage_enabled())
3546 __clear_page_guard(zone, page, order, migratetype);
3549 #else /* CONFIG_DEBUG_PAGEALLOC */
3550 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3551 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3552 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3553 static inline bool debug_guardpage_enabled(void) { return false; }
3554 static inline bool page_is_guard(struct page *page) { return false; }
3555 static inline bool set_page_guard(struct zone *zone, struct page *page,
3556 unsigned int order, int migratetype) { return false; }
3557 static inline void clear_page_guard(struct zone *zone, struct page *page,
3558 unsigned int order, int migratetype) {}
3559 #endif /* CONFIG_DEBUG_PAGEALLOC */
3561 #ifdef __HAVE_ARCH_GATE_AREA
3562 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3563 extern int in_gate_area_no_mm(unsigned long addr);
3564 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3566 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3570 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3571 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3575 #endif /* __HAVE_ARCH_GATE_AREA */
3577 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3579 #ifdef CONFIG_SYSCTL
3580 extern int sysctl_drop_caches;
3581 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3585 void drop_slab(void);
3588 #define randomize_va_space 0
3590 extern int randomize_va_space;
3593 const char * arch_vma_name(struct vm_area_struct *vma);
3595 void print_vma_addr(char *prefix, unsigned long rip);
3597 static inline void print_vma_addr(char *prefix, unsigned long rip)
3602 void *sparse_buffer_alloc(unsigned long size);
3603 struct page * __populate_section_memmap(unsigned long pfn,
3604 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3605 struct dev_pagemap *pgmap);
3606 void pmd_init(void *addr);
3607 void pud_init(void *addr);
3608 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3609 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3610 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3611 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3612 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3613 struct vmem_altmap *altmap, struct page *reuse);
3614 void *vmemmap_alloc_block(unsigned long size, int node);
3616 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3617 struct vmem_altmap *altmap);
3618 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3619 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3620 unsigned long addr, unsigned long next);
3621 int vmemmap_check_pmd(pmd_t *pmd, int node,
3622 unsigned long addr, unsigned long next);
3623 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3624 int node, struct vmem_altmap *altmap);
3625 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3626 int node, struct vmem_altmap *altmap);
3627 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3628 struct vmem_altmap *altmap);
3629 void vmemmap_populate_print_last(void);
3630 #ifdef CONFIG_MEMORY_HOTPLUG
3631 void vmemmap_free(unsigned long start, unsigned long end,
3632 struct vmem_altmap *altmap);
3635 #define VMEMMAP_RESERVE_NR 2
3636 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP
3637 static inline bool __vmemmap_can_optimize(struct vmem_altmap *altmap,
3638 struct dev_pagemap *pgmap)
3640 unsigned long nr_pages;
3641 unsigned long nr_vmemmap_pages;
3643 if (!pgmap || !is_power_of_2(sizeof(struct page)))
3646 nr_pages = pgmap_vmemmap_nr(pgmap);
3647 nr_vmemmap_pages = ((nr_pages * sizeof(struct page)) >> PAGE_SHIFT);
3649 * For vmemmap optimization with DAX we need minimum 2 vmemmap
3650 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
3652 return !altmap && (nr_vmemmap_pages > VMEMMAP_RESERVE_NR);
3655 * If we don't have an architecture override, use the generic rule
3657 #ifndef vmemmap_can_optimize
3658 #define vmemmap_can_optimize __vmemmap_can_optimize
3662 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3663 struct dev_pagemap *pgmap)
3669 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3670 unsigned long nr_pages);
3673 MF_COUNT_INCREASED = 1 << 0,
3674 MF_ACTION_REQUIRED = 1 << 1,
3675 MF_MUST_KILL = 1 << 2,
3676 MF_SOFT_OFFLINE = 1 << 3,
3677 MF_UNPOISON = 1 << 4,
3678 MF_SW_SIMULATED = 1 << 5,
3679 MF_NO_RETRY = 1 << 6,
3681 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3682 unsigned long count, int mf_flags);
3683 extern int memory_failure(unsigned long pfn, int flags);
3684 extern void memory_failure_queue_kick(int cpu);
3685 extern int unpoison_memory(unsigned long pfn);
3686 extern void shake_page(struct page *p);
3687 extern atomic_long_t num_poisoned_pages __read_mostly;
3688 extern int soft_offline_page(unsigned long pfn, int flags);
3689 #ifdef CONFIG_MEMORY_FAILURE
3691 * Sysfs entries for memory failure handling statistics.
3693 extern const struct attribute_group memory_failure_attr_group;
3694 extern void memory_failure_queue(unsigned long pfn, int flags);
3695 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3696 bool *migratable_cleared);
3697 void num_poisoned_pages_inc(unsigned long pfn);
3698 void num_poisoned_pages_sub(unsigned long pfn, long i);
3699 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3701 static inline void memory_failure_queue(unsigned long pfn, int flags)
3705 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3706 bool *migratable_cleared)
3711 static inline void num_poisoned_pages_inc(unsigned long pfn)
3715 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3720 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3721 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3722 struct vm_area_struct *vma, struct list_head *to_kill,
3723 unsigned long ksm_addr);
3726 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3727 extern void memblk_nr_poison_inc(unsigned long pfn);
3728 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3730 static inline void memblk_nr_poison_inc(unsigned long pfn)
3734 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3739 #ifndef arch_memory_failure
3740 static inline int arch_memory_failure(unsigned long pfn, int flags)
3746 #ifndef arch_is_platform_page
3747 static inline bool arch_is_platform_page(u64 paddr)
3754 * Error handlers for various types of pages.
3757 MF_IGNORED, /* Error: cannot be handled */
3758 MF_FAILED, /* Error: handling failed */
3759 MF_DELAYED, /* Will be handled later */
3760 MF_RECOVERED, /* Successfully recovered */
3763 enum mf_action_page_type {
3765 MF_MSG_KERNEL_HIGH_ORDER,
3767 MF_MSG_DIFFERENT_COMPOUND,
3770 MF_MSG_UNMAP_FAILED,
3771 MF_MSG_DIRTY_SWAPCACHE,
3772 MF_MSG_CLEAN_SWAPCACHE,
3773 MF_MSG_DIRTY_MLOCKED_LRU,
3774 MF_MSG_CLEAN_MLOCKED_LRU,
3775 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3776 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3779 MF_MSG_TRUNCATED_LRU,
3786 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3787 extern void clear_huge_page(struct page *page,
3788 unsigned long addr_hint,
3789 unsigned int pages_per_huge_page);
3790 int copy_user_large_folio(struct folio *dst, struct folio *src,
3791 unsigned long addr_hint,
3792 struct vm_area_struct *vma);
3793 long copy_folio_from_user(struct folio *dst_folio,
3794 const void __user *usr_src,
3795 bool allow_pagefault);
3798 * vma_is_special_huge - Are transhuge page-table entries considered special?
3799 * @vma: Pointer to the struct vm_area_struct to consider
3801 * Whether transhuge page-table entries are considered "special" following
3802 * the definition in vm_normal_page().
3804 * Return: true if transhuge page-table entries should be considered special,
3807 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3809 return vma_is_dax(vma) || (vma->vm_file &&
3810 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3813 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3815 #if MAX_NUMNODES > 1
3816 void __init setup_nr_node_ids(void);
3818 static inline void setup_nr_node_ids(void) {}
3821 extern int memcmp_pages(struct page *page1, struct page *page2);
3823 static inline int pages_identical(struct page *page1, struct page *page2)
3825 return !memcmp_pages(page1, page2);
3828 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3829 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3830 pgoff_t first_index, pgoff_t nr,
3831 pgoff_t bitmap_pgoff,
3832 unsigned long *bitmap,
3836 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3837 pgoff_t first_index, pgoff_t nr);
3840 extern int sysctl_nr_trim_pages;
3842 #ifdef CONFIG_PRINTK
3843 void mem_dump_obj(void *object);
3845 static inline void mem_dump_obj(void *object) {}
3849 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3850 * @seals: the seals to check
3851 * @vma: the vma to operate on
3853 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3854 * the vma flags. Return 0 if check pass, or <0 for errors.
3856 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3858 if (seals & F_SEAL_FUTURE_WRITE) {
3860 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3861 * "future write" seal active.
3863 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3867 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3868 * MAP_SHARED and read-only, take care to not allow mprotect to
3869 * revert protections on such mappings. Do this only for shared
3870 * mappings. For private mappings, don't need to mask
3871 * VM_MAYWRITE as we still want them to be COW-writable.
3873 if (vma->vm_flags & VM_SHARED)
3874 vm_flags_clear(vma, VM_MAYWRITE);
3880 #ifdef CONFIG_ANON_VMA_NAME
3881 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3882 unsigned long len_in,
3883 struct anon_vma_name *anon_name);
3886 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3887 unsigned long len_in, struct anon_vma_name *anon_name) {
3892 #ifdef CONFIG_UNACCEPTED_MEMORY
3894 bool range_contains_unaccepted_memory(phys_addr_t start, phys_addr_t end);
3895 void accept_memory(phys_addr_t start, phys_addr_t end);
3899 static inline bool range_contains_unaccepted_memory(phys_addr_t start,
3905 static inline void accept_memory(phys_addr_t start, phys_addr_t end)
3911 #endif /* _LINUX_MM_H */