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 *);
262 extern struct rb_root nommu_region_tree;
263 extern struct rw_semaphore nommu_region_sem;
265 extern unsigned int kobjsize(const void *objp);
269 * vm_flags in vm_area_struct, see mm_types.h.
270 * When changing, update also include/trace/events/mmflags.h
272 #define VM_NONE 0x00000000
274 #define VM_READ 0x00000001 /* currently active flags */
275 #define VM_WRITE 0x00000002
276 #define VM_EXEC 0x00000004
277 #define VM_SHARED 0x00000008
279 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
280 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
281 #define VM_MAYWRITE 0x00000020
282 #define VM_MAYEXEC 0x00000040
283 #define VM_MAYSHARE 0x00000080
285 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
287 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
288 #else /* CONFIG_MMU */
289 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
290 #define VM_UFFD_MISSING 0
291 #endif /* CONFIG_MMU */
292 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
293 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
295 #define VM_LOCKED 0x00002000
296 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
298 /* Used by sys_madvise() */
299 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
300 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
302 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
303 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
304 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
305 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
306 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
307 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
308 #define VM_SYNC 0x00800000 /* Synchronous page faults */
309 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
310 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
311 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
313 #ifdef CONFIG_MEM_SOFT_DIRTY
314 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
316 # define VM_SOFTDIRTY 0
319 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
320 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
321 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
322 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
324 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
325 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
326 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
327 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
328 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
329 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
330 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
331 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
332 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
333 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
334 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
335 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
337 #ifdef CONFIG_ARCH_HAS_PKEYS
338 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
339 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
340 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
341 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
342 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
344 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
346 # define VM_PKEY_BIT4 0
348 #endif /* CONFIG_ARCH_HAS_PKEYS */
350 #if defined(CONFIG_X86)
351 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
352 #elif defined(CONFIG_PPC)
353 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
354 #elif defined(CONFIG_PARISC)
355 # define VM_GROWSUP VM_ARCH_1
356 #elif defined(CONFIG_IA64)
357 # define VM_GROWSUP VM_ARCH_1
358 #elif defined(CONFIG_SPARC64)
359 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
360 # define VM_ARCH_CLEAR VM_SPARC_ADI
361 #elif defined(CONFIG_ARM64)
362 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
363 # define VM_ARCH_CLEAR VM_ARM64_BTI
364 #elif !defined(CONFIG_MMU)
365 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
368 #if defined(CONFIG_ARM64_MTE)
369 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
370 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
372 # define VM_MTE VM_NONE
373 # define VM_MTE_ALLOWED VM_NONE
377 # define VM_GROWSUP VM_NONE
380 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
381 # define VM_UFFD_MINOR_BIT 37
382 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
383 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
384 # define VM_UFFD_MINOR VM_NONE
385 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
387 /* Bits set in the VMA until the stack is in its final location */
388 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
390 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
392 /* Common data flag combinations */
393 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
394 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
395 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
396 VM_MAYWRITE | VM_MAYEXEC)
397 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
398 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
400 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
401 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
404 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
405 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
408 #ifdef CONFIG_STACK_GROWSUP
409 #define VM_STACK VM_GROWSUP
411 #define VM_STACK VM_GROWSDOWN
414 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
416 /* VMA basic access permission flags */
417 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
421 * Special vmas that are non-mergable, non-mlock()able.
423 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
425 /* This mask prevents VMA from being scanned with khugepaged */
426 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
428 /* This mask defines which mm->def_flags a process can inherit its parent */
429 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
431 /* This mask represents all the VMA flag bits used by mlock */
432 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
434 /* Arch-specific flags to clear when updating VM flags on protection change */
435 #ifndef VM_ARCH_CLEAR
436 # define VM_ARCH_CLEAR VM_NONE
438 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
441 * mapping from the currently active vm_flags protection bits (the
442 * low four bits) to a page protection mask..
446 * The default fault flags that should be used by most of the
447 * arch-specific page fault handlers.
449 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
450 FAULT_FLAG_KILLABLE | \
451 FAULT_FLAG_INTERRUPTIBLE)
454 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
455 * @flags: Fault flags.
457 * This is mostly used for places where we want to try to avoid taking
458 * the mmap_lock for too long a time when waiting for another condition
459 * to change, in which case we can try to be polite to release the
460 * mmap_lock in the first round to avoid potential starvation of other
461 * processes that would also want the mmap_lock.
463 * Return: true if the page fault allows retry and this is the first
464 * attempt of the fault handling; false otherwise.
466 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
468 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
469 (!(flags & FAULT_FLAG_TRIED));
472 #define FAULT_FLAG_TRACE \
473 { FAULT_FLAG_WRITE, "WRITE" }, \
474 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
475 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
476 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
477 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
478 { FAULT_FLAG_TRIED, "TRIED" }, \
479 { FAULT_FLAG_USER, "USER" }, \
480 { FAULT_FLAG_REMOTE, "REMOTE" }, \
481 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
482 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
483 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
486 * vm_fault is filled by the pagefault handler and passed to the vma's
487 * ->fault function. The vma's ->fault is responsible for returning a bitmask
488 * of VM_FAULT_xxx flags that give details about how the fault was handled.
490 * MM layer fills up gfp_mask for page allocations but fault handler might
491 * alter it if its implementation requires a different allocation context.
493 * pgoff should be used in favour of virtual_address, if possible.
497 struct vm_area_struct *vma; /* Target VMA */
498 gfp_t gfp_mask; /* gfp mask to be used for allocations */
499 pgoff_t pgoff; /* Logical page offset based on vma */
500 unsigned long address; /* Faulting virtual address - masked */
501 unsigned long real_address; /* Faulting virtual address - unmasked */
503 enum fault_flag flags; /* FAULT_FLAG_xxx flags
504 * XXX: should really be 'const' */
505 pmd_t *pmd; /* Pointer to pmd entry matching
507 pud_t *pud; /* Pointer to pud entry matching
511 pte_t orig_pte; /* Value of PTE at the time of fault */
512 pmd_t orig_pmd; /* Value of PMD at the time of fault,
513 * used by PMD fault only.
517 struct page *cow_page; /* Page handler may use for COW fault */
518 struct page *page; /* ->fault handlers should return a
519 * page here, unless VM_FAULT_NOPAGE
520 * is set (which is also implied by
523 /* These three entries are valid only while holding ptl lock */
524 pte_t *pte; /* Pointer to pte entry matching
525 * the 'address'. NULL if the page
526 * table hasn't been allocated.
528 spinlock_t *ptl; /* Page table lock.
529 * Protects pte page table if 'pte'
530 * is not NULL, otherwise pmd.
532 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
533 * vm_ops->map_pages() sets up a page
534 * table from atomic context.
535 * do_fault_around() pre-allocates
536 * page table to avoid allocation from
541 /* page entry size for vm->huge_fault() */
542 enum page_entry_size {
549 * These are the virtual MM functions - opening of an area, closing and
550 * unmapping it (needed to keep files on disk up-to-date etc), pointer
551 * to the functions called when a no-page or a wp-page exception occurs.
553 struct vm_operations_struct {
554 void (*open)(struct vm_area_struct * area);
556 * @close: Called when the VMA is being removed from the MM.
557 * Context: User context. May sleep. Caller holds mmap_lock.
559 void (*close)(struct vm_area_struct * area);
560 /* Called any time before splitting to check if it's allowed */
561 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
562 int (*mremap)(struct vm_area_struct *area);
564 * Called by mprotect() to make driver-specific permission
565 * checks before mprotect() is finalised. The VMA must not
566 * be modified. Returns 0 if mprotect() can proceed.
568 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
569 unsigned long end, unsigned long newflags);
570 vm_fault_t (*fault)(struct vm_fault *vmf);
571 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
572 enum page_entry_size pe_size);
573 vm_fault_t (*map_pages)(struct vm_fault *vmf,
574 pgoff_t start_pgoff, pgoff_t end_pgoff);
575 unsigned long (*pagesize)(struct vm_area_struct * area);
577 /* notification that a previously read-only page is about to become
578 * writable, if an error is returned it will cause a SIGBUS */
579 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
581 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
582 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
584 /* called by access_process_vm when get_user_pages() fails, typically
585 * for use by special VMAs. See also generic_access_phys() for a generic
586 * implementation useful for any iomem mapping.
588 int (*access)(struct vm_area_struct *vma, unsigned long addr,
589 void *buf, int len, int write);
591 /* Called by the /proc/PID/maps code to ask the vma whether it
592 * has a special name. Returning non-NULL will also cause this
593 * vma to be dumped unconditionally. */
594 const char *(*name)(struct vm_area_struct *vma);
598 * set_policy() op must add a reference to any non-NULL @new mempolicy
599 * to hold the policy upon return. Caller should pass NULL @new to
600 * remove a policy and fall back to surrounding context--i.e. do not
601 * install a MPOL_DEFAULT policy, nor the task or system default
604 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
607 * get_policy() op must add reference [mpol_get()] to any policy at
608 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
609 * in mm/mempolicy.c will do this automatically.
610 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
611 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
612 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
613 * must return NULL--i.e., do not "fallback" to task or system default
616 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
620 * Called by vm_normal_page() for special PTEs to find the
621 * page for @addr. This is useful if the default behavior
622 * (using pte_page()) would not find the correct page.
624 struct page *(*find_special_page)(struct vm_area_struct *vma,
628 #ifdef CONFIG_PER_VMA_LOCK
629 static inline void vma_init_lock(struct vm_area_struct *vma)
631 init_rwsem(&vma->lock);
632 vma->vm_lock_seq = -1;
636 * Try to read-lock a vma. The function is allowed to occasionally yield false
637 * locked result to avoid performance overhead, in which case we fall back to
638 * using mmap_lock. The function should never yield false unlocked result.
640 static inline bool vma_start_read(struct vm_area_struct *vma)
642 /* Check before locking. A race might cause false locked result. */
643 if (vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))
646 if (unlikely(down_read_trylock(&vma->lock) == 0))
650 * Overflow might produce false locked result.
651 * False unlocked result is impossible because we modify and check
652 * vma->vm_lock_seq under vma->lock protection and mm->mm_lock_seq
653 * modification invalidates all existing locks.
655 if (unlikely(vma->vm_lock_seq == READ_ONCE(vma->vm_mm->mm_lock_seq))) {
662 static inline void vma_end_read(struct vm_area_struct *vma)
664 rcu_read_lock(); /* keeps vma alive till the end of up_read */
669 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
671 mmap_assert_write_locked(vma->vm_mm);
674 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
675 * mm->mm_lock_seq can't be concurrently modified.
677 *mm_lock_seq = READ_ONCE(vma->vm_mm->mm_lock_seq);
678 return (vma->vm_lock_seq == *mm_lock_seq);
681 static inline void vma_start_write(struct vm_area_struct *vma)
685 if (__is_vma_write_locked(vma, &mm_lock_seq))
688 down_write(&vma->lock);
689 vma->vm_lock_seq = mm_lock_seq;
690 up_write(&vma->lock);
693 static inline bool vma_try_start_write(struct vm_area_struct *vma)
697 if (__is_vma_write_locked(vma, &mm_lock_seq))
700 if (!down_write_trylock(&vma->vm_lock->lock))
703 vma->vm_lock_seq = mm_lock_seq;
704 up_write(&vma->vm_lock->lock);
708 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
712 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
715 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
717 /* When detaching vma should be write-locked */
719 vma_assert_write_locked(vma);
720 vma->detached = detached;
723 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
724 unsigned long address);
726 #else /* CONFIG_PER_VMA_LOCK */
728 static inline void vma_init_lock(struct vm_area_struct *vma) {}
729 static inline bool vma_start_read(struct vm_area_struct *vma)
731 static inline void vma_end_read(struct vm_area_struct *vma) {}
732 static inline void vma_start_write(struct vm_area_struct *vma) {}
733 static inline bool vma_try_start_write(struct vm_area_struct *vma)
735 static inline void vma_assert_write_locked(struct vm_area_struct *vma) {}
736 static inline void vma_mark_detached(struct vm_area_struct *vma,
739 #endif /* CONFIG_PER_VMA_LOCK */
741 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
743 static const struct vm_operations_struct dummy_vm_ops = {};
745 memset(vma, 0, sizeof(*vma));
747 vma->vm_ops = &dummy_vm_ops;
748 INIT_LIST_HEAD(&vma->anon_vma_chain);
749 vma_mark_detached(vma, false);
753 /* Use when VMA is not part of the VMA tree and needs no locking */
754 static inline void vm_flags_init(struct vm_area_struct *vma,
757 ACCESS_PRIVATE(vma, __vm_flags) = flags;
760 /* Use when VMA is part of the VMA tree and modifications need coordination */
761 static inline void vm_flags_reset(struct vm_area_struct *vma,
764 vma_start_write(vma);
765 vm_flags_init(vma, flags);
768 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
771 vma_start_write(vma);
772 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
775 static inline void vm_flags_set(struct vm_area_struct *vma,
778 vma_start_write(vma);
779 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
782 static inline void vm_flags_clear(struct vm_area_struct *vma,
785 vma_start_write(vma);
786 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
790 * Use only if VMA is not part of the VMA tree or has no other users and
791 * therefore needs no locking.
793 static inline void __vm_flags_mod(struct vm_area_struct *vma,
794 vm_flags_t set, vm_flags_t clear)
796 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
800 * Use only when the order of set/clear operations is unimportant, otherwise
801 * use vm_flags_{set|clear} explicitly.
803 static inline void vm_flags_mod(struct vm_area_struct *vma,
804 vm_flags_t set, vm_flags_t clear)
806 vma_start_write(vma);
807 __vm_flags_mod(vma, set, clear);
810 static inline void vma_set_anonymous(struct vm_area_struct *vma)
815 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
820 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
822 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
827 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
828 VM_STACK_INCOMPLETE_SETUP)
834 static inline bool vma_is_foreign(struct vm_area_struct *vma)
839 if (current->mm != vma->vm_mm)
845 static inline bool vma_is_accessible(struct vm_area_struct *vma)
847 return vma->vm_flags & VM_ACCESS_FLAGS;
851 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
853 return mas_find(&vmi->mas, max - 1);
856 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
859 * Uses mas_find() to get the first VMA when the iterator starts.
860 * Calling mas_next() could skip the first entry.
862 return mas_find(&vmi->mas, ULONG_MAX);
865 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
867 return mas_prev(&vmi->mas, 0);
870 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
872 return vmi->mas.index;
875 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
877 return vmi->mas.last + 1;
879 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
882 return mas_expected_entries(&vmi->mas, count);
885 /* Free any unused preallocations */
886 static inline void vma_iter_free(struct vma_iterator *vmi)
888 mas_destroy(&vmi->mas);
891 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
892 struct vm_area_struct *vma)
894 vmi->mas.index = vma->vm_start;
895 vmi->mas.last = vma->vm_end - 1;
896 mas_store(&vmi->mas, vma);
897 if (unlikely(mas_is_err(&vmi->mas)))
903 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
905 mas_pause(&vmi->mas);
908 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
910 mas_set(&vmi->mas, addr);
913 #define for_each_vma(__vmi, __vma) \
914 while (((__vma) = vma_next(&(__vmi))) != NULL)
916 /* The MM code likes to work with exclusive end addresses */
917 #define for_each_vma_range(__vmi, __vma, __end) \
918 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
922 * The vma_is_shmem is not inline because it is used only by slow
923 * paths in userfault.
925 bool vma_is_shmem(struct vm_area_struct *vma);
926 bool vma_is_anon_shmem(struct vm_area_struct *vma);
928 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
929 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
932 int vma_is_stack_for_current(struct vm_area_struct *vma);
934 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
935 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
941 * compound_order() can be called without holding a reference, which means
942 * that niceties like page_folio() don't work. These callers should be
943 * prepared to handle wild return values. For example, PG_head may be
944 * set before _folio_order is initialised, or this may be a tail page.
945 * See compaction.c for some good examples.
947 static inline unsigned int compound_order(struct page *page)
949 struct folio *folio = (struct folio *)page;
951 if (!test_bit(PG_head, &folio->flags))
953 return folio->_folio_order;
957 * folio_order - The allocation order of a folio.
960 * A folio is composed of 2^order pages. See get_order() for the definition
963 * Return: The order of the folio.
965 static inline unsigned int folio_order(struct folio *folio)
967 if (!folio_test_large(folio))
969 return folio->_folio_order;
972 #include <linux/huge_mm.h>
975 * Methods to modify the page usage count.
977 * What counts for a page usage:
978 * - cache mapping (page->mapping)
979 * - private data (page->private)
980 * - page mapped in a task's page tables, each mapping
981 * is counted separately
983 * Also, many kernel routines increase the page count before a critical
984 * routine so they can be sure the page doesn't go away from under them.
988 * Drop a ref, return true if the refcount fell to zero (the page has no users)
990 static inline int put_page_testzero(struct page *page)
992 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
993 return page_ref_dec_and_test(page);
996 static inline int folio_put_testzero(struct folio *folio)
998 return put_page_testzero(&folio->page);
1002 * Try to grab a ref unless the page has a refcount of zero, return false if
1004 * This can be called when MMU is off so it must not access
1005 * any of the virtual mappings.
1007 static inline bool get_page_unless_zero(struct page *page)
1009 return page_ref_add_unless(page, 1, 0);
1012 static inline struct folio *folio_get_nontail_page(struct page *page)
1014 if (unlikely(!get_page_unless_zero(page)))
1016 return (struct folio *)page;
1019 extern int page_is_ram(unsigned long pfn);
1027 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1028 unsigned long desc);
1030 /* Support for virtually mapped pages */
1031 struct page *vmalloc_to_page(const void *addr);
1032 unsigned long vmalloc_to_pfn(const void *addr);
1035 * Determine if an address is within the vmalloc range
1037 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1038 * is no special casing required.
1041 #ifndef is_ioremap_addr
1042 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1046 extern bool is_vmalloc_addr(const void *x);
1047 extern int is_vmalloc_or_module_addr(const void *x);
1049 static inline bool is_vmalloc_addr(const void *x)
1053 static inline int is_vmalloc_or_module_addr(const void *x)
1060 * How many times the entire folio is mapped as a single unit (eg by a
1061 * PMD or PUD entry). This is probably not what you want, except for
1062 * debugging purposes - it does not include PTE-mapped sub-pages; look
1063 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1065 static inline int folio_entire_mapcount(struct folio *folio)
1067 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1068 return atomic_read(&folio->_entire_mapcount) + 1;
1072 * The atomic page->_mapcount, starts from -1: so that transitions
1073 * both from it and to it can be tracked, using atomic_inc_and_test
1074 * and atomic_add_negative(-1).
1076 static inline void page_mapcount_reset(struct page *page)
1078 atomic_set(&(page)->_mapcount, -1);
1082 * page_mapcount() - Number of times this precise page is mapped.
1085 * The number of times this page is mapped. If this page is part of
1086 * a large folio, it includes the number of times this page is mapped
1087 * as part of that folio.
1089 * The result is undefined for pages which cannot be mapped into userspace.
1090 * For example SLAB or special types of pages. See function page_has_type().
1091 * They use this field in struct page differently.
1093 static inline int page_mapcount(struct page *page)
1095 int mapcount = atomic_read(&page->_mapcount) + 1;
1097 if (unlikely(PageCompound(page)))
1098 mapcount += folio_entire_mapcount(page_folio(page));
1103 int folio_total_mapcount(struct folio *folio);
1106 * folio_mapcount() - Calculate the number of mappings of this folio.
1107 * @folio: The folio.
1109 * A large folio tracks both how many times the entire folio is mapped,
1110 * and how many times each individual page in the folio is mapped.
1111 * This function calculates the total number of times the folio is
1114 * Return: The number of times this folio is mapped.
1116 static inline int folio_mapcount(struct folio *folio)
1118 if (likely(!folio_test_large(folio)))
1119 return atomic_read(&folio->_mapcount) + 1;
1120 return folio_total_mapcount(folio);
1123 static inline int total_mapcount(struct page *page)
1125 if (likely(!PageCompound(page)))
1126 return atomic_read(&page->_mapcount) + 1;
1127 return folio_total_mapcount(page_folio(page));
1130 static inline bool folio_large_is_mapped(struct folio *folio)
1133 * Reading _entire_mapcount below could be omitted if hugetlb
1134 * participated in incrementing nr_pages_mapped when compound mapped.
1136 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1137 atomic_read(&folio->_entire_mapcount) >= 0;
1141 * folio_mapped - Is this folio mapped into userspace?
1142 * @folio: The folio.
1144 * Return: True if any page in this folio is referenced by user page tables.
1146 static inline bool folio_mapped(struct folio *folio)
1148 if (likely(!folio_test_large(folio)))
1149 return atomic_read(&folio->_mapcount) >= 0;
1150 return folio_large_is_mapped(folio);
1154 * Return true if this page is mapped into pagetables.
1155 * For compound page it returns true if any sub-page of compound page is mapped,
1156 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1158 static inline bool page_mapped(struct page *page)
1160 if (likely(!PageCompound(page)))
1161 return atomic_read(&page->_mapcount) >= 0;
1162 return folio_large_is_mapped(page_folio(page));
1165 static inline struct page *virt_to_head_page(const void *x)
1167 struct page *page = virt_to_page(x);
1169 return compound_head(page);
1172 static inline struct folio *virt_to_folio(const void *x)
1174 struct page *page = virt_to_page(x);
1176 return page_folio(page);
1179 void __folio_put(struct folio *folio);
1181 void put_pages_list(struct list_head *pages);
1183 void split_page(struct page *page, unsigned int order);
1184 void folio_copy(struct folio *dst, struct folio *src);
1186 unsigned long nr_free_buffer_pages(void);
1189 * Compound pages have a destructor function. Provide a
1190 * prototype for that function and accessor functions.
1191 * These are _only_ valid on the head of a compound page.
1193 typedef void compound_page_dtor(struct page *);
1195 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1196 enum compound_dtor_id {
1199 #ifdef CONFIG_HUGETLB_PAGE
1202 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1203 TRANSHUGE_PAGE_DTOR,
1207 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1209 static inline void set_compound_page_dtor(struct page *page,
1210 enum compound_dtor_id compound_dtor)
1212 struct folio *folio = (struct folio *)page;
1214 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1215 VM_BUG_ON_PAGE(!PageHead(page), page);
1216 folio->_folio_dtor = compound_dtor;
1219 static inline void folio_set_compound_dtor(struct folio *folio,
1220 enum compound_dtor_id compound_dtor)
1222 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1223 folio->_folio_dtor = compound_dtor;
1226 void destroy_large_folio(struct folio *folio);
1228 static inline void set_compound_order(struct page *page, unsigned int order)
1230 struct folio *folio = (struct folio *)page;
1232 folio->_folio_order = order;
1234 folio->_folio_nr_pages = 1U << order;
1238 /* Returns the number of bytes in this potentially compound page. */
1239 static inline unsigned long page_size(struct page *page)
1241 return PAGE_SIZE << compound_order(page);
1244 /* Returns the number of bits needed for the number of bytes in a page */
1245 static inline unsigned int page_shift(struct page *page)
1247 return PAGE_SHIFT + compound_order(page);
1251 * thp_order - Order of a transparent huge page.
1252 * @page: Head page of a transparent huge page.
1254 static inline unsigned int thp_order(struct page *page)
1256 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1257 return compound_order(page);
1261 * thp_size - Size of a transparent huge page.
1262 * @page: Head page of a transparent huge page.
1264 * Return: Number of bytes in this page.
1266 static inline unsigned long thp_size(struct page *page)
1268 return PAGE_SIZE << thp_order(page);
1271 void free_compound_page(struct page *page);
1275 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1276 * servicing faults for write access. In the normal case, do always want
1277 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1278 * that do not have writing enabled, when used by access_process_vm.
1280 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1282 if (likely(vma->vm_flags & VM_WRITE))
1283 pte = pte_mkwrite(pte);
1287 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1288 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1290 vm_fault_t finish_fault(struct vm_fault *vmf);
1291 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1295 * Multiple processes may "see" the same page. E.g. for untouched
1296 * mappings of /dev/null, all processes see the same page full of
1297 * zeroes, and text pages of executables and shared libraries have
1298 * only one copy in memory, at most, normally.
1300 * For the non-reserved pages, page_count(page) denotes a reference count.
1301 * page_count() == 0 means the page is free. page->lru is then used for
1302 * freelist management in the buddy allocator.
1303 * page_count() > 0 means the page has been allocated.
1305 * Pages are allocated by the slab allocator in order to provide memory
1306 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1307 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1308 * unless a particular usage is carefully commented. (the responsibility of
1309 * freeing the kmalloc memory is the caller's, of course).
1311 * A page may be used by anyone else who does a __get_free_page().
1312 * In this case, page_count still tracks the references, and should only
1313 * be used through the normal accessor functions. The top bits of page->flags
1314 * and page->virtual store page management information, but all other fields
1315 * are unused and could be used privately, carefully. The management of this
1316 * page is the responsibility of the one who allocated it, and those who have
1317 * subsequently been given references to it.
1319 * The other pages (we may call them "pagecache pages") are completely
1320 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1321 * The following discussion applies only to them.
1323 * A pagecache page contains an opaque `private' member, which belongs to the
1324 * page's address_space. Usually, this is the address of a circular list of
1325 * the page's disk buffers. PG_private must be set to tell the VM to call
1326 * into the filesystem to release these pages.
1328 * A page may belong to an inode's memory mapping. In this case, page->mapping
1329 * is the pointer to the inode, and page->index is the file offset of the page,
1330 * in units of PAGE_SIZE.
1332 * If pagecache pages are not associated with an inode, they are said to be
1333 * anonymous pages. These may become associated with the swapcache, and in that
1334 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1336 * In either case (swapcache or inode backed), the pagecache itself holds one
1337 * reference to the page. Setting PG_private should also increment the
1338 * refcount. The each user mapping also has a reference to the page.
1340 * The pagecache pages are stored in a per-mapping radix tree, which is
1341 * rooted at mapping->i_pages, and indexed by offset.
1342 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1343 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1345 * All pagecache pages may be subject to I/O:
1346 * - inode pages may need to be read from disk,
1347 * - inode pages which have been modified and are MAP_SHARED may need
1348 * to be written back to the inode on disk,
1349 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1350 * modified may need to be swapped out to swap space and (later) to be read
1354 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1355 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1357 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1358 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1360 if (!static_branch_unlikely(&devmap_managed_key))
1362 if (!is_zone_device_page(page))
1364 return __put_devmap_managed_page_refs(page, refs);
1366 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1367 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1371 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1373 static inline bool put_devmap_managed_page(struct page *page)
1375 return put_devmap_managed_page_refs(page, 1);
1378 /* 127: arbitrary random number, small enough to assemble well */
1379 #define folio_ref_zero_or_close_to_overflow(folio) \
1380 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1383 * folio_get - Increment the reference count on a folio.
1384 * @folio: The folio.
1386 * Context: May be called in any context, as long as you know that
1387 * you have a refcount on the folio. If you do not already have one,
1388 * folio_try_get() may be the right interface for you to use.
1390 static inline void folio_get(struct folio *folio)
1392 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1393 folio_ref_inc(folio);
1396 static inline void get_page(struct page *page)
1398 folio_get(page_folio(page));
1401 static inline __must_check bool try_get_page(struct page *page)
1403 page = compound_head(page);
1404 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1411 * folio_put - Decrement the reference count on a folio.
1412 * @folio: The folio.
1414 * If the folio's reference count reaches zero, the memory will be
1415 * released back to the page allocator and may be used by another
1416 * allocation immediately. Do not access the memory or the struct folio
1417 * after calling folio_put() unless you can be sure that it wasn't the
1420 * Context: May be called in process or interrupt context, but not in NMI
1421 * context. May be called while holding a spinlock.
1423 static inline void folio_put(struct folio *folio)
1425 if (folio_put_testzero(folio))
1430 * folio_put_refs - Reduce the reference count on a folio.
1431 * @folio: The folio.
1432 * @refs: The amount to subtract from the folio's reference count.
1434 * If the folio's reference count reaches zero, the memory will be
1435 * released back to the page allocator and may be used by another
1436 * allocation immediately. Do not access the memory or the struct folio
1437 * after calling folio_put_refs() unless you can be sure that these weren't
1438 * the last references.
1440 * Context: May be called in process or interrupt context, but not in NMI
1441 * context. May be called while holding a spinlock.
1443 static inline void folio_put_refs(struct folio *folio, int refs)
1445 if (folio_ref_sub_and_test(folio, refs))
1450 * union release_pages_arg - an array of pages or folios
1452 * release_pages() releases a simple array of multiple pages, and
1453 * accepts various different forms of said page array: either
1454 * a regular old boring array of pages, an array of folios, or
1455 * an array of encoded page pointers.
1457 * The transparent union syntax for this kind of "any of these
1458 * argument types" is all kinds of ugly, so look away.
1461 struct page **pages;
1462 struct folio **folios;
1463 struct encoded_page **encoded_pages;
1464 } release_pages_arg __attribute__ ((__transparent_union__));
1466 void release_pages(release_pages_arg, int nr);
1469 * folios_put - Decrement the reference count on an array of folios.
1470 * @folios: The folios.
1471 * @nr: How many folios there are.
1473 * Like folio_put(), but for an array of folios. This is more efficient
1474 * than writing the loop yourself as it will optimise the locks which
1475 * need to be taken if the folios are freed.
1477 * Context: May be called in process or interrupt context, but not in NMI
1478 * context. May be called while holding a spinlock.
1480 static inline void folios_put(struct folio **folios, unsigned int nr)
1482 release_pages(folios, nr);
1485 static inline void put_page(struct page *page)
1487 struct folio *folio = page_folio(page);
1490 * For some devmap managed pages we need to catch refcount transition
1493 if (put_devmap_managed_page(&folio->page))
1499 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1500 * the page's refcount so that two separate items are tracked: the original page
1501 * reference count, and also a new count of how many pin_user_pages() calls were
1502 * made against the page. ("gup-pinned" is another term for the latter).
1504 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1505 * distinct from normal pages. As such, the unpin_user_page() call (and its
1506 * variants) must be used in order to release gup-pinned pages.
1510 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1511 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1512 * simpler, due to the fact that adding an even power of two to the page
1513 * refcount has the effect of using only the upper N bits, for the code that
1514 * counts up using the bias value. This means that the lower bits are left for
1515 * the exclusive use of the original code that increments and decrements by one
1516 * (or at least, by much smaller values than the bias value).
1518 * Of course, once the lower bits overflow into the upper bits (and this is
1519 * OK, because subtraction recovers the original values), then visual inspection
1520 * no longer suffices to directly view the separate counts. However, for normal
1521 * applications that don't have huge page reference counts, this won't be an
1524 * Locking: the lockless algorithm described in folio_try_get_rcu()
1525 * provides safe operation for get_user_pages(), page_mkclean() and
1526 * other calls that race to set up page table entries.
1528 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1530 void unpin_user_page(struct page *page);
1531 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1533 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1535 void unpin_user_pages(struct page **pages, unsigned long npages);
1537 static inline bool is_cow_mapping(vm_flags_t flags)
1539 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1543 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1546 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1547 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1548 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1549 * underlying memory if ptrace is active, so this is only possible if
1550 * ptrace does not apply. Note that there is no mprotect() to upgrade
1551 * write permissions later.
1553 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1557 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1558 #define SECTION_IN_PAGE_FLAGS
1562 * The identification function is mainly used by the buddy allocator for
1563 * determining if two pages could be buddies. We are not really identifying
1564 * the zone since we could be using the section number id if we do not have
1565 * node id available in page flags.
1566 * We only guarantee that it will return the same value for two combinable
1569 static inline int page_zone_id(struct page *page)
1571 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1574 #ifdef NODE_NOT_IN_PAGE_FLAGS
1575 extern int page_to_nid(const struct page *page);
1577 static inline int page_to_nid(const struct page *page)
1579 struct page *p = (struct page *)page;
1581 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1585 static inline int folio_nid(const struct folio *folio)
1587 return page_to_nid(&folio->page);
1590 #ifdef CONFIG_NUMA_BALANCING
1591 /* page access time bits needs to hold at least 4 seconds */
1592 #define PAGE_ACCESS_TIME_MIN_BITS 12
1593 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1594 #define PAGE_ACCESS_TIME_BUCKETS \
1595 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1597 #define PAGE_ACCESS_TIME_BUCKETS 0
1600 #define PAGE_ACCESS_TIME_MASK \
1601 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1603 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1605 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1608 static inline int cpupid_to_pid(int cpupid)
1610 return cpupid & LAST__PID_MASK;
1613 static inline int cpupid_to_cpu(int cpupid)
1615 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1618 static inline int cpupid_to_nid(int cpupid)
1620 return cpu_to_node(cpupid_to_cpu(cpupid));
1623 static inline bool cpupid_pid_unset(int cpupid)
1625 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1628 static inline bool cpupid_cpu_unset(int cpupid)
1630 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1633 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1635 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1638 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1639 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1640 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1642 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1645 static inline int page_cpupid_last(struct page *page)
1647 return page->_last_cpupid;
1649 static inline void page_cpupid_reset_last(struct page *page)
1651 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1654 static inline int page_cpupid_last(struct page *page)
1656 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1659 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1661 static inline void page_cpupid_reset_last(struct page *page)
1663 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1665 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1667 static inline int xchg_page_access_time(struct page *page, int time)
1671 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1672 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1674 #else /* !CONFIG_NUMA_BALANCING */
1675 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1677 return page_to_nid(page); /* XXX */
1680 static inline int xchg_page_access_time(struct page *page, int time)
1685 static inline int page_cpupid_last(struct page *page)
1687 return page_to_nid(page); /* XXX */
1690 static inline int cpupid_to_nid(int cpupid)
1695 static inline int cpupid_to_pid(int cpupid)
1700 static inline int cpupid_to_cpu(int cpupid)
1705 static inline int cpu_pid_to_cpupid(int nid, int pid)
1710 static inline bool cpupid_pid_unset(int cpupid)
1715 static inline void page_cpupid_reset_last(struct page *page)
1719 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1723 #endif /* CONFIG_NUMA_BALANCING */
1725 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1728 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1729 * setting tags for all pages to native kernel tag value 0xff, as the default
1730 * value 0x00 maps to 0xff.
1733 static inline u8 page_kasan_tag(const struct page *page)
1737 if (kasan_enabled()) {
1738 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1745 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1747 unsigned long old_flags, flags;
1749 if (!kasan_enabled())
1753 old_flags = READ_ONCE(page->flags);
1756 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1757 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1758 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1761 static inline void page_kasan_tag_reset(struct page *page)
1763 if (kasan_enabled())
1764 page_kasan_tag_set(page, 0xff);
1767 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1769 static inline u8 page_kasan_tag(const struct page *page)
1774 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1775 static inline void page_kasan_tag_reset(struct page *page) { }
1777 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1779 static inline struct zone *page_zone(const struct page *page)
1781 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1784 static inline pg_data_t *page_pgdat(const struct page *page)
1786 return NODE_DATA(page_to_nid(page));
1789 static inline struct zone *folio_zone(const struct folio *folio)
1791 return page_zone(&folio->page);
1794 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1796 return page_pgdat(&folio->page);
1799 #ifdef SECTION_IN_PAGE_FLAGS
1800 static inline void set_page_section(struct page *page, unsigned long section)
1802 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1803 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1806 static inline unsigned long page_to_section(const struct page *page)
1808 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1813 * folio_pfn - Return the Page Frame Number of a folio.
1814 * @folio: The folio.
1816 * A folio may contain multiple pages. The pages have consecutive
1817 * Page Frame Numbers.
1819 * Return: The Page Frame Number of the first page in the folio.
1821 static inline unsigned long folio_pfn(struct folio *folio)
1823 return page_to_pfn(&folio->page);
1826 static inline struct folio *pfn_folio(unsigned long pfn)
1828 return page_folio(pfn_to_page(pfn));
1832 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1833 * @folio: The folio.
1835 * This function checks if a folio has been pinned via a call to
1836 * a function in the pin_user_pages() family.
1838 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1839 * because it means "definitely not pinned for DMA", but true means "probably
1840 * pinned for DMA, but possibly a false positive due to having at least
1841 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1843 * False positives are OK, because: a) it's unlikely for a folio to
1844 * get that many refcounts, and b) all the callers of this routine are
1845 * expected to be able to deal gracefully with a false positive.
1847 * For large folios, the result will be exactly correct. That's because
1848 * we have more tracking data available: the _pincount field is used
1849 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1851 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1853 * Return: True, if it is likely that the page has been "dma-pinned".
1854 * False, if the page is definitely not dma-pinned.
1856 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1858 if (folio_test_large(folio))
1859 return atomic_read(&folio->_pincount) > 0;
1862 * folio_ref_count() is signed. If that refcount overflows, then
1863 * folio_ref_count() returns a negative value, and callers will avoid
1864 * further incrementing the refcount.
1866 * Here, for that overflow case, use the sign bit to count a little
1867 * bit higher via unsigned math, and thus still get an accurate result.
1869 return ((unsigned int)folio_ref_count(folio)) >=
1870 GUP_PIN_COUNTING_BIAS;
1873 static inline bool page_maybe_dma_pinned(struct page *page)
1875 return folio_maybe_dma_pinned(page_folio(page));
1879 * This should most likely only be called during fork() to see whether we
1880 * should break the cow immediately for an anon page on the src mm.
1882 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1884 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1887 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1889 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1892 return page_maybe_dma_pinned(page);
1895 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1896 #ifdef CONFIG_MIGRATION
1897 static inline bool is_longterm_pinnable_page(struct page *page)
1900 int mt = get_pageblock_migratetype(page);
1902 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1905 /* The zero page may always be pinned */
1906 if (is_zero_pfn(page_to_pfn(page)))
1909 /* Coherent device memory must always allow eviction. */
1910 if (is_device_coherent_page(page))
1913 /* Otherwise, non-movable zone pages can be pinned. */
1914 return !is_zone_movable_page(page);
1917 static inline bool is_longterm_pinnable_page(struct page *page)
1923 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1925 return is_longterm_pinnable_page(&folio->page);
1928 static inline void set_page_zone(struct page *page, enum zone_type zone)
1930 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1931 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1934 static inline void set_page_node(struct page *page, unsigned long node)
1936 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1937 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1940 static inline void set_page_links(struct page *page, enum zone_type zone,
1941 unsigned long node, unsigned long pfn)
1943 set_page_zone(page, zone);
1944 set_page_node(page, node);
1945 #ifdef SECTION_IN_PAGE_FLAGS
1946 set_page_section(page, pfn_to_section_nr(pfn));
1951 * folio_nr_pages - The number of pages in the folio.
1952 * @folio: The folio.
1954 * Return: A positive power of two.
1956 static inline long folio_nr_pages(struct folio *folio)
1958 if (!folio_test_large(folio))
1961 return folio->_folio_nr_pages;
1963 return 1L << folio->_folio_order;
1968 * compound_nr() returns the number of pages in this potentially compound
1969 * page. compound_nr() can be called on a tail page, and is defined to
1970 * return 1 in that case.
1972 static inline unsigned long compound_nr(struct page *page)
1974 struct folio *folio = (struct folio *)page;
1976 if (!test_bit(PG_head, &folio->flags))
1979 return folio->_folio_nr_pages;
1981 return 1L << folio->_folio_order;
1986 * thp_nr_pages - The number of regular pages in this huge page.
1987 * @page: The head page of a huge page.
1989 static inline int thp_nr_pages(struct page *page)
1991 return folio_nr_pages((struct folio *)page);
1995 * folio_next - Move to the next physical folio.
1996 * @folio: The folio we're currently operating on.
1998 * If you have physically contiguous memory which may span more than
1999 * one folio (eg a &struct bio_vec), use this function to move from one
2000 * folio to the next. Do not use it if the memory is only virtually
2001 * contiguous as the folios are almost certainly not adjacent to each
2002 * other. This is the folio equivalent to writing ``page++``.
2004 * Context: We assume that the folios are refcounted and/or locked at a
2005 * higher level and do not adjust the reference counts.
2006 * Return: The next struct folio.
2008 static inline struct folio *folio_next(struct folio *folio)
2010 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2014 * folio_shift - The size of the memory described by this folio.
2015 * @folio: The folio.
2017 * A folio represents a number of bytes which is a power-of-two in size.
2018 * This function tells you which power-of-two the folio is. See also
2019 * folio_size() and folio_order().
2021 * Context: The caller should have a reference on the folio to prevent
2022 * it from being split. It is not necessary for the folio to be locked.
2023 * Return: The base-2 logarithm of the size of this folio.
2025 static inline unsigned int folio_shift(struct folio *folio)
2027 return PAGE_SHIFT + folio_order(folio);
2031 * folio_size - The number of bytes in a folio.
2032 * @folio: The folio.
2034 * Context: The caller should have a reference on the folio to prevent
2035 * it from being split. It is not necessary for the folio to be locked.
2036 * Return: The number of bytes in this folio.
2038 static inline size_t folio_size(struct folio *folio)
2040 return PAGE_SIZE << folio_order(folio);
2044 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2045 * @folio: The folio.
2047 * folio_estimated_sharers() aims to serve as a function to efficiently
2048 * estimate the number of processes sharing a folio. This is done by
2049 * looking at the precise mapcount of the first subpage in the folio, and
2050 * assuming the other subpages are the same. This may not be true for large
2051 * folios. If you want exact mapcounts for exact calculations, look at
2052 * page_mapcount() or folio_total_mapcount().
2054 * Return: The estimated number of processes sharing a folio.
2056 static inline int folio_estimated_sharers(struct folio *folio)
2058 return page_mapcount(folio_page(folio, 0));
2061 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2062 static inline int arch_make_page_accessible(struct page *page)
2068 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2069 static inline int arch_make_folio_accessible(struct folio *folio)
2072 long i, nr = folio_nr_pages(folio);
2074 for (i = 0; i < nr; i++) {
2075 ret = arch_make_page_accessible(folio_page(folio, i));
2085 * Some inline functions in vmstat.h depend on page_zone()
2087 #include <linux/vmstat.h>
2089 static __always_inline void *lowmem_page_address(const struct page *page)
2091 return page_to_virt(page);
2094 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2095 #define HASHED_PAGE_VIRTUAL
2098 #if defined(WANT_PAGE_VIRTUAL)
2099 static inline void *page_address(const struct page *page)
2101 return page->virtual;
2103 static inline void set_page_address(struct page *page, void *address)
2105 page->virtual = address;
2107 #define page_address_init() do { } while(0)
2110 #if defined(HASHED_PAGE_VIRTUAL)
2111 void *page_address(const struct page *page);
2112 void set_page_address(struct page *page, void *virtual);
2113 void page_address_init(void);
2116 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2117 #define page_address(page) lowmem_page_address(page)
2118 #define set_page_address(page, address) do { } while(0)
2119 #define page_address_init() do { } while(0)
2122 static inline void *folio_address(const struct folio *folio)
2124 return page_address(&folio->page);
2127 extern void *page_rmapping(struct page *page);
2128 extern pgoff_t __page_file_index(struct page *page);
2131 * Return the pagecache index of the passed page. Regular pagecache pages
2132 * use ->index whereas swapcache pages use swp_offset(->private)
2134 static inline pgoff_t page_index(struct page *page)
2136 if (unlikely(PageSwapCache(page)))
2137 return __page_file_index(page);
2142 * Return true only if the page has been allocated with
2143 * ALLOC_NO_WATERMARKS and the low watermark was not
2144 * met implying that the system is under some pressure.
2146 static inline bool page_is_pfmemalloc(const struct page *page)
2149 * lru.next has bit 1 set if the page is allocated from the
2150 * pfmemalloc reserves. Callers may simply overwrite it if
2151 * they do not need to preserve that information.
2153 return (uintptr_t)page->lru.next & BIT(1);
2157 * Return true only if the folio has been allocated with
2158 * ALLOC_NO_WATERMARKS and the low watermark was not
2159 * met implying that the system is under some pressure.
2161 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2164 * lru.next has bit 1 set if the page is allocated from the
2165 * pfmemalloc reserves. Callers may simply overwrite it if
2166 * they do not need to preserve that information.
2168 return (uintptr_t)folio->lru.next & BIT(1);
2172 * Only to be called by the page allocator on a freshly allocated
2175 static inline void set_page_pfmemalloc(struct page *page)
2177 page->lru.next = (void *)BIT(1);
2180 static inline void clear_page_pfmemalloc(struct page *page)
2182 page->lru.next = NULL;
2186 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2188 extern void pagefault_out_of_memory(void);
2190 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2191 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2192 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2195 * Flags passed to show_mem() and show_free_areas() to suppress output in
2198 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2200 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2201 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2203 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2207 * Parameter block passed down to zap_pte_range in exceptional cases.
2209 struct zap_details {
2210 struct folio *single_folio; /* Locked folio to be unmapped */
2211 bool even_cows; /* Zap COWed private pages too? */
2212 zap_flags_t zap_flags; /* Extra flags for zapping */
2216 * Whether to drop the pte markers, for example, the uffd-wp information for
2217 * file-backed memory. This should only be specified when we will completely
2218 * drop the page in the mm, either by truncation or unmapping of the vma. By
2219 * default, the flag is not set.
2221 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2222 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2223 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2225 #ifdef CONFIG_SCHED_MM_CID
2226 void sched_mm_cid_before_execve(struct task_struct *t);
2227 void sched_mm_cid_after_execve(struct task_struct *t);
2228 void sched_mm_cid_fork(struct task_struct *t);
2229 void sched_mm_cid_exit_signals(struct task_struct *t);
2230 static inline int task_mm_cid(struct task_struct *t)
2235 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
2236 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
2237 static inline void sched_mm_cid_fork(struct task_struct *t) { }
2238 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
2239 static inline int task_mm_cid(struct task_struct *t)
2242 * Use the processor id as a fall-back when the mm cid feature is
2243 * disabled. This provides functional per-cpu data structure accesses
2244 * in user-space, althrough it won't provide the memory usage benefits.
2246 return raw_smp_processor_id();
2251 extern bool can_do_mlock(void);
2253 static inline bool can_do_mlock(void) { return false; }
2255 extern int user_shm_lock(size_t, struct ucounts *);
2256 extern void user_shm_unlock(size_t, struct ucounts *);
2258 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2260 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2262 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2265 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2266 unsigned long size);
2267 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2268 unsigned long size, struct zap_details *details);
2269 static inline void zap_vma_pages(struct vm_area_struct *vma)
2271 zap_page_range_single(vma, vma->vm_start,
2272 vma->vm_end - vma->vm_start, NULL);
2274 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2275 struct vm_area_struct *start_vma, unsigned long start,
2276 unsigned long end, bool mm_wr_locked);
2278 struct mmu_notifier_range;
2280 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2281 unsigned long end, unsigned long floor, unsigned long ceiling);
2283 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2284 int follow_pte(struct mm_struct *mm, unsigned long address,
2285 pte_t **ptepp, spinlock_t **ptlp);
2286 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2287 unsigned long *pfn);
2288 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2289 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2290 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2291 void *buf, int len, int write);
2293 extern void truncate_pagecache(struct inode *inode, loff_t new);
2294 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2295 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2296 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2297 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2300 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2301 unsigned long address, unsigned int flags,
2302 struct pt_regs *regs);
2303 extern int fixup_user_fault(struct mm_struct *mm,
2304 unsigned long address, unsigned int fault_flags,
2306 void unmap_mapping_pages(struct address_space *mapping,
2307 pgoff_t start, pgoff_t nr, bool even_cows);
2308 void unmap_mapping_range(struct address_space *mapping,
2309 loff_t const holebegin, loff_t const holelen, int even_cows);
2311 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2312 unsigned long address, unsigned int flags,
2313 struct pt_regs *regs)
2315 /* should never happen if there's no MMU */
2317 return VM_FAULT_SIGBUS;
2319 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2320 unsigned int fault_flags, bool *unlocked)
2322 /* should never happen if there's no MMU */
2326 static inline void unmap_mapping_pages(struct address_space *mapping,
2327 pgoff_t start, pgoff_t nr, bool even_cows) { }
2328 static inline void unmap_mapping_range(struct address_space *mapping,
2329 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2332 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2333 loff_t const holebegin, loff_t const holelen)
2335 unmap_mapping_range(mapping, holebegin, holelen, 0);
2338 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2339 void *buf, int len, unsigned int gup_flags);
2340 extern int access_remote_vm(struct mm_struct *mm, 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);
2345 long get_user_pages_remote(struct mm_struct *mm,
2346 unsigned long start, unsigned long nr_pages,
2347 unsigned int gup_flags, struct page **pages,
2348 struct vm_area_struct **vmas, int *locked);
2349 long pin_user_pages_remote(struct mm_struct *mm,
2350 unsigned long start, unsigned long nr_pages,
2351 unsigned int gup_flags, struct page **pages,
2352 struct vm_area_struct **vmas, int *locked);
2353 long get_user_pages(unsigned long start, unsigned long nr_pages,
2354 unsigned int gup_flags, struct page **pages,
2355 struct vm_area_struct **vmas);
2356 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2357 unsigned int gup_flags, struct page **pages,
2358 struct vm_area_struct **vmas);
2359 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2360 struct page **pages, unsigned int gup_flags);
2361 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2362 struct page **pages, unsigned int gup_flags);
2364 int get_user_pages_fast(unsigned long start, int nr_pages,
2365 unsigned int gup_flags, struct page **pages);
2366 int pin_user_pages_fast(unsigned long start, int nr_pages,
2367 unsigned int gup_flags, struct page **pages);
2369 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2370 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2371 struct task_struct *task, bool bypass_rlim);
2374 struct page *get_dump_page(unsigned long addr);
2376 bool folio_mark_dirty(struct folio *folio);
2377 bool set_page_dirty(struct page *page);
2378 int set_page_dirty_lock(struct page *page);
2380 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2382 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2383 unsigned long old_addr, struct vm_area_struct *new_vma,
2384 unsigned long new_addr, unsigned long len,
2385 bool need_rmap_locks);
2388 * Flags used by change_protection(). For now we make it a bitmap so
2389 * that we can pass in multiple flags just like parameters. However
2390 * for now all the callers are only use one of the flags at the same
2394 * Whether we should manually check if we can map individual PTEs writable,
2395 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2396 * PTEs automatically in a writable mapping.
2398 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2399 /* Whether this protection change is for NUMA hints */
2400 #define MM_CP_PROT_NUMA (1UL << 1)
2401 /* Whether this change is for write protecting */
2402 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2403 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2404 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2405 MM_CP_UFFD_WP_RESOLVE)
2407 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2408 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2411 * We want to check manually if we can change individual PTEs writable
2412 * if we can't do that automatically for all PTEs in a mapping. For
2413 * private mappings, that's always the case when we have write
2414 * permissions as we properly have to handle COW.
2416 if (vma->vm_flags & VM_SHARED)
2417 return vma_wants_writenotify(vma, vma->vm_page_prot);
2418 return !!(vma->vm_flags & VM_WRITE);
2421 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2423 extern long change_protection(struct mmu_gather *tlb,
2424 struct vm_area_struct *vma, unsigned long start,
2425 unsigned long end, unsigned long cp_flags);
2426 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2427 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2428 unsigned long start, unsigned long end, unsigned long newflags);
2431 * doesn't attempt to fault and will return short.
2433 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2434 unsigned int gup_flags, struct page **pages);
2436 static inline bool get_user_page_fast_only(unsigned long addr,
2437 unsigned int gup_flags, struct page **pagep)
2439 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2442 * per-process(per-mm_struct) statistics.
2444 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2446 return percpu_counter_read_positive(&mm->rss_stat[member]);
2449 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2451 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2453 percpu_counter_add(&mm->rss_stat[member], value);
2455 mm_trace_rss_stat(mm, member);
2458 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2460 percpu_counter_inc(&mm->rss_stat[member]);
2462 mm_trace_rss_stat(mm, member);
2465 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2467 percpu_counter_dec(&mm->rss_stat[member]);
2469 mm_trace_rss_stat(mm, member);
2472 /* Optimized variant when page is already known not to be PageAnon */
2473 static inline int mm_counter_file(struct page *page)
2475 if (PageSwapBacked(page))
2476 return MM_SHMEMPAGES;
2477 return MM_FILEPAGES;
2480 static inline int mm_counter(struct page *page)
2483 return MM_ANONPAGES;
2484 return mm_counter_file(page);
2487 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2489 return get_mm_counter(mm, MM_FILEPAGES) +
2490 get_mm_counter(mm, MM_ANONPAGES) +
2491 get_mm_counter(mm, MM_SHMEMPAGES);
2494 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2496 return max(mm->hiwater_rss, get_mm_rss(mm));
2499 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2501 return max(mm->hiwater_vm, mm->total_vm);
2504 static inline void update_hiwater_rss(struct mm_struct *mm)
2506 unsigned long _rss = get_mm_rss(mm);
2508 if ((mm)->hiwater_rss < _rss)
2509 (mm)->hiwater_rss = _rss;
2512 static inline void update_hiwater_vm(struct mm_struct *mm)
2514 if (mm->hiwater_vm < mm->total_vm)
2515 mm->hiwater_vm = mm->total_vm;
2518 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2520 mm->hiwater_rss = get_mm_rss(mm);
2523 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2524 struct mm_struct *mm)
2526 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2528 if (*maxrss < hiwater_rss)
2529 *maxrss = hiwater_rss;
2532 #if defined(SPLIT_RSS_COUNTING)
2533 void sync_mm_rss(struct mm_struct *mm);
2535 static inline void sync_mm_rss(struct mm_struct *mm)
2540 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2541 static inline int pte_special(pte_t pte)
2546 static inline pte_t pte_mkspecial(pte_t pte)
2552 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2553 static inline int pte_devmap(pte_t pte)
2559 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2561 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2565 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2569 #ifdef __PAGETABLE_P4D_FOLDED
2570 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2571 unsigned long address)
2576 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2579 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2580 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2581 unsigned long address)
2585 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2586 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2589 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2591 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2593 if (mm_pud_folded(mm))
2595 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2598 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2600 if (mm_pud_folded(mm))
2602 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2606 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2607 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2608 unsigned long address)
2613 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2614 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2617 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2619 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2621 if (mm_pmd_folded(mm))
2623 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2626 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2628 if (mm_pmd_folded(mm))
2630 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2635 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2637 atomic_long_set(&mm->pgtables_bytes, 0);
2640 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2642 return atomic_long_read(&mm->pgtables_bytes);
2645 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2647 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2650 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2652 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2656 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2657 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2662 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2663 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2666 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2667 int __pte_alloc_kernel(pmd_t *pmd);
2669 #if defined(CONFIG_MMU)
2671 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2672 unsigned long address)
2674 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2675 NULL : p4d_offset(pgd, address);
2678 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2679 unsigned long address)
2681 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2682 NULL : pud_offset(p4d, address);
2685 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2687 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2688 NULL: pmd_offset(pud, address);
2690 #endif /* CONFIG_MMU */
2692 #if USE_SPLIT_PTE_PTLOCKS
2693 #if ALLOC_SPLIT_PTLOCKS
2694 void __init ptlock_cache_init(void);
2695 extern bool ptlock_alloc(struct page *page);
2696 extern void ptlock_free(struct page *page);
2698 static inline spinlock_t *ptlock_ptr(struct page *page)
2702 #else /* ALLOC_SPLIT_PTLOCKS */
2703 static inline void ptlock_cache_init(void)
2707 static inline bool ptlock_alloc(struct page *page)
2712 static inline void ptlock_free(struct page *page)
2716 static inline spinlock_t *ptlock_ptr(struct page *page)
2720 #endif /* ALLOC_SPLIT_PTLOCKS */
2722 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2724 return ptlock_ptr(pmd_page(*pmd));
2727 static inline bool ptlock_init(struct page *page)
2730 * prep_new_page() initialize page->private (and therefore page->ptl)
2731 * with 0. Make sure nobody took it in use in between.
2733 * It can happen if arch try to use slab for page table allocation:
2734 * slab code uses page->slab_cache, which share storage with page->ptl.
2736 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2737 if (!ptlock_alloc(page))
2739 spin_lock_init(ptlock_ptr(page));
2743 #else /* !USE_SPLIT_PTE_PTLOCKS */
2745 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2747 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2749 return &mm->page_table_lock;
2751 static inline void ptlock_cache_init(void) {}
2752 static inline bool ptlock_init(struct page *page) { return true; }
2753 static inline void ptlock_free(struct page *page) {}
2754 #endif /* USE_SPLIT_PTE_PTLOCKS */
2756 static inline bool pgtable_pte_page_ctor(struct page *page)
2758 if (!ptlock_init(page))
2760 __SetPageTable(page);
2761 inc_lruvec_page_state(page, NR_PAGETABLE);
2765 static inline void pgtable_pte_page_dtor(struct page *page)
2768 __ClearPageTable(page);
2769 dec_lruvec_page_state(page, NR_PAGETABLE);
2772 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2774 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2775 pte_t *__pte = pte_offset_map(pmd, address); \
2781 #define pte_unmap_unlock(pte, ptl) do { \
2786 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2788 #define pte_alloc_map(mm, pmd, address) \
2789 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2791 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2792 (pte_alloc(mm, pmd) ? \
2793 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2795 #define pte_alloc_kernel(pmd, address) \
2796 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2797 NULL: pte_offset_kernel(pmd, address))
2799 #if USE_SPLIT_PMD_PTLOCKS
2801 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2803 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2804 return virt_to_page((void *)((unsigned long) pmd & mask));
2807 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2809 return ptlock_ptr(pmd_pgtable_page(pmd));
2812 static inline bool pmd_ptlock_init(struct page *page)
2814 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2815 page->pmd_huge_pte = NULL;
2817 return ptlock_init(page);
2820 static inline void pmd_ptlock_free(struct page *page)
2822 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2823 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2828 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2832 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2834 return &mm->page_table_lock;
2837 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2838 static inline void pmd_ptlock_free(struct page *page) {}
2840 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2844 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2846 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2851 static inline bool pgtable_pmd_page_ctor(struct page *page)
2853 if (!pmd_ptlock_init(page))
2855 __SetPageTable(page);
2856 inc_lruvec_page_state(page, NR_PAGETABLE);
2860 static inline void pgtable_pmd_page_dtor(struct page *page)
2862 pmd_ptlock_free(page);
2863 __ClearPageTable(page);
2864 dec_lruvec_page_state(page, NR_PAGETABLE);
2868 * No scalability reason to split PUD locks yet, but follow the same pattern
2869 * as the PMD locks to make it easier if we decide to. The VM should not be
2870 * considered ready to switch to split PUD locks yet; there may be places
2871 * which need to be converted from page_table_lock.
2873 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2875 return &mm->page_table_lock;
2878 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2880 spinlock_t *ptl = pud_lockptr(mm, pud);
2886 extern void __init pagecache_init(void);
2887 extern void free_initmem(void);
2890 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2891 * into the buddy system. The freed pages will be poisoned with pattern
2892 * "poison" if it's within range [0, UCHAR_MAX].
2893 * Return pages freed into the buddy system.
2895 extern unsigned long free_reserved_area(void *start, void *end,
2896 int poison, const char *s);
2898 extern void adjust_managed_page_count(struct page *page, long count);
2900 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2902 /* Free the reserved page into the buddy system, so it gets managed. */
2903 static inline void free_reserved_page(struct page *page)
2905 ClearPageReserved(page);
2906 init_page_count(page);
2908 adjust_managed_page_count(page, 1);
2910 #define free_highmem_page(page) free_reserved_page(page)
2912 static inline void mark_page_reserved(struct page *page)
2914 SetPageReserved(page);
2915 adjust_managed_page_count(page, -1);
2919 * Default method to free all the __init memory into the buddy system.
2920 * The freed pages will be poisoned with pattern "poison" if it's within
2921 * range [0, UCHAR_MAX].
2922 * Return pages freed into the buddy system.
2924 static inline unsigned long free_initmem_default(int poison)
2926 extern char __init_begin[], __init_end[];
2928 return free_reserved_area(&__init_begin, &__init_end,
2929 poison, "unused kernel image (initmem)");
2932 static inline unsigned long get_num_physpages(void)
2935 unsigned long phys_pages = 0;
2937 for_each_online_node(nid)
2938 phys_pages += node_present_pages(nid);
2944 * Using memblock node mappings, an architecture may initialise its
2945 * zones, allocate the backing mem_map and account for memory holes in an
2946 * architecture independent manner.
2948 * An architecture is expected to register range of page frames backed by
2949 * physical memory with memblock_add[_node]() before calling
2950 * free_area_init() passing in the PFN each zone ends at. At a basic
2951 * usage, an architecture is expected to do something like
2953 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2955 * for_each_valid_physical_page_range()
2956 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2957 * free_area_init(max_zone_pfns);
2959 void free_area_init(unsigned long *max_zone_pfn);
2960 unsigned long node_map_pfn_alignment(void);
2961 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2962 unsigned long end_pfn);
2963 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2964 unsigned long end_pfn);
2965 extern void get_pfn_range_for_nid(unsigned int nid,
2966 unsigned long *start_pfn, unsigned long *end_pfn);
2969 static inline int early_pfn_to_nid(unsigned long pfn)
2974 /* please see mm/page_alloc.c */
2975 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2978 extern void set_dma_reserve(unsigned long new_dma_reserve);
2979 extern void memmap_init_range(unsigned long, int, unsigned long,
2980 unsigned long, unsigned long, enum meminit_context,
2981 struct vmem_altmap *, int migratetype);
2982 extern void setup_per_zone_wmarks(void);
2983 extern void calculate_min_free_kbytes(void);
2984 extern int __meminit init_per_zone_wmark_min(void);
2985 extern void mem_init(void);
2986 extern void __init mmap_init(void);
2988 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2989 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2991 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2993 extern long si_mem_available(void);
2994 extern void si_meminfo(struct sysinfo * val);
2995 extern void si_meminfo_node(struct sysinfo *val, int nid);
2996 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2997 extern unsigned long arch_reserved_kernel_pages(void);
3000 extern __printf(3, 4)
3001 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3003 extern void setup_per_cpu_pageset(void);
3006 extern int min_free_kbytes;
3007 extern int watermark_boost_factor;
3008 extern int watermark_scale_factor;
3009 extern bool arch_has_descending_max_zone_pfns(void);
3012 extern atomic_long_t mmap_pages_allocated;
3013 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3015 /* interval_tree.c */
3016 void vma_interval_tree_insert(struct vm_area_struct *node,
3017 struct rb_root_cached *root);
3018 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3019 struct vm_area_struct *prev,
3020 struct rb_root_cached *root);
3021 void vma_interval_tree_remove(struct vm_area_struct *node,
3022 struct rb_root_cached *root);
3023 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3024 unsigned long start, unsigned long last);
3025 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3026 unsigned long start, unsigned long last);
3028 #define vma_interval_tree_foreach(vma, root, start, last) \
3029 for (vma = vma_interval_tree_iter_first(root, start, last); \
3030 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3032 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3033 struct rb_root_cached *root);
3034 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3035 struct rb_root_cached *root);
3036 struct anon_vma_chain *
3037 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3038 unsigned long start, unsigned long last);
3039 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3040 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3041 #ifdef CONFIG_DEBUG_VM_RB
3042 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3045 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3046 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3047 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3050 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3051 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3052 unsigned long start, unsigned long end, pgoff_t pgoff,
3053 struct vm_area_struct *next);
3054 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3055 unsigned long start, unsigned long end, pgoff_t pgoff);
3056 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3057 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3058 unsigned long end, unsigned long vm_flags, struct anon_vma *,
3059 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3060 struct anon_vma_name *);
3061 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3062 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3063 unsigned long addr, int new_below);
3064 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3065 unsigned long addr, int new_below);
3066 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3067 extern void unlink_file_vma(struct vm_area_struct *);
3068 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3069 unsigned long addr, unsigned long len, pgoff_t pgoff,
3070 bool *need_rmap_locks);
3071 extern void exit_mmap(struct mm_struct *);
3073 static inline int check_data_rlimit(unsigned long rlim,
3075 unsigned long start,
3076 unsigned long end_data,
3077 unsigned long start_data)
3079 if (rlim < RLIM_INFINITY) {
3080 if (((new - start) + (end_data - start_data)) > rlim)
3087 extern int mm_take_all_locks(struct mm_struct *mm);
3088 extern void mm_drop_all_locks(struct mm_struct *mm);
3090 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3091 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3092 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3093 extern struct file *get_task_exe_file(struct task_struct *task);
3095 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3096 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3098 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3099 const struct vm_special_mapping *sm);
3100 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3101 unsigned long addr, unsigned long len,
3102 unsigned long flags,
3103 const struct vm_special_mapping *spec);
3104 /* This is an obsolete alternative to _install_special_mapping. */
3105 extern int install_special_mapping(struct mm_struct *mm,
3106 unsigned long addr, unsigned long len,
3107 unsigned long flags, struct page **pages);
3109 unsigned long randomize_stack_top(unsigned long stack_top);
3110 unsigned long randomize_page(unsigned long start, unsigned long range);
3112 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3114 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3115 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3116 struct list_head *uf);
3117 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3118 unsigned long len, unsigned long prot, unsigned long flags,
3119 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
3120 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3121 unsigned long start, size_t len, struct list_head *uf,
3123 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3124 struct list_head *uf);
3125 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3128 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3129 unsigned long start, unsigned long end,
3130 struct list_head *uf, bool downgrade);
3131 extern int __mm_populate(unsigned long addr, unsigned long len,
3133 static inline void mm_populate(unsigned long addr, unsigned long len)
3136 (void) __mm_populate(addr, len, 1);
3139 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3142 /* These take the mm semaphore themselves */
3143 extern int __must_check vm_brk(unsigned long, unsigned long);
3144 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3145 extern int vm_munmap(unsigned long, size_t);
3146 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3147 unsigned long, unsigned long,
3148 unsigned long, unsigned long);
3150 struct vm_unmapped_area_info {
3151 #define VM_UNMAPPED_AREA_TOPDOWN 1
3152 unsigned long flags;
3153 unsigned long length;
3154 unsigned long low_limit;
3155 unsigned long high_limit;
3156 unsigned long align_mask;
3157 unsigned long align_offset;
3160 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3163 extern void truncate_inode_pages(struct address_space *, loff_t);
3164 extern void truncate_inode_pages_range(struct address_space *,
3165 loff_t lstart, loff_t lend);
3166 extern void truncate_inode_pages_final(struct address_space *);
3168 /* generic vm_area_ops exported for stackable file systems */
3169 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3170 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3171 pgoff_t start_pgoff, pgoff_t end_pgoff);
3172 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3174 extern unsigned long stack_guard_gap;
3175 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3176 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
3178 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3179 extern int expand_downwards(struct vm_area_struct *vma,
3180 unsigned long address);
3182 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
3184 #define expand_upwards(vma, address) (0)
3187 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3188 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3189 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3190 struct vm_area_struct **pprev);
3193 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3194 * NULL if none. Assume start_addr < end_addr.
3196 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3197 unsigned long start_addr, unsigned long end_addr);
3200 * vma_lookup() - Find a VMA at a specific address
3201 * @mm: The process address space.
3202 * @addr: The user address.
3204 * Return: The vm_area_struct at the given address, %NULL otherwise.
3207 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3209 return mtree_load(&mm->mm_mt, addr);
3212 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3214 unsigned long vm_start = vma->vm_start;
3216 if (vma->vm_flags & VM_GROWSDOWN) {
3217 vm_start -= stack_guard_gap;
3218 if (vm_start > vma->vm_start)
3224 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3226 unsigned long vm_end = vma->vm_end;
3228 if (vma->vm_flags & VM_GROWSUP) {
3229 vm_end += stack_guard_gap;
3230 if (vm_end < vma->vm_end)
3231 vm_end = -PAGE_SIZE;
3236 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3238 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3241 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3242 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3243 unsigned long vm_start, unsigned long vm_end)
3245 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3247 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3253 static inline bool range_in_vma(struct vm_area_struct *vma,
3254 unsigned long start, unsigned long end)
3256 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3260 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3261 void vma_set_page_prot(struct vm_area_struct *vma);
3263 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3267 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3269 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3273 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3275 #ifdef CONFIG_NUMA_BALANCING
3276 unsigned long change_prot_numa(struct vm_area_struct *vma,
3277 unsigned long start, unsigned long end);
3280 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3281 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3282 unsigned long pfn, unsigned long size, pgprot_t);
3283 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3284 unsigned long pfn, unsigned long size, pgprot_t prot);
3285 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3286 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3287 struct page **pages, unsigned long *num);
3288 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3290 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3292 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3294 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3295 unsigned long pfn, pgprot_t pgprot);
3296 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3298 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3299 unsigned long addr, pfn_t pfn);
3300 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3302 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3303 unsigned long addr, struct page *page)
3305 int err = vm_insert_page(vma, addr, page);
3308 return VM_FAULT_OOM;
3309 if (err < 0 && err != -EBUSY)
3310 return VM_FAULT_SIGBUS;
3312 return VM_FAULT_NOPAGE;
3315 #ifndef io_remap_pfn_range
3316 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3317 unsigned long addr, unsigned long pfn,
3318 unsigned long size, pgprot_t prot)
3320 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3324 static inline vm_fault_t vmf_error(int err)
3327 return VM_FAULT_OOM;
3328 return VM_FAULT_SIGBUS;
3331 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3332 unsigned int foll_flags);
3334 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3336 if (vm_fault & VM_FAULT_OOM)
3338 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3339 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3340 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3346 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3347 * a (NUMA hinting) fault is required.
3349 static inline bool gup_can_follow_protnone(unsigned int flags)
3352 * FOLL_FORCE has to be able to make progress even if the VMA is
3353 * inaccessible. Further, FOLL_FORCE access usually does not represent
3354 * application behaviour and we should avoid triggering NUMA hinting
3357 return flags & FOLL_FORCE;
3360 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3361 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3362 unsigned long size, pte_fn_t fn, void *data);
3363 extern int apply_to_existing_page_range(struct mm_struct *mm,
3364 unsigned long address, unsigned long size,
3365 pte_fn_t fn, void *data);
3367 #ifdef CONFIG_PAGE_POISONING
3368 extern void __kernel_poison_pages(struct page *page, int numpages);
3369 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3370 extern bool _page_poisoning_enabled_early;
3371 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3372 static inline bool page_poisoning_enabled(void)
3374 return _page_poisoning_enabled_early;
3377 * For use in fast paths after init_mem_debugging() has run, or when a
3378 * false negative result is not harmful when called too early.
3380 static inline bool page_poisoning_enabled_static(void)
3382 return static_branch_unlikely(&_page_poisoning_enabled);
3384 static inline void kernel_poison_pages(struct page *page, int numpages)
3386 if (page_poisoning_enabled_static())
3387 __kernel_poison_pages(page, numpages);
3389 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3391 if (page_poisoning_enabled_static())
3392 __kernel_unpoison_pages(page, numpages);
3395 static inline bool page_poisoning_enabled(void) { return false; }
3396 static inline bool page_poisoning_enabled_static(void) { return false; }
3397 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3398 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3399 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3402 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3403 static inline bool want_init_on_alloc(gfp_t flags)
3405 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3408 return flags & __GFP_ZERO;
3411 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3412 static inline bool want_init_on_free(void)
3414 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3418 extern bool _debug_pagealloc_enabled_early;
3419 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3421 static inline bool debug_pagealloc_enabled(void)
3423 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3424 _debug_pagealloc_enabled_early;
3428 * For use in fast paths after init_debug_pagealloc() has run, or when a
3429 * false negative result is not harmful when called too early.
3431 static inline bool debug_pagealloc_enabled_static(void)
3433 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3436 return static_branch_unlikely(&_debug_pagealloc_enabled);
3439 #ifdef CONFIG_DEBUG_PAGEALLOC
3441 * To support DEBUG_PAGEALLOC architecture must ensure that
3442 * __kernel_map_pages() never fails
3444 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3446 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3448 if (debug_pagealloc_enabled_static())
3449 __kernel_map_pages(page, numpages, 1);
3452 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3454 if (debug_pagealloc_enabled_static())
3455 __kernel_map_pages(page, numpages, 0);
3457 #else /* CONFIG_DEBUG_PAGEALLOC */
3458 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3459 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3460 #endif /* CONFIG_DEBUG_PAGEALLOC */
3462 #ifdef __HAVE_ARCH_GATE_AREA
3463 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3464 extern int in_gate_area_no_mm(unsigned long addr);
3465 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3467 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3471 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3472 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3476 #endif /* __HAVE_ARCH_GATE_AREA */
3478 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3480 #ifdef CONFIG_SYSCTL
3481 extern int sysctl_drop_caches;
3482 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3486 void drop_slab(void);
3489 #define randomize_va_space 0
3491 extern int randomize_va_space;
3494 const char * arch_vma_name(struct vm_area_struct *vma);
3496 void print_vma_addr(char *prefix, unsigned long rip);
3498 static inline void print_vma_addr(char *prefix, unsigned long rip)
3503 void *sparse_buffer_alloc(unsigned long size);
3504 struct page * __populate_section_memmap(unsigned long pfn,
3505 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3506 struct dev_pagemap *pgmap);
3507 void pmd_init(void *addr);
3508 void pud_init(void *addr);
3509 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3510 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3511 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3512 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3513 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3514 struct vmem_altmap *altmap, struct page *reuse);
3515 void *vmemmap_alloc_block(unsigned long size, int node);
3517 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3518 struct vmem_altmap *altmap);
3519 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3520 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3521 unsigned long addr, unsigned long next);
3522 int vmemmap_check_pmd(pmd_t *pmd, int node,
3523 unsigned long addr, unsigned long next);
3524 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3525 int node, struct vmem_altmap *altmap);
3526 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3527 int node, struct vmem_altmap *altmap);
3528 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3529 struct vmem_altmap *altmap);
3530 void vmemmap_populate_print_last(void);
3531 #ifdef CONFIG_MEMORY_HOTPLUG
3532 void vmemmap_free(unsigned long start, unsigned long end,
3533 struct vmem_altmap *altmap);
3535 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3536 unsigned long nr_pages);
3539 MF_COUNT_INCREASED = 1 << 0,
3540 MF_ACTION_REQUIRED = 1 << 1,
3541 MF_MUST_KILL = 1 << 2,
3542 MF_SOFT_OFFLINE = 1 << 3,
3543 MF_UNPOISON = 1 << 4,
3544 MF_SW_SIMULATED = 1 << 5,
3545 MF_NO_RETRY = 1 << 6,
3547 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3548 unsigned long count, int mf_flags);
3549 extern int memory_failure(unsigned long pfn, int flags);
3550 extern void memory_failure_queue_kick(int cpu);
3551 extern int unpoison_memory(unsigned long pfn);
3552 extern int sysctl_memory_failure_early_kill;
3553 extern int sysctl_memory_failure_recovery;
3554 extern void shake_page(struct page *p);
3555 extern atomic_long_t num_poisoned_pages __read_mostly;
3556 extern int soft_offline_page(unsigned long pfn, int flags);
3557 #ifdef CONFIG_MEMORY_FAILURE
3558 extern void memory_failure_queue(unsigned long pfn, int flags);
3559 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3560 bool *migratable_cleared);
3561 void num_poisoned_pages_inc(unsigned long pfn);
3562 void num_poisoned_pages_sub(unsigned long pfn, long i);
3564 static inline void memory_failure_queue(unsigned long pfn, int flags)
3568 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3569 bool *migratable_cleared)
3574 static inline void num_poisoned_pages_inc(unsigned long pfn)
3578 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3583 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3584 extern void memblk_nr_poison_inc(unsigned long pfn);
3585 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3587 static inline void memblk_nr_poison_inc(unsigned long pfn)
3591 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3596 #ifndef arch_memory_failure
3597 static inline int arch_memory_failure(unsigned long pfn, int flags)
3603 #ifndef arch_is_platform_page
3604 static inline bool arch_is_platform_page(u64 paddr)
3611 * Error handlers for various types of pages.
3614 MF_IGNORED, /* Error: cannot be handled */
3615 MF_FAILED, /* Error: handling failed */
3616 MF_DELAYED, /* Will be handled later */
3617 MF_RECOVERED, /* Successfully recovered */
3620 enum mf_action_page_type {
3622 MF_MSG_KERNEL_HIGH_ORDER,
3624 MF_MSG_DIFFERENT_COMPOUND,
3627 MF_MSG_UNMAP_FAILED,
3628 MF_MSG_DIRTY_SWAPCACHE,
3629 MF_MSG_CLEAN_SWAPCACHE,
3630 MF_MSG_DIRTY_MLOCKED_LRU,
3631 MF_MSG_CLEAN_MLOCKED_LRU,
3632 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3633 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3636 MF_MSG_TRUNCATED_LRU,
3644 * Sysfs entries for memory failure handling statistics.
3646 extern const struct attribute_group memory_failure_attr_group;
3648 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3649 extern void clear_huge_page(struct page *page,
3650 unsigned long addr_hint,
3651 unsigned int pages_per_huge_page);
3652 extern void copy_user_huge_page(struct page *dst, struct page *src,
3653 unsigned long addr_hint,
3654 struct vm_area_struct *vma,
3655 unsigned int pages_per_huge_page);
3656 extern long copy_huge_page_from_user(struct page *dst_page,
3657 const void __user *usr_src,
3658 unsigned int pages_per_huge_page,
3659 bool allow_pagefault);
3662 * vma_is_special_huge - Are transhuge page-table entries considered special?
3663 * @vma: Pointer to the struct vm_area_struct to consider
3665 * Whether transhuge page-table entries are considered "special" following
3666 * the definition in vm_normal_page().
3668 * Return: true if transhuge page-table entries should be considered special,
3671 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3673 return vma_is_dax(vma) || (vma->vm_file &&
3674 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3677 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3679 #ifdef CONFIG_DEBUG_PAGEALLOC
3680 extern unsigned int _debug_guardpage_minorder;
3681 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3683 static inline unsigned int debug_guardpage_minorder(void)
3685 return _debug_guardpage_minorder;
3688 static inline bool debug_guardpage_enabled(void)
3690 return static_branch_unlikely(&_debug_guardpage_enabled);
3693 static inline bool page_is_guard(struct page *page)
3695 if (!debug_guardpage_enabled())
3698 return PageGuard(page);
3701 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3702 static inline bool debug_guardpage_enabled(void) { return false; }
3703 static inline bool page_is_guard(struct page *page) { return false; }
3704 #endif /* CONFIG_DEBUG_PAGEALLOC */
3706 #if MAX_NUMNODES > 1
3707 void __init setup_nr_node_ids(void);
3709 static inline void setup_nr_node_ids(void) {}
3712 extern int memcmp_pages(struct page *page1, struct page *page2);
3714 static inline int pages_identical(struct page *page1, struct page *page2)
3716 return !memcmp_pages(page1, page2);
3719 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3720 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3721 pgoff_t first_index, pgoff_t nr,
3722 pgoff_t bitmap_pgoff,
3723 unsigned long *bitmap,
3727 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3728 pgoff_t first_index, pgoff_t nr);
3731 extern int sysctl_nr_trim_pages;
3733 #ifdef CONFIG_PRINTK
3734 void mem_dump_obj(void *object);
3736 static inline void mem_dump_obj(void *object) {}
3740 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3741 * @seals: the seals to check
3742 * @vma: the vma to operate on
3744 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3745 * the vma flags. Return 0 if check pass, or <0 for errors.
3747 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3749 if (seals & F_SEAL_FUTURE_WRITE) {
3751 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3752 * "future write" seal active.
3754 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3758 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3759 * MAP_SHARED and read-only, take care to not allow mprotect to
3760 * revert protections on such mappings. Do this only for shared
3761 * mappings. For private mappings, don't need to mask
3762 * VM_MAYWRITE as we still want them to be COW-writable.
3764 if (vma->vm_flags & VM_SHARED)
3765 vm_flags_clear(vma, VM_MAYWRITE);
3771 #ifdef CONFIG_ANON_VMA_NAME
3772 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3773 unsigned long len_in,
3774 struct anon_vma_name *anon_name);
3777 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3778 unsigned long len_in, struct anon_vma_name *anon_name) {
3783 #endif /* _LINUX_MM_H */