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 { mmap_assert_write_locked(vma->vm_mm); }
743 static inline void vma_mark_detached(struct vm_area_struct *vma,
746 static inline struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
747 unsigned long address)
752 #endif /* CONFIG_PER_VMA_LOCK */
754 extern const struct vm_operations_struct vma_dummy_vm_ops;
757 * WARNING: vma_init does not initialize vma->vm_lock.
758 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
760 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
762 memset(vma, 0, sizeof(*vma));
764 vma->vm_ops = &vma_dummy_vm_ops;
765 INIT_LIST_HEAD(&vma->anon_vma_chain);
766 vma_mark_detached(vma, false);
767 vma_numab_state_init(vma);
770 /* Use when VMA is not part of the VMA tree and needs no locking */
771 static inline void vm_flags_init(struct vm_area_struct *vma,
774 ACCESS_PRIVATE(vma, __vm_flags) = flags;
778 * Use when VMA is part of the VMA tree and modifications need coordination
779 * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
780 * it should be locked explicitly beforehand.
782 static inline void vm_flags_reset(struct vm_area_struct *vma,
785 vma_assert_write_locked(vma);
786 vm_flags_init(vma, flags);
789 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
792 vma_assert_write_locked(vma);
793 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
796 static inline void vm_flags_set(struct vm_area_struct *vma,
799 vma_start_write(vma);
800 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
803 static inline void vm_flags_clear(struct vm_area_struct *vma,
806 vma_start_write(vma);
807 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
811 * Use only if VMA is not part of the VMA tree or has no other users and
812 * therefore needs no locking.
814 static inline void __vm_flags_mod(struct vm_area_struct *vma,
815 vm_flags_t set, vm_flags_t clear)
817 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
821 * Use only when the order of set/clear operations is unimportant, otherwise
822 * use vm_flags_{set|clear} explicitly.
824 static inline void vm_flags_mod(struct vm_area_struct *vma,
825 vm_flags_t set, vm_flags_t clear)
827 vma_start_write(vma);
828 __vm_flags_mod(vma, set, clear);
831 static inline void vma_set_anonymous(struct vm_area_struct *vma)
836 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
842 * Indicate if the VMA is a heap for the given task; for
843 * /proc/PID/maps that is the heap of the main task.
845 static inline bool vma_is_initial_heap(const struct vm_area_struct *vma)
847 return vma->vm_start <= vma->vm_mm->brk &&
848 vma->vm_end >= vma->vm_mm->start_brk;
852 * Indicate if the VMA is a stack for the given task; for
853 * /proc/PID/maps that is the stack of the main task.
855 static inline bool vma_is_initial_stack(const struct vm_area_struct *vma)
858 * We make no effort to guess what a given thread considers to be
859 * its "stack". It's not even well-defined for programs written
862 return vma->vm_start <= vma->vm_mm->start_stack &&
863 vma->vm_end >= vma->vm_mm->start_stack;
866 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
868 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
873 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
874 VM_STACK_INCOMPLETE_SETUP)
880 static inline bool vma_is_foreign(struct vm_area_struct *vma)
885 if (current->mm != vma->vm_mm)
891 static inline bool vma_is_accessible(struct vm_area_struct *vma)
893 return vma->vm_flags & VM_ACCESS_FLAGS;
897 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
899 return mas_find(&vmi->mas, max - 1);
902 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
905 * Uses mas_find() to get the first VMA when the iterator starts.
906 * Calling mas_next() could skip the first entry.
908 return mas_find(&vmi->mas, ULONG_MAX);
912 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
914 return mas_next_range(&vmi->mas, ULONG_MAX);
918 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
920 return mas_prev(&vmi->mas, 0);
924 struct vm_area_struct *vma_iter_prev_range(struct vma_iterator *vmi)
926 return mas_prev_range(&vmi->mas, 0);
929 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
931 return vmi->mas.index;
934 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
936 return vmi->mas.last + 1;
938 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
941 return mas_expected_entries(&vmi->mas, count);
944 /* Free any unused preallocations */
945 static inline void vma_iter_free(struct vma_iterator *vmi)
947 mas_destroy(&vmi->mas);
950 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
951 struct vm_area_struct *vma)
953 vmi->mas.index = vma->vm_start;
954 vmi->mas.last = vma->vm_end - 1;
955 mas_store(&vmi->mas, vma);
956 if (unlikely(mas_is_err(&vmi->mas)))
962 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
964 mas_pause(&vmi->mas);
967 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
969 mas_set(&vmi->mas, addr);
972 #define for_each_vma(__vmi, __vma) \
973 while (((__vma) = vma_next(&(__vmi))) != NULL)
975 /* The MM code likes to work with exclusive end addresses */
976 #define for_each_vma_range(__vmi, __vma, __end) \
977 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
981 * The vma_is_shmem is not inline because it is used only by slow
982 * paths in userfault.
984 bool vma_is_shmem(struct vm_area_struct *vma);
985 bool vma_is_anon_shmem(struct vm_area_struct *vma);
987 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
988 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
991 int vma_is_stack_for_current(struct vm_area_struct *vma);
993 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
994 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
1000 * compound_order() can be called without holding a reference, which means
1001 * that niceties like page_folio() don't work. These callers should be
1002 * prepared to handle wild return values. For example, PG_head may be
1003 * set before _folio_order is initialised, or this may be a tail page.
1004 * See compaction.c for some good examples.
1006 static inline unsigned int compound_order(struct page *page)
1008 struct folio *folio = (struct folio *)page;
1010 if (!test_bit(PG_head, &folio->flags))
1012 return folio->_folio_order;
1016 * folio_order - The allocation order of a folio.
1017 * @folio: The folio.
1019 * A folio is composed of 2^order pages. See get_order() for the definition
1022 * Return: The order of the folio.
1024 static inline unsigned int folio_order(struct folio *folio)
1026 if (!folio_test_large(folio))
1028 return folio->_folio_order;
1031 #include <linux/huge_mm.h>
1034 * Methods to modify the page usage count.
1036 * What counts for a page usage:
1037 * - cache mapping (page->mapping)
1038 * - private data (page->private)
1039 * - page mapped in a task's page tables, each mapping
1040 * is counted separately
1042 * Also, many kernel routines increase the page count before a critical
1043 * routine so they can be sure the page doesn't go away from under them.
1047 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1049 static inline int put_page_testzero(struct page *page)
1051 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1052 return page_ref_dec_and_test(page);
1055 static inline int folio_put_testzero(struct folio *folio)
1057 return put_page_testzero(&folio->page);
1061 * Try to grab a ref unless the page has a refcount of zero, return false if
1063 * This can be called when MMU is off so it must not access
1064 * any of the virtual mappings.
1066 static inline bool get_page_unless_zero(struct page *page)
1068 return page_ref_add_unless(page, 1, 0);
1071 static inline struct folio *folio_get_nontail_page(struct page *page)
1073 if (unlikely(!get_page_unless_zero(page)))
1075 return (struct folio *)page;
1078 extern int page_is_ram(unsigned long pfn);
1086 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1087 unsigned long desc);
1089 /* Support for virtually mapped pages */
1090 struct page *vmalloc_to_page(const void *addr);
1091 unsigned long vmalloc_to_pfn(const void *addr);
1094 * Determine if an address is within the vmalloc range
1096 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1097 * is no special casing required.
1100 extern bool is_vmalloc_addr(const void *x);
1101 extern int is_vmalloc_or_module_addr(const void *x);
1103 static inline bool is_vmalloc_addr(const void *x)
1107 static inline int is_vmalloc_or_module_addr(const void *x)
1114 * How many times the entire folio is mapped as a single unit (eg by a
1115 * PMD or PUD entry). This is probably not what you want, except for
1116 * debugging purposes - it does not include PTE-mapped sub-pages; look
1117 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1119 static inline int folio_entire_mapcount(struct folio *folio)
1121 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1122 return atomic_read(&folio->_entire_mapcount) + 1;
1126 * The atomic page->_mapcount, starts from -1: so that transitions
1127 * both from it and to it can be tracked, using atomic_inc_and_test
1128 * and atomic_add_negative(-1).
1130 static inline void page_mapcount_reset(struct page *page)
1132 atomic_set(&(page)->_mapcount, -1);
1136 * page_mapcount() - Number of times this precise page is mapped.
1139 * The number of times this page is mapped. If this page is part of
1140 * a large folio, it includes the number of times this page is mapped
1141 * as part of that folio.
1143 * The result is undefined for pages which cannot be mapped into userspace.
1144 * For example SLAB or special types of pages. See function page_has_type().
1145 * They use this field in struct page differently.
1147 static inline int page_mapcount(struct page *page)
1149 int mapcount = atomic_read(&page->_mapcount) + 1;
1151 if (unlikely(PageCompound(page)))
1152 mapcount += folio_entire_mapcount(page_folio(page));
1157 int folio_total_mapcount(struct folio *folio);
1160 * folio_mapcount() - Calculate the number of mappings of this folio.
1161 * @folio: The folio.
1163 * A large folio tracks both how many times the entire folio is mapped,
1164 * and how many times each individual page in the folio is mapped.
1165 * This function calculates the total number of times the folio is
1168 * Return: The number of times this folio is mapped.
1170 static inline int folio_mapcount(struct folio *folio)
1172 if (likely(!folio_test_large(folio)))
1173 return atomic_read(&folio->_mapcount) + 1;
1174 return folio_total_mapcount(folio);
1177 static inline int total_mapcount(struct page *page)
1179 if (likely(!PageCompound(page)))
1180 return atomic_read(&page->_mapcount) + 1;
1181 return folio_total_mapcount(page_folio(page));
1184 static inline bool folio_large_is_mapped(struct folio *folio)
1187 * Reading _entire_mapcount below could be omitted if hugetlb
1188 * participated in incrementing nr_pages_mapped when compound mapped.
1190 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1191 atomic_read(&folio->_entire_mapcount) >= 0;
1195 * folio_mapped - Is this folio mapped into userspace?
1196 * @folio: The folio.
1198 * Return: True if any page in this folio is referenced by user page tables.
1200 static inline bool folio_mapped(struct folio *folio)
1202 if (likely(!folio_test_large(folio)))
1203 return atomic_read(&folio->_mapcount) >= 0;
1204 return folio_large_is_mapped(folio);
1208 * Return true if this page is mapped into pagetables.
1209 * For compound page it returns true if any sub-page of compound page is mapped,
1210 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1212 static inline bool page_mapped(struct page *page)
1214 if (likely(!PageCompound(page)))
1215 return atomic_read(&page->_mapcount) >= 0;
1216 return folio_large_is_mapped(page_folio(page));
1219 static inline struct page *virt_to_head_page(const void *x)
1221 struct page *page = virt_to_page(x);
1223 return compound_head(page);
1226 static inline struct folio *virt_to_folio(const void *x)
1228 struct page *page = virt_to_page(x);
1230 return page_folio(page);
1233 void __folio_put(struct folio *folio);
1235 void put_pages_list(struct list_head *pages);
1237 void split_page(struct page *page, unsigned int order);
1238 void folio_copy(struct folio *dst, struct folio *src);
1240 unsigned long nr_free_buffer_pages(void);
1243 * Compound pages have a destructor function. Provide a
1244 * prototype for that function and accessor functions.
1245 * These are _only_ valid on the head of a compound page.
1247 typedef void compound_page_dtor(struct page *);
1249 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1250 enum compound_dtor_id {
1253 #ifdef CONFIG_HUGETLB_PAGE
1256 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1257 TRANSHUGE_PAGE_DTOR,
1262 static inline void folio_set_compound_dtor(struct folio *folio,
1263 enum compound_dtor_id compound_dtor)
1265 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1266 folio->_folio_dtor = compound_dtor;
1269 void destroy_large_folio(struct folio *folio);
1271 /* Returns the number of bytes in this potentially compound page. */
1272 static inline unsigned long page_size(struct page *page)
1274 return PAGE_SIZE << compound_order(page);
1277 /* Returns the number of bits needed for the number of bytes in a page */
1278 static inline unsigned int page_shift(struct page *page)
1280 return PAGE_SHIFT + compound_order(page);
1284 * thp_order - Order of a transparent huge page.
1285 * @page: Head page of a transparent huge page.
1287 static inline unsigned int thp_order(struct page *page)
1289 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1290 return compound_order(page);
1294 * thp_size - Size of a transparent huge page.
1295 * @page: Head page of a transparent huge page.
1297 * Return: Number of bytes in this page.
1299 static inline unsigned long thp_size(struct page *page)
1301 return PAGE_SIZE << thp_order(page);
1304 void free_compound_page(struct page *page);
1308 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1309 * servicing faults for write access. In the normal case, do always want
1310 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1311 * that do not have writing enabled, when used by access_process_vm.
1313 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1315 if (likely(vma->vm_flags & VM_WRITE))
1316 pte = pte_mkwrite(pte);
1320 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1321 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1323 vm_fault_t finish_fault(struct vm_fault *vmf);
1324 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1328 * Multiple processes may "see" the same page. E.g. for untouched
1329 * mappings of /dev/null, all processes see the same page full of
1330 * zeroes, and text pages of executables and shared libraries have
1331 * only one copy in memory, at most, normally.
1333 * For the non-reserved pages, page_count(page) denotes a reference count.
1334 * page_count() == 0 means the page is free. page->lru is then used for
1335 * freelist management in the buddy allocator.
1336 * page_count() > 0 means the page has been allocated.
1338 * Pages are allocated by the slab allocator in order to provide memory
1339 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1340 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1341 * unless a particular usage is carefully commented. (the responsibility of
1342 * freeing the kmalloc memory is the caller's, of course).
1344 * A page may be used by anyone else who does a __get_free_page().
1345 * In this case, page_count still tracks the references, and should only
1346 * be used through the normal accessor functions. The top bits of page->flags
1347 * and page->virtual store page management information, but all other fields
1348 * are unused and could be used privately, carefully. The management of this
1349 * page is the responsibility of the one who allocated it, and those who have
1350 * subsequently been given references to it.
1352 * The other pages (we may call them "pagecache pages") are completely
1353 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1354 * The following discussion applies only to them.
1356 * A pagecache page contains an opaque `private' member, which belongs to the
1357 * page's address_space. Usually, this is the address of a circular list of
1358 * the page's disk buffers. PG_private must be set to tell the VM to call
1359 * into the filesystem to release these pages.
1361 * A page may belong to an inode's memory mapping. In this case, page->mapping
1362 * is the pointer to the inode, and page->index is the file offset of the page,
1363 * in units of PAGE_SIZE.
1365 * If pagecache pages are not associated with an inode, they are said to be
1366 * anonymous pages. These may become associated with the swapcache, and in that
1367 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1369 * In either case (swapcache or inode backed), the pagecache itself holds one
1370 * reference to the page. Setting PG_private should also increment the
1371 * refcount. The each user mapping also has a reference to the page.
1373 * The pagecache pages are stored in a per-mapping radix tree, which is
1374 * rooted at mapping->i_pages, and indexed by offset.
1375 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1376 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1378 * All pagecache pages may be subject to I/O:
1379 * - inode pages may need to be read from disk,
1380 * - inode pages which have been modified and are MAP_SHARED may need
1381 * to be written back to the inode on disk,
1382 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1383 * modified may need to be swapped out to swap space and (later) to be read
1387 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1388 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1390 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1391 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1393 if (!static_branch_unlikely(&devmap_managed_key))
1395 if (!is_zone_device_page(page))
1397 return __put_devmap_managed_page_refs(page, refs);
1399 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1400 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1404 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1406 static inline bool put_devmap_managed_page(struct page *page)
1408 return put_devmap_managed_page_refs(page, 1);
1411 /* 127: arbitrary random number, small enough to assemble well */
1412 #define folio_ref_zero_or_close_to_overflow(folio) \
1413 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1416 * folio_get - Increment the reference count on a folio.
1417 * @folio: The folio.
1419 * Context: May be called in any context, as long as you know that
1420 * you have a refcount on the folio. If you do not already have one,
1421 * folio_try_get() may be the right interface for you to use.
1423 static inline void folio_get(struct folio *folio)
1425 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1426 folio_ref_inc(folio);
1429 static inline void get_page(struct page *page)
1431 folio_get(page_folio(page));
1434 static inline __must_check bool try_get_page(struct page *page)
1436 page = compound_head(page);
1437 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1444 * folio_put - Decrement the reference count on a folio.
1445 * @folio: The folio.
1447 * If the folio's reference count reaches zero, the memory will be
1448 * released back to the page allocator and may be used by another
1449 * allocation immediately. Do not access the memory or the struct folio
1450 * after calling folio_put() unless you can be sure that it wasn't the
1453 * Context: May be called in process or interrupt context, but not in NMI
1454 * context. May be called while holding a spinlock.
1456 static inline void folio_put(struct folio *folio)
1458 if (folio_put_testzero(folio))
1463 * folio_put_refs - Reduce the reference count on a folio.
1464 * @folio: The folio.
1465 * @refs: The amount to subtract from the folio's reference count.
1467 * If the folio's reference count reaches zero, the memory will be
1468 * released back to the page allocator and may be used by another
1469 * allocation immediately. Do not access the memory or the struct folio
1470 * after calling folio_put_refs() unless you can be sure that these weren't
1471 * the last references.
1473 * Context: May be called in process or interrupt context, but not in NMI
1474 * context. May be called while holding a spinlock.
1476 static inline void folio_put_refs(struct folio *folio, int refs)
1478 if (folio_ref_sub_and_test(folio, refs))
1483 * union release_pages_arg - an array of pages or folios
1485 * release_pages() releases a simple array of multiple pages, and
1486 * accepts various different forms of said page array: either
1487 * a regular old boring array of pages, an array of folios, or
1488 * an array of encoded page pointers.
1490 * The transparent union syntax for this kind of "any of these
1491 * argument types" is all kinds of ugly, so look away.
1494 struct page **pages;
1495 struct folio **folios;
1496 struct encoded_page **encoded_pages;
1497 } release_pages_arg __attribute__ ((__transparent_union__));
1499 void release_pages(release_pages_arg, int nr);
1502 * folios_put - Decrement the reference count on an array of folios.
1503 * @folios: The folios.
1504 * @nr: How many folios there are.
1506 * Like folio_put(), but for an array of folios. This is more efficient
1507 * than writing the loop yourself as it will optimise the locks which
1508 * need to be taken if the folios are freed.
1510 * Context: May be called in process or interrupt context, but not in NMI
1511 * context. May be called while holding a spinlock.
1513 static inline void folios_put(struct folio **folios, unsigned int nr)
1515 release_pages(folios, nr);
1518 static inline void put_page(struct page *page)
1520 struct folio *folio = page_folio(page);
1523 * For some devmap managed pages we need to catch refcount transition
1526 if (put_devmap_managed_page(&folio->page))
1532 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1533 * the page's refcount so that two separate items are tracked: the original page
1534 * reference count, and also a new count of how many pin_user_pages() calls were
1535 * made against the page. ("gup-pinned" is another term for the latter).
1537 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1538 * distinct from normal pages. As such, the unpin_user_page() call (and its
1539 * variants) must be used in order to release gup-pinned pages.
1543 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1544 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1545 * simpler, due to the fact that adding an even power of two to the page
1546 * refcount has the effect of using only the upper N bits, for the code that
1547 * counts up using the bias value. This means that the lower bits are left for
1548 * the exclusive use of the original code that increments and decrements by one
1549 * (or at least, by much smaller values than the bias value).
1551 * Of course, once the lower bits overflow into the upper bits (and this is
1552 * OK, because subtraction recovers the original values), then visual inspection
1553 * no longer suffices to directly view the separate counts. However, for normal
1554 * applications that don't have huge page reference counts, this won't be an
1557 * Locking: the lockless algorithm described in folio_try_get_rcu()
1558 * provides safe operation for get_user_pages(), page_mkclean() and
1559 * other calls that race to set up page table entries.
1561 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1563 void unpin_user_page(struct page *page);
1564 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1566 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1568 void unpin_user_pages(struct page **pages, unsigned long npages);
1570 static inline bool is_cow_mapping(vm_flags_t flags)
1572 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1576 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1579 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1580 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1581 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1582 * underlying memory if ptrace is active, so this is only possible if
1583 * ptrace does not apply. Note that there is no mprotect() to upgrade
1584 * write permissions later.
1586 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1590 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1591 #define SECTION_IN_PAGE_FLAGS
1595 * The identification function is mainly used by the buddy allocator for
1596 * determining if two pages could be buddies. We are not really identifying
1597 * the zone since we could be using the section number id if we do not have
1598 * node id available in page flags.
1599 * We only guarantee that it will return the same value for two combinable
1602 static inline int page_zone_id(struct page *page)
1604 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1607 #ifdef NODE_NOT_IN_PAGE_FLAGS
1608 extern int page_to_nid(const struct page *page);
1610 static inline int page_to_nid(const struct page *page)
1612 struct page *p = (struct page *)page;
1614 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1618 static inline int folio_nid(const struct folio *folio)
1620 return page_to_nid(&folio->page);
1623 #ifdef CONFIG_NUMA_BALANCING
1624 /* page access time bits needs to hold at least 4 seconds */
1625 #define PAGE_ACCESS_TIME_MIN_BITS 12
1626 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1627 #define PAGE_ACCESS_TIME_BUCKETS \
1628 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1630 #define PAGE_ACCESS_TIME_BUCKETS 0
1633 #define PAGE_ACCESS_TIME_MASK \
1634 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1636 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1638 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1641 static inline int cpupid_to_pid(int cpupid)
1643 return cpupid & LAST__PID_MASK;
1646 static inline int cpupid_to_cpu(int cpupid)
1648 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1651 static inline int cpupid_to_nid(int cpupid)
1653 return cpu_to_node(cpupid_to_cpu(cpupid));
1656 static inline bool cpupid_pid_unset(int cpupid)
1658 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1661 static inline bool cpupid_cpu_unset(int cpupid)
1663 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1666 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1668 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1671 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1672 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1673 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1675 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1678 static inline int page_cpupid_last(struct page *page)
1680 return page->_last_cpupid;
1682 static inline void page_cpupid_reset_last(struct page *page)
1684 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1687 static inline int page_cpupid_last(struct page *page)
1689 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1692 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1694 static inline void page_cpupid_reset_last(struct page *page)
1696 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1698 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1700 static inline int xchg_page_access_time(struct page *page, int time)
1704 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1705 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1708 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1710 unsigned int pid_bit;
1712 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1713 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) {
1714 __set_bit(pid_bit, &vma->numab_state->access_pids[1]);
1717 #else /* !CONFIG_NUMA_BALANCING */
1718 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1720 return page_to_nid(page); /* XXX */
1723 static inline int xchg_page_access_time(struct page *page, int time)
1728 static inline int page_cpupid_last(struct page *page)
1730 return page_to_nid(page); /* XXX */
1733 static inline int cpupid_to_nid(int cpupid)
1738 static inline int cpupid_to_pid(int cpupid)
1743 static inline int cpupid_to_cpu(int cpupid)
1748 static inline int cpu_pid_to_cpupid(int nid, int pid)
1753 static inline bool cpupid_pid_unset(int cpupid)
1758 static inline void page_cpupid_reset_last(struct page *page)
1762 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1767 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1770 #endif /* CONFIG_NUMA_BALANCING */
1772 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1775 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1776 * setting tags for all pages to native kernel tag value 0xff, as the default
1777 * value 0x00 maps to 0xff.
1780 static inline u8 page_kasan_tag(const struct page *page)
1784 if (kasan_enabled()) {
1785 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1792 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1794 unsigned long old_flags, flags;
1796 if (!kasan_enabled())
1800 old_flags = READ_ONCE(page->flags);
1803 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1804 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1805 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1808 static inline void page_kasan_tag_reset(struct page *page)
1810 if (kasan_enabled())
1811 page_kasan_tag_set(page, 0xff);
1814 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1816 static inline u8 page_kasan_tag(const struct page *page)
1821 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1822 static inline void page_kasan_tag_reset(struct page *page) { }
1824 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1826 static inline struct zone *page_zone(const struct page *page)
1828 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1831 static inline pg_data_t *page_pgdat(const struct page *page)
1833 return NODE_DATA(page_to_nid(page));
1836 static inline struct zone *folio_zone(const struct folio *folio)
1838 return page_zone(&folio->page);
1841 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1843 return page_pgdat(&folio->page);
1846 #ifdef SECTION_IN_PAGE_FLAGS
1847 static inline void set_page_section(struct page *page, unsigned long section)
1849 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1850 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1853 static inline unsigned long page_to_section(const struct page *page)
1855 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1860 * folio_pfn - Return the Page Frame Number of a folio.
1861 * @folio: The folio.
1863 * A folio may contain multiple pages. The pages have consecutive
1864 * Page Frame Numbers.
1866 * Return: The Page Frame Number of the first page in the folio.
1868 static inline unsigned long folio_pfn(struct folio *folio)
1870 return page_to_pfn(&folio->page);
1873 static inline struct folio *pfn_folio(unsigned long pfn)
1875 return page_folio(pfn_to_page(pfn));
1879 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1880 * @folio: The folio.
1882 * This function checks if a folio has been pinned via a call to
1883 * a function in the pin_user_pages() family.
1885 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1886 * because it means "definitely not pinned for DMA", but true means "probably
1887 * pinned for DMA, but possibly a false positive due to having at least
1888 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1890 * False positives are OK, because: a) it's unlikely for a folio to
1891 * get that many refcounts, and b) all the callers of this routine are
1892 * expected to be able to deal gracefully with a false positive.
1894 * For large folios, the result will be exactly correct. That's because
1895 * we have more tracking data available: the _pincount field is used
1896 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1898 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1900 * Return: True, if it is likely that the page has been "dma-pinned".
1901 * False, if the page is definitely not dma-pinned.
1903 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1905 if (folio_test_large(folio))
1906 return atomic_read(&folio->_pincount) > 0;
1909 * folio_ref_count() is signed. If that refcount overflows, then
1910 * folio_ref_count() returns a negative value, and callers will avoid
1911 * further incrementing the refcount.
1913 * Here, for that overflow case, use the sign bit to count a little
1914 * bit higher via unsigned math, and thus still get an accurate result.
1916 return ((unsigned int)folio_ref_count(folio)) >=
1917 GUP_PIN_COUNTING_BIAS;
1920 static inline bool page_maybe_dma_pinned(struct page *page)
1922 return folio_maybe_dma_pinned(page_folio(page));
1926 * This should most likely only be called during fork() to see whether we
1927 * should break the cow immediately for an anon page on the src mm.
1929 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1931 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1934 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1936 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1939 return page_maybe_dma_pinned(page);
1943 * is_zero_page - Query if a page is a zero page
1944 * @page: The page to query
1946 * This returns true if @page is one of the permanent zero pages.
1948 static inline bool is_zero_page(const struct page *page)
1950 return is_zero_pfn(page_to_pfn(page));
1954 * is_zero_folio - Query if a folio is a zero page
1955 * @folio: The folio to query
1957 * This returns true if @folio is one of the permanent zero pages.
1959 static inline bool is_zero_folio(const struct folio *folio)
1961 return is_zero_page(&folio->page);
1964 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1965 #ifdef CONFIG_MIGRATION
1966 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1969 int mt = folio_migratetype(folio);
1971 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1974 /* The zero page can be "pinned" but gets special handling. */
1975 if (is_zero_folio(folio))
1978 /* Coherent device memory must always allow eviction. */
1979 if (folio_is_device_coherent(folio))
1982 /* Otherwise, non-movable zone folios can be pinned. */
1983 return !folio_is_zone_movable(folio);
1987 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1993 static inline void set_page_zone(struct page *page, enum zone_type zone)
1995 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1996 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1999 static inline void set_page_node(struct page *page, unsigned long node)
2001 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
2002 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
2005 static inline void set_page_links(struct page *page, enum zone_type zone,
2006 unsigned long node, unsigned long pfn)
2008 set_page_zone(page, zone);
2009 set_page_node(page, node);
2010 #ifdef SECTION_IN_PAGE_FLAGS
2011 set_page_section(page, pfn_to_section_nr(pfn));
2016 * folio_nr_pages - The number of pages in the folio.
2017 * @folio: The folio.
2019 * Return: A positive power of two.
2021 static inline long folio_nr_pages(struct folio *folio)
2023 if (!folio_test_large(folio))
2026 return folio->_folio_nr_pages;
2028 return 1L << folio->_folio_order;
2033 * compound_nr() returns the number of pages in this potentially compound
2034 * page. compound_nr() can be called on a tail page, and is defined to
2035 * return 1 in that case.
2037 static inline unsigned long compound_nr(struct page *page)
2039 struct folio *folio = (struct folio *)page;
2041 if (!test_bit(PG_head, &folio->flags))
2044 return folio->_folio_nr_pages;
2046 return 1L << folio->_folio_order;
2051 * thp_nr_pages - The number of regular pages in this huge page.
2052 * @page: The head page of a huge page.
2054 static inline int thp_nr_pages(struct page *page)
2056 return folio_nr_pages((struct folio *)page);
2060 * folio_next - Move to the next physical folio.
2061 * @folio: The folio we're currently operating on.
2063 * If you have physically contiguous memory which may span more than
2064 * one folio (eg a &struct bio_vec), use this function to move from one
2065 * folio to the next. Do not use it if the memory is only virtually
2066 * contiguous as the folios are almost certainly not adjacent to each
2067 * other. This is the folio equivalent to writing ``page++``.
2069 * Context: We assume that the folios are refcounted and/or locked at a
2070 * higher level and do not adjust the reference counts.
2071 * Return: The next struct folio.
2073 static inline struct folio *folio_next(struct folio *folio)
2075 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2079 * folio_shift - The size of the memory described by this folio.
2080 * @folio: The folio.
2082 * A folio represents a number of bytes which is a power-of-two in size.
2083 * This function tells you which power-of-two the folio is. See also
2084 * folio_size() and folio_order().
2086 * Context: The caller should have a reference on the folio to prevent
2087 * it from being split. It is not necessary for the folio to be locked.
2088 * Return: The base-2 logarithm of the size of this folio.
2090 static inline unsigned int folio_shift(struct folio *folio)
2092 return PAGE_SHIFT + folio_order(folio);
2096 * folio_size - The number of bytes in a folio.
2097 * @folio: The folio.
2099 * Context: The caller should have a reference on the folio to prevent
2100 * it from being split. It is not necessary for the folio to be locked.
2101 * Return: The number of bytes in this folio.
2103 static inline size_t folio_size(struct folio *folio)
2105 return PAGE_SIZE << folio_order(folio);
2109 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2110 * @folio: The folio.
2112 * folio_estimated_sharers() aims to serve as a function to efficiently
2113 * estimate the number of processes sharing a folio. This is done by
2114 * looking at the precise mapcount of the first subpage in the folio, and
2115 * assuming the other subpages are the same. This may not be true for large
2116 * folios. If you want exact mapcounts for exact calculations, look at
2117 * page_mapcount() or folio_total_mapcount().
2119 * Return: The estimated number of processes sharing a folio.
2121 static inline int folio_estimated_sharers(struct folio *folio)
2123 return page_mapcount(folio_page(folio, 0));
2126 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2127 static inline int arch_make_page_accessible(struct page *page)
2133 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2134 static inline int arch_make_folio_accessible(struct folio *folio)
2137 long i, nr = folio_nr_pages(folio);
2139 for (i = 0; i < nr; i++) {
2140 ret = arch_make_page_accessible(folio_page(folio, i));
2150 * Some inline functions in vmstat.h depend on page_zone()
2152 #include <linux/vmstat.h>
2154 static __always_inline void *lowmem_page_address(const struct page *page)
2156 return page_to_virt(page);
2159 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2160 #define HASHED_PAGE_VIRTUAL
2163 #if defined(WANT_PAGE_VIRTUAL)
2164 static inline void *page_address(const struct page *page)
2166 return page->virtual;
2168 static inline void set_page_address(struct page *page, void *address)
2170 page->virtual = address;
2172 #define page_address_init() do { } while(0)
2175 #if defined(HASHED_PAGE_VIRTUAL)
2176 void *page_address(const struct page *page);
2177 void set_page_address(struct page *page, void *virtual);
2178 void page_address_init(void);
2181 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2182 #define page_address(page) lowmem_page_address(page)
2183 #define set_page_address(page, address) do { } while(0)
2184 #define page_address_init() do { } while(0)
2187 static inline void *folio_address(const struct folio *folio)
2189 return page_address(&folio->page);
2192 extern pgoff_t __page_file_index(struct page *page);
2195 * Return the pagecache index of the passed page. Regular pagecache pages
2196 * use ->index whereas swapcache pages use swp_offset(->private)
2198 static inline pgoff_t page_index(struct page *page)
2200 if (unlikely(PageSwapCache(page)))
2201 return __page_file_index(page);
2206 * Return true only if the page has been allocated with
2207 * ALLOC_NO_WATERMARKS and the low watermark was not
2208 * met implying that the system is under some pressure.
2210 static inline bool page_is_pfmemalloc(const struct page *page)
2213 * lru.next has bit 1 set if the page is allocated from the
2214 * pfmemalloc reserves. Callers may simply overwrite it if
2215 * they do not need to preserve that information.
2217 return (uintptr_t)page->lru.next & BIT(1);
2221 * Return true only if the folio has been allocated with
2222 * ALLOC_NO_WATERMARKS and the low watermark was not
2223 * met implying that the system is under some pressure.
2225 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2228 * lru.next has bit 1 set if the page is allocated from the
2229 * pfmemalloc reserves. Callers may simply overwrite it if
2230 * they do not need to preserve that information.
2232 return (uintptr_t)folio->lru.next & BIT(1);
2236 * Only to be called by the page allocator on a freshly allocated
2239 static inline void set_page_pfmemalloc(struct page *page)
2241 page->lru.next = (void *)BIT(1);
2244 static inline void clear_page_pfmemalloc(struct page *page)
2246 page->lru.next = NULL;
2250 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2252 extern void pagefault_out_of_memory(void);
2254 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2255 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2256 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2259 * Parameter block passed down to zap_pte_range in exceptional cases.
2261 struct zap_details {
2262 struct folio *single_folio; /* Locked folio to be unmapped */
2263 bool even_cows; /* Zap COWed private pages too? */
2264 zap_flags_t zap_flags; /* Extra flags for zapping */
2268 * Whether to drop the pte markers, for example, the uffd-wp information for
2269 * file-backed memory. This should only be specified when we will completely
2270 * drop the page in the mm, either by truncation or unmapping of the vma. By
2271 * default, the flag is not set.
2273 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2274 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2275 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2277 #ifdef CONFIG_SCHED_MM_CID
2278 void sched_mm_cid_before_execve(struct task_struct *t);
2279 void sched_mm_cid_after_execve(struct task_struct *t);
2280 void sched_mm_cid_fork(struct task_struct *t);
2281 void sched_mm_cid_exit_signals(struct task_struct *t);
2282 static inline int task_mm_cid(struct task_struct *t)
2287 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2288 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2289 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2290 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2291 static inline int task_mm_cid(struct task_struct *t)
2294 * Use the processor id as a fall-back when the mm cid feature is
2295 * disabled. This provides functional per-cpu data structure accesses
2296 * in user-space, althrough it won't provide the memory usage benefits.
2298 return raw_smp_processor_id();
2303 extern bool can_do_mlock(void);
2305 static inline bool can_do_mlock(void) { return false; }
2307 extern int user_shm_lock(size_t, struct ucounts *);
2308 extern void user_shm_unlock(size_t, struct ucounts *);
2310 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2312 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2314 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2317 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2318 unsigned long size);
2319 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2320 unsigned long size, struct zap_details *details);
2321 static inline void zap_vma_pages(struct vm_area_struct *vma)
2323 zap_page_range_single(vma, vma->vm_start,
2324 vma->vm_end - vma->vm_start, NULL);
2326 void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
2327 struct vm_area_struct *start_vma, unsigned long start,
2328 unsigned long end, unsigned long tree_end, bool mm_wr_locked);
2330 struct mmu_notifier_range;
2332 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2333 unsigned long end, unsigned long floor, unsigned long ceiling);
2335 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2336 int follow_pte(struct mm_struct *mm, unsigned long address,
2337 pte_t **ptepp, spinlock_t **ptlp);
2338 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2339 unsigned long *pfn);
2340 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2341 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2342 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2343 void *buf, int len, int write);
2345 extern void truncate_pagecache(struct inode *inode, loff_t new);
2346 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2347 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2348 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2349 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2351 struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
2352 unsigned long address, struct pt_regs *regs);
2355 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2356 unsigned long address, unsigned int flags,
2357 struct pt_regs *regs);
2358 extern int fixup_user_fault(struct mm_struct *mm,
2359 unsigned long address, unsigned int fault_flags,
2361 void unmap_mapping_pages(struct address_space *mapping,
2362 pgoff_t start, pgoff_t nr, bool even_cows);
2363 void unmap_mapping_range(struct address_space *mapping,
2364 loff_t const holebegin, loff_t const holelen, int even_cows);
2366 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2367 unsigned long address, unsigned int flags,
2368 struct pt_regs *regs)
2370 /* should never happen if there's no MMU */
2372 return VM_FAULT_SIGBUS;
2374 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2375 unsigned int fault_flags, bool *unlocked)
2377 /* should never happen if there's no MMU */
2381 static inline void unmap_mapping_pages(struct address_space *mapping,
2382 pgoff_t start, pgoff_t nr, bool even_cows) { }
2383 static inline void unmap_mapping_range(struct address_space *mapping,
2384 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2387 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2388 loff_t const holebegin, loff_t const holelen)
2390 unmap_mapping_range(mapping, holebegin, holelen, 0);
2393 static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
2394 unsigned long addr);
2396 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2397 void *buf, int len, unsigned int gup_flags);
2398 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2399 void *buf, int len, unsigned int gup_flags);
2400 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2401 void *buf, int len, unsigned int gup_flags);
2403 long get_user_pages_remote(struct mm_struct *mm,
2404 unsigned long start, unsigned long nr_pages,
2405 unsigned int gup_flags, struct page **pages,
2407 long pin_user_pages_remote(struct mm_struct *mm,
2408 unsigned long start, unsigned long nr_pages,
2409 unsigned int gup_flags, struct page **pages,
2412 static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
2415 struct vm_area_struct **vmap)
2418 struct vm_area_struct *vma;
2419 int got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
2422 return ERR_PTR(got);
2426 vma = vma_lookup(mm, addr);
2427 if (WARN_ON_ONCE(!vma)) {
2429 return ERR_PTR(-EINVAL);
2436 long get_user_pages(unsigned long start, unsigned long nr_pages,
2437 unsigned int gup_flags, struct page **pages);
2438 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2439 unsigned int gup_flags, struct page **pages);
2440 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2441 struct page **pages, unsigned int gup_flags);
2442 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2443 struct page **pages, unsigned int gup_flags);
2445 int get_user_pages_fast(unsigned long start, int nr_pages,
2446 unsigned int gup_flags, struct page **pages);
2447 int pin_user_pages_fast(unsigned long start, int nr_pages,
2448 unsigned int gup_flags, struct page **pages);
2449 void folio_add_pin(struct folio *folio);
2451 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2452 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2453 struct task_struct *task, bool bypass_rlim);
2456 struct page *get_dump_page(unsigned long addr);
2458 bool folio_mark_dirty(struct folio *folio);
2459 bool set_page_dirty(struct page *page);
2460 int set_page_dirty_lock(struct page *page);
2462 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2464 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2465 unsigned long old_addr, struct vm_area_struct *new_vma,
2466 unsigned long new_addr, unsigned long len,
2467 bool need_rmap_locks);
2470 * Flags used by change_protection(). For now we make it a bitmap so
2471 * that we can pass in multiple flags just like parameters. However
2472 * for now all the callers are only use one of the flags at the same
2476 * Whether we should manually check if we can map individual PTEs writable,
2477 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2478 * PTEs automatically in a writable mapping.
2480 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2481 /* Whether this protection change is for NUMA hints */
2482 #define MM_CP_PROT_NUMA (1UL << 1)
2483 /* Whether this change is for write protecting */
2484 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2485 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2486 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2487 MM_CP_UFFD_WP_RESOLVE)
2489 bool vma_needs_dirty_tracking(struct vm_area_struct *vma);
2490 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2491 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2494 * We want to check manually if we can change individual PTEs writable
2495 * if we can't do that automatically for all PTEs in a mapping. For
2496 * private mappings, that's always the case when we have write
2497 * permissions as we properly have to handle COW.
2499 if (vma->vm_flags & VM_SHARED)
2500 return vma_wants_writenotify(vma, vma->vm_page_prot);
2501 return !!(vma->vm_flags & VM_WRITE);
2504 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2506 extern long change_protection(struct mmu_gather *tlb,
2507 struct vm_area_struct *vma, unsigned long start,
2508 unsigned long end, unsigned long cp_flags);
2509 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2510 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2511 unsigned long start, unsigned long end, unsigned long newflags);
2514 * doesn't attempt to fault and will return short.
2516 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2517 unsigned int gup_flags, struct page **pages);
2519 static inline bool get_user_page_fast_only(unsigned long addr,
2520 unsigned int gup_flags, struct page **pagep)
2522 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2525 * per-process(per-mm_struct) statistics.
2527 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2529 return percpu_counter_read_positive(&mm->rss_stat[member]);
2532 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2534 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2536 percpu_counter_add(&mm->rss_stat[member], value);
2538 mm_trace_rss_stat(mm, member);
2541 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2543 percpu_counter_inc(&mm->rss_stat[member]);
2545 mm_trace_rss_stat(mm, member);
2548 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2550 percpu_counter_dec(&mm->rss_stat[member]);
2552 mm_trace_rss_stat(mm, member);
2555 /* Optimized variant when page is already known not to be PageAnon */
2556 static inline int mm_counter_file(struct page *page)
2558 if (PageSwapBacked(page))
2559 return MM_SHMEMPAGES;
2560 return MM_FILEPAGES;
2563 static inline int mm_counter(struct page *page)
2566 return MM_ANONPAGES;
2567 return mm_counter_file(page);
2570 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2572 return get_mm_counter(mm, MM_FILEPAGES) +
2573 get_mm_counter(mm, MM_ANONPAGES) +
2574 get_mm_counter(mm, MM_SHMEMPAGES);
2577 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2579 return max(mm->hiwater_rss, get_mm_rss(mm));
2582 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2584 return max(mm->hiwater_vm, mm->total_vm);
2587 static inline void update_hiwater_rss(struct mm_struct *mm)
2589 unsigned long _rss = get_mm_rss(mm);
2591 if ((mm)->hiwater_rss < _rss)
2592 (mm)->hiwater_rss = _rss;
2595 static inline void update_hiwater_vm(struct mm_struct *mm)
2597 if (mm->hiwater_vm < mm->total_vm)
2598 mm->hiwater_vm = mm->total_vm;
2601 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2603 mm->hiwater_rss = get_mm_rss(mm);
2606 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2607 struct mm_struct *mm)
2609 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2611 if (*maxrss < hiwater_rss)
2612 *maxrss = hiwater_rss;
2615 #if defined(SPLIT_RSS_COUNTING)
2616 void sync_mm_rss(struct mm_struct *mm);
2618 static inline void sync_mm_rss(struct mm_struct *mm)
2623 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2624 static inline int pte_special(pte_t pte)
2629 static inline pte_t pte_mkspecial(pte_t pte)
2635 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2636 static inline int pte_devmap(pte_t pte)
2642 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2644 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2648 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2652 #ifdef __PAGETABLE_P4D_FOLDED
2653 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2654 unsigned long address)
2659 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2662 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2663 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2664 unsigned long address)
2668 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2669 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2672 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2674 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2676 if (mm_pud_folded(mm))
2678 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2681 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2683 if (mm_pud_folded(mm))
2685 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2689 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2690 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2691 unsigned long address)
2696 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2697 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2700 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2702 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2704 if (mm_pmd_folded(mm))
2706 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2709 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2711 if (mm_pmd_folded(mm))
2713 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2718 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2720 atomic_long_set(&mm->pgtables_bytes, 0);
2723 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2725 return atomic_long_read(&mm->pgtables_bytes);
2728 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2730 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2733 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2735 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2739 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2740 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2745 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2746 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2749 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2750 int __pte_alloc_kernel(pmd_t *pmd);
2752 #if defined(CONFIG_MMU)
2754 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2755 unsigned long address)
2757 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2758 NULL : p4d_offset(pgd, address);
2761 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2762 unsigned long address)
2764 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2765 NULL : pud_offset(p4d, address);
2768 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2770 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2771 NULL: pmd_offset(pud, address);
2773 #endif /* CONFIG_MMU */
2775 static inline struct ptdesc *virt_to_ptdesc(const void *x)
2777 return page_ptdesc(virt_to_page(x));
2780 static inline void *ptdesc_to_virt(const struct ptdesc *pt)
2782 return page_to_virt(ptdesc_page(pt));
2785 static inline void *ptdesc_address(const struct ptdesc *pt)
2787 return folio_address(ptdesc_folio(pt));
2790 static inline bool pagetable_is_reserved(struct ptdesc *pt)
2792 return folio_test_reserved(ptdesc_folio(pt));
2796 * pagetable_alloc - Allocate pagetables
2798 * @order: desired pagetable order
2800 * pagetable_alloc allocates memory for page tables as well as a page table
2801 * descriptor to describe that memory.
2803 * Return: The ptdesc describing the allocated page tables.
2805 static inline struct ptdesc *pagetable_alloc(gfp_t gfp, unsigned int order)
2807 struct page *page = alloc_pages(gfp | __GFP_COMP, order);
2809 return page_ptdesc(page);
2813 * pagetable_free - Free pagetables
2814 * @pt: The page table descriptor
2816 * pagetable_free frees the memory of all page tables described by a page
2817 * table descriptor and the memory for the descriptor itself.
2819 static inline void pagetable_free(struct ptdesc *pt)
2821 struct page *page = ptdesc_page(pt);
2823 __free_pages(page, compound_order(page));
2826 #if USE_SPLIT_PTE_PTLOCKS
2827 #if ALLOC_SPLIT_PTLOCKS
2828 void __init ptlock_cache_init(void);
2829 bool ptlock_alloc(struct ptdesc *ptdesc);
2830 void ptlock_free(struct ptdesc *ptdesc);
2832 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2836 #else /* ALLOC_SPLIT_PTLOCKS */
2837 static inline void ptlock_cache_init(void)
2841 static inline bool ptlock_alloc(struct ptdesc *ptdesc)
2846 static inline void ptlock_free(struct ptdesc *ptdesc)
2850 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2852 return &ptdesc->ptl;
2854 #endif /* ALLOC_SPLIT_PTLOCKS */
2856 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2858 return ptlock_ptr(page_ptdesc(pmd_page(*pmd)));
2861 static inline bool ptlock_init(struct ptdesc *ptdesc)
2864 * prep_new_page() initialize page->private (and therefore page->ptl)
2865 * with 0. Make sure nobody took it in use in between.
2867 * It can happen if arch try to use slab for page table allocation:
2868 * slab code uses page->slab_cache, which share storage with page->ptl.
2870 VM_BUG_ON_PAGE(*(unsigned long *)&ptdesc->ptl, ptdesc_page(ptdesc));
2871 if (!ptlock_alloc(ptdesc))
2873 spin_lock_init(ptlock_ptr(ptdesc));
2877 #else /* !USE_SPLIT_PTE_PTLOCKS */
2879 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2881 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2883 return &mm->page_table_lock;
2885 static inline void ptlock_cache_init(void) {}
2886 static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
2887 static inline void ptlock_free(struct ptdesc *ptdesc) {}
2888 #endif /* USE_SPLIT_PTE_PTLOCKS */
2890 static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
2892 struct folio *folio = ptdesc_folio(ptdesc);
2894 if (!ptlock_init(ptdesc))
2896 __folio_set_pgtable(folio);
2897 lruvec_stat_add_folio(folio, NR_PAGETABLE);
2901 static inline bool pgtable_pte_page_ctor(struct page *page)
2903 return pagetable_pte_ctor(page_ptdesc(page));
2906 static inline void pagetable_pte_dtor(struct ptdesc *ptdesc)
2908 struct folio *folio = ptdesc_folio(ptdesc);
2910 ptlock_free(ptdesc);
2911 __folio_clear_pgtable(folio);
2912 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
2915 static inline void pgtable_pte_page_dtor(struct page *page)
2917 pagetable_pte_dtor(page_ptdesc(page));
2920 pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
2921 static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
2923 return __pte_offset_map(pmd, addr, NULL);
2926 pte_t *__pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2927 unsigned long addr, spinlock_t **ptlp);
2928 static inline pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2929 unsigned long addr, spinlock_t **ptlp)
2933 __cond_lock(*ptlp, pte = __pte_offset_map_lock(mm, pmd, addr, ptlp));
2937 pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
2938 unsigned long addr, spinlock_t **ptlp);
2940 #define pte_unmap_unlock(pte, ptl) do { \
2945 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2947 #define pte_alloc_map(mm, pmd, address) \
2948 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2950 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2951 (pte_alloc(mm, pmd) ? \
2952 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2954 #define pte_alloc_kernel(pmd, address) \
2955 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2956 NULL: pte_offset_kernel(pmd, address))
2958 #if USE_SPLIT_PMD_PTLOCKS
2960 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2962 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2963 return virt_to_page((void *)((unsigned long) pmd & mask));
2966 static inline struct ptdesc *pmd_ptdesc(pmd_t *pmd)
2968 return page_ptdesc(pmd_pgtable_page(pmd));
2971 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2973 return ptlock_ptr(pmd_ptdesc(pmd));
2976 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc)
2978 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2979 ptdesc->pmd_huge_pte = NULL;
2981 return ptlock_init(ptdesc);
2984 static inline void pmd_ptlock_free(struct ptdesc *ptdesc)
2986 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2987 VM_BUG_ON_PAGE(ptdesc->pmd_huge_pte, ptdesc_page(ptdesc));
2989 ptlock_free(ptdesc);
2992 #define pmd_huge_pte(mm, pmd) (pmd_ptdesc(pmd)->pmd_huge_pte)
2996 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2998 return &mm->page_table_lock;
3001 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc) { return true; }
3002 static inline void pmd_ptlock_free(struct ptdesc *ptdesc) {}
3004 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
3008 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
3010 spinlock_t *ptl = pmd_lockptr(mm, pmd);
3015 static inline bool pagetable_pmd_ctor(struct ptdesc *ptdesc)
3017 struct folio *folio = ptdesc_folio(ptdesc);
3019 if (!pmd_ptlock_init(ptdesc))
3021 __folio_set_pgtable(folio);
3022 lruvec_stat_add_folio(folio, NR_PAGETABLE);
3026 static inline bool pgtable_pmd_page_ctor(struct page *page)
3028 return pagetable_pmd_ctor(page_ptdesc(page));
3031 static inline void pagetable_pmd_dtor(struct ptdesc *ptdesc)
3033 struct folio *folio = ptdesc_folio(ptdesc);
3035 pmd_ptlock_free(ptdesc);
3036 __folio_clear_pgtable(folio);
3037 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3040 static inline void pgtable_pmd_page_dtor(struct page *page)
3042 pagetable_pmd_dtor(page_ptdesc(page));
3046 * No scalability reason to split PUD locks yet, but follow the same pattern
3047 * as the PMD locks to make it easier if we decide to. The VM should not be
3048 * considered ready to switch to split PUD locks yet; there may be places
3049 * which need to be converted from page_table_lock.
3051 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
3053 return &mm->page_table_lock;
3056 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
3058 spinlock_t *ptl = pud_lockptr(mm, pud);
3064 extern void __init pagecache_init(void);
3065 extern void free_initmem(void);
3068 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
3069 * into the buddy system. The freed pages will be poisoned with pattern
3070 * "poison" if it's within range [0, UCHAR_MAX].
3071 * Return pages freed into the buddy system.
3073 extern unsigned long free_reserved_area(void *start, void *end,
3074 int poison, const char *s);
3076 extern void adjust_managed_page_count(struct page *page, long count);
3078 extern void reserve_bootmem_region(phys_addr_t start,
3079 phys_addr_t end, int nid);
3081 /* Free the reserved page into the buddy system, so it gets managed. */
3082 static inline void free_reserved_page(struct page *page)
3084 ClearPageReserved(page);
3085 init_page_count(page);
3087 adjust_managed_page_count(page, 1);
3089 #define free_highmem_page(page) free_reserved_page(page)
3091 static inline void mark_page_reserved(struct page *page)
3093 SetPageReserved(page);
3094 adjust_managed_page_count(page, -1);
3097 static inline void free_reserved_ptdesc(struct ptdesc *pt)
3099 free_reserved_page(ptdesc_page(pt));
3103 * Default method to free all the __init memory into the buddy system.
3104 * The freed pages will be poisoned with pattern "poison" if it's within
3105 * range [0, UCHAR_MAX].
3106 * Return pages freed into the buddy system.
3108 static inline unsigned long free_initmem_default(int poison)
3110 extern char __init_begin[], __init_end[];
3112 return free_reserved_area(&__init_begin, &__init_end,
3113 poison, "unused kernel image (initmem)");
3116 static inline unsigned long get_num_physpages(void)
3119 unsigned long phys_pages = 0;
3121 for_each_online_node(nid)
3122 phys_pages += node_present_pages(nid);
3128 * Using memblock node mappings, an architecture may initialise its
3129 * zones, allocate the backing mem_map and account for memory holes in an
3130 * architecture independent manner.
3132 * An architecture is expected to register range of page frames backed by
3133 * physical memory with memblock_add[_node]() before calling
3134 * free_area_init() passing in the PFN each zone ends at. At a basic
3135 * usage, an architecture is expected to do something like
3137 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3139 * for_each_valid_physical_page_range()
3140 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3141 * free_area_init(max_zone_pfns);
3143 void free_area_init(unsigned long *max_zone_pfn);
3144 unsigned long node_map_pfn_alignment(void);
3145 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
3146 unsigned long end_pfn);
3147 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3148 unsigned long end_pfn);
3149 extern void get_pfn_range_for_nid(unsigned int nid,
3150 unsigned long *start_pfn, unsigned long *end_pfn);
3153 static inline int early_pfn_to_nid(unsigned long pfn)
3158 /* please see mm/page_alloc.c */
3159 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3162 extern void set_dma_reserve(unsigned long new_dma_reserve);
3163 extern void mem_init(void);
3164 extern void __init mmap_init(void);
3166 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
3167 static inline void show_mem(void)
3169 __show_mem(0, NULL, MAX_NR_ZONES - 1);
3171 extern long si_mem_available(void);
3172 extern void si_meminfo(struct sysinfo * val);
3173 extern void si_meminfo_node(struct sysinfo *val, int nid);
3174 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3175 extern unsigned long arch_reserved_kernel_pages(void);
3178 extern __printf(3, 4)
3179 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3181 extern void setup_per_cpu_pageset(void);
3184 extern atomic_long_t mmap_pages_allocated;
3185 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3187 /* interval_tree.c */
3188 void vma_interval_tree_insert(struct vm_area_struct *node,
3189 struct rb_root_cached *root);
3190 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3191 struct vm_area_struct *prev,
3192 struct rb_root_cached *root);
3193 void vma_interval_tree_remove(struct vm_area_struct *node,
3194 struct rb_root_cached *root);
3195 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3196 unsigned long start, unsigned long last);
3197 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3198 unsigned long start, unsigned long last);
3200 #define vma_interval_tree_foreach(vma, root, start, last) \
3201 for (vma = vma_interval_tree_iter_first(root, start, last); \
3202 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3204 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3205 struct rb_root_cached *root);
3206 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3207 struct rb_root_cached *root);
3208 struct anon_vma_chain *
3209 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3210 unsigned long start, unsigned long last);
3211 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3212 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3213 #ifdef CONFIG_DEBUG_VM_RB
3214 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3217 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3218 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3219 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3222 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3223 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3224 unsigned long start, unsigned long end, pgoff_t pgoff,
3225 struct vm_area_struct *next);
3226 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3227 unsigned long start, unsigned long end, pgoff_t pgoff);
3228 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3229 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3230 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3231 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3232 struct anon_vma_name *);
3233 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3234 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3235 unsigned long addr, int new_below);
3236 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3237 unsigned long addr, int new_below);
3238 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3239 extern void unlink_file_vma(struct vm_area_struct *);
3240 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3241 unsigned long addr, unsigned long len, pgoff_t pgoff,
3242 bool *need_rmap_locks);
3243 extern void exit_mmap(struct mm_struct *);
3245 static inline int check_data_rlimit(unsigned long rlim,
3247 unsigned long start,
3248 unsigned long end_data,
3249 unsigned long start_data)
3251 if (rlim < RLIM_INFINITY) {
3252 if (((new - start) + (end_data - start_data)) > rlim)
3259 extern int mm_take_all_locks(struct mm_struct *mm);
3260 extern void mm_drop_all_locks(struct mm_struct *mm);
3262 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3263 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3264 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3265 extern struct file *get_task_exe_file(struct task_struct *task);
3267 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3268 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3270 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3271 const struct vm_special_mapping *sm);
3272 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3273 unsigned long addr, unsigned long len,
3274 unsigned long flags,
3275 const struct vm_special_mapping *spec);
3276 /* This is an obsolete alternative to _install_special_mapping. */
3277 extern int install_special_mapping(struct mm_struct *mm,
3278 unsigned long addr, unsigned long len,
3279 unsigned long flags, struct page **pages);
3281 unsigned long randomize_stack_top(unsigned long stack_top);
3282 unsigned long randomize_page(unsigned long start, unsigned long range);
3284 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3286 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3287 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3288 struct list_head *uf);
3289 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3290 unsigned long len, unsigned long prot, unsigned long flags,
3291 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3292 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3293 unsigned long start, size_t len, struct list_head *uf,
3295 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3296 struct list_head *uf);
3297 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3300 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3301 unsigned long start, unsigned long end,
3302 struct list_head *uf, bool unlock);
3303 extern int __mm_populate(unsigned long addr, unsigned long len,
3305 static inline void mm_populate(unsigned long addr, unsigned long len)
3308 (void) __mm_populate(addr, len, 1);
3311 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3314 /* These take the mm semaphore themselves */
3315 extern int __must_check vm_brk(unsigned long, unsigned long);
3316 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3317 extern int vm_munmap(unsigned long, size_t);
3318 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3319 unsigned long, unsigned long,
3320 unsigned long, unsigned long);
3322 struct vm_unmapped_area_info {
3323 #define VM_UNMAPPED_AREA_TOPDOWN 1
3324 unsigned long flags;
3325 unsigned long length;
3326 unsigned long low_limit;
3327 unsigned long high_limit;
3328 unsigned long align_mask;
3329 unsigned long align_offset;
3332 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3335 extern void truncate_inode_pages(struct address_space *, loff_t);
3336 extern void truncate_inode_pages_range(struct address_space *,
3337 loff_t lstart, loff_t lend);
3338 extern void truncate_inode_pages_final(struct address_space *);
3340 /* generic vm_area_ops exported for stackable file systems */
3341 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3342 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3343 pgoff_t start_pgoff, pgoff_t end_pgoff);
3344 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3346 extern unsigned long stack_guard_gap;
3347 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3348 int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
3349 struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
3351 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3352 int expand_downwards(struct vm_area_struct *vma, unsigned long address);
3354 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3355 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3356 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3357 struct vm_area_struct **pprev);
3360 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3361 * NULL if none. Assume start_addr < end_addr.
3363 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3364 unsigned long start_addr, unsigned long end_addr);
3367 * vma_lookup() - Find a VMA at a specific address
3368 * @mm: The process address space.
3369 * @addr: The user address.
3371 * Return: The vm_area_struct at the given address, %NULL otherwise.
3374 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3376 return mtree_load(&mm->mm_mt, addr);
3379 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3381 unsigned long vm_start = vma->vm_start;
3383 if (vma->vm_flags & VM_GROWSDOWN) {
3384 vm_start -= stack_guard_gap;
3385 if (vm_start > vma->vm_start)
3391 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3393 unsigned long vm_end = vma->vm_end;
3395 if (vma->vm_flags & VM_GROWSUP) {
3396 vm_end += stack_guard_gap;
3397 if (vm_end < vma->vm_end)
3398 vm_end = -PAGE_SIZE;
3403 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3405 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3408 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3409 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3410 unsigned long vm_start, unsigned long vm_end)
3412 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3414 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3420 static inline bool range_in_vma(struct vm_area_struct *vma,
3421 unsigned long start, unsigned long end)
3423 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3427 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3428 void vma_set_page_prot(struct vm_area_struct *vma);
3430 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3434 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3436 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3440 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3442 #ifdef CONFIG_NUMA_BALANCING
3443 unsigned long change_prot_numa(struct vm_area_struct *vma,
3444 unsigned long start, unsigned long end);
3447 struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
3448 unsigned long addr);
3449 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3450 unsigned long pfn, unsigned long size, pgprot_t);
3451 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3452 unsigned long pfn, unsigned long size, pgprot_t prot);
3453 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3454 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3455 struct page **pages, unsigned long *num);
3456 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3458 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3460 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3462 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3463 unsigned long pfn, pgprot_t pgprot);
3464 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3466 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3467 unsigned long addr, pfn_t pfn);
3468 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3470 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3471 unsigned long addr, struct page *page)
3473 int err = vm_insert_page(vma, addr, page);
3476 return VM_FAULT_OOM;
3477 if (err < 0 && err != -EBUSY)
3478 return VM_FAULT_SIGBUS;
3480 return VM_FAULT_NOPAGE;
3483 #ifndef io_remap_pfn_range
3484 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3485 unsigned long addr, unsigned long pfn,
3486 unsigned long size, pgprot_t prot)
3488 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3492 static inline vm_fault_t vmf_error(int err)
3495 return VM_FAULT_OOM;
3496 else if (err == -EHWPOISON)
3497 return VM_FAULT_HWPOISON;
3498 return VM_FAULT_SIGBUS;
3501 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3502 unsigned int foll_flags);
3504 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3506 if (vm_fault & VM_FAULT_OOM)
3508 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3509 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3510 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3516 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3517 * a (NUMA hinting) fault is required.
3519 static inline bool gup_can_follow_protnone(unsigned int flags)
3522 * FOLL_FORCE has to be able to make progress even if the VMA is
3523 * inaccessible. Further, FOLL_FORCE access usually does not represent
3524 * application behaviour and we should avoid triggering NUMA hinting
3527 return flags & FOLL_FORCE;
3530 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3531 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3532 unsigned long size, pte_fn_t fn, void *data);
3533 extern int apply_to_existing_page_range(struct mm_struct *mm,
3534 unsigned long address, unsigned long size,
3535 pte_fn_t fn, void *data);
3537 #ifdef CONFIG_PAGE_POISONING
3538 extern void __kernel_poison_pages(struct page *page, int numpages);
3539 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3540 extern bool _page_poisoning_enabled_early;
3541 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3542 static inline bool page_poisoning_enabled(void)
3544 return _page_poisoning_enabled_early;
3547 * For use in fast paths after init_mem_debugging() has run, or when a
3548 * false negative result is not harmful when called too early.
3550 static inline bool page_poisoning_enabled_static(void)
3552 return static_branch_unlikely(&_page_poisoning_enabled);
3554 static inline void kernel_poison_pages(struct page *page, int numpages)
3556 if (page_poisoning_enabled_static())
3557 __kernel_poison_pages(page, numpages);
3559 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3561 if (page_poisoning_enabled_static())
3562 __kernel_unpoison_pages(page, numpages);
3565 static inline bool page_poisoning_enabled(void) { return false; }
3566 static inline bool page_poisoning_enabled_static(void) { return false; }
3567 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3568 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3569 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3572 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3573 static inline bool want_init_on_alloc(gfp_t flags)
3575 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3578 return flags & __GFP_ZERO;
3581 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3582 static inline bool want_init_on_free(void)
3584 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3588 extern bool _debug_pagealloc_enabled_early;
3589 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3591 static inline bool debug_pagealloc_enabled(void)
3593 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3594 _debug_pagealloc_enabled_early;
3598 * For use in fast paths after mem_debugging_and_hardening_init() has run,
3599 * or when a false negative result is not harmful when called too early.
3601 static inline bool debug_pagealloc_enabled_static(void)
3603 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3606 return static_branch_unlikely(&_debug_pagealloc_enabled);
3610 * To support DEBUG_PAGEALLOC architecture must ensure that
3611 * __kernel_map_pages() never fails
3613 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3614 #ifdef CONFIG_DEBUG_PAGEALLOC
3615 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3617 if (debug_pagealloc_enabled_static())
3618 __kernel_map_pages(page, numpages, 1);
3621 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3623 if (debug_pagealloc_enabled_static())
3624 __kernel_map_pages(page, numpages, 0);
3627 extern unsigned int _debug_guardpage_minorder;
3628 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3630 static inline unsigned int debug_guardpage_minorder(void)
3632 return _debug_guardpage_minorder;
3635 static inline bool debug_guardpage_enabled(void)
3637 return static_branch_unlikely(&_debug_guardpage_enabled);
3640 static inline bool page_is_guard(struct page *page)
3642 if (!debug_guardpage_enabled())
3645 return PageGuard(page);
3648 bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order,
3650 static inline bool set_page_guard(struct zone *zone, struct page *page,
3651 unsigned int order, int migratetype)
3653 if (!debug_guardpage_enabled())
3655 return __set_page_guard(zone, page, order, migratetype);
3658 void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order,
3660 static inline void clear_page_guard(struct zone *zone, struct page *page,
3661 unsigned int order, int migratetype)
3663 if (!debug_guardpage_enabled())
3665 __clear_page_guard(zone, page, order, migratetype);
3668 #else /* CONFIG_DEBUG_PAGEALLOC */
3669 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3670 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3671 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3672 static inline bool debug_guardpage_enabled(void) { return false; }
3673 static inline bool page_is_guard(struct page *page) { return false; }
3674 static inline bool set_page_guard(struct zone *zone, struct page *page,
3675 unsigned int order, int migratetype) { return false; }
3676 static inline void clear_page_guard(struct zone *zone, struct page *page,
3677 unsigned int order, int migratetype) {}
3678 #endif /* CONFIG_DEBUG_PAGEALLOC */
3680 #ifdef __HAVE_ARCH_GATE_AREA
3681 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3682 extern int in_gate_area_no_mm(unsigned long addr);
3683 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3685 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3689 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3690 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3694 #endif /* __HAVE_ARCH_GATE_AREA */
3696 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3698 #ifdef CONFIG_SYSCTL
3699 extern int sysctl_drop_caches;
3700 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3704 void drop_slab(void);
3707 #define randomize_va_space 0
3709 extern int randomize_va_space;
3712 const char * arch_vma_name(struct vm_area_struct *vma);
3714 void print_vma_addr(char *prefix, unsigned long rip);
3716 static inline void print_vma_addr(char *prefix, unsigned long rip)
3721 void *sparse_buffer_alloc(unsigned long size);
3722 struct page * __populate_section_memmap(unsigned long pfn,
3723 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3724 struct dev_pagemap *pgmap);
3725 void pmd_init(void *addr);
3726 void pud_init(void *addr);
3727 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3728 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3729 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3730 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3731 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3732 struct vmem_altmap *altmap, struct page *reuse);
3733 void *vmemmap_alloc_block(unsigned long size, int node);
3735 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3736 struct vmem_altmap *altmap);
3737 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3738 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3739 unsigned long addr, unsigned long next);
3740 int vmemmap_check_pmd(pmd_t *pmd, int node,
3741 unsigned long addr, unsigned long next);
3742 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3743 int node, struct vmem_altmap *altmap);
3744 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3745 int node, struct vmem_altmap *altmap);
3746 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3747 struct vmem_altmap *altmap);
3748 void vmemmap_populate_print_last(void);
3749 #ifdef CONFIG_MEMORY_HOTPLUG
3750 void vmemmap_free(unsigned long start, unsigned long end,
3751 struct vmem_altmap *altmap);
3754 #define VMEMMAP_RESERVE_NR 2
3755 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
3756 static inline bool __vmemmap_can_optimize(struct vmem_altmap *altmap,
3757 struct dev_pagemap *pgmap)
3759 unsigned long nr_pages;
3760 unsigned long nr_vmemmap_pages;
3762 if (!pgmap || !is_power_of_2(sizeof(struct page)))
3765 nr_pages = pgmap_vmemmap_nr(pgmap);
3766 nr_vmemmap_pages = ((nr_pages * sizeof(struct page)) >> PAGE_SHIFT);
3768 * For vmemmap optimization with DAX we need minimum 2 vmemmap
3769 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
3771 return !altmap && (nr_vmemmap_pages > VMEMMAP_RESERVE_NR);
3774 * If we don't have an architecture override, use the generic rule
3776 #ifndef vmemmap_can_optimize
3777 #define vmemmap_can_optimize __vmemmap_can_optimize
3781 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3782 struct dev_pagemap *pgmap)
3788 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3789 unsigned long nr_pages);
3792 MF_COUNT_INCREASED = 1 << 0,
3793 MF_ACTION_REQUIRED = 1 << 1,
3794 MF_MUST_KILL = 1 << 2,
3795 MF_SOFT_OFFLINE = 1 << 3,
3796 MF_UNPOISON = 1 << 4,
3797 MF_SW_SIMULATED = 1 << 5,
3798 MF_NO_RETRY = 1 << 6,
3800 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3801 unsigned long count, int mf_flags);
3802 extern int memory_failure(unsigned long pfn, int flags);
3803 extern void memory_failure_queue_kick(int cpu);
3804 extern int unpoison_memory(unsigned long pfn);
3805 extern void shake_page(struct page *p);
3806 extern atomic_long_t num_poisoned_pages __read_mostly;
3807 extern int soft_offline_page(unsigned long pfn, int flags);
3808 #ifdef CONFIG_MEMORY_FAILURE
3810 * Sysfs entries for memory failure handling statistics.
3812 extern const struct attribute_group memory_failure_attr_group;
3813 extern void memory_failure_queue(unsigned long pfn, int flags);
3814 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3815 bool *migratable_cleared);
3816 void num_poisoned_pages_inc(unsigned long pfn);
3817 void num_poisoned_pages_sub(unsigned long pfn, long i);
3818 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3820 static inline void memory_failure_queue(unsigned long pfn, int flags)
3824 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3825 bool *migratable_cleared)
3830 static inline void num_poisoned_pages_inc(unsigned long pfn)
3834 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3839 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3840 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3841 struct vm_area_struct *vma, struct list_head *to_kill,
3842 unsigned long ksm_addr);
3845 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3846 extern void memblk_nr_poison_inc(unsigned long pfn);
3847 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3849 static inline void memblk_nr_poison_inc(unsigned long pfn)
3853 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3858 #ifndef arch_memory_failure
3859 static inline int arch_memory_failure(unsigned long pfn, int flags)
3865 #ifndef arch_is_platform_page
3866 static inline bool arch_is_platform_page(u64 paddr)
3873 * Error handlers for various types of pages.
3876 MF_IGNORED, /* Error: cannot be handled */
3877 MF_FAILED, /* Error: handling failed */
3878 MF_DELAYED, /* Will be handled later */
3879 MF_RECOVERED, /* Successfully recovered */
3882 enum mf_action_page_type {
3884 MF_MSG_KERNEL_HIGH_ORDER,
3886 MF_MSG_DIFFERENT_COMPOUND,
3889 MF_MSG_UNMAP_FAILED,
3890 MF_MSG_DIRTY_SWAPCACHE,
3891 MF_MSG_CLEAN_SWAPCACHE,
3892 MF_MSG_DIRTY_MLOCKED_LRU,
3893 MF_MSG_CLEAN_MLOCKED_LRU,
3894 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3895 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3898 MF_MSG_TRUNCATED_LRU,
3905 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3906 extern void clear_huge_page(struct page *page,
3907 unsigned long addr_hint,
3908 unsigned int pages_per_huge_page);
3909 int copy_user_large_folio(struct folio *dst, struct folio *src,
3910 unsigned long addr_hint,
3911 struct vm_area_struct *vma);
3912 long copy_folio_from_user(struct folio *dst_folio,
3913 const void __user *usr_src,
3914 bool allow_pagefault);
3917 * vma_is_special_huge - Are transhuge page-table entries considered special?
3918 * @vma: Pointer to the struct vm_area_struct to consider
3920 * Whether transhuge page-table entries are considered "special" following
3921 * the definition in vm_normal_page().
3923 * Return: true if transhuge page-table entries should be considered special,
3926 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3928 return vma_is_dax(vma) || (vma->vm_file &&
3929 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3932 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3934 #if MAX_NUMNODES > 1
3935 void __init setup_nr_node_ids(void);
3937 static inline void setup_nr_node_ids(void) {}
3940 extern int memcmp_pages(struct page *page1, struct page *page2);
3942 static inline int pages_identical(struct page *page1, struct page *page2)
3944 return !memcmp_pages(page1, page2);
3947 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3948 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3949 pgoff_t first_index, pgoff_t nr,
3950 pgoff_t bitmap_pgoff,
3951 unsigned long *bitmap,
3955 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3956 pgoff_t first_index, pgoff_t nr);
3959 extern int sysctl_nr_trim_pages;
3961 #ifdef CONFIG_PRINTK
3962 void mem_dump_obj(void *object);
3964 static inline void mem_dump_obj(void *object) {}
3968 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3969 * @seals: the seals to check
3970 * @vma: the vma to operate on
3972 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3973 * the vma flags. Return 0 if check pass, or <0 for errors.
3975 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3977 if (seals & F_SEAL_FUTURE_WRITE) {
3979 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3980 * "future write" seal active.
3982 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3986 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3987 * MAP_SHARED and read-only, take care to not allow mprotect to
3988 * revert protections on such mappings. Do this only for shared
3989 * mappings. For private mappings, don't need to mask
3990 * VM_MAYWRITE as we still want them to be COW-writable.
3992 if (vma->vm_flags & VM_SHARED)
3993 vm_flags_clear(vma, VM_MAYWRITE);
3999 #ifdef CONFIG_ANON_VMA_NAME
4000 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4001 unsigned long len_in,
4002 struct anon_vma_name *anon_name);
4005 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4006 unsigned long len_in, struct anon_vma_name *anon_name) {
4011 #ifdef CONFIG_UNACCEPTED_MEMORY
4013 bool range_contains_unaccepted_memory(phys_addr_t start, phys_addr_t end);
4014 void accept_memory(phys_addr_t start, phys_addr_t end);
4018 static inline bool range_contains_unaccepted_memory(phys_addr_t start,
4024 static inline void accept_memory(phys_addr_t start, phys_addr_t end)
4030 #endif /* _LINUX_MM_H */