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>
35 struct anon_vma_chain;
39 extern int sysctl_page_lock_unfairness;
41 void mm_core_init(void);
42 void init_mm_internals(void);
44 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
45 extern unsigned long max_mapnr;
47 static inline void set_max_mapnr(unsigned long limit)
52 static inline void set_max_mapnr(unsigned long limit) { }
55 extern atomic_long_t _totalram_pages;
56 static inline unsigned long totalram_pages(void)
58 return (unsigned long)atomic_long_read(&_totalram_pages);
61 static inline void totalram_pages_inc(void)
63 atomic_long_inc(&_totalram_pages);
66 static inline void totalram_pages_dec(void)
68 atomic_long_dec(&_totalram_pages);
71 static inline void totalram_pages_add(long count)
73 atomic_long_add(count, &_totalram_pages);
76 extern void * high_memory;
77 extern int page_cluster;
78 extern const int page_cluster_max;
81 extern int sysctl_legacy_va_layout;
83 #define sysctl_legacy_va_layout 0
86 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
87 extern const int mmap_rnd_bits_min;
88 extern const int mmap_rnd_bits_max;
89 extern int mmap_rnd_bits __read_mostly;
91 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
92 extern const int mmap_rnd_compat_bits_min;
93 extern const int mmap_rnd_compat_bits_max;
94 extern int mmap_rnd_compat_bits __read_mostly;
98 #include <asm/processor.h>
101 * Architectures that support memory tagging (assigning tags to memory regions,
102 * embedding these tags into addresses that point to these memory regions, and
103 * checking that the memory and the pointer tags match on memory accesses)
104 * redefine this macro to strip tags from pointers.
105 * It's defined as noop for architectures that don't support memory tagging.
107 #ifndef untagged_addr
108 #define untagged_addr(addr) (addr)
112 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
116 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
120 #define lm_alias(x) __va(__pa_symbol(x))
124 * To prevent common memory management code establishing
125 * a zero page mapping on a read fault.
126 * This macro should be defined within <asm/pgtable.h>.
127 * s390 does this to prevent multiplexing of hardware bits
128 * related to the physical page in case of virtualization.
130 #ifndef mm_forbids_zeropage
131 #define mm_forbids_zeropage(X) (0)
135 * On some architectures it is expensive to call memset() for small sizes.
136 * If an architecture decides to implement their own version of
137 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
138 * define their own version of this macro in <asm/pgtable.h>
140 #if BITS_PER_LONG == 64
141 /* This function must be updated when the size of struct page grows above 96
142 * or reduces below 56. The idea that compiler optimizes out switch()
143 * statement, and only leaves move/store instructions. Also the compiler can
144 * combine write statements if they are both assignments and can be reordered,
145 * this can result in several of the writes here being dropped.
147 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
148 static inline void __mm_zero_struct_page(struct page *page)
150 unsigned long *_pp = (void *)page;
152 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
153 BUILD_BUG_ON(sizeof(struct page) & 7);
154 BUILD_BUG_ON(sizeof(struct page) < 56);
155 BUILD_BUG_ON(sizeof(struct page) > 96);
157 switch (sizeof(struct page)) {
184 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
188 * Default maximum number of active map areas, this limits the number of vmas
189 * per mm struct. Users can overwrite this number by sysctl but there is a
192 * When a program's coredump is generated as ELF format, a section is created
193 * per a vma. In ELF, the number of sections is represented in unsigned short.
194 * This means the number of sections should be smaller than 65535 at coredump.
195 * Because the kernel adds some informative sections to a image of program at
196 * generating coredump, we need some margin. The number of extra sections is
197 * 1-3 now and depends on arch. We use "5" as safe margin, here.
199 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
200 * not a hard limit any more. Although some userspace tools can be surprised by
203 #define MAPCOUNT_ELF_CORE_MARGIN (5)
204 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
206 extern int sysctl_max_map_count;
208 extern unsigned long sysctl_user_reserve_kbytes;
209 extern unsigned long sysctl_admin_reserve_kbytes;
211 extern int sysctl_overcommit_memory;
212 extern int sysctl_overcommit_ratio;
213 extern unsigned long sysctl_overcommit_kbytes;
215 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
217 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
219 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
222 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
223 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
224 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
226 #define nth_page(page,n) ((page) + (n))
227 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
230 /* to align the pointer to the (next) page boundary */
231 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
233 /* to align the pointer to the (prev) page boundary */
234 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
236 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
237 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
239 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
240 static inline struct folio *lru_to_folio(struct list_head *head)
242 return list_entry((head)->prev, struct folio, lru);
245 void setup_initial_init_mm(void *start_code, void *end_code,
246 void *end_data, void *brk);
249 * Linux kernel virtual memory manager primitives.
250 * The idea being to have a "virtual" mm in the same way
251 * we have a virtual fs - giving a cleaner interface to the
252 * mm details, and allowing different kinds of memory mappings
253 * (from shared memory to executable loading to arbitrary
257 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
258 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
259 void vm_area_free(struct vm_area_struct *);
260 /* Use only if VMA has no other users */
261 void __vm_area_free(struct vm_area_struct *vma);
264 extern struct rb_root nommu_region_tree;
265 extern struct rw_semaphore nommu_region_sem;
267 extern unsigned int kobjsize(const void *objp);
271 * vm_flags in vm_area_struct, see mm_types.h.
272 * When changing, update also include/trace/events/mmflags.h
274 #define VM_NONE 0x00000000
276 #define VM_READ 0x00000001 /* currently active flags */
277 #define VM_WRITE 0x00000002
278 #define VM_EXEC 0x00000004
279 #define VM_SHARED 0x00000008
281 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
282 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
283 #define VM_MAYWRITE 0x00000020
284 #define VM_MAYEXEC 0x00000040
285 #define VM_MAYSHARE 0x00000080
287 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
289 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
290 #else /* CONFIG_MMU */
291 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
292 #define VM_UFFD_MISSING 0
293 #endif /* CONFIG_MMU */
294 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
295 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
297 #define VM_LOCKED 0x00002000
298 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
300 /* Used by sys_madvise() */
301 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
302 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
304 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
305 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
306 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
307 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
308 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
309 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
310 #define VM_SYNC 0x00800000 /* Synchronous page faults */
311 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
312 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
313 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
315 #ifdef CONFIG_MEM_SOFT_DIRTY
316 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
318 # define VM_SOFTDIRTY 0
321 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
322 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
323 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
324 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
326 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
327 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
328 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
329 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
330 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
331 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
332 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
333 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
334 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
335 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
336 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
337 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
339 #ifdef CONFIG_ARCH_HAS_PKEYS
340 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
341 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
342 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
343 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
344 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
346 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
348 # define VM_PKEY_BIT4 0
350 #endif /* CONFIG_ARCH_HAS_PKEYS */
352 #if defined(CONFIG_X86)
353 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
354 #elif defined(CONFIG_PPC)
355 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
356 #elif defined(CONFIG_PARISC)
357 # define VM_GROWSUP VM_ARCH_1
358 #elif defined(CONFIG_IA64)
359 # define VM_GROWSUP VM_ARCH_1
360 #elif defined(CONFIG_SPARC64)
361 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
362 # define VM_ARCH_CLEAR VM_SPARC_ADI
363 #elif defined(CONFIG_ARM64)
364 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
365 # define VM_ARCH_CLEAR VM_ARM64_BTI
366 #elif !defined(CONFIG_MMU)
367 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
370 #if defined(CONFIG_ARM64_MTE)
371 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
372 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
374 # define VM_MTE VM_NONE
375 # define VM_MTE_ALLOWED VM_NONE
379 # define VM_GROWSUP VM_NONE
382 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
383 # define VM_UFFD_MINOR_BIT 37
384 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
385 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
386 # define VM_UFFD_MINOR VM_NONE
387 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
389 /* Bits set in the VMA until the stack is in its final location */
390 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
392 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
394 /* Common data flag combinations */
395 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
396 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
397 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
398 VM_MAYWRITE | VM_MAYEXEC)
399 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
400 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
402 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
403 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
406 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
407 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
410 #ifdef CONFIG_STACK_GROWSUP
411 #define VM_STACK VM_GROWSUP
413 #define VM_STACK VM_GROWSDOWN
416 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
418 /* VMA basic access permission flags */
419 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
423 * Special vmas that are non-mergable, non-mlock()able.
425 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
427 /* This mask prevents VMA from being scanned with khugepaged */
428 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
430 /* This mask defines which mm->def_flags a process can inherit its parent */
431 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
433 /* This mask represents all the VMA flag bits used by mlock */
434 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
436 /* Arch-specific flags to clear when updating VM flags on protection change */
437 #ifndef VM_ARCH_CLEAR
438 # define VM_ARCH_CLEAR VM_NONE
440 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
443 * mapping from the currently active vm_flags protection bits (the
444 * low four bits) to a page protection mask..
448 * The default fault flags that should be used by most of the
449 * arch-specific page fault handlers.
451 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
452 FAULT_FLAG_KILLABLE | \
453 FAULT_FLAG_INTERRUPTIBLE)
456 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
457 * @flags: Fault flags.
459 * This is mostly used for places where we want to try to avoid taking
460 * the mmap_lock for too long a time when waiting for another condition
461 * to change, in which case we can try to be polite to release the
462 * mmap_lock in the first round to avoid potential starvation of other
463 * processes that would also want the mmap_lock.
465 * Return: true if the page fault allows retry and this is the first
466 * attempt of the fault handling; false otherwise.
468 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
470 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
471 (!(flags & FAULT_FLAG_TRIED));
474 #define FAULT_FLAG_TRACE \
475 { FAULT_FLAG_WRITE, "WRITE" }, \
476 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
477 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
478 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
479 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
480 { FAULT_FLAG_TRIED, "TRIED" }, \
481 { FAULT_FLAG_USER, "USER" }, \
482 { FAULT_FLAG_REMOTE, "REMOTE" }, \
483 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
484 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
485 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
488 * vm_fault is filled by the pagefault handler and passed to the vma's
489 * ->fault function. The vma's ->fault is responsible for returning a bitmask
490 * of VM_FAULT_xxx flags that give details about how the fault was handled.
492 * MM layer fills up gfp_mask for page allocations but fault handler might
493 * alter it if its implementation requires a different allocation context.
495 * pgoff should be used in favour of virtual_address, if possible.
499 struct vm_area_struct *vma; /* Target VMA */
500 gfp_t gfp_mask; /* gfp mask to be used for allocations */
501 pgoff_t pgoff; /* Logical page offset based on vma */
502 unsigned long address; /* Faulting virtual address - masked */
503 unsigned long real_address; /* Faulting virtual address - unmasked */
505 enum fault_flag flags; /* FAULT_FLAG_xxx flags
506 * XXX: should really be 'const' */
507 pmd_t *pmd; /* Pointer to pmd entry matching
509 pud_t *pud; /* Pointer to pud entry matching
513 pte_t orig_pte; /* Value of PTE at the time of fault */
514 pmd_t orig_pmd; /* Value of PMD at the time of fault,
515 * used by PMD fault only.
519 struct page *cow_page; /* Page handler may use for COW fault */
520 struct page *page; /* ->fault handlers should return a
521 * page here, unless VM_FAULT_NOPAGE
522 * is set (which is also implied by
525 /* These three entries are valid only while holding ptl lock */
526 pte_t *pte; /* Pointer to pte entry matching
527 * the 'address'. NULL if the page
528 * table hasn't been allocated.
530 spinlock_t *ptl; /* Page table lock.
531 * Protects pte page table if 'pte'
532 * is not NULL, otherwise pmd.
534 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
535 * vm_ops->map_pages() sets up a page
536 * table from atomic context.
537 * do_fault_around() pre-allocates
538 * page table to avoid allocation from
543 /* page entry size for vm->huge_fault() */
544 enum page_entry_size {
551 * These are the virtual MM functions - opening of an area, closing and
552 * unmapping it (needed to keep files on disk up-to-date etc), pointer
553 * to the functions called when a no-page or a wp-page exception occurs.
555 struct vm_operations_struct {
556 void (*open)(struct vm_area_struct * area);
558 * @close: Called when the VMA is being removed from the MM.
559 * Context: User context. May sleep. Caller holds mmap_lock.
561 void (*close)(struct vm_area_struct * area);
562 /* Called any time before splitting to check if it's allowed */
563 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
564 int (*mremap)(struct vm_area_struct *area);
566 * Called by mprotect() to make driver-specific permission
567 * checks before mprotect() is finalised. The VMA must not
568 * be modified. Returns 0 if mprotect() can proceed.
570 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
571 unsigned long end, unsigned long newflags);
572 vm_fault_t (*fault)(struct vm_fault *vmf);
573 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
574 enum page_entry_size pe_size);
575 vm_fault_t (*map_pages)(struct vm_fault *vmf,
576 pgoff_t start_pgoff, pgoff_t end_pgoff);
577 unsigned long (*pagesize)(struct vm_area_struct * area);
579 /* notification that a previously read-only page is about to become
580 * writable, if an error is returned it will cause a SIGBUS */
581 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
583 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
584 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
586 /* called by access_process_vm when get_user_pages() fails, typically
587 * for use by special VMAs. See also generic_access_phys() for a generic
588 * implementation useful for any iomem mapping.
590 int (*access)(struct vm_area_struct *vma, unsigned long addr,
591 void *buf, int len, int write);
593 /* Called by the /proc/PID/maps code to ask the vma whether it
594 * has a special name. Returning non-NULL will also cause this
595 * vma to be dumped unconditionally. */
596 const char *(*name)(struct vm_area_struct *vma);
600 * set_policy() op must add a reference to any non-NULL @new mempolicy
601 * to hold the policy upon return. Caller should pass NULL @new to
602 * remove a policy and fall back to surrounding context--i.e. do not
603 * install a MPOL_DEFAULT policy, nor the task or system default
606 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
609 * get_policy() op must add reference [mpol_get()] to any policy at
610 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
611 * in mm/mempolicy.c will do this automatically.
612 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
613 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
614 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
615 * must return NULL--i.e., do not "fallback" to task or system default
618 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
622 * Called by vm_normal_page() for special PTEs to find the
623 * page for @addr. This is useful if the default behavior
624 * (using pte_page()) would not find the correct page.
626 struct page *(*find_special_page)(struct vm_area_struct *vma,
630 #ifdef CONFIG_PER_VMA_LOCK
631 static inline void vma_init_lock(struct vm_area_struct *vma)
633 init_rwsem(&vma->lock);
634 vma->vm_lock_seq = -1;
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->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->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))) {
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 */
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->lock);
691 vma->vm_lock_seq = mm_lock_seq;
692 up_write(&vma->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 void vma_init_lock(struct vm_area_struct *vma) {}
731 static inline bool vma_start_read(struct vm_area_struct *vma)
733 static inline void vma_end_read(struct vm_area_struct *vma) {}
734 static inline void vma_start_write(struct vm_area_struct *vma) {}
735 static inline bool vma_try_start_write(struct vm_area_struct *vma)
737 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
738 static inline void vma_mark_detached(struct vm_area_struct *vma,
741 #endif /* CONFIG_PER_VMA_LOCK */
743 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
745 static const struct vm_operations_struct dummy_vm_ops = {};
747 memset(vma, 0, sizeof(*vma));
749 vma->vm_ops = &dummy_vm_ops;
750 INIT_LIST_HEAD(&vma->anon_vma_chain);
751 vma_mark_detached(vma, false);
755 /* Use when VMA is not part of the VMA tree and needs no locking */
756 static inline void vm_flags_init(struct vm_area_struct *vma,
759 ACCESS_PRIVATE(vma, __vm_flags) = flags;
762 /* Use when VMA is part of the VMA tree and modifications need coordination */
763 static inline void vm_flags_reset(struct vm_area_struct *vma,
766 vma_start_write(vma);
767 vm_flags_init(vma, flags);
770 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
773 vma_start_write(vma);
774 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
777 static inline void vm_flags_set(struct vm_area_struct *vma,
780 vma_start_write(vma);
781 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
784 static inline void vm_flags_clear(struct vm_area_struct *vma,
787 vma_start_write(vma);
788 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
792 * Use only if VMA is not part of the VMA tree or has no other users and
793 * therefore needs no locking.
795 static inline void __vm_flags_mod(struct vm_area_struct *vma,
796 vm_flags_t set, vm_flags_t clear)
798 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
802 * Use only when the order of set/clear operations is unimportant, otherwise
803 * use vm_flags_{set|clear} explicitly.
805 static inline void vm_flags_mod(struct vm_area_struct *vma,
806 vm_flags_t set, vm_flags_t clear)
808 vma_start_write(vma);
809 __vm_flags_mod(vma, set, clear);
812 static inline void vma_set_anonymous(struct vm_area_struct *vma)
817 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
822 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
824 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
829 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
830 VM_STACK_INCOMPLETE_SETUP)
836 static inline bool vma_is_foreign(struct vm_area_struct *vma)
841 if (current->mm != vma->vm_mm)
847 static inline bool vma_is_accessible(struct vm_area_struct *vma)
849 return vma->vm_flags & VM_ACCESS_FLAGS;
853 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
855 return mas_find(&vmi->mas, max - 1);
858 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
861 * Uses mas_find() to get the first VMA when the iterator starts.
862 * Calling mas_next() could skip the first entry.
864 return mas_find(&vmi->mas, ULONG_MAX);
867 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
869 return mas_prev(&vmi->mas, 0);
872 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
874 return vmi->mas.index;
877 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
879 return vmi->mas.last + 1;
881 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
884 return mas_expected_entries(&vmi->mas, count);
887 /* Free any unused preallocations */
888 static inline void vma_iter_free(struct vma_iterator *vmi)
890 mas_destroy(&vmi->mas);
893 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
894 struct vm_area_struct *vma)
896 vmi->mas.index = vma->vm_start;
897 vmi->mas.last = vma->vm_end - 1;
898 mas_store(&vmi->mas, vma);
899 if (unlikely(mas_is_err(&vmi->mas)))
905 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
907 mas_pause(&vmi->mas);
910 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
912 mas_set(&vmi->mas, addr);
915 #define for_each_vma(__vmi, __vma) \
916 while (((__vma) = vma_next(&(__vmi))) != NULL)
918 /* The MM code likes to work with exclusive end addresses */
919 #define for_each_vma_range(__vmi, __vma, __end) \
920 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
924 * The vma_is_shmem is not inline because it is used only by slow
925 * paths in userfault.
927 bool vma_is_shmem(struct vm_area_struct *vma);
928 bool vma_is_anon_shmem(struct vm_area_struct *vma);
930 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
931 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
934 int vma_is_stack_for_current(struct vm_area_struct *vma);
936 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
937 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
943 * compound_order() can be called without holding a reference, which means
944 * that niceties like page_folio() don't work. These callers should be
945 * prepared to handle wild return values. For example, PG_head may be
946 * set before _folio_order is initialised, or this may be a tail page.
947 * See compaction.c for some good examples.
949 static inline unsigned int compound_order(struct page *page)
951 struct folio *folio = (struct folio *)page;
953 if (!test_bit(PG_head, &folio->flags))
955 return folio->_folio_order;
959 * folio_order - The allocation order of a folio.
962 * A folio is composed of 2^order pages. See get_order() for the definition
965 * Return: The order of the folio.
967 static inline unsigned int folio_order(struct folio *folio)
969 if (!folio_test_large(folio))
971 return folio->_folio_order;
974 #include <linux/huge_mm.h>
977 * Methods to modify the page usage count.
979 * What counts for a page usage:
980 * - cache mapping (page->mapping)
981 * - private data (page->private)
982 * - page mapped in a task's page tables, each mapping
983 * is counted separately
985 * Also, many kernel routines increase the page count before a critical
986 * routine so they can be sure the page doesn't go away from under them.
990 * Drop a ref, return true if the refcount fell to zero (the page has no users)
992 static inline int put_page_testzero(struct page *page)
994 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
995 return page_ref_dec_and_test(page);
998 static inline int folio_put_testzero(struct folio *folio)
1000 return put_page_testzero(&folio->page);
1004 * Try to grab a ref unless the page has a refcount of zero, return false if
1006 * This can be called when MMU is off so it must not access
1007 * any of the virtual mappings.
1009 static inline bool get_page_unless_zero(struct page *page)
1011 return page_ref_add_unless(page, 1, 0);
1014 static inline struct folio *folio_get_nontail_page(struct page *page)
1016 if (unlikely(!get_page_unless_zero(page)))
1018 return (struct folio *)page;
1021 extern int page_is_ram(unsigned long pfn);
1029 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1030 unsigned long desc);
1032 /* Support for virtually mapped pages */
1033 struct page *vmalloc_to_page(const void *addr);
1034 unsigned long vmalloc_to_pfn(const void *addr);
1037 * Determine if an address is within the vmalloc range
1039 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1040 * is no special casing required.
1043 #ifndef is_ioremap_addr
1044 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1048 extern bool is_vmalloc_addr(const void *x);
1049 extern int is_vmalloc_or_module_addr(const void *x);
1051 static inline bool is_vmalloc_addr(const void *x)
1055 static inline int is_vmalloc_or_module_addr(const void *x)
1062 * How many times the entire folio is mapped as a single unit (eg by a
1063 * PMD or PUD entry). This is probably not what you want, except for
1064 * debugging purposes - it does not include PTE-mapped sub-pages; look
1065 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1067 static inline int folio_entire_mapcount(struct folio *folio)
1069 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1070 return atomic_read(&folio->_entire_mapcount) + 1;
1074 * The atomic page->_mapcount, starts from -1: so that transitions
1075 * both from it and to it can be tracked, using atomic_inc_and_test
1076 * and atomic_add_negative(-1).
1078 static inline void page_mapcount_reset(struct page *page)
1080 atomic_set(&(page)->_mapcount, -1);
1084 * page_mapcount() - Number of times this precise page is mapped.
1087 * The number of times this page is mapped. If this page is part of
1088 * a large folio, it includes the number of times this page is mapped
1089 * as part of that folio.
1091 * The result is undefined for pages which cannot be mapped into userspace.
1092 * For example SLAB or special types of pages. See function page_has_type().
1093 * They use this field in struct page differently.
1095 static inline int page_mapcount(struct page *page)
1097 int mapcount = atomic_read(&page->_mapcount) + 1;
1099 if (unlikely(PageCompound(page)))
1100 mapcount += folio_entire_mapcount(page_folio(page));
1105 int folio_total_mapcount(struct folio *folio);
1108 * folio_mapcount() - Calculate the number of mappings of this folio.
1109 * @folio: The folio.
1111 * A large folio tracks both how many times the entire folio is mapped,
1112 * and how many times each individual page in the folio is mapped.
1113 * This function calculates the total number of times the folio is
1116 * Return: The number of times this folio is mapped.
1118 static inline int folio_mapcount(struct folio *folio)
1120 if (likely(!folio_test_large(folio)))
1121 return atomic_read(&folio->_mapcount) + 1;
1122 return folio_total_mapcount(folio);
1125 static inline int total_mapcount(struct page *page)
1127 if (likely(!PageCompound(page)))
1128 return atomic_read(&page->_mapcount) + 1;
1129 return folio_total_mapcount(page_folio(page));
1132 static inline bool folio_large_is_mapped(struct folio *folio)
1135 * Reading _entire_mapcount below could be omitted if hugetlb
1136 * participated in incrementing nr_pages_mapped when compound mapped.
1138 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1139 atomic_read(&folio->_entire_mapcount) >= 0;
1143 * folio_mapped - Is this folio mapped into userspace?
1144 * @folio: The folio.
1146 * Return: True if any page in this folio is referenced by user page tables.
1148 static inline bool folio_mapped(struct folio *folio)
1150 if (likely(!folio_test_large(folio)))
1151 return atomic_read(&folio->_mapcount) >= 0;
1152 return folio_large_is_mapped(folio);
1156 * Return true if this page is mapped into pagetables.
1157 * For compound page it returns true if any sub-page of compound page is mapped,
1158 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1160 static inline bool page_mapped(struct page *page)
1162 if (likely(!PageCompound(page)))
1163 return atomic_read(&page->_mapcount) >= 0;
1164 return folio_large_is_mapped(page_folio(page));
1167 static inline struct page *virt_to_head_page(const void *x)
1169 struct page *page = virt_to_page(x);
1171 return compound_head(page);
1174 static inline struct folio *virt_to_folio(const void *x)
1176 struct page *page = virt_to_page(x);
1178 return page_folio(page);
1181 void __folio_put(struct folio *folio);
1183 void put_pages_list(struct list_head *pages);
1185 void split_page(struct page *page, unsigned int order);
1186 void folio_copy(struct folio *dst, struct folio *src);
1188 unsigned long nr_free_buffer_pages(void);
1191 * Compound pages have a destructor function. Provide a
1192 * prototype for that function and accessor functions.
1193 * These are _only_ valid on the head of a compound page.
1195 typedef void compound_page_dtor(struct page *);
1197 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1198 enum compound_dtor_id {
1201 #ifdef CONFIG_HUGETLB_PAGE
1204 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1205 TRANSHUGE_PAGE_DTOR,
1209 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1211 static inline void set_compound_page_dtor(struct page *page,
1212 enum compound_dtor_id compound_dtor)
1214 struct folio *folio = (struct folio *)page;
1216 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1217 VM_BUG_ON_PAGE(!PageHead(page), page);
1218 folio->_folio_dtor = compound_dtor;
1221 static inline void folio_set_compound_dtor(struct folio *folio,
1222 enum compound_dtor_id compound_dtor)
1224 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1225 folio->_folio_dtor = compound_dtor;
1228 void destroy_large_folio(struct folio *folio);
1230 static inline void set_compound_order(struct page *page, unsigned int order)
1232 struct folio *folio = (struct folio *)page;
1234 folio->_folio_order = order;
1236 folio->_folio_nr_pages = 1U << order;
1240 /* Returns the number of bytes in this potentially compound page. */
1241 static inline unsigned long page_size(struct page *page)
1243 return PAGE_SIZE << compound_order(page);
1246 /* Returns the number of bits needed for the number of bytes in a page */
1247 static inline unsigned int page_shift(struct page *page)
1249 return PAGE_SHIFT + compound_order(page);
1253 * thp_order - Order of a transparent huge page.
1254 * @page: Head page of a transparent huge page.
1256 static inline unsigned int thp_order(struct page *page)
1258 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1259 return compound_order(page);
1263 * thp_size - Size of a transparent huge page.
1264 * @page: Head page of a transparent huge page.
1266 * Return: Number of bytes in this page.
1268 static inline unsigned long thp_size(struct page *page)
1270 return PAGE_SIZE << thp_order(page);
1273 void free_compound_page(struct page *page);
1277 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1278 * servicing faults for write access. In the normal case, do always want
1279 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1280 * that do not have writing enabled, when used by access_process_vm.
1282 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1284 if (likely(vma->vm_flags & VM_WRITE))
1285 pte = pte_mkwrite(pte);
1289 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1290 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1292 vm_fault_t finish_fault(struct vm_fault *vmf);
1293 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1297 * Multiple processes may "see" the same page. E.g. for untouched
1298 * mappings of /dev/null, all processes see the same page full of
1299 * zeroes, and text pages of executables and shared libraries have
1300 * only one copy in memory, at most, normally.
1302 * For the non-reserved pages, page_count(page) denotes a reference count.
1303 * page_count() == 0 means the page is free. page->lru is then used for
1304 * freelist management in the buddy allocator.
1305 * page_count() > 0 means the page has been allocated.
1307 * Pages are allocated by the slab allocator in order to provide memory
1308 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1309 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1310 * unless a particular usage is carefully commented. (the responsibility of
1311 * freeing the kmalloc memory is the caller's, of course).
1313 * A page may be used by anyone else who does a __get_free_page().
1314 * In this case, page_count still tracks the references, and should only
1315 * be used through the normal accessor functions. The top bits of page->flags
1316 * and page->virtual store page management information, but all other fields
1317 * are unused and could be used privately, carefully. The management of this
1318 * page is the responsibility of the one who allocated it, and those who have
1319 * subsequently been given references to it.
1321 * The other pages (we may call them "pagecache pages") are completely
1322 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1323 * The following discussion applies only to them.
1325 * A pagecache page contains an opaque `private' member, which belongs to the
1326 * page's address_space. Usually, this is the address of a circular list of
1327 * the page's disk buffers. PG_private must be set to tell the VM to call
1328 * into the filesystem to release these pages.
1330 * A page may belong to an inode's memory mapping. In this case, page->mapping
1331 * is the pointer to the inode, and page->index is the file offset of the page,
1332 * in units of PAGE_SIZE.
1334 * If pagecache pages are not associated with an inode, they are said to be
1335 * anonymous pages. These may become associated with the swapcache, and in that
1336 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1338 * In either case (swapcache or inode backed), the pagecache itself holds one
1339 * reference to the page. Setting PG_private should also increment the
1340 * refcount. The each user mapping also has a reference to the page.
1342 * The pagecache pages are stored in a per-mapping radix tree, which is
1343 * rooted at mapping->i_pages, and indexed by offset.
1344 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1345 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1347 * All pagecache pages may be subject to I/O:
1348 * - inode pages may need to be read from disk,
1349 * - inode pages which have been modified and are MAP_SHARED may need
1350 * to be written back to the inode on disk,
1351 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1352 * modified may need to be swapped out to swap space and (later) to be read
1356 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1357 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1359 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1360 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1362 if (!static_branch_unlikely(&devmap_managed_key))
1364 if (!is_zone_device_page(page))
1366 return __put_devmap_managed_page_refs(page, refs);
1368 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1369 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1373 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1375 static inline bool put_devmap_managed_page(struct page *page)
1377 return put_devmap_managed_page_refs(page, 1);
1380 /* 127: arbitrary random number, small enough to assemble well */
1381 #define folio_ref_zero_or_close_to_overflow(folio) \
1382 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1385 * folio_get - Increment the reference count on a folio.
1386 * @folio: The folio.
1388 * Context: May be called in any context, as long as you know that
1389 * you have a refcount on the folio. If you do not already have one,
1390 * folio_try_get() may be the right interface for you to use.
1392 static inline void folio_get(struct folio *folio)
1394 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1395 folio_ref_inc(folio);
1398 static inline void get_page(struct page *page)
1400 folio_get(page_folio(page));
1403 static inline __must_check bool try_get_page(struct page *page)
1405 page = compound_head(page);
1406 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1413 * folio_put - Decrement the reference count on a folio.
1414 * @folio: The folio.
1416 * If the folio's reference count reaches zero, the memory will be
1417 * released back to the page allocator and may be used by another
1418 * allocation immediately. Do not access the memory or the struct folio
1419 * after calling folio_put() unless you can be sure that it wasn't the
1422 * Context: May be called in process or interrupt context, but not in NMI
1423 * context. May be called while holding a spinlock.
1425 static inline void folio_put(struct folio *folio)
1427 if (folio_put_testzero(folio))
1432 * folio_put_refs - Reduce the reference count on a folio.
1433 * @folio: The folio.
1434 * @refs: The amount to subtract from the folio's reference count.
1436 * If the folio's reference count reaches zero, the memory will be
1437 * released back to the page allocator and may be used by another
1438 * allocation immediately. Do not access the memory or the struct folio
1439 * after calling folio_put_refs() unless you can be sure that these weren't
1440 * the last references.
1442 * Context: May be called in process or interrupt context, but not in NMI
1443 * context. May be called while holding a spinlock.
1445 static inline void folio_put_refs(struct folio *folio, int refs)
1447 if (folio_ref_sub_and_test(folio, refs))
1452 * union release_pages_arg - an array of pages or folios
1454 * release_pages() releases a simple array of multiple pages, and
1455 * accepts various different forms of said page array: either
1456 * a regular old boring array of pages, an array of folios, or
1457 * an array of encoded page pointers.
1459 * The transparent union syntax for this kind of "any of these
1460 * argument types" is all kinds of ugly, so look away.
1463 struct page **pages;
1464 struct folio **folios;
1465 struct encoded_page **encoded_pages;
1466 } release_pages_arg __attribute__ ((__transparent_union__));
1468 void release_pages(release_pages_arg, int nr);
1471 * folios_put - Decrement the reference count on an array of folios.
1472 * @folios: The folios.
1473 * @nr: How many folios there are.
1475 * Like folio_put(), but for an array of folios. This is more efficient
1476 * than writing the loop yourself as it will optimise the locks which
1477 * need to be taken if the folios are freed.
1479 * Context: May be called in process or interrupt context, but not in NMI
1480 * context. May be called while holding a spinlock.
1482 static inline void folios_put(struct folio **folios, unsigned int nr)
1484 release_pages(folios, nr);
1487 static inline void put_page(struct page *page)
1489 struct folio *folio = page_folio(page);
1492 * For some devmap managed pages we need to catch refcount transition
1495 if (put_devmap_managed_page(&folio->page))
1501 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1502 * the page's refcount so that two separate items are tracked: the original page
1503 * reference count, and also a new count of how many pin_user_pages() calls were
1504 * made against the page. ("gup-pinned" is another term for the latter).
1506 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1507 * distinct from normal pages. As such, the unpin_user_page() call (and its
1508 * variants) must be used in order to release gup-pinned pages.
1512 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1513 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1514 * simpler, due to the fact that adding an even power of two to the page
1515 * refcount has the effect of using only the upper N bits, for the code that
1516 * counts up using the bias value. This means that the lower bits are left for
1517 * the exclusive use of the original code that increments and decrements by one
1518 * (or at least, by much smaller values than the bias value).
1520 * Of course, once the lower bits overflow into the upper bits (and this is
1521 * OK, because subtraction recovers the original values), then visual inspection
1522 * no longer suffices to directly view the separate counts. However, for normal
1523 * applications that don't have huge page reference counts, this won't be an
1526 * Locking: the lockless algorithm described in folio_try_get_rcu()
1527 * provides safe operation for get_user_pages(), page_mkclean() and
1528 * other calls that race to set up page table entries.
1530 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1532 void unpin_user_page(struct page *page);
1533 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1535 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1537 void unpin_user_pages(struct page **pages, unsigned long npages);
1539 static inline bool is_cow_mapping(vm_flags_t flags)
1541 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1545 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1548 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1549 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1550 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1551 * underlying memory if ptrace is active, so this is only possible if
1552 * ptrace does not apply. Note that there is no mprotect() to upgrade
1553 * write permissions later.
1555 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1559 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1560 #define SECTION_IN_PAGE_FLAGS
1564 * The identification function is mainly used by the buddy allocator for
1565 * determining if two pages could be buddies. We are not really identifying
1566 * the zone since we could be using the section number id if we do not have
1567 * node id available in page flags.
1568 * We only guarantee that it will return the same value for two combinable
1571 static inline int page_zone_id(struct page *page)
1573 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1576 #ifdef NODE_NOT_IN_PAGE_FLAGS
1577 extern int page_to_nid(const struct page *page);
1579 static inline int page_to_nid(const struct page *page)
1581 struct page *p = (struct page *)page;
1583 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1587 static inline int folio_nid(const struct folio *folio)
1589 return page_to_nid(&folio->page);
1592 #ifdef CONFIG_NUMA_BALANCING
1593 /* page access time bits needs to hold at least 4 seconds */
1594 #define PAGE_ACCESS_TIME_MIN_BITS 12
1595 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1596 #define PAGE_ACCESS_TIME_BUCKETS \
1597 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1599 #define PAGE_ACCESS_TIME_BUCKETS 0
1602 #define PAGE_ACCESS_TIME_MASK \
1603 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1605 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1607 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1610 static inline int cpupid_to_pid(int cpupid)
1612 return cpupid & LAST__PID_MASK;
1615 static inline int cpupid_to_cpu(int cpupid)
1617 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1620 static inline int cpupid_to_nid(int cpupid)
1622 return cpu_to_node(cpupid_to_cpu(cpupid));
1625 static inline bool cpupid_pid_unset(int cpupid)
1627 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1630 static inline bool cpupid_cpu_unset(int cpupid)
1632 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1635 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1637 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1640 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1641 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1642 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1644 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1647 static inline int page_cpupid_last(struct page *page)
1649 return page->_last_cpupid;
1651 static inline void page_cpupid_reset_last(struct page *page)
1653 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1656 static inline int page_cpupid_last(struct page *page)
1658 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1661 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1663 static inline void page_cpupid_reset_last(struct page *page)
1665 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1667 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1669 static inline int xchg_page_access_time(struct page *page, int time)
1673 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1674 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1676 #else /* !CONFIG_NUMA_BALANCING */
1677 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1679 return page_to_nid(page); /* XXX */
1682 static inline int xchg_page_access_time(struct page *page, int time)
1687 static inline int page_cpupid_last(struct page *page)
1689 return page_to_nid(page); /* XXX */
1692 static inline int cpupid_to_nid(int cpupid)
1697 static inline int cpupid_to_pid(int cpupid)
1702 static inline int cpupid_to_cpu(int cpupid)
1707 static inline int cpu_pid_to_cpupid(int nid, int pid)
1712 static inline bool cpupid_pid_unset(int cpupid)
1717 static inline void page_cpupid_reset_last(struct page *page)
1721 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1725 #endif /* CONFIG_NUMA_BALANCING */
1727 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1730 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1731 * setting tags for all pages to native kernel tag value 0xff, as the default
1732 * value 0x00 maps to 0xff.
1735 static inline u8 page_kasan_tag(const struct page *page)
1739 if (kasan_enabled()) {
1740 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1747 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1749 unsigned long old_flags, flags;
1751 if (!kasan_enabled())
1755 old_flags = READ_ONCE(page->flags);
1758 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1759 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1760 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1763 static inline void page_kasan_tag_reset(struct page *page)
1765 if (kasan_enabled())
1766 page_kasan_tag_set(page, 0xff);
1769 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1771 static inline u8 page_kasan_tag(const struct page *page)
1776 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1777 static inline void page_kasan_tag_reset(struct page *page) { }
1779 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1781 static inline struct zone *page_zone(const struct page *page)
1783 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1786 static inline pg_data_t *page_pgdat(const struct page *page)
1788 return NODE_DATA(page_to_nid(page));
1791 static inline struct zone *folio_zone(const struct folio *folio)
1793 return page_zone(&folio->page);
1796 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1798 return page_pgdat(&folio->page);
1801 #ifdef SECTION_IN_PAGE_FLAGS
1802 static inline void set_page_section(struct page *page, unsigned long section)
1804 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1805 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1808 static inline unsigned long page_to_section(const struct page *page)
1810 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1815 * folio_pfn - Return the Page Frame Number of a folio.
1816 * @folio: The folio.
1818 * A folio may contain multiple pages. The pages have consecutive
1819 * Page Frame Numbers.
1821 * Return: The Page Frame Number of the first page in the folio.
1823 static inline unsigned long folio_pfn(struct folio *folio)
1825 return page_to_pfn(&folio->page);
1828 static inline struct folio *pfn_folio(unsigned long pfn)
1830 return page_folio(pfn_to_page(pfn));
1834 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1835 * @folio: The folio.
1837 * This function checks if a folio has been pinned via a call to
1838 * a function in the pin_user_pages() family.
1840 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1841 * because it means "definitely not pinned for DMA", but true means "probably
1842 * pinned for DMA, but possibly a false positive due to having at least
1843 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1845 * False positives are OK, because: a) it's unlikely for a folio to
1846 * get that many refcounts, and b) all the callers of this routine are
1847 * expected to be able to deal gracefully with a false positive.
1849 * For large folios, the result will be exactly correct. That's because
1850 * we have more tracking data available: the _pincount field is used
1851 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1853 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1855 * Return: True, if it is likely that the page has been "dma-pinned".
1856 * False, if the page is definitely not dma-pinned.
1858 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1860 if (folio_test_large(folio))
1861 return atomic_read(&folio->_pincount) > 0;
1864 * folio_ref_count() is signed. If that refcount overflows, then
1865 * folio_ref_count() returns a negative value, and callers will avoid
1866 * further incrementing the refcount.
1868 * Here, for that overflow case, use the sign bit to count a little
1869 * bit higher via unsigned math, and thus still get an accurate result.
1871 return ((unsigned int)folio_ref_count(folio)) >=
1872 GUP_PIN_COUNTING_BIAS;
1875 static inline bool page_maybe_dma_pinned(struct page *page)
1877 return folio_maybe_dma_pinned(page_folio(page));
1881 * This should most likely only be called during fork() to see whether we
1882 * should break the cow immediately for an anon page on the src mm.
1884 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1886 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1889 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1891 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1894 return page_maybe_dma_pinned(page);
1897 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1898 #ifdef CONFIG_MIGRATION
1899 static inline bool is_longterm_pinnable_page(struct page *page)
1902 int mt = get_pageblock_migratetype(page);
1904 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1907 /* The zero page may always be pinned */
1908 if (is_zero_pfn(page_to_pfn(page)))
1911 /* Coherent device memory must always allow eviction. */
1912 if (is_device_coherent_page(page))
1915 /* Otherwise, non-movable zone pages can be pinned. */
1916 return !is_zone_movable_page(page);
1919 static inline bool is_longterm_pinnable_page(struct page *page)
1925 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1927 return is_longterm_pinnable_page(&folio->page);
1930 static inline void set_page_zone(struct page *page, enum zone_type zone)
1932 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1933 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1936 static inline void set_page_node(struct page *page, unsigned long node)
1938 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1939 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1942 static inline void set_page_links(struct page *page, enum zone_type zone,
1943 unsigned long node, unsigned long pfn)
1945 set_page_zone(page, zone);
1946 set_page_node(page, node);
1947 #ifdef SECTION_IN_PAGE_FLAGS
1948 set_page_section(page, pfn_to_section_nr(pfn));
1953 * folio_nr_pages - The number of pages in the folio.
1954 * @folio: The folio.
1956 * Return: A positive power of two.
1958 static inline long folio_nr_pages(struct folio *folio)
1960 if (!folio_test_large(folio))
1963 return folio->_folio_nr_pages;
1965 return 1L << folio->_folio_order;
1970 * compound_nr() returns the number of pages in this potentially compound
1971 * page. compound_nr() can be called on a tail page, and is defined to
1972 * return 1 in that case.
1974 static inline unsigned long compound_nr(struct page *page)
1976 struct folio *folio = (struct folio *)page;
1978 if (!test_bit(PG_head, &folio->flags))
1981 return folio->_folio_nr_pages;
1983 return 1L << folio->_folio_order;
1988 * thp_nr_pages - The number of regular pages in this huge page.
1989 * @page: The head page of a huge page.
1991 static inline int thp_nr_pages(struct page *page)
1993 return folio_nr_pages((struct folio *)page);
1997 * folio_next - Move to the next physical folio.
1998 * @folio: The folio we're currently operating on.
2000 * If you have physically contiguous memory which may span more than
2001 * one folio (eg a &struct bio_vec), use this function to move from one
2002 * folio to the next. Do not use it if the memory is only virtually
2003 * contiguous as the folios are almost certainly not adjacent to each
2004 * other. This is the folio equivalent to writing ``page++``.
2006 * Context: We assume that the folios are refcounted and/or locked at a
2007 * higher level and do not adjust the reference counts.
2008 * Return: The next struct folio.
2010 static inline struct folio *folio_next(struct folio *folio)
2012 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2016 * folio_shift - The size of the memory described by this folio.
2017 * @folio: The folio.
2019 * A folio represents a number of bytes which is a power-of-two in size.
2020 * This function tells you which power-of-two the folio is. See also
2021 * folio_size() and folio_order().
2023 * Context: The caller should have a reference on the folio to prevent
2024 * it from being split. It is not necessary for the folio to be locked.
2025 * Return: The base-2 logarithm of the size of this folio.
2027 static inline unsigned int folio_shift(struct folio *folio)
2029 return PAGE_SHIFT + folio_order(folio);
2033 * folio_size - The number of bytes in a folio.
2034 * @folio: The folio.
2036 * Context: The caller should have a reference on the folio to prevent
2037 * it from being split. It is not necessary for the folio to be locked.
2038 * Return: The number of bytes in this folio.
2040 static inline size_t folio_size(struct folio *folio)
2042 return PAGE_SIZE << folio_order(folio);
2046 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2047 * @folio: The folio.
2049 * folio_estimated_sharers() aims to serve as a function to efficiently
2050 * estimate the number of processes sharing a folio. This is done by
2051 * looking at the precise mapcount of the first subpage in the folio, and
2052 * assuming the other subpages are the same. This may not be true for large
2053 * folios. If you want exact mapcounts for exact calculations, look at
2054 * page_mapcount() or folio_total_mapcount().
2056 * Return: The estimated number of processes sharing a folio.
2058 static inline int folio_estimated_sharers(struct folio *folio)
2060 return page_mapcount(folio_page(folio, 0));
2063 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2064 static inline int arch_make_page_accessible(struct page *page)
2070 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2071 static inline int arch_make_folio_accessible(struct folio *folio)
2074 long i, nr = folio_nr_pages(folio);
2076 for (i = 0; i < nr; i++) {
2077 ret = arch_make_page_accessible(folio_page(folio, i));
2087 * Some inline functions in vmstat.h depend on page_zone()
2089 #include <linux/vmstat.h>
2091 static __always_inline void *lowmem_page_address(const struct page *page)
2093 return page_to_virt(page);
2096 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2097 #define HASHED_PAGE_VIRTUAL
2100 #if defined(WANT_PAGE_VIRTUAL)
2101 static inline void *page_address(const struct page *page)
2103 return page->virtual;
2105 static inline void set_page_address(struct page *page, void *address)
2107 page->virtual = address;
2109 #define page_address_init() do { } while(0)
2112 #if defined(HASHED_PAGE_VIRTUAL)
2113 void *page_address(const struct page *page);
2114 void set_page_address(struct page *page, void *virtual);
2115 void page_address_init(void);
2118 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2119 #define page_address(page) lowmem_page_address(page)
2120 #define set_page_address(page, address) do { } while(0)
2121 #define page_address_init() do { } while(0)
2124 static inline void *folio_address(const struct folio *folio)
2126 return page_address(&folio->page);
2129 extern void *page_rmapping(struct page *page);
2130 extern pgoff_t __page_file_index(struct page *page);
2133 * Return the pagecache index of the passed page. Regular pagecache pages
2134 * use ->index whereas swapcache pages use swp_offset(->private)
2136 static inline pgoff_t page_index(struct page *page)
2138 if (unlikely(PageSwapCache(page)))
2139 return __page_file_index(page);
2144 * Return true only if the page has been allocated with
2145 * ALLOC_NO_WATERMARKS and the low watermark was not
2146 * met implying that the system is under some pressure.
2148 static inline bool page_is_pfmemalloc(const struct page *page)
2151 * lru.next has bit 1 set if the page is allocated from the
2152 * pfmemalloc reserves. Callers may simply overwrite it if
2153 * they do not need to preserve that information.
2155 return (uintptr_t)page->lru.next & BIT(1);
2159 * Return true only if the folio has been allocated with
2160 * ALLOC_NO_WATERMARKS and the low watermark was not
2161 * met implying that the system is under some pressure.
2163 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2166 * lru.next has bit 1 set if the page is allocated from the
2167 * pfmemalloc reserves. Callers may simply overwrite it if
2168 * they do not need to preserve that information.
2170 return (uintptr_t)folio->lru.next & BIT(1);
2174 * Only to be called by the page allocator on a freshly allocated
2177 static inline void set_page_pfmemalloc(struct page *page)
2179 page->lru.next = (void *)BIT(1);
2182 static inline void clear_page_pfmemalloc(struct page *page)
2184 page->lru.next = NULL;
2188 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2190 extern void pagefault_out_of_memory(void);
2192 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2193 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2194 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2197 * Flags passed to show_mem() and show_free_areas() to suppress output in
2200 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2202 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2203 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2205 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2209 * Parameter block passed down to zap_pte_range in exceptional cases.
2211 struct zap_details {
2212 struct folio *single_folio; /* Locked folio to be unmapped */
2213 bool even_cows; /* Zap COWed private pages too? */
2214 zap_flags_t zap_flags; /* Extra flags for zapping */
2218 * Whether to drop the pte markers, for example, the uffd-wp information for
2219 * file-backed memory. This should only be specified when we will completely
2220 * drop the page in the mm, either by truncation or unmapping of the vma. By
2221 * default, the flag is not set.
2223 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2224 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2225 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2227 #ifdef CONFIG_SCHED_MM_CID
2228 void sched_mm_cid_before_execve(struct task_struct *t);
2229 void sched_mm_cid_after_execve(struct task_struct *t);
2230 void sched_mm_cid_fork(struct task_struct *t);
2231 void sched_mm_cid_exit_signals(struct task_struct *t);
2232 static inline int task_mm_cid(struct task_struct *t)
2237 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2238 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2239 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2240 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2241 static inline int task_mm_cid(struct task_struct *t)
2244 * Use the processor id as a fall-back when the mm cid feature is
2245 * disabled. This provides functional per-cpu data structure accesses
2246 * in user-space, althrough it won't provide the memory usage benefits.
2248 return raw_smp_processor_id();
2253 extern bool can_do_mlock(void);
2255 static inline bool can_do_mlock(void) { return false; }
2257 extern int user_shm_lock(size_t, struct ucounts *);
2258 extern void user_shm_unlock(size_t, struct ucounts *);
2260 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2262 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2264 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2267 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2268 unsigned long size);
2269 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2270 unsigned long size, struct zap_details *details);
2271 static inline void zap_vma_pages(struct vm_area_struct *vma)
2273 zap_page_range_single(vma, vma->vm_start,
2274 vma->vm_end - vma->vm_start, NULL);
2276 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2277 struct vm_area_struct *start_vma, unsigned long start,
2278 unsigned long end, bool mm_wr_locked);
2280 struct mmu_notifier_range;
2282 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2283 unsigned long end, unsigned long floor, unsigned long ceiling);
2285 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2286 int follow_pte(struct mm_struct *mm, unsigned long address,
2287 pte_t **ptepp, spinlock_t **ptlp);
2288 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2289 unsigned long *pfn);
2290 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2291 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2292 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2293 void *buf, int len, int write);
2295 extern void truncate_pagecache(struct inode *inode, loff_t new);
2296 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2297 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2298 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2299 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2302 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2303 unsigned long address, unsigned int flags,
2304 struct pt_regs *regs);
2305 extern int fixup_user_fault(struct mm_struct *mm,
2306 unsigned long address, unsigned int fault_flags,
2308 void unmap_mapping_pages(struct address_space *mapping,
2309 pgoff_t start, pgoff_t nr, bool even_cows);
2310 void unmap_mapping_range(struct address_space *mapping,
2311 loff_t const holebegin, loff_t const holelen, int even_cows);
2313 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2314 unsigned long address, unsigned int flags,
2315 struct pt_regs *regs)
2317 /* should never happen if there's no MMU */
2319 return VM_FAULT_SIGBUS;
2321 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2322 unsigned int fault_flags, bool *unlocked)
2324 /* should never happen if there's no MMU */
2328 static inline void unmap_mapping_pages(struct address_space *mapping,
2329 pgoff_t start, pgoff_t nr, bool even_cows) { }
2330 static inline void unmap_mapping_range(struct address_space *mapping,
2331 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2334 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2335 loff_t const holebegin, loff_t const holelen)
2337 unmap_mapping_range(mapping, holebegin, holelen, 0);
2340 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2341 void *buf, int len, unsigned int gup_flags);
2342 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2343 void *buf, int len, unsigned int gup_flags);
2344 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2345 void *buf, int len, unsigned int gup_flags);
2347 long get_user_pages_remote(struct mm_struct *mm,
2348 unsigned long start, unsigned long nr_pages,
2349 unsigned int gup_flags, struct page **pages,
2350 struct vm_area_struct **vmas, int *locked);
2351 long pin_user_pages_remote(struct mm_struct *mm,
2352 unsigned long start, unsigned long nr_pages,
2353 unsigned int gup_flags, struct page **pages,
2354 struct vm_area_struct **vmas, int *locked);
2355 long get_user_pages(unsigned long start, unsigned long nr_pages,
2356 unsigned int gup_flags, struct page **pages,
2357 struct vm_area_struct **vmas);
2358 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2359 unsigned int gup_flags, struct page **pages,
2360 struct vm_area_struct **vmas);
2361 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2362 struct page **pages, unsigned int gup_flags);
2363 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2364 struct page **pages, unsigned int gup_flags);
2366 int get_user_pages_fast(unsigned long start, int nr_pages,
2367 unsigned int gup_flags, struct page **pages);
2368 int pin_user_pages_fast(unsigned long start, int nr_pages,
2369 unsigned int gup_flags, struct page **pages);
2371 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2372 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2373 struct task_struct *task, bool bypass_rlim);
2376 struct page *get_dump_page(unsigned long addr);
2378 bool folio_mark_dirty(struct folio *folio);
2379 bool set_page_dirty(struct page *page);
2380 int set_page_dirty_lock(struct page *page);
2382 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2384 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2385 unsigned long old_addr, struct vm_area_struct *new_vma,
2386 unsigned long new_addr, unsigned long len,
2387 bool need_rmap_locks);
2390 * Flags used by change_protection(). For now we make it a bitmap so
2391 * that we can pass in multiple flags just like parameters. However
2392 * for now all the callers are only use one of the flags at the same
2396 * Whether we should manually check if we can map individual PTEs writable,
2397 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2398 * PTEs automatically in a writable mapping.
2400 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2401 /* Whether this protection change is for NUMA hints */
2402 #define MM_CP_PROT_NUMA (1UL << 1)
2403 /* Whether this change is for write protecting */
2404 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2405 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2406 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2407 MM_CP_UFFD_WP_RESOLVE)
2409 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2410 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2413 * We want to check manually if we can change individual PTEs writable
2414 * if we can't do that automatically for all PTEs in a mapping. For
2415 * private mappings, that's always the case when we have write
2416 * permissions as we properly have to handle COW.
2418 if (vma->vm_flags & VM_SHARED)
2419 return vma_wants_writenotify(vma, vma->vm_page_prot);
2420 return !!(vma->vm_flags & VM_WRITE);
2423 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2425 extern long change_protection(struct mmu_gather *tlb,
2426 struct vm_area_struct *vma, unsigned long start,
2427 unsigned long end, unsigned long cp_flags);
2428 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2429 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2430 unsigned long start, unsigned long end, unsigned long newflags);
2433 * doesn't attempt to fault and will return short.
2435 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2436 unsigned int gup_flags, struct page **pages);
2438 static inline bool get_user_page_fast_only(unsigned long addr,
2439 unsigned int gup_flags, struct page **pagep)
2441 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2444 * per-process(per-mm_struct) statistics.
2446 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2448 return percpu_counter_read_positive(&mm->rss_stat[member]);
2451 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2453 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2455 percpu_counter_add(&mm->rss_stat[member], value);
2457 mm_trace_rss_stat(mm, member);
2460 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2462 percpu_counter_inc(&mm->rss_stat[member]);
2464 mm_trace_rss_stat(mm, member);
2467 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2469 percpu_counter_dec(&mm->rss_stat[member]);
2471 mm_trace_rss_stat(mm, member);
2474 /* Optimized variant when page is already known not to be PageAnon */
2475 static inline int mm_counter_file(struct page *page)
2477 if (PageSwapBacked(page))
2478 return MM_SHMEMPAGES;
2479 return MM_FILEPAGES;
2482 static inline int mm_counter(struct page *page)
2485 return MM_ANONPAGES;
2486 return mm_counter_file(page);
2489 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2491 return get_mm_counter(mm, MM_FILEPAGES) +
2492 get_mm_counter(mm, MM_ANONPAGES) +
2493 get_mm_counter(mm, MM_SHMEMPAGES);
2496 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2498 return max(mm->hiwater_rss, get_mm_rss(mm));
2501 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2503 return max(mm->hiwater_vm, mm->total_vm);
2506 static inline void update_hiwater_rss(struct mm_struct *mm)
2508 unsigned long _rss = get_mm_rss(mm);
2510 if ((mm)->hiwater_rss < _rss)
2511 (mm)->hiwater_rss = _rss;
2514 static inline void update_hiwater_vm(struct mm_struct *mm)
2516 if (mm->hiwater_vm < mm->total_vm)
2517 mm->hiwater_vm = mm->total_vm;
2520 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2522 mm->hiwater_rss = get_mm_rss(mm);
2525 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2526 struct mm_struct *mm)
2528 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2530 if (*maxrss < hiwater_rss)
2531 *maxrss = hiwater_rss;
2534 #if defined(SPLIT_RSS_COUNTING)
2535 void sync_mm_rss(struct mm_struct *mm);
2537 static inline void sync_mm_rss(struct mm_struct *mm)
2542 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2543 static inline int pte_special(pte_t pte)
2548 static inline pte_t pte_mkspecial(pte_t pte)
2554 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2555 static inline int pte_devmap(pte_t pte)
2561 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2563 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2567 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2571 #ifdef __PAGETABLE_P4D_FOLDED
2572 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2573 unsigned long address)
2578 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2581 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2582 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2583 unsigned long address)
2587 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2588 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2591 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2593 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2595 if (mm_pud_folded(mm))
2597 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2600 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2602 if (mm_pud_folded(mm))
2604 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2608 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2609 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2610 unsigned long address)
2615 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2616 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2619 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2621 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2623 if (mm_pmd_folded(mm))
2625 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2628 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2630 if (mm_pmd_folded(mm))
2632 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2637 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2639 atomic_long_set(&mm->pgtables_bytes, 0);
2642 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2644 return atomic_long_read(&mm->pgtables_bytes);
2647 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2649 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2652 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2654 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2658 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2659 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2664 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2665 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2668 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2669 int __pte_alloc_kernel(pmd_t *pmd);
2671 #if defined(CONFIG_MMU)
2673 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2674 unsigned long address)
2676 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2677 NULL : p4d_offset(pgd, address);
2680 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2681 unsigned long address)
2683 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2684 NULL : pud_offset(p4d, address);
2687 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2689 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2690 NULL: pmd_offset(pud, address);
2692 #endif /* CONFIG_MMU */
2694 #if USE_SPLIT_PTE_PTLOCKS
2695 #if ALLOC_SPLIT_PTLOCKS
2696 void __init ptlock_cache_init(void);
2697 extern bool ptlock_alloc(struct page *page);
2698 extern void ptlock_free(struct page *page);
2700 static inline spinlock_t *ptlock_ptr(struct page *page)
2704 #else /* ALLOC_SPLIT_PTLOCKS */
2705 static inline void ptlock_cache_init(void)
2709 static inline bool ptlock_alloc(struct page *page)
2714 static inline void ptlock_free(struct page *page)
2718 static inline spinlock_t *ptlock_ptr(struct page *page)
2722 #endif /* ALLOC_SPLIT_PTLOCKS */
2724 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2726 return ptlock_ptr(pmd_page(*pmd));
2729 static inline bool ptlock_init(struct page *page)
2732 * prep_new_page() initialize page->private (and therefore page->ptl)
2733 * with 0. Make sure nobody took it in use in between.
2735 * It can happen if arch try to use slab for page table allocation:
2736 * slab code uses page->slab_cache, which share storage with page->ptl.
2738 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2739 if (!ptlock_alloc(page))
2741 spin_lock_init(ptlock_ptr(page));
2745 #else /* !USE_SPLIT_PTE_PTLOCKS */
2747 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2749 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2751 return &mm->page_table_lock;
2753 static inline void ptlock_cache_init(void) {}
2754 static inline bool ptlock_init(struct page *page) { return true; }
2755 static inline void ptlock_free(struct page *page) {}
2756 #endif /* USE_SPLIT_PTE_PTLOCKS */
2758 static inline bool pgtable_pte_page_ctor(struct page *page)
2760 if (!ptlock_init(page))
2762 __SetPageTable(page);
2763 inc_lruvec_page_state(page, NR_PAGETABLE);
2767 static inline void pgtable_pte_page_dtor(struct page *page)
2770 __ClearPageTable(page);
2771 dec_lruvec_page_state(page, NR_PAGETABLE);
2774 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2776 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2777 pte_t *__pte = pte_offset_map(pmd, address); \
2783 #define pte_unmap_unlock(pte, ptl) do { \
2788 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2790 #define pte_alloc_map(mm, pmd, address) \
2791 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2793 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2794 (pte_alloc(mm, pmd) ? \
2795 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2797 #define pte_alloc_kernel(pmd, address) \
2798 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2799 NULL: pte_offset_kernel(pmd, address))
2801 #if USE_SPLIT_PMD_PTLOCKS
2803 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2805 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2806 return virt_to_page((void *)((unsigned long) pmd & mask));
2809 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2811 return ptlock_ptr(pmd_pgtable_page(pmd));
2814 static inline bool pmd_ptlock_init(struct page *page)
2816 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2817 page->pmd_huge_pte = NULL;
2819 return ptlock_init(page);
2822 static inline void pmd_ptlock_free(struct page *page)
2824 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2825 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2830 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2834 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2836 return &mm->page_table_lock;
2839 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2840 static inline void pmd_ptlock_free(struct page *page) {}
2842 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2846 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2848 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2853 static inline bool pgtable_pmd_page_ctor(struct page *page)
2855 if (!pmd_ptlock_init(page))
2857 __SetPageTable(page);
2858 inc_lruvec_page_state(page, NR_PAGETABLE);
2862 static inline void pgtable_pmd_page_dtor(struct page *page)
2864 pmd_ptlock_free(page);
2865 __ClearPageTable(page);
2866 dec_lruvec_page_state(page, NR_PAGETABLE);
2870 * No scalability reason to split PUD locks yet, but follow the same pattern
2871 * as the PMD locks to make it easier if we decide to. The VM should not be
2872 * considered ready to switch to split PUD locks yet; there may be places
2873 * which need to be converted from page_table_lock.
2875 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2877 return &mm->page_table_lock;
2880 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2882 spinlock_t *ptl = pud_lockptr(mm, pud);
2888 extern void __init pagecache_init(void);
2889 extern void free_initmem(void);
2892 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2893 * into the buddy system. The freed pages will be poisoned with pattern
2894 * "poison" if it's within range [0, UCHAR_MAX].
2895 * Return pages freed into the buddy system.
2897 extern unsigned long free_reserved_area(void *start, void *end,
2898 int poison, const char *s);
2900 extern void adjust_managed_page_count(struct page *page, long count);
2902 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2904 /* Free the reserved page into the buddy system, so it gets managed. */
2905 static inline void free_reserved_page(struct page *page)
2907 ClearPageReserved(page);
2908 init_page_count(page);
2910 adjust_managed_page_count(page, 1);
2912 #define free_highmem_page(page) free_reserved_page(page)
2914 static inline void mark_page_reserved(struct page *page)
2916 SetPageReserved(page);
2917 adjust_managed_page_count(page, -1);
2921 * Default method to free all the __init memory into the buddy system.
2922 * The freed pages will be poisoned with pattern "poison" if it's within
2923 * range [0, UCHAR_MAX].
2924 * Return pages freed into the buddy system.
2926 static inline unsigned long free_initmem_default(int poison)
2928 extern char __init_begin[], __init_end[];
2930 return free_reserved_area(&__init_begin, &__init_end,
2931 poison, "unused kernel image (initmem)");
2934 static inline unsigned long get_num_physpages(void)
2937 unsigned long phys_pages = 0;
2939 for_each_online_node(nid)
2940 phys_pages += node_present_pages(nid);
2946 * Using memblock node mappings, an architecture may initialise its
2947 * zones, allocate the backing mem_map and account for memory holes in an
2948 * architecture independent manner.
2950 * An architecture is expected to register range of page frames backed by
2951 * physical memory with memblock_add[_node]() before calling
2952 * free_area_init() passing in the PFN each zone ends at. At a basic
2953 * usage, an architecture is expected to do something like
2955 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2957 * for_each_valid_physical_page_range()
2958 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2959 * free_area_init(max_zone_pfns);
2961 void free_area_init(unsigned long *max_zone_pfn);
2962 unsigned long node_map_pfn_alignment(void);
2963 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2964 unsigned long end_pfn);
2965 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2966 unsigned long end_pfn);
2967 extern void get_pfn_range_for_nid(unsigned int nid,
2968 unsigned long *start_pfn, unsigned long *end_pfn);
2971 static inline int early_pfn_to_nid(unsigned long pfn)
2976 /* please see mm/page_alloc.c */
2977 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2980 extern void set_dma_reserve(unsigned long new_dma_reserve);
2981 extern void memmap_init_range(unsigned long, int, unsigned long,
2982 unsigned long, unsigned long, enum meminit_context,
2983 struct vmem_altmap *, int migratetype);
2984 extern void setup_per_zone_wmarks(void);
2985 extern void calculate_min_free_kbytes(void);
2986 extern int __meminit init_per_zone_wmark_min(void);
2987 extern void mem_init(void);
2988 extern void __init mmap_init(void);
2990 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2991 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2993 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2995 extern long si_mem_available(void);
2996 extern void si_meminfo(struct sysinfo * val);
2997 extern void si_meminfo_node(struct sysinfo *val, int nid);
2998 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2999 extern unsigned long arch_reserved_kernel_pages(void);
3002 extern __printf(3, 4)
3003 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3005 extern void setup_per_cpu_pageset(void);
3008 extern int min_free_kbytes;
3009 extern int watermark_boost_factor;
3010 extern int watermark_scale_factor;
3011 extern bool arch_has_descending_max_zone_pfns(void);
3014 extern atomic_long_t mmap_pages_allocated;
3015 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3017 /* interval_tree.c */
3018 void vma_interval_tree_insert(struct vm_area_struct *node,
3019 struct rb_root_cached *root);
3020 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3021 struct vm_area_struct *prev,
3022 struct rb_root_cached *root);
3023 void vma_interval_tree_remove(struct vm_area_struct *node,
3024 struct rb_root_cached *root);
3025 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3026 unsigned long start, unsigned long last);
3027 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3028 unsigned long start, unsigned long last);
3030 #define vma_interval_tree_foreach(vma, root, start, last) \
3031 for (vma = vma_interval_tree_iter_first(root, start, last); \
3032 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3034 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3035 struct rb_root_cached *root);
3036 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3037 struct rb_root_cached *root);
3038 struct anon_vma_chain *
3039 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3040 unsigned long start, unsigned long last);
3041 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3042 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3043 #ifdef CONFIG_DEBUG_VM_RB
3044 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3047 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3048 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3049 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3052 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3053 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3054 unsigned long start, unsigned long end, pgoff_t pgoff,
3055 struct vm_area_struct *next);
3056 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3057 unsigned long start, unsigned long end, pgoff_t pgoff);
3058 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3059 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3060 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3061 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3062 struct anon_vma_name *);
3063 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3064 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3065 unsigned long addr, int new_below);
3066 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3067 unsigned long addr, int new_below);
3068 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3069 extern void unlink_file_vma(struct vm_area_struct *);
3070 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3071 unsigned long addr, unsigned long len, pgoff_t pgoff,
3072 bool *need_rmap_locks);
3073 extern void exit_mmap(struct mm_struct *);
3075 static inline int check_data_rlimit(unsigned long rlim,
3077 unsigned long start,
3078 unsigned long end_data,
3079 unsigned long start_data)
3081 if (rlim < RLIM_INFINITY) {
3082 if (((new - start) + (end_data - start_data)) > rlim)
3089 extern int mm_take_all_locks(struct mm_struct *mm);
3090 extern void mm_drop_all_locks(struct mm_struct *mm);
3092 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3093 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3094 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3095 extern struct file *get_task_exe_file(struct task_struct *task);
3097 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3098 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3100 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3101 const struct vm_special_mapping *sm);
3102 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3103 unsigned long addr, unsigned long len,
3104 unsigned long flags,
3105 const struct vm_special_mapping *spec);
3106 /* This is an obsolete alternative to _install_special_mapping. */
3107 extern int install_special_mapping(struct mm_struct *mm,
3108 unsigned long addr, unsigned long len,
3109 unsigned long flags, struct page **pages);
3111 unsigned long randomize_stack_top(unsigned long stack_top);
3112 unsigned long randomize_page(unsigned long start, unsigned long range);
3114 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3116 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3117 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3118 struct list_head *uf);
3119 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3120 unsigned long len, unsigned long prot, unsigned long flags,
3121 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3122 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3123 unsigned long start, size_t len, struct list_head *uf,
3125 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3126 struct list_head *uf);
3127 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3130 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3131 unsigned long start, unsigned long end,
3132 struct list_head *uf, bool downgrade);
3133 extern int __mm_populate(unsigned long addr, unsigned long len,
3135 static inline void mm_populate(unsigned long addr, unsigned long len)
3138 (void) __mm_populate(addr, len, 1);
3141 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3144 /* These take the mm semaphore themselves */
3145 extern int __must_check vm_brk(unsigned long, unsigned long);
3146 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3147 extern int vm_munmap(unsigned long, size_t);
3148 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3149 unsigned long, unsigned long,
3150 unsigned long, unsigned long);
3152 struct vm_unmapped_area_info {
3153 #define VM_UNMAPPED_AREA_TOPDOWN 1
3154 unsigned long flags;
3155 unsigned long length;
3156 unsigned long low_limit;
3157 unsigned long high_limit;
3158 unsigned long align_mask;
3159 unsigned long align_offset;
3162 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3165 extern void truncate_inode_pages(struct address_space *, loff_t);
3166 extern void truncate_inode_pages_range(struct address_space *,
3167 loff_t lstart, loff_t lend);
3168 extern void truncate_inode_pages_final(struct address_space *);
3170 /* generic vm_area_ops exported for stackable file systems */
3171 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3172 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3173 pgoff_t start_pgoff, pgoff_t end_pgoff);
3174 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3176 extern unsigned long stack_guard_gap;
3177 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3178 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
3180 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3181 extern int expand_downwards(struct vm_area_struct *vma,
3182 unsigned long address);
3184 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
3186 #define expand_upwards(vma, address) (0)
3189 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3190 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3191 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3192 struct vm_area_struct **pprev);
3195 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3196 * NULL if none. Assume start_addr < end_addr.
3198 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3199 unsigned long start_addr, unsigned long end_addr);
3202 * vma_lookup() - Find a VMA at a specific address
3203 * @mm: The process address space.
3204 * @addr: The user address.
3206 * Return: The vm_area_struct at the given address, %NULL otherwise.
3209 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3211 return mtree_load(&mm->mm_mt, addr);
3214 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3216 unsigned long vm_start = vma->vm_start;
3218 if (vma->vm_flags & VM_GROWSDOWN) {
3219 vm_start -= stack_guard_gap;
3220 if (vm_start > vma->vm_start)
3226 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3228 unsigned long vm_end = vma->vm_end;
3230 if (vma->vm_flags & VM_GROWSUP) {
3231 vm_end += stack_guard_gap;
3232 if (vm_end < vma->vm_end)
3233 vm_end = -PAGE_SIZE;
3238 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3240 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3243 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3244 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3245 unsigned long vm_start, unsigned long vm_end)
3247 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3249 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3255 static inline bool range_in_vma(struct vm_area_struct *vma,
3256 unsigned long start, unsigned long end)
3258 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3262 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3263 void vma_set_page_prot(struct vm_area_struct *vma);
3265 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3269 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3271 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3275 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3277 #ifdef CONFIG_NUMA_BALANCING
3278 unsigned long change_prot_numa(struct vm_area_struct *vma,
3279 unsigned long start, unsigned long end);
3282 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3283 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3284 unsigned long pfn, unsigned long size, pgprot_t);
3285 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3286 unsigned long pfn, unsigned long size, pgprot_t prot);
3287 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3288 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3289 struct page **pages, unsigned long *num);
3290 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3292 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3294 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3296 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3297 unsigned long pfn, pgprot_t pgprot);
3298 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3300 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3301 unsigned long addr, pfn_t pfn);
3302 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3304 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3305 unsigned long addr, struct page *page)
3307 int err = vm_insert_page(vma, addr, page);
3310 return VM_FAULT_OOM;
3311 if (err < 0 && err != -EBUSY)
3312 return VM_FAULT_SIGBUS;
3314 return VM_FAULT_NOPAGE;
3317 #ifndef io_remap_pfn_range
3318 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3319 unsigned long addr, unsigned long pfn,
3320 unsigned long size, pgprot_t prot)
3322 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3326 static inline vm_fault_t vmf_error(int err)
3329 return VM_FAULT_OOM;
3330 return VM_FAULT_SIGBUS;
3333 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3334 unsigned int foll_flags);
3336 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3338 if (vm_fault & VM_FAULT_OOM)
3340 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3341 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3342 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3348 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3349 * a (NUMA hinting) fault is required.
3351 static inline bool gup_can_follow_protnone(unsigned int flags)
3354 * FOLL_FORCE has to be able to make progress even if the VMA is
3355 * inaccessible. Further, FOLL_FORCE access usually does not represent
3356 * application behaviour and we should avoid triggering NUMA hinting
3359 return flags & FOLL_FORCE;
3362 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3363 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3364 unsigned long size, pte_fn_t fn, void *data);
3365 extern int apply_to_existing_page_range(struct mm_struct *mm,
3366 unsigned long address, unsigned long size,
3367 pte_fn_t fn, void *data);
3369 #ifdef CONFIG_PAGE_POISONING
3370 extern void __kernel_poison_pages(struct page *page, int numpages);
3371 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3372 extern bool _page_poisoning_enabled_early;
3373 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3374 static inline bool page_poisoning_enabled(void)
3376 return _page_poisoning_enabled_early;
3379 * For use in fast paths after init_mem_debugging() has run, or when a
3380 * false negative result is not harmful when called too early.
3382 static inline bool page_poisoning_enabled_static(void)
3384 return static_branch_unlikely(&_page_poisoning_enabled);
3386 static inline void kernel_poison_pages(struct page *page, int numpages)
3388 if (page_poisoning_enabled_static())
3389 __kernel_poison_pages(page, numpages);
3391 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3393 if (page_poisoning_enabled_static())
3394 __kernel_unpoison_pages(page, numpages);
3397 static inline bool page_poisoning_enabled(void) { return false; }
3398 static inline bool page_poisoning_enabled_static(void) { return false; }
3399 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3400 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3401 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3404 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3405 static inline bool want_init_on_alloc(gfp_t flags)
3407 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3410 return flags & __GFP_ZERO;
3413 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3414 static inline bool want_init_on_free(void)
3416 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3420 extern bool _debug_pagealloc_enabled_early;
3421 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3423 static inline bool debug_pagealloc_enabled(void)
3425 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3426 _debug_pagealloc_enabled_early;
3430 * For use in fast paths after init_debug_pagealloc() has run, or when a
3431 * false negative result is not harmful when called too early.
3433 static inline bool debug_pagealloc_enabled_static(void)
3435 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3438 return static_branch_unlikely(&_debug_pagealloc_enabled);
3441 #ifdef CONFIG_DEBUG_PAGEALLOC
3443 * To support DEBUG_PAGEALLOC architecture must ensure that
3444 * __kernel_map_pages() never fails
3446 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3448 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3450 if (debug_pagealloc_enabled_static())
3451 __kernel_map_pages(page, numpages, 1);
3454 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3456 if (debug_pagealloc_enabled_static())
3457 __kernel_map_pages(page, numpages, 0);
3459 #else /* CONFIG_DEBUG_PAGEALLOC */
3460 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3461 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3462 #endif /* CONFIG_DEBUG_PAGEALLOC */
3464 #ifdef __HAVE_ARCH_GATE_AREA
3465 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3466 extern int in_gate_area_no_mm(unsigned long addr);
3467 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3469 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3473 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3474 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3478 #endif /* __HAVE_ARCH_GATE_AREA */
3480 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3482 #ifdef CONFIG_SYSCTL
3483 extern int sysctl_drop_caches;
3484 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3488 void drop_slab(void);
3491 #define randomize_va_space 0
3493 extern int randomize_va_space;
3496 const char * arch_vma_name(struct vm_area_struct *vma);
3498 void print_vma_addr(char *prefix, unsigned long rip);
3500 static inline void print_vma_addr(char *prefix, unsigned long rip)
3505 void *sparse_buffer_alloc(unsigned long size);
3506 struct page * __populate_section_memmap(unsigned long pfn,
3507 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3508 struct dev_pagemap *pgmap);
3509 void pmd_init(void *addr);
3510 void pud_init(void *addr);
3511 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3512 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3513 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3514 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3515 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3516 struct vmem_altmap *altmap, struct page *reuse);
3517 void *vmemmap_alloc_block(unsigned long size, int node);
3519 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3520 struct vmem_altmap *altmap);
3521 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3522 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3523 unsigned long addr, unsigned long next);
3524 int vmemmap_check_pmd(pmd_t *pmd, int node,
3525 unsigned long addr, unsigned long next);
3526 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3527 int node, struct vmem_altmap *altmap);
3528 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3529 int node, struct vmem_altmap *altmap);
3530 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3531 struct vmem_altmap *altmap);
3532 void vmemmap_populate_print_last(void);
3533 #ifdef CONFIG_MEMORY_HOTPLUG
3534 void vmemmap_free(unsigned long start, unsigned long end,
3535 struct vmem_altmap *altmap);
3537 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3538 unsigned long nr_pages);
3541 MF_COUNT_INCREASED = 1 << 0,
3542 MF_ACTION_REQUIRED = 1 << 1,
3543 MF_MUST_KILL = 1 << 2,
3544 MF_SOFT_OFFLINE = 1 << 3,
3545 MF_UNPOISON = 1 << 4,
3546 MF_SW_SIMULATED = 1 << 5,
3547 MF_NO_RETRY = 1 << 6,
3549 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3550 unsigned long count, int mf_flags);
3551 extern int memory_failure(unsigned long pfn, int flags);
3552 extern void memory_failure_queue_kick(int cpu);
3553 extern int unpoison_memory(unsigned long pfn);
3554 extern int sysctl_memory_failure_early_kill;
3555 extern int sysctl_memory_failure_recovery;
3556 extern void shake_page(struct page *p);
3557 extern atomic_long_t num_poisoned_pages __read_mostly;
3558 extern int soft_offline_page(unsigned long pfn, int flags);
3559 #ifdef CONFIG_MEMORY_FAILURE
3560 extern void memory_failure_queue(unsigned long pfn, int flags);
3561 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3562 bool *migratable_cleared);
3563 void num_poisoned_pages_inc(unsigned long pfn);
3564 void num_poisoned_pages_sub(unsigned long pfn, long i);
3566 static inline void memory_failure_queue(unsigned long pfn, int flags)
3570 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3571 bool *migratable_cleared)
3576 static inline void num_poisoned_pages_inc(unsigned long pfn)
3580 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3585 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3586 extern void memblk_nr_poison_inc(unsigned long pfn);
3587 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3589 static inline void memblk_nr_poison_inc(unsigned long pfn)
3593 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3598 #ifndef arch_memory_failure
3599 static inline int arch_memory_failure(unsigned long pfn, int flags)
3605 #ifndef arch_is_platform_page
3606 static inline bool arch_is_platform_page(u64 paddr)
3613 * Error handlers for various types of pages.
3616 MF_IGNORED, /* Error: cannot be handled */
3617 MF_FAILED, /* Error: handling failed */
3618 MF_DELAYED, /* Will be handled later */
3619 MF_RECOVERED, /* Successfully recovered */
3622 enum mf_action_page_type {
3624 MF_MSG_KERNEL_HIGH_ORDER,
3626 MF_MSG_DIFFERENT_COMPOUND,
3629 MF_MSG_UNMAP_FAILED,
3630 MF_MSG_DIRTY_SWAPCACHE,
3631 MF_MSG_CLEAN_SWAPCACHE,
3632 MF_MSG_DIRTY_MLOCKED_LRU,
3633 MF_MSG_CLEAN_MLOCKED_LRU,
3634 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3635 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3638 MF_MSG_TRUNCATED_LRU,
3646 * Sysfs entries for memory failure handling statistics.
3648 extern const struct attribute_group memory_failure_attr_group;
3650 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3651 extern void clear_huge_page(struct page *page,
3652 unsigned long addr_hint,
3653 unsigned int pages_per_huge_page);
3654 extern void copy_user_huge_page(struct page *dst, struct page *src,
3655 unsigned long addr_hint,
3656 struct vm_area_struct *vma,
3657 unsigned int pages_per_huge_page);
3658 extern long copy_huge_page_from_user(struct page *dst_page,
3659 const void __user *usr_src,
3660 unsigned int pages_per_huge_page,
3661 bool allow_pagefault);
3664 * vma_is_special_huge - Are transhuge page-table entries considered special?
3665 * @vma: Pointer to the struct vm_area_struct to consider
3667 * Whether transhuge page-table entries are considered "special" following
3668 * the definition in vm_normal_page().
3670 * Return: true if transhuge page-table entries should be considered special,
3673 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3675 return vma_is_dax(vma) || (vma->vm_file &&
3676 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3679 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3681 #ifdef CONFIG_DEBUG_PAGEALLOC
3682 extern unsigned int _debug_guardpage_minorder;
3683 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3685 static inline unsigned int debug_guardpage_minorder(void)
3687 return _debug_guardpage_minorder;
3690 static inline bool debug_guardpage_enabled(void)
3692 return static_branch_unlikely(&_debug_guardpage_enabled);
3695 static inline bool page_is_guard(struct page *page)
3697 if (!debug_guardpage_enabled())
3700 return PageGuard(page);
3703 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3704 static inline bool debug_guardpage_enabled(void) { return false; }
3705 static inline bool page_is_guard(struct page *page) { return false; }
3706 #endif /* CONFIG_DEBUG_PAGEALLOC */
3708 #if MAX_NUMNODES > 1
3709 void __init setup_nr_node_ids(void);
3711 static inline void setup_nr_node_ids(void) {}
3714 extern int memcmp_pages(struct page *page1, struct page *page2);
3716 static inline int pages_identical(struct page *page1, struct page *page2)
3718 return !memcmp_pages(page1, page2);
3721 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3722 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3723 pgoff_t first_index, pgoff_t nr,
3724 pgoff_t bitmap_pgoff,
3725 unsigned long *bitmap,
3729 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3730 pgoff_t first_index, pgoff_t nr);
3733 extern int sysctl_nr_trim_pages;
3735 #ifdef CONFIG_PRINTK
3736 void mem_dump_obj(void *object);
3738 static inline void mem_dump_obj(void *object) {}
3742 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3743 * @seals: the seals to check
3744 * @vma: the vma to operate on
3746 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3747 * the vma flags. Return 0 if check pass, or <0 for errors.
3749 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3751 if (seals & F_SEAL_FUTURE_WRITE) {
3753 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3754 * "future write" seal active.
3756 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3760 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3761 * MAP_SHARED and read-only, take care to not allow mprotect to
3762 * revert protections on such mappings. Do this only for shared
3763 * mappings. For private mappings, don't need to mask
3764 * VM_MAYWRITE as we still want them to be COW-writable.
3766 if (vma->vm_flags & VM_SHARED)
3767 vm_flags_clear(vma, VM_MAYWRITE);
3773 #ifdef CONFIG_ANON_VMA_NAME
3774 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3775 unsigned long len_in,
3776 struct anon_vma_name *anon_name);
3779 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3780 unsigned long len_in, struct anon_vma_name *anon_name) {
3785 #endif /* _LINUX_MM_H */