1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
6 #include <linux/mmdebug.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/atomic.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/mmap_lock.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/percpu-refcount.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
24 #include <linux/page-flags.h>
25 #include <linux/page_ref.h>
26 #include <linux/overflow.h>
27 #include <linux/sizes.h>
28 #include <linux/sched.h>
29 #include <linux/pgtable.h>
30 #include <linux/kasan.h>
34 struct anon_vma_chain;
38 extern int sysctl_page_lock_unfairness;
40 void init_mm_internals(void);
42 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
43 extern unsigned long max_mapnr;
45 static inline void set_max_mapnr(unsigned long limit)
50 static inline void set_max_mapnr(unsigned long limit) { }
53 extern atomic_long_t _totalram_pages;
54 static inline unsigned long totalram_pages(void)
56 return (unsigned long)atomic_long_read(&_totalram_pages);
59 static inline void totalram_pages_inc(void)
61 atomic_long_inc(&_totalram_pages);
64 static inline void totalram_pages_dec(void)
66 atomic_long_dec(&_totalram_pages);
69 static inline void totalram_pages_add(long count)
71 atomic_long_add(count, &_totalram_pages);
74 extern void * high_memory;
75 extern int page_cluster;
78 extern int sysctl_legacy_va_layout;
80 #define sysctl_legacy_va_layout 0
83 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
84 extern const int mmap_rnd_bits_min;
85 extern const int mmap_rnd_bits_max;
86 extern int mmap_rnd_bits __read_mostly;
88 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
89 extern const int mmap_rnd_compat_bits_min;
90 extern const int mmap_rnd_compat_bits_max;
91 extern int mmap_rnd_compat_bits __read_mostly;
95 #include <asm/processor.h>
98 * Architectures that support memory tagging (assigning tags to memory regions,
99 * embedding these tags into addresses that point to these memory regions, and
100 * checking that the memory and the pointer tags match on memory accesses)
101 * redefine this macro to strip tags from pointers.
102 * It's defined as noop for architectures that don't support memory tagging.
104 #ifndef untagged_addr
105 #define untagged_addr(addr) (addr)
109 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
113 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
117 #define lm_alias(x) __va(__pa_symbol(x))
121 * To prevent common memory management code establishing
122 * a zero page mapping on a read fault.
123 * This macro should be defined within <asm/pgtable.h>.
124 * s390 does this to prevent multiplexing of hardware bits
125 * related to the physical page in case of virtualization.
127 #ifndef mm_forbids_zeropage
128 #define mm_forbids_zeropage(X) (0)
132 * On some architectures it is expensive to call memset() for small sizes.
133 * If an architecture decides to implement their own version of
134 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
135 * define their own version of this macro in <asm/pgtable.h>
137 #if BITS_PER_LONG == 64
138 /* This function must be updated when the size of struct page grows above 80
139 * or reduces below 56. The idea that compiler optimizes out switch()
140 * statement, and only leaves move/store instructions. Also the compiler can
141 * combine write statements if they are both assignments and can be reordered,
142 * this can result in several of the writes here being dropped.
144 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
145 static inline void __mm_zero_struct_page(struct page *page)
147 unsigned long *_pp = (void *)page;
149 /* Check that struct page is either 56, 64, 72, or 80 bytes */
150 BUILD_BUG_ON(sizeof(struct page) & 7);
151 BUILD_BUG_ON(sizeof(struct page) < 56);
152 BUILD_BUG_ON(sizeof(struct page) > 80);
154 switch (sizeof(struct page)) {
175 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
179 * Default maximum number of active map areas, this limits the number of vmas
180 * per mm struct. Users can overwrite this number by sysctl but there is a
183 * When a program's coredump is generated as ELF format, a section is created
184 * per a vma. In ELF, the number of sections is represented in unsigned short.
185 * This means the number of sections should be smaller than 65535 at coredump.
186 * Because the kernel adds some informative sections to a image of program at
187 * generating coredump, we need some margin. The number of extra sections is
188 * 1-3 now and depends on arch. We use "5" as safe margin, here.
190 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
191 * not a hard limit any more. Although some userspace tools can be surprised by
194 #define MAPCOUNT_ELF_CORE_MARGIN (5)
195 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
197 extern int sysctl_max_map_count;
199 extern unsigned long sysctl_user_reserve_kbytes;
200 extern unsigned long sysctl_admin_reserve_kbytes;
202 extern int sysctl_overcommit_memory;
203 extern int sysctl_overcommit_ratio;
204 extern unsigned long sysctl_overcommit_kbytes;
206 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
208 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
210 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
213 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
214 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
215 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
217 #define nth_page(page,n) ((page) + (n))
218 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
221 /* to align the pointer to the (next) page boundary */
222 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
224 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
225 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
227 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
228 static inline struct folio *lru_to_folio(struct list_head *head)
230 return list_entry((head)->prev, struct folio, lru);
233 void setup_initial_init_mm(void *start_code, void *end_code,
234 void *end_data, void *brk);
237 * Linux kernel virtual memory manager primitives.
238 * The idea being to have a "virtual" mm in the same way
239 * we have a virtual fs - giving a cleaner interface to the
240 * mm details, and allowing different kinds of memory mappings
241 * (from shared memory to executable loading to arbitrary
245 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
246 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
247 void vm_area_free(struct vm_area_struct *);
250 extern struct rb_root nommu_region_tree;
251 extern struct rw_semaphore nommu_region_sem;
253 extern unsigned int kobjsize(const void *objp);
257 * vm_flags in vm_area_struct, see mm_types.h.
258 * When changing, update also include/trace/events/mmflags.h
260 #define VM_NONE 0x00000000
262 #define VM_READ 0x00000001 /* currently active flags */
263 #define VM_WRITE 0x00000002
264 #define VM_EXEC 0x00000004
265 #define VM_SHARED 0x00000008
267 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
268 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
269 #define VM_MAYWRITE 0x00000020
270 #define VM_MAYEXEC 0x00000040
271 #define VM_MAYSHARE 0x00000080
273 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
274 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
275 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
276 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
278 #define VM_LOCKED 0x00002000
279 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
281 /* Used by sys_madvise() */
282 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
283 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
285 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
286 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
287 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
288 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
289 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
290 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
291 #define VM_SYNC 0x00800000 /* Synchronous page faults */
292 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
293 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
294 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
296 #ifdef CONFIG_MEM_SOFT_DIRTY
297 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
299 # define VM_SOFTDIRTY 0
302 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
303 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
304 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
305 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
307 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
308 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
309 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
310 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
311 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
312 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
313 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
314 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
315 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
316 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
317 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
318 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
320 #ifdef CONFIG_ARCH_HAS_PKEYS
321 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
322 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
323 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
324 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
325 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
327 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
329 # define VM_PKEY_BIT4 0
331 #endif /* CONFIG_ARCH_HAS_PKEYS */
333 #if defined(CONFIG_X86)
334 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
335 #elif defined(CONFIG_PPC)
336 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
337 #elif defined(CONFIG_PARISC)
338 # define VM_GROWSUP VM_ARCH_1
339 #elif defined(CONFIG_IA64)
340 # define VM_GROWSUP VM_ARCH_1
341 #elif defined(CONFIG_SPARC64)
342 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
343 # define VM_ARCH_CLEAR VM_SPARC_ADI
344 #elif defined(CONFIG_ARM64)
345 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
346 # define VM_ARCH_CLEAR VM_ARM64_BTI
347 #elif !defined(CONFIG_MMU)
348 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
351 #if defined(CONFIG_ARM64_MTE)
352 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
353 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
355 # define VM_MTE VM_NONE
356 # define VM_MTE_ALLOWED VM_NONE
360 # define VM_GROWSUP VM_NONE
363 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
364 # define VM_UFFD_MINOR_BIT 37
365 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
366 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
367 # define VM_UFFD_MINOR VM_NONE
368 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
370 /* Bits set in the VMA until the stack is in its final location */
371 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
373 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
375 /* Common data flag combinations */
376 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
377 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
378 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
379 VM_MAYWRITE | VM_MAYEXEC)
380 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
381 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
383 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
384 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
387 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
388 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
391 #ifdef CONFIG_STACK_GROWSUP
392 #define VM_STACK VM_GROWSUP
394 #define VM_STACK VM_GROWSDOWN
397 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
399 /* VMA basic access permission flags */
400 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
404 * Special vmas that are non-mergable, non-mlock()able.
406 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
408 /* This mask prevents VMA from being scanned with khugepaged */
409 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
411 /* This mask defines which mm->def_flags a process can inherit its parent */
412 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
414 /* This mask is used to clear all the VMA flags used by mlock */
415 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
417 /* Arch-specific flags to clear when updating VM flags on protection change */
418 #ifndef VM_ARCH_CLEAR
419 # define VM_ARCH_CLEAR VM_NONE
421 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
424 * mapping from the currently active vm_flags protection bits (the
425 * low four bits) to a page protection mask..
427 extern pgprot_t protection_map[16];
430 * The default fault flags that should be used by most of the
431 * arch-specific page fault handlers.
433 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
434 FAULT_FLAG_KILLABLE | \
435 FAULT_FLAG_INTERRUPTIBLE)
438 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
439 * @flags: Fault flags.
441 * This is mostly used for places where we want to try to avoid taking
442 * the mmap_lock for too long a time when waiting for another condition
443 * to change, in which case we can try to be polite to release the
444 * mmap_lock in the first round to avoid potential starvation of other
445 * processes that would also want the mmap_lock.
447 * Return: true if the page fault allows retry and this is the first
448 * attempt of the fault handling; false otherwise.
450 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
452 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
453 (!(flags & FAULT_FLAG_TRIED));
456 #define FAULT_FLAG_TRACE \
457 { FAULT_FLAG_WRITE, "WRITE" }, \
458 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
459 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
460 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
461 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
462 { FAULT_FLAG_TRIED, "TRIED" }, \
463 { FAULT_FLAG_USER, "USER" }, \
464 { FAULT_FLAG_REMOTE, "REMOTE" }, \
465 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
466 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
469 * vm_fault is filled by the pagefault handler and passed to the vma's
470 * ->fault function. The vma's ->fault is responsible for returning a bitmask
471 * of VM_FAULT_xxx flags that give details about how the fault was handled.
473 * MM layer fills up gfp_mask for page allocations but fault handler might
474 * alter it if its implementation requires a different allocation context.
476 * pgoff should be used in favour of virtual_address, if possible.
480 struct vm_area_struct *vma; /* Target VMA */
481 gfp_t gfp_mask; /* gfp mask to be used for allocations */
482 pgoff_t pgoff; /* Logical page offset based on vma */
483 unsigned long address; /* Faulting virtual address - masked */
484 unsigned long real_address; /* Faulting virtual address - unmasked */
486 enum fault_flag flags; /* FAULT_FLAG_xxx flags
487 * XXX: should really be 'const' */
488 pmd_t *pmd; /* Pointer to pmd entry matching
490 pud_t *pud; /* Pointer to pud entry matching
494 pte_t orig_pte; /* Value of PTE at the time of fault */
495 pmd_t orig_pmd; /* Value of PMD at the time of fault,
496 * used by PMD fault only.
500 struct page *cow_page; /* Page handler may use for COW fault */
501 struct page *page; /* ->fault handlers should return a
502 * page here, unless VM_FAULT_NOPAGE
503 * is set (which is also implied by
506 /* These three entries are valid only while holding ptl lock */
507 pte_t *pte; /* Pointer to pte entry matching
508 * the 'address'. NULL if the page
509 * table hasn't been allocated.
511 spinlock_t *ptl; /* Page table lock.
512 * Protects pte page table if 'pte'
513 * is not NULL, otherwise pmd.
515 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
516 * vm_ops->map_pages() sets up a page
517 * table from atomic context.
518 * do_fault_around() pre-allocates
519 * page table to avoid allocation from
524 /* page entry size for vm->huge_fault() */
525 enum page_entry_size {
532 * These are the virtual MM functions - opening of an area, closing and
533 * unmapping it (needed to keep files on disk up-to-date etc), pointer
534 * to the functions called when a no-page or a wp-page exception occurs.
536 struct vm_operations_struct {
537 void (*open)(struct vm_area_struct * area);
539 * @close: Called when the VMA is being removed from the MM.
540 * Context: User context. May sleep. Caller holds mmap_lock.
542 void (*close)(struct vm_area_struct * area);
543 /* Called any time before splitting to check if it's allowed */
544 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
545 int (*mremap)(struct vm_area_struct *area);
547 * Called by mprotect() to make driver-specific permission
548 * checks before mprotect() is finalised. The VMA must not
549 * be modified. Returns 0 if eprotect() can proceed.
551 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
552 unsigned long end, unsigned long newflags);
553 vm_fault_t (*fault)(struct vm_fault *vmf);
554 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
555 enum page_entry_size pe_size);
556 vm_fault_t (*map_pages)(struct vm_fault *vmf,
557 pgoff_t start_pgoff, pgoff_t end_pgoff);
558 unsigned long (*pagesize)(struct vm_area_struct * area);
560 /* notification that a previously read-only page is about to become
561 * writable, if an error is returned it will cause a SIGBUS */
562 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
564 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
565 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
567 /* called by access_process_vm when get_user_pages() fails, typically
568 * for use by special VMAs. See also generic_access_phys() for a generic
569 * implementation useful for any iomem mapping.
571 int (*access)(struct vm_area_struct *vma, unsigned long addr,
572 void *buf, int len, int write);
574 /* Called by the /proc/PID/maps code to ask the vma whether it
575 * has a special name. Returning non-NULL will also cause this
576 * vma to be dumped unconditionally. */
577 const char *(*name)(struct vm_area_struct *vma);
581 * set_policy() op must add a reference to any non-NULL @new mempolicy
582 * to hold the policy upon return. Caller should pass NULL @new to
583 * remove a policy and fall back to surrounding context--i.e. do not
584 * install a MPOL_DEFAULT policy, nor the task or system default
587 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
590 * get_policy() op must add reference [mpol_get()] to any policy at
591 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
592 * in mm/mempolicy.c will do this automatically.
593 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
594 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
595 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
596 * must return NULL--i.e., do not "fallback" to task or system default
599 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
603 * Called by vm_normal_page() for special PTEs to find the
604 * page for @addr. This is useful if the default behavior
605 * (using pte_page()) would not find the correct page.
607 struct page *(*find_special_page)(struct vm_area_struct *vma,
611 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
613 static const struct vm_operations_struct dummy_vm_ops = {};
615 memset(vma, 0, sizeof(*vma));
617 vma->vm_ops = &dummy_vm_ops;
618 INIT_LIST_HEAD(&vma->anon_vma_chain);
621 static inline void vma_set_anonymous(struct vm_area_struct *vma)
626 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
631 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
633 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
638 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
639 VM_STACK_INCOMPLETE_SETUP)
645 static inline bool vma_is_foreign(struct vm_area_struct *vma)
650 if (current->mm != vma->vm_mm)
656 static inline bool vma_is_accessible(struct vm_area_struct *vma)
658 return vma->vm_flags & VM_ACCESS_FLAGS;
663 * The vma_is_shmem is not inline because it is used only by slow
664 * paths in userfault.
666 bool vma_is_shmem(struct vm_area_struct *vma);
668 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
671 int vma_is_stack_for_current(struct vm_area_struct *vma);
673 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
674 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
679 static inline unsigned int compound_order(struct page *page)
683 return page[1].compound_order;
687 * folio_order - The allocation order of a folio.
690 * A folio is composed of 2^order pages. See get_order() for the definition
693 * Return: The order of the folio.
695 static inline unsigned int folio_order(struct folio *folio)
697 return compound_order(&folio->page);
700 #include <linux/huge_mm.h>
703 * Methods to modify the page usage count.
705 * What counts for a page usage:
706 * - cache mapping (page->mapping)
707 * - private data (page->private)
708 * - page mapped in a task's page tables, each mapping
709 * is counted separately
711 * Also, many kernel routines increase the page count before a critical
712 * routine so they can be sure the page doesn't go away from under them.
716 * Drop a ref, return true if the refcount fell to zero (the page has no users)
718 static inline int put_page_testzero(struct page *page)
720 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
721 return page_ref_dec_and_test(page);
724 static inline int folio_put_testzero(struct folio *folio)
726 return put_page_testzero(&folio->page);
730 * Try to grab a ref unless the page has a refcount of zero, return false if
732 * This can be called when MMU is off so it must not access
733 * any of the virtual mappings.
735 static inline bool get_page_unless_zero(struct page *page)
737 return page_ref_add_unless(page, 1, 0);
740 extern int page_is_ram(unsigned long pfn);
748 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
751 /* Support for virtually mapped pages */
752 struct page *vmalloc_to_page(const void *addr);
753 unsigned long vmalloc_to_pfn(const void *addr);
756 * Determine if an address is within the vmalloc range
758 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
759 * is no special casing required.
762 #ifndef is_ioremap_addr
763 #define is_ioremap_addr(x) is_vmalloc_addr(x)
767 extern bool is_vmalloc_addr(const void *x);
768 extern int is_vmalloc_or_module_addr(const void *x);
770 static inline bool is_vmalloc_addr(const void *x)
774 static inline int is_vmalloc_or_module_addr(const void *x)
781 * How many times the entire folio is mapped as a single unit (eg by a
782 * PMD or PUD entry). This is probably not what you want, except for
783 * debugging purposes; look at folio_mapcount() or page_mapcount()
786 static inline int folio_entire_mapcount(struct folio *folio)
788 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
789 return atomic_read(folio_mapcount_ptr(folio)) + 1;
793 * Mapcount of compound page as a whole, does not include mapped sub-pages.
795 * Must be called only for compound pages.
797 static inline int compound_mapcount(struct page *page)
799 return folio_entire_mapcount(page_folio(page));
803 * The atomic page->_mapcount, starts from -1: so that transitions
804 * both from it and to it can be tracked, using atomic_inc_and_test
805 * and atomic_add_negative(-1).
807 static inline void page_mapcount_reset(struct page *page)
809 atomic_set(&(page)->_mapcount, -1);
812 int __page_mapcount(struct page *page);
815 * Mapcount of 0-order page; when compound sub-page, includes
816 * compound_mapcount().
818 * Result is undefined for pages which cannot be mapped into userspace.
819 * For example SLAB or special types of pages. See function page_has_type().
820 * They use this place in struct page differently.
822 static inline int page_mapcount(struct page *page)
824 if (unlikely(PageCompound(page)))
825 return __page_mapcount(page);
826 return atomic_read(&page->_mapcount) + 1;
829 int folio_mapcount(struct folio *folio);
831 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
832 static inline int total_mapcount(struct page *page)
834 return folio_mapcount(page_folio(page));
838 static inline int total_mapcount(struct page *page)
840 return page_mapcount(page);
844 static inline struct page *virt_to_head_page(const void *x)
846 struct page *page = virt_to_page(x);
848 return compound_head(page);
851 static inline struct folio *virt_to_folio(const void *x)
853 struct page *page = virt_to_page(x);
855 return page_folio(page);
858 void __put_page(struct page *page);
860 void put_pages_list(struct list_head *pages);
862 void split_page(struct page *page, unsigned int order);
863 void folio_copy(struct folio *dst, struct folio *src);
865 unsigned long nr_free_buffer_pages(void);
868 * Compound pages have a destructor function. Provide a
869 * prototype for that function and accessor functions.
870 * These are _only_ valid on the head of a compound page.
872 typedef void compound_page_dtor(struct page *);
874 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
875 enum compound_dtor_id {
878 #ifdef CONFIG_HUGETLB_PAGE
881 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
886 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
888 static inline void set_compound_page_dtor(struct page *page,
889 enum compound_dtor_id compound_dtor)
891 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
892 page[1].compound_dtor = compound_dtor;
895 static inline void destroy_compound_page(struct page *page)
897 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
898 compound_page_dtors[page[1].compound_dtor](page);
901 static inline int head_compound_pincount(struct page *head)
903 return atomic_read(compound_pincount_ptr(head));
906 static inline void set_compound_order(struct page *page, unsigned int order)
908 page[1].compound_order = order;
910 page[1].compound_nr = 1U << order;
914 /* Returns the number of pages in this potentially compound page. */
915 static inline unsigned long compound_nr(struct page *page)
920 return page[1].compound_nr;
922 return 1UL << compound_order(page);
926 /* Returns the number of bytes in this potentially compound page. */
927 static inline unsigned long page_size(struct page *page)
929 return PAGE_SIZE << compound_order(page);
932 /* Returns the number of bits needed for the number of bytes in a page */
933 static inline unsigned int page_shift(struct page *page)
935 return PAGE_SHIFT + compound_order(page);
939 * thp_order - Order of a transparent huge page.
940 * @page: Head page of a transparent huge page.
942 static inline unsigned int thp_order(struct page *page)
944 VM_BUG_ON_PGFLAGS(PageTail(page), page);
945 return compound_order(page);
949 * thp_nr_pages - The number of regular pages in this huge page.
950 * @page: The head page of a huge page.
952 static inline int thp_nr_pages(struct page *page)
954 VM_BUG_ON_PGFLAGS(PageTail(page), page);
955 return compound_nr(page);
959 * thp_size - Size of a transparent huge page.
960 * @page: Head page of a transparent huge page.
962 * Return: Number of bytes in this page.
964 static inline unsigned long thp_size(struct page *page)
966 return PAGE_SIZE << thp_order(page);
969 void free_compound_page(struct page *page);
973 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
974 * servicing faults for write access. In the normal case, do always want
975 * pte_mkwrite. But get_user_pages can cause write faults for mappings
976 * that do not have writing enabled, when used by access_process_vm.
978 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
980 if (likely(vma->vm_flags & VM_WRITE))
981 pte = pte_mkwrite(pte);
985 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
986 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
988 vm_fault_t finish_fault(struct vm_fault *vmf);
989 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
993 * Multiple processes may "see" the same page. E.g. for untouched
994 * mappings of /dev/null, all processes see the same page full of
995 * zeroes, and text pages of executables and shared libraries have
996 * only one copy in memory, at most, normally.
998 * For the non-reserved pages, page_count(page) denotes a reference count.
999 * page_count() == 0 means the page is free. page->lru is then used for
1000 * freelist management in the buddy allocator.
1001 * page_count() > 0 means the page has been allocated.
1003 * Pages are allocated by the slab allocator in order to provide memory
1004 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1005 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1006 * unless a particular usage is carefully commented. (the responsibility of
1007 * freeing the kmalloc memory is the caller's, of course).
1009 * A page may be used by anyone else who does a __get_free_page().
1010 * In this case, page_count still tracks the references, and should only
1011 * be used through the normal accessor functions. The top bits of page->flags
1012 * and page->virtual store page management information, but all other fields
1013 * are unused and could be used privately, carefully. The management of this
1014 * page is the responsibility of the one who allocated it, and those who have
1015 * subsequently been given references to it.
1017 * The other pages (we may call them "pagecache pages") are completely
1018 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1019 * The following discussion applies only to them.
1021 * A pagecache page contains an opaque `private' member, which belongs to the
1022 * page's address_space. Usually, this is the address of a circular list of
1023 * the page's disk buffers. PG_private must be set to tell the VM to call
1024 * into the filesystem to release these pages.
1026 * A page may belong to an inode's memory mapping. In this case, page->mapping
1027 * is the pointer to the inode, and page->index is the file offset of the page,
1028 * in units of PAGE_SIZE.
1030 * If pagecache pages are not associated with an inode, they are said to be
1031 * anonymous pages. These may become associated with the swapcache, and in that
1032 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1034 * In either case (swapcache or inode backed), the pagecache itself holds one
1035 * reference to the page. Setting PG_private should also increment the
1036 * refcount. The each user mapping also has a reference to the page.
1038 * The pagecache pages are stored in a per-mapping radix tree, which is
1039 * rooted at mapping->i_pages, and indexed by offset.
1040 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1041 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1043 * All pagecache pages may be subject to I/O:
1044 * - inode pages may need to be read from disk,
1045 * - inode pages which have been modified and are MAP_SHARED may need
1046 * to be written back to the inode on disk,
1047 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1048 * modified may need to be swapped out to swap space and (later) to be read
1053 * The zone field is never updated after free_area_init_core()
1054 * sets it, so none of the operations on it need to be atomic.
1057 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1058 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1059 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1060 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1061 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1062 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1065 * Define the bit shifts to access each section. For non-existent
1066 * sections we define the shift as 0; that plus a 0 mask ensures
1067 * the compiler will optimise away reference to them.
1069 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1070 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1071 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1072 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1073 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1075 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1076 #ifdef NODE_NOT_IN_PAGE_FLAGS
1077 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1078 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1079 SECTIONS_PGOFF : ZONES_PGOFF)
1081 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1082 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1083 NODES_PGOFF : ZONES_PGOFF)
1086 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1088 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1089 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1090 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1091 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1092 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1093 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1095 static inline enum zone_type page_zonenum(const struct page *page)
1097 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1098 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1101 static inline enum zone_type folio_zonenum(const struct folio *folio)
1103 return page_zonenum(&folio->page);
1106 #ifdef CONFIG_ZONE_DEVICE
1107 static inline bool is_zone_device_page(const struct page *page)
1109 return page_zonenum(page) == ZONE_DEVICE;
1111 extern void memmap_init_zone_device(struct zone *, unsigned long,
1112 unsigned long, struct dev_pagemap *);
1114 static inline bool is_zone_device_page(const struct page *page)
1120 static inline bool folio_is_zone_device(const struct folio *folio)
1122 return is_zone_device_page(&folio->page);
1125 static inline bool is_zone_movable_page(const struct page *page)
1127 return page_zonenum(page) == ZONE_MOVABLE;
1130 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1131 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1133 bool __put_devmap_managed_page(struct page *page);
1134 static inline bool put_devmap_managed_page(struct page *page)
1136 if (!static_branch_unlikely(&devmap_managed_key))
1138 if (!is_zone_device_page(page))
1140 return __put_devmap_managed_page(page);
1143 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1144 static inline bool put_devmap_managed_page(struct page *page)
1148 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1150 /* 127: arbitrary random number, small enough to assemble well */
1151 #define folio_ref_zero_or_close_to_overflow(folio) \
1152 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1155 * folio_get - Increment the reference count on a folio.
1156 * @folio: The folio.
1158 * Context: May be called in any context, as long as you know that
1159 * you have a refcount on the folio. If you do not already have one,
1160 * folio_try_get() may be the right interface for you to use.
1162 static inline void folio_get(struct folio *folio)
1164 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1165 folio_ref_inc(folio);
1168 static inline void get_page(struct page *page)
1170 folio_get(page_folio(page));
1173 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1175 static inline __must_check bool try_get_page(struct page *page)
1177 page = compound_head(page);
1178 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1185 * folio_put - Decrement the reference count on a folio.
1186 * @folio: The folio.
1188 * If the folio's reference count reaches zero, the memory will be
1189 * released back to the page allocator and may be used by another
1190 * allocation immediately. Do not access the memory or the struct folio
1191 * after calling folio_put() unless you can be sure that it wasn't the
1194 * Context: May be called in process or interrupt context, but not in NMI
1195 * context. May be called while holding a spinlock.
1197 static inline void folio_put(struct folio *folio)
1199 if (folio_put_testzero(folio))
1200 __put_page(&folio->page);
1204 * folio_put_refs - Reduce the reference count on a folio.
1205 * @folio: The folio.
1206 * @refs: The amount to subtract from the folio's reference count.
1208 * If the folio's reference count reaches zero, the memory will be
1209 * released back to the page allocator and may be used by another
1210 * allocation immediately. Do not access the memory or the struct folio
1211 * after calling folio_put_refs() unless you can be sure that these weren't
1212 * the last references.
1214 * Context: May be called in process or interrupt context, but not in NMI
1215 * context. May be called while holding a spinlock.
1217 static inline void folio_put_refs(struct folio *folio, int refs)
1219 if (folio_ref_sub_and_test(folio, refs))
1220 __put_page(&folio->page);
1223 static inline void put_page(struct page *page)
1225 struct folio *folio = page_folio(page);
1228 * For some devmap managed pages we need to catch refcount transition
1231 if (put_devmap_managed_page(&folio->page))
1237 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1238 * the page's refcount so that two separate items are tracked: the original page
1239 * reference count, and also a new count of how many pin_user_pages() calls were
1240 * made against the page. ("gup-pinned" is another term for the latter).
1242 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1243 * distinct from normal pages. As such, the unpin_user_page() call (and its
1244 * variants) must be used in order to release gup-pinned pages.
1248 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1249 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1250 * simpler, due to the fact that adding an even power of two to the page
1251 * refcount has the effect of using only the upper N bits, for the code that
1252 * counts up using the bias value. This means that the lower bits are left for
1253 * the exclusive use of the original code that increments and decrements by one
1254 * (or at least, by much smaller values than the bias value).
1256 * Of course, once the lower bits overflow into the upper bits (and this is
1257 * OK, because subtraction recovers the original values), then visual inspection
1258 * no longer suffices to directly view the separate counts. However, for normal
1259 * applications that don't have huge page reference counts, this won't be an
1262 * Locking: the lockless algorithm described in folio_try_get_rcu()
1263 * provides safe operation for get_user_pages(), page_mkclean() and
1264 * other calls that race to set up page table entries.
1266 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1268 void unpin_user_page(struct page *page);
1269 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1271 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1273 void unpin_user_pages(struct page **pages, unsigned long npages);
1275 static inline bool is_cow_mapping(vm_flags_t flags)
1277 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1280 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1281 #define SECTION_IN_PAGE_FLAGS
1285 * The identification function is mainly used by the buddy allocator for
1286 * determining if two pages could be buddies. We are not really identifying
1287 * the zone since we could be using the section number id if we do not have
1288 * node id available in page flags.
1289 * We only guarantee that it will return the same value for two combinable
1292 static inline int page_zone_id(struct page *page)
1294 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1297 #ifdef NODE_NOT_IN_PAGE_FLAGS
1298 extern int page_to_nid(const struct page *page);
1300 static inline int page_to_nid(const struct page *page)
1302 struct page *p = (struct page *)page;
1304 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1308 static inline int folio_nid(const struct folio *folio)
1310 return page_to_nid(&folio->page);
1313 #ifdef CONFIG_NUMA_BALANCING
1314 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1316 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1319 static inline int cpupid_to_pid(int cpupid)
1321 return cpupid & LAST__PID_MASK;
1324 static inline int cpupid_to_cpu(int cpupid)
1326 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1329 static inline int cpupid_to_nid(int cpupid)
1331 return cpu_to_node(cpupid_to_cpu(cpupid));
1334 static inline bool cpupid_pid_unset(int cpupid)
1336 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1339 static inline bool cpupid_cpu_unset(int cpupid)
1341 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1344 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1346 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1349 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1350 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1351 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1353 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1356 static inline int page_cpupid_last(struct page *page)
1358 return page->_last_cpupid;
1360 static inline void page_cpupid_reset_last(struct page *page)
1362 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1365 static inline int page_cpupid_last(struct page *page)
1367 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1370 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1372 static inline void page_cpupid_reset_last(struct page *page)
1374 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1376 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1377 #else /* !CONFIG_NUMA_BALANCING */
1378 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1380 return page_to_nid(page); /* XXX */
1383 static inline int page_cpupid_last(struct page *page)
1385 return page_to_nid(page); /* XXX */
1388 static inline int cpupid_to_nid(int cpupid)
1393 static inline int cpupid_to_pid(int cpupid)
1398 static inline int cpupid_to_cpu(int cpupid)
1403 static inline int cpu_pid_to_cpupid(int nid, int pid)
1408 static inline bool cpupid_pid_unset(int cpupid)
1413 static inline void page_cpupid_reset_last(struct page *page)
1417 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1421 #endif /* CONFIG_NUMA_BALANCING */
1423 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1426 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1427 * setting tags for all pages to native kernel tag value 0xff, as the default
1428 * value 0x00 maps to 0xff.
1431 static inline u8 page_kasan_tag(const struct page *page)
1435 if (kasan_enabled()) {
1436 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1443 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1445 unsigned long old_flags, flags;
1447 if (!kasan_enabled())
1451 old_flags = READ_ONCE(page->flags);
1454 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1455 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1456 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1459 static inline void page_kasan_tag_reset(struct page *page)
1461 if (kasan_enabled())
1462 page_kasan_tag_set(page, 0xff);
1465 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1467 static inline u8 page_kasan_tag(const struct page *page)
1472 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1473 static inline void page_kasan_tag_reset(struct page *page) { }
1475 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1477 static inline struct zone *page_zone(const struct page *page)
1479 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1482 static inline pg_data_t *page_pgdat(const struct page *page)
1484 return NODE_DATA(page_to_nid(page));
1487 static inline struct zone *folio_zone(const struct folio *folio)
1489 return page_zone(&folio->page);
1492 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1494 return page_pgdat(&folio->page);
1497 #ifdef SECTION_IN_PAGE_FLAGS
1498 static inline void set_page_section(struct page *page, unsigned long section)
1500 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1501 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1504 static inline unsigned long page_to_section(const struct page *page)
1506 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1511 * folio_pfn - Return the Page Frame Number of a folio.
1512 * @folio: The folio.
1514 * A folio may contain multiple pages. The pages have consecutive
1515 * Page Frame Numbers.
1517 * Return: The Page Frame Number of the first page in the folio.
1519 static inline unsigned long folio_pfn(struct folio *folio)
1521 return page_to_pfn(&folio->page);
1524 static inline atomic_t *folio_pincount_ptr(struct folio *folio)
1526 return &folio_page(folio, 1)->compound_pincount;
1530 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1531 * @folio: The folio.
1533 * This function checks if a folio has been pinned via a call to
1534 * a function in the pin_user_pages() family.
1536 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1537 * because it means "definitely not pinned for DMA", but true means "probably
1538 * pinned for DMA, but possibly a false positive due to having at least
1539 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1541 * False positives are OK, because: a) it's unlikely for a folio to
1542 * get that many refcounts, and b) all the callers of this routine are
1543 * expected to be able to deal gracefully with a false positive.
1545 * For large folios, the result will be exactly correct. That's because
1546 * we have more tracking data available: the compound_pincount is used
1547 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1549 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1551 * Return: True, if it is likely that the page has been "dma-pinned".
1552 * False, if the page is definitely not dma-pinned.
1554 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1556 if (folio_test_large(folio))
1557 return atomic_read(folio_pincount_ptr(folio)) > 0;
1560 * folio_ref_count() is signed. If that refcount overflows, then
1561 * folio_ref_count() returns a negative value, and callers will avoid
1562 * further incrementing the refcount.
1564 * Here, for that overflow case, use the sign bit to count a little
1565 * bit higher via unsigned math, and thus still get an accurate result.
1567 return ((unsigned int)folio_ref_count(folio)) >=
1568 GUP_PIN_COUNTING_BIAS;
1571 static inline bool page_maybe_dma_pinned(struct page *page)
1573 return folio_maybe_dma_pinned(page_folio(page));
1577 * This should most likely only be called during fork() to see whether we
1578 * should break the cow immediately for a page on the src mm.
1580 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1583 if (!is_cow_mapping(vma->vm_flags))
1586 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1589 return page_maybe_dma_pinned(page);
1592 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1593 #ifdef CONFIG_MIGRATION
1594 static inline bool is_pinnable_page(struct page *page)
1596 return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) ||
1597 is_zero_pfn(page_to_pfn(page));
1600 static inline bool is_pinnable_page(struct page *page)
1606 static inline bool folio_is_pinnable(struct folio *folio)
1608 return is_pinnable_page(&folio->page);
1611 static inline void set_page_zone(struct page *page, enum zone_type zone)
1613 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1614 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1617 static inline void set_page_node(struct page *page, unsigned long node)
1619 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1620 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1623 static inline void set_page_links(struct page *page, enum zone_type zone,
1624 unsigned long node, unsigned long pfn)
1626 set_page_zone(page, zone);
1627 set_page_node(page, node);
1628 #ifdef SECTION_IN_PAGE_FLAGS
1629 set_page_section(page, pfn_to_section_nr(pfn));
1634 * folio_nr_pages - The number of pages in the folio.
1635 * @folio: The folio.
1637 * Return: A positive power of two.
1639 static inline long folio_nr_pages(struct folio *folio)
1641 return compound_nr(&folio->page);
1645 * folio_next - Move to the next physical folio.
1646 * @folio: The folio we're currently operating on.
1648 * If you have physically contiguous memory which may span more than
1649 * one folio (eg a &struct bio_vec), use this function to move from one
1650 * folio to the next. Do not use it if the memory is only virtually
1651 * contiguous as the folios are almost certainly not adjacent to each
1652 * other. This is the folio equivalent to writing ``page++``.
1654 * Context: We assume that the folios are refcounted and/or locked at a
1655 * higher level and do not adjust the reference counts.
1656 * Return: The next struct folio.
1658 static inline struct folio *folio_next(struct folio *folio)
1660 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1664 * folio_shift - The size of the memory described by this folio.
1665 * @folio: The folio.
1667 * A folio represents a number of bytes which is a power-of-two in size.
1668 * This function tells you which power-of-two the folio is. See also
1669 * folio_size() and folio_order().
1671 * Context: The caller should have a reference on the folio to prevent
1672 * it from being split. It is not necessary for the folio to be locked.
1673 * Return: The base-2 logarithm of the size of this folio.
1675 static inline unsigned int folio_shift(struct folio *folio)
1677 return PAGE_SHIFT + folio_order(folio);
1681 * folio_size - The number of bytes in a folio.
1682 * @folio: The folio.
1684 * Context: The caller should have a reference on the folio to prevent
1685 * it from being split. It is not necessary for the folio to be locked.
1686 * Return: The number of bytes in this folio.
1688 static inline size_t folio_size(struct folio *folio)
1690 return PAGE_SIZE << folio_order(folio);
1693 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1694 static inline int arch_make_page_accessible(struct page *page)
1700 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1701 static inline int arch_make_folio_accessible(struct folio *folio)
1704 long i, nr = folio_nr_pages(folio);
1706 for (i = 0; i < nr; i++) {
1707 ret = arch_make_page_accessible(folio_page(folio, i));
1717 * Some inline functions in vmstat.h depend on page_zone()
1719 #include <linux/vmstat.h>
1721 static __always_inline void *lowmem_page_address(const struct page *page)
1723 return page_to_virt(page);
1726 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1727 #define HASHED_PAGE_VIRTUAL
1730 #if defined(WANT_PAGE_VIRTUAL)
1731 static inline void *page_address(const struct page *page)
1733 return page->virtual;
1735 static inline void set_page_address(struct page *page, void *address)
1737 page->virtual = address;
1739 #define page_address_init() do { } while(0)
1742 #if defined(HASHED_PAGE_VIRTUAL)
1743 void *page_address(const struct page *page);
1744 void set_page_address(struct page *page, void *virtual);
1745 void page_address_init(void);
1748 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1749 #define page_address(page) lowmem_page_address(page)
1750 #define set_page_address(page, address) do { } while(0)
1751 #define page_address_init() do { } while(0)
1754 static inline void *folio_address(const struct folio *folio)
1756 return page_address(&folio->page);
1759 extern void *page_rmapping(struct page *page);
1760 extern pgoff_t __page_file_index(struct page *page);
1763 * Return the pagecache index of the passed page. Regular pagecache pages
1764 * use ->index whereas swapcache pages use swp_offset(->private)
1766 static inline pgoff_t page_index(struct page *page)
1768 if (unlikely(PageSwapCache(page)))
1769 return __page_file_index(page);
1773 bool page_mapped(struct page *page);
1774 bool folio_mapped(struct folio *folio);
1777 * Return true only if the page has been allocated with
1778 * ALLOC_NO_WATERMARKS and the low watermark was not
1779 * met implying that the system is under some pressure.
1781 static inline bool page_is_pfmemalloc(const struct page *page)
1784 * lru.next has bit 1 set if the page is allocated from the
1785 * pfmemalloc reserves. Callers may simply overwrite it if
1786 * they do not need to preserve that information.
1788 return (uintptr_t)page->lru.next & BIT(1);
1792 * Only to be called by the page allocator on a freshly allocated
1795 static inline void set_page_pfmemalloc(struct page *page)
1797 page->lru.next = (void *)BIT(1);
1800 static inline void clear_page_pfmemalloc(struct page *page)
1802 page->lru.next = NULL;
1806 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1808 extern void pagefault_out_of_memory(void);
1810 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1811 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1812 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1815 * Flags passed to show_mem() and show_free_areas() to suppress output in
1818 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1820 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1823 extern bool can_do_mlock(void);
1825 static inline bool can_do_mlock(void) { return false; }
1827 extern int user_shm_lock(size_t, struct ucounts *);
1828 extern void user_shm_unlock(size_t, struct ucounts *);
1830 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1832 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1835 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1836 unsigned long size);
1837 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1838 unsigned long size);
1839 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1840 unsigned long start, unsigned long end);
1842 struct mmu_notifier_range;
1844 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1845 unsigned long end, unsigned long floor, unsigned long ceiling);
1847 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1848 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1849 struct mmu_notifier_range *range, pte_t **ptepp,
1850 pmd_t **pmdpp, spinlock_t **ptlp);
1851 int follow_pte(struct mm_struct *mm, unsigned long address,
1852 pte_t **ptepp, spinlock_t **ptlp);
1853 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1854 unsigned long *pfn);
1855 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1856 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1857 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1858 void *buf, int len, int write);
1860 extern void truncate_pagecache(struct inode *inode, loff_t new);
1861 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1862 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1863 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1864 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1867 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1868 unsigned long address, unsigned int flags,
1869 struct pt_regs *regs);
1870 extern int fixup_user_fault(struct mm_struct *mm,
1871 unsigned long address, unsigned int fault_flags,
1873 void unmap_mapping_pages(struct address_space *mapping,
1874 pgoff_t start, pgoff_t nr, bool even_cows);
1875 void unmap_mapping_range(struct address_space *mapping,
1876 loff_t const holebegin, loff_t const holelen, int even_cows);
1878 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1879 unsigned long address, unsigned int flags,
1880 struct pt_regs *regs)
1882 /* should never happen if there's no MMU */
1884 return VM_FAULT_SIGBUS;
1886 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1887 unsigned int fault_flags, bool *unlocked)
1889 /* should never happen if there's no MMU */
1893 static inline void unmap_mapping_pages(struct address_space *mapping,
1894 pgoff_t start, pgoff_t nr, bool even_cows) { }
1895 static inline void unmap_mapping_range(struct address_space *mapping,
1896 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1899 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1900 loff_t const holebegin, loff_t const holelen)
1902 unmap_mapping_range(mapping, holebegin, holelen, 0);
1905 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1906 void *buf, int len, unsigned int gup_flags);
1907 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1908 void *buf, int len, unsigned int gup_flags);
1909 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1910 void *buf, int len, unsigned int gup_flags);
1912 long get_user_pages_remote(struct mm_struct *mm,
1913 unsigned long start, unsigned long nr_pages,
1914 unsigned int gup_flags, struct page **pages,
1915 struct vm_area_struct **vmas, int *locked);
1916 long pin_user_pages_remote(struct mm_struct *mm,
1917 unsigned long start, unsigned long nr_pages,
1918 unsigned int gup_flags, struct page **pages,
1919 struct vm_area_struct **vmas, int *locked);
1920 long get_user_pages(unsigned long start, unsigned long nr_pages,
1921 unsigned int gup_flags, struct page **pages,
1922 struct vm_area_struct **vmas);
1923 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1924 unsigned int gup_flags, struct page **pages,
1925 struct vm_area_struct **vmas);
1926 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1927 struct page **pages, unsigned int gup_flags);
1928 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1929 struct page **pages, unsigned int gup_flags);
1931 int get_user_pages_fast(unsigned long start, int nr_pages,
1932 unsigned int gup_flags, struct page **pages);
1933 int pin_user_pages_fast(unsigned long start, int nr_pages,
1934 unsigned int gup_flags, struct page **pages);
1936 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1937 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1938 struct task_struct *task, bool bypass_rlim);
1941 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1942 struct page **pages);
1943 struct page *get_dump_page(unsigned long addr);
1945 bool folio_mark_dirty(struct folio *folio);
1946 bool set_page_dirty(struct page *page);
1947 int set_page_dirty_lock(struct page *page);
1949 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1951 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1952 unsigned long old_addr, struct vm_area_struct *new_vma,
1953 unsigned long new_addr, unsigned long len,
1954 bool need_rmap_locks);
1957 * Flags used by change_protection(). For now we make it a bitmap so
1958 * that we can pass in multiple flags just like parameters. However
1959 * for now all the callers are only use one of the flags at the same
1962 /* Whether we should allow dirty bit accounting */
1963 #define MM_CP_DIRTY_ACCT (1UL << 0)
1964 /* Whether this protection change is for NUMA hints */
1965 #define MM_CP_PROT_NUMA (1UL << 1)
1966 /* Whether this change is for write protecting */
1967 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1968 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1969 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1970 MM_CP_UFFD_WP_RESOLVE)
1972 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1973 unsigned long end, pgprot_t newprot,
1974 unsigned long cp_flags);
1975 extern int mprotect_fixup(struct vm_area_struct *vma,
1976 struct vm_area_struct **pprev, unsigned long start,
1977 unsigned long end, unsigned long newflags);
1980 * doesn't attempt to fault and will return short.
1982 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1983 unsigned int gup_flags, struct page **pages);
1984 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1985 unsigned int gup_flags, struct page **pages);
1987 static inline bool get_user_page_fast_only(unsigned long addr,
1988 unsigned int gup_flags, struct page **pagep)
1990 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1993 * per-process(per-mm_struct) statistics.
1995 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1997 long val = atomic_long_read(&mm->rss_stat.count[member]);
1999 #ifdef SPLIT_RSS_COUNTING
2001 * counter is updated in asynchronous manner and may go to minus.
2002 * But it's never be expected number for users.
2007 return (unsigned long)val;
2010 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
2012 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2014 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
2016 mm_trace_rss_stat(mm, member, count);
2019 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2021 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
2023 mm_trace_rss_stat(mm, member, count);
2026 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2028 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
2030 mm_trace_rss_stat(mm, member, count);
2033 /* Optimized variant when page is already known not to be PageAnon */
2034 static inline int mm_counter_file(struct page *page)
2036 if (PageSwapBacked(page))
2037 return MM_SHMEMPAGES;
2038 return MM_FILEPAGES;
2041 static inline int mm_counter(struct page *page)
2044 return MM_ANONPAGES;
2045 return mm_counter_file(page);
2048 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2050 return get_mm_counter(mm, MM_FILEPAGES) +
2051 get_mm_counter(mm, MM_ANONPAGES) +
2052 get_mm_counter(mm, MM_SHMEMPAGES);
2055 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2057 return max(mm->hiwater_rss, get_mm_rss(mm));
2060 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2062 return max(mm->hiwater_vm, mm->total_vm);
2065 static inline void update_hiwater_rss(struct mm_struct *mm)
2067 unsigned long _rss = get_mm_rss(mm);
2069 if ((mm)->hiwater_rss < _rss)
2070 (mm)->hiwater_rss = _rss;
2073 static inline void update_hiwater_vm(struct mm_struct *mm)
2075 if (mm->hiwater_vm < mm->total_vm)
2076 mm->hiwater_vm = mm->total_vm;
2079 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2081 mm->hiwater_rss = get_mm_rss(mm);
2084 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2085 struct mm_struct *mm)
2087 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2089 if (*maxrss < hiwater_rss)
2090 *maxrss = hiwater_rss;
2093 #if defined(SPLIT_RSS_COUNTING)
2094 void sync_mm_rss(struct mm_struct *mm);
2096 static inline void sync_mm_rss(struct mm_struct *mm)
2101 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2102 static inline int pte_special(pte_t pte)
2107 static inline pte_t pte_mkspecial(pte_t pte)
2113 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2114 static inline int pte_devmap(pte_t pte)
2120 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2122 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2124 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2128 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2132 #ifdef __PAGETABLE_P4D_FOLDED
2133 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2134 unsigned long address)
2139 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2142 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2143 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2144 unsigned long address)
2148 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2149 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2152 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2154 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2156 if (mm_pud_folded(mm))
2158 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2161 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2163 if (mm_pud_folded(mm))
2165 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2169 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2170 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2171 unsigned long address)
2176 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2177 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2180 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2182 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2184 if (mm_pmd_folded(mm))
2186 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2189 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2191 if (mm_pmd_folded(mm))
2193 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2198 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2200 atomic_long_set(&mm->pgtables_bytes, 0);
2203 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2205 return atomic_long_read(&mm->pgtables_bytes);
2208 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2210 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2213 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2215 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2219 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2220 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2225 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2226 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2229 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2230 int __pte_alloc_kernel(pmd_t *pmd);
2232 #if defined(CONFIG_MMU)
2234 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2235 unsigned long address)
2237 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2238 NULL : p4d_offset(pgd, address);
2241 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2242 unsigned long address)
2244 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2245 NULL : pud_offset(p4d, address);
2248 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2250 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2251 NULL: pmd_offset(pud, address);
2253 #endif /* CONFIG_MMU */
2255 #if USE_SPLIT_PTE_PTLOCKS
2256 #if ALLOC_SPLIT_PTLOCKS
2257 void __init ptlock_cache_init(void);
2258 extern bool ptlock_alloc(struct page *page);
2259 extern void ptlock_free(struct page *page);
2261 static inline spinlock_t *ptlock_ptr(struct page *page)
2265 #else /* ALLOC_SPLIT_PTLOCKS */
2266 static inline void ptlock_cache_init(void)
2270 static inline bool ptlock_alloc(struct page *page)
2275 static inline void ptlock_free(struct page *page)
2279 static inline spinlock_t *ptlock_ptr(struct page *page)
2283 #endif /* ALLOC_SPLIT_PTLOCKS */
2285 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2287 return ptlock_ptr(pmd_page(*pmd));
2290 static inline bool ptlock_init(struct page *page)
2293 * prep_new_page() initialize page->private (and therefore page->ptl)
2294 * with 0. Make sure nobody took it in use in between.
2296 * It can happen if arch try to use slab for page table allocation:
2297 * slab code uses page->slab_cache, which share storage with page->ptl.
2299 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2300 if (!ptlock_alloc(page))
2302 spin_lock_init(ptlock_ptr(page));
2306 #else /* !USE_SPLIT_PTE_PTLOCKS */
2308 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2310 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2312 return &mm->page_table_lock;
2314 static inline void ptlock_cache_init(void) {}
2315 static inline bool ptlock_init(struct page *page) { return true; }
2316 static inline void ptlock_free(struct page *page) {}
2317 #endif /* USE_SPLIT_PTE_PTLOCKS */
2319 static inline void pgtable_init(void)
2321 ptlock_cache_init();
2322 pgtable_cache_init();
2325 static inline bool pgtable_pte_page_ctor(struct page *page)
2327 if (!ptlock_init(page))
2329 __SetPageTable(page);
2330 inc_lruvec_page_state(page, NR_PAGETABLE);
2334 static inline void pgtable_pte_page_dtor(struct page *page)
2337 __ClearPageTable(page);
2338 dec_lruvec_page_state(page, NR_PAGETABLE);
2341 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2343 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2344 pte_t *__pte = pte_offset_map(pmd, address); \
2350 #define pte_unmap_unlock(pte, ptl) do { \
2355 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2357 #define pte_alloc_map(mm, pmd, address) \
2358 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2360 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2361 (pte_alloc(mm, pmd) ? \
2362 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2364 #define pte_alloc_kernel(pmd, address) \
2365 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2366 NULL: pte_offset_kernel(pmd, address))
2368 #if USE_SPLIT_PMD_PTLOCKS
2370 static struct page *pmd_to_page(pmd_t *pmd)
2372 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2373 return virt_to_page((void *)((unsigned long) pmd & mask));
2376 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2378 return ptlock_ptr(pmd_to_page(pmd));
2381 static inline bool pmd_ptlock_init(struct page *page)
2383 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2384 page->pmd_huge_pte = NULL;
2386 return ptlock_init(page);
2389 static inline void pmd_ptlock_free(struct page *page)
2391 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2392 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2397 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2401 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2403 return &mm->page_table_lock;
2406 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2407 static inline void pmd_ptlock_free(struct page *page) {}
2409 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2413 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2415 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2420 static inline bool pgtable_pmd_page_ctor(struct page *page)
2422 if (!pmd_ptlock_init(page))
2424 __SetPageTable(page);
2425 inc_lruvec_page_state(page, NR_PAGETABLE);
2429 static inline void pgtable_pmd_page_dtor(struct page *page)
2431 pmd_ptlock_free(page);
2432 __ClearPageTable(page);
2433 dec_lruvec_page_state(page, NR_PAGETABLE);
2437 * No scalability reason to split PUD locks yet, but follow the same pattern
2438 * as the PMD locks to make it easier if we decide to. The VM should not be
2439 * considered ready to switch to split PUD locks yet; there may be places
2440 * which need to be converted from page_table_lock.
2442 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2444 return &mm->page_table_lock;
2447 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2449 spinlock_t *ptl = pud_lockptr(mm, pud);
2455 extern void __init pagecache_init(void);
2456 extern void free_initmem(void);
2459 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2460 * into the buddy system. The freed pages will be poisoned with pattern
2461 * "poison" if it's within range [0, UCHAR_MAX].
2462 * Return pages freed into the buddy system.
2464 extern unsigned long free_reserved_area(void *start, void *end,
2465 int poison, const char *s);
2467 extern void adjust_managed_page_count(struct page *page, long count);
2468 extern void mem_init_print_info(void);
2470 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2472 /* Free the reserved page into the buddy system, so it gets managed. */
2473 static inline void free_reserved_page(struct page *page)
2475 ClearPageReserved(page);
2476 init_page_count(page);
2478 adjust_managed_page_count(page, 1);
2480 #define free_highmem_page(page) free_reserved_page(page)
2482 static inline void mark_page_reserved(struct page *page)
2484 SetPageReserved(page);
2485 adjust_managed_page_count(page, -1);
2489 * Default method to free all the __init memory into the buddy system.
2490 * The freed pages will be poisoned with pattern "poison" if it's within
2491 * range [0, UCHAR_MAX].
2492 * Return pages freed into the buddy system.
2494 static inline unsigned long free_initmem_default(int poison)
2496 extern char __init_begin[], __init_end[];
2498 return free_reserved_area(&__init_begin, &__init_end,
2499 poison, "unused kernel image (initmem)");
2502 static inline unsigned long get_num_physpages(void)
2505 unsigned long phys_pages = 0;
2507 for_each_online_node(nid)
2508 phys_pages += node_present_pages(nid);
2514 * Using memblock node mappings, an architecture may initialise its
2515 * zones, allocate the backing mem_map and account for memory holes in an
2516 * architecture independent manner.
2518 * An architecture is expected to register range of page frames backed by
2519 * physical memory with memblock_add[_node]() before calling
2520 * free_area_init() passing in the PFN each zone ends at. At a basic
2521 * usage, an architecture is expected to do something like
2523 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2525 * for_each_valid_physical_page_range()
2526 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2527 * free_area_init(max_zone_pfns);
2529 void free_area_init(unsigned long *max_zone_pfn);
2530 unsigned long node_map_pfn_alignment(void);
2531 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2532 unsigned long end_pfn);
2533 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2534 unsigned long end_pfn);
2535 extern void get_pfn_range_for_nid(unsigned int nid,
2536 unsigned long *start_pfn, unsigned long *end_pfn);
2537 extern unsigned long find_min_pfn_with_active_regions(void);
2540 static inline int early_pfn_to_nid(unsigned long pfn)
2545 /* please see mm/page_alloc.c */
2546 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2549 extern void set_dma_reserve(unsigned long new_dma_reserve);
2550 extern void memmap_init_range(unsigned long, int, unsigned long,
2551 unsigned long, unsigned long, enum meminit_context,
2552 struct vmem_altmap *, int migratetype);
2553 extern void setup_per_zone_wmarks(void);
2554 extern void calculate_min_free_kbytes(void);
2555 extern int __meminit init_per_zone_wmark_min(void);
2556 extern void mem_init(void);
2557 extern void __init mmap_init(void);
2558 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2559 extern long si_mem_available(void);
2560 extern void si_meminfo(struct sysinfo * val);
2561 extern void si_meminfo_node(struct sysinfo *val, int nid);
2562 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2563 extern unsigned long arch_reserved_kernel_pages(void);
2566 extern __printf(3, 4)
2567 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2569 extern void setup_per_cpu_pageset(void);
2572 extern int min_free_kbytes;
2573 extern int watermark_boost_factor;
2574 extern int watermark_scale_factor;
2575 extern bool arch_has_descending_max_zone_pfns(void);
2578 extern atomic_long_t mmap_pages_allocated;
2579 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2581 /* interval_tree.c */
2582 void vma_interval_tree_insert(struct vm_area_struct *node,
2583 struct rb_root_cached *root);
2584 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2585 struct vm_area_struct *prev,
2586 struct rb_root_cached *root);
2587 void vma_interval_tree_remove(struct vm_area_struct *node,
2588 struct rb_root_cached *root);
2589 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2590 unsigned long start, unsigned long last);
2591 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2592 unsigned long start, unsigned long last);
2594 #define vma_interval_tree_foreach(vma, root, start, last) \
2595 for (vma = vma_interval_tree_iter_first(root, start, last); \
2596 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2598 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2599 struct rb_root_cached *root);
2600 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2601 struct rb_root_cached *root);
2602 struct anon_vma_chain *
2603 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2604 unsigned long start, unsigned long last);
2605 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2606 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2607 #ifdef CONFIG_DEBUG_VM_RB
2608 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2611 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2612 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2613 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2616 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2617 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2618 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2619 struct vm_area_struct *expand);
2620 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2621 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2623 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2625 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2626 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2627 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2628 struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2629 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2630 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2631 unsigned long addr, int new_below);
2632 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2633 unsigned long addr, int new_below);
2634 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2635 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2636 struct rb_node **, struct rb_node *);
2637 extern void unlink_file_vma(struct vm_area_struct *);
2638 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2639 unsigned long addr, unsigned long len, pgoff_t pgoff,
2640 bool *need_rmap_locks);
2641 extern void exit_mmap(struct mm_struct *);
2643 static inline int check_data_rlimit(unsigned long rlim,
2645 unsigned long start,
2646 unsigned long end_data,
2647 unsigned long start_data)
2649 if (rlim < RLIM_INFINITY) {
2650 if (((new - start) + (end_data - start_data)) > rlim)
2657 extern int mm_take_all_locks(struct mm_struct *mm);
2658 extern void mm_drop_all_locks(struct mm_struct *mm);
2660 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2661 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2662 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2663 extern struct file *get_task_exe_file(struct task_struct *task);
2665 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2666 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2668 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2669 const struct vm_special_mapping *sm);
2670 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2671 unsigned long addr, unsigned long len,
2672 unsigned long flags,
2673 const struct vm_special_mapping *spec);
2674 /* This is an obsolete alternative to _install_special_mapping. */
2675 extern int install_special_mapping(struct mm_struct *mm,
2676 unsigned long addr, unsigned long len,
2677 unsigned long flags, struct page **pages);
2679 unsigned long randomize_stack_top(unsigned long stack_top);
2681 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2683 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2684 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2685 struct list_head *uf);
2686 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2687 unsigned long len, unsigned long prot, unsigned long flags,
2688 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2689 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2690 struct list_head *uf, bool downgrade);
2691 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2692 struct list_head *uf);
2693 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2696 extern int __mm_populate(unsigned long addr, unsigned long len,
2698 static inline void mm_populate(unsigned long addr, unsigned long len)
2701 (void) __mm_populate(addr, len, 1);
2704 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2707 /* These take the mm semaphore themselves */
2708 extern int __must_check vm_brk(unsigned long, unsigned long);
2709 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2710 extern int vm_munmap(unsigned long, size_t);
2711 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2712 unsigned long, unsigned long,
2713 unsigned long, unsigned long);
2715 struct vm_unmapped_area_info {
2716 #define VM_UNMAPPED_AREA_TOPDOWN 1
2717 unsigned long flags;
2718 unsigned long length;
2719 unsigned long low_limit;
2720 unsigned long high_limit;
2721 unsigned long align_mask;
2722 unsigned long align_offset;
2725 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2728 extern void truncate_inode_pages(struct address_space *, loff_t);
2729 extern void truncate_inode_pages_range(struct address_space *,
2730 loff_t lstart, loff_t lend);
2731 extern void truncate_inode_pages_final(struct address_space *);
2733 /* generic vm_area_ops exported for stackable file systems */
2734 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2735 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2736 pgoff_t start_pgoff, pgoff_t end_pgoff);
2737 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2739 extern unsigned long stack_guard_gap;
2740 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2741 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2743 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2744 extern int expand_downwards(struct vm_area_struct *vma,
2745 unsigned long address);
2747 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2749 #define expand_upwards(vma, address) (0)
2752 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2753 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2754 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2755 struct vm_area_struct **pprev);
2758 * find_vma_intersection() - Look up the first VMA which intersects the interval
2759 * @mm: The process address space.
2760 * @start_addr: The inclusive start user address.
2761 * @end_addr: The exclusive end user address.
2763 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2764 * start_addr < end_addr.
2767 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2768 unsigned long start_addr,
2769 unsigned long end_addr)
2771 struct vm_area_struct *vma = find_vma(mm, start_addr);
2773 if (vma && end_addr <= vma->vm_start)
2779 * vma_lookup() - Find a VMA at a specific address
2780 * @mm: The process address space.
2781 * @addr: The user address.
2783 * Return: The vm_area_struct at the given address, %NULL otherwise.
2786 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2788 struct vm_area_struct *vma = find_vma(mm, addr);
2790 if (vma && addr < vma->vm_start)
2796 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2798 unsigned long vm_start = vma->vm_start;
2800 if (vma->vm_flags & VM_GROWSDOWN) {
2801 vm_start -= stack_guard_gap;
2802 if (vm_start > vma->vm_start)
2808 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2810 unsigned long vm_end = vma->vm_end;
2812 if (vma->vm_flags & VM_GROWSUP) {
2813 vm_end += stack_guard_gap;
2814 if (vm_end < vma->vm_end)
2815 vm_end = -PAGE_SIZE;
2820 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2822 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2825 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2826 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2827 unsigned long vm_start, unsigned long vm_end)
2829 struct vm_area_struct *vma = find_vma(mm, vm_start);
2831 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2837 static inline bool range_in_vma(struct vm_area_struct *vma,
2838 unsigned long start, unsigned long end)
2840 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2844 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2845 void vma_set_page_prot(struct vm_area_struct *vma);
2847 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2851 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2853 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2857 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2859 #ifdef CONFIG_NUMA_BALANCING
2860 unsigned long change_prot_numa(struct vm_area_struct *vma,
2861 unsigned long start, unsigned long end);
2864 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2865 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2866 unsigned long pfn, unsigned long size, pgprot_t);
2867 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2868 unsigned long pfn, unsigned long size, pgprot_t prot);
2869 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2870 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2871 struct page **pages, unsigned long *num);
2872 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2874 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2876 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2878 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2879 unsigned long pfn, pgprot_t pgprot);
2880 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2882 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2883 pfn_t pfn, pgprot_t pgprot);
2884 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2885 unsigned long addr, pfn_t pfn);
2886 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2888 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2889 unsigned long addr, struct page *page)
2891 int err = vm_insert_page(vma, addr, page);
2894 return VM_FAULT_OOM;
2895 if (err < 0 && err != -EBUSY)
2896 return VM_FAULT_SIGBUS;
2898 return VM_FAULT_NOPAGE;
2901 #ifndef io_remap_pfn_range
2902 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2903 unsigned long addr, unsigned long pfn,
2904 unsigned long size, pgprot_t prot)
2906 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2910 static inline vm_fault_t vmf_error(int err)
2913 return VM_FAULT_OOM;
2914 return VM_FAULT_SIGBUS;
2917 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2918 unsigned int foll_flags);
2920 #define FOLL_WRITE 0x01 /* check pte is writable */
2921 #define FOLL_TOUCH 0x02 /* mark page accessed */
2922 #define FOLL_GET 0x04 /* do get_page on page */
2923 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2924 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2925 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2926 * and return without waiting upon it */
2927 #define FOLL_NOFAULT 0x80 /* do not fault in pages */
2928 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2929 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2930 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2931 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2932 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2933 #define FOLL_COW 0x4000 /* internal GUP flag */
2934 #define FOLL_ANON 0x8000 /* don't do file mappings */
2935 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2936 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2937 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2938 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2941 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2942 * other. Here is what they mean, and how to use them:
2944 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2945 * period _often_ under userspace control. This is in contrast to
2946 * iov_iter_get_pages(), whose usages are transient.
2948 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2949 * lifetime enforced by the filesystem and we need guarantees that longterm
2950 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2951 * the filesystem. Ideas for this coordination include revoking the longterm
2952 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2953 * added after the problem with filesystems was found FS DAX VMAs are
2954 * specifically failed. Filesystem pages are still subject to bugs and use of
2955 * FOLL_LONGTERM should be avoided on those pages.
2957 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2958 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2959 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2960 * is due to an incompatibility with the FS DAX check and
2961 * FAULT_FLAG_ALLOW_RETRY.
2963 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2964 * that region. And so, CMA attempts to migrate the page before pinning, when
2965 * FOLL_LONGTERM is specified.
2967 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2968 * but an additional pin counting system) will be invoked. This is intended for
2969 * anything that gets a page reference and then touches page data (for example,
2970 * Direct IO). This lets the filesystem know that some non-file-system entity is
2971 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2972 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2973 * a call to unpin_user_page().
2975 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2976 * and separate refcounting mechanisms, however, and that means that each has
2977 * its own acquire and release mechanisms:
2979 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2981 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2983 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2984 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2985 * calls applied to them, and that's perfectly OK. This is a constraint on the
2986 * callers, not on the pages.)
2988 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2989 * directly by the caller. That's in order to help avoid mismatches when
2990 * releasing pages: get_user_pages*() pages must be released via put_page(),
2991 * while pin_user_pages*() pages must be released via unpin_user_page().
2993 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2996 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2998 if (vm_fault & VM_FAULT_OOM)
3000 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3001 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3002 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3007 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3008 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3009 unsigned long size, pte_fn_t fn, void *data);
3010 extern int apply_to_existing_page_range(struct mm_struct *mm,
3011 unsigned long address, unsigned long size,
3012 pte_fn_t fn, void *data);
3014 extern void init_mem_debugging_and_hardening(void);
3015 #ifdef CONFIG_PAGE_POISONING
3016 extern void __kernel_poison_pages(struct page *page, int numpages);
3017 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3018 extern bool _page_poisoning_enabled_early;
3019 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3020 static inline bool page_poisoning_enabled(void)
3022 return _page_poisoning_enabled_early;
3025 * For use in fast paths after init_mem_debugging() has run, or when a
3026 * false negative result is not harmful when called too early.
3028 static inline bool page_poisoning_enabled_static(void)
3030 return static_branch_unlikely(&_page_poisoning_enabled);
3032 static inline void kernel_poison_pages(struct page *page, int numpages)
3034 if (page_poisoning_enabled_static())
3035 __kernel_poison_pages(page, numpages);
3037 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3039 if (page_poisoning_enabled_static())
3040 __kernel_unpoison_pages(page, numpages);
3043 static inline bool page_poisoning_enabled(void) { return false; }
3044 static inline bool page_poisoning_enabled_static(void) { return false; }
3045 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3046 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3047 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3050 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3051 static inline bool want_init_on_alloc(gfp_t flags)
3053 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3056 return flags & __GFP_ZERO;
3059 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3060 static inline bool want_init_on_free(void)
3062 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3066 extern bool _debug_pagealloc_enabled_early;
3067 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3069 static inline bool debug_pagealloc_enabled(void)
3071 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3072 _debug_pagealloc_enabled_early;
3076 * For use in fast paths after init_debug_pagealloc() has run, or when a
3077 * false negative result is not harmful when called too early.
3079 static inline bool debug_pagealloc_enabled_static(void)
3081 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3084 return static_branch_unlikely(&_debug_pagealloc_enabled);
3087 #ifdef CONFIG_DEBUG_PAGEALLOC
3089 * To support DEBUG_PAGEALLOC architecture must ensure that
3090 * __kernel_map_pages() never fails
3092 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3094 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3096 if (debug_pagealloc_enabled_static())
3097 __kernel_map_pages(page, numpages, 1);
3100 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3102 if (debug_pagealloc_enabled_static())
3103 __kernel_map_pages(page, numpages, 0);
3105 #else /* CONFIG_DEBUG_PAGEALLOC */
3106 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3107 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3108 #endif /* CONFIG_DEBUG_PAGEALLOC */
3110 #ifdef __HAVE_ARCH_GATE_AREA
3111 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3112 extern int in_gate_area_no_mm(unsigned long addr);
3113 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3115 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3119 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3120 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3124 #endif /* __HAVE_ARCH_GATE_AREA */
3126 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3128 #ifdef CONFIG_SYSCTL
3129 extern int sysctl_drop_caches;
3130 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3134 void drop_slab(void);
3137 #define randomize_va_space 0
3139 extern int randomize_va_space;
3142 const char * arch_vma_name(struct vm_area_struct *vma);
3144 void print_vma_addr(char *prefix, unsigned long rip);
3146 static inline void print_vma_addr(char *prefix, unsigned long rip)
3151 #ifdef CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
3152 int vmemmap_remap_free(unsigned long start, unsigned long end,
3153 unsigned long reuse);
3154 int vmemmap_remap_alloc(unsigned long start, unsigned long end,
3155 unsigned long reuse, gfp_t gfp_mask);
3158 void *sparse_buffer_alloc(unsigned long size);
3159 struct page * __populate_section_memmap(unsigned long pfn,
3160 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3161 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3162 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3163 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3164 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3165 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3166 struct vmem_altmap *altmap);
3167 void *vmemmap_alloc_block(unsigned long size, int node);
3169 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3170 struct vmem_altmap *altmap);
3171 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3172 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3173 int node, struct vmem_altmap *altmap);
3174 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3175 struct vmem_altmap *altmap);
3176 void vmemmap_populate_print_last(void);
3177 #ifdef CONFIG_MEMORY_HOTPLUG
3178 void vmemmap_free(unsigned long start, unsigned long end,
3179 struct vmem_altmap *altmap);
3181 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3182 unsigned long nr_pages);
3185 MF_COUNT_INCREASED = 1 << 0,
3186 MF_ACTION_REQUIRED = 1 << 1,
3187 MF_MUST_KILL = 1 << 2,
3188 MF_SOFT_OFFLINE = 1 << 3,
3189 MF_UNPOISON = 1 << 4,
3191 extern int memory_failure(unsigned long pfn, int flags);
3192 extern void memory_failure_queue(unsigned long pfn, int flags);
3193 extern void memory_failure_queue_kick(int cpu);
3194 extern int unpoison_memory(unsigned long pfn);
3195 extern int sysctl_memory_failure_early_kill;
3196 extern int sysctl_memory_failure_recovery;
3197 extern void shake_page(struct page *p);
3198 extern atomic_long_t num_poisoned_pages __read_mostly;
3199 extern int soft_offline_page(unsigned long pfn, int flags);
3200 #ifdef CONFIG_MEMORY_FAILURE
3201 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags);
3203 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags)
3209 #ifndef arch_memory_failure
3210 static inline int arch_memory_failure(unsigned long pfn, int flags)
3216 #ifndef arch_is_platform_page
3217 static inline bool arch_is_platform_page(u64 paddr)
3224 * Error handlers for various types of pages.
3227 MF_IGNORED, /* Error: cannot be handled */
3228 MF_FAILED, /* Error: handling failed */
3229 MF_DELAYED, /* Will be handled later */
3230 MF_RECOVERED, /* Successfully recovered */
3233 enum mf_action_page_type {
3235 MF_MSG_KERNEL_HIGH_ORDER,
3237 MF_MSG_DIFFERENT_COMPOUND,
3240 MF_MSG_NON_PMD_HUGE,
3241 MF_MSG_UNMAP_FAILED,
3242 MF_MSG_DIRTY_SWAPCACHE,
3243 MF_MSG_CLEAN_SWAPCACHE,
3244 MF_MSG_DIRTY_MLOCKED_LRU,
3245 MF_MSG_CLEAN_MLOCKED_LRU,
3246 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3247 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3250 MF_MSG_TRUNCATED_LRU,
3254 MF_MSG_DIFFERENT_PAGE_SIZE,
3258 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3259 extern void clear_huge_page(struct page *page,
3260 unsigned long addr_hint,
3261 unsigned int pages_per_huge_page);
3262 extern void copy_user_huge_page(struct page *dst, struct page *src,
3263 unsigned long addr_hint,
3264 struct vm_area_struct *vma,
3265 unsigned int pages_per_huge_page);
3266 extern long copy_huge_page_from_user(struct page *dst_page,
3267 const void __user *usr_src,
3268 unsigned int pages_per_huge_page,
3269 bool allow_pagefault);
3272 * vma_is_special_huge - Are transhuge page-table entries considered special?
3273 * @vma: Pointer to the struct vm_area_struct to consider
3275 * Whether transhuge page-table entries are considered "special" following
3276 * the definition in vm_normal_page().
3278 * Return: true if transhuge page-table entries should be considered special,
3281 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3283 return vma_is_dax(vma) || (vma->vm_file &&
3284 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3287 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3289 #ifdef CONFIG_DEBUG_PAGEALLOC
3290 extern unsigned int _debug_guardpage_minorder;
3291 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3293 static inline unsigned int debug_guardpage_minorder(void)
3295 return _debug_guardpage_minorder;
3298 static inline bool debug_guardpage_enabled(void)
3300 return static_branch_unlikely(&_debug_guardpage_enabled);
3303 static inline bool page_is_guard(struct page *page)
3305 if (!debug_guardpage_enabled())
3308 return PageGuard(page);
3311 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3312 static inline bool debug_guardpage_enabled(void) { return false; }
3313 static inline bool page_is_guard(struct page *page) { return false; }
3314 #endif /* CONFIG_DEBUG_PAGEALLOC */
3316 #if MAX_NUMNODES > 1
3317 void __init setup_nr_node_ids(void);
3319 static inline void setup_nr_node_ids(void) {}
3322 extern int memcmp_pages(struct page *page1, struct page *page2);
3324 static inline int pages_identical(struct page *page1, struct page *page2)
3326 return !memcmp_pages(page1, page2);
3329 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3330 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3331 pgoff_t first_index, pgoff_t nr,
3332 pgoff_t bitmap_pgoff,
3333 unsigned long *bitmap,
3337 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3338 pgoff_t first_index, pgoff_t nr);
3341 extern int sysctl_nr_trim_pages;
3343 #ifdef CONFIG_PRINTK
3344 void mem_dump_obj(void *object);
3346 static inline void mem_dump_obj(void *object) {}
3350 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3351 * @seals: the seals to check
3352 * @vma: the vma to operate on
3354 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3355 * the vma flags. Return 0 if check pass, or <0 for errors.
3357 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3359 if (seals & F_SEAL_FUTURE_WRITE) {
3361 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3362 * "future write" seal active.
3364 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3368 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3369 * MAP_SHARED and read-only, take care to not allow mprotect to
3370 * revert protections on such mappings. Do this only for shared
3371 * mappings. For private mappings, don't need to mask
3372 * VM_MAYWRITE as we still want them to be COW-writable.
3374 if (vma->vm_flags & VM_SHARED)
3375 vma->vm_flags &= ~(VM_MAYWRITE);
3381 #ifdef CONFIG_ANON_VMA_NAME
3382 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3383 unsigned long len_in,
3384 struct anon_vma_name *anon_name);
3387 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3388 unsigned long len_in, struct anon_vma_name *anon_name) {
3393 #endif /* _LINUX_MM_H */