1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
34 #include <linux/kasan.h>
38 struct anon_vma_chain;
42 extern int sysctl_page_lock_unfairness;
44 void init_mm_internals(void);
46 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
47 extern unsigned long max_mapnr;
49 static inline void set_max_mapnr(unsigned long limit)
54 static inline void set_max_mapnr(unsigned long limit) { }
57 extern atomic_long_t _totalram_pages;
58 static inline unsigned long totalram_pages(void)
60 return (unsigned long)atomic_long_read(&_totalram_pages);
63 static inline void totalram_pages_inc(void)
65 atomic_long_inc(&_totalram_pages);
68 static inline void totalram_pages_dec(void)
70 atomic_long_dec(&_totalram_pages);
73 static inline void totalram_pages_add(long count)
75 atomic_long_add(count, &_totalram_pages);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/processor.h>
102 * Architectures that support memory tagging (assigning tags to memory regions,
103 * embedding these tags into addresses that point to these memory regions, and
104 * checking that the memory and the pointer tags match on memory accesses)
105 * redefine this macro to strip tags from pointers.
106 * It's defined as noop for architectures that don't support memory tagging.
108 #ifndef untagged_addr
109 #define untagged_addr(addr) (addr)
113 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 #define lm_alias(x) __va(__pa_symbol(x))
125 * To prevent common memory management code establishing
126 * a zero page mapping on a read fault.
127 * This macro should be defined within <asm/pgtable.h>.
128 * s390 does this to prevent multiplexing of hardware bits
129 * related to the physical page in case of virtualization.
131 #ifndef mm_forbids_zeropage
132 #define mm_forbids_zeropage(X) (0)
136 * On some architectures it is expensive to call memset() for small sizes.
137 * If an architecture decides to implement their own version of
138 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
139 * define their own version of this macro in <asm/pgtable.h>
141 #if BITS_PER_LONG == 64
142 /* This function must be updated when the size of struct page grows above 80
143 * or reduces below 56. The idea that compiler optimizes out switch()
144 * statement, and only leaves move/store instructions. Also the compiler can
145 * combine write statements if they are both assignments and can be reordered,
146 * this can result in several of the writes here being dropped.
148 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
149 static inline void __mm_zero_struct_page(struct page *page)
151 unsigned long *_pp = (void *)page;
153 /* Check that struct page is either 56, 64, 72, or 80 bytes */
154 BUILD_BUG_ON(sizeof(struct page) & 7);
155 BUILD_BUG_ON(sizeof(struct page) < 56);
156 BUILD_BUG_ON(sizeof(struct page) > 80);
158 switch (sizeof(struct page)) {
179 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
183 * Default maximum number of active map areas, this limits the number of vmas
184 * per mm struct. Users can overwrite this number by sysctl but there is a
187 * When a program's coredump is generated as ELF format, a section is created
188 * per a vma. In ELF, the number of sections is represented in unsigned short.
189 * This means the number of sections should be smaller than 65535 at coredump.
190 * Because the kernel adds some informative sections to a image of program at
191 * generating coredump, we need some margin. The number of extra sections is
192 * 1-3 now and depends on arch. We use "5" as safe margin, here.
194 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
195 * not a hard limit any more. Although some userspace tools can be surprised by
198 #define MAPCOUNT_ELF_CORE_MARGIN (5)
199 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
201 extern int sysctl_max_map_count;
203 extern unsigned long sysctl_user_reserve_kbytes;
204 extern unsigned long sysctl_admin_reserve_kbytes;
206 extern int sysctl_overcommit_memory;
207 extern int sysctl_overcommit_ratio;
208 extern unsigned long sysctl_overcommit_kbytes;
210 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
212 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
214 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
217 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
218 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
220 #define nth_page(page,n) ((page) + (n))
223 /* to align the pointer to the (next) page boundary */
224 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
226 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
227 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
229 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
231 void setup_initial_init_mm(void *start_code, void *end_code,
232 void *end_data, void *brk);
235 * Linux kernel virtual memory manager primitives.
236 * The idea being to have a "virtual" mm in the same way
237 * we have a virtual fs - giving a cleaner interface to the
238 * mm details, and allowing different kinds of memory mappings
239 * (from shared memory to executable loading to arbitrary
243 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
244 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
245 void vm_area_free(struct vm_area_struct *);
248 extern struct rb_root nommu_region_tree;
249 extern struct rw_semaphore nommu_region_sem;
251 extern unsigned int kobjsize(const void *objp);
255 * vm_flags in vm_area_struct, see mm_types.h.
256 * When changing, update also include/trace/events/mmflags.h
258 #define VM_NONE 0x00000000
260 #define VM_READ 0x00000001 /* currently active flags */
261 #define VM_WRITE 0x00000002
262 #define VM_EXEC 0x00000004
263 #define VM_SHARED 0x00000008
265 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
266 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
267 #define VM_MAYWRITE 0x00000020
268 #define VM_MAYEXEC 0x00000040
269 #define VM_MAYSHARE 0x00000080
271 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
272 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
273 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
274 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
276 #define VM_LOCKED 0x00002000
277 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
279 /* Used by sys_madvise() */
280 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
281 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
283 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
284 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
285 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
286 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
287 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
288 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
289 #define VM_SYNC 0x00800000 /* Synchronous page faults */
290 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
291 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
292 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
294 #ifdef CONFIG_MEM_SOFT_DIRTY
295 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
297 # define VM_SOFTDIRTY 0
300 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
301 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
302 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
303 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
305 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
306 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
307 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
308 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
309 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
310 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
311 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
312 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
313 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
314 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
315 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
316 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
318 #ifdef CONFIG_ARCH_HAS_PKEYS
319 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
320 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
321 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
322 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
323 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
325 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
327 # define VM_PKEY_BIT4 0
329 #endif /* CONFIG_ARCH_HAS_PKEYS */
331 #if defined(CONFIG_X86)
332 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
333 #elif defined(CONFIG_PPC)
334 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
335 #elif defined(CONFIG_PARISC)
336 # define VM_GROWSUP VM_ARCH_1
337 #elif defined(CONFIG_IA64)
338 # define VM_GROWSUP VM_ARCH_1
339 #elif defined(CONFIG_SPARC64)
340 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
341 # define VM_ARCH_CLEAR VM_SPARC_ADI
342 #elif defined(CONFIG_ARM64)
343 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
344 # define VM_ARCH_CLEAR VM_ARM64_BTI
345 #elif !defined(CONFIG_MMU)
346 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
349 #if defined(CONFIG_ARM64_MTE)
350 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
351 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
353 # define VM_MTE VM_NONE
354 # define VM_MTE_ALLOWED VM_NONE
358 # define VM_GROWSUP VM_NONE
361 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
362 # define VM_UFFD_MINOR_BIT 37
363 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
364 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
365 # define VM_UFFD_MINOR VM_NONE
366 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
368 /* Bits set in the VMA until the stack is in its final location */
369 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
371 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
373 /* Common data flag combinations */
374 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
375 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
376 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
377 VM_MAYWRITE | VM_MAYEXEC)
378 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
379 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
381 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
382 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
385 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
386 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
389 #ifdef CONFIG_STACK_GROWSUP
390 #define VM_STACK VM_GROWSUP
392 #define VM_STACK VM_GROWSDOWN
395 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
397 /* VMA basic access permission flags */
398 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
402 * Special vmas that are non-mergable, non-mlock()able.
404 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
406 /* This mask prevents VMA from being scanned with khugepaged */
407 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
409 /* This mask defines which mm->def_flags a process can inherit its parent */
410 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
412 /* This mask is used to clear all the VMA flags used by mlock */
413 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
415 /* Arch-specific flags to clear when updating VM flags on protection change */
416 #ifndef VM_ARCH_CLEAR
417 # define VM_ARCH_CLEAR VM_NONE
419 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
422 * mapping from the currently active vm_flags protection bits (the
423 * low four bits) to a page protection mask..
425 extern pgprot_t protection_map[16];
428 * The default fault flags that should be used by most of the
429 * arch-specific page fault handlers.
431 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
432 FAULT_FLAG_KILLABLE | \
433 FAULT_FLAG_INTERRUPTIBLE)
436 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
437 * @flags: Fault flags.
439 * This is mostly used for places where we want to try to avoid taking
440 * the mmap_lock for too long a time when waiting for another condition
441 * to change, in which case we can try to be polite to release the
442 * mmap_lock in the first round to avoid potential starvation of other
443 * processes that would also want the mmap_lock.
445 * Return: true if the page fault allows retry and this is the first
446 * attempt of the fault handling; false otherwise.
448 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
450 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
451 (!(flags & FAULT_FLAG_TRIED));
454 #define FAULT_FLAG_TRACE \
455 { FAULT_FLAG_WRITE, "WRITE" }, \
456 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
457 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
458 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
459 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
460 { FAULT_FLAG_TRIED, "TRIED" }, \
461 { FAULT_FLAG_USER, "USER" }, \
462 { FAULT_FLAG_REMOTE, "REMOTE" }, \
463 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
464 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
467 * vm_fault is filled by the pagefault handler and passed to the vma's
468 * ->fault function. The vma's ->fault is responsible for returning a bitmask
469 * of VM_FAULT_xxx flags that give details about how the fault was handled.
471 * MM layer fills up gfp_mask for page allocations but fault handler might
472 * alter it if its implementation requires a different allocation context.
474 * pgoff should be used in favour of virtual_address, if possible.
478 struct vm_area_struct *vma; /* Target VMA */
479 gfp_t gfp_mask; /* gfp mask to be used for allocations */
480 pgoff_t pgoff; /* Logical page offset based on vma */
481 unsigned long address; /* Faulting virtual address */
483 enum fault_flag flags; /* FAULT_FLAG_xxx flags
484 * XXX: should really be 'const' */
485 pmd_t *pmd; /* Pointer to pmd entry matching
487 pud_t *pud; /* Pointer to pud entry matching
491 pte_t orig_pte; /* Value of PTE at the time of fault */
492 pmd_t orig_pmd; /* Value of PMD at the time of fault,
493 * used by PMD fault only.
497 struct page *cow_page; /* Page handler may use for COW fault */
498 struct page *page; /* ->fault handlers should return a
499 * page here, unless VM_FAULT_NOPAGE
500 * is set (which is also implied by
503 /* These three entries are valid only while holding ptl lock */
504 pte_t *pte; /* Pointer to pte entry matching
505 * the 'address'. NULL if the page
506 * table hasn't been allocated.
508 spinlock_t *ptl; /* Page table lock.
509 * Protects pte page table if 'pte'
510 * is not NULL, otherwise pmd.
512 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
513 * vm_ops->map_pages() sets up a page
514 * table from atomic context.
515 * do_fault_around() pre-allocates
516 * page table to avoid allocation from
521 /* page entry size for vm->huge_fault() */
522 enum page_entry_size {
529 * These are the virtual MM functions - opening of an area, closing and
530 * unmapping it (needed to keep files on disk up-to-date etc), pointer
531 * to the functions called when a no-page or a wp-page exception occurs.
533 struct vm_operations_struct {
534 void (*open)(struct vm_area_struct * area);
536 * @close: Called when the VMA is being removed from the MM.
537 * Context: User context. May sleep. Caller holds mmap_lock.
539 void (*close)(struct vm_area_struct * area);
540 /* Called any time before splitting to check if it's allowed */
541 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
542 int (*mremap)(struct vm_area_struct *area);
544 * Called by mprotect() to make driver-specific permission
545 * checks before mprotect() is finalised. The VMA must not
546 * be modified. Returns 0 if eprotect() can proceed.
548 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
549 unsigned long end, unsigned long newflags);
550 vm_fault_t (*fault)(struct vm_fault *vmf);
551 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
552 enum page_entry_size pe_size);
553 vm_fault_t (*map_pages)(struct vm_fault *vmf,
554 pgoff_t start_pgoff, pgoff_t end_pgoff);
555 unsigned long (*pagesize)(struct vm_area_struct * area);
557 /* notification that a previously read-only page is about to become
558 * writable, if an error is returned it will cause a SIGBUS */
559 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
561 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
562 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
564 /* called by access_process_vm when get_user_pages() fails, typically
565 * for use by special VMAs. See also generic_access_phys() for a generic
566 * implementation useful for any iomem mapping.
568 int (*access)(struct vm_area_struct *vma, unsigned long addr,
569 void *buf, int len, int write);
571 /* Called by the /proc/PID/maps code to ask the vma whether it
572 * has a special name. Returning non-NULL will also cause this
573 * vma to be dumped unconditionally. */
574 const char *(*name)(struct vm_area_struct *vma);
578 * set_policy() op must add a reference to any non-NULL @new mempolicy
579 * to hold the policy upon return. Caller should pass NULL @new to
580 * remove a policy and fall back to surrounding context--i.e. do not
581 * install a MPOL_DEFAULT policy, nor the task or system default
584 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
587 * get_policy() op must add reference [mpol_get()] to any policy at
588 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
589 * in mm/mempolicy.c will do this automatically.
590 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
591 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
592 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
593 * must return NULL--i.e., do not "fallback" to task or system default
596 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
600 * Called by vm_normal_page() for special PTEs to find the
601 * page for @addr. This is useful if the default behavior
602 * (using pte_page()) would not find the correct page.
604 struct page *(*find_special_page)(struct vm_area_struct *vma,
608 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
610 static const struct vm_operations_struct dummy_vm_ops = {};
612 memset(vma, 0, sizeof(*vma));
614 vma->vm_ops = &dummy_vm_ops;
615 INIT_LIST_HEAD(&vma->anon_vma_chain);
618 static inline void vma_set_anonymous(struct vm_area_struct *vma)
623 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
628 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
630 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
635 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
636 VM_STACK_INCOMPLETE_SETUP)
642 static inline bool vma_is_foreign(struct vm_area_struct *vma)
647 if (current->mm != vma->vm_mm)
653 static inline bool vma_is_accessible(struct vm_area_struct *vma)
655 return vma->vm_flags & VM_ACCESS_FLAGS;
660 * The vma_is_shmem is not inline because it is used only by slow
661 * paths in userfault.
663 bool vma_is_shmem(struct vm_area_struct *vma);
665 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
668 int vma_is_stack_for_current(struct vm_area_struct *vma);
670 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
671 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
676 static inline unsigned int compound_order(struct page *page)
680 return page[1].compound_order;
684 * folio_order - The allocation order of a folio.
687 * A folio is composed of 2^order pages. See get_order() for the definition
690 * Return: The order of the folio.
692 static inline unsigned int folio_order(struct folio *folio)
694 return compound_order(&folio->page);
697 #include <linux/huge_mm.h>
700 * Methods to modify the page usage count.
702 * What counts for a page usage:
703 * - cache mapping (page->mapping)
704 * - private data (page->private)
705 * - page mapped in a task's page tables, each mapping
706 * is counted separately
708 * Also, many kernel routines increase the page count before a critical
709 * routine so they can be sure the page doesn't go away from under them.
713 * Drop a ref, return true if the refcount fell to zero (the page has no users)
715 static inline int put_page_testzero(struct page *page)
717 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
718 return page_ref_dec_and_test(page);
721 static inline int folio_put_testzero(struct folio *folio)
723 return put_page_testzero(&folio->page);
727 * Try to grab a ref unless the page has a refcount of zero, return false if
729 * This can be called when MMU is off so it must not access
730 * any of the virtual mappings.
732 static inline bool get_page_unless_zero(struct page *page)
734 return page_ref_add_unless(page, 1, 0);
737 extern int page_is_ram(unsigned long pfn);
745 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
748 /* Support for virtually mapped pages */
749 struct page *vmalloc_to_page(const void *addr);
750 unsigned long vmalloc_to_pfn(const void *addr);
753 * Determine if an address is within the vmalloc range
755 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
756 * is no special casing required.
759 #ifndef is_ioremap_addr
760 #define is_ioremap_addr(x) is_vmalloc_addr(x)
764 extern bool is_vmalloc_addr(const void *x);
765 extern int is_vmalloc_or_module_addr(const void *x);
767 static inline bool is_vmalloc_addr(const void *x)
771 static inline int is_vmalloc_or_module_addr(const void *x)
777 static inline int head_compound_mapcount(struct page *head)
779 return atomic_read(compound_mapcount_ptr(head)) + 1;
783 * Mapcount of compound page as a whole, does not include mapped sub-pages.
785 * Must be called only for compound pages or any their tail sub-pages.
787 static inline int compound_mapcount(struct page *page)
789 VM_BUG_ON_PAGE(!PageCompound(page), page);
790 page = compound_head(page);
791 return head_compound_mapcount(page);
795 * The atomic page->_mapcount, starts from -1: so that transitions
796 * both from it and to it can be tracked, using atomic_inc_and_test
797 * and atomic_add_negative(-1).
799 static inline void page_mapcount_reset(struct page *page)
801 atomic_set(&(page)->_mapcount, -1);
804 int __page_mapcount(struct page *page);
807 * Mapcount of 0-order page; when compound sub-page, includes
808 * compound_mapcount().
810 * Result is undefined for pages which cannot be mapped into userspace.
811 * For example SLAB or special types of pages. See function page_has_type().
812 * They use this place in struct page differently.
814 static inline int page_mapcount(struct page *page)
816 if (unlikely(PageCompound(page)))
817 return __page_mapcount(page);
818 return atomic_read(&page->_mapcount) + 1;
821 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
822 int total_mapcount(struct page *page);
823 int page_trans_huge_mapcount(struct page *page);
825 static inline int total_mapcount(struct page *page)
827 return page_mapcount(page);
829 static inline int page_trans_huge_mapcount(struct page *page)
831 return page_mapcount(page);
835 static inline struct page *virt_to_head_page(const void *x)
837 struct page *page = virt_to_page(x);
839 return compound_head(page);
842 static inline struct folio *virt_to_folio(const void *x)
844 struct page *page = virt_to_page(x);
846 return page_folio(page);
849 void __put_page(struct page *page);
851 void put_pages_list(struct list_head *pages);
853 void split_page(struct page *page, unsigned int order);
854 void folio_copy(struct folio *dst, struct folio *src);
856 unsigned long nr_free_buffer_pages(void);
859 * Compound pages have a destructor function. Provide a
860 * prototype for that function and accessor functions.
861 * These are _only_ valid on the head of a compound page.
863 typedef void compound_page_dtor(struct page *);
865 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
866 enum compound_dtor_id {
869 #ifdef CONFIG_HUGETLB_PAGE
872 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
877 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
879 static inline void set_compound_page_dtor(struct page *page,
880 enum compound_dtor_id compound_dtor)
882 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
883 page[1].compound_dtor = compound_dtor;
886 static inline void destroy_compound_page(struct page *page)
888 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
889 compound_page_dtors[page[1].compound_dtor](page);
892 static inline bool hpage_pincount_available(struct page *page)
895 * Can the page->hpage_pinned_refcount field be used? That field is in
896 * the 3rd page of the compound page, so the smallest (2-page) compound
897 * pages cannot support it.
899 page = compound_head(page);
900 return PageCompound(page) && compound_order(page) > 1;
903 static inline int head_compound_pincount(struct page *head)
905 return atomic_read(compound_pincount_ptr(head));
908 static inline int compound_pincount(struct page *page)
910 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
911 page = compound_head(page);
912 return head_compound_pincount(page);
915 static inline void set_compound_order(struct page *page, unsigned int order)
917 page[1].compound_order = order;
918 page[1].compound_nr = 1U << order;
921 /* Returns the number of pages in this potentially compound page. */
922 static inline unsigned long compound_nr(struct page *page)
926 return page[1].compound_nr;
929 /* Returns the number of bytes in this potentially compound page. */
930 static inline unsigned long page_size(struct page *page)
932 return PAGE_SIZE << compound_order(page);
935 /* Returns the number of bits needed for the number of bytes in a page */
936 static inline unsigned int page_shift(struct page *page)
938 return PAGE_SHIFT + compound_order(page);
941 void free_compound_page(struct page *page);
945 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
946 * servicing faults for write access. In the normal case, do always want
947 * pte_mkwrite. But get_user_pages can cause write faults for mappings
948 * that do not have writing enabled, when used by access_process_vm.
950 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
952 if (likely(vma->vm_flags & VM_WRITE))
953 pte = pte_mkwrite(pte);
957 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
958 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
960 vm_fault_t finish_fault(struct vm_fault *vmf);
961 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
965 * Multiple processes may "see" the same page. E.g. for untouched
966 * mappings of /dev/null, all processes see the same page full of
967 * zeroes, and text pages of executables and shared libraries have
968 * only one copy in memory, at most, normally.
970 * For the non-reserved pages, page_count(page) denotes a reference count.
971 * page_count() == 0 means the page is free. page->lru is then used for
972 * freelist management in the buddy allocator.
973 * page_count() > 0 means the page has been allocated.
975 * Pages are allocated by the slab allocator in order to provide memory
976 * to kmalloc and kmem_cache_alloc. In this case, the management of the
977 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
978 * unless a particular usage is carefully commented. (the responsibility of
979 * freeing the kmalloc memory is the caller's, of course).
981 * A page may be used by anyone else who does a __get_free_page().
982 * In this case, page_count still tracks the references, and should only
983 * be used through the normal accessor functions. The top bits of page->flags
984 * and page->virtual store page management information, but all other fields
985 * are unused and could be used privately, carefully. The management of this
986 * page is the responsibility of the one who allocated it, and those who have
987 * subsequently been given references to it.
989 * The other pages (we may call them "pagecache pages") are completely
990 * managed by the Linux memory manager: I/O, buffers, swapping etc.
991 * The following discussion applies only to them.
993 * A pagecache page contains an opaque `private' member, which belongs to the
994 * page's address_space. Usually, this is the address of a circular list of
995 * the page's disk buffers. PG_private must be set to tell the VM to call
996 * into the filesystem to release these pages.
998 * A page may belong to an inode's memory mapping. In this case, page->mapping
999 * is the pointer to the inode, and page->index is the file offset of the page,
1000 * in units of PAGE_SIZE.
1002 * If pagecache pages are not associated with an inode, they are said to be
1003 * anonymous pages. These may become associated with the swapcache, and in that
1004 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1006 * In either case (swapcache or inode backed), the pagecache itself holds one
1007 * reference to the page. Setting PG_private should also increment the
1008 * refcount. The each user mapping also has a reference to the page.
1010 * The pagecache pages are stored in a per-mapping radix tree, which is
1011 * rooted at mapping->i_pages, and indexed by offset.
1012 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1013 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1015 * All pagecache pages may be subject to I/O:
1016 * - inode pages may need to be read from disk,
1017 * - inode pages which have been modified and are MAP_SHARED may need
1018 * to be written back to the inode on disk,
1019 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1020 * modified may need to be swapped out to swap space and (later) to be read
1025 * The zone field is never updated after free_area_init_core()
1026 * sets it, so none of the operations on it need to be atomic.
1029 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1030 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1031 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1032 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1033 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1034 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1037 * Define the bit shifts to access each section. For non-existent
1038 * sections we define the shift as 0; that plus a 0 mask ensures
1039 * the compiler will optimise away reference to them.
1041 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1042 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1043 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1044 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1045 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1047 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1048 #ifdef NODE_NOT_IN_PAGE_FLAGS
1049 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1050 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1051 SECTIONS_PGOFF : ZONES_PGOFF)
1053 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1054 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1055 NODES_PGOFF : ZONES_PGOFF)
1058 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1060 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1061 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1062 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1063 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1064 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1065 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1067 static inline enum zone_type page_zonenum(const struct page *page)
1069 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1070 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1073 static inline enum zone_type folio_zonenum(const struct folio *folio)
1075 return page_zonenum(&folio->page);
1078 #ifdef CONFIG_ZONE_DEVICE
1079 static inline bool is_zone_device_page(const struct page *page)
1081 return page_zonenum(page) == ZONE_DEVICE;
1083 extern void memmap_init_zone_device(struct zone *, unsigned long,
1084 unsigned long, struct dev_pagemap *);
1086 static inline bool is_zone_device_page(const struct page *page)
1092 static inline bool is_zone_movable_page(const struct page *page)
1094 return page_zonenum(page) == ZONE_MOVABLE;
1097 #ifdef CONFIG_DEV_PAGEMAP_OPS
1098 void free_devmap_managed_page(struct page *page);
1099 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1101 static inline bool page_is_devmap_managed(struct page *page)
1103 if (!static_branch_unlikely(&devmap_managed_key))
1105 if (!is_zone_device_page(page))
1107 switch (page->pgmap->type) {
1108 case MEMORY_DEVICE_PRIVATE:
1109 case MEMORY_DEVICE_FS_DAX:
1117 void put_devmap_managed_page(struct page *page);
1119 #else /* CONFIG_DEV_PAGEMAP_OPS */
1120 static inline bool page_is_devmap_managed(struct page *page)
1125 static inline void put_devmap_managed_page(struct page *page)
1128 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1130 static inline bool is_device_private_page(const struct page *page)
1132 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1133 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1134 is_zone_device_page(page) &&
1135 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1138 static inline bool is_pci_p2pdma_page(const struct page *page)
1140 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1141 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1142 is_zone_device_page(page) &&
1143 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1146 /* 127: arbitrary random number, small enough to assemble well */
1147 #define folio_ref_zero_or_close_to_overflow(folio) \
1148 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1151 * folio_get - Increment the reference count on a folio.
1152 * @folio: The folio.
1154 * Context: May be called in any context, as long as you know that
1155 * you have a refcount on the folio. If you do not already have one,
1156 * folio_try_get() may be the right interface for you to use.
1158 static inline void folio_get(struct folio *folio)
1160 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1161 folio_ref_inc(folio);
1164 static inline void get_page(struct page *page)
1166 folio_get(page_folio(page));
1169 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1170 struct page *try_grab_compound_head(struct page *page, int refs,
1171 unsigned int flags);
1174 static inline __must_check bool try_get_page(struct page *page)
1176 page = compound_head(page);
1177 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1184 * folio_put - Decrement the reference count on a folio.
1185 * @folio: The folio.
1187 * If the folio's reference count reaches zero, the memory will be
1188 * released back to the page allocator and may be used by another
1189 * allocation immediately. Do not access the memory or the struct folio
1190 * after calling folio_put() unless you can be sure that it wasn't the
1193 * Context: May be called in process or interrupt context, but not in NMI
1194 * context. May be called while holding a spinlock.
1196 static inline void folio_put(struct folio *folio)
1198 if (folio_put_testzero(folio))
1199 __put_page(&folio->page);
1202 static inline void put_page(struct page *page)
1204 struct folio *folio = page_folio(page);
1207 * For devmap managed pages we need to catch refcount transition from
1208 * 2 to 1, when refcount reach one it means the page is free and we
1209 * need to inform the device driver through callback. See
1210 * include/linux/memremap.h and HMM for details.
1212 if (page_is_devmap_managed(&folio->page)) {
1213 put_devmap_managed_page(&folio->page);
1221 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1222 * the page's refcount so that two separate items are tracked: the original page
1223 * reference count, and also a new count of how many pin_user_pages() calls were
1224 * made against the page. ("gup-pinned" is another term for the latter).
1226 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1227 * distinct from normal pages. As such, the unpin_user_page() call (and its
1228 * variants) must be used in order to release gup-pinned pages.
1232 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1233 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1234 * simpler, due to the fact that adding an even power of two to the page
1235 * refcount has the effect of using only the upper N bits, for the code that
1236 * counts up using the bias value. This means that the lower bits are left for
1237 * the exclusive use of the original code that increments and decrements by one
1238 * (or at least, by much smaller values than the bias value).
1240 * Of course, once the lower bits overflow into the upper bits (and this is
1241 * OK, because subtraction recovers the original values), then visual inspection
1242 * no longer suffices to directly view the separate counts. However, for normal
1243 * applications that don't have huge page reference counts, this won't be an
1246 * Locking: the lockless algorithm described in page_cache_get_speculative()
1247 * and page_cache_gup_pin_speculative() provides safe operation for
1248 * get_user_pages and page_mkclean and other calls that race to set up page
1251 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1253 void unpin_user_page(struct page *page);
1254 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1256 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1258 void unpin_user_pages(struct page **pages, unsigned long npages);
1261 * page_maybe_dma_pinned - Report if a page is pinned for DMA.
1264 * This function checks if a page has been pinned via a call to
1265 * a function in the pin_user_pages() family.
1267 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1268 * because it means "definitely not pinned for DMA", but true means "probably
1269 * pinned for DMA, but possibly a false positive due to having at least
1270 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1272 * False positives are OK, because: a) it's unlikely for a page to get that many
1273 * refcounts, and b) all the callers of this routine are expected to be able to
1274 * deal gracefully with a false positive.
1276 * For huge pages, the result will be exactly correct. That's because we have
1277 * more tracking data available: the 3rd struct page in the compound page is
1278 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1281 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1283 * Return: True, if it is likely that the page has been "dma-pinned".
1284 * False, if the page is definitely not dma-pinned.
1286 static inline bool page_maybe_dma_pinned(struct page *page)
1288 if (hpage_pincount_available(page))
1289 return compound_pincount(page) > 0;
1292 * page_ref_count() is signed. If that refcount overflows, then
1293 * page_ref_count() returns a negative value, and callers will avoid
1294 * further incrementing the refcount.
1296 * Here, for that overflow case, use the signed bit to count a little
1297 * bit higher via unsigned math, and thus still get an accurate result.
1299 return ((unsigned int)page_ref_count(compound_head(page))) >=
1300 GUP_PIN_COUNTING_BIAS;
1303 static inline bool is_cow_mapping(vm_flags_t flags)
1305 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1309 * This should most likely only be called during fork() to see whether we
1310 * should break the cow immediately for a page on the src mm.
1312 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1315 if (!is_cow_mapping(vma->vm_flags))
1318 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1321 return page_maybe_dma_pinned(page);
1324 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1325 #define SECTION_IN_PAGE_FLAGS
1329 * The identification function is mainly used by the buddy allocator for
1330 * determining if two pages could be buddies. We are not really identifying
1331 * the zone since we could be using the section number id if we do not have
1332 * node id available in page flags.
1333 * We only guarantee that it will return the same value for two combinable
1336 static inline int page_zone_id(struct page *page)
1338 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1341 #ifdef NODE_NOT_IN_PAGE_FLAGS
1342 extern int page_to_nid(const struct page *page);
1344 static inline int page_to_nid(const struct page *page)
1346 struct page *p = (struct page *)page;
1348 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1352 static inline int folio_nid(const struct folio *folio)
1354 return page_to_nid(&folio->page);
1357 #ifdef CONFIG_NUMA_BALANCING
1358 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1360 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1363 static inline int cpupid_to_pid(int cpupid)
1365 return cpupid & LAST__PID_MASK;
1368 static inline int cpupid_to_cpu(int cpupid)
1370 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1373 static inline int cpupid_to_nid(int cpupid)
1375 return cpu_to_node(cpupid_to_cpu(cpupid));
1378 static inline bool cpupid_pid_unset(int cpupid)
1380 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1383 static inline bool cpupid_cpu_unset(int cpupid)
1385 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1388 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1390 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1393 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1394 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1395 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1397 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1400 static inline int page_cpupid_last(struct page *page)
1402 return page->_last_cpupid;
1404 static inline void page_cpupid_reset_last(struct page *page)
1406 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1409 static inline int page_cpupid_last(struct page *page)
1411 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1414 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1416 static inline void page_cpupid_reset_last(struct page *page)
1418 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1420 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1421 #else /* !CONFIG_NUMA_BALANCING */
1422 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1424 return page_to_nid(page); /* XXX */
1427 static inline int page_cpupid_last(struct page *page)
1429 return page_to_nid(page); /* XXX */
1432 static inline int cpupid_to_nid(int cpupid)
1437 static inline int cpupid_to_pid(int cpupid)
1442 static inline int cpupid_to_cpu(int cpupid)
1447 static inline int cpu_pid_to_cpupid(int nid, int pid)
1452 static inline bool cpupid_pid_unset(int cpupid)
1457 static inline void page_cpupid_reset_last(struct page *page)
1461 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1465 #endif /* CONFIG_NUMA_BALANCING */
1467 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1470 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1471 * setting tags for all pages to native kernel tag value 0xff, as the default
1472 * value 0x00 maps to 0xff.
1475 static inline u8 page_kasan_tag(const struct page *page)
1479 if (kasan_enabled()) {
1480 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1487 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1489 if (kasan_enabled()) {
1491 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1492 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1496 static inline void page_kasan_tag_reset(struct page *page)
1498 if (kasan_enabled())
1499 page_kasan_tag_set(page, 0xff);
1502 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1504 static inline u8 page_kasan_tag(const struct page *page)
1509 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1510 static inline void page_kasan_tag_reset(struct page *page) { }
1512 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1514 static inline struct zone *page_zone(const struct page *page)
1516 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1519 static inline pg_data_t *page_pgdat(const struct page *page)
1521 return NODE_DATA(page_to_nid(page));
1524 static inline struct zone *folio_zone(const struct folio *folio)
1526 return page_zone(&folio->page);
1529 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1531 return page_pgdat(&folio->page);
1534 #ifdef SECTION_IN_PAGE_FLAGS
1535 static inline void set_page_section(struct page *page, unsigned long section)
1537 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1538 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1541 static inline unsigned long page_to_section(const struct page *page)
1543 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1548 * folio_pfn - Return the Page Frame Number of a folio.
1549 * @folio: The folio.
1551 * A folio may contain multiple pages. The pages have consecutive
1552 * Page Frame Numbers.
1554 * Return: The Page Frame Number of the first page in the folio.
1556 static inline unsigned long folio_pfn(struct folio *folio)
1558 return page_to_pfn(&folio->page);
1561 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1562 #ifdef CONFIG_MIGRATION
1563 static inline bool is_pinnable_page(struct page *page)
1565 return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) ||
1566 is_zero_pfn(page_to_pfn(page));
1569 static inline bool is_pinnable_page(struct page *page)
1575 static inline void set_page_zone(struct page *page, enum zone_type zone)
1577 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1578 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1581 static inline void set_page_node(struct page *page, unsigned long node)
1583 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1584 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1587 static inline void set_page_links(struct page *page, enum zone_type zone,
1588 unsigned long node, unsigned long pfn)
1590 set_page_zone(page, zone);
1591 set_page_node(page, node);
1592 #ifdef SECTION_IN_PAGE_FLAGS
1593 set_page_section(page, pfn_to_section_nr(pfn));
1598 * folio_nr_pages - The number of pages in the folio.
1599 * @folio: The folio.
1601 * Return: A positive power of two.
1603 static inline long folio_nr_pages(struct folio *folio)
1605 return compound_nr(&folio->page);
1609 * folio_next - Move to the next physical folio.
1610 * @folio: The folio we're currently operating on.
1612 * If you have physically contiguous memory which may span more than
1613 * one folio (eg a &struct bio_vec), use this function to move from one
1614 * folio to the next. Do not use it if the memory is only virtually
1615 * contiguous as the folios are almost certainly not adjacent to each
1616 * other. This is the folio equivalent to writing ``page++``.
1618 * Context: We assume that the folios are refcounted and/or locked at a
1619 * higher level and do not adjust the reference counts.
1620 * Return: The next struct folio.
1622 static inline struct folio *folio_next(struct folio *folio)
1624 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1628 * folio_shift - The size of the memory described by this folio.
1629 * @folio: The folio.
1631 * A folio represents a number of bytes which is a power-of-two in size.
1632 * This function tells you which power-of-two the folio is. See also
1633 * folio_size() and folio_order().
1635 * Context: The caller should have a reference on the folio to prevent
1636 * it from being split. It is not necessary for the folio to be locked.
1637 * Return: The base-2 logarithm of the size of this folio.
1639 static inline unsigned int folio_shift(struct folio *folio)
1641 return PAGE_SHIFT + folio_order(folio);
1645 * folio_size - The number of bytes in a folio.
1646 * @folio: The folio.
1648 * Context: The caller should have a reference on the folio to prevent
1649 * it from being split. It is not necessary for the folio to be locked.
1650 * Return: The number of bytes in this folio.
1652 static inline size_t folio_size(struct folio *folio)
1654 return PAGE_SIZE << folio_order(folio);
1657 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1658 static inline int arch_make_page_accessible(struct page *page)
1664 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1665 static inline int arch_make_folio_accessible(struct folio *folio)
1668 long i, nr = folio_nr_pages(folio);
1670 for (i = 0; i < nr; i++) {
1671 ret = arch_make_page_accessible(folio_page(folio, i));
1681 * Some inline functions in vmstat.h depend on page_zone()
1683 #include <linux/vmstat.h>
1685 static __always_inline void *lowmem_page_address(const struct page *page)
1687 return page_to_virt(page);
1690 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1691 #define HASHED_PAGE_VIRTUAL
1694 #if defined(WANT_PAGE_VIRTUAL)
1695 static inline void *page_address(const struct page *page)
1697 return page->virtual;
1699 static inline void set_page_address(struct page *page, void *address)
1701 page->virtual = address;
1703 #define page_address_init() do { } while(0)
1706 #if defined(HASHED_PAGE_VIRTUAL)
1707 void *page_address(const struct page *page);
1708 void set_page_address(struct page *page, void *virtual);
1709 void page_address_init(void);
1712 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1713 #define page_address(page) lowmem_page_address(page)
1714 #define set_page_address(page, address) do { } while(0)
1715 #define page_address_init() do { } while(0)
1718 static inline void *folio_address(const struct folio *folio)
1720 return page_address(&folio->page);
1723 extern void *page_rmapping(struct page *page);
1724 extern struct anon_vma *page_anon_vma(struct page *page);
1725 extern pgoff_t __page_file_index(struct page *page);
1728 * Return the pagecache index of the passed page. Regular pagecache pages
1729 * use ->index whereas swapcache pages use swp_offset(->private)
1731 static inline pgoff_t page_index(struct page *page)
1733 if (unlikely(PageSwapCache(page)))
1734 return __page_file_index(page);
1738 bool page_mapped(struct page *page);
1739 bool folio_mapped(struct folio *folio);
1742 * Return true only if the page has been allocated with
1743 * ALLOC_NO_WATERMARKS and the low watermark was not
1744 * met implying that the system is under some pressure.
1746 static inline bool page_is_pfmemalloc(const struct page *page)
1749 * lru.next has bit 1 set if the page is allocated from the
1750 * pfmemalloc reserves. Callers may simply overwrite it if
1751 * they do not need to preserve that information.
1753 return (uintptr_t)page->lru.next & BIT(1);
1757 * Only to be called by the page allocator on a freshly allocated
1760 static inline void set_page_pfmemalloc(struct page *page)
1762 page->lru.next = (void *)BIT(1);
1765 static inline void clear_page_pfmemalloc(struct page *page)
1767 page->lru.next = NULL;
1771 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1773 extern void pagefault_out_of_memory(void);
1775 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1776 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1777 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1780 * Flags passed to show_mem() and show_free_areas() to suppress output in
1783 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1785 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1788 extern bool can_do_mlock(void);
1790 static inline bool can_do_mlock(void) { return false; }
1792 extern int user_shm_lock(size_t, struct ucounts *);
1793 extern void user_shm_unlock(size_t, struct ucounts *);
1795 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1797 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1800 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1801 unsigned long size);
1802 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1803 unsigned long size);
1804 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1805 unsigned long start, unsigned long end);
1807 struct mmu_notifier_range;
1809 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1810 unsigned long end, unsigned long floor, unsigned long ceiling);
1812 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1813 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1814 struct mmu_notifier_range *range, pte_t **ptepp,
1815 pmd_t **pmdpp, spinlock_t **ptlp);
1816 int follow_pte(struct mm_struct *mm, unsigned long address,
1817 pte_t **ptepp, spinlock_t **ptlp);
1818 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1819 unsigned long *pfn);
1820 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1821 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1822 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1823 void *buf, int len, int write);
1825 extern void truncate_pagecache(struct inode *inode, loff_t new);
1826 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1827 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1828 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1829 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1830 int invalidate_inode_page(struct page *page);
1833 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1834 unsigned long address, unsigned int flags,
1835 struct pt_regs *regs);
1836 extern int fixup_user_fault(struct mm_struct *mm,
1837 unsigned long address, unsigned int fault_flags,
1839 void unmap_mapping_pages(struct address_space *mapping,
1840 pgoff_t start, pgoff_t nr, bool even_cows);
1841 void unmap_mapping_range(struct address_space *mapping,
1842 loff_t const holebegin, loff_t const holelen, int even_cows);
1844 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1845 unsigned long address, unsigned int flags,
1846 struct pt_regs *regs)
1848 /* should never happen if there's no MMU */
1850 return VM_FAULT_SIGBUS;
1852 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1853 unsigned int fault_flags, bool *unlocked)
1855 /* should never happen if there's no MMU */
1859 static inline void unmap_mapping_pages(struct address_space *mapping,
1860 pgoff_t start, pgoff_t nr, bool even_cows) { }
1861 static inline void unmap_mapping_range(struct address_space *mapping,
1862 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1865 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1866 loff_t const holebegin, loff_t const holelen)
1868 unmap_mapping_range(mapping, holebegin, holelen, 0);
1871 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1872 void *buf, int len, unsigned int gup_flags);
1873 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1874 void *buf, int len, unsigned int gup_flags);
1875 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1876 void *buf, int len, unsigned int gup_flags);
1878 long get_user_pages_remote(struct mm_struct *mm,
1879 unsigned long start, unsigned long nr_pages,
1880 unsigned int gup_flags, struct page **pages,
1881 struct vm_area_struct **vmas, int *locked);
1882 long pin_user_pages_remote(struct mm_struct *mm,
1883 unsigned long start, unsigned long nr_pages,
1884 unsigned int gup_flags, struct page **pages,
1885 struct vm_area_struct **vmas, int *locked);
1886 long get_user_pages(unsigned long start, unsigned long nr_pages,
1887 unsigned int gup_flags, struct page **pages,
1888 struct vm_area_struct **vmas);
1889 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1890 unsigned int gup_flags, struct page **pages,
1891 struct vm_area_struct **vmas);
1892 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1893 unsigned int gup_flags, struct page **pages, int *locked);
1894 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1895 unsigned int gup_flags, struct page **pages, int *locked);
1896 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1897 struct page **pages, unsigned int gup_flags);
1898 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1899 struct page **pages, unsigned int gup_flags);
1901 int get_user_pages_fast(unsigned long start, int nr_pages,
1902 unsigned int gup_flags, struct page **pages);
1903 int pin_user_pages_fast(unsigned long start, int nr_pages,
1904 unsigned int gup_flags, struct page **pages);
1906 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1907 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1908 struct task_struct *task, bool bypass_rlim);
1911 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1912 struct page **pages);
1913 struct page *get_dump_page(unsigned long addr);
1915 extern void do_invalidatepage(struct page *page, unsigned int offset,
1916 unsigned int length);
1918 bool folio_mark_dirty(struct folio *folio);
1919 bool set_page_dirty(struct page *page);
1920 int set_page_dirty_lock(struct page *page);
1922 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1924 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1925 unsigned long old_addr, struct vm_area_struct *new_vma,
1926 unsigned long new_addr, unsigned long len,
1927 bool need_rmap_locks);
1930 * Flags used by change_protection(). For now we make it a bitmap so
1931 * that we can pass in multiple flags just like parameters. However
1932 * for now all the callers are only use one of the flags at the same
1935 /* Whether we should allow dirty bit accounting */
1936 #define MM_CP_DIRTY_ACCT (1UL << 0)
1937 /* Whether this protection change is for NUMA hints */
1938 #define MM_CP_PROT_NUMA (1UL << 1)
1939 /* Whether this change is for write protecting */
1940 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1941 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1942 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1943 MM_CP_UFFD_WP_RESOLVE)
1945 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1946 unsigned long end, pgprot_t newprot,
1947 unsigned long cp_flags);
1948 extern int mprotect_fixup(struct vm_area_struct *vma,
1949 struct vm_area_struct **pprev, unsigned long start,
1950 unsigned long end, unsigned long newflags);
1953 * doesn't attempt to fault and will return short.
1955 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1956 unsigned int gup_flags, struct page **pages);
1957 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1958 unsigned int gup_flags, struct page **pages);
1960 static inline bool get_user_page_fast_only(unsigned long addr,
1961 unsigned int gup_flags, struct page **pagep)
1963 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1966 * per-process(per-mm_struct) statistics.
1968 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1970 long val = atomic_long_read(&mm->rss_stat.count[member]);
1972 #ifdef SPLIT_RSS_COUNTING
1974 * counter is updated in asynchronous manner and may go to minus.
1975 * But it's never be expected number for users.
1980 return (unsigned long)val;
1983 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1985 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1987 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1989 mm_trace_rss_stat(mm, member, count);
1992 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1994 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1996 mm_trace_rss_stat(mm, member, count);
1999 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2001 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
2003 mm_trace_rss_stat(mm, member, count);
2006 /* Optimized variant when page is already known not to be PageAnon */
2007 static inline int mm_counter_file(struct page *page)
2009 if (PageSwapBacked(page))
2010 return MM_SHMEMPAGES;
2011 return MM_FILEPAGES;
2014 static inline int mm_counter(struct page *page)
2017 return MM_ANONPAGES;
2018 return mm_counter_file(page);
2021 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2023 return get_mm_counter(mm, MM_FILEPAGES) +
2024 get_mm_counter(mm, MM_ANONPAGES) +
2025 get_mm_counter(mm, MM_SHMEMPAGES);
2028 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2030 return max(mm->hiwater_rss, get_mm_rss(mm));
2033 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2035 return max(mm->hiwater_vm, mm->total_vm);
2038 static inline void update_hiwater_rss(struct mm_struct *mm)
2040 unsigned long _rss = get_mm_rss(mm);
2042 if ((mm)->hiwater_rss < _rss)
2043 (mm)->hiwater_rss = _rss;
2046 static inline void update_hiwater_vm(struct mm_struct *mm)
2048 if (mm->hiwater_vm < mm->total_vm)
2049 mm->hiwater_vm = mm->total_vm;
2052 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2054 mm->hiwater_rss = get_mm_rss(mm);
2057 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2058 struct mm_struct *mm)
2060 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2062 if (*maxrss < hiwater_rss)
2063 *maxrss = hiwater_rss;
2066 #if defined(SPLIT_RSS_COUNTING)
2067 void sync_mm_rss(struct mm_struct *mm);
2069 static inline void sync_mm_rss(struct mm_struct *mm)
2074 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2075 static inline int pte_special(pte_t pte)
2080 static inline pte_t pte_mkspecial(pte_t pte)
2086 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2087 static inline int pte_devmap(pte_t pte)
2093 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2095 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2097 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2101 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2105 #ifdef __PAGETABLE_P4D_FOLDED
2106 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2107 unsigned long address)
2112 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2115 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2116 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2117 unsigned long address)
2121 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2122 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2125 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2127 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2129 if (mm_pud_folded(mm))
2131 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2134 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2136 if (mm_pud_folded(mm))
2138 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2142 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2143 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2144 unsigned long address)
2149 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2150 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2153 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2155 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2157 if (mm_pmd_folded(mm))
2159 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2162 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2164 if (mm_pmd_folded(mm))
2166 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2171 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2173 atomic_long_set(&mm->pgtables_bytes, 0);
2176 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2178 return atomic_long_read(&mm->pgtables_bytes);
2181 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2183 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2186 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2188 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2192 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2193 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2198 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2199 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2202 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2203 int __pte_alloc_kernel(pmd_t *pmd);
2205 #if defined(CONFIG_MMU)
2207 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2208 unsigned long address)
2210 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2211 NULL : p4d_offset(pgd, address);
2214 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2215 unsigned long address)
2217 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2218 NULL : pud_offset(p4d, address);
2221 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2223 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2224 NULL: pmd_offset(pud, address);
2226 #endif /* CONFIG_MMU */
2228 #if USE_SPLIT_PTE_PTLOCKS
2229 #if ALLOC_SPLIT_PTLOCKS
2230 void __init ptlock_cache_init(void);
2231 extern bool ptlock_alloc(struct page *page);
2232 extern void ptlock_free(struct page *page);
2234 static inline spinlock_t *ptlock_ptr(struct page *page)
2238 #else /* ALLOC_SPLIT_PTLOCKS */
2239 static inline void ptlock_cache_init(void)
2243 static inline bool ptlock_alloc(struct page *page)
2248 static inline void ptlock_free(struct page *page)
2252 static inline spinlock_t *ptlock_ptr(struct page *page)
2256 #endif /* ALLOC_SPLIT_PTLOCKS */
2258 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2260 return ptlock_ptr(pmd_page(*pmd));
2263 static inline bool ptlock_init(struct page *page)
2266 * prep_new_page() initialize page->private (and therefore page->ptl)
2267 * with 0. Make sure nobody took it in use in between.
2269 * It can happen if arch try to use slab for page table allocation:
2270 * slab code uses page->slab_cache, which share storage with page->ptl.
2272 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2273 if (!ptlock_alloc(page))
2275 spin_lock_init(ptlock_ptr(page));
2279 #else /* !USE_SPLIT_PTE_PTLOCKS */
2281 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2283 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2285 return &mm->page_table_lock;
2287 static inline void ptlock_cache_init(void) {}
2288 static inline bool ptlock_init(struct page *page) { return true; }
2289 static inline void ptlock_free(struct page *page) {}
2290 #endif /* USE_SPLIT_PTE_PTLOCKS */
2292 static inline void pgtable_init(void)
2294 ptlock_cache_init();
2295 pgtable_cache_init();
2298 static inline bool pgtable_pte_page_ctor(struct page *page)
2300 if (!ptlock_init(page))
2302 __SetPageTable(page);
2303 inc_lruvec_page_state(page, NR_PAGETABLE);
2307 static inline void pgtable_pte_page_dtor(struct page *page)
2310 __ClearPageTable(page);
2311 dec_lruvec_page_state(page, NR_PAGETABLE);
2314 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2316 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2317 pte_t *__pte = pte_offset_map(pmd, address); \
2323 #define pte_unmap_unlock(pte, ptl) do { \
2328 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2330 #define pte_alloc_map(mm, pmd, address) \
2331 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2333 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2334 (pte_alloc(mm, pmd) ? \
2335 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2337 #define pte_alloc_kernel(pmd, address) \
2338 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2339 NULL: pte_offset_kernel(pmd, address))
2341 #if USE_SPLIT_PMD_PTLOCKS
2343 static struct page *pmd_to_page(pmd_t *pmd)
2345 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2346 return virt_to_page((void *)((unsigned long) pmd & mask));
2349 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2351 return ptlock_ptr(pmd_to_page(pmd));
2354 static inline bool pmd_ptlock_init(struct page *page)
2356 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2357 page->pmd_huge_pte = NULL;
2359 return ptlock_init(page);
2362 static inline void pmd_ptlock_free(struct page *page)
2364 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2365 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2370 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2374 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2376 return &mm->page_table_lock;
2379 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2380 static inline void pmd_ptlock_free(struct page *page) {}
2382 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2386 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2388 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2393 static inline bool pgtable_pmd_page_ctor(struct page *page)
2395 if (!pmd_ptlock_init(page))
2397 __SetPageTable(page);
2398 inc_lruvec_page_state(page, NR_PAGETABLE);
2402 static inline void pgtable_pmd_page_dtor(struct page *page)
2404 pmd_ptlock_free(page);
2405 __ClearPageTable(page);
2406 dec_lruvec_page_state(page, NR_PAGETABLE);
2410 * No scalability reason to split PUD locks yet, but follow the same pattern
2411 * as the PMD locks to make it easier if we decide to. The VM should not be
2412 * considered ready to switch to split PUD locks yet; there may be places
2413 * which need to be converted from page_table_lock.
2415 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2417 return &mm->page_table_lock;
2420 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2422 spinlock_t *ptl = pud_lockptr(mm, pud);
2428 extern void __init pagecache_init(void);
2429 extern void __init free_area_init_memoryless_node(int nid);
2430 extern void free_initmem(void);
2433 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2434 * into the buddy system. The freed pages will be poisoned with pattern
2435 * "poison" if it's within range [0, UCHAR_MAX].
2436 * Return pages freed into the buddy system.
2438 extern unsigned long free_reserved_area(void *start, void *end,
2439 int poison, const char *s);
2441 extern void adjust_managed_page_count(struct page *page, long count);
2442 extern void mem_init_print_info(void);
2444 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2446 /* Free the reserved page into the buddy system, so it gets managed. */
2447 static inline void free_reserved_page(struct page *page)
2449 ClearPageReserved(page);
2450 init_page_count(page);
2452 adjust_managed_page_count(page, 1);
2454 #define free_highmem_page(page) free_reserved_page(page)
2456 static inline void mark_page_reserved(struct page *page)
2458 SetPageReserved(page);
2459 adjust_managed_page_count(page, -1);
2463 * Default method to free all the __init memory into the buddy system.
2464 * The freed pages will be poisoned with pattern "poison" if it's within
2465 * range [0, UCHAR_MAX].
2466 * Return pages freed into the buddy system.
2468 static inline unsigned long free_initmem_default(int poison)
2470 extern char __init_begin[], __init_end[];
2472 return free_reserved_area(&__init_begin, &__init_end,
2473 poison, "unused kernel image (initmem)");
2476 static inline unsigned long get_num_physpages(void)
2479 unsigned long phys_pages = 0;
2481 for_each_online_node(nid)
2482 phys_pages += node_present_pages(nid);
2488 * Using memblock node mappings, an architecture may initialise its
2489 * zones, allocate the backing mem_map and account for memory holes in an
2490 * architecture independent manner.
2492 * An architecture is expected to register range of page frames backed by
2493 * physical memory with memblock_add[_node]() before calling
2494 * free_area_init() passing in the PFN each zone ends at. At a basic
2495 * usage, an architecture is expected to do something like
2497 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2499 * for_each_valid_physical_page_range()
2500 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2501 * free_area_init(max_zone_pfns);
2503 void free_area_init(unsigned long *max_zone_pfn);
2504 unsigned long node_map_pfn_alignment(void);
2505 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2506 unsigned long end_pfn);
2507 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2508 unsigned long end_pfn);
2509 extern void get_pfn_range_for_nid(unsigned int nid,
2510 unsigned long *start_pfn, unsigned long *end_pfn);
2511 extern unsigned long find_min_pfn_with_active_regions(void);
2514 static inline int early_pfn_to_nid(unsigned long pfn)
2519 /* please see mm/page_alloc.c */
2520 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2523 extern void set_dma_reserve(unsigned long new_dma_reserve);
2524 extern void memmap_init_range(unsigned long, int, unsigned long,
2525 unsigned long, unsigned long, enum meminit_context,
2526 struct vmem_altmap *, int migratetype);
2527 extern void setup_per_zone_wmarks(void);
2528 extern void calculate_min_free_kbytes(void);
2529 extern int __meminit init_per_zone_wmark_min(void);
2530 extern void mem_init(void);
2531 extern void __init mmap_init(void);
2532 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2533 extern long si_mem_available(void);
2534 extern void si_meminfo(struct sysinfo * val);
2535 extern void si_meminfo_node(struct sysinfo *val, int nid);
2536 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2537 extern unsigned long arch_reserved_kernel_pages(void);
2540 extern __printf(3, 4)
2541 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2543 extern void setup_per_cpu_pageset(void);
2546 extern int min_free_kbytes;
2547 extern int watermark_boost_factor;
2548 extern int watermark_scale_factor;
2549 extern bool arch_has_descending_max_zone_pfns(void);
2552 extern atomic_long_t mmap_pages_allocated;
2553 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2555 /* interval_tree.c */
2556 void vma_interval_tree_insert(struct vm_area_struct *node,
2557 struct rb_root_cached *root);
2558 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2559 struct vm_area_struct *prev,
2560 struct rb_root_cached *root);
2561 void vma_interval_tree_remove(struct vm_area_struct *node,
2562 struct rb_root_cached *root);
2563 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2564 unsigned long start, unsigned long last);
2565 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2566 unsigned long start, unsigned long last);
2568 #define vma_interval_tree_foreach(vma, root, start, last) \
2569 for (vma = vma_interval_tree_iter_first(root, start, last); \
2570 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2572 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2573 struct rb_root_cached *root);
2574 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2575 struct rb_root_cached *root);
2576 struct anon_vma_chain *
2577 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2578 unsigned long start, unsigned long last);
2579 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2580 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2581 #ifdef CONFIG_DEBUG_VM_RB
2582 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2585 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2586 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2587 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2590 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2591 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2592 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2593 struct vm_area_struct *expand);
2594 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2595 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2597 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2599 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2600 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2601 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2602 struct mempolicy *, struct vm_userfaultfd_ctx, const char *);
2603 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2604 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2605 unsigned long addr, int new_below);
2606 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2607 unsigned long addr, int new_below);
2608 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2609 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2610 struct rb_node **, struct rb_node *);
2611 extern void unlink_file_vma(struct vm_area_struct *);
2612 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2613 unsigned long addr, unsigned long len, pgoff_t pgoff,
2614 bool *need_rmap_locks);
2615 extern void exit_mmap(struct mm_struct *);
2617 static inline int check_data_rlimit(unsigned long rlim,
2619 unsigned long start,
2620 unsigned long end_data,
2621 unsigned long start_data)
2623 if (rlim < RLIM_INFINITY) {
2624 if (((new - start) + (end_data - start_data)) > rlim)
2631 extern int mm_take_all_locks(struct mm_struct *mm);
2632 extern void mm_drop_all_locks(struct mm_struct *mm);
2634 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2635 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2636 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2637 extern struct file *get_task_exe_file(struct task_struct *task);
2639 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2640 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2642 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2643 const struct vm_special_mapping *sm);
2644 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2645 unsigned long addr, unsigned long len,
2646 unsigned long flags,
2647 const struct vm_special_mapping *spec);
2648 /* This is an obsolete alternative to _install_special_mapping. */
2649 extern int install_special_mapping(struct mm_struct *mm,
2650 unsigned long addr, unsigned long len,
2651 unsigned long flags, struct page **pages);
2653 unsigned long randomize_stack_top(unsigned long stack_top);
2655 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2657 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2658 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2659 struct list_head *uf);
2660 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2661 unsigned long len, unsigned long prot, unsigned long flags,
2662 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2663 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2664 struct list_head *uf, bool downgrade);
2665 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2666 struct list_head *uf);
2667 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2670 extern int __mm_populate(unsigned long addr, unsigned long len,
2672 static inline void mm_populate(unsigned long addr, unsigned long len)
2675 (void) __mm_populate(addr, len, 1);
2678 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2681 /* These take the mm semaphore themselves */
2682 extern int __must_check vm_brk(unsigned long, unsigned long);
2683 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2684 extern int vm_munmap(unsigned long, size_t);
2685 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2686 unsigned long, unsigned long,
2687 unsigned long, unsigned long);
2689 struct vm_unmapped_area_info {
2690 #define VM_UNMAPPED_AREA_TOPDOWN 1
2691 unsigned long flags;
2692 unsigned long length;
2693 unsigned long low_limit;
2694 unsigned long high_limit;
2695 unsigned long align_mask;
2696 unsigned long align_offset;
2699 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2702 extern void truncate_inode_pages(struct address_space *, loff_t);
2703 extern void truncate_inode_pages_range(struct address_space *,
2704 loff_t lstart, loff_t lend);
2705 extern void truncate_inode_pages_final(struct address_space *);
2707 /* generic vm_area_ops exported for stackable file systems */
2708 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2709 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2710 pgoff_t start_pgoff, pgoff_t end_pgoff);
2711 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2713 extern unsigned long stack_guard_gap;
2714 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2715 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2717 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2718 extern int expand_downwards(struct vm_area_struct *vma,
2719 unsigned long address);
2721 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2723 #define expand_upwards(vma, address) (0)
2726 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2727 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2728 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2729 struct vm_area_struct **pprev);
2732 * find_vma_intersection() - Look up the first VMA which intersects the interval
2733 * @mm: The process address space.
2734 * @start_addr: The inclusive start user address.
2735 * @end_addr: The exclusive end user address.
2737 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2738 * start_addr < end_addr.
2741 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2742 unsigned long start_addr,
2743 unsigned long end_addr)
2745 struct vm_area_struct *vma = find_vma(mm, start_addr);
2747 if (vma && end_addr <= vma->vm_start)
2753 * vma_lookup() - Find a VMA at a specific address
2754 * @mm: The process address space.
2755 * @addr: The user address.
2757 * Return: The vm_area_struct at the given address, %NULL otherwise.
2760 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2762 struct vm_area_struct *vma = find_vma(mm, addr);
2764 if (vma && addr < vma->vm_start)
2770 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2772 unsigned long vm_start = vma->vm_start;
2774 if (vma->vm_flags & VM_GROWSDOWN) {
2775 vm_start -= stack_guard_gap;
2776 if (vm_start > vma->vm_start)
2782 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2784 unsigned long vm_end = vma->vm_end;
2786 if (vma->vm_flags & VM_GROWSUP) {
2787 vm_end += stack_guard_gap;
2788 if (vm_end < vma->vm_end)
2789 vm_end = -PAGE_SIZE;
2794 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2796 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2799 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2800 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2801 unsigned long vm_start, unsigned long vm_end)
2803 struct vm_area_struct *vma = find_vma(mm, vm_start);
2805 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2811 static inline bool range_in_vma(struct vm_area_struct *vma,
2812 unsigned long start, unsigned long end)
2814 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2818 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2819 void vma_set_page_prot(struct vm_area_struct *vma);
2821 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2825 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2827 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2831 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2833 #ifdef CONFIG_NUMA_BALANCING
2834 unsigned long change_prot_numa(struct vm_area_struct *vma,
2835 unsigned long start, unsigned long end);
2838 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2839 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2840 unsigned long pfn, unsigned long size, pgprot_t);
2841 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2842 unsigned long pfn, unsigned long size, pgprot_t prot);
2843 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2844 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2845 struct page **pages, unsigned long *num);
2846 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2848 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2850 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2852 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2853 unsigned long pfn, pgprot_t pgprot);
2854 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2856 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2857 pfn_t pfn, pgprot_t pgprot);
2858 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2859 unsigned long addr, pfn_t pfn);
2860 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2862 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2863 unsigned long addr, struct page *page)
2865 int err = vm_insert_page(vma, addr, page);
2868 return VM_FAULT_OOM;
2869 if (err < 0 && err != -EBUSY)
2870 return VM_FAULT_SIGBUS;
2872 return VM_FAULT_NOPAGE;
2875 #ifndef io_remap_pfn_range
2876 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2877 unsigned long addr, unsigned long pfn,
2878 unsigned long size, pgprot_t prot)
2880 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2884 static inline vm_fault_t vmf_error(int err)
2887 return VM_FAULT_OOM;
2888 return VM_FAULT_SIGBUS;
2891 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2892 unsigned int foll_flags);
2894 #define FOLL_WRITE 0x01 /* check pte is writable */
2895 #define FOLL_TOUCH 0x02 /* mark page accessed */
2896 #define FOLL_GET 0x04 /* do get_page on page */
2897 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2898 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2899 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2900 * and return without waiting upon it */
2901 #define FOLL_POPULATE 0x40 /* fault in pages (with FOLL_MLOCK) */
2902 #define FOLL_NOFAULT 0x80 /* do not fault in pages */
2903 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2904 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2905 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2906 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2907 #define FOLL_MLOCK 0x1000 /* lock present pages */
2908 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2909 #define FOLL_COW 0x4000 /* internal GUP flag */
2910 #define FOLL_ANON 0x8000 /* don't do file mappings */
2911 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2912 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2913 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2914 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2917 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2918 * other. Here is what they mean, and how to use them:
2920 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2921 * period _often_ under userspace control. This is in contrast to
2922 * iov_iter_get_pages(), whose usages are transient.
2924 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2925 * lifetime enforced by the filesystem and we need guarantees that longterm
2926 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2927 * the filesystem. Ideas for this coordination include revoking the longterm
2928 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2929 * added after the problem with filesystems was found FS DAX VMAs are
2930 * specifically failed. Filesystem pages are still subject to bugs and use of
2931 * FOLL_LONGTERM should be avoided on those pages.
2933 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2934 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2935 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2936 * is due to an incompatibility with the FS DAX check and
2937 * FAULT_FLAG_ALLOW_RETRY.
2939 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2940 * that region. And so, CMA attempts to migrate the page before pinning, when
2941 * FOLL_LONGTERM is specified.
2943 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2944 * but an additional pin counting system) will be invoked. This is intended for
2945 * anything that gets a page reference and then touches page data (for example,
2946 * Direct IO). This lets the filesystem know that some non-file-system entity is
2947 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2948 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2949 * a call to unpin_user_page().
2951 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2952 * and separate refcounting mechanisms, however, and that means that each has
2953 * its own acquire and release mechanisms:
2955 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2957 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2959 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2960 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2961 * calls applied to them, and that's perfectly OK. This is a constraint on the
2962 * callers, not on the pages.)
2964 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2965 * directly by the caller. That's in order to help avoid mismatches when
2966 * releasing pages: get_user_pages*() pages must be released via put_page(),
2967 * while pin_user_pages*() pages must be released via unpin_user_page().
2969 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2972 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2974 if (vm_fault & VM_FAULT_OOM)
2976 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2977 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2978 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2983 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2984 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2985 unsigned long size, pte_fn_t fn, void *data);
2986 extern int apply_to_existing_page_range(struct mm_struct *mm,
2987 unsigned long address, unsigned long size,
2988 pte_fn_t fn, void *data);
2990 extern void init_mem_debugging_and_hardening(void);
2991 #ifdef CONFIG_PAGE_POISONING
2992 extern void __kernel_poison_pages(struct page *page, int numpages);
2993 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2994 extern bool _page_poisoning_enabled_early;
2995 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2996 static inline bool page_poisoning_enabled(void)
2998 return _page_poisoning_enabled_early;
3001 * For use in fast paths after init_mem_debugging() has run, or when a
3002 * false negative result is not harmful when called too early.
3004 static inline bool page_poisoning_enabled_static(void)
3006 return static_branch_unlikely(&_page_poisoning_enabled);
3008 static inline void kernel_poison_pages(struct page *page, int numpages)
3010 if (page_poisoning_enabled_static())
3011 __kernel_poison_pages(page, numpages);
3013 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3015 if (page_poisoning_enabled_static())
3016 __kernel_unpoison_pages(page, numpages);
3019 static inline bool page_poisoning_enabled(void) { return false; }
3020 static inline bool page_poisoning_enabled_static(void) { return false; }
3021 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3022 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3023 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3026 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3027 static inline bool want_init_on_alloc(gfp_t flags)
3029 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3032 return flags & __GFP_ZERO;
3035 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3036 static inline bool want_init_on_free(void)
3038 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3042 extern bool _debug_pagealloc_enabled_early;
3043 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3045 static inline bool debug_pagealloc_enabled(void)
3047 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3048 _debug_pagealloc_enabled_early;
3052 * For use in fast paths after init_debug_pagealloc() has run, or when a
3053 * false negative result is not harmful when called too early.
3055 static inline bool debug_pagealloc_enabled_static(void)
3057 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3060 return static_branch_unlikely(&_debug_pagealloc_enabled);
3063 #ifdef CONFIG_DEBUG_PAGEALLOC
3065 * To support DEBUG_PAGEALLOC architecture must ensure that
3066 * __kernel_map_pages() never fails
3068 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3070 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3072 if (debug_pagealloc_enabled_static())
3073 __kernel_map_pages(page, numpages, 1);
3076 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3078 if (debug_pagealloc_enabled_static())
3079 __kernel_map_pages(page, numpages, 0);
3081 #else /* CONFIG_DEBUG_PAGEALLOC */
3082 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3083 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3084 #endif /* CONFIG_DEBUG_PAGEALLOC */
3086 #ifdef __HAVE_ARCH_GATE_AREA
3087 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3088 extern int in_gate_area_no_mm(unsigned long addr);
3089 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3091 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3095 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3096 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3100 #endif /* __HAVE_ARCH_GATE_AREA */
3102 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3104 #ifdef CONFIG_SYSCTL
3105 extern int sysctl_drop_caches;
3106 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3110 void drop_slab(void);
3113 #define randomize_va_space 0
3115 extern int randomize_va_space;
3118 const char * arch_vma_name(struct vm_area_struct *vma);
3120 void print_vma_addr(char *prefix, unsigned long rip);
3122 static inline void print_vma_addr(char *prefix, unsigned long rip)
3127 int vmemmap_remap_free(unsigned long start, unsigned long end,
3128 unsigned long reuse);
3129 int vmemmap_remap_alloc(unsigned long start, unsigned long end,
3130 unsigned long reuse, gfp_t gfp_mask);
3132 void *sparse_buffer_alloc(unsigned long size);
3133 struct page * __populate_section_memmap(unsigned long pfn,
3134 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3135 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3136 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3137 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3138 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3139 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3140 struct vmem_altmap *altmap);
3141 void *vmemmap_alloc_block(unsigned long size, int node);
3143 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3144 struct vmem_altmap *altmap);
3145 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3146 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3147 int node, struct vmem_altmap *altmap);
3148 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3149 struct vmem_altmap *altmap);
3150 void vmemmap_populate_print_last(void);
3151 #ifdef CONFIG_MEMORY_HOTPLUG
3152 void vmemmap_free(unsigned long start, unsigned long end,
3153 struct vmem_altmap *altmap);
3155 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3156 unsigned long nr_pages);
3159 MF_COUNT_INCREASED = 1 << 0,
3160 MF_ACTION_REQUIRED = 1 << 1,
3161 MF_MUST_KILL = 1 << 2,
3162 MF_SOFT_OFFLINE = 1 << 3,
3163 MF_UNPOISON = 1 << 4,
3165 extern int memory_failure(unsigned long pfn, int flags);
3166 extern void memory_failure_queue(unsigned long pfn, int flags);
3167 extern void memory_failure_queue_kick(int cpu);
3168 extern int unpoison_memory(unsigned long pfn);
3169 extern int sysctl_memory_failure_early_kill;
3170 extern int sysctl_memory_failure_recovery;
3171 extern void shake_page(struct page *p);
3172 extern atomic_long_t num_poisoned_pages __read_mostly;
3173 extern int soft_offline_page(unsigned long pfn, int flags);
3175 #ifndef arch_memory_failure
3176 static inline int arch_memory_failure(unsigned long pfn, int flags)
3182 #ifndef arch_is_platform_page
3183 static inline bool arch_is_platform_page(u64 paddr)
3190 * Error handlers for various types of pages.
3193 MF_IGNORED, /* Error: cannot be handled */
3194 MF_FAILED, /* Error: handling failed */
3195 MF_DELAYED, /* Will be handled later */
3196 MF_RECOVERED, /* Successfully recovered */
3199 enum mf_action_page_type {
3201 MF_MSG_KERNEL_HIGH_ORDER,
3203 MF_MSG_DIFFERENT_COMPOUND,
3206 MF_MSG_NON_PMD_HUGE,
3207 MF_MSG_UNMAP_FAILED,
3208 MF_MSG_DIRTY_SWAPCACHE,
3209 MF_MSG_CLEAN_SWAPCACHE,
3210 MF_MSG_DIRTY_MLOCKED_LRU,
3211 MF_MSG_CLEAN_MLOCKED_LRU,
3212 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3213 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3216 MF_MSG_TRUNCATED_LRU,
3223 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3224 extern void clear_huge_page(struct page *page,
3225 unsigned long addr_hint,
3226 unsigned int pages_per_huge_page);
3227 extern void copy_user_huge_page(struct page *dst, struct page *src,
3228 unsigned long addr_hint,
3229 struct vm_area_struct *vma,
3230 unsigned int pages_per_huge_page);
3231 extern long copy_huge_page_from_user(struct page *dst_page,
3232 const void __user *usr_src,
3233 unsigned int pages_per_huge_page,
3234 bool allow_pagefault);
3237 * vma_is_special_huge - Are transhuge page-table entries considered special?
3238 * @vma: Pointer to the struct vm_area_struct to consider
3240 * Whether transhuge page-table entries are considered "special" following
3241 * the definition in vm_normal_page().
3243 * Return: true if transhuge page-table entries should be considered special,
3246 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3248 return vma_is_dax(vma) || (vma->vm_file &&
3249 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3252 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3254 #ifdef CONFIG_DEBUG_PAGEALLOC
3255 extern unsigned int _debug_guardpage_minorder;
3256 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3258 static inline unsigned int debug_guardpage_minorder(void)
3260 return _debug_guardpage_minorder;
3263 static inline bool debug_guardpage_enabled(void)
3265 return static_branch_unlikely(&_debug_guardpage_enabled);
3268 static inline bool page_is_guard(struct page *page)
3270 if (!debug_guardpage_enabled())
3273 return PageGuard(page);
3276 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3277 static inline bool debug_guardpage_enabled(void) { return false; }
3278 static inline bool page_is_guard(struct page *page) { return false; }
3279 #endif /* CONFIG_DEBUG_PAGEALLOC */
3281 #if MAX_NUMNODES > 1
3282 void __init setup_nr_node_ids(void);
3284 static inline void setup_nr_node_ids(void) {}
3287 extern int memcmp_pages(struct page *page1, struct page *page2);
3289 static inline int pages_identical(struct page *page1, struct page *page2)
3291 return !memcmp_pages(page1, page2);
3294 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3295 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3296 pgoff_t first_index, pgoff_t nr,
3297 pgoff_t bitmap_pgoff,
3298 unsigned long *bitmap,
3302 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3303 pgoff_t first_index, pgoff_t nr);
3306 extern int sysctl_nr_trim_pages;
3308 #ifdef CONFIG_PRINTK
3309 void mem_dump_obj(void *object);
3311 static inline void mem_dump_obj(void *object) {}
3315 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3316 * @seals: the seals to check
3317 * @vma: the vma to operate on
3319 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3320 * the vma flags. Return 0 if check pass, or <0 for errors.
3322 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3324 if (seals & F_SEAL_FUTURE_WRITE) {
3326 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3327 * "future write" seal active.
3329 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3333 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3334 * MAP_SHARED and read-only, take care to not allow mprotect to
3335 * revert protections on such mappings. Do this only for shared
3336 * mappings. For private mappings, don't need to mask
3337 * VM_MAYWRITE as we still want them to be COW-writable.
3339 if (vma->vm_flags & VM_SHARED)
3340 vma->vm_flags &= ~(VM_MAYWRITE);
3346 #ifdef CONFIG_ANON_VMA_NAME
3347 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3348 unsigned long len_in, const char *name);
3351 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3352 unsigned long len_in, const char *name) {
3357 #endif /* __KERNEL__ */
3358 #endif /* _LINUX_MM_H */