1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
34 #include <linux/kasan.h>
38 struct anon_vma_chain;
42 extern int sysctl_page_lock_unfairness;
44 void init_mm_internals(void);
46 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
47 extern unsigned long max_mapnr;
49 static inline void set_max_mapnr(unsigned long limit)
54 static inline void set_max_mapnr(unsigned long limit) { }
57 extern atomic_long_t _totalram_pages;
58 static inline unsigned long totalram_pages(void)
60 return (unsigned long)atomic_long_read(&_totalram_pages);
63 static inline void totalram_pages_inc(void)
65 atomic_long_inc(&_totalram_pages);
68 static inline void totalram_pages_dec(void)
70 atomic_long_dec(&_totalram_pages);
73 static inline void totalram_pages_add(long count)
75 atomic_long_add(count, &_totalram_pages);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/processor.h>
102 * Architectures that support memory tagging (assigning tags to memory regions,
103 * embedding these tags into addresses that point to these memory regions, and
104 * checking that the memory and the pointer tags match on memory accesses)
105 * redefine this macro to strip tags from pointers.
106 * It's defined as noop for architectures that don't support memory tagging.
108 #ifndef untagged_addr
109 #define untagged_addr(addr) (addr)
113 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 #define lm_alias(x) __va(__pa_symbol(x))
125 * To prevent common memory management code establishing
126 * a zero page mapping on a read fault.
127 * This macro should be defined within <asm/pgtable.h>.
128 * s390 does this to prevent multiplexing of hardware bits
129 * related to the physical page in case of virtualization.
131 #ifndef mm_forbids_zeropage
132 #define mm_forbids_zeropage(X) (0)
136 * On some architectures it is expensive to call memset() for small sizes.
137 * If an architecture decides to implement their own version of
138 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
139 * define their own version of this macro in <asm/pgtable.h>
141 #if BITS_PER_LONG == 64
142 /* This function must be updated when the size of struct page grows above 80
143 * or reduces below 56. The idea that compiler optimizes out switch()
144 * statement, and only leaves move/store instructions. Also the compiler can
145 * combine write statements if they are both assignments and can be reordered,
146 * this can result in several of the writes here being dropped.
148 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
149 static inline void __mm_zero_struct_page(struct page *page)
151 unsigned long *_pp = (void *)page;
153 /* Check that struct page is either 56, 64, 72, or 80 bytes */
154 BUILD_BUG_ON(sizeof(struct page) & 7);
155 BUILD_BUG_ON(sizeof(struct page) < 56);
156 BUILD_BUG_ON(sizeof(struct page) > 80);
158 switch (sizeof(struct page)) {
179 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
183 * Default maximum number of active map areas, this limits the number of vmas
184 * per mm struct. Users can overwrite this number by sysctl but there is a
187 * When a program's coredump is generated as ELF format, a section is created
188 * per a vma. In ELF, the number of sections is represented in unsigned short.
189 * This means the number of sections should be smaller than 65535 at coredump.
190 * Because the kernel adds some informative sections to a image of program at
191 * generating coredump, we need some margin. The number of extra sections is
192 * 1-3 now and depends on arch. We use "5" as safe margin, here.
194 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
195 * not a hard limit any more. Although some userspace tools can be surprised by
198 #define MAPCOUNT_ELF_CORE_MARGIN (5)
199 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
201 extern int sysctl_max_map_count;
203 extern unsigned long sysctl_user_reserve_kbytes;
204 extern unsigned long sysctl_admin_reserve_kbytes;
206 extern int sysctl_overcommit_memory;
207 extern int sysctl_overcommit_ratio;
208 extern unsigned long sysctl_overcommit_kbytes;
210 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
212 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
214 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
217 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
218 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
220 #define nth_page(page,n) ((page) + (n))
223 /* to align the pointer to the (next) page boundary */
224 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
226 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
227 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
229 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
231 void setup_initial_init_mm(void *start_code, void *end_code,
232 void *end_data, void *brk);
235 * Linux kernel virtual memory manager primitives.
236 * The idea being to have a "virtual" mm in the same way
237 * we have a virtual fs - giving a cleaner interface to the
238 * mm details, and allowing different kinds of memory mappings
239 * (from shared memory to executable loading to arbitrary
243 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
244 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
245 void vm_area_free(struct vm_area_struct *);
248 extern struct rb_root nommu_region_tree;
249 extern struct rw_semaphore nommu_region_sem;
251 extern unsigned int kobjsize(const void *objp);
255 * vm_flags in vm_area_struct, see mm_types.h.
256 * When changing, update also include/trace/events/mmflags.h
258 #define VM_NONE 0x00000000
260 #define VM_READ 0x00000001 /* currently active flags */
261 #define VM_WRITE 0x00000002
262 #define VM_EXEC 0x00000004
263 #define VM_SHARED 0x00000008
265 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
266 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
267 #define VM_MAYWRITE 0x00000020
268 #define VM_MAYEXEC 0x00000040
269 #define VM_MAYSHARE 0x00000080
271 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
272 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
273 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
274 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
276 #define VM_LOCKED 0x00002000
277 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
279 /* Used by sys_madvise() */
280 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
281 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
283 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
284 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
285 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
286 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
287 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
288 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
289 #define VM_SYNC 0x00800000 /* Synchronous page faults */
290 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
291 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
292 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
294 #ifdef CONFIG_MEM_SOFT_DIRTY
295 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
297 # define VM_SOFTDIRTY 0
300 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
301 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
302 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
303 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
305 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
306 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
307 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
308 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
309 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
310 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
311 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
312 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
313 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
314 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
315 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
316 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
318 #ifdef CONFIG_ARCH_HAS_PKEYS
319 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
320 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
321 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
322 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
323 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
325 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
327 # define VM_PKEY_BIT4 0
329 #endif /* CONFIG_ARCH_HAS_PKEYS */
331 #if defined(CONFIG_X86)
332 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
333 #elif defined(CONFIG_PPC)
334 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
335 #elif defined(CONFIG_PARISC)
336 # define VM_GROWSUP VM_ARCH_1
337 #elif defined(CONFIG_IA64)
338 # define VM_GROWSUP VM_ARCH_1
339 #elif defined(CONFIG_SPARC64)
340 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
341 # define VM_ARCH_CLEAR VM_SPARC_ADI
342 #elif defined(CONFIG_ARM64)
343 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
344 # define VM_ARCH_CLEAR VM_ARM64_BTI
345 #elif !defined(CONFIG_MMU)
346 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
349 #if defined(CONFIG_ARM64_MTE)
350 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
351 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
353 # define VM_MTE VM_NONE
354 # define VM_MTE_ALLOWED VM_NONE
358 # define VM_GROWSUP VM_NONE
361 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
362 # define VM_UFFD_MINOR_BIT 37
363 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
364 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
365 # define VM_UFFD_MINOR VM_NONE
366 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
368 /* Bits set in the VMA until the stack is in its final location */
369 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
371 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
373 /* Common data flag combinations */
374 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
375 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
376 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
377 VM_MAYWRITE | VM_MAYEXEC)
378 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
379 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
381 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
382 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
385 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
386 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
389 #ifdef CONFIG_STACK_GROWSUP
390 #define VM_STACK VM_GROWSUP
392 #define VM_STACK VM_GROWSDOWN
395 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
397 /* VMA basic access permission flags */
398 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
402 * Special vmas that are non-mergable, non-mlock()able.
404 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
406 /* This mask prevents VMA from being scanned with khugepaged */
407 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
409 /* This mask defines which mm->def_flags a process can inherit its parent */
410 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
412 /* This mask is used to clear all the VMA flags used by mlock */
413 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
415 /* Arch-specific flags to clear when updating VM flags on protection change */
416 #ifndef VM_ARCH_CLEAR
417 # define VM_ARCH_CLEAR VM_NONE
419 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
422 * mapping from the currently active vm_flags protection bits (the
423 * low four bits) to a page protection mask..
425 extern pgprot_t protection_map[16];
428 * enum fault_flag - Fault flag definitions.
429 * @FAULT_FLAG_WRITE: Fault was a write fault.
430 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
431 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
432 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
433 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
434 * @FAULT_FLAG_TRIED: The fault has been tried once.
435 * @FAULT_FLAG_USER: The fault originated in userspace.
436 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
437 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
438 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
440 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
441 * whether we would allow page faults to retry by specifying these two
442 * fault flags correctly. Currently there can be three legal combinations:
444 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
445 * this is the first try
447 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
448 * we've already tried at least once
450 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
452 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
453 * be used. Note that page faults can be allowed to retry for multiple times,
454 * in which case we'll have an initial fault with flags (a) then later on
455 * continuous faults with flags (b). We should always try to detect pending
456 * signals before a retry to make sure the continuous page faults can still be
457 * interrupted if necessary.
460 FAULT_FLAG_WRITE = 1 << 0,
461 FAULT_FLAG_MKWRITE = 1 << 1,
462 FAULT_FLAG_ALLOW_RETRY = 1 << 2,
463 FAULT_FLAG_RETRY_NOWAIT = 1 << 3,
464 FAULT_FLAG_KILLABLE = 1 << 4,
465 FAULT_FLAG_TRIED = 1 << 5,
466 FAULT_FLAG_USER = 1 << 6,
467 FAULT_FLAG_REMOTE = 1 << 7,
468 FAULT_FLAG_INSTRUCTION = 1 << 8,
469 FAULT_FLAG_INTERRUPTIBLE = 1 << 9,
473 * The default fault flags that should be used by most of the
474 * arch-specific page fault handlers.
476 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
477 FAULT_FLAG_KILLABLE | \
478 FAULT_FLAG_INTERRUPTIBLE)
481 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
482 * @flags: Fault flags.
484 * This is mostly used for places where we want to try to avoid taking
485 * the mmap_lock for too long a time when waiting for another condition
486 * to change, in which case we can try to be polite to release the
487 * mmap_lock in the first round to avoid potential starvation of other
488 * processes that would also want the mmap_lock.
490 * Return: true if the page fault allows retry and this is the first
491 * attempt of the fault handling; false otherwise.
493 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
495 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
496 (!(flags & FAULT_FLAG_TRIED));
499 #define FAULT_FLAG_TRACE \
500 { FAULT_FLAG_WRITE, "WRITE" }, \
501 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
502 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
503 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
504 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
505 { FAULT_FLAG_TRIED, "TRIED" }, \
506 { FAULT_FLAG_USER, "USER" }, \
507 { FAULT_FLAG_REMOTE, "REMOTE" }, \
508 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
509 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
512 * vm_fault is filled by the pagefault handler and passed to the vma's
513 * ->fault function. The vma's ->fault is responsible for returning a bitmask
514 * of VM_FAULT_xxx flags that give details about how the fault was handled.
516 * MM layer fills up gfp_mask for page allocations but fault handler might
517 * alter it if its implementation requires a different allocation context.
519 * pgoff should be used in favour of virtual_address, if possible.
523 struct vm_area_struct *vma; /* Target VMA */
524 gfp_t gfp_mask; /* gfp mask to be used for allocations */
525 pgoff_t pgoff; /* Logical page offset based on vma */
526 unsigned long address; /* Faulting virtual address */
528 enum fault_flag flags; /* FAULT_FLAG_xxx flags
529 * XXX: should really be 'const' */
530 pmd_t *pmd; /* Pointer to pmd entry matching
532 pud_t *pud; /* Pointer to pud entry matching
536 pte_t orig_pte; /* Value of PTE at the time of fault */
537 pmd_t orig_pmd; /* Value of PMD at the time of fault,
538 * used by PMD fault only.
542 struct page *cow_page; /* Page handler may use for COW fault */
543 struct page *page; /* ->fault handlers should return a
544 * page here, unless VM_FAULT_NOPAGE
545 * is set (which is also implied by
548 /* These three entries are valid only while holding ptl lock */
549 pte_t *pte; /* Pointer to pte entry matching
550 * the 'address'. NULL if the page
551 * table hasn't been allocated.
553 spinlock_t *ptl; /* Page table lock.
554 * Protects pte page table if 'pte'
555 * is not NULL, otherwise pmd.
557 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
558 * vm_ops->map_pages() sets up a page
559 * table from atomic context.
560 * do_fault_around() pre-allocates
561 * page table to avoid allocation from
566 /* page entry size for vm->huge_fault() */
567 enum page_entry_size {
574 * These are the virtual MM functions - opening of an area, closing and
575 * unmapping it (needed to keep files on disk up-to-date etc), pointer
576 * to the functions called when a no-page or a wp-page exception occurs.
578 struct vm_operations_struct {
579 void (*open)(struct vm_area_struct * area);
580 void (*close)(struct vm_area_struct * area);
581 /* Called any time before splitting to check if it's allowed */
582 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
583 int (*mremap)(struct vm_area_struct *area);
585 * Called by mprotect() to make driver-specific permission
586 * checks before mprotect() is finalised. The VMA must not
587 * be modified. Returns 0 if eprotect() can proceed.
589 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
590 unsigned long end, unsigned long newflags);
591 vm_fault_t (*fault)(struct vm_fault *vmf);
592 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
593 enum page_entry_size pe_size);
594 vm_fault_t (*map_pages)(struct vm_fault *vmf,
595 pgoff_t start_pgoff, pgoff_t end_pgoff);
596 unsigned long (*pagesize)(struct vm_area_struct * area);
598 /* notification that a previously read-only page is about to become
599 * writable, if an error is returned it will cause a SIGBUS */
600 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
602 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
603 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
605 /* called by access_process_vm when get_user_pages() fails, typically
606 * for use by special VMAs. See also generic_access_phys() for a generic
607 * implementation useful for any iomem mapping.
609 int (*access)(struct vm_area_struct *vma, unsigned long addr,
610 void *buf, int len, int write);
612 /* Called by the /proc/PID/maps code to ask the vma whether it
613 * has a special name. Returning non-NULL will also cause this
614 * vma to be dumped unconditionally. */
615 const char *(*name)(struct vm_area_struct *vma);
619 * set_policy() op must add a reference to any non-NULL @new mempolicy
620 * to hold the policy upon return. Caller should pass NULL @new to
621 * remove a policy and fall back to surrounding context--i.e. do not
622 * install a MPOL_DEFAULT policy, nor the task or system default
625 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
628 * get_policy() op must add reference [mpol_get()] to any policy at
629 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
630 * in mm/mempolicy.c will do this automatically.
631 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
632 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
633 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
634 * must return NULL--i.e., do not "fallback" to task or system default
637 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
641 * Called by vm_normal_page() for special PTEs to find the
642 * page for @addr. This is useful if the default behavior
643 * (using pte_page()) would not find the correct page.
645 struct page *(*find_special_page)(struct vm_area_struct *vma,
649 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
651 static const struct vm_operations_struct dummy_vm_ops = {};
653 memset(vma, 0, sizeof(*vma));
655 vma->vm_ops = &dummy_vm_ops;
656 INIT_LIST_HEAD(&vma->anon_vma_chain);
659 static inline void vma_set_anonymous(struct vm_area_struct *vma)
664 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
669 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
671 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
676 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
677 VM_STACK_INCOMPLETE_SETUP)
683 static inline bool vma_is_foreign(struct vm_area_struct *vma)
688 if (current->mm != vma->vm_mm)
694 static inline bool vma_is_accessible(struct vm_area_struct *vma)
696 return vma->vm_flags & VM_ACCESS_FLAGS;
701 * The vma_is_shmem is not inline because it is used only by slow
702 * paths in userfault.
704 bool vma_is_shmem(struct vm_area_struct *vma);
706 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
709 int vma_is_stack_for_current(struct vm_area_struct *vma);
711 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
712 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
717 static inline unsigned int compound_order(struct page *page)
721 return page[1].compound_order;
725 * folio_order - The allocation order of a folio.
728 * A folio is composed of 2^order pages. See get_order() for the definition
731 * Return: The order of the folio.
733 static inline unsigned int folio_order(struct folio *folio)
735 return compound_order(&folio->page);
738 #include <linux/huge_mm.h>
741 * Methods to modify the page usage count.
743 * What counts for a page usage:
744 * - cache mapping (page->mapping)
745 * - private data (page->private)
746 * - page mapped in a task's page tables, each mapping
747 * is counted separately
749 * Also, many kernel routines increase the page count before a critical
750 * routine so they can be sure the page doesn't go away from under them.
754 * Drop a ref, return true if the refcount fell to zero (the page has no users)
756 static inline int put_page_testzero(struct page *page)
758 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
759 return page_ref_dec_and_test(page);
762 static inline int folio_put_testzero(struct folio *folio)
764 return put_page_testzero(&folio->page);
768 * Try to grab a ref unless the page has a refcount of zero, return false if
770 * This can be called when MMU is off so it must not access
771 * any of the virtual mappings.
773 static inline bool get_page_unless_zero(struct page *page)
775 return page_ref_add_unless(page, 1, 0);
778 extern int page_is_ram(unsigned long pfn);
786 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
789 /* Support for virtually mapped pages */
790 struct page *vmalloc_to_page(const void *addr);
791 unsigned long vmalloc_to_pfn(const void *addr);
794 * Determine if an address is within the vmalloc range
796 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
797 * is no special casing required.
800 #ifndef is_ioremap_addr
801 #define is_ioremap_addr(x) is_vmalloc_addr(x)
805 extern bool is_vmalloc_addr(const void *x);
806 extern int is_vmalloc_or_module_addr(const void *x);
808 static inline bool is_vmalloc_addr(const void *x)
812 static inline int is_vmalloc_or_module_addr(const void *x)
818 static inline int head_compound_mapcount(struct page *head)
820 return atomic_read(compound_mapcount_ptr(head)) + 1;
824 * Mapcount of compound page as a whole, does not include mapped sub-pages.
826 * Must be called only for compound pages or any their tail sub-pages.
828 static inline int compound_mapcount(struct page *page)
830 VM_BUG_ON_PAGE(!PageCompound(page), page);
831 page = compound_head(page);
832 return head_compound_mapcount(page);
836 * The atomic page->_mapcount, starts from -1: so that transitions
837 * both from it and to it can be tracked, using atomic_inc_and_test
838 * and atomic_add_negative(-1).
840 static inline void page_mapcount_reset(struct page *page)
842 atomic_set(&(page)->_mapcount, -1);
845 int __page_mapcount(struct page *page);
848 * Mapcount of 0-order page; when compound sub-page, includes
849 * compound_mapcount().
851 * Result is undefined for pages which cannot be mapped into userspace.
852 * For example SLAB or special types of pages. See function page_has_type().
853 * They use this place in struct page differently.
855 static inline int page_mapcount(struct page *page)
857 if (unlikely(PageCompound(page)))
858 return __page_mapcount(page);
859 return atomic_read(&page->_mapcount) + 1;
862 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
863 int total_mapcount(struct page *page);
864 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
866 static inline int total_mapcount(struct page *page)
868 return page_mapcount(page);
870 static inline int page_trans_huge_mapcount(struct page *page,
873 int mapcount = page_mapcount(page);
875 *total_mapcount = mapcount;
880 static inline struct page *virt_to_head_page(const void *x)
882 struct page *page = virt_to_page(x);
884 return compound_head(page);
887 static inline struct folio *virt_to_folio(const void *x)
889 struct page *page = virt_to_page(x);
891 return page_folio(page);
894 void __put_page(struct page *page);
896 void put_pages_list(struct list_head *pages);
898 void split_page(struct page *page, unsigned int order);
899 void folio_copy(struct folio *dst, struct folio *src);
901 unsigned long nr_free_buffer_pages(void);
904 * Compound pages have a destructor function. Provide a
905 * prototype for that function and accessor functions.
906 * These are _only_ valid on the head of a compound page.
908 typedef void compound_page_dtor(struct page *);
910 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
911 enum compound_dtor_id {
914 #ifdef CONFIG_HUGETLB_PAGE
917 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
922 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
924 static inline void set_compound_page_dtor(struct page *page,
925 enum compound_dtor_id compound_dtor)
927 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
928 page[1].compound_dtor = compound_dtor;
931 static inline void destroy_compound_page(struct page *page)
933 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
934 compound_page_dtors[page[1].compound_dtor](page);
937 static inline bool hpage_pincount_available(struct page *page)
940 * Can the page->hpage_pinned_refcount field be used? That field is in
941 * the 3rd page of the compound page, so the smallest (2-page) compound
942 * pages cannot support it.
944 page = compound_head(page);
945 return PageCompound(page) && compound_order(page) > 1;
948 static inline int head_compound_pincount(struct page *head)
950 return atomic_read(compound_pincount_ptr(head));
953 static inline int compound_pincount(struct page *page)
955 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
956 page = compound_head(page);
957 return head_compound_pincount(page);
960 static inline void set_compound_order(struct page *page, unsigned int order)
962 page[1].compound_order = order;
963 page[1].compound_nr = 1U << order;
966 /* Returns the number of pages in this potentially compound page. */
967 static inline unsigned long compound_nr(struct page *page)
971 return page[1].compound_nr;
974 /* Returns the number of bytes in this potentially compound page. */
975 static inline unsigned long page_size(struct page *page)
977 return PAGE_SIZE << compound_order(page);
980 /* Returns the number of bits needed for the number of bytes in a page */
981 static inline unsigned int page_shift(struct page *page)
983 return PAGE_SHIFT + compound_order(page);
986 void free_compound_page(struct page *page);
990 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
991 * servicing faults for write access. In the normal case, do always want
992 * pte_mkwrite. But get_user_pages can cause write faults for mappings
993 * that do not have writing enabled, when used by access_process_vm.
995 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
997 if (likely(vma->vm_flags & VM_WRITE))
998 pte = pte_mkwrite(pte);
1002 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1003 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1005 vm_fault_t finish_fault(struct vm_fault *vmf);
1006 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1010 * Multiple processes may "see" the same page. E.g. for untouched
1011 * mappings of /dev/null, all processes see the same page full of
1012 * zeroes, and text pages of executables and shared libraries have
1013 * only one copy in memory, at most, normally.
1015 * For the non-reserved pages, page_count(page) denotes a reference count.
1016 * page_count() == 0 means the page is free. page->lru is then used for
1017 * freelist management in the buddy allocator.
1018 * page_count() > 0 means the page has been allocated.
1020 * Pages are allocated by the slab allocator in order to provide memory
1021 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1022 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1023 * unless a particular usage is carefully commented. (the responsibility of
1024 * freeing the kmalloc memory is the caller's, of course).
1026 * A page may be used by anyone else who does a __get_free_page().
1027 * In this case, page_count still tracks the references, and should only
1028 * be used through the normal accessor functions. The top bits of page->flags
1029 * and page->virtual store page management information, but all other fields
1030 * are unused and could be used privately, carefully. The management of this
1031 * page is the responsibility of the one who allocated it, and those who have
1032 * subsequently been given references to it.
1034 * The other pages (we may call them "pagecache pages") are completely
1035 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1036 * The following discussion applies only to them.
1038 * A pagecache page contains an opaque `private' member, which belongs to the
1039 * page's address_space. Usually, this is the address of a circular list of
1040 * the page's disk buffers. PG_private must be set to tell the VM to call
1041 * into the filesystem to release these pages.
1043 * A page may belong to an inode's memory mapping. In this case, page->mapping
1044 * is the pointer to the inode, and page->index is the file offset of the page,
1045 * in units of PAGE_SIZE.
1047 * If pagecache pages are not associated with an inode, they are said to be
1048 * anonymous pages. These may become associated with the swapcache, and in that
1049 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1051 * In either case (swapcache or inode backed), the pagecache itself holds one
1052 * reference to the page. Setting PG_private should also increment the
1053 * refcount. The each user mapping also has a reference to the page.
1055 * The pagecache pages are stored in a per-mapping radix tree, which is
1056 * rooted at mapping->i_pages, and indexed by offset.
1057 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1058 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1060 * All pagecache pages may be subject to I/O:
1061 * - inode pages may need to be read from disk,
1062 * - inode pages which have been modified and are MAP_SHARED may need
1063 * to be written back to the inode on disk,
1064 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1065 * modified may need to be swapped out to swap space and (later) to be read
1070 * The zone field is never updated after free_area_init_core()
1071 * sets it, so none of the operations on it need to be atomic.
1074 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1075 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1076 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1077 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1078 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1079 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1082 * Define the bit shifts to access each section. For non-existent
1083 * sections we define the shift as 0; that plus a 0 mask ensures
1084 * the compiler will optimise away reference to them.
1086 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1087 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1088 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1089 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1090 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1092 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1093 #ifdef NODE_NOT_IN_PAGE_FLAGS
1094 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1095 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1096 SECTIONS_PGOFF : ZONES_PGOFF)
1098 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1099 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1100 NODES_PGOFF : ZONES_PGOFF)
1103 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1105 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1106 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1107 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1108 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1109 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1110 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1112 static inline enum zone_type page_zonenum(const struct page *page)
1114 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1115 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1118 static inline enum zone_type folio_zonenum(const struct folio *folio)
1120 return page_zonenum(&folio->page);
1123 #ifdef CONFIG_ZONE_DEVICE
1124 static inline bool is_zone_device_page(const struct page *page)
1126 return page_zonenum(page) == ZONE_DEVICE;
1128 extern void memmap_init_zone_device(struct zone *, unsigned long,
1129 unsigned long, struct dev_pagemap *);
1131 static inline bool is_zone_device_page(const struct page *page)
1137 static inline bool is_zone_movable_page(const struct page *page)
1139 return page_zonenum(page) == ZONE_MOVABLE;
1142 #ifdef CONFIG_DEV_PAGEMAP_OPS
1143 void free_devmap_managed_page(struct page *page);
1144 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1146 static inline bool page_is_devmap_managed(struct page *page)
1148 if (!static_branch_unlikely(&devmap_managed_key))
1150 if (!is_zone_device_page(page))
1152 switch (page->pgmap->type) {
1153 case MEMORY_DEVICE_PRIVATE:
1154 case MEMORY_DEVICE_FS_DAX:
1162 void put_devmap_managed_page(struct page *page);
1164 #else /* CONFIG_DEV_PAGEMAP_OPS */
1165 static inline bool page_is_devmap_managed(struct page *page)
1170 static inline void put_devmap_managed_page(struct page *page)
1173 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1175 static inline bool is_device_private_page(const struct page *page)
1177 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1178 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1179 is_zone_device_page(page) &&
1180 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1183 static inline bool is_pci_p2pdma_page(const struct page *page)
1185 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1186 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1187 is_zone_device_page(page) &&
1188 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1191 /* 127: arbitrary random number, small enough to assemble well */
1192 #define folio_ref_zero_or_close_to_overflow(folio) \
1193 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1196 * folio_get - Increment the reference count on a folio.
1197 * @folio: The folio.
1199 * Context: May be called in any context, as long as you know that
1200 * you have a refcount on the folio. If you do not already have one,
1201 * folio_try_get() may be the right interface for you to use.
1203 static inline void folio_get(struct folio *folio)
1205 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1206 folio_ref_inc(folio);
1209 static inline void get_page(struct page *page)
1211 folio_get(page_folio(page));
1214 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1215 struct page *try_grab_compound_head(struct page *page, int refs,
1216 unsigned int flags);
1219 static inline __must_check bool try_get_page(struct page *page)
1221 page = compound_head(page);
1222 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1229 * folio_put - Decrement the reference count on a folio.
1230 * @folio: The folio.
1232 * If the folio's reference count reaches zero, the memory will be
1233 * released back to the page allocator and may be used by another
1234 * allocation immediately. Do not access the memory or the struct folio
1235 * after calling folio_put() unless you can be sure that it wasn't the
1238 * Context: May be called in process or interrupt context, but not in NMI
1239 * context. May be called while holding a spinlock.
1241 static inline void folio_put(struct folio *folio)
1243 if (folio_put_testzero(folio))
1244 __put_page(&folio->page);
1247 static inline void put_page(struct page *page)
1249 struct folio *folio = page_folio(page);
1252 * For devmap managed pages we need to catch refcount transition from
1253 * 2 to 1, when refcount reach one it means the page is free and we
1254 * need to inform the device driver through callback. See
1255 * include/linux/memremap.h and HMM for details.
1257 if (page_is_devmap_managed(&folio->page)) {
1258 put_devmap_managed_page(&folio->page);
1266 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1267 * the page's refcount so that two separate items are tracked: the original page
1268 * reference count, and also a new count of how many pin_user_pages() calls were
1269 * made against the page. ("gup-pinned" is another term for the latter).
1271 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1272 * distinct from normal pages. As such, the unpin_user_page() call (and its
1273 * variants) must be used in order to release gup-pinned pages.
1277 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1278 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1279 * simpler, due to the fact that adding an even power of two to the page
1280 * refcount has the effect of using only the upper N bits, for the code that
1281 * counts up using the bias value. This means that the lower bits are left for
1282 * the exclusive use of the original code that increments and decrements by one
1283 * (or at least, by much smaller values than the bias value).
1285 * Of course, once the lower bits overflow into the upper bits (and this is
1286 * OK, because subtraction recovers the original values), then visual inspection
1287 * no longer suffices to directly view the separate counts. However, for normal
1288 * applications that don't have huge page reference counts, this won't be an
1291 * Locking: the lockless algorithm described in page_cache_get_speculative()
1292 * and page_cache_gup_pin_speculative() provides safe operation for
1293 * get_user_pages and page_mkclean and other calls that race to set up page
1296 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1298 void unpin_user_page(struct page *page);
1299 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1301 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1303 void unpin_user_pages(struct page **pages, unsigned long npages);
1306 * page_maybe_dma_pinned - Report if a page is pinned for DMA.
1309 * This function checks if a page has been pinned via a call to
1310 * a function in the pin_user_pages() family.
1312 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1313 * because it means "definitely not pinned for DMA", but true means "probably
1314 * pinned for DMA, but possibly a false positive due to having at least
1315 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1317 * False positives are OK, because: a) it's unlikely for a page to get that many
1318 * refcounts, and b) all the callers of this routine are expected to be able to
1319 * deal gracefully with a false positive.
1321 * For huge pages, the result will be exactly correct. That's because we have
1322 * more tracking data available: the 3rd struct page in the compound page is
1323 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1326 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1328 * Return: True, if it is likely that the page has been "dma-pinned".
1329 * False, if the page is definitely not dma-pinned.
1331 static inline bool page_maybe_dma_pinned(struct page *page)
1333 if (hpage_pincount_available(page))
1334 return compound_pincount(page) > 0;
1337 * page_ref_count() is signed. If that refcount overflows, then
1338 * page_ref_count() returns a negative value, and callers will avoid
1339 * further incrementing the refcount.
1341 * Here, for that overflow case, use the signed bit to count a little
1342 * bit higher via unsigned math, and thus still get an accurate result.
1344 return ((unsigned int)page_ref_count(compound_head(page))) >=
1345 GUP_PIN_COUNTING_BIAS;
1348 static inline bool is_cow_mapping(vm_flags_t flags)
1350 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1354 * This should most likely only be called during fork() to see whether we
1355 * should break the cow immediately for a page on the src mm.
1357 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1360 if (!is_cow_mapping(vma->vm_flags))
1363 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1366 return page_maybe_dma_pinned(page);
1369 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1370 #define SECTION_IN_PAGE_FLAGS
1374 * The identification function is mainly used by the buddy allocator for
1375 * determining if two pages could be buddies. We are not really identifying
1376 * the zone since we could be using the section number id if we do not have
1377 * node id available in page flags.
1378 * We only guarantee that it will return the same value for two combinable
1381 static inline int page_zone_id(struct page *page)
1383 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1386 #ifdef NODE_NOT_IN_PAGE_FLAGS
1387 extern int page_to_nid(const struct page *page);
1389 static inline int page_to_nid(const struct page *page)
1391 struct page *p = (struct page *)page;
1393 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1397 static inline int folio_nid(const struct folio *folio)
1399 return page_to_nid(&folio->page);
1402 #ifdef CONFIG_NUMA_BALANCING
1403 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1405 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1408 static inline int cpupid_to_pid(int cpupid)
1410 return cpupid & LAST__PID_MASK;
1413 static inline int cpupid_to_cpu(int cpupid)
1415 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1418 static inline int cpupid_to_nid(int cpupid)
1420 return cpu_to_node(cpupid_to_cpu(cpupid));
1423 static inline bool cpupid_pid_unset(int cpupid)
1425 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1428 static inline bool cpupid_cpu_unset(int cpupid)
1430 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1433 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1435 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1438 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1439 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1440 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1442 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1445 static inline int page_cpupid_last(struct page *page)
1447 return page->_last_cpupid;
1449 static inline void page_cpupid_reset_last(struct page *page)
1451 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1454 static inline int page_cpupid_last(struct page *page)
1456 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1459 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1461 static inline void page_cpupid_reset_last(struct page *page)
1463 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1465 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1466 #else /* !CONFIG_NUMA_BALANCING */
1467 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1469 return page_to_nid(page); /* XXX */
1472 static inline int page_cpupid_last(struct page *page)
1474 return page_to_nid(page); /* XXX */
1477 static inline int cpupid_to_nid(int cpupid)
1482 static inline int cpupid_to_pid(int cpupid)
1487 static inline int cpupid_to_cpu(int cpupid)
1492 static inline int cpu_pid_to_cpupid(int nid, int pid)
1497 static inline bool cpupid_pid_unset(int cpupid)
1502 static inline void page_cpupid_reset_last(struct page *page)
1506 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1510 #endif /* CONFIG_NUMA_BALANCING */
1512 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1515 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1516 * setting tags for all pages to native kernel tag value 0xff, as the default
1517 * value 0x00 maps to 0xff.
1520 static inline u8 page_kasan_tag(const struct page *page)
1524 if (kasan_enabled()) {
1525 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1532 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1534 if (kasan_enabled()) {
1536 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1537 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1541 static inline void page_kasan_tag_reset(struct page *page)
1543 if (kasan_enabled())
1544 page_kasan_tag_set(page, 0xff);
1547 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1549 static inline u8 page_kasan_tag(const struct page *page)
1554 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1555 static inline void page_kasan_tag_reset(struct page *page) { }
1557 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1559 static inline struct zone *page_zone(const struct page *page)
1561 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1564 static inline pg_data_t *page_pgdat(const struct page *page)
1566 return NODE_DATA(page_to_nid(page));
1569 static inline struct zone *folio_zone(const struct folio *folio)
1571 return page_zone(&folio->page);
1574 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1576 return page_pgdat(&folio->page);
1579 #ifdef SECTION_IN_PAGE_FLAGS
1580 static inline void set_page_section(struct page *page, unsigned long section)
1582 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1583 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1586 static inline unsigned long page_to_section(const struct page *page)
1588 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1593 * folio_pfn - Return the Page Frame Number of a folio.
1594 * @folio: The folio.
1596 * A folio may contain multiple pages. The pages have consecutive
1597 * Page Frame Numbers.
1599 * Return: The Page Frame Number of the first page in the folio.
1601 static inline unsigned long folio_pfn(struct folio *folio)
1603 return page_to_pfn(&folio->page);
1606 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1607 #ifdef CONFIG_MIGRATION
1608 static inline bool is_pinnable_page(struct page *page)
1610 return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) ||
1611 is_zero_pfn(page_to_pfn(page));
1614 static inline bool is_pinnable_page(struct page *page)
1620 static inline void set_page_zone(struct page *page, enum zone_type zone)
1622 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1623 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1626 static inline void set_page_node(struct page *page, unsigned long node)
1628 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1629 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1632 static inline void set_page_links(struct page *page, enum zone_type zone,
1633 unsigned long node, unsigned long pfn)
1635 set_page_zone(page, zone);
1636 set_page_node(page, node);
1637 #ifdef SECTION_IN_PAGE_FLAGS
1638 set_page_section(page, pfn_to_section_nr(pfn));
1643 * folio_nr_pages - The number of pages in the folio.
1644 * @folio: The folio.
1646 * Return: A positive power of two.
1648 static inline long folio_nr_pages(struct folio *folio)
1650 return compound_nr(&folio->page);
1654 * folio_next - Move to the next physical folio.
1655 * @folio: The folio we're currently operating on.
1657 * If you have physically contiguous memory which may span more than
1658 * one folio (eg a &struct bio_vec), use this function to move from one
1659 * folio to the next. Do not use it if the memory is only virtually
1660 * contiguous as the folios are almost certainly not adjacent to each
1661 * other. This is the folio equivalent to writing ``page++``.
1663 * Context: We assume that the folios are refcounted and/or locked at a
1664 * higher level and do not adjust the reference counts.
1665 * Return: The next struct folio.
1667 static inline struct folio *folio_next(struct folio *folio)
1669 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1673 * folio_shift - The size of the memory described by this folio.
1674 * @folio: The folio.
1676 * A folio represents a number of bytes which is a power-of-two in size.
1677 * This function tells you which power-of-two the folio is. See also
1678 * folio_size() and folio_order().
1680 * Context: The caller should have a reference on the folio to prevent
1681 * it from being split. It is not necessary for the folio to be locked.
1682 * Return: The base-2 logarithm of the size of this folio.
1684 static inline unsigned int folio_shift(struct folio *folio)
1686 return PAGE_SHIFT + folio_order(folio);
1690 * folio_size - The number of bytes in a folio.
1691 * @folio: The folio.
1693 * Context: The caller should have a reference on the folio to prevent
1694 * it from being split. It is not necessary for the folio to be locked.
1695 * Return: The number of bytes in this folio.
1697 static inline size_t folio_size(struct folio *folio)
1699 return PAGE_SIZE << folio_order(folio);
1702 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1703 static inline int arch_make_page_accessible(struct page *page)
1709 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1710 static inline int arch_make_folio_accessible(struct folio *folio)
1713 long i, nr = folio_nr_pages(folio);
1715 for (i = 0; i < nr; i++) {
1716 ret = arch_make_page_accessible(folio_page(folio, i));
1726 * Some inline functions in vmstat.h depend on page_zone()
1728 #include <linux/vmstat.h>
1730 static __always_inline void *lowmem_page_address(const struct page *page)
1732 return page_to_virt(page);
1735 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1736 #define HASHED_PAGE_VIRTUAL
1739 #if defined(WANT_PAGE_VIRTUAL)
1740 static inline void *page_address(const struct page *page)
1742 return page->virtual;
1744 static inline void set_page_address(struct page *page, void *address)
1746 page->virtual = address;
1748 #define page_address_init() do { } while(0)
1751 #if defined(HASHED_PAGE_VIRTUAL)
1752 void *page_address(const struct page *page);
1753 void set_page_address(struct page *page, void *virtual);
1754 void page_address_init(void);
1757 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1758 #define page_address(page) lowmem_page_address(page)
1759 #define set_page_address(page, address) do { } while(0)
1760 #define page_address_init() do { } while(0)
1763 static inline void *folio_address(const struct folio *folio)
1765 return page_address(&folio->page);
1768 extern void *page_rmapping(struct page *page);
1769 extern struct anon_vma *page_anon_vma(struct page *page);
1770 extern pgoff_t __page_file_index(struct page *page);
1773 * Return the pagecache index of the passed page. Regular pagecache pages
1774 * use ->index whereas swapcache pages use swp_offset(->private)
1776 static inline pgoff_t page_index(struct page *page)
1778 if (unlikely(PageSwapCache(page)))
1779 return __page_file_index(page);
1783 bool page_mapped(struct page *page);
1784 bool folio_mapped(struct folio *folio);
1787 * Return true only if the page has been allocated with
1788 * ALLOC_NO_WATERMARKS and the low watermark was not
1789 * met implying that the system is under some pressure.
1791 static inline bool page_is_pfmemalloc(const struct page *page)
1794 * lru.next has bit 1 set if the page is allocated from the
1795 * pfmemalloc reserves. Callers may simply overwrite it if
1796 * they do not need to preserve that information.
1798 return (uintptr_t)page->lru.next & BIT(1);
1802 * Only to be called by the page allocator on a freshly allocated
1805 static inline void set_page_pfmemalloc(struct page *page)
1807 page->lru.next = (void *)BIT(1);
1810 static inline void clear_page_pfmemalloc(struct page *page)
1812 page->lru.next = NULL;
1816 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1818 extern void pagefault_out_of_memory(void);
1820 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1821 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1822 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1825 * Flags passed to show_mem() and show_free_areas() to suppress output in
1828 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1830 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1833 extern bool can_do_mlock(void);
1835 static inline bool can_do_mlock(void) { return false; }
1837 extern int user_shm_lock(size_t, struct ucounts *);
1838 extern void user_shm_unlock(size_t, struct ucounts *);
1840 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1842 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1845 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1846 unsigned long size);
1847 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1848 unsigned long size);
1849 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1850 unsigned long start, unsigned long end);
1852 struct mmu_notifier_range;
1854 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1855 unsigned long end, unsigned long floor, unsigned long ceiling);
1857 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1858 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1859 struct mmu_notifier_range *range, pte_t **ptepp,
1860 pmd_t **pmdpp, spinlock_t **ptlp);
1861 int follow_pte(struct mm_struct *mm, unsigned long address,
1862 pte_t **ptepp, spinlock_t **ptlp);
1863 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1864 unsigned long *pfn);
1865 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1866 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1867 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1868 void *buf, int len, int write);
1870 extern void truncate_pagecache(struct inode *inode, loff_t new);
1871 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1872 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1873 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1874 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1875 int invalidate_inode_page(struct page *page);
1878 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1879 unsigned long address, unsigned int flags,
1880 struct pt_regs *regs);
1881 extern int fixup_user_fault(struct mm_struct *mm,
1882 unsigned long address, unsigned int fault_flags,
1884 void unmap_mapping_pages(struct address_space *mapping,
1885 pgoff_t start, pgoff_t nr, bool even_cows);
1886 void unmap_mapping_range(struct address_space *mapping,
1887 loff_t const holebegin, loff_t const holelen, int even_cows);
1889 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1890 unsigned long address, unsigned int flags,
1891 struct pt_regs *regs)
1893 /* should never happen if there's no MMU */
1895 return VM_FAULT_SIGBUS;
1897 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1898 unsigned int fault_flags, bool *unlocked)
1900 /* should never happen if there's no MMU */
1904 static inline void unmap_mapping_pages(struct address_space *mapping,
1905 pgoff_t start, pgoff_t nr, bool even_cows) { }
1906 static inline void unmap_mapping_range(struct address_space *mapping,
1907 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1910 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1911 loff_t const holebegin, loff_t const holelen)
1913 unmap_mapping_range(mapping, holebegin, holelen, 0);
1916 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1917 void *buf, int len, unsigned int gup_flags);
1918 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1919 void *buf, int len, unsigned int gup_flags);
1920 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1921 void *buf, int len, unsigned int gup_flags);
1923 long get_user_pages_remote(struct mm_struct *mm,
1924 unsigned long start, unsigned long nr_pages,
1925 unsigned int gup_flags, struct page **pages,
1926 struct vm_area_struct **vmas, int *locked);
1927 long pin_user_pages_remote(struct mm_struct *mm,
1928 unsigned long start, unsigned long nr_pages,
1929 unsigned int gup_flags, struct page **pages,
1930 struct vm_area_struct **vmas, int *locked);
1931 long get_user_pages(unsigned long start, unsigned long nr_pages,
1932 unsigned int gup_flags, struct page **pages,
1933 struct vm_area_struct **vmas);
1934 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1935 unsigned int gup_flags, struct page **pages,
1936 struct vm_area_struct **vmas);
1937 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1938 unsigned int gup_flags, struct page **pages, int *locked);
1939 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1940 unsigned int gup_flags, struct page **pages, int *locked);
1941 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1942 struct page **pages, unsigned int gup_flags);
1943 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1944 struct page **pages, unsigned int gup_flags);
1946 int get_user_pages_fast(unsigned long start, int nr_pages,
1947 unsigned int gup_flags, struct page **pages);
1948 int pin_user_pages_fast(unsigned long start, int nr_pages,
1949 unsigned int gup_flags, struct page **pages);
1951 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1952 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1953 struct task_struct *task, bool bypass_rlim);
1956 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1957 struct page **pages);
1958 struct page *get_dump_page(unsigned long addr);
1960 extern void do_invalidatepage(struct page *page, unsigned int offset,
1961 unsigned int length);
1963 bool folio_mark_dirty(struct folio *folio);
1964 bool set_page_dirty(struct page *page);
1965 int set_page_dirty_lock(struct page *page);
1967 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1969 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1970 unsigned long old_addr, struct vm_area_struct *new_vma,
1971 unsigned long new_addr, unsigned long len,
1972 bool need_rmap_locks);
1975 * Flags used by change_protection(). For now we make it a bitmap so
1976 * that we can pass in multiple flags just like parameters. However
1977 * for now all the callers are only use one of the flags at the same
1980 /* Whether we should allow dirty bit accounting */
1981 #define MM_CP_DIRTY_ACCT (1UL << 0)
1982 /* Whether this protection change is for NUMA hints */
1983 #define MM_CP_PROT_NUMA (1UL << 1)
1984 /* Whether this change is for write protecting */
1985 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1986 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1987 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1988 MM_CP_UFFD_WP_RESOLVE)
1990 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1991 unsigned long end, pgprot_t newprot,
1992 unsigned long cp_flags);
1993 extern int mprotect_fixup(struct vm_area_struct *vma,
1994 struct vm_area_struct **pprev, unsigned long start,
1995 unsigned long end, unsigned long newflags);
1998 * doesn't attempt to fault and will return short.
2000 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2001 unsigned int gup_flags, struct page **pages);
2002 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
2003 unsigned int gup_flags, struct page **pages);
2005 static inline bool get_user_page_fast_only(unsigned long addr,
2006 unsigned int gup_flags, struct page **pagep)
2008 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2011 * per-process(per-mm_struct) statistics.
2013 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2015 long val = atomic_long_read(&mm->rss_stat.count[member]);
2017 #ifdef SPLIT_RSS_COUNTING
2019 * counter is updated in asynchronous manner and may go to minus.
2020 * But it's never be expected number for users.
2025 return (unsigned long)val;
2028 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
2030 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2032 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
2034 mm_trace_rss_stat(mm, member, count);
2037 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2039 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
2041 mm_trace_rss_stat(mm, member, count);
2044 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2046 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
2048 mm_trace_rss_stat(mm, member, count);
2051 /* Optimized variant when page is already known not to be PageAnon */
2052 static inline int mm_counter_file(struct page *page)
2054 if (PageSwapBacked(page))
2055 return MM_SHMEMPAGES;
2056 return MM_FILEPAGES;
2059 static inline int mm_counter(struct page *page)
2062 return MM_ANONPAGES;
2063 return mm_counter_file(page);
2066 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2068 return get_mm_counter(mm, MM_FILEPAGES) +
2069 get_mm_counter(mm, MM_ANONPAGES) +
2070 get_mm_counter(mm, MM_SHMEMPAGES);
2073 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2075 return max(mm->hiwater_rss, get_mm_rss(mm));
2078 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2080 return max(mm->hiwater_vm, mm->total_vm);
2083 static inline void update_hiwater_rss(struct mm_struct *mm)
2085 unsigned long _rss = get_mm_rss(mm);
2087 if ((mm)->hiwater_rss < _rss)
2088 (mm)->hiwater_rss = _rss;
2091 static inline void update_hiwater_vm(struct mm_struct *mm)
2093 if (mm->hiwater_vm < mm->total_vm)
2094 mm->hiwater_vm = mm->total_vm;
2097 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2099 mm->hiwater_rss = get_mm_rss(mm);
2102 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2103 struct mm_struct *mm)
2105 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2107 if (*maxrss < hiwater_rss)
2108 *maxrss = hiwater_rss;
2111 #if defined(SPLIT_RSS_COUNTING)
2112 void sync_mm_rss(struct mm_struct *mm);
2114 static inline void sync_mm_rss(struct mm_struct *mm)
2119 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2120 static inline int pte_special(pte_t pte)
2125 static inline pte_t pte_mkspecial(pte_t pte)
2131 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2132 static inline int pte_devmap(pte_t pte)
2138 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2140 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2142 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2146 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2150 #ifdef __PAGETABLE_P4D_FOLDED
2151 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2152 unsigned long address)
2157 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2160 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2161 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2162 unsigned long address)
2166 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2167 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2170 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2172 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2174 if (mm_pud_folded(mm))
2176 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2179 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2181 if (mm_pud_folded(mm))
2183 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2187 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2188 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2189 unsigned long address)
2194 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2195 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2198 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2200 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2202 if (mm_pmd_folded(mm))
2204 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2207 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2209 if (mm_pmd_folded(mm))
2211 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2216 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2218 atomic_long_set(&mm->pgtables_bytes, 0);
2221 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2223 return atomic_long_read(&mm->pgtables_bytes);
2226 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2228 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2231 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2233 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2237 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2238 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2243 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2244 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2247 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2248 int __pte_alloc_kernel(pmd_t *pmd);
2250 #if defined(CONFIG_MMU)
2252 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2253 unsigned long address)
2255 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2256 NULL : p4d_offset(pgd, address);
2259 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2260 unsigned long address)
2262 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2263 NULL : pud_offset(p4d, address);
2266 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2268 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2269 NULL: pmd_offset(pud, address);
2271 #endif /* CONFIG_MMU */
2273 #if USE_SPLIT_PTE_PTLOCKS
2274 #if ALLOC_SPLIT_PTLOCKS
2275 void __init ptlock_cache_init(void);
2276 extern bool ptlock_alloc(struct page *page);
2277 extern void ptlock_free(struct page *page);
2279 static inline spinlock_t *ptlock_ptr(struct page *page)
2283 #else /* ALLOC_SPLIT_PTLOCKS */
2284 static inline void ptlock_cache_init(void)
2288 static inline bool ptlock_alloc(struct page *page)
2293 static inline void ptlock_free(struct page *page)
2297 static inline spinlock_t *ptlock_ptr(struct page *page)
2301 #endif /* ALLOC_SPLIT_PTLOCKS */
2303 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2305 return ptlock_ptr(pmd_page(*pmd));
2308 static inline bool ptlock_init(struct page *page)
2311 * prep_new_page() initialize page->private (and therefore page->ptl)
2312 * with 0. Make sure nobody took it in use in between.
2314 * It can happen if arch try to use slab for page table allocation:
2315 * slab code uses page->slab_cache, which share storage with page->ptl.
2317 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2318 if (!ptlock_alloc(page))
2320 spin_lock_init(ptlock_ptr(page));
2324 #else /* !USE_SPLIT_PTE_PTLOCKS */
2326 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2328 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2330 return &mm->page_table_lock;
2332 static inline void ptlock_cache_init(void) {}
2333 static inline bool ptlock_init(struct page *page) { return true; }
2334 static inline void ptlock_free(struct page *page) {}
2335 #endif /* USE_SPLIT_PTE_PTLOCKS */
2337 static inline void pgtable_init(void)
2339 ptlock_cache_init();
2340 pgtable_cache_init();
2343 static inline bool pgtable_pte_page_ctor(struct page *page)
2345 if (!ptlock_init(page))
2347 __SetPageTable(page);
2348 inc_lruvec_page_state(page, NR_PAGETABLE);
2352 static inline void pgtable_pte_page_dtor(struct page *page)
2355 __ClearPageTable(page);
2356 dec_lruvec_page_state(page, NR_PAGETABLE);
2359 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2361 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2362 pte_t *__pte = pte_offset_map(pmd, address); \
2368 #define pte_unmap_unlock(pte, ptl) do { \
2373 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2375 #define pte_alloc_map(mm, pmd, address) \
2376 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2378 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2379 (pte_alloc(mm, pmd) ? \
2380 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2382 #define pte_alloc_kernel(pmd, address) \
2383 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2384 NULL: pte_offset_kernel(pmd, address))
2386 #if USE_SPLIT_PMD_PTLOCKS
2388 static struct page *pmd_to_page(pmd_t *pmd)
2390 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2391 return virt_to_page((void *)((unsigned long) pmd & mask));
2394 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2396 return ptlock_ptr(pmd_to_page(pmd));
2399 static inline bool pmd_ptlock_init(struct page *page)
2401 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2402 page->pmd_huge_pte = NULL;
2404 return ptlock_init(page);
2407 static inline void pmd_ptlock_free(struct page *page)
2409 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2410 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2415 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2419 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2421 return &mm->page_table_lock;
2424 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2425 static inline void pmd_ptlock_free(struct page *page) {}
2427 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2431 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2433 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2438 static inline bool pgtable_pmd_page_ctor(struct page *page)
2440 if (!pmd_ptlock_init(page))
2442 __SetPageTable(page);
2443 inc_lruvec_page_state(page, NR_PAGETABLE);
2447 static inline void pgtable_pmd_page_dtor(struct page *page)
2449 pmd_ptlock_free(page);
2450 __ClearPageTable(page);
2451 dec_lruvec_page_state(page, NR_PAGETABLE);
2455 * No scalability reason to split PUD locks yet, but follow the same pattern
2456 * as the PMD locks to make it easier if we decide to. The VM should not be
2457 * considered ready to switch to split PUD locks yet; there may be places
2458 * which need to be converted from page_table_lock.
2460 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2462 return &mm->page_table_lock;
2465 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2467 spinlock_t *ptl = pud_lockptr(mm, pud);
2473 extern void __init pagecache_init(void);
2474 extern void __init free_area_init_memoryless_node(int nid);
2475 extern void free_initmem(void);
2478 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2479 * into the buddy system. The freed pages will be poisoned with pattern
2480 * "poison" if it's within range [0, UCHAR_MAX].
2481 * Return pages freed into the buddy system.
2483 extern unsigned long free_reserved_area(void *start, void *end,
2484 int poison, const char *s);
2486 extern void adjust_managed_page_count(struct page *page, long count);
2487 extern void mem_init_print_info(void);
2489 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2491 /* Free the reserved page into the buddy system, so it gets managed. */
2492 static inline void free_reserved_page(struct page *page)
2494 ClearPageReserved(page);
2495 init_page_count(page);
2497 adjust_managed_page_count(page, 1);
2499 #define free_highmem_page(page) free_reserved_page(page)
2501 static inline void mark_page_reserved(struct page *page)
2503 SetPageReserved(page);
2504 adjust_managed_page_count(page, -1);
2508 * Default method to free all the __init memory into the buddy system.
2509 * The freed pages will be poisoned with pattern "poison" if it's within
2510 * range [0, UCHAR_MAX].
2511 * Return pages freed into the buddy system.
2513 static inline unsigned long free_initmem_default(int poison)
2515 extern char __init_begin[], __init_end[];
2517 return free_reserved_area(&__init_begin, &__init_end,
2518 poison, "unused kernel image (initmem)");
2521 static inline unsigned long get_num_physpages(void)
2524 unsigned long phys_pages = 0;
2526 for_each_online_node(nid)
2527 phys_pages += node_present_pages(nid);
2533 * Using memblock node mappings, an architecture may initialise its
2534 * zones, allocate the backing mem_map and account for memory holes in an
2535 * architecture independent manner.
2537 * An architecture is expected to register range of page frames backed by
2538 * physical memory with memblock_add[_node]() before calling
2539 * free_area_init() passing in the PFN each zone ends at. At a basic
2540 * usage, an architecture is expected to do something like
2542 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2544 * for_each_valid_physical_page_range()
2545 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2546 * free_area_init(max_zone_pfns);
2548 void free_area_init(unsigned long *max_zone_pfn);
2549 unsigned long node_map_pfn_alignment(void);
2550 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2551 unsigned long end_pfn);
2552 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2553 unsigned long end_pfn);
2554 extern void get_pfn_range_for_nid(unsigned int nid,
2555 unsigned long *start_pfn, unsigned long *end_pfn);
2556 extern unsigned long find_min_pfn_with_active_regions(void);
2559 static inline int early_pfn_to_nid(unsigned long pfn)
2564 /* please see mm/page_alloc.c */
2565 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2568 extern void set_dma_reserve(unsigned long new_dma_reserve);
2569 extern void memmap_init_range(unsigned long, int, unsigned long,
2570 unsigned long, unsigned long, enum meminit_context,
2571 struct vmem_altmap *, int migratetype);
2572 extern void setup_per_zone_wmarks(void);
2573 extern void calculate_min_free_kbytes(void);
2574 extern int __meminit init_per_zone_wmark_min(void);
2575 extern void mem_init(void);
2576 extern void __init mmap_init(void);
2577 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2578 extern long si_mem_available(void);
2579 extern void si_meminfo(struct sysinfo * val);
2580 extern void si_meminfo_node(struct sysinfo *val, int nid);
2581 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2582 extern unsigned long arch_reserved_kernel_pages(void);
2585 extern __printf(3, 4)
2586 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2588 extern void setup_per_cpu_pageset(void);
2591 extern int min_free_kbytes;
2592 extern int watermark_boost_factor;
2593 extern int watermark_scale_factor;
2594 extern bool arch_has_descending_max_zone_pfns(void);
2597 extern atomic_long_t mmap_pages_allocated;
2598 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2600 /* interval_tree.c */
2601 void vma_interval_tree_insert(struct vm_area_struct *node,
2602 struct rb_root_cached *root);
2603 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2604 struct vm_area_struct *prev,
2605 struct rb_root_cached *root);
2606 void vma_interval_tree_remove(struct vm_area_struct *node,
2607 struct rb_root_cached *root);
2608 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2609 unsigned long start, unsigned long last);
2610 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2611 unsigned long start, unsigned long last);
2613 #define vma_interval_tree_foreach(vma, root, start, last) \
2614 for (vma = vma_interval_tree_iter_first(root, start, last); \
2615 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2617 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2618 struct rb_root_cached *root);
2619 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2620 struct rb_root_cached *root);
2621 struct anon_vma_chain *
2622 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2623 unsigned long start, unsigned long last);
2624 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2625 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2626 #ifdef CONFIG_DEBUG_VM_RB
2627 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2630 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2631 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2632 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2635 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2636 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2637 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2638 struct vm_area_struct *expand);
2639 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2640 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2642 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2644 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2645 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2646 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2647 struct mempolicy *, struct vm_userfaultfd_ctx);
2648 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2649 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2650 unsigned long addr, int new_below);
2651 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2652 unsigned long addr, int new_below);
2653 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2654 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2655 struct rb_node **, struct rb_node *);
2656 extern void unlink_file_vma(struct vm_area_struct *);
2657 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2658 unsigned long addr, unsigned long len, pgoff_t pgoff,
2659 bool *need_rmap_locks);
2660 extern void exit_mmap(struct mm_struct *);
2662 static inline int check_data_rlimit(unsigned long rlim,
2664 unsigned long start,
2665 unsigned long end_data,
2666 unsigned long start_data)
2668 if (rlim < RLIM_INFINITY) {
2669 if (((new - start) + (end_data - start_data)) > rlim)
2676 extern int mm_take_all_locks(struct mm_struct *mm);
2677 extern void mm_drop_all_locks(struct mm_struct *mm);
2679 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2680 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2681 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2682 extern struct file *get_task_exe_file(struct task_struct *task);
2684 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2685 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2687 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2688 const struct vm_special_mapping *sm);
2689 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2690 unsigned long addr, unsigned long len,
2691 unsigned long flags,
2692 const struct vm_special_mapping *spec);
2693 /* This is an obsolete alternative to _install_special_mapping. */
2694 extern int install_special_mapping(struct mm_struct *mm,
2695 unsigned long addr, unsigned long len,
2696 unsigned long flags, struct page **pages);
2698 unsigned long randomize_stack_top(unsigned long stack_top);
2700 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2702 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2703 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2704 struct list_head *uf);
2705 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2706 unsigned long len, unsigned long prot, unsigned long flags,
2707 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2708 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2709 struct list_head *uf, bool downgrade);
2710 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2711 struct list_head *uf);
2712 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2715 extern int __mm_populate(unsigned long addr, unsigned long len,
2717 static inline void mm_populate(unsigned long addr, unsigned long len)
2720 (void) __mm_populate(addr, len, 1);
2723 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2726 /* These take the mm semaphore themselves */
2727 extern int __must_check vm_brk(unsigned long, unsigned long);
2728 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2729 extern int vm_munmap(unsigned long, size_t);
2730 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2731 unsigned long, unsigned long,
2732 unsigned long, unsigned long);
2734 struct vm_unmapped_area_info {
2735 #define VM_UNMAPPED_AREA_TOPDOWN 1
2736 unsigned long flags;
2737 unsigned long length;
2738 unsigned long low_limit;
2739 unsigned long high_limit;
2740 unsigned long align_mask;
2741 unsigned long align_offset;
2744 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2747 extern void truncate_inode_pages(struct address_space *, loff_t);
2748 extern void truncate_inode_pages_range(struct address_space *,
2749 loff_t lstart, loff_t lend);
2750 extern void truncate_inode_pages_final(struct address_space *);
2752 /* generic vm_area_ops exported for stackable file systems */
2753 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2754 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2755 pgoff_t start_pgoff, pgoff_t end_pgoff);
2756 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2758 extern unsigned long stack_guard_gap;
2759 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2760 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2762 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2763 extern int expand_downwards(struct vm_area_struct *vma,
2764 unsigned long address);
2766 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2768 #define expand_upwards(vma, address) (0)
2771 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2772 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2773 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2774 struct vm_area_struct **pprev);
2777 * find_vma_intersection() - Look up the first VMA which intersects the interval
2778 * @mm: The process address space.
2779 * @start_addr: The inclusive start user address.
2780 * @end_addr: The exclusive end user address.
2782 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2783 * start_addr < end_addr.
2786 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2787 unsigned long start_addr,
2788 unsigned long end_addr)
2790 struct vm_area_struct *vma = find_vma(mm, start_addr);
2792 if (vma && end_addr <= vma->vm_start)
2798 * vma_lookup() - Find a VMA at a specific address
2799 * @mm: The process address space.
2800 * @addr: The user address.
2802 * Return: The vm_area_struct at the given address, %NULL otherwise.
2805 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2807 struct vm_area_struct *vma = find_vma(mm, addr);
2809 if (vma && addr < vma->vm_start)
2815 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2817 unsigned long vm_start = vma->vm_start;
2819 if (vma->vm_flags & VM_GROWSDOWN) {
2820 vm_start -= stack_guard_gap;
2821 if (vm_start > vma->vm_start)
2827 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2829 unsigned long vm_end = vma->vm_end;
2831 if (vma->vm_flags & VM_GROWSUP) {
2832 vm_end += stack_guard_gap;
2833 if (vm_end < vma->vm_end)
2834 vm_end = -PAGE_SIZE;
2839 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2841 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2844 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2845 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2846 unsigned long vm_start, unsigned long vm_end)
2848 struct vm_area_struct *vma = find_vma(mm, vm_start);
2850 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2856 static inline bool range_in_vma(struct vm_area_struct *vma,
2857 unsigned long start, unsigned long end)
2859 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2863 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2864 void vma_set_page_prot(struct vm_area_struct *vma);
2866 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2870 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2872 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2876 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2878 #ifdef CONFIG_NUMA_BALANCING
2879 unsigned long change_prot_numa(struct vm_area_struct *vma,
2880 unsigned long start, unsigned long end);
2883 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2884 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2885 unsigned long pfn, unsigned long size, pgprot_t);
2886 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2887 unsigned long pfn, unsigned long size, pgprot_t prot);
2888 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2889 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2890 struct page **pages, unsigned long *num);
2891 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2893 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2895 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2897 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2898 unsigned long pfn, pgprot_t pgprot);
2899 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2901 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2902 pfn_t pfn, pgprot_t pgprot);
2903 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2904 unsigned long addr, pfn_t pfn);
2905 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2907 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2908 unsigned long addr, struct page *page)
2910 int err = vm_insert_page(vma, addr, page);
2913 return VM_FAULT_OOM;
2914 if (err < 0 && err != -EBUSY)
2915 return VM_FAULT_SIGBUS;
2917 return VM_FAULT_NOPAGE;
2920 #ifndef io_remap_pfn_range
2921 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2922 unsigned long addr, unsigned long pfn,
2923 unsigned long size, pgprot_t prot)
2925 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2929 static inline vm_fault_t vmf_error(int err)
2932 return VM_FAULT_OOM;
2933 return VM_FAULT_SIGBUS;
2936 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2937 unsigned int foll_flags);
2939 #define FOLL_WRITE 0x01 /* check pte is writable */
2940 #define FOLL_TOUCH 0x02 /* mark page accessed */
2941 #define FOLL_GET 0x04 /* do get_page on page */
2942 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2943 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2944 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2945 * and return without waiting upon it */
2946 #define FOLL_POPULATE 0x40 /* fault in pages (with FOLL_MLOCK) */
2947 #define FOLL_NOFAULT 0x80 /* do not fault in pages */
2948 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2949 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2950 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2951 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2952 #define FOLL_MLOCK 0x1000 /* lock present pages */
2953 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2954 #define FOLL_COW 0x4000 /* internal GUP flag */
2955 #define FOLL_ANON 0x8000 /* don't do file mappings */
2956 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2957 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2958 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2959 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2962 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2963 * other. Here is what they mean, and how to use them:
2965 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2966 * period _often_ under userspace control. This is in contrast to
2967 * iov_iter_get_pages(), whose usages are transient.
2969 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2970 * lifetime enforced by the filesystem and we need guarantees that longterm
2971 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2972 * the filesystem. Ideas for this coordination include revoking the longterm
2973 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2974 * added after the problem with filesystems was found FS DAX VMAs are
2975 * specifically failed. Filesystem pages are still subject to bugs and use of
2976 * FOLL_LONGTERM should be avoided on those pages.
2978 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2979 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2980 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2981 * is due to an incompatibility with the FS DAX check and
2982 * FAULT_FLAG_ALLOW_RETRY.
2984 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2985 * that region. And so, CMA attempts to migrate the page before pinning, when
2986 * FOLL_LONGTERM is specified.
2988 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2989 * but an additional pin counting system) will be invoked. This is intended for
2990 * anything that gets a page reference and then touches page data (for example,
2991 * Direct IO). This lets the filesystem know that some non-file-system entity is
2992 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2993 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2994 * a call to unpin_user_page().
2996 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2997 * and separate refcounting mechanisms, however, and that means that each has
2998 * its own acquire and release mechanisms:
3000 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
3002 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
3004 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
3005 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
3006 * calls applied to them, and that's perfectly OK. This is a constraint on the
3007 * callers, not on the pages.)
3009 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
3010 * directly by the caller. That's in order to help avoid mismatches when
3011 * releasing pages: get_user_pages*() pages must be released via put_page(),
3012 * while pin_user_pages*() pages must be released via unpin_user_page().
3014 * Please see Documentation/core-api/pin_user_pages.rst for more information.
3017 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3019 if (vm_fault & VM_FAULT_OOM)
3021 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3022 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3023 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3028 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3029 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3030 unsigned long size, pte_fn_t fn, void *data);
3031 extern int apply_to_existing_page_range(struct mm_struct *mm,
3032 unsigned long address, unsigned long size,
3033 pte_fn_t fn, void *data);
3035 extern void init_mem_debugging_and_hardening(void);
3036 #ifdef CONFIG_PAGE_POISONING
3037 extern void __kernel_poison_pages(struct page *page, int numpages);
3038 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3039 extern bool _page_poisoning_enabled_early;
3040 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3041 static inline bool page_poisoning_enabled(void)
3043 return _page_poisoning_enabled_early;
3046 * For use in fast paths after init_mem_debugging() has run, or when a
3047 * false negative result is not harmful when called too early.
3049 static inline bool page_poisoning_enabled_static(void)
3051 return static_branch_unlikely(&_page_poisoning_enabled);
3053 static inline void kernel_poison_pages(struct page *page, int numpages)
3055 if (page_poisoning_enabled_static())
3056 __kernel_poison_pages(page, numpages);
3058 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3060 if (page_poisoning_enabled_static())
3061 __kernel_unpoison_pages(page, numpages);
3064 static inline bool page_poisoning_enabled(void) { return false; }
3065 static inline bool page_poisoning_enabled_static(void) { return false; }
3066 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3067 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3068 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3071 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3072 static inline bool want_init_on_alloc(gfp_t flags)
3074 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3077 return flags & __GFP_ZERO;
3080 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3081 static inline bool want_init_on_free(void)
3083 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3087 extern bool _debug_pagealloc_enabled_early;
3088 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3090 static inline bool debug_pagealloc_enabled(void)
3092 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3093 _debug_pagealloc_enabled_early;
3097 * For use in fast paths after init_debug_pagealloc() has run, or when a
3098 * false negative result is not harmful when called too early.
3100 static inline bool debug_pagealloc_enabled_static(void)
3102 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3105 return static_branch_unlikely(&_debug_pagealloc_enabled);
3108 #ifdef CONFIG_DEBUG_PAGEALLOC
3110 * To support DEBUG_PAGEALLOC architecture must ensure that
3111 * __kernel_map_pages() never fails
3113 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3115 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3117 if (debug_pagealloc_enabled_static())
3118 __kernel_map_pages(page, numpages, 1);
3121 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3123 if (debug_pagealloc_enabled_static())
3124 __kernel_map_pages(page, numpages, 0);
3126 #else /* CONFIG_DEBUG_PAGEALLOC */
3127 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3128 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3129 #endif /* CONFIG_DEBUG_PAGEALLOC */
3131 #ifdef __HAVE_ARCH_GATE_AREA
3132 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3133 extern int in_gate_area_no_mm(unsigned long addr);
3134 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3136 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3140 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3141 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3145 #endif /* __HAVE_ARCH_GATE_AREA */
3147 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3149 #ifdef CONFIG_SYSCTL
3150 extern int sysctl_drop_caches;
3151 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3155 void drop_slab(void);
3156 void drop_slab_node(int nid);
3159 #define randomize_va_space 0
3161 extern int randomize_va_space;
3164 const char * arch_vma_name(struct vm_area_struct *vma);
3166 void print_vma_addr(char *prefix, unsigned long rip);
3168 static inline void print_vma_addr(char *prefix, unsigned long rip)
3173 int vmemmap_remap_free(unsigned long start, unsigned long end,
3174 unsigned long reuse);
3175 int vmemmap_remap_alloc(unsigned long start, unsigned long end,
3176 unsigned long reuse, gfp_t gfp_mask);
3178 void *sparse_buffer_alloc(unsigned long size);
3179 struct page * __populate_section_memmap(unsigned long pfn,
3180 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3181 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3182 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3183 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3184 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3185 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3186 struct vmem_altmap *altmap);
3187 void *vmemmap_alloc_block(unsigned long size, int node);
3189 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3190 struct vmem_altmap *altmap);
3191 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3192 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3193 int node, struct vmem_altmap *altmap);
3194 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3195 struct vmem_altmap *altmap);
3196 void vmemmap_populate_print_last(void);
3197 #ifdef CONFIG_MEMORY_HOTPLUG
3198 void vmemmap_free(unsigned long start, unsigned long end,
3199 struct vmem_altmap *altmap);
3201 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3202 unsigned long nr_pages);
3205 MF_COUNT_INCREASED = 1 << 0,
3206 MF_ACTION_REQUIRED = 1 << 1,
3207 MF_MUST_KILL = 1 << 2,
3208 MF_SOFT_OFFLINE = 1 << 3,
3210 extern int memory_failure(unsigned long pfn, int flags);
3211 extern void memory_failure_queue(unsigned long pfn, int flags);
3212 extern void memory_failure_queue_kick(int cpu);
3213 extern int unpoison_memory(unsigned long pfn);
3214 extern int sysctl_memory_failure_early_kill;
3215 extern int sysctl_memory_failure_recovery;
3216 extern void shake_page(struct page *p);
3217 extern atomic_long_t num_poisoned_pages __read_mostly;
3218 extern int soft_offline_page(unsigned long pfn, int flags);
3220 #ifndef arch_memory_failure
3221 static inline int arch_memory_failure(unsigned long pfn, int flags)
3227 #ifndef arch_is_platform_page
3228 static inline bool arch_is_platform_page(u64 paddr)
3235 * Error handlers for various types of pages.
3238 MF_IGNORED, /* Error: cannot be handled */
3239 MF_FAILED, /* Error: handling failed */
3240 MF_DELAYED, /* Will be handled later */
3241 MF_RECOVERED, /* Successfully recovered */
3244 enum mf_action_page_type {
3246 MF_MSG_KERNEL_HIGH_ORDER,
3248 MF_MSG_DIFFERENT_COMPOUND,
3249 MF_MSG_POISONED_HUGE,
3252 MF_MSG_NON_PMD_HUGE,
3253 MF_MSG_UNMAP_FAILED,
3254 MF_MSG_DIRTY_SWAPCACHE,
3255 MF_MSG_CLEAN_SWAPCACHE,
3256 MF_MSG_DIRTY_MLOCKED_LRU,
3257 MF_MSG_CLEAN_MLOCKED_LRU,
3258 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3259 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3262 MF_MSG_TRUNCATED_LRU,
3270 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3271 extern void clear_huge_page(struct page *page,
3272 unsigned long addr_hint,
3273 unsigned int pages_per_huge_page);
3274 extern void copy_user_huge_page(struct page *dst, struct page *src,
3275 unsigned long addr_hint,
3276 struct vm_area_struct *vma,
3277 unsigned int pages_per_huge_page);
3278 extern long copy_huge_page_from_user(struct page *dst_page,
3279 const void __user *usr_src,
3280 unsigned int pages_per_huge_page,
3281 bool allow_pagefault);
3284 * vma_is_special_huge - Are transhuge page-table entries considered special?
3285 * @vma: Pointer to the struct vm_area_struct to consider
3287 * Whether transhuge page-table entries are considered "special" following
3288 * the definition in vm_normal_page().
3290 * Return: true if transhuge page-table entries should be considered special,
3293 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3295 return vma_is_dax(vma) || (vma->vm_file &&
3296 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3299 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3301 #ifdef CONFIG_DEBUG_PAGEALLOC
3302 extern unsigned int _debug_guardpage_minorder;
3303 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3305 static inline unsigned int debug_guardpage_minorder(void)
3307 return _debug_guardpage_minorder;
3310 static inline bool debug_guardpage_enabled(void)
3312 return static_branch_unlikely(&_debug_guardpage_enabled);
3315 static inline bool page_is_guard(struct page *page)
3317 if (!debug_guardpage_enabled())
3320 return PageGuard(page);
3323 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3324 static inline bool debug_guardpage_enabled(void) { return false; }
3325 static inline bool page_is_guard(struct page *page) { return false; }
3326 #endif /* CONFIG_DEBUG_PAGEALLOC */
3328 #if MAX_NUMNODES > 1
3329 void __init setup_nr_node_ids(void);
3331 static inline void setup_nr_node_ids(void) {}
3334 extern int memcmp_pages(struct page *page1, struct page *page2);
3336 static inline int pages_identical(struct page *page1, struct page *page2)
3338 return !memcmp_pages(page1, page2);
3341 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3342 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3343 pgoff_t first_index, pgoff_t nr,
3344 pgoff_t bitmap_pgoff,
3345 unsigned long *bitmap,
3349 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3350 pgoff_t first_index, pgoff_t nr);
3353 extern int sysctl_nr_trim_pages;
3355 #ifdef CONFIG_PRINTK
3356 void mem_dump_obj(void *object);
3358 static inline void mem_dump_obj(void *object) {}
3362 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3363 * @seals: the seals to check
3364 * @vma: the vma to operate on
3366 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3367 * the vma flags. Return 0 if check pass, or <0 for errors.
3369 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3371 if (seals & F_SEAL_FUTURE_WRITE) {
3373 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3374 * "future write" seal active.
3376 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3380 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3381 * MAP_SHARED and read-only, take care to not allow mprotect to
3382 * revert protections on such mappings. Do this only for shared
3383 * mappings. For private mappings, don't need to mask
3384 * VM_MAYWRITE as we still want them to be COW-writable.
3386 if (vma->vm_flags & VM_SHARED)
3387 vma->vm_flags &= ~(VM_MAYWRITE);
3393 #endif /* __KERNEL__ */
3394 #endif /* _LINUX_MM_H */