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)
644 /* Check before locking. A race might cause false locked result. */
645 if (vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))
648 if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
652 * Overflow might produce false locked result.
653 * False unlocked result is impossible because we modify and check
654 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
655 * modification invalidates all existing locks.
657 if (unlikely(vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))) {
658 up_read(&vma->vm_lock->lock);
664 static inline void vma_end_read(struct vm_area_struct *vma)
666 rcu_read_lock(); /* keeps vma alive till the end of up_read */
667 up_read(&vma->vm_lock->lock);
671 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
673 mmap_assert_write_locked(vma->vm_mm);
676 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
677 * mm->mm_lock_seq can't be concurrently modified.
679 *mm_lock_seq = READ_ONCE(vma->vm_mm->mm_lock_seq);
680 return (vma->vm_lock_seq == *mm_lock_seq);
683 static inline void vma_start_write(struct vm_area_struct *vma)
687 if (__is_vma_write_locked(vma, &mm_lock_seq))
690 down_write(&vma->vm_lock->lock);
691 vma->vm_lock_seq = mm_lock_seq;
692 up_write(&vma->vm_lock->lock);
695 static inline bool vma_try_start_write(struct vm_area_struct *vma)
699 if (__is_vma_write_locked(vma, &mm_lock_seq))
702 if (!down_write_trylock(&vma->vm_lock->lock))
705 vma->vm_lock_seq = mm_lock_seq;
706 up_write(&vma->vm_lock->lock);
710 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
714 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
717 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
719 /* When detaching vma should be write-locked */
721 vma_assert_write_locked(vma);
722 vma->detached = detached;
725 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
726 unsigned long address);
728 #else /* CONFIG_PER_VMA_LOCK */
730 static inline bool vma_start_read(struct vm_area_struct *vma)
732 static inline void vma_end_read(struct vm_area_struct *vma) {}
733 static inline void vma_start_write(struct vm_area_struct *vma) {}
734 static inline bool vma_try_start_write(struct vm_area_struct *vma)
736 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
737 static inline void vma_mark_detached(struct vm_area_struct *vma,
740 #endif /* CONFIG_PER_VMA_LOCK */
743 * WARNING: vma_init does not initialize vma->vm_lock.
744 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
746 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
748 static const struct vm_operations_struct dummy_vm_ops = {};
750 memset(vma, 0, sizeof(*vma));
752 vma->vm_ops = &dummy_vm_ops;
753 INIT_LIST_HEAD(&vma->anon_vma_chain);
754 vma_mark_detached(vma, false);
755 vma_numab_state_init(vma);
758 /* Use when VMA is not part of the VMA tree and needs no locking */
759 static inline void vm_flags_init(struct vm_area_struct *vma,
762 ACCESS_PRIVATE(vma, __vm_flags) = flags;
765 /* Use when VMA is part of the VMA tree and modifications need coordination */
766 static inline void vm_flags_reset(struct vm_area_struct *vma,
769 vma_start_write(vma);
770 vm_flags_init(vma, flags);
773 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
776 vma_start_write(vma);
777 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
780 static inline void vm_flags_set(struct vm_area_struct *vma,
783 vma_start_write(vma);
784 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
787 static inline void vm_flags_clear(struct vm_area_struct *vma,
790 vma_start_write(vma);
791 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
795 * Use only if VMA is not part of the VMA tree or has no other users and
796 * therefore needs no locking.
798 static inline void __vm_flags_mod(struct vm_area_struct *vma,
799 vm_flags_t set, vm_flags_t clear)
801 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
805 * Use only when the order of set/clear operations is unimportant, otherwise
806 * use vm_flags_{set|clear} explicitly.
808 static inline void vm_flags_mod(struct vm_area_struct *vma,
809 vm_flags_t set, vm_flags_t clear)
811 vma_start_write(vma);
812 __vm_flags_mod(vma, set, clear);
815 static inline void vma_set_anonymous(struct vm_area_struct *vma)
820 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
825 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
827 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
832 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
833 VM_STACK_INCOMPLETE_SETUP)
839 static inline bool vma_is_foreign(struct vm_area_struct *vma)
844 if (current->mm != vma->vm_mm)
850 static inline bool vma_is_accessible(struct vm_area_struct *vma)
852 return vma->vm_flags & VM_ACCESS_FLAGS;
856 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
858 return mas_find(&vmi->mas, max - 1);
861 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
864 * Uses mas_find() to get the first VMA when the iterator starts.
865 * Calling mas_next() could skip the first entry.
867 return mas_find(&vmi->mas, ULONG_MAX);
871 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
873 return mas_next_range(&vmi->mas, ULONG_MAX);
877 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
879 return mas_prev(&vmi->mas, 0);
883 struct vm_area_struct *vma_iter_prev_range(struct vma_iterator *vmi)
885 return mas_prev_range(&vmi->mas, 0);
888 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
890 return vmi->mas.index;
893 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
895 return vmi->mas.last + 1;
897 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
900 return mas_expected_entries(&vmi->mas, count);
903 /* Free any unused preallocations */
904 static inline void vma_iter_free(struct vma_iterator *vmi)
906 mas_destroy(&vmi->mas);
909 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
910 struct vm_area_struct *vma)
912 vmi->mas.index = vma->vm_start;
913 vmi->mas.last = vma->vm_end - 1;
914 mas_store(&vmi->mas, vma);
915 if (unlikely(mas_is_err(&vmi->mas)))
921 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
923 mas_pause(&vmi->mas);
926 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
928 mas_set(&vmi->mas, addr);
931 #define for_each_vma(__vmi, __vma) \
932 while (((__vma) = vma_next(&(__vmi))) != NULL)
934 /* The MM code likes to work with exclusive end addresses */
935 #define for_each_vma_range(__vmi, __vma, __end) \
936 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
940 * The vma_is_shmem is not inline because it is used only by slow
941 * paths in userfault.
943 bool vma_is_shmem(struct vm_area_struct *vma);
944 bool vma_is_anon_shmem(struct vm_area_struct *vma);
946 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
947 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
950 int vma_is_stack_for_current(struct vm_area_struct *vma);
952 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
953 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
959 * compound_order() can be called without holding a reference, which means
960 * that niceties like page_folio() don't work. These callers should be
961 * prepared to handle wild return values. For example, PG_head may be
962 * set before _folio_order is initialised, or this may be a tail page.
963 * See compaction.c for some good examples.
965 static inline unsigned int compound_order(struct page *page)
967 struct folio *folio = (struct folio *)page;
969 if (!test_bit(PG_head, &folio->flags))
971 return folio->_folio_order;
975 * folio_order - The allocation order of a folio.
978 * A folio is composed of 2^order pages. See get_order() for the definition
981 * Return: The order of the folio.
983 static inline unsigned int folio_order(struct folio *folio)
985 if (!folio_test_large(folio))
987 return folio->_folio_order;
990 #include <linux/huge_mm.h>
993 * Methods to modify the page usage count.
995 * What counts for a page usage:
996 * - cache mapping (page->mapping)
997 * - private data (page->private)
998 * - page mapped in a task's page tables, each mapping
999 * is counted separately
1001 * Also, many kernel routines increase the page count before a critical
1002 * routine so they can be sure the page doesn't go away from under them.
1006 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1008 static inline int put_page_testzero(struct page *page)
1010 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1011 return page_ref_dec_and_test(page);
1014 static inline int folio_put_testzero(struct folio *folio)
1016 return put_page_testzero(&folio->page);
1020 * Try to grab a ref unless the page has a refcount of zero, return false if
1022 * This can be called when MMU is off so it must not access
1023 * any of the virtual mappings.
1025 static inline bool get_page_unless_zero(struct page *page)
1027 return page_ref_add_unless(page, 1, 0);
1030 static inline struct folio *folio_get_nontail_page(struct page *page)
1032 if (unlikely(!get_page_unless_zero(page)))
1034 return (struct folio *)page;
1037 extern int page_is_ram(unsigned long pfn);
1045 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1046 unsigned long desc);
1048 /* Support for virtually mapped pages */
1049 struct page *vmalloc_to_page(const void *addr);
1050 unsigned long vmalloc_to_pfn(const void *addr);
1053 * Determine if an address is within the vmalloc range
1055 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1056 * is no special casing required.
1059 #ifndef is_ioremap_addr
1060 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1064 extern bool is_vmalloc_addr(const void *x);
1065 extern int is_vmalloc_or_module_addr(const void *x);
1067 static inline bool is_vmalloc_addr(const void *x)
1071 static inline int is_vmalloc_or_module_addr(const void *x)
1078 * How many times the entire folio is mapped as a single unit (eg by a
1079 * PMD or PUD entry). This is probably not what you want, except for
1080 * debugging purposes - it does not include PTE-mapped sub-pages; look
1081 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1083 static inline int folio_entire_mapcount(struct folio *folio)
1085 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1086 return atomic_read(&folio->_entire_mapcount) + 1;
1090 * The atomic page->_mapcount, starts from -1: so that transitions
1091 * both from it and to it can be tracked, using atomic_inc_and_test
1092 * and atomic_add_negative(-1).
1094 static inline void page_mapcount_reset(struct page *page)
1096 atomic_set(&(page)->_mapcount, -1);
1100 * page_mapcount() - Number of times this precise page is mapped.
1103 * The number of times this page is mapped. If this page is part of
1104 * a large folio, it includes the number of times this page is mapped
1105 * as part of that folio.
1107 * The result is undefined for pages which cannot be mapped into userspace.
1108 * For example SLAB or special types of pages. See function page_has_type().
1109 * They use this field in struct page differently.
1111 static inline int page_mapcount(struct page *page)
1113 int mapcount = atomic_read(&page->_mapcount) + 1;
1115 if (unlikely(PageCompound(page)))
1116 mapcount += folio_entire_mapcount(page_folio(page));
1121 int folio_total_mapcount(struct folio *folio);
1124 * folio_mapcount() - Calculate the number of mappings of this folio.
1125 * @folio: The folio.
1127 * A large folio tracks both how many times the entire folio is mapped,
1128 * and how many times each individual page in the folio is mapped.
1129 * This function calculates the total number of times the folio is
1132 * Return: The number of times this folio is mapped.
1134 static inline int folio_mapcount(struct folio *folio)
1136 if (likely(!folio_test_large(folio)))
1137 return atomic_read(&folio->_mapcount) + 1;
1138 return folio_total_mapcount(folio);
1141 static inline int total_mapcount(struct page *page)
1143 if (likely(!PageCompound(page)))
1144 return atomic_read(&page->_mapcount) + 1;
1145 return folio_total_mapcount(page_folio(page));
1148 static inline bool folio_large_is_mapped(struct folio *folio)
1151 * Reading _entire_mapcount below could be omitted if hugetlb
1152 * participated in incrementing nr_pages_mapped when compound mapped.
1154 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1155 atomic_read(&folio->_entire_mapcount) >= 0;
1159 * folio_mapped - Is this folio mapped into userspace?
1160 * @folio: The folio.
1162 * Return: True if any page in this folio is referenced by user page tables.
1164 static inline bool folio_mapped(struct folio *folio)
1166 if (likely(!folio_test_large(folio)))
1167 return atomic_read(&folio->_mapcount) >= 0;
1168 return folio_large_is_mapped(folio);
1172 * Return true if this page is mapped into pagetables.
1173 * For compound page it returns true if any sub-page of compound page is mapped,
1174 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1176 static inline bool page_mapped(struct page *page)
1178 if (likely(!PageCompound(page)))
1179 return atomic_read(&page->_mapcount) >= 0;
1180 return folio_large_is_mapped(page_folio(page));
1183 static inline struct page *virt_to_head_page(const void *x)
1185 struct page *page = virt_to_page(x);
1187 return compound_head(page);
1190 static inline struct folio *virt_to_folio(const void *x)
1192 struct page *page = virt_to_page(x);
1194 return page_folio(page);
1197 void __folio_put(struct folio *folio);
1199 void put_pages_list(struct list_head *pages);
1201 void split_page(struct page *page, unsigned int order);
1202 void folio_copy(struct folio *dst, struct folio *src);
1204 unsigned long nr_free_buffer_pages(void);
1207 * Compound pages have a destructor function. Provide a
1208 * prototype for that function and accessor functions.
1209 * These are _only_ valid on the head of a compound page.
1211 typedef void compound_page_dtor(struct page *);
1213 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1214 enum compound_dtor_id {
1217 #ifdef CONFIG_HUGETLB_PAGE
1220 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1221 TRANSHUGE_PAGE_DTOR,
1226 static inline void folio_set_compound_dtor(struct folio *folio,
1227 enum compound_dtor_id compound_dtor)
1229 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1230 folio->_folio_dtor = compound_dtor;
1233 void destroy_large_folio(struct folio *folio);
1235 /* Returns the number of bytes in this potentially compound page. */
1236 static inline unsigned long page_size(struct page *page)
1238 return PAGE_SIZE << compound_order(page);
1241 /* Returns the number of bits needed for the number of bytes in a page */
1242 static inline unsigned int page_shift(struct page *page)
1244 return PAGE_SHIFT + compound_order(page);
1248 * thp_order - Order of a transparent huge page.
1249 * @page: Head page of a transparent huge page.
1251 static inline unsigned int thp_order(struct page *page)
1253 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1254 return compound_order(page);
1258 * thp_size - Size of a transparent huge page.
1259 * @page: Head page of a transparent huge page.
1261 * Return: Number of bytes in this page.
1263 static inline unsigned long thp_size(struct page *page)
1265 return PAGE_SIZE << thp_order(page);
1268 void free_compound_page(struct page *page);
1272 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1273 * servicing faults for write access. In the normal case, do always want
1274 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1275 * that do not have writing enabled, when used by access_process_vm.
1277 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1279 if (likely(vma->vm_flags & VM_WRITE))
1280 pte = pte_mkwrite(pte);
1284 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1285 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1287 vm_fault_t finish_fault(struct vm_fault *vmf);
1288 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1292 * Multiple processes may "see" the same page. E.g. for untouched
1293 * mappings of /dev/null, all processes see the same page full of
1294 * zeroes, and text pages of executables and shared libraries have
1295 * only one copy in memory, at most, normally.
1297 * For the non-reserved pages, page_count(page) denotes a reference count.
1298 * page_count() == 0 means the page is free. page->lru is then used for
1299 * freelist management in the buddy allocator.
1300 * page_count() > 0 means the page has been allocated.
1302 * Pages are allocated by the slab allocator in order to provide memory
1303 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1304 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1305 * unless a particular usage is carefully commented. (the responsibility of
1306 * freeing the kmalloc memory is the caller's, of course).
1308 * A page may be used by anyone else who does a __get_free_page().
1309 * In this case, page_count still tracks the references, and should only
1310 * be used through the normal accessor functions. The top bits of page->flags
1311 * and page->virtual store page management information, but all other fields
1312 * are unused and could be used privately, carefully. The management of this
1313 * page is the responsibility of the one who allocated it, and those who have
1314 * subsequently been given references to it.
1316 * The other pages (we may call them "pagecache pages") are completely
1317 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1318 * The following discussion applies only to them.
1320 * A pagecache page contains an opaque `private' member, which belongs to the
1321 * page's address_space. Usually, this is the address of a circular list of
1322 * the page's disk buffers. PG_private must be set to tell the VM to call
1323 * into the filesystem to release these pages.
1325 * A page may belong to an inode's memory mapping. In this case, page->mapping
1326 * is the pointer to the inode, and page->index is the file offset of the page,
1327 * in units of PAGE_SIZE.
1329 * If pagecache pages are not associated with an inode, they are said to be
1330 * anonymous pages. These may become associated with the swapcache, and in that
1331 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1333 * In either case (swapcache or inode backed), the pagecache itself holds one
1334 * reference to the page. Setting PG_private should also increment the
1335 * refcount. The each user mapping also has a reference to the page.
1337 * The pagecache pages are stored in a per-mapping radix tree, which is
1338 * rooted at mapping->i_pages, and indexed by offset.
1339 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1340 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1342 * All pagecache pages may be subject to I/O:
1343 * - inode pages may need to be read from disk,
1344 * - inode pages which have been modified and are MAP_SHARED may need
1345 * to be written back to the inode on disk,
1346 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1347 * modified may need to be swapped out to swap space and (later) to be read
1351 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1352 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1354 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1355 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1357 if (!static_branch_unlikely(&devmap_managed_key))
1359 if (!is_zone_device_page(page))
1361 return __put_devmap_managed_page_refs(page, refs);
1363 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1364 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1368 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1370 static inline bool put_devmap_managed_page(struct page *page)
1372 return put_devmap_managed_page_refs(page, 1);
1375 /* 127: arbitrary random number, small enough to assemble well */
1376 #define folio_ref_zero_or_close_to_overflow(folio) \
1377 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1380 * folio_get - Increment the reference count on a folio.
1381 * @folio: The folio.
1383 * Context: May be called in any context, as long as you know that
1384 * you have a refcount on the folio. If you do not already have one,
1385 * folio_try_get() may be the right interface for you to use.
1387 static inline void folio_get(struct folio *folio)
1389 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1390 folio_ref_inc(folio);
1393 static inline void get_page(struct page *page)
1395 folio_get(page_folio(page));
1398 static inline __must_check bool try_get_page(struct page *page)
1400 page = compound_head(page);
1401 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1408 * folio_put - Decrement the reference count on a folio.
1409 * @folio: The folio.
1411 * If the folio's reference count reaches zero, the memory will be
1412 * released back to the page allocator and may be used by another
1413 * allocation immediately. Do not access the memory or the struct folio
1414 * after calling folio_put() unless you can be sure that it wasn't the
1417 * Context: May be called in process or interrupt context, but not in NMI
1418 * context. May be called while holding a spinlock.
1420 static inline void folio_put(struct folio *folio)
1422 if (folio_put_testzero(folio))
1427 * folio_put_refs - Reduce the reference count on a folio.
1428 * @folio: The folio.
1429 * @refs: The amount to subtract from the folio's reference count.
1431 * If the folio's reference count reaches zero, the memory will be
1432 * released back to the page allocator and may be used by another
1433 * allocation immediately. Do not access the memory or the struct folio
1434 * after calling folio_put_refs() unless you can be sure that these weren't
1435 * the last references.
1437 * Context: May be called in process or interrupt context, but not in NMI
1438 * context. May be called while holding a spinlock.
1440 static inline void folio_put_refs(struct folio *folio, int refs)
1442 if (folio_ref_sub_and_test(folio, refs))
1447 * union release_pages_arg - an array of pages or folios
1449 * release_pages() releases a simple array of multiple pages, and
1450 * accepts various different forms of said page array: either
1451 * a regular old boring array of pages, an array of folios, or
1452 * an array of encoded page pointers.
1454 * The transparent union syntax for this kind of "any of these
1455 * argument types" is all kinds of ugly, so look away.
1458 struct page **pages;
1459 struct folio **folios;
1460 struct encoded_page **encoded_pages;
1461 } release_pages_arg __attribute__ ((__transparent_union__));
1463 void release_pages(release_pages_arg, int nr);
1466 * folios_put - Decrement the reference count on an array of folios.
1467 * @folios: The folios.
1468 * @nr: How many folios there are.
1470 * Like folio_put(), but for an array of folios. This is more efficient
1471 * than writing the loop yourself as it will optimise the locks which
1472 * need to be taken if the folios are freed.
1474 * Context: May be called in process or interrupt context, but not in NMI
1475 * context. May be called while holding a spinlock.
1477 static inline void folios_put(struct folio **folios, unsigned int nr)
1479 release_pages(folios, nr);
1482 static inline void put_page(struct page *page)
1484 struct folio *folio = page_folio(page);
1487 * For some devmap managed pages we need to catch refcount transition
1490 if (put_devmap_managed_page(&folio->page))
1496 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1497 * the page's refcount so that two separate items are tracked: the original page
1498 * reference count, and also a new count of how many pin_user_pages() calls were
1499 * made against the page. ("gup-pinned" is another term for the latter).
1501 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1502 * distinct from normal pages. As such, the unpin_user_page() call (and its
1503 * variants) must be used in order to release gup-pinned pages.
1507 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1508 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1509 * simpler, due to the fact that adding an even power of two to the page
1510 * refcount has the effect of using only the upper N bits, for the code that
1511 * counts up using the bias value. This means that the lower bits are left for
1512 * the exclusive use of the original code that increments and decrements by one
1513 * (or at least, by much smaller values than the bias value).
1515 * Of course, once the lower bits overflow into the upper bits (and this is
1516 * OK, because subtraction recovers the original values), then visual inspection
1517 * no longer suffices to directly view the separate counts. However, for normal
1518 * applications that don't have huge page reference counts, this won't be an
1521 * Locking: the lockless algorithm described in folio_try_get_rcu()
1522 * provides safe operation for get_user_pages(), page_mkclean() and
1523 * other calls that race to set up page table entries.
1525 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1527 void unpin_user_page(struct page *page);
1528 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1530 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1532 void unpin_user_pages(struct page **pages, unsigned long npages);
1534 static inline bool is_cow_mapping(vm_flags_t flags)
1536 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1540 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1543 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1544 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1545 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1546 * underlying memory if ptrace is active, so this is only possible if
1547 * ptrace does not apply. Note that there is no mprotect() to upgrade
1548 * write permissions later.
1550 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1554 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1555 #define SECTION_IN_PAGE_FLAGS
1559 * The identification function is mainly used by the buddy allocator for
1560 * determining if two pages could be buddies. We are not really identifying
1561 * the zone since we could be using the section number id if we do not have
1562 * node id available in page flags.
1563 * We only guarantee that it will return the same value for two combinable
1566 static inline int page_zone_id(struct page *page)
1568 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1571 #ifdef NODE_NOT_IN_PAGE_FLAGS
1572 extern int page_to_nid(const struct page *page);
1574 static inline int page_to_nid(const struct page *page)
1576 struct page *p = (struct page *)page;
1578 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1582 static inline int folio_nid(const struct folio *folio)
1584 return page_to_nid(&folio->page);
1587 #ifdef CONFIG_NUMA_BALANCING
1588 /* page access time bits needs to hold at least 4 seconds */
1589 #define PAGE_ACCESS_TIME_MIN_BITS 12
1590 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1591 #define PAGE_ACCESS_TIME_BUCKETS \
1592 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1594 #define PAGE_ACCESS_TIME_BUCKETS 0
1597 #define PAGE_ACCESS_TIME_MASK \
1598 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1600 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1602 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1605 static inline int cpupid_to_pid(int cpupid)
1607 return cpupid & LAST__PID_MASK;
1610 static inline int cpupid_to_cpu(int cpupid)
1612 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1615 static inline int cpupid_to_nid(int cpupid)
1617 return cpu_to_node(cpupid_to_cpu(cpupid));
1620 static inline bool cpupid_pid_unset(int cpupid)
1622 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1625 static inline bool cpupid_cpu_unset(int cpupid)
1627 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1630 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1632 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1635 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1636 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1637 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1639 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1642 static inline int page_cpupid_last(struct page *page)
1644 return page->_last_cpupid;
1646 static inline void page_cpupid_reset_last(struct page *page)
1648 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1651 static inline int page_cpupid_last(struct page *page)
1653 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1656 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1658 static inline void page_cpupid_reset_last(struct page *page)
1660 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1662 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1664 static inline int xchg_page_access_time(struct page *page, int time)
1668 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1669 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1672 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1674 unsigned int pid_bit;
1676 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1677 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->access_pids[1])) {
1678 __set_bit(pid_bit, &vma->numab_state->access_pids[1]);
1681 #else /* !CONFIG_NUMA_BALANCING */
1682 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1684 return page_to_nid(page); /* XXX */
1687 static inline int xchg_page_access_time(struct page *page, int time)
1692 static inline int page_cpupid_last(struct page *page)
1694 return page_to_nid(page); /* XXX */
1697 static inline int cpupid_to_nid(int cpupid)
1702 static inline int cpupid_to_pid(int cpupid)
1707 static inline int cpupid_to_cpu(int cpupid)
1712 static inline int cpu_pid_to_cpupid(int nid, int pid)
1717 static inline bool cpupid_pid_unset(int cpupid)
1722 static inline void page_cpupid_reset_last(struct page *page)
1726 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1731 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1734 #endif /* CONFIG_NUMA_BALANCING */
1736 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1739 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1740 * setting tags for all pages to native kernel tag value 0xff, as the default
1741 * value 0x00 maps to 0xff.
1744 static inline u8 page_kasan_tag(const struct page *page)
1748 if (kasan_enabled()) {
1749 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1756 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1758 unsigned long old_flags, flags;
1760 if (!kasan_enabled())
1764 old_flags = READ_ONCE(page->flags);
1767 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1768 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1769 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1772 static inline void page_kasan_tag_reset(struct page *page)
1774 if (kasan_enabled())
1775 page_kasan_tag_set(page, 0xff);
1778 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1780 static inline u8 page_kasan_tag(const struct page *page)
1785 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1786 static inline void page_kasan_tag_reset(struct page *page) { }
1788 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1790 static inline struct zone *page_zone(const struct page *page)
1792 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1795 static inline pg_data_t *page_pgdat(const struct page *page)
1797 return NODE_DATA(page_to_nid(page));
1800 static inline struct zone *folio_zone(const struct folio *folio)
1802 return page_zone(&folio->page);
1805 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1807 return page_pgdat(&folio->page);
1810 #ifdef SECTION_IN_PAGE_FLAGS
1811 static inline void set_page_section(struct page *page, unsigned long section)
1813 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1814 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1817 static inline unsigned long page_to_section(const struct page *page)
1819 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1824 * folio_pfn - Return the Page Frame Number of a folio.
1825 * @folio: The folio.
1827 * A folio may contain multiple pages. The pages have consecutive
1828 * Page Frame Numbers.
1830 * Return: The Page Frame Number of the first page in the folio.
1832 static inline unsigned long folio_pfn(struct folio *folio)
1834 return page_to_pfn(&folio->page);
1837 static inline struct folio *pfn_folio(unsigned long pfn)
1839 return page_folio(pfn_to_page(pfn));
1843 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1844 * @folio: The folio.
1846 * This function checks if a folio has been pinned via a call to
1847 * a function in the pin_user_pages() family.
1849 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1850 * because it means "definitely not pinned for DMA", but true means "probably
1851 * pinned for DMA, but possibly a false positive due to having at least
1852 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1854 * False positives are OK, because: a) it's unlikely for a folio to
1855 * get that many refcounts, and b) all the callers of this routine are
1856 * expected to be able to deal gracefully with a false positive.
1858 * For large folios, the result will be exactly correct. That's because
1859 * we have more tracking data available: the _pincount field is used
1860 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1862 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1864 * Return: True, if it is likely that the page has been "dma-pinned".
1865 * False, if the page is definitely not dma-pinned.
1867 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1869 if (folio_test_large(folio))
1870 return atomic_read(&folio->_pincount) > 0;
1873 * folio_ref_count() is signed. If that refcount overflows, then
1874 * folio_ref_count() returns a negative value, and callers will avoid
1875 * further incrementing the refcount.
1877 * Here, for that overflow case, use the sign bit to count a little
1878 * bit higher via unsigned math, and thus still get an accurate result.
1880 return ((unsigned int)folio_ref_count(folio)) >=
1881 GUP_PIN_COUNTING_BIAS;
1884 static inline bool page_maybe_dma_pinned(struct page *page)
1886 return folio_maybe_dma_pinned(page_folio(page));
1890 * This should most likely only be called during fork() to see whether we
1891 * should break the cow immediately for an anon page on the src mm.
1893 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1895 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1898 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1900 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1903 return page_maybe_dma_pinned(page);
1907 * is_zero_page - Query if a page is a zero page
1908 * @page: The page to query
1910 * This returns true if @page is one of the permanent zero pages.
1912 static inline bool is_zero_page(const struct page *page)
1914 return is_zero_pfn(page_to_pfn(page));
1918 * is_zero_folio - Query if a folio is a zero page
1919 * @folio: The folio to query
1921 * This returns true if @folio is one of the permanent zero pages.
1923 static inline bool is_zero_folio(const struct folio *folio)
1925 return is_zero_page(&folio->page);
1928 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1929 #ifdef CONFIG_MIGRATION
1930 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1933 int mt = folio_migratetype(folio);
1935 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1938 /* The zero page can be "pinned" but gets special handling. */
1939 if (is_zero_folio(folio))
1942 /* Coherent device memory must always allow eviction. */
1943 if (folio_is_device_coherent(folio))
1946 /* Otherwise, non-movable zone folios can be pinned. */
1947 return !folio_is_zone_movable(folio);
1951 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1957 static inline void set_page_zone(struct page *page, enum zone_type zone)
1959 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1960 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1963 static inline void set_page_node(struct page *page, unsigned long node)
1965 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1966 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1969 static inline void set_page_links(struct page *page, enum zone_type zone,
1970 unsigned long node, unsigned long pfn)
1972 set_page_zone(page, zone);
1973 set_page_node(page, node);
1974 #ifdef SECTION_IN_PAGE_FLAGS
1975 set_page_section(page, pfn_to_section_nr(pfn));
1980 * folio_nr_pages - The number of pages in the folio.
1981 * @folio: The folio.
1983 * Return: A positive power of two.
1985 static inline long folio_nr_pages(struct folio *folio)
1987 if (!folio_test_large(folio))
1990 return folio->_folio_nr_pages;
1992 return 1L << folio->_folio_order;
1997 * compound_nr() returns the number of pages in this potentially compound
1998 * page. compound_nr() can be called on a tail page, and is defined to
1999 * return 1 in that case.
2001 static inline unsigned long compound_nr(struct page *page)
2003 struct folio *folio = (struct folio *)page;
2005 if (!test_bit(PG_head, &folio->flags))
2008 return folio->_folio_nr_pages;
2010 return 1L << folio->_folio_order;
2015 * thp_nr_pages - The number of regular pages in this huge page.
2016 * @page: The head page of a huge page.
2018 static inline int thp_nr_pages(struct page *page)
2020 return folio_nr_pages((struct folio *)page);
2024 * folio_next - Move to the next physical folio.
2025 * @folio: The folio we're currently operating on.
2027 * If you have physically contiguous memory which may span more than
2028 * one folio (eg a &struct bio_vec), use this function to move from one
2029 * folio to the next. Do not use it if the memory is only virtually
2030 * contiguous as the folios are almost certainly not adjacent to each
2031 * other. This is the folio equivalent to writing ``page++``.
2033 * Context: We assume that the folios are refcounted and/or locked at a
2034 * higher level and do not adjust the reference counts.
2035 * Return: The next struct folio.
2037 static inline struct folio *folio_next(struct folio *folio)
2039 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2043 * folio_shift - The size of the memory described by this folio.
2044 * @folio: The folio.
2046 * A folio represents a number of bytes which is a power-of-two in size.
2047 * This function tells you which power-of-two the folio is. See also
2048 * folio_size() and folio_order().
2050 * Context: The caller should have a reference on the folio to prevent
2051 * it from being split. It is not necessary for the folio to be locked.
2052 * Return: The base-2 logarithm of the size of this folio.
2054 static inline unsigned int folio_shift(struct folio *folio)
2056 return PAGE_SHIFT + folio_order(folio);
2060 * folio_size - The number of bytes in a folio.
2061 * @folio: The folio.
2063 * Context: The caller should have a reference on the folio to prevent
2064 * it from being split. It is not necessary for the folio to be locked.
2065 * Return: The number of bytes in this folio.
2067 static inline size_t folio_size(struct folio *folio)
2069 return PAGE_SIZE << folio_order(folio);
2073 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2074 * @folio: The folio.
2076 * folio_estimated_sharers() aims to serve as a function to efficiently
2077 * estimate the number of processes sharing a folio. This is done by
2078 * looking at the precise mapcount of the first subpage in the folio, and
2079 * assuming the other subpages are the same. This may not be true for large
2080 * folios. If you want exact mapcounts for exact calculations, look at
2081 * page_mapcount() or folio_total_mapcount().
2083 * Return: The estimated number of processes sharing a folio.
2085 static inline int folio_estimated_sharers(struct folio *folio)
2087 return page_mapcount(folio_page(folio, 0));
2090 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2091 static inline int arch_make_page_accessible(struct page *page)
2097 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2098 static inline int arch_make_folio_accessible(struct folio *folio)
2101 long i, nr = folio_nr_pages(folio);
2103 for (i = 0; i < nr; i++) {
2104 ret = arch_make_page_accessible(folio_page(folio, i));
2114 * Some inline functions in vmstat.h depend on page_zone()
2116 #include <linux/vmstat.h>
2118 static __always_inline void *lowmem_page_address(const struct page *page)
2120 return page_to_virt(page);
2123 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2124 #define HASHED_PAGE_VIRTUAL
2127 #if defined(WANT_PAGE_VIRTUAL)
2128 static inline void *page_address(const struct page *page)
2130 return page->virtual;
2132 static inline void set_page_address(struct page *page, void *address)
2134 page->virtual = address;
2136 #define page_address_init() do { } while(0)
2139 #if defined(HASHED_PAGE_VIRTUAL)
2140 void *page_address(const struct page *page);
2141 void set_page_address(struct page *page, void *virtual);
2142 void page_address_init(void);
2145 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2146 #define page_address(page) lowmem_page_address(page)
2147 #define set_page_address(page, address) do { } while(0)
2148 #define page_address_init() do { } while(0)
2151 static inline void *folio_address(const struct folio *folio)
2153 return page_address(&folio->page);
2156 extern void *page_rmapping(struct page *page);
2157 extern pgoff_t __page_file_index(struct page *page);
2160 * Return the pagecache index of the passed page. Regular pagecache pages
2161 * use ->index whereas swapcache pages use swp_offset(->private)
2163 static inline pgoff_t page_index(struct page *page)
2165 if (unlikely(PageSwapCache(page)))
2166 return __page_file_index(page);
2171 * Return true only if the page has been allocated with
2172 * ALLOC_NO_WATERMARKS and the low watermark was not
2173 * met implying that the system is under some pressure.
2175 static inline bool page_is_pfmemalloc(const struct page *page)
2178 * lru.next has bit 1 set if the page is allocated from the
2179 * pfmemalloc reserves. Callers may simply overwrite it if
2180 * they do not need to preserve that information.
2182 return (uintptr_t)page->lru.next & BIT(1);
2186 * Return true only if the folio has been allocated with
2187 * ALLOC_NO_WATERMARKS and the low watermark was not
2188 * met implying that the system is under some pressure.
2190 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2193 * lru.next has bit 1 set if the page is allocated from the
2194 * pfmemalloc reserves. Callers may simply overwrite it if
2195 * they do not need to preserve that information.
2197 return (uintptr_t)folio->lru.next & BIT(1);
2201 * Only to be called by the page allocator on a freshly allocated
2204 static inline void set_page_pfmemalloc(struct page *page)
2206 page->lru.next = (void *)BIT(1);
2209 static inline void clear_page_pfmemalloc(struct page *page)
2211 page->lru.next = NULL;
2215 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2217 extern void pagefault_out_of_memory(void);
2219 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2220 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2221 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2224 * Flags passed to show_mem() and show_free_areas() to suppress output in
2227 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2229 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2230 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2232 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2236 * Parameter block passed down to zap_pte_range in exceptional cases.
2238 struct zap_details {
2239 struct folio *single_folio; /* Locked folio to be unmapped */
2240 bool even_cows; /* Zap COWed private pages too? */
2241 zap_flags_t zap_flags; /* Extra flags for zapping */
2245 * Whether to drop the pte markers, for example, the uffd-wp information for
2246 * file-backed memory. This should only be specified when we will completely
2247 * drop the page in the mm, either by truncation or unmapping of the vma. By
2248 * default, the flag is not set.
2250 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2251 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2252 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2254 #ifdef CONFIG_SCHED_MM_CID
2255 void sched_mm_cid_before_execve(struct task_struct *t);
2256 void sched_mm_cid_after_execve(struct task_struct *t);
2257 void sched_mm_cid_fork(struct task_struct *t);
2258 void sched_mm_cid_exit_signals(struct task_struct *t);
2259 static inline int task_mm_cid(struct task_struct *t)
2264 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2265 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2266 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2267 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2268 static inline int task_mm_cid(struct task_struct *t)
2271 * Use the processor id as a fall-back when the mm cid feature is
2272 * disabled. This provides functional per-cpu data structure accesses
2273 * in user-space, althrough it won't provide the memory usage benefits.
2275 return raw_smp_processor_id();
2280 extern bool can_do_mlock(void);
2282 static inline bool can_do_mlock(void) { return false; }
2284 extern int user_shm_lock(size_t, struct ucounts *);
2285 extern void user_shm_unlock(size_t, struct ucounts *);
2287 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2289 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2291 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2294 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2295 unsigned long size);
2296 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2297 unsigned long size, struct zap_details *details);
2298 static inline void zap_vma_pages(struct vm_area_struct *vma)
2300 zap_page_range_single(vma, vma->vm_start,
2301 vma->vm_end - vma->vm_start, NULL);
2303 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2304 struct vm_area_struct *start_vma, unsigned long start,
2305 unsigned long end, bool mm_wr_locked);
2307 struct mmu_notifier_range;
2309 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2310 unsigned long end, unsigned long floor, unsigned long ceiling);
2312 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2313 int follow_pte(struct mm_struct *mm, unsigned long address,
2314 pte_t **ptepp, spinlock_t **ptlp);
2315 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2316 unsigned long *pfn);
2317 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2318 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2319 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2320 void *buf, int len, int write);
2322 extern void truncate_pagecache(struct inode *inode, loff_t new);
2323 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2324 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2325 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2326 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2328 struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
2329 unsigned long address, struct pt_regs *regs);
2332 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2333 unsigned long address, unsigned int flags,
2334 struct pt_regs *regs);
2335 extern int fixup_user_fault(struct mm_struct *mm,
2336 unsigned long address, unsigned int fault_flags,
2338 void unmap_mapping_pages(struct address_space *mapping,
2339 pgoff_t start, pgoff_t nr, bool even_cows);
2340 void unmap_mapping_range(struct address_space *mapping,
2341 loff_t const holebegin, loff_t const holelen, int even_cows);
2343 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2344 unsigned long address, unsigned int flags,
2345 struct pt_regs *regs)
2347 /* should never happen if there's no MMU */
2349 return VM_FAULT_SIGBUS;
2351 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2352 unsigned int fault_flags, bool *unlocked)
2354 /* should never happen if there's no MMU */
2358 static inline void unmap_mapping_pages(struct address_space *mapping,
2359 pgoff_t start, pgoff_t nr, bool even_cows) { }
2360 static inline void unmap_mapping_range(struct address_space *mapping,
2361 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2364 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2365 loff_t const holebegin, loff_t const holelen)
2367 unmap_mapping_range(mapping, holebegin, holelen, 0);
2370 static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
2371 unsigned long addr);
2373 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2374 void *buf, int len, unsigned int gup_flags);
2375 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2376 void *buf, int len, unsigned int gup_flags);
2377 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2378 void *buf, int len, unsigned int gup_flags);
2380 long get_user_pages_remote(struct mm_struct *mm,
2381 unsigned long start, unsigned long nr_pages,
2382 unsigned int gup_flags, struct page **pages,
2384 long pin_user_pages_remote(struct mm_struct *mm,
2385 unsigned long start, unsigned long nr_pages,
2386 unsigned int gup_flags, struct page **pages,
2389 static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
2392 struct vm_area_struct **vmap)
2395 struct vm_area_struct *vma;
2396 int got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
2399 return ERR_PTR(got);
2403 vma = vma_lookup(mm, addr);
2404 if (WARN_ON_ONCE(!vma)) {
2406 return ERR_PTR(-EINVAL);
2413 long get_user_pages(unsigned long start, unsigned long nr_pages,
2414 unsigned int gup_flags, struct page **pages);
2415 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2416 unsigned int gup_flags, struct page **pages);
2417 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2418 struct page **pages, unsigned int gup_flags);
2419 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2420 struct page **pages, unsigned int gup_flags);
2422 int get_user_pages_fast(unsigned long start, int nr_pages,
2423 unsigned int gup_flags, struct page **pages);
2424 int pin_user_pages_fast(unsigned long start, int nr_pages,
2425 unsigned int gup_flags, struct page **pages);
2426 void folio_add_pin(struct folio *folio);
2428 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2429 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2430 struct task_struct *task, bool bypass_rlim);
2433 struct page *get_dump_page(unsigned long addr);
2435 bool folio_mark_dirty(struct folio *folio);
2436 bool set_page_dirty(struct page *page);
2437 int set_page_dirty_lock(struct page *page);
2439 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2441 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2442 unsigned long old_addr, struct vm_area_struct *new_vma,
2443 unsigned long new_addr, unsigned long len,
2444 bool need_rmap_locks);
2447 * Flags used by change_protection(). For now we make it a bitmap so
2448 * that we can pass in multiple flags just like parameters. However
2449 * for now all the callers are only use one of the flags at the same
2453 * Whether we should manually check if we can map individual PTEs writable,
2454 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2455 * PTEs automatically in a writable mapping.
2457 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2458 /* Whether this protection change is for NUMA hints */
2459 #define MM_CP_PROT_NUMA (1UL << 1)
2460 /* Whether this change is for write protecting */
2461 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2462 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2463 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2464 MM_CP_UFFD_WP_RESOLVE)
2466 bool vma_needs_dirty_tracking(struct vm_area_struct *vma);
2467 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2468 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2471 * We want to check manually if we can change individual PTEs writable
2472 * if we can't do that automatically for all PTEs in a mapping. For
2473 * private mappings, that's always the case when we have write
2474 * permissions as we properly have to handle COW.
2476 if (vma->vm_flags & VM_SHARED)
2477 return vma_wants_writenotify(vma, vma->vm_page_prot);
2478 return !!(vma->vm_flags & VM_WRITE);
2481 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2483 extern long change_protection(struct mmu_gather *tlb,
2484 struct vm_area_struct *vma, unsigned long start,
2485 unsigned long end, unsigned long cp_flags);
2486 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2487 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2488 unsigned long start, unsigned long end, unsigned long newflags);
2491 * doesn't attempt to fault and will return short.
2493 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2494 unsigned int gup_flags, struct page **pages);
2496 static inline bool get_user_page_fast_only(unsigned long addr,
2497 unsigned int gup_flags, struct page **pagep)
2499 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2502 * per-process(per-mm_struct) statistics.
2504 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2506 return percpu_counter_read_positive(&mm->rss_stat[member]);
2509 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2511 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2513 percpu_counter_add(&mm->rss_stat[member], value);
2515 mm_trace_rss_stat(mm, member);
2518 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2520 percpu_counter_inc(&mm->rss_stat[member]);
2522 mm_trace_rss_stat(mm, member);
2525 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2527 percpu_counter_dec(&mm->rss_stat[member]);
2529 mm_trace_rss_stat(mm, member);
2532 /* Optimized variant when page is already known not to be PageAnon */
2533 static inline int mm_counter_file(struct page *page)
2535 if (PageSwapBacked(page))
2536 return MM_SHMEMPAGES;
2537 return MM_FILEPAGES;
2540 static inline int mm_counter(struct page *page)
2543 return MM_ANONPAGES;
2544 return mm_counter_file(page);
2547 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2549 return get_mm_counter(mm, MM_FILEPAGES) +
2550 get_mm_counter(mm, MM_ANONPAGES) +
2551 get_mm_counter(mm, MM_SHMEMPAGES);
2554 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2556 return max(mm->hiwater_rss, get_mm_rss(mm));
2559 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2561 return max(mm->hiwater_vm, mm->total_vm);
2564 static inline void update_hiwater_rss(struct mm_struct *mm)
2566 unsigned long _rss = get_mm_rss(mm);
2568 if ((mm)->hiwater_rss < _rss)
2569 (mm)->hiwater_rss = _rss;
2572 static inline void update_hiwater_vm(struct mm_struct *mm)
2574 if (mm->hiwater_vm < mm->total_vm)
2575 mm->hiwater_vm = mm->total_vm;
2578 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2580 mm->hiwater_rss = get_mm_rss(mm);
2583 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2584 struct mm_struct *mm)
2586 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2588 if (*maxrss < hiwater_rss)
2589 *maxrss = hiwater_rss;
2592 #if defined(SPLIT_RSS_COUNTING)
2593 void sync_mm_rss(struct mm_struct *mm);
2595 static inline void sync_mm_rss(struct mm_struct *mm)
2600 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2601 static inline int pte_special(pte_t pte)
2606 static inline pte_t pte_mkspecial(pte_t pte)
2612 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2613 static inline int pte_devmap(pte_t pte)
2619 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2621 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2625 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2629 #ifdef __PAGETABLE_P4D_FOLDED
2630 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2631 unsigned long address)
2636 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2639 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2640 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2641 unsigned long address)
2645 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2646 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2649 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2651 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2653 if (mm_pud_folded(mm))
2655 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2658 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2660 if (mm_pud_folded(mm))
2662 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2666 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2667 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2668 unsigned long address)
2673 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2674 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2677 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2679 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2681 if (mm_pmd_folded(mm))
2683 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2686 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2688 if (mm_pmd_folded(mm))
2690 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2695 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2697 atomic_long_set(&mm->pgtables_bytes, 0);
2700 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2702 return atomic_long_read(&mm->pgtables_bytes);
2705 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2707 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2710 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2712 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2716 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2717 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2722 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2723 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2726 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2727 int __pte_alloc_kernel(pmd_t *pmd);
2729 #if defined(CONFIG_MMU)
2731 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2732 unsigned long address)
2734 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2735 NULL : p4d_offset(pgd, address);
2738 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2739 unsigned long address)
2741 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2742 NULL : pud_offset(p4d, address);
2745 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2747 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2748 NULL: pmd_offset(pud, address);
2750 #endif /* CONFIG_MMU */
2752 #if USE_SPLIT_PTE_PTLOCKS
2753 #if ALLOC_SPLIT_PTLOCKS
2754 void __init ptlock_cache_init(void);
2755 extern bool ptlock_alloc(struct page *page);
2756 extern void ptlock_free(struct page *page);
2758 static inline spinlock_t *ptlock_ptr(struct page *page)
2762 #else /* ALLOC_SPLIT_PTLOCKS */
2763 static inline void ptlock_cache_init(void)
2767 static inline bool ptlock_alloc(struct page *page)
2772 static inline void ptlock_free(struct page *page)
2776 static inline spinlock_t *ptlock_ptr(struct page *page)
2780 #endif /* ALLOC_SPLIT_PTLOCKS */
2782 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2784 return ptlock_ptr(pmd_page(*pmd));
2787 static inline bool ptlock_init(struct page *page)
2790 * prep_new_page() initialize page->private (and therefore page->ptl)
2791 * with 0. Make sure nobody took it in use in between.
2793 * It can happen if arch try to use slab for page table allocation:
2794 * slab code uses page->slab_cache, which share storage with page->ptl.
2796 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2797 if (!ptlock_alloc(page))
2799 spin_lock_init(ptlock_ptr(page));
2803 #else /* !USE_SPLIT_PTE_PTLOCKS */
2805 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2807 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2809 return &mm->page_table_lock;
2811 static inline void ptlock_cache_init(void) {}
2812 static inline bool ptlock_init(struct page *page) { return true; }
2813 static inline void ptlock_free(struct page *page) {}
2814 #endif /* USE_SPLIT_PTE_PTLOCKS */
2816 static inline bool pgtable_pte_page_ctor(struct page *page)
2818 if (!ptlock_init(page))
2820 __SetPageTable(page);
2821 inc_lruvec_page_state(page, NR_PAGETABLE);
2825 static inline void pgtable_pte_page_dtor(struct page *page)
2828 __ClearPageTable(page);
2829 dec_lruvec_page_state(page, NR_PAGETABLE);
2832 pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
2833 static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
2835 return __pte_offset_map(pmd, addr, NULL);
2838 pte_t *__pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2839 unsigned long addr, spinlock_t **ptlp);
2840 static inline pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2841 unsigned long addr, spinlock_t **ptlp)
2845 __cond_lock(*ptlp, pte = __pte_offset_map_lock(mm, pmd, addr, ptlp));
2849 pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
2850 unsigned long addr, spinlock_t **ptlp);
2852 #define pte_unmap_unlock(pte, ptl) do { \
2857 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2859 #define pte_alloc_map(mm, pmd, address) \
2860 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2862 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2863 (pte_alloc(mm, pmd) ? \
2864 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2866 #define pte_alloc_kernel(pmd, address) \
2867 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2868 NULL: pte_offset_kernel(pmd, address))
2870 #if USE_SPLIT_PMD_PTLOCKS
2872 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2874 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2875 return virt_to_page((void *)((unsigned long) pmd & mask));
2878 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2880 return ptlock_ptr(pmd_pgtable_page(pmd));
2883 static inline bool pmd_ptlock_init(struct page *page)
2885 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2886 page->pmd_huge_pte = NULL;
2888 return ptlock_init(page);
2891 static inline void pmd_ptlock_free(struct page *page)
2893 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2894 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2899 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2903 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2905 return &mm->page_table_lock;
2908 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2909 static inline void pmd_ptlock_free(struct page *page) {}
2911 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2915 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2917 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2922 static inline bool pgtable_pmd_page_ctor(struct page *page)
2924 if (!pmd_ptlock_init(page))
2926 __SetPageTable(page);
2927 inc_lruvec_page_state(page, NR_PAGETABLE);
2931 static inline void pgtable_pmd_page_dtor(struct page *page)
2933 pmd_ptlock_free(page);
2934 __ClearPageTable(page);
2935 dec_lruvec_page_state(page, NR_PAGETABLE);
2939 * No scalability reason to split PUD locks yet, but follow the same pattern
2940 * as the PMD locks to make it easier if we decide to. The VM should not be
2941 * considered ready to switch to split PUD locks yet; there may be places
2942 * which need to be converted from page_table_lock.
2944 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2946 return &mm->page_table_lock;
2949 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2951 spinlock_t *ptl = pud_lockptr(mm, pud);
2957 extern void __init pagecache_init(void);
2958 extern void free_initmem(void);
2961 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2962 * into the buddy system. The freed pages will be poisoned with pattern
2963 * "poison" if it's within range [0, UCHAR_MAX].
2964 * Return pages freed into the buddy system.
2966 extern unsigned long free_reserved_area(void *start, void *end,
2967 int poison, const char *s);
2969 extern void adjust_managed_page_count(struct page *page, long count);
2971 extern void reserve_bootmem_region(phys_addr_t start,
2972 phys_addr_t end, int nid);
2974 /* Free the reserved page into the buddy system, so it gets managed. */
2975 static inline void free_reserved_page(struct page *page)
2977 ClearPageReserved(page);
2978 init_page_count(page);
2980 adjust_managed_page_count(page, 1);
2982 #define free_highmem_page(page) free_reserved_page(page)
2984 static inline void mark_page_reserved(struct page *page)
2986 SetPageReserved(page);
2987 adjust_managed_page_count(page, -1);
2991 * Default method to free all the __init memory into the buddy system.
2992 * The freed pages will be poisoned with pattern "poison" if it's within
2993 * range [0, UCHAR_MAX].
2994 * Return pages freed into the buddy system.
2996 static inline unsigned long free_initmem_default(int poison)
2998 extern char __init_begin[], __init_end[];
3000 return free_reserved_area(&__init_begin, &__init_end,
3001 poison, "unused kernel image (initmem)");
3004 static inline unsigned long get_num_physpages(void)
3007 unsigned long phys_pages = 0;
3009 for_each_online_node(nid)
3010 phys_pages += node_present_pages(nid);
3016 * Using memblock node mappings, an architecture may initialise its
3017 * zones, allocate the backing mem_map and account for memory holes in an
3018 * architecture independent manner.
3020 * An architecture is expected to register range of page frames backed by
3021 * physical memory with memblock_add[_node]() before calling
3022 * free_area_init() passing in the PFN each zone ends at. At a basic
3023 * usage, an architecture is expected to do something like
3025 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3027 * for_each_valid_physical_page_range()
3028 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3029 * free_area_init(max_zone_pfns);
3031 void free_area_init(unsigned long *max_zone_pfn);
3032 unsigned long node_map_pfn_alignment(void);
3033 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
3034 unsigned long end_pfn);
3035 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3036 unsigned long end_pfn);
3037 extern void get_pfn_range_for_nid(unsigned int nid,
3038 unsigned long *start_pfn, unsigned long *end_pfn);
3041 static inline int early_pfn_to_nid(unsigned long pfn)
3046 /* please see mm/page_alloc.c */
3047 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3050 extern void set_dma_reserve(unsigned long new_dma_reserve);
3051 extern void mem_init(void);
3052 extern void __init mmap_init(void);
3054 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
3055 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
3057 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
3059 extern long si_mem_available(void);
3060 extern void si_meminfo(struct sysinfo * val);
3061 extern void si_meminfo_node(struct sysinfo *val, int nid);
3062 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3063 extern unsigned long arch_reserved_kernel_pages(void);
3066 extern __printf(3, 4)
3067 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3069 extern void setup_per_cpu_pageset(void);
3072 extern atomic_long_t mmap_pages_allocated;
3073 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3075 /* interval_tree.c */
3076 void vma_interval_tree_insert(struct vm_area_struct *node,
3077 struct rb_root_cached *root);
3078 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3079 struct vm_area_struct *prev,
3080 struct rb_root_cached *root);
3081 void vma_interval_tree_remove(struct vm_area_struct *node,
3082 struct rb_root_cached *root);
3083 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3084 unsigned long start, unsigned long last);
3085 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3086 unsigned long start, unsigned long last);
3088 #define vma_interval_tree_foreach(vma, root, start, last) \
3089 for (vma = vma_interval_tree_iter_first(root, start, last); \
3090 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3092 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3093 struct rb_root_cached *root);
3094 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3095 struct rb_root_cached *root);
3096 struct anon_vma_chain *
3097 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3098 unsigned long start, unsigned long last);
3099 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3100 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3101 #ifdef CONFIG_DEBUG_VM_RB
3102 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3105 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3106 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3107 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3110 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3111 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3112 unsigned long start, unsigned long end, pgoff_t pgoff,
3113 struct vm_area_struct *next);
3114 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3115 unsigned long start, unsigned long end, pgoff_t pgoff);
3116 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3117 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3118 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3119 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3120 struct anon_vma_name *);
3121 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3122 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3123 unsigned long addr, int new_below);
3124 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3125 unsigned long addr, int new_below);
3126 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3127 extern void unlink_file_vma(struct vm_area_struct *);
3128 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3129 unsigned long addr, unsigned long len, pgoff_t pgoff,
3130 bool *need_rmap_locks);
3131 extern void exit_mmap(struct mm_struct *);
3133 static inline int check_data_rlimit(unsigned long rlim,
3135 unsigned long start,
3136 unsigned long end_data,
3137 unsigned long start_data)
3139 if (rlim < RLIM_INFINITY) {
3140 if (((new - start) + (end_data - start_data)) > rlim)
3147 extern int mm_take_all_locks(struct mm_struct *mm);
3148 extern void mm_drop_all_locks(struct mm_struct *mm);
3150 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3151 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3152 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3153 extern struct file *get_task_exe_file(struct task_struct *task);
3155 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3156 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3158 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3159 const struct vm_special_mapping *sm);
3160 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3161 unsigned long addr, unsigned long len,
3162 unsigned long flags,
3163 const struct vm_special_mapping *spec);
3164 /* This is an obsolete alternative to _install_special_mapping. */
3165 extern int install_special_mapping(struct mm_struct *mm,
3166 unsigned long addr, unsigned long len,
3167 unsigned long flags, struct page **pages);
3169 unsigned long randomize_stack_top(unsigned long stack_top);
3170 unsigned long randomize_page(unsigned long start, unsigned long range);
3172 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3174 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3175 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3176 struct list_head *uf);
3177 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3178 unsigned long len, unsigned long prot, unsigned long flags,
3179 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3180 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3181 unsigned long start, size_t len, struct list_head *uf,
3183 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3184 struct list_head *uf);
3185 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3188 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3189 unsigned long start, unsigned long end,
3190 struct list_head *uf, bool unlock);
3191 extern int __mm_populate(unsigned long addr, unsigned long len,
3193 static inline void mm_populate(unsigned long addr, unsigned long len)
3196 (void) __mm_populate(addr, len, 1);
3199 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3202 /* These take the mm semaphore themselves */
3203 extern int __must_check vm_brk(unsigned long, unsigned long);
3204 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3205 extern int vm_munmap(unsigned long, size_t);
3206 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3207 unsigned long, unsigned long,
3208 unsigned long, unsigned long);
3210 struct vm_unmapped_area_info {
3211 #define VM_UNMAPPED_AREA_TOPDOWN 1
3212 unsigned long flags;
3213 unsigned long length;
3214 unsigned long low_limit;
3215 unsigned long high_limit;
3216 unsigned long align_mask;
3217 unsigned long align_offset;
3220 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3223 extern void truncate_inode_pages(struct address_space *, loff_t);
3224 extern void truncate_inode_pages_range(struct address_space *,
3225 loff_t lstart, loff_t lend);
3226 extern void truncate_inode_pages_final(struct address_space *);
3228 /* generic vm_area_ops exported for stackable file systems */
3229 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3230 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3231 pgoff_t start_pgoff, pgoff_t end_pgoff);
3232 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3234 extern unsigned long stack_guard_gap;
3235 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3236 int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
3237 struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
3239 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3240 int expand_downwards(struct vm_area_struct *vma, unsigned long address);
3242 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3243 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3244 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3245 struct vm_area_struct **pprev);
3248 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3249 * NULL if none. Assume start_addr < end_addr.
3251 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3252 unsigned long start_addr, unsigned long end_addr);
3255 * vma_lookup() - Find a VMA at a specific address
3256 * @mm: The process address space.
3257 * @addr: The user address.
3259 * Return: The vm_area_struct at the given address, %NULL otherwise.
3262 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3264 return mtree_load(&mm->mm_mt, addr);
3267 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3269 unsigned long vm_start = vma->vm_start;
3271 if (vma->vm_flags & VM_GROWSDOWN) {
3272 vm_start -= stack_guard_gap;
3273 if (vm_start > vma->vm_start)
3279 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3281 unsigned long vm_end = vma->vm_end;
3283 if (vma->vm_flags & VM_GROWSUP) {
3284 vm_end += stack_guard_gap;
3285 if (vm_end < vma->vm_end)
3286 vm_end = -PAGE_SIZE;
3291 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3293 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3296 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3297 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3298 unsigned long vm_start, unsigned long vm_end)
3300 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3302 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3308 static inline bool range_in_vma(struct vm_area_struct *vma,
3309 unsigned long start, unsigned long end)
3311 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3315 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3316 void vma_set_page_prot(struct vm_area_struct *vma);
3318 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3322 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3324 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3328 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3330 #ifdef CONFIG_NUMA_BALANCING
3331 unsigned long change_prot_numa(struct vm_area_struct *vma,
3332 unsigned long start, unsigned long end);
3335 struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
3336 unsigned long addr);
3337 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3338 unsigned long pfn, unsigned long size, pgprot_t);
3339 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3340 unsigned long pfn, unsigned long size, pgprot_t prot);
3341 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3342 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3343 struct page **pages, unsigned long *num);
3344 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3346 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3348 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3350 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3351 unsigned long pfn, pgprot_t pgprot);
3352 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3354 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3355 unsigned long addr, pfn_t pfn);
3356 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3358 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3359 unsigned long addr, struct page *page)
3361 int err = vm_insert_page(vma, addr, page);
3364 return VM_FAULT_OOM;
3365 if (err < 0 && err != -EBUSY)
3366 return VM_FAULT_SIGBUS;
3368 return VM_FAULT_NOPAGE;
3371 #ifndef io_remap_pfn_range
3372 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3373 unsigned long addr, unsigned long pfn,
3374 unsigned long size, pgprot_t prot)
3376 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3380 static inline vm_fault_t vmf_error(int err)
3383 return VM_FAULT_OOM;
3384 else if (err == -EHWPOISON)
3385 return VM_FAULT_HWPOISON;
3386 return VM_FAULT_SIGBUS;
3389 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3390 unsigned int foll_flags);
3392 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3394 if (vm_fault & VM_FAULT_OOM)
3396 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3397 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3398 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3404 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3405 * a (NUMA hinting) fault is required.
3407 static inline bool gup_can_follow_protnone(unsigned int flags)
3410 * FOLL_FORCE has to be able to make progress even if the VMA is
3411 * inaccessible. Further, FOLL_FORCE access usually does not represent
3412 * application behaviour and we should avoid triggering NUMA hinting
3415 return flags & FOLL_FORCE;
3418 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3419 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3420 unsigned long size, pte_fn_t fn, void *data);
3421 extern int apply_to_existing_page_range(struct mm_struct *mm,
3422 unsigned long address, unsigned long size,
3423 pte_fn_t fn, void *data);
3425 #ifdef CONFIG_PAGE_POISONING
3426 extern void __kernel_poison_pages(struct page *page, int numpages);
3427 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3428 extern bool _page_poisoning_enabled_early;
3429 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3430 static inline bool page_poisoning_enabled(void)
3432 return _page_poisoning_enabled_early;
3435 * For use in fast paths after init_mem_debugging() has run, or when a
3436 * false negative result is not harmful when called too early.
3438 static inline bool page_poisoning_enabled_static(void)
3440 return static_branch_unlikely(&_page_poisoning_enabled);
3442 static inline void kernel_poison_pages(struct page *page, int numpages)
3444 if (page_poisoning_enabled_static())
3445 __kernel_poison_pages(page, numpages);
3447 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3449 if (page_poisoning_enabled_static())
3450 __kernel_unpoison_pages(page, numpages);
3453 static inline bool page_poisoning_enabled(void) { return false; }
3454 static inline bool page_poisoning_enabled_static(void) { return false; }
3455 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3456 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3457 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3460 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3461 static inline bool want_init_on_alloc(gfp_t flags)
3463 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3466 return flags & __GFP_ZERO;
3469 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3470 static inline bool want_init_on_free(void)
3472 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3476 extern bool _debug_pagealloc_enabled_early;
3477 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3479 static inline bool debug_pagealloc_enabled(void)
3481 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3482 _debug_pagealloc_enabled_early;
3486 * For use in fast paths after init_debug_pagealloc() has run, or when a
3487 * false negative result is not harmful when called too early.
3489 static inline bool debug_pagealloc_enabled_static(void)
3491 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3494 return static_branch_unlikely(&_debug_pagealloc_enabled);
3498 * To support DEBUG_PAGEALLOC architecture must ensure that
3499 * __kernel_map_pages() never fails
3501 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3502 #ifdef CONFIG_DEBUG_PAGEALLOC
3503 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3505 if (debug_pagealloc_enabled_static())
3506 __kernel_map_pages(page, numpages, 1);
3509 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3511 if (debug_pagealloc_enabled_static())
3512 __kernel_map_pages(page, numpages, 0);
3515 extern unsigned int _debug_guardpage_minorder;
3516 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3518 static inline unsigned int debug_guardpage_minorder(void)
3520 return _debug_guardpage_minorder;
3523 static inline bool debug_guardpage_enabled(void)
3525 return static_branch_unlikely(&_debug_guardpage_enabled);
3528 static inline bool page_is_guard(struct page *page)
3530 if (!debug_guardpage_enabled())
3533 return PageGuard(page);
3536 bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order,
3538 static inline bool set_page_guard(struct zone *zone, struct page *page,
3539 unsigned int order, int migratetype)
3541 if (!debug_guardpage_enabled())
3543 return __set_page_guard(zone, page, order, migratetype);
3546 void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order,
3548 static inline void clear_page_guard(struct zone *zone, struct page *page,
3549 unsigned int order, int migratetype)
3551 if (!debug_guardpage_enabled())
3553 __clear_page_guard(zone, page, order, migratetype);
3556 #else /* CONFIG_DEBUG_PAGEALLOC */
3557 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3558 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3559 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3560 static inline bool debug_guardpage_enabled(void) { return false; }
3561 static inline bool page_is_guard(struct page *page) { return false; }
3562 static inline bool set_page_guard(struct zone *zone, struct page *page,
3563 unsigned int order, int migratetype) { return false; }
3564 static inline void clear_page_guard(struct zone *zone, struct page *page,
3565 unsigned int order, int migratetype) {}
3566 #endif /* CONFIG_DEBUG_PAGEALLOC */
3568 #ifdef __HAVE_ARCH_GATE_AREA
3569 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3570 extern int in_gate_area_no_mm(unsigned long addr);
3571 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3573 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3577 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3578 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3582 #endif /* __HAVE_ARCH_GATE_AREA */
3584 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3586 #ifdef CONFIG_SYSCTL
3587 extern int sysctl_drop_caches;
3588 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3592 void drop_slab(void);
3595 #define randomize_va_space 0
3597 extern int randomize_va_space;
3600 const char * arch_vma_name(struct vm_area_struct *vma);
3602 void print_vma_addr(char *prefix, unsigned long rip);
3604 static inline void print_vma_addr(char *prefix, unsigned long rip)
3609 void *sparse_buffer_alloc(unsigned long size);
3610 struct page * __populate_section_memmap(unsigned long pfn,
3611 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3612 struct dev_pagemap *pgmap);
3613 void pmd_init(void *addr);
3614 void pud_init(void *addr);
3615 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3616 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3617 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3618 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3619 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3620 struct vmem_altmap *altmap, struct page *reuse);
3621 void *vmemmap_alloc_block(unsigned long size, int node);
3623 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3624 struct vmem_altmap *altmap);
3625 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3626 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3627 unsigned long addr, unsigned long next);
3628 int vmemmap_check_pmd(pmd_t *pmd, int node,
3629 unsigned long addr, unsigned long next);
3630 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3631 int node, struct vmem_altmap *altmap);
3632 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3633 int node, struct vmem_altmap *altmap);
3634 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3635 struct vmem_altmap *altmap);
3636 void vmemmap_populate_print_last(void);
3637 #ifdef CONFIG_MEMORY_HOTPLUG
3638 void vmemmap_free(unsigned long start, unsigned long end,
3639 struct vmem_altmap *altmap);
3642 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP
3643 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3644 struct dev_pagemap *pgmap)
3646 return is_power_of_2(sizeof(struct page)) &&
3647 pgmap && (pgmap_vmemmap_nr(pgmap) > 1) && !altmap;
3650 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3651 struct dev_pagemap *pgmap)
3657 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3658 unsigned long nr_pages);
3661 MF_COUNT_INCREASED = 1 << 0,
3662 MF_ACTION_REQUIRED = 1 << 1,
3663 MF_MUST_KILL = 1 << 2,
3664 MF_SOFT_OFFLINE = 1 << 3,
3665 MF_UNPOISON = 1 << 4,
3666 MF_SW_SIMULATED = 1 << 5,
3667 MF_NO_RETRY = 1 << 6,
3669 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3670 unsigned long count, int mf_flags);
3671 extern int memory_failure(unsigned long pfn, int flags);
3672 extern void memory_failure_queue_kick(int cpu);
3673 extern int unpoison_memory(unsigned long pfn);
3674 extern void shake_page(struct page *p);
3675 extern atomic_long_t num_poisoned_pages __read_mostly;
3676 extern int soft_offline_page(unsigned long pfn, int flags);
3677 #ifdef CONFIG_MEMORY_FAILURE
3679 * Sysfs entries for memory failure handling statistics.
3681 extern const struct attribute_group memory_failure_attr_group;
3682 extern void memory_failure_queue(unsigned long pfn, int flags);
3683 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3684 bool *migratable_cleared);
3685 void num_poisoned_pages_inc(unsigned long pfn);
3686 void num_poisoned_pages_sub(unsigned long pfn, long i);
3687 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3689 static inline void memory_failure_queue(unsigned long pfn, int flags)
3693 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3694 bool *migratable_cleared)
3699 static inline void num_poisoned_pages_inc(unsigned long pfn)
3703 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3708 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3709 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3710 struct vm_area_struct *vma, struct list_head *to_kill,
3711 unsigned long ksm_addr);
3714 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3715 extern void memblk_nr_poison_inc(unsigned long pfn);
3716 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3718 static inline void memblk_nr_poison_inc(unsigned long pfn)
3722 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3727 #ifndef arch_memory_failure
3728 static inline int arch_memory_failure(unsigned long pfn, int flags)
3734 #ifndef arch_is_platform_page
3735 static inline bool arch_is_platform_page(u64 paddr)
3742 * Error handlers for various types of pages.
3745 MF_IGNORED, /* Error: cannot be handled */
3746 MF_FAILED, /* Error: handling failed */
3747 MF_DELAYED, /* Will be handled later */
3748 MF_RECOVERED, /* Successfully recovered */
3751 enum mf_action_page_type {
3753 MF_MSG_KERNEL_HIGH_ORDER,
3755 MF_MSG_DIFFERENT_COMPOUND,
3758 MF_MSG_UNMAP_FAILED,
3759 MF_MSG_DIRTY_SWAPCACHE,
3760 MF_MSG_CLEAN_SWAPCACHE,
3761 MF_MSG_DIRTY_MLOCKED_LRU,
3762 MF_MSG_CLEAN_MLOCKED_LRU,
3763 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3764 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3767 MF_MSG_TRUNCATED_LRU,
3774 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3775 extern void clear_huge_page(struct page *page,
3776 unsigned long addr_hint,
3777 unsigned int pages_per_huge_page);
3778 int copy_user_large_folio(struct folio *dst, struct folio *src,
3779 unsigned long addr_hint,
3780 struct vm_area_struct *vma);
3781 long copy_folio_from_user(struct folio *dst_folio,
3782 const void __user *usr_src,
3783 bool allow_pagefault);
3786 * vma_is_special_huge - Are transhuge page-table entries considered special?
3787 * @vma: Pointer to the struct vm_area_struct to consider
3789 * Whether transhuge page-table entries are considered "special" following
3790 * the definition in vm_normal_page().
3792 * Return: true if transhuge page-table entries should be considered special,
3795 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3797 return vma_is_dax(vma) || (vma->vm_file &&
3798 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3801 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3803 #if MAX_NUMNODES > 1
3804 void __init setup_nr_node_ids(void);
3806 static inline void setup_nr_node_ids(void) {}
3809 extern int memcmp_pages(struct page *page1, struct page *page2);
3811 static inline int pages_identical(struct page *page1, struct page *page2)
3813 return !memcmp_pages(page1, page2);
3816 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3817 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3818 pgoff_t first_index, pgoff_t nr,
3819 pgoff_t bitmap_pgoff,
3820 unsigned long *bitmap,
3824 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3825 pgoff_t first_index, pgoff_t nr);
3828 extern int sysctl_nr_trim_pages;
3830 #ifdef CONFIG_PRINTK
3831 void mem_dump_obj(void *object);
3833 static inline void mem_dump_obj(void *object) {}
3837 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3838 * @seals: the seals to check
3839 * @vma: the vma to operate on
3841 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3842 * the vma flags. Return 0 if check pass, or <0 for errors.
3844 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3846 if (seals & F_SEAL_FUTURE_WRITE) {
3848 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3849 * "future write" seal active.
3851 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3855 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3856 * MAP_SHARED and read-only, take care to not allow mprotect to
3857 * revert protections on such mappings. Do this only for shared
3858 * mappings. For private mappings, don't need to mask
3859 * VM_MAYWRITE as we still want them to be COW-writable.
3861 if (vma->vm_flags & VM_SHARED)
3862 vm_flags_clear(vma, VM_MAYWRITE);
3868 #ifdef CONFIG_ANON_VMA_NAME
3869 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3870 unsigned long len_in,
3871 struct anon_vma_name *anon_name);
3874 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3875 unsigned long len_in, struct anon_vma_name *anon_name) {
3880 #ifdef CONFIG_UNACCEPTED_MEMORY
3882 bool range_contains_unaccepted_memory(phys_addr_t start, phys_addr_t end);
3883 void accept_memory(phys_addr_t start, phys_addr_t end);
3887 static inline bool range_contains_unaccepted_memory(phys_addr_t start,
3893 static inline void accept_memory(phys_addr_t start, phys_addr_t end)
3899 #endif /* _LINUX_MM_H */