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 init_mm_internals(void);
43 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
44 extern unsigned long max_mapnr;
46 static inline void set_max_mapnr(unsigned long limit)
51 static inline void set_max_mapnr(unsigned long limit) { }
54 extern atomic_long_t _totalram_pages;
55 static inline unsigned long totalram_pages(void)
57 return (unsigned long)atomic_long_read(&_totalram_pages);
60 static inline void totalram_pages_inc(void)
62 atomic_long_inc(&_totalram_pages);
65 static inline void totalram_pages_dec(void)
67 atomic_long_dec(&_totalram_pages);
70 static inline void totalram_pages_add(long count)
72 atomic_long_add(count, &_totalram_pages);
75 extern void * high_memory;
76 extern int page_cluster;
77 extern const int page_cluster_max;
80 extern int sysctl_legacy_va_layout;
82 #define sysctl_legacy_va_layout 0
85 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
86 extern const int mmap_rnd_bits_min;
87 extern const int mmap_rnd_bits_max;
88 extern int mmap_rnd_bits __read_mostly;
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
91 extern const int mmap_rnd_compat_bits_min;
92 extern const int mmap_rnd_compat_bits_max;
93 extern int mmap_rnd_compat_bits __read_mostly;
97 #include <asm/processor.h>
100 * Architectures that support memory tagging (assigning tags to memory regions,
101 * embedding these tags into addresses that point to these memory regions, and
102 * checking that the memory and the pointer tags match on memory accesses)
103 * redefine this macro to strip tags from pointers.
104 * It's defined as noop for architectures that don't support memory tagging.
106 #ifndef untagged_addr
107 #define untagged_addr(addr) (addr)
111 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
115 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 #define lm_alias(x) __va(__pa_symbol(x))
123 * To prevent common memory management code establishing
124 * a zero page mapping on a read fault.
125 * This macro should be defined within <asm/pgtable.h>.
126 * s390 does this to prevent multiplexing of hardware bits
127 * related to the physical page in case of virtualization.
129 #ifndef mm_forbids_zeropage
130 #define mm_forbids_zeropage(X) (0)
134 * On some architectures it is expensive to call memset() for small sizes.
135 * If an architecture decides to implement their own version of
136 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
137 * define their own version of this macro in <asm/pgtable.h>
139 #if BITS_PER_LONG == 64
140 /* This function must be updated when the size of struct page grows above 80
141 * or reduces below 56. The idea that compiler optimizes out switch()
142 * statement, and only leaves move/store instructions. Also the compiler can
143 * combine write statements if they are both assignments and can be reordered,
144 * this can result in several of the writes here being dropped.
146 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
147 static inline void __mm_zero_struct_page(struct page *page)
149 unsigned long *_pp = (void *)page;
151 /* Check that struct page is either 56, 64, 72, or 80 bytes */
152 BUILD_BUG_ON(sizeof(struct page) & 7);
153 BUILD_BUG_ON(sizeof(struct page) < 56);
154 BUILD_BUG_ON(sizeof(struct page) > 80);
156 switch (sizeof(struct page)) {
177 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
181 * Default maximum number of active map areas, this limits the number of vmas
182 * per mm struct. Users can overwrite this number by sysctl but there is a
185 * When a program's coredump is generated as ELF format, a section is created
186 * per a vma. In ELF, the number of sections is represented in unsigned short.
187 * This means the number of sections should be smaller than 65535 at coredump.
188 * Because the kernel adds some informative sections to a image of program at
189 * generating coredump, we need some margin. The number of extra sections is
190 * 1-3 now and depends on arch. We use "5" as safe margin, here.
192 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
193 * not a hard limit any more. Although some userspace tools can be surprised by
196 #define MAPCOUNT_ELF_CORE_MARGIN (5)
197 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
199 extern int sysctl_max_map_count;
201 extern unsigned long sysctl_user_reserve_kbytes;
202 extern unsigned long sysctl_admin_reserve_kbytes;
204 extern int sysctl_overcommit_memory;
205 extern int sysctl_overcommit_ratio;
206 extern unsigned long sysctl_overcommit_kbytes;
208 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
210 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
212 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
215 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
216 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
217 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
219 #define nth_page(page,n) ((page) + (n))
220 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
223 /* to align the pointer to the (next) page boundary */
224 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
226 /* to align the pointer to the (prev) page boundary */
227 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
229 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
230 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
232 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
233 static inline struct folio *lru_to_folio(struct list_head *head)
235 return list_entry((head)->prev, struct folio, lru);
238 void setup_initial_init_mm(void *start_code, void *end_code,
239 void *end_data, void *brk);
242 * Linux kernel virtual memory manager primitives.
243 * The idea being to have a "virtual" mm in the same way
244 * we have a virtual fs - giving a cleaner interface to the
245 * mm details, and allowing different kinds of memory mappings
246 * (from shared memory to executable loading to arbitrary
250 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
251 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
252 void vm_area_free(struct vm_area_struct *);
255 extern struct rb_root nommu_region_tree;
256 extern struct rw_semaphore nommu_region_sem;
258 extern unsigned int kobjsize(const void *objp);
262 * vm_flags in vm_area_struct, see mm_types.h.
263 * When changing, update also include/trace/events/mmflags.h
265 #define VM_NONE 0x00000000
267 #define VM_READ 0x00000001 /* currently active flags */
268 #define VM_WRITE 0x00000002
269 #define VM_EXEC 0x00000004
270 #define VM_SHARED 0x00000008
272 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
273 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
274 #define VM_MAYWRITE 0x00000020
275 #define VM_MAYEXEC 0x00000040
276 #define VM_MAYSHARE 0x00000080
278 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
280 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
281 #else /* CONFIG_MMU */
282 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
283 #define VM_UFFD_MISSING 0
284 #endif /* CONFIG_MMU */
285 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
286 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
288 #define VM_LOCKED 0x00002000
289 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
291 /* Used by sys_madvise() */
292 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
293 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
295 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
296 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
297 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
298 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
299 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
300 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
301 #define VM_SYNC 0x00800000 /* Synchronous page faults */
302 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
303 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
304 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
306 #ifdef CONFIG_MEM_SOFT_DIRTY
307 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
309 # define VM_SOFTDIRTY 0
312 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
313 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
314 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
315 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
317 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
318 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
323 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
324 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
325 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
326 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
327 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
328 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
330 #ifdef CONFIG_ARCH_HAS_PKEYS
331 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
332 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
333 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
334 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
335 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
337 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
339 # define VM_PKEY_BIT4 0
341 #endif /* CONFIG_ARCH_HAS_PKEYS */
343 #if defined(CONFIG_X86)
344 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
345 #elif defined(CONFIG_PPC)
346 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
347 #elif defined(CONFIG_PARISC)
348 # define VM_GROWSUP VM_ARCH_1
349 #elif defined(CONFIG_IA64)
350 # define VM_GROWSUP VM_ARCH_1
351 #elif defined(CONFIG_SPARC64)
352 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
353 # define VM_ARCH_CLEAR VM_SPARC_ADI
354 #elif defined(CONFIG_ARM64)
355 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
356 # define VM_ARCH_CLEAR VM_ARM64_BTI
357 #elif !defined(CONFIG_MMU)
358 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
361 #if defined(CONFIG_ARM64_MTE)
362 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
363 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
365 # define VM_MTE VM_NONE
366 # define VM_MTE_ALLOWED VM_NONE
370 # define VM_GROWSUP VM_NONE
373 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
374 # define VM_UFFD_MINOR_BIT 37
375 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
376 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
377 # define VM_UFFD_MINOR VM_NONE
378 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
380 /* Bits set in the VMA until the stack is in its final location */
381 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
383 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
385 /* Common data flag combinations */
386 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
387 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
388 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
389 VM_MAYWRITE | VM_MAYEXEC)
390 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
391 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
393 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
394 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
397 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
398 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
401 #ifdef CONFIG_STACK_GROWSUP
402 #define VM_STACK VM_GROWSUP
404 #define VM_STACK VM_GROWSDOWN
407 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
409 /* VMA basic access permission flags */
410 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
414 * Special vmas that are non-mergable, non-mlock()able.
416 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
418 /* This mask prevents VMA from being scanned with khugepaged */
419 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
421 /* This mask defines which mm->def_flags a process can inherit its parent */
422 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
424 /* This mask is used to clear all the VMA flags used by mlock */
425 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
427 /* Arch-specific flags to clear when updating VM flags on protection change */
428 #ifndef VM_ARCH_CLEAR
429 # define VM_ARCH_CLEAR VM_NONE
431 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
434 * mapping from the currently active vm_flags protection bits (the
435 * low four bits) to a page protection mask..
439 * The default fault flags that should be used by most of the
440 * arch-specific page fault handlers.
442 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
443 FAULT_FLAG_KILLABLE | \
444 FAULT_FLAG_INTERRUPTIBLE)
447 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
448 * @flags: Fault flags.
450 * This is mostly used for places where we want to try to avoid taking
451 * the mmap_lock for too long a time when waiting for another condition
452 * to change, in which case we can try to be polite to release the
453 * mmap_lock in the first round to avoid potential starvation of other
454 * processes that would also want the mmap_lock.
456 * Return: true if the page fault allows retry and this is the first
457 * attempt of the fault handling; false otherwise.
459 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
461 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
462 (!(flags & FAULT_FLAG_TRIED));
465 #define FAULT_FLAG_TRACE \
466 { FAULT_FLAG_WRITE, "WRITE" }, \
467 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
468 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
469 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
470 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
471 { FAULT_FLAG_TRIED, "TRIED" }, \
472 { FAULT_FLAG_USER, "USER" }, \
473 { FAULT_FLAG_REMOTE, "REMOTE" }, \
474 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
475 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
478 * vm_fault is filled by the pagefault handler and passed to the vma's
479 * ->fault function. The vma's ->fault is responsible for returning a bitmask
480 * of VM_FAULT_xxx flags that give details about how the fault was handled.
482 * MM layer fills up gfp_mask for page allocations but fault handler might
483 * alter it if its implementation requires a different allocation context.
485 * pgoff should be used in favour of virtual_address, if possible.
489 struct vm_area_struct *vma; /* Target VMA */
490 gfp_t gfp_mask; /* gfp mask to be used for allocations */
491 pgoff_t pgoff; /* Logical page offset based on vma */
492 unsigned long address; /* Faulting virtual address - masked */
493 unsigned long real_address; /* Faulting virtual address - unmasked */
495 enum fault_flag flags; /* FAULT_FLAG_xxx flags
496 * XXX: should really be 'const' */
497 pmd_t *pmd; /* Pointer to pmd entry matching
499 pud_t *pud; /* Pointer to pud entry matching
503 pte_t orig_pte; /* Value of PTE at the time of fault */
504 pmd_t orig_pmd; /* Value of PMD at the time of fault,
505 * used by PMD fault only.
509 struct page *cow_page; /* Page handler may use for COW fault */
510 struct page *page; /* ->fault handlers should return a
511 * page here, unless VM_FAULT_NOPAGE
512 * is set (which is also implied by
515 /* These three entries are valid only while holding ptl lock */
516 pte_t *pte; /* Pointer to pte entry matching
517 * the 'address'. NULL if the page
518 * table hasn't been allocated.
520 spinlock_t *ptl; /* Page table lock.
521 * Protects pte page table if 'pte'
522 * is not NULL, otherwise pmd.
524 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
525 * vm_ops->map_pages() sets up a page
526 * table from atomic context.
527 * do_fault_around() pre-allocates
528 * page table to avoid allocation from
533 /* page entry size for vm->huge_fault() */
534 enum page_entry_size {
541 * These are the virtual MM functions - opening of an area, closing and
542 * unmapping it (needed to keep files on disk up-to-date etc), pointer
543 * to the functions called when a no-page or a wp-page exception occurs.
545 struct vm_operations_struct {
546 void (*open)(struct vm_area_struct * area);
548 * @close: Called when the VMA is being removed from the MM.
549 * Context: User context. May sleep. Caller holds mmap_lock.
551 void (*close)(struct vm_area_struct * area);
552 /* Called any time before splitting to check if it's allowed */
553 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
554 int (*mremap)(struct vm_area_struct *area);
556 * Called by mprotect() to make driver-specific permission
557 * checks before mprotect() is finalised. The VMA must not
558 * be modified. Returns 0 if mprotect() can proceed.
560 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
561 unsigned long end, unsigned long newflags);
562 vm_fault_t (*fault)(struct vm_fault *vmf);
563 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
564 enum page_entry_size pe_size);
565 vm_fault_t (*map_pages)(struct vm_fault *vmf,
566 pgoff_t start_pgoff, pgoff_t end_pgoff);
567 unsigned long (*pagesize)(struct vm_area_struct * area);
569 /* notification that a previously read-only page is about to become
570 * writable, if an error is returned it will cause a SIGBUS */
571 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
573 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
574 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
576 /* called by access_process_vm when get_user_pages() fails, typically
577 * for use by special VMAs. See also generic_access_phys() for a generic
578 * implementation useful for any iomem mapping.
580 int (*access)(struct vm_area_struct *vma, unsigned long addr,
581 void *buf, int len, int write);
583 /* Called by the /proc/PID/maps code to ask the vma whether it
584 * has a special name. Returning non-NULL will also cause this
585 * vma to be dumped unconditionally. */
586 const char *(*name)(struct vm_area_struct *vma);
590 * set_policy() op must add a reference to any non-NULL @new mempolicy
591 * to hold the policy upon return. Caller should pass NULL @new to
592 * remove a policy and fall back to surrounding context--i.e. do not
593 * install a MPOL_DEFAULT policy, nor the task or system default
596 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
599 * get_policy() op must add reference [mpol_get()] to any policy at
600 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
601 * in mm/mempolicy.c will do this automatically.
602 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
603 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
604 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
605 * must return NULL--i.e., do not "fallback" to task or system default
608 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
612 * Called by vm_normal_page() for special PTEs to find the
613 * page for @addr. This is useful if the default behavior
614 * (using pte_page()) would not find the correct page.
616 struct page *(*find_special_page)(struct vm_area_struct *vma,
620 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
622 static const struct vm_operations_struct dummy_vm_ops = {};
624 memset(vma, 0, sizeof(*vma));
626 vma->vm_ops = &dummy_vm_ops;
627 INIT_LIST_HEAD(&vma->anon_vma_chain);
630 static inline void vma_set_anonymous(struct vm_area_struct *vma)
635 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
640 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
642 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
647 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
648 VM_STACK_INCOMPLETE_SETUP)
654 static inline bool vma_is_foreign(struct vm_area_struct *vma)
659 if (current->mm != vma->vm_mm)
665 static inline bool vma_is_accessible(struct vm_area_struct *vma)
667 return vma->vm_flags & VM_ACCESS_FLAGS;
671 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
673 return mas_find(&vmi->mas, max - 1);
676 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
679 * Uses mas_find() to get the first VMA when the iterator starts.
680 * Calling mas_next() could skip the first entry.
682 return mas_find(&vmi->mas, ULONG_MAX);
685 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
687 return mas_prev(&vmi->mas, 0);
690 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
692 return vmi->mas.index;
695 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
697 return vmi->mas.last + 1;
699 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
702 return mas_expected_entries(&vmi->mas, count);
705 /* Free any unused preallocations */
706 static inline void vma_iter_free(struct vma_iterator *vmi)
708 mas_destroy(&vmi->mas);
711 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
712 struct vm_area_struct *vma)
714 vmi->mas.index = vma->vm_start;
715 vmi->mas.last = vma->vm_end - 1;
716 mas_store(&vmi->mas, vma);
717 if (unlikely(mas_is_err(&vmi->mas)))
723 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
725 mas_pause(&vmi->mas);
728 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
730 mas_set(&vmi->mas, addr);
733 #define for_each_vma(__vmi, __vma) \
734 while (((__vma) = vma_next(&(__vmi))) != NULL)
736 /* The MM code likes to work with exclusive end addresses */
737 #define for_each_vma_range(__vmi, __vma, __end) \
738 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
742 * The vma_is_shmem is not inline because it is used only by slow
743 * paths in userfault.
745 bool vma_is_shmem(struct vm_area_struct *vma);
746 bool vma_is_anon_shmem(struct vm_area_struct *vma);
748 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
749 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
752 int vma_is_stack_for_current(struct vm_area_struct *vma);
754 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
755 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
761 * compound_order() can be called without holding a reference, which means
762 * that niceties like page_folio() don't work. These callers should be
763 * prepared to handle wild return values. For example, PG_head may be
764 * set before _folio_order is initialised, or this may be a tail page.
765 * See compaction.c for some good examples.
767 static inline unsigned int compound_order(struct page *page)
769 struct folio *folio = (struct folio *)page;
771 if (!test_bit(PG_head, &folio->flags))
773 return folio->_folio_order;
777 * folio_order - The allocation order of a folio.
780 * A folio is composed of 2^order pages. See get_order() for the definition
783 * Return: The order of the folio.
785 static inline unsigned int folio_order(struct folio *folio)
787 if (!folio_test_large(folio))
789 return folio->_folio_order;
792 #include <linux/huge_mm.h>
795 * Methods to modify the page usage count.
797 * What counts for a page usage:
798 * - cache mapping (page->mapping)
799 * - private data (page->private)
800 * - page mapped in a task's page tables, each mapping
801 * is counted separately
803 * Also, many kernel routines increase the page count before a critical
804 * routine so they can be sure the page doesn't go away from under them.
808 * Drop a ref, return true if the refcount fell to zero (the page has no users)
810 static inline int put_page_testzero(struct page *page)
812 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
813 return page_ref_dec_and_test(page);
816 static inline int folio_put_testzero(struct folio *folio)
818 return put_page_testzero(&folio->page);
822 * Try to grab a ref unless the page has a refcount of zero, return false if
824 * This can be called when MMU is off so it must not access
825 * any of the virtual mappings.
827 static inline bool get_page_unless_zero(struct page *page)
829 return page_ref_add_unless(page, 1, 0);
832 extern int page_is_ram(unsigned long pfn);
840 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
843 /* Support for virtually mapped pages */
844 struct page *vmalloc_to_page(const void *addr);
845 unsigned long vmalloc_to_pfn(const void *addr);
848 * Determine if an address is within the vmalloc range
850 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
851 * is no special casing required.
854 #ifndef is_ioremap_addr
855 #define is_ioremap_addr(x) is_vmalloc_addr(x)
859 extern bool is_vmalloc_addr(const void *x);
860 extern int is_vmalloc_or_module_addr(const void *x);
862 static inline bool is_vmalloc_addr(const void *x)
866 static inline int is_vmalloc_or_module_addr(const void *x)
873 * How many times the entire folio is mapped as a single unit (eg by a
874 * PMD or PUD entry). This is probably not what you want, except for
875 * debugging purposes - it does not include PTE-mapped sub-pages; look
876 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
878 static inline int folio_entire_mapcount(struct folio *folio)
880 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
881 return atomic_read(&folio->_entire_mapcount) + 1;
885 * The atomic page->_mapcount, starts from -1: so that transitions
886 * both from it and to it can be tracked, using atomic_inc_and_test
887 * and atomic_add_negative(-1).
889 static inline void page_mapcount_reset(struct page *page)
891 atomic_set(&(page)->_mapcount, -1);
895 * page_mapcount() - Number of times this precise page is mapped.
898 * The number of times this page is mapped. If this page is part of
899 * a large folio, it includes the number of times this page is mapped
900 * as part of that folio.
902 * The result is undefined for pages which cannot be mapped into userspace.
903 * For example SLAB or special types of pages. See function page_has_type().
904 * They use this field in struct page differently.
906 static inline int page_mapcount(struct page *page)
908 int mapcount = atomic_read(&page->_mapcount) + 1;
910 if (unlikely(PageCompound(page)))
911 mapcount += folio_entire_mapcount(page_folio(page));
916 int folio_total_mapcount(struct folio *folio);
919 * folio_mapcount() - Calculate the number of mappings of this folio.
922 * A large folio tracks both how many times the entire folio is mapped,
923 * and how many times each individual page in the folio is mapped.
924 * This function calculates the total number of times the folio is
927 * Return: The number of times this folio is mapped.
929 static inline int folio_mapcount(struct folio *folio)
931 if (likely(!folio_test_large(folio)))
932 return atomic_read(&folio->_mapcount) + 1;
933 return folio_total_mapcount(folio);
936 static inline int total_mapcount(struct page *page)
938 if (likely(!PageCompound(page)))
939 return atomic_read(&page->_mapcount) + 1;
940 return folio_total_mapcount(page_folio(page));
943 static inline bool folio_large_is_mapped(struct folio *folio)
946 * Reading _entire_mapcount below could be omitted if hugetlb
947 * participated in incrementing nr_pages_mapped when compound mapped.
949 return atomic_read(&folio->_nr_pages_mapped) > 0 ||
950 atomic_read(&folio->_entire_mapcount) >= 0;
954 * folio_mapped - Is this folio mapped into userspace?
957 * Return: True if any page in this folio is referenced by user page tables.
959 static inline bool folio_mapped(struct folio *folio)
961 if (likely(!folio_test_large(folio)))
962 return atomic_read(&folio->_mapcount) >= 0;
963 return folio_large_is_mapped(folio);
967 * Return true if this page is mapped into pagetables.
968 * For compound page it returns true if any sub-page of compound page is mapped,
969 * even if this particular sub-page is not itself mapped by any PTE or PMD.
971 static inline bool page_mapped(struct page *page)
973 if (likely(!PageCompound(page)))
974 return atomic_read(&page->_mapcount) >= 0;
975 return folio_large_is_mapped(page_folio(page));
978 static inline struct page *virt_to_head_page(const void *x)
980 struct page *page = virt_to_page(x);
982 return compound_head(page);
985 static inline struct folio *virt_to_folio(const void *x)
987 struct page *page = virt_to_page(x);
989 return page_folio(page);
992 void __folio_put(struct folio *folio);
994 void put_pages_list(struct list_head *pages);
996 void split_page(struct page *page, unsigned int order);
997 void folio_copy(struct folio *dst, struct folio *src);
999 unsigned long nr_free_buffer_pages(void);
1002 * Compound pages have a destructor function. Provide a
1003 * prototype for that function and accessor functions.
1004 * These are _only_ valid on the head of a compound page.
1006 typedef void compound_page_dtor(struct page *);
1008 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1009 enum compound_dtor_id {
1012 #ifdef CONFIG_HUGETLB_PAGE
1015 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1016 TRANSHUGE_PAGE_DTOR,
1020 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
1022 static inline void set_compound_page_dtor(struct page *page,
1023 enum compound_dtor_id compound_dtor)
1025 struct folio *folio = (struct folio *)page;
1027 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
1028 VM_BUG_ON_PAGE(!PageHead(page), page);
1029 folio->_folio_dtor = compound_dtor;
1032 static inline void folio_set_compound_dtor(struct folio *folio,
1033 enum compound_dtor_id compound_dtor)
1035 VM_BUG_ON_FOLIO(compound_dtor >= NR_COMPOUND_DTORS, folio);
1036 folio->_folio_dtor = compound_dtor;
1039 void destroy_large_folio(struct folio *folio);
1041 static inline void set_compound_order(struct page *page, unsigned int order)
1043 struct folio *folio = (struct folio *)page;
1045 folio->_folio_order = order;
1047 folio->_folio_nr_pages = 1U << order;
1051 /* Returns the number of bytes in this potentially compound page. */
1052 static inline unsigned long page_size(struct page *page)
1054 return PAGE_SIZE << compound_order(page);
1057 /* Returns the number of bits needed for the number of bytes in a page */
1058 static inline unsigned int page_shift(struct page *page)
1060 return PAGE_SHIFT + compound_order(page);
1064 * thp_order - Order of a transparent huge page.
1065 * @page: Head page of a transparent huge page.
1067 static inline unsigned int thp_order(struct page *page)
1069 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1070 return compound_order(page);
1074 * thp_size - Size of a transparent huge page.
1075 * @page: Head page of a transparent huge page.
1077 * Return: Number of bytes in this page.
1079 static inline unsigned long thp_size(struct page *page)
1081 return PAGE_SIZE << thp_order(page);
1084 void free_compound_page(struct page *page);
1088 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1089 * servicing faults for write access. In the normal case, do always want
1090 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1091 * that do not have writing enabled, when used by access_process_vm.
1093 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1095 if (likely(vma->vm_flags & VM_WRITE))
1096 pte = pte_mkwrite(pte);
1100 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1101 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1103 vm_fault_t finish_fault(struct vm_fault *vmf);
1104 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1108 * Multiple processes may "see" the same page. E.g. for untouched
1109 * mappings of /dev/null, all processes see the same page full of
1110 * zeroes, and text pages of executables and shared libraries have
1111 * only one copy in memory, at most, normally.
1113 * For the non-reserved pages, page_count(page) denotes a reference count.
1114 * page_count() == 0 means the page is free. page->lru is then used for
1115 * freelist management in the buddy allocator.
1116 * page_count() > 0 means the page has been allocated.
1118 * Pages are allocated by the slab allocator in order to provide memory
1119 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1120 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1121 * unless a particular usage is carefully commented. (the responsibility of
1122 * freeing the kmalloc memory is the caller's, of course).
1124 * A page may be used by anyone else who does a __get_free_page().
1125 * In this case, page_count still tracks the references, and should only
1126 * be used through the normal accessor functions. The top bits of page->flags
1127 * and page->virtual store page management information, but all other fields
1128 * are unused and could be used privately, carefully. The management of this
1129 * page is the responsibility of the one who allocated it, and those who have
1130 * subsequently been given references to it.
1132 * The other pages (we may call them "pagecache pages") are completely
1133 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1134 * The following discussion applies only to them.
1136 * A pagecache page contains an opaque `private' member, which belongs to the
1137 * page's address_space. Usually, this is the address of a circular list of
1138 * the page's disk buffers. PG_private must be set to tell the VM to call
1139 * into the filesystem to release these pages.
1141 * A page may belong to an inode's memory mapping. In this case, page->mapping
1142 * is the pointer to the inode, and page->index is the file offset of the page,
1143 * in units of PAGE_SIZE.
1145 * If pagecache pages are not associated with an inode, they are said to be
1146 * anonymous pages. These may become associated with the swapcache, and in that
1147 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1149 * In either case (swapcache or inode backed), the pagecache itself holds one
1150 * reference to the page. Setting PG_private should also increment the
1151 * refcount. The each user mapping also has a reference to the page.
1153 * The pagecache pages are stored in a per-mapping radix tree, which is
1154 * rooted at mapping->i_pages, and indexed by offset.
1155 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1156 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1158 * All pagecache pages may be subject to I/O:
1159 * - inode pages may need to be read from disk,
1160 * - inode pages which have been modified and are MAP_SHARED may need
1161 * to be written back to the inode on disk,
1162 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1163 * modified may need to be swapped out to swap space and (later) to be read
1167 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1168 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1170 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1171 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1173 if (!static_branch_unlikely(&devmap_managed_key))
1175 if (!is_zone_device_page(page))
1177 return __put_devmap_managed_page_refs(page, refs);
1179 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1180 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1184 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1186 static inline bool put_devmap_managed_page(struct page *page)
1188 return put_devmap_managed_page_refs(page, 1);
1191 /* 127: arbitrary random number, small enough to assemble well */
1192 #define folio_ref_zero_or_close_to_overflow(folio) \
1193 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1196 * folio_get - Increment the reference count on a folio.
1197 * @folio: The folio.
1199 * Context: May be called in any context, as long as you know that
1200 * you have a refcount on the folio. If you do not already have one,
1201 * folio_try_get() may be the right interface for you to use.
1203 static inline void folio_get(struct folio *folio)
1205 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1206 folio_ref_inc(folio);
1209 static inline void get_page(struct page *page)
1211 folio_get(page_folio(page));
1214 int __must_check try_grab_page(struct page *page, unsigned int flags);
1216 static inline __must_check bool try_get_page(struct page *page)
1218 page = compound_head(page);
1219 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1226 * folio_put - Decrement the reference count on a folio.
1227 * @folio: The folio.
1229 * If the folio's reference count reaches zero, the memory will be
1230 * released back to the page allocator and may be used by another
1231 * allocation immediately. Do not access the memory or the struct folio
1232 * after calling folio_put() unless you can be sure that it wasn't the
1235 * Context: May be called in process or interrupt context, but not in NMI
1236 * context. May be called while holding a spinlock.
1238 static inline void folio_put(struct folio *folio)
1240 if (folio_put_testzero(folio))
1245 * folio_put_refs - Reduce the reference count on a folio.
1246 * @folio: The folio.
1247 * @refs: The amount to subtract from the folio's reference count.
1249 * If the folio's reference count reaches zero, the memory will be
1250 * released back to the page allocator and may be used by another
1251 * allocation immediately. Do not access the memory or the struct folio
1252 * after calling folio_put_refs() unless you can be sure that these weren't
1253 * the last references.
1255 * Context: May be called in process or interrupt context, but not in NMI
1256 * context. May be called while holding a spinlock.
1258 static inline void folio_put_refs(struct folio *folio, int refs)
1260 if (folio_ref_sub_and_test(folio, refs))
1265 * union release_pages_arg - an array of pages or folios
1267 * release_pages() releases a simple array of multiple pages, and
1268 * accepts various different forms of said page array: either
1269 * a regular old boring array of pages, an array of folios, or
1270 * an array of encoded page pointers.
1272 * The transparent union syntax for this kind of "any of these
1273 * argument types" is all kinds of ugly, so look away.
1276 struct page **pages;
1277 struct folio **folios;
1278 struct encoded_page **encoded_pages;
1279 } release_pages_arg __attribute__ ((__transparent_union__));
1281 void release_pages(release_pages_arg, int nr);
1284 * folios_put - Decrement the reference count on an array of folios.
1285 * @folios: The folios.
1286 * @nr: How many folios there are.
1288 * Like folio_put(), but for an array of folios. This is more efficient
1289 * than writing the loop yourself as it will optimise the locks which
1290 * need to be taken if the folios are freed.
1292 * Context: May be called in process or interrupt context, but not in NMI
1293 * context. May be called while holding a spinlock.
1295 static inline void folios_put(struct folio **folios, unsigned int nr)
1297 release_pages(folios, nr);
1300 static inline void put_page(struct page *page)
1302 struct folio *folio = page_folio(page);
1305 * For some devmap managed pages we need to catch refcount transition
1308 if (put_devmap_managed_page(&folio->page))
1314 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1315 * the page's refcount so that two separate items are tracked: the original page
1316 * reference count, and also a new count of how many pin_user_pages() calls were
1317 * made against the page. ("gup-pinned" is another term for the latter).
1319 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1320 * distinct from normal pages. As such, the unpin_user_page() call (and its
1321 * variants) must be used in order to release gup-pinned pages.
1325 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1326 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1327 * simpler, due to the fact that adding an even power of two to the page
1328 * refcount has the effect of using only the upper N bits, for the code that
1329 * counts up using the bias value. This means that the lower bits are left for
1330 * the exclusive use of the original code that increments and decrements by one
1331 * (or at least, by much smaller values than the bias value).
1333 * Of course, once the lower bits overflow into the upper bits (and this is
1334 * OK, because subtraction recovers the original values), then visual inspection
1335 * no longer suffices to directly view the separate counts. However, for normal
1336 * applications that don't have huge page reference counts, this won't be an
1339 * Locking: the lockless algorithm described in folio_try_get_rcu()
1340 * provides safe operation for get_user_pages(), page_mkclean() and
1341 * other calls that race to set up page table entries.
1343 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1345 void unpin_user_page(struct page *page);
1346 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1348 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1350 void unpin_user_pages(struct page **pages, unsigned long npages);
1352 static inline bool is_cow_mapping(vm_flags_t flags)
1354 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1358 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1361 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1362 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1363 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1364 * underlying memory if ptrace is active, so this is only possible if
1365 * ptrace does not apply. Note that there is no mprotect() to upgrade
1366 * write permissions later.
1368 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1372 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1373 #define SECTION_IN_PAGE_FLAGS
1377 * The identification function is mainly used by the buddy allocator for
1378 * determining if two pages could be buddies. We are not really identifying
1379 * the zone since we could be using the section number id if we do not have
1380 * node id available in page flags.
1381 * We only guarantee that it will return the same value for two combinable
1384 static inline int page_zone_id(struct page *page)
1386 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1389 #ifdef NODE_NOT_IN_PAGE_FLAGS
1390 extern int page_to_nid(const struct page *page);
1392 static inline int page_to_nid(const struct page *page)
1394 struct page *p = (struct page *)page;
1396 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1400 static inline int folio_nid(const struct folio *folio)
1402 return page_to_nid(&folio->page);
1405 #ifdef CONFIG_NUMA_BALANCING
1406 /* page access time bits needs to hold at least 4 seconds */
1407 #define PAGE_ACCESS_TIME_MIN_BITS 12
1408 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1409 #define PAGE_ACCESS_TIME_BUCKETS \
1410 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1412 #define PAGE_ACCESS_TIME_BUCKETS 0
1415 #define PAGE_ACCESS_TIME_MASK \
1416 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1418 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1420 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1423 static inline int cpupid_to_pid(int cpupid)
1425 return cpupid & LAST__PID_MASK;
1428 static inline int cpupid_to_cpu(int cpupid)
1430 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1433 static inline int cpupid_to_nid(int cpupid)
1435 return cpu_to_node(cpupid_to_cpu(cpupid));
1438 static inline bool cpupid_pid_unset(int cpupid)
1440 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1443 static inline bool cpupid_cpu_unset(int cpupid)
1445 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1448 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1450 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1453 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1454 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1455 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1457 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1460 static inline int page_cpupid_last(struct page *page)
1462 return page->_last_cpupid;
1464 static inline void page_cpupid_reset_last(struct page *page)
1466 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1469 static inline int page_cpupid_last(struct page *page)
1471 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1474 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1476 static inline void page_cpupid_reset_last(struct page *page)
1478 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1480 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1482 static inline int xchg_page_access_time(struct page *page, int time)
1486 last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1487 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1489 #else /* !CONFIG_NUMA_BALANCING */
1490 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1492 return page_to_nid(page); /* XXX */
1495 static inline int xchg_page_access_time(struct page *page, int time)
1500 static inline int page_cpupid_last(struct page *page)
1502 return page_to_nid(page); /* XXX */
1505 static inline int cpupid_to_nid(int cpupid)
1510 static inline int cpupid_to_pid(int cpupid)
1515 static inline int cpupid_to_cpu(int cpupid)
1520 static inline int cpu_pid_to_cpupid(int nid, int pid)
1525 static inline bool cpupid_pid_unset(int cpupid)
1530 static inline void page_cpupid_reset_last(struct page *page)
1534 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1538 #endif /* CONFIG_NUMA_BALANCING */
1540 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1543 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1544 * setting tags for all pages to native kernel tag value 0xff, as the default
1545 * value 0x00 maps to 0xff.
1548 static inline u8 page_kasan_tag(const struct page *page)
1552 if (kasan_enabled()) {
1553 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1560 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1562 unsigned long old_flags, flags;
1564 if (!kasan_enabled())
1568 old_flags = READ_ONCE(page->flags);
1571 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1572 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1573 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1576 static inline void page_kasan_tag_reset(struct page *page)
1578 if (kasan_enabled())
1579 page_kasan_tag_set(page, 0xff);
1582 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1584 static inline u8 page_kasan_tag(const struct page *page)
1589 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1590 static inline void page_kasan_tag_reset(struct page *page) { }
1592 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1594 static inline struct zone *page_zone(const struct page *page)
1596 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1599 static inline pg_data_t *page_pgdat(const struct page *page)
1601 return NODE_DATA(page_to_nid(page));
1604 static inline struct zone *folio_zone(const struct folio *folio)
1606 return page_zone(&folio->page);
1609 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1611 return page_pgdat(&folio->page);
1614 #ifdef SECTION_IN_PAGE_FLAGS
1615 static inline void set_page_section(struct page *page, unsigned long section)
1617 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1618 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1621 static inline unsigned long page_to_section(const struct page *page)
1623 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1628 * folio_pfn - Return the Page Frame Number of a folio.
1629 * @folio: The folio.
1631 * A folio may contain multiple pages. The pages have consecutive
1632 * Page Frame Numbers.
1634 * Return: The Page Frame Number of the first page in the folio.
1636 static inline unsigned long folio_pfn(struct folio *folio)
1638 return page_to_pfn(&folio->page);
1641 static inline struct folio *pfn_folio(unsigned long pfn)
1643 return page_folio(pfn_to_page(pfn));
1647 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1648 * @folio: The folio.
1650 * This function checks if a folio has been pinned via a call to
1651 * a function in the pin_user_pages() family.
1653 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1654 * because it means "definitely not pinned for DMA", but true means "probably
1655 * pinned for DMA, but possibly a false positive due to having at least
1656 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1658 * False positives are OK, because: a) it's unlikely for a folio to
1659 * get that many refcounts, and b) all the callers of this routine are
1660 * expected to be able to deal gracefully with a false positive.
1662 * For large folios, the result will be exactly correct. That's because
1663 * we have more tracking data available: the _pincount field is used
1664 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1666 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1668 * Return: True, if it is likely that the page has been "dma-pinned".
1669 * False, if the page is definitely not dma-pinned.
1671 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1673 if (folio_test_large(folio))
1674 return atomic_read(&folio->_pincount) > 0;
1677 * folio_ref_count() is signed. If that refcount overflows, then
1678 * folio_ref_count() returns a negative value, and callers will avoid
1679 * further incrementing the refcount.
1681 * Here, for that overflow case, use the sign bit to count a little
1682 * bit higher via unsigned math, and thus still get an accurate result.
1684 return ((unsigned int)folio_ref_count(folio)) >=
1685 GUP_PIN_COUNTING_BIAS;
1688 static inline bool page_maybe_dma_pinned(struct page *page)
1690 return folio_maybe_dma_pinned(page_folio(page));
1694 * This should most likely only be called during fork() to see whether we
1695 * should break the cow immediately for an anon page on the src mm.
1697 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1699 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1702 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1704 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1707 return page_maybe_dma_pinned(page);
1710 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1711 #ifdef CONFIG_MIGRATION
1712 static inline bool is_longterm_pinnable_page(struct page *page)
1715 int mt = get_pageblock_migratetype(page);
1717 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1720 /* The zero page may always be pinned */
1721 if (is_zero_pfn(page_to_pfn(page)))
1724 /* Coherent device memory must always allow eviction. */
1725 if (is_device_coherent_page(page))
1728 /* Otherwise, non-movable zone pages can be pinned. */
1729 return !is_zone_movable_page(page);
1732 static inline bool is_longterm_pinnable_page(struct page *page)
1738 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1740 return is_longterm_pinnable_page(&folio->page);
1743 static inline void set_page_zone(struct page *page, enum zone_type zone)
1745 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1746 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1749 static inline void set_page_node(struct page *page, unsigned long node)
1751 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1752 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1755 static inline void set_page_links(struct page *page, enum zone_type zone,
1756 unsigned long node, unsigned long pfn)
1758 set_page_zone(page, zone);
1759 set_page_node(page, node);
1760 #ifdef SECTION_IN_PAGE_FLAGS
1761 set_page_section(page, pfn_to_section_nr(pfn));
1766 * folio_nr_pages - The number of pages in the folio.
1767 * @folio: The folio.
1769 * Return: A positive power of two.
1771 static inline long folio_nr_pages(struct folio *folio)
1773 if (!folio_test_large(folio))
1776 return folio->_folio_nr_pages;
1778 return 1L << folio->_folio_order;
1783 * compound_nr() returns the number of pages in this potentially compound
1784 * page. compound_nr() can be called on a tail page, and is defined to
1785 * return 1 in that case.
1787 static inline unsigned long compound_nr(struct page *page)
1789 struct folio *folio = (struct folio *)page;
1791 if (!test_bit(PG_head, &folio->flags))
1794 return folio->_folio_nr_pages;
1796 return 1L << folio->_folio_order;
1801 * thp_nr_pages - The number of regular pages in this huge page.
1802 * @page: The head page of a huge page.
1804 static inline int thp_nr_pages(struct page *page)
1806 return folio_nr_pages((struct folio *)page);
1810 * folio_next - Move to the next physical folio.
1811 * @folio: The folio we're currently operating on.
1813 * If you have physically contiguous memory which may span more than
1814 * one folio (eg a &struct bio_vec), use this function to move from one
1815 * folio to the next. Do not use it if the memory is only virtually
1816 * contiguous as the folios are almost certainly not adjacent to each
1817 * other. This is the folio equivalent to writing ``page++``.
1819 * Context: We assume that the folios are refcounted and/or locked at a
1820 * higher level and do not adjust the reference counts.
1821 * Return: The next struct folio.
1823 static inline struct folio *folio_next(struct folio *folio)
1825 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1829 * folio_shift - The size of the memory described by this folio.
1830 * @folio: The folio.
1832 * A folio represents a number of bytes which is a power-of-two in size.
1833 * This function tells you which power-of-two the folio is. See also
1834 * folio_size() and folio_order().
1836 * Context: The caller should have a reference on the folio to prevent
1837 * it from being split. It is not necessary for the folio to be locked.
1838 * Return: The base-2 logarithm of the size of this folio.
1840 static inline unsigned int folio_shift(struct folio *folio)
1842 return PAGE_SHIFT + folio_order(folio);
1846 * folio_size - The number of bytes in a folio.
1847 * @folio: The folio.
1849 * Context: The caller should have a reference on the folio to prevent
1850 * it from being split. It is not necessary for the folio to be locked.
1851 * Return: The number of bytes in this folio.
1853 static inline size_t folio_size(struct folio *folio)
1855 return PAGE_SIZE << folio_order(folio);
1858 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1859 static inline int arch_make_page_accessible(struct page *page)
1865 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1866 static inline int arch_make_folio_accessible(struct folio *folio)
1869 long i, nr = folio_nr_pages(folio);
1871 for (i = 0; i < nr; i++) {
1872 ret = arch_make_page_accessible(folio_page(folio, i));
1882 * Some inline functions in vmstat.h depend on page_zone()
1884 #include <linux/vmstat.h>
1886 static __always_inline void *lowmem_page_address(const struct page *page)
1888 return page_to_virt(page);
1891 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1892 #define HASHED_PAGE_VIRTUAL
1895 #if defined(WANT_PAGE_VIRTUAL)
1896 static inline void *page_address(const struct page *page)
1898 return page->virtual;
1900 static inline void set_page_address(struct page *page, void *address)
1902 page->virtual = address;
1904 #define page_address_init() do { } while(0)
1907 #if defined(HASHED_PAGE_VIRTUAL)
1908 void *page_address(const struct page *page);
1909 void set_page_address(struct page *page, void *virtual);
1910 void page_address_init(void);
1913 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1914 #define page_address(page) lowmem_page_address(page)
1915 #define set_page_address(page, address) do { } while(0)
1916 #define page_address_init() do { } while(0)
1919 static inline void *folio_address(const struct folio *folio)
1921 return page_address(&folio->page);
1924 extern void *page_rmapping(struct page *page);
1925 extern pgoff_t __page_file_index(struct page *page);
1928 * Return the pagecache index of the passed page. Regular pagecache pages
1929 * use ->index whereas swapcache pages use swp_offset(->private)
1931 static inline pgoff_t page_index(struct page *page)
1933 if (unlikely(PageSwapCache(page)))
1934 return __page_file_index(page);
1939 * Return true only if the page has been allocated with
1940 * ALLOC_NO_WATERMARKS and the low watermark was not
1941 * met implying that the system is under some pressure.
1943 static inline bool page_is_pfmemalloc(const struct page *page)
1946 * lru.next has bit 1 set if the page is allocated from the
1947 * pfmemalloc reserves. Callers may simply overwrite it if
1948 * they do not need to preserve that information.
1950 return (uintptr_t)page->lru.next & BIT(1);
1954 * Return true only if the folio has been allocated with
1955 * ALLOC_NO_WATERMARKS and the low watermark was not
1956 * met implying that the system is under some pressure.
1958 static inline bool folio_is_pfmemalloc(const struct folio *folio)
1961 * lru.next has bit 1 set if the page is allocated from the
1962 * pfmemalloc reserves. Callers may simply overwrite it if
1963 * they do not need to preserve that information.
1965 return (uintptr_t)folio->lru.next & BIT(1);
1969 * Only to be called by the page allocator on a freshly allocated
1972 static inline void set_page_pfmemalloc(struct page *page)
1974 page->lru.next = (void *)BIT(1);
1977 static inline void clear_page_pfmemalloc(struct page *page)
1979 page->lru.next = NULL;
1983 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1985 extern void pagefault_out_of_memory(void);
1987 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1988 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1989 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1992 * Flags passed to show_mem() and show_free_areas() to suppress output in
1995 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1997 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
1998 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
2000 __show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
2004 * Parameter block passed down to zap_pte_range in exceptional cases.
2006 struct zap_details {
2007 struct folio *single_folio; /* Locked folio to be unmapped */
2008 bool even_cows; /* Zap COWed private pages too? */
2009 zap_flags_t zap_flags; /* Extra flags for zapping */
2013 * Whether to drop the pte markers, for example, the uffd-wp information for
2014 * file-backed memory. This should only be specified when we will completely
2015 * drop the page in the mm, either by truncation or unmapping of the vma. By
2016 * default, the flag is not set.
2018 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2019 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2020 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2023 extern bool can_do_mlock(void);
2025 static inline bool can_do_mlock(void) { return false; }
2027 extern int user_shm_lock(size_t, struct ucounts *);
2028 extern void user_shm_unlock(size_t, struct ucounts *);
2030 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2032 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2034 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2037 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2038 unsigned long size);
2039 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2040 unsigned long size, struct zap_details *details);
2041 static inline void zap_vma_pages(struct vm_area_struct *vma)
2043 zap_page_range_single(vma, vma->vm_start,
2044 vma->vm_end - vma->vm_start, NULL);
2046 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
2047 struct vm_area_struct *start_vma, unsigned long start,
2050 struct mmu_notifier_range;
2052 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2053 unsigned long end, unsigned long floor, unsigned long ceiling);
2055 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2056 int follow_pte(struct mm_struct *mm, unsigned long address,
2057 pte_t **ptepp, spinlock_t **ptlp);
2058 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2059 unsigned long *pfn);
2060 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2061 unsigned int flags, unsigned long *prot, resource_size_t *phys);
2062 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2063 void *buf, int len, int write);
2065 extern void truncate_pagecache(struct inode *inode, loff_t new);
2066 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2067 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2068 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2069 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2072 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2073 unsigned long address, unsigned int flags,
2074 struct pt_regs *regs);
2075 extern int fixup_user_fault(struct mm_struct *mm,
2076 unsigned long address, unsigned int fault_flags,
2078 void unmap_mapping_pages(struct address_space *mapping,
2079 pgoff_t start, pgoff_t nr, bool even_cows);
2080 void unmap_mapping_range(struct address_space *mapping,
2081 loff_t const holebegin, loff_t const holelen, int even_cows);
2083 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2084 unsigned long address, unsigned int flags,
2085 struct pt_regs *regs)
2087 /* should never happen if there's no MMU */
2089 return VM_FAULT_SIGBUS;
2091 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2092 unsigned int fault_flags, bool *unlocked)
2094 /* should never happen if there's no MMU */
2098 static inline void unmap_mapping_pages(struct address_space *mapping,
2099 pgoff_t start, pgoff_t nr, bool even_cows) { }
2100 static inline void unmap_mapping_range(struct address_space *mapping,
2101 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2104 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2105 loff_t const holebegin, loff_t const holelen)
2107 unmap_mapping_range(mapping, holebegin, holelen, 0);
2110 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2111 void *buf, int len, unsigned int gup_flags);
2112 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2113 void *buf, int len, unsigned int gup_flags);
2114 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2115 void *buf, int len, unsigned int gup_flags);
2117 long get_user_pages_remote(struct mm_struct *mm,
2118 unsigned long start, unsigned long nr_pages,
2119 unsigned int gup_flags, struct page **pages,
2120 struct vm_area_struct **vmas, int *locked);
2121 long pin_user_pages_remote(struct mm_struct *mm,
2122 unsigned long start, unsigned long nr_pages,
2123 unsigned int gup_flags, struct page **pages,
2124 struct vm_area_struct **vmas, int *locked);
2125 long get_user_pages(unsigned long start, unsigned long nr_pages,
2126 unsigned int gup_flags, struct page **pages,
2127 struct vm_area_struct **vmas);
2128 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2129 unsigned int gup_flags, struct page **pages,
2130 struct vm_area_struct **vmas);
2131 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2132 struct page **pages, unsigned int gup_flags);
2133 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2134 struct page **pages, unsigned int gup_flags);
2136 int get_user_pages_fast(unsigned long start, int nr_pages,
2137 unsigned int gup_flags, struct page **pages);
2138 int pin_user_pages_fast(unsigned long start, int nr_pages,
2139 unsigned int gup_flags, struct page **pages);
2141 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2142 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2143 struct task_struct *task, bool bypass_rlim);
2146 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
2147 struct page **pages);
2148 struct page *get_dump_page(unsigned long addr);
2150 bool folio_mark_dirty(struct folio *folio);
2151 bool set_page_dirty(struct page *page);
2152 int set_page_dirty_lock(struct page *page);
2154 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2156 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2157 unsigned long old_addr, struct vm_area_struct *new_vma,
2158 unsigned long new_addr, unsigned long len,
2159 bool need_rmap_locks);
2162 * Flags used by change_protection(). For now we make it a bitmap so
2163 * that we can pass in multiple flags just like parameters. However
2164 * for now all the callers are only use one of the flags at the same
2168 * Whether we should manually check if we can map individual PTEs writable,
2169 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2170 * PTEs automatically in a writable mapping.
2172 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2173 /* Whether this protection change is for NUMA hints */
2174 #define MM_CP_PROT_NUMA (1UL << 1)
2175 /* Whether this change is for write protecting */
2176 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2177 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2178 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2179 MM_CP_UFFD_WP_RESOLVE)
2181 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2182 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2185 * We want to check manually if we can change individual PTEs writable
2186 * if we can't do that automatically for all PTEs in a mapping. For
2187 * private mappings, that's always the case when we have write
2188 * permissions as we properly have to handle COW.
2190 if (vma->vm_flags & VM_SHARED)
2191 return vma_wants_writenotify(vma, vma->vm_page_prot);
2192 return !!(vma->vm_flags & VM_WRITE);
2195 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2197 extern long change_protection(struct mmu_gather *tlb,
2198 struct vm_area_struct *vma, unsigned long start,
2199 unsigned long end, unsigned long cp_flags);
2200 extern int mprotect_fixup(struct mmu_gather *tlb, struct vm_area_struct *vma,
2201 struct vm_area_struct **pprev, unsigned long start,
2202 unsigned long end, unsigned long newflags);
2205 * doesn't attempt to fault and will return short.
2207 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2208 unsigned int gup_flags, struct page **pages);
2209 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
2210 unsigned int gup_flags, struct page **pages);
2212 static inline bool get_user_page_fast_only(unsigned long addr,
2213 unsigned int gup_flags, struct page **pagep)
2215 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2218 * per-process(per-mm_struct) statistics.
2220 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2222 return percpu_counter_read_positive(&mm->rss_stat[member]);
2225 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2227 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2229 percpu_counter_add(&mm->rss_stat[member], value);
2231 mm_trace_rss_stat(mm, member);
2234 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2236 percpu_counter_inc(&mm->rss_stat[member]);
2238 mm_trace_rss_stat(mm, member);
2241 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2243 percpu_counter_dec(&mm->rss_stat[member]);
2245 mm_trace_rss_stat(mm, member);
2248 /* Optimized variant when page is already known not to be PageAnon */
2249 static inline int mm_counter_file(struct page *page)
2251 if (PageSwapBacked(page))
2252 return MM_SHMEMPAGES;
2253 return MM_FILEPAGES;
2256 static inline int mm_counter(struct page *page)
2259 return MM_ANONPAGES;
2260 return mm_counter_file(page);
2263 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2265 return get_mm_counter(mm, MM_FILEPAGES) +
2266 get_mm_counter(mm, MM_ANONPAGES) +
2267 get_mm_counter(mm, MM_SHMEMPAGES);
2270 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2272 return max(mm->hiwater_rss, get_mm_rss(mm));
2275 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2277 return max(mm->hiwater_vm, mm->total_vm);
2280 static inline void update_hiwater_rss(struct mm_struct *mm)
2282 unsigned long _rss = get_mm_rss(mm);
2284 if ((mm)->hiwater_rss < _rss)
2285 (mm)->hiwater_rss = _rss;
2288 static inline void update_hiwater_vm(struct mm_struct *mm)
2290 if (mm->hiwater_vm < mm->total_vm)
2291 mm->hiwater_vm = mm->total_vm;
2294 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2296 mm->hiwater_rss = get_mm_rss(mm);
2299 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2300 struct mm_struct *mm)
2302 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2304 if (*maxrss < hiwater_rss)
2305 *maxrss = hiwater_rss;
2308 #if defined(SPLIT_RSS_COUNTING)
2309 void sync_mm_rss(struct mm_struct *mm);
2311 static inline void sync_mm_rss(struct mm_struct *mm)
2316 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2317 static inline int pte_special(pte_t pte)
2322 static inline pte_t pte_mkspecial(pte_t pte)
2328 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2329 static inline int pte_devmap(pte_t pte)
2335 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2337 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2341 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2345 #ifdef __PAGETABLE_P4D_FOLDED
2346 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2347 unsigned long address)
2352 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2355 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2356 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2357 unsigned long address)
2361 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2362 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2365 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2367 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2369 if (mm_pud_folded(mm))
2371 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2374 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2376 if (mm_pud_folded(mm))
2378 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2382 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2383 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2384 unsigned long address)
2389 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2390 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2393 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2395 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2397 if (mm_pmd_folded(mm))
2399 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2402 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2404 if (mm_pmd_folded(mm))
2406 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2411 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2413 atomic_long_set(&mm->pgtables_bytes, 0);
2416 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2418 return atomic_long_read(&mm->pgtables_bytes);
2421 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2423 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2426 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2428 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2432 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2433 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2438 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2439 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2442 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2443 int __pte_alloc_kernel(pmd_t *pmd);
2445 #if defined(CONFIG_MMU)
2447 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2448 unsigned long address)
2450 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2451 NULL : p4d_offset(pgd, address);
2454 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2455 unsigned long address)
2457 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2458 NULL : pud_offset(p4d, address);
2461 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2463 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2464 NULL: pmd_offset(pud, address);
2466 #endif /* CONFIG_MMU */
2468 #if USE_SPLIT_PTE_PTLOCKS
2469 #if ALLOC_SPLIT_PTLOCKS
2470 void __init ptlock_cache_init(void);
2471 extern bool ptlock_alloc(struct page *page);
2472 extern void ptlock_free(struct page *page);
2474 static inline spinlock_t *ptlock_ptr(struct page *page)
2478 #else /* ALLOC_SPLIT_PTLOCKS */
2479 static inline void ptlock_cache_init(void)
2483 static inline bool ptlock_alloc(struct page *page)
2488 static inline void ptlock_free(struct page *page)
2492 static inline spinlock_t *ptlock_ptr(struct page *page)
2496 #endif /* ALLOC_SPLIT_PTLOCKS */
2498 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2500 return ptlock_ptr(pmd_page(*pmd));
2503 static inline bool ptlock_init(struct page *page)
2506 * prep_new_page() initialize page->private (and therefore page->ptl)
2507 * with 0. Make sure nobody took it in use in between.
2509 * It can happen if arch try to use slab for page table allocation:
2510 * slab code uses page->slab_cache, which share storage with page->ptl.
2512 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2513 if (!ptlock_alloc(page))
2515 spin_lock_init(ptlock_ptr(page));
2519 #else /* !USE_SPLIT_PTE_PTLOCKS */
2521 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2523 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2525 return &mm->page_table_lock;
2527 static inline void ptlock_cache_init(void) {}
2528 static inline bool ptlock_init(struct page *page) { return true; }
2529 static inline void ptlock_free(struct page *page) {}
2530 #endif /* USE_SPLIT_PTE_PTLOCKS */
2532 static inline void pgtable_init(void)
2534 ptlock_cache_init();
2535 pgtable_cache_init();
2538 static inline bool pgtable_pte_page_ctor(struct page *page)
2540 if (!ptlock_init(page))
2542 __SetPageTable(page);
2543 inc_lruvec_page_state(page, NR_PAGETABLE);
2547 static inline void pgtable_pte_page_dtor(struct page *page)
2550 __ClearPageTable(page);
2551 dec_lruvec_page_state(page, NR_PAGETABLE);
2554 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2556 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2557 pte_t *__pte = pte_offset_map(pmd, address); \
2563 #define pte_unmap_unlock(pte, ptl) do { \
2568 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2570 #define pte_alloc_map(mm, pmd, address) \
2571 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2573 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2574 (pte_alloc(mm, pmd) ? \
2575 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2577 #define pte_alloc_kernel(pmd, address) \
2578 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2579 NULL: pte_offset_kernel(pmd, address))
2581 #if USE_SPLIT_PMD_PTLOCKS
2583 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2585 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2586 return virt_to_page((void *)((unsigned long) pmd & mask));
2589 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2591 return ptlock_ptr(pmd_pgtable_page(pmd));
2594 static inline bool pmd_ptlock_init(struct page *page)
2596 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2597 page->pmd_huge_pte = NULL;
2599 return ptlock_init(page);
2602 static inline void pmd_ptlock_free(struct page *page)
2604 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2605 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2610 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2614 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2616 return &mm->page_table_lock;
2619 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2620 static inline void pmd_ptlock_free(struct page *page) {}
2622 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2626 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2628 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2633 static inline bool pgtable_pmd_page_ctor(struct page *page)
2635 if (!pmd_ptlock_init(page))
2637 __SetPageTable(page);
2638 inc_lruvec_page_state(page, NR_PAGETABLE);
2642 static inline void pgtable_pmd_page_dtor(struct page *page)
2644 pmd_ptlock_free(page);
2645 __ClearPageTable(page);
2646 dec_lruvec_page_state(page, NR_PAGETABLE);
2650 * No scalability reason to split PUD locks yet, but follow the same pattern
2651 * as the PMD locks to make it easier if we decide to. The VM should not be
2652 * considered ready to switch to split PUD locks yet; there may be places
2653 * which need to be converted from page_table_lock.
2655 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2657 return &mm->page_table_lock;
2660 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2662 spinlock_t *ptl = pud_lockptr(mm, pud);
2668 extern void __init pagecache_init(void);
2669 extern void free_initmem(void);
2672 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2673 * into the buddy system. The freed pages will be poisoned with pattern
2674 * "poison" if it's within range [0, UCHAR_MAX].
2675 * Return pages freed into the buddy system.
2677 extern unsigned long free_reserved_area(void *start, void *end,
2678 int poison, const char *s);
2680 extern void adjust_managed_page_count(struct page *page, long count);
2681 extern void mem_init_print_info(void);
2683 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2685 /* Free the reserved page into the buddy system, so it gets managed. */
2686 static inline void free_reserved_page(struct page *page)
2688 ClearPageReserved(page);
2689 init_page_count(page);
2691 adjust_managed_page_count(page, 1);
2693 #define free_highmem_page(page) free_reserved_page(page)
2695 static inline void mark_page_reserved(struct page *page)
2697 SetPageReserved(page);
2698 adjust_managed_page_count(page, -1);
2702 * Default method to free all the __init memory into the buddy system.
2703 * The freed pages will be poisoned with pattern "poison" if it's within
2704 * range [0, UCHAR_MAX].
2705 * Return pages freed into the buddy system.
2707 static inline unsigned long free_initmem_default(int poison)
2709 extern char __init_begin[], __init_end[];
2711 return free_reserved_area(&__init_begin, &__init_end,
2712 poison, "unused kernel image (initmem)");
2715 static inline unsigned long get_num_physpages(void)
2718 unsigned long phys_pages = 0;
2720 for_each_online_node(nid)
2721 phys_pages += node_present_pages(nid);
2727 * Using memblock node mappings, an architecture may initialise its
2728 * zones, allocate the backing mem_map and account for memory holes in an
2729 * architecture independent manner.
2731 * An architecture is expected to register range of page frames backed by
2732 * physical memory with memblock_add[_node]() before calling
2733 * free_area_init() passing in the PFN each zone ends at. At a basic
2734 * usage, an architecture is expected to do something like
2736 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2738 * for_each_valid_physical_page_range()
2739 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2740 * free_area_init(max_zone_pfns);
2742 void free_area_init(unsigned long *max_zone_pfn);
2743 unsigned long node_map_pfn_alignment(void);
2744 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2745 unsigned long end_pfn);
2746 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2747 unsigned long end_pfn);
2748 extern void get_pfn_range_for_nid(unsigned int nid,
2749 unsigned long *start_pfn, unsigned long *end_pfn);
2752 static inline int early_pfn_to_nid(unsigned long pfn)
2757 /* please see mm/page_alloc.c */
2758 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2761 extern void set_dma_reserve(unsigned long new_dma_reserve);
2762 extern void memmap_init_range(unsigned long, int, unsigned long,
2763 unsigned long, unsigned long, enum meminit_context,
2764 struct vmem_altmap *, int migratetype);
2765 extern void setup_per_zone_wmarks(void);
2766 extern void calculate_min_free_kbytes(void);
2767 extern int __meminit init_per_zone_wmark_min(void);
2768 extern void mem_init(void);
2769 extern void __init mmap_init(void);
2771 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
2772 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2774 __show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2776 extern long si_mem_available(void);
2777 extern void si_meminfo(struct sysinfo * val);
2778 extern void si_meminfo_node(struct sysinfo *val, int nid);
2779 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2780 extern unsigned long arch_reserved_kernel_pages(void);
2783 extern __printf(3, 4)
2784 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2786 extern void setup_per_cpu_pageset(void);
2789 extern int min_free_kbytes;
2790 extern int watermark_boost_factor;
2791 extern int watermark_scale_factor;
2792 extern bool arch_has_descending_max_zone_pfns(void);
2795 extern atomic_long_t mmap_pages_allocated;
2796 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2798 /* interval_tree.c */
2799 void vma_interval_tree_insert(struct vm_area_struct *node,
2800 struct rb_root_cached *root);
2801 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2802 struct vm_area_struct *prev,
2803 struct rb_root_cached *root);
2804 void vma_interval_tree_remove(struct vm_area_struct *node,
2805 struct rb_root_cached *root);
2806 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2807 unsigned long start, unsigned long last);
2808 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2809 unsigned long start, unsigned long last);
2811 #define vma_interval_tree_foreach(vma, root, start, last) \
2812 for (vma = vma_interval_tree_iter_first(root, start, last); \
2813 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2815 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2816 struct rb_root_cached *root);
2817 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2818 struct rb_root_cached *root);
2819 struct anon_vma_chain *
2820 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2821 unsigned long start, unsigned long last);
2822 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2823 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2824 #ifdef CONFIG_DEBUG_VM_RB
2825 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2828 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2829 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2830 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2833 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2834 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2835 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2836 struct vm_area_struct *expand);
2837 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2838 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2840 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2842 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2843 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2844 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2845 struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2846 extern struct vm_area_struct *vmi_vma_merge(struct vma_iterator *vmi,
2847 struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
2848 unsigned long end, unsigned long vm_flags, struct anon_vma *,
2849 struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
2850 struct anon_vma_name *);
2851 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2852 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2853 unsigned long addr, int new_below);
2854 extern int vmi__split_vma(struct vma_iterator *vmi, struct mm_struct *,
2855 struct vm_area_struct *, unsigned long addr, int new_below);
2856 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2857 unsigned long addr, int new_below);
2858 extern int vmi_split_vma(struct vma_iterator *vmi, struct mm_struct *,
2859 struct vm_area_struct *, unsigned long addr, int new_below);
2860 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2861 extern void unlink_file_vma(struct vm_area_struct *);
2862 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2863 unsigned long addr, unsigned long len, pgoff_t pgoff,
2864 bool *need_rmap_locks);
2865 extern void exit_mmap(struct mm_struct *);
2867 void vma_mas_store(struct vm_area_struct *vma, struct ma_state *mas);
2868 void vma_mas_remove(struct vm_area_struct *vma, struct ma_state *mas);
2870 static inline int check_data_rlimit(unsigned long rlim,
2872 unsigned long start,
2873 unsigned long end_data,
2874 unsigned long start_data)
2876 if (rlim < RLIM_INFINITY) {
2877 if (((new - start) + (end_data - start_data)) > rlim)
2884 extern int mm_take_all_locks(struct mm_struct *mm);
2885 extern void mm_drop_all_locks(struct mm_struct *mm);
2887 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2888 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2889 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2890 extern struct file *get_task_exe_file(struct task_struct *task);
2892 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2893 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2895 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2896 const struct vm_special_mapping *sm);
2897 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2898 unsigned long addr, unsigned long len,
2899 unsigned long flags,
2900 const struct vm_special_mapping *spec);
2901 /* This is an obsolete alternative to _install_special_mapping. */
2902 extern int install_special_mapping(struct mm_struct *mm,
2903 unsigned long addr, unsigned long len,
2904 unsigned long flags, struct page **pages);
2906 unsigned long randomize_stack_top(unsigned long stack_top);
2907 unsigned long randomize_page(unsigned long start, unsigned long range);
2909 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2911 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2912 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2913 struct list_head *uf);
2914 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2915 unsigned long len, unsigned long prot, unsigned long flags,
2916 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2917 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
2918 unsigned long start, size_t len, struct list_head *uf,
2920 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2921 struct list_head *uf);
2922 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2925 extern int __mm_populate(unsigned long addr, unsigned long len,
2927 static inline void mm_populate(unsigned long addr, unsigned long len)
2930 (void) __mm_populate(addr, len, 1);
2933 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2936 /* These take the mm semaphore themselves */
2937 extern int __must_check vm_brk(unsigned long, unsigned long);
2938 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2939 extern int vm_munmap(unsigned long, size_t);
2940 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2941 unsigned long, unsigned long,
2942 unsigned long, unsigned long);
2944 struct vm_unmapped_area_info {
2945 #define VM_UNMAPPED_AREA_TOPDOWN 1
2946 unsigned long flags;
2947 unsigned long length;
2948 unsigned long low_limit;
2949 unsigned long high_limit;
2950 unsigned long align_mask;
2951 unsigned long align_offset;
2954 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2957 extern void truncate_inode_pages(struct address_space *, loff_t);
2958 extern void truncate_inode_pages_range(struct address_space *,
2959 loff_t lstart, loff_t lend);
2960 extern void truncate_inode_pages_final(struct address_space *);
2962 /* generic vm_area_ops exported for stackable file systems */
2963 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2964 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2965 pgoff_t start_pgoff, pgoff_t end_pgoff);
2966 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2968 extern unsigned long stack_guard_gap;
2969 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2970 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2972 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2973 extern int expand_downwards(struct vm_area_struct *vma,
2974 unsigned long address);
2976 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2978 #define expand_upwards(vma, address) (0)
2981 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2982 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2983 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2984 struct vm_area_struct **pprev);
2987 * Look up the first VMA which intersects the interval [start_addr, end_addr)
2988 * NULL if none. Assume start_addr < end_addr.
2990 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2991 unsigned long start_addr, unsigned long end_addr);
2994 * vma_lookup() - Find a VMA at a specific address
2995 * @mm: The process address space.
2996 * @addr: The user address.
2998 * Return: The vm_area_struct at the given address, %NULL otherwise.
3001 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3003 return mtree_load(&mm->mm_mt, addr);
3006 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3008 unsigned long vm_start = vma->vm_start;
3010 if (vma->vm_flags & VM_GROWSDOWN) {
3011 vm_start -= stack_guard_gap;
3012 if (vm_start > vma->vm_start)
3018 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3020 unsigned long vm_end = vma->vm_end;
3022 if (vma->vm_flags & VM_GROWSUP) {
3023 vm_end += stack_guard_gap;
3024 if (vm_end < vma->vm_end)
3025 vm_end = -PAGE_SIZE;
3030 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3032 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3035 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3036 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3037 unsigned long vm_start, unsigned long vm_end)
3039 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3041 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3047 static inline bool range_in_vma(struct vm_area_struct *vma,
3048 unsigned long start, unsigned long end)
3050 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3054 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3055 void vma_set_page_prot(struct vm_area_struct *vma);
3057 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3061 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3063 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3067 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3069 #ifdef CONFIG_NUMA_BALANCING
3070 unsigned long change_prot_numa(struct vm_area_struct *vma,
3071 unsigned long start, unsigned long end);
3074 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
3075 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3076 unsigned long pfn, unsigned long size, pgprot_t);
3077 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3078 unsigned long pfn, unsigned long size, pgprot_t prot);
3079 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3080 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3081 struct page **pages, unsigned long *num);
3082 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3084 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3086 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3088 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3089 unsigned long pfn, pgprot_t pgprot);
3090 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3092 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
3093 pfn_t pfn, pgprot_t pgprot);
3094 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3095 unsigned long addr, pfn_t pfn);
3096 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3098 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3099 unsigned long addr, struct page *page)
3101 int err = vm_insert_page(vma, addr, page);
3104 return VM_FAULT_OOM;
3105 if (err < 0 && err != -EBUSY)
3106 return VM_FAULT_SIGBUS;
3108 return VM_FAULT_NOPAGE;
3111 #ifndef io_remap_pfn_range
3112 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3113 unsigned long addr, unsigned long pfn,
3114 unsigned long size, pgprot_t prot)
3116 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3120 static inline vm_fault_t vmf_error(int err)
3123 return VM_FAULT_OOM;
3124 return VM_FAULT_SIGBUS;
3127 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3128 unsigned int foll_flags);
3130 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3132 if (vm_fault & VM_FAULT_OOM)
3134 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3135 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3136 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3142 * Indicates for which pages that are write-protected in the page table,
3143 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
3144 * GUP pin will remain consistent with the pages mapped into the page tables
3147 * Temporary unmapping of PageAnonExclusive() pages or clearing of
3148 * PageAnonExclusive() has to protect against concurrent GUP:
3149 * * Ordinary GUP: Using the PT lock
3150 * * GUP-fast and fork(): mm->write_protect_seq
3151 * * GUP-fast and KSM or temporary unmapping (swap, migration): see
3152 * page_try_share_anon_rmap()
3154 * Must be called with the (sub)page that's actually referenced via the
3155 * page table entry, which might not necessarily be the head page for a
3158 * If the vma is NULL, we're coming from the GUP-fast path and might have
3159 * to fallback to the slow path just to lookup the vma.
3161 static inline bool gup_must_unshare(struct vm_area_struct *vma,
3162 unsigned int flags, struct page *page)
3165 * FOLL_WRITE is implicitly handled correctly as the page table entry
3166 * has to be writable -- and if it references (part of) an anonymous
3167 * folio, that part is required to be marked exclusive.
3169 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
3172 * Note: PageAnon(page) is stable until the page is actually getting
3175 if (!PageAnon(page)) {
3177 * We only care about R/O long-term pining: R/O short-term
3178 * pinning does not have the semantics to observe successive
3179 * changes through the process page tables.
3181 if (!(flags & FOLL_LONGTERM))
3184 /* We really need the vma ... */
3189 * ... because we only care about writable private ("COW")
3190 * mappings where we have to break COW early.
3192 return is_cow_mapping(vma->vm_flags);
3195 /* Paired with a memory barrier in page_try_share_anon_rmap(). */
3196 if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
3200 * Note that PageKsm() pages cannot be exclusive, and consequently,
3201 * cannot get pinned.
3203 return !PageAnonExclusive(page);
3207 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3208 * a (NUMA hinting) fault is required.
3210 static inline bool gup_can_follow_protnone(unsigned int flags)
3213 * FOLL_FORCE has to be able to make progress even if the VMA is
3214 * inaccessible. Further, FOLL_FORCE access usually does not represent
3215 * application behaviour and we should avoid triggering NUMA hinting
3218 return flags & FOLL_FORCE;
3221 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3222 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3223 unsigned long size, pte_fn_t fn, void *data);
3224 extern int apply_to_existing_page_range(struct mm_struct *mm,
3225 unsigned long address, unsigned long size,
3226 pte_fn_t fn, void *data);
3228 extern void __init init_mem_debugging_and_hardening(void);
3229 #ifdef CONFIG_PAGE_POISONING
3230 extern void __kernel_poison_pages(struct page *page, int numpages);
3231 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3232 extern bool _page_poisoning_enabled_early;
3233 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3234 static inline bool page_poisoning_enabled(void)
3236 return _page_poisoning_enabled_early;
3239 * For use in fast paths after init_mem_debugging() has run, or when a
3240 * false negative result is not harmful when called too early.
3242 static inline bool page_poisoning_enabled_static(void)
3244 return static_branch_unlikely(&_page_poisoning_enabled);
3246 static inline void kernel_poison_pages(struct page *page, int numpages)
3248 if (page_poisoning_enabled_static())
3249 __kernel_poison_pages(page, numpages);
3251 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3253 if (page_poisoning_enabled_static())
3254 __kernel_unpoison_pages(page, numpages);
3257 static inline bool page_poisoning_enabled(void) { return false; }
3258 static inline bool page_poisoning_enabled_static(void) { return false; }
3259 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3260 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3261 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3264 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3265 static inline bool want_init_on_alloc(gfp_t flags)
3267 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3270 return flags & __GFP_ZERO;
3273 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3274 static inline bool want_init_on_free(void)
3276 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3280 extern bool _debug_pagealloc_enabled_early;
3281 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3283 static inline bool debug_pagealloc_enabled(void)
3285 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3286 _debug_pagealloc_enabled_early;
3290 * For use in fast paths after init_debug_pagealloc() has run, or when a
3291 * false negative result is not harmful when called too early.
3293 static inline bool debug_pagealloc_enabled_static(void)
3295 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3298 return static_branch_unlikely(&_debug_pagealloc_enabled);
3301 #ifdef CONFIG_DEBUG_PAGEALLOC
3303 * To support DEBUG_PAGEALLOC architecture must ensure that
3304 * __kernel_map_pages() never fails
3306 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3308 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3310 if (debug_pagealloc_enabled_static())
3311 __kernel_map_pages(page, numpages, 1);
3314 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3316 if (debug_pagealloc_enabled_static())
3317 __kernel_map_pages(page, numpages, 0);
3319 #else /* CONFIG_DEBUG_PAGEALLOC */
3320 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3321 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3322 #endif /* CONFIG_DEBUG_PAGEALLOC */
3324 #ifdef __HAVE_ARCH_GATE_AREA
3325 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3326 extern int in_gate_area_no_mm(unsigned long addr);
3327 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3329 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3333 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3334 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3338 #endif /* __HAVE_ARCH_GATE_AREA */
3340 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3342 #ifdef CONFIG_SYSCTL
3343 extern int sysctl_drop_caches;
3344 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3348 void drop_slab(void);
3351 #define randomize_va_space 0
3353 extern int randomize_va_space;
3356 const char * arch_vma_name(struct vm_area_struct *vma);
3358 void print_vma_addr(char *prefix, unsigned long rip);
3360 static inline void print_vma_addr(char *prefix, unsigned long rip)
3365 void *sparse_buffer_alloc(unsigned long size);
3366 struct page * __populate_section_memmap(unsigned long pfn,
3367 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3368 struct dev_pagemap *pgmap);
3369 void pmd_init(void *addr);
3370 void pud_init(void *addr);
3371 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3372 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3373 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3374 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3375 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3376 struct vmem_altmap *altmap, struct page *reuse);
3377 void *vmemmap_alloc_block(unsigned long size, int node);
3379 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3380 struct vmem_altmap *altmap);
3381 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3382 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3383 unsigned long addr, unsigned long next);
3384 int vmemmap_check_pmd(pmd_t *pmd, int node,
3385 unsigned long addr, unsigned long next);
3386 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3387 int node, struct vmem_altmap *altmap);
3388 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3389 int node, struct vmem_altmap *altmap);
3390 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3391 struct vmem_altmap *altmap);
3392 void vmemmap_populate_print_last(void);
3393 #ifdef CONFIG_MEMORY_HOTPLUG
3394 void vmemmap_free(unsigned long start, unsigned long end,
3395 struct vmem_altmap *altmap);
3397 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3398 unsigned long nr_pages);
3401 MF_COUNT_INCREASED = 1 << 0,
3402 MF_ACTION_REQUIRED = 1 << 1,
3403 MF_MUST_KILL = 1 << 2,
3404 MF_SOFT_OFFLINE = 1 << 3,
3405 MF_UNPOISON = 1 << 4,
3406 MF_SW_SIMULATED = 1 << 5,
3407 MF_NO_RETRY = 1 << 6,
3409 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3410 unsigned long count, int mf_flags);
3411 extern int memory_failure(unsigned long pfn, int flags);
3412 extern void memory_failure_queue_kick(int cpu);
3413 extern int unpoison_memory(unsigned long pfn);
3414 extern int sysctl_memory_failure_early_kill;
3415 extern int sysctl_memory_failure_recovery;
3416 extern void shake_page(struct page *p);
3417 extern atomic_long_t num_poisoned_pages __read_mostly;
3418 extern int soft_offline_page(unsigned long pfn, int flags);
3419 #ifdef CONFIG_MEMORY_FAILURE
3420 extern void memory_failure_queue(unsigned long pfn, int flags);
3421 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3422 bool *migratable_cleared);
3423 void num_poisoned_pages_inc(unsigned long pfn);
3424 void num_poisoned_pages_sub(unsigned long pfn, long i);
3426 static inline void memory_failure_queue(unsigned long pfn, int flags)
3430 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3431 bool *migratable_cleared)
3436 static inline void num_poisoned_pages_inc(unsigned long pfn)
3440 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3445 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3446 extern void memblk_nr_poison_inc(unsigned long pfn);
3447 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3449 static inline void memblk_nr_poison_inc(unsigned long pfn)
3453 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3458 #ifndef arch_memory_failure
3459 static inline int arch_memory_failure(unsigned long pfn, int flags)
3465 #ifndef arch_is_platform_page
3466 static inline bool arch_is_platform_page(u64 paddr)
3473 * Error handlers for various types of pages.
3476 MF_IGNORED, /* Error: cannot be handled */
3477 MF_FAILED, /* Error: handling failed */
3478 MF_DELAYED, /* Will be handled later */
3479 MF_RECOVERED, /* Successfully recovered */
3482 enum mf_action_page_type {
3484 MF_MSG_KERNEL_HIGH_ORDER,
3486 MF_MSG_DIFFERENT_COMPOUND,
3489 MF_MSG_UNMAP_FAILED,
3490 MF_MSG_DIRTY_SWAPCACHE,
3491 MF_MSG_CLEAN_SWAPCACHE,
3492 MF_MSG_DIRTY_MLOCKED_LRU,
3493 MF_MSG_CLEAN_MLOCKED_LRU,
3494 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3495 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3498 MF_MSG_TRUNCATED_LRU,
3506 * Sysfs entries for memory failure handling statistics.
3508 extern const struct attribute_group memory_failure_attr_group;
3510 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3511 extern void clear_huge_page(struct page *page,
3512 unsigned long addr_hint,
3513 unsigned int pages_per_huge_page);
3514 extern void copy_user_huge_page(struct page *dst, struct page *src,
3515 unsigned long addr_hint,
3516 struct vm_area_struct *vma,
3517 unsigned int pages_per_huge_page);
3518 extern long copy_huge_page_from_user(struct page *dst_page,
3519 const void __user *usr_src,
3520 unsigned int pages_per_huge_page,
3521 bool allow_pagefault);
3524 * vma_is_special_huge - Are transhuge page-table entries considered special?
3525 * @vma: Pointer to the struct vm_area_struct to consider
3527 * Whether transhuge page-table entries are considered "special" following
3528 * the definition in vm_normal_page().
3530 * Return: true if transhuge page-table entries should be considered special,
3533 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3535 return vma_is_dax(vma) || (vma->vm_file &&
3536 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3539 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3541 #ifdef CONFIG_DEBUG_PAGEALLOC
3542 extern unsigned int _debug_guardpage_minorder;
3543 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3545 static inline unsigned int debug_guardpage_minorder(void)
3547 return _debug_guardpage_minorder;
3550 static inline bool debug_guardpage_enabled(void)
3552 return static_branch_unlikely(&_debug_guardpage_enabled);
3555 static inline bool page_is_guard(struct page *page)
3557 if (!debug_guardpage_enabled())
3560 return PageGuard(page);
3563 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3564 static inline bool debug_guardpage_enabled(void) { return false; }
3565 static inline bool page_is_guard(struct page *page) { return false; }
3566 #endif /* CONFIG_DEBUG_PAGEALLOC */
3568 #if MAX_NUMNODES > 1
3569 void __init setup_nr_node_ids(void);
3571 static inline void setup_nr_node_ids(void) {}
3574 extern int memcmp_pages(struct page *page1, struct page *page2);
3576 static inline int pages_identical(struct page *page1, struct page *page2)
3578 return !memcmp_pages(page1, page2);
3581 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3582 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3583 pgoff_t first_index, pgoff_t nr,
3584 pgoff_t bitmap_pgoff,
3585 unsigned long *bitmap,
3589 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3590 pgoff_t first_index, pgoff_t nr);
3593 extern int sysctl_nr_trim_pages;
3595 #ifdef CONFIG_PRINTK
3596 void mem_dump_obj(void *object);
3598 static inline void mem_dump_obj(void *object) {}
3602 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3603 * @seals: the seals to check
3604 * @vma: the vma to operate on
3606 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3607 * the vma flags. Return 0 if check pass, or <0 for errors.
3609 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3611 if (seals & F_SEAL_FUTURE_WRITE) {
3613 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3614 * "future write" seal active.
3616 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3620 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3621 * MAP_SHARED and read-only, take care to not allow mprotect to
3622 * revert protections on such mappings. Do this only for shared
3623 * mappings. For private mappings, don't need to mask
3624 * VM_MAYWRITE as we still want them to be COW-writable.
3626 if (vma->vm_flags & VM_SHARED)
3627 vma->vm_flags &= ~(VM_MAYWRITE);
3633 #ifdef CONFIG_ANON_VMA_NAME
3634 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3635 unsigned long len_in,
3636 struct anon_vma_name *anon_name);
3639 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3640 unsigned long len_in, struct anon_vma_name *anon_name) {
3645 #endif /* _LINUX_MM_H */