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;
41 struct writeback_control;
45 extern int sysctl_page_lock_unfairness;
47 void init_mm_internals(void);
49 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
50 extern unsigned long max_mapnr;
52 static inline void set_max_mapnr(unsigned long limit)
57 static inline void set_max_mapnr(unsigned long limit) { }
60 extern atomic_long_t _totalram_pages;
61 static inline unsigned long totalram_pages(void)
63 return (unsigned long)atomic_long_read(&_totalram_pages);
66 static inline void totalram_pages_inc(void)
68 atomic_long_inc(&_totalram_pages);
71 static inline void totalram_pages_dec(void)
73 atomic_long_dec(&_totalram_pages);
76 static inline void totalram_pages_add(long count)
78 atomic_long_add(count, &_totalram_pages);
81 extern void * high_memory;
82 extern int page_cluster;
85 extern int sysctl_legacy_va_layout;
87 #define sysctl_legacy_va_layout 0
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
91 extern const int mmap_rnd_bits_min;
92 extern const int mmap_rnd_bits_max;
93 extern int mmap_rnd_bits __read_mostly;
95 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
96 extern const int mmap_rnd_compat_bits_min;
97 extern const int mmap_rnd_compat_bits_max;
98 extern int mmap_rnd_compat_bits __read_mostly;
101 #include <asm/page.h>
102 #include <asm/processor.h>
105 * Architectures that support memory tagging (assigning tags to memory regions,
106 * embedding these tags into addresses that point to these memory regions, and
107 * checking that the memory and the pointer tags match on memory accesses)
108 * redefine this macro to strip tags from pointers.
109 * It's defined as noop for architectures that don't support memory tagging.
111 #ifndef untagged_addr
112 #define untagged_addr(addr) (addr)
116 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
120 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
124 #define lm_alias(x) __va(__pa_symbol(x))
128 * To prevent common memory management code establishing
129 * a zero page mapping on a read fault.
130 * This macro should be defined within <asm/pgtable.h>.
131 * s390 does this to prevent multiplexing of hardware bits
132 * related to the physical page in case of virtualization.
134 #ifndef mm_forbids_zeropage
135 #define mm_forbids_zeropage(X) (0)
139 * On some architectures it is expensive to call memset() for small sizes.
140 * If an architecture decides to implement their own version of
141 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 * define their own version of this macro in <asm/pgtable.h>
144 #if BITS_PER_LONG == 64
145 /* This function must be updated when the size of struct page grows above 80
146 * or reduces below 56. The idea that compiler optimizes out switch()
147 * statement, and only leaves move/store instructions. Also the compiler can
148 * combine write statments if they are both assignments and can be reordered,
149 * this can result in several of the writes here being dropped.
151 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152 static inline void __mm_zero_struct_page(struct page *page)
154 unsigned long *_pp = (void *)page;
156 /* Check that struct page is either 56, 64, 72, or 80 bytes */
157 BUILD_BUG_ON(sizeof(struct page) & 7);
158 BUILD_BUG_ON(sizeof(struct page) < 56);
159 BUILD_BUG_ON(sizeof(struct page) > 80);
161 switch (sizeof(struct page)) {
182 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
186 * Default maximum number of active map areas, this limits the number of vmas
187 * per mm struct. Users can overwrite this number by sysctl but there is a
190 * When a program's coredump is generated as ELF format, a section is created
191 * per a vma. In ELF, the number of sections is represented in unsigned short.
192 * This means the number of sections should be smaller than 65535 at coredump.
193 * Because the kernel adds some informative sections to a image of program at
194 * generating coredump, we need some margin. The number of extra sections is
195 * 1-3 now and depends on arch. We use "5" as safe margin, here.
197 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
198 * not a hard limit any more. Although some userspace tools can be surprised by
201 #define MAPCOUNT_ELF_CORE_MARGIN (5)
202 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
204 extern int sysctl_max_map_count;
206 extern unsigned long sysctl_user_reserve_kbytes;
207 extern unsigned long sysctl_admin_reserve_kbytes;
209 extern int sysctl_overcommit_memory;
210 extern int sysctl_overcommit_ratio;
211 extern unsigned long sysctl_overcommit_kbytes;
213 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
215 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
217 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
220 * Any attempt to mark this function as static leads to build failure
221 * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
222 * is referred to by BPF code. This must be visible for error injection.
224 int __add_to_page_cache_locked(struct page *page, struct address_space *mapping,
225 pgoff_t index, gfp_t gfp, void **shadowp);
227 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
228 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
230 #define nth_page(page,n) ((page) + (n))
233 /* to align the pointer to the (next) page boundary */
234 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
236 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
237 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
239 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
242 * Linux kernel virtual memory manager primitives.
243 * The idea being to have a "virtual" mm in the same way
244 * we have a virtual fs - giving a cleaner interface to the
245 * mm details, and allowing different kinds of memory mappings
246 * (from shared memory to executable loading to arbitrary
250 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
251 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
252 void vm_area_free(struct vm_area_struct *);
255 extern struct rb_root nommu_region_tree;
256 extern struct rw_semaphore nommu_region_sem;
258 extern unsigned int kobjsize(const void *objp);
262 * vm_flags in vm_area_struct, see mm_types.h.
263 * When changing, update also include/trace/events/mmflags.h
265 #define VM_NONE 0x00000000
267 #define VM_READ 0x00000001 /* currently active flags */
268 #define VM_WRITE 0x00000002
269 #define VM_EXEC 0x00000004
270 #define VM_SHARED 0x00000008
272 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
273 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
274 #define VM_MAYWRITE 0x00000020
275 #define VM_MAYEXEC 0x00000040
276 #define VM_MAYSHARE 0x00000080
278 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
279 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
280 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
281 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
282 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
284 #define VM_LOCKED 0x00002000
285 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
287 /* Used by sys_madvise() */
288 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
289 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
291 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
292 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
293 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
294 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
295 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
296 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
297 #define VM_SYNC 0x00800000 /* Synchronous page faults */
298 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
299 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
300 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
302 #ifdef CONFIG_MEM_SOFT_DIRTY
303 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
305 # define VM_SOFTDIRTY 0
308 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
309 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
310 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
311 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
313 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
314 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
315 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
316 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
317 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
320 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
321 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
322 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
323 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
324 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
326 #ifdef CONFIG_ARCH_HAS_PKEYS
327 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
328 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
329 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
330 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
331 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
333 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
335 # define VM_PKEY_BIT4 0
337 #endif /* CONFIG_ARCH_HAS_PKEYS */
339 #if defined(CONFIG_X86)
340 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
341 #elif defined(CONFIG_PPC)
342 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
343 #elif defined(CONFIG_PARISC)
344 # define VM_GROWSUP VM_ARCH_1
345 #elif defined(CONFIG_IA64)
346 # define VM_GROWSUP VM_ARCH_1
347 #elif defined(CONFIG_SPARC64)
348 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
349 # define VM_ARCH_CLEAR VM_SPARC_ADI
350 #elif defined(CONFIG_ARM64)
351 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
352 # define VM_ARCH_CLEAR VM_ARM64_BTI
353 #elif !defined(CONFIG_MMU)
354 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
357 #if defined(CONFIG_ARM64_MTE)
358 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
359 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
361 # define VM_MTE VM_NONE
362 # define VM_MTE_ALLOWED VM_NONE
366 # define VM_GROWSUP VM_NONE
369 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
370 # define VM_UFFD_MINOR_BIT 37
371 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
372 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
373 # define VM_UFFD_MINOR VM_NONE
374 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
376 /* Bits set in the VMA until the stack is in its final location */
377 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
379 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
381 /* Common data flag combinations */
382 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
383 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
384 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
385 VM_MAYWRITE | VM_MAYEXEC)
386 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
387 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
389 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
390 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
393 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
394 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
397 #ifdef CONFIG_STACK_GROWSUP
398 #define VM_STACK VM_GROWSUP
400 #define VM_STACK VM_GROWSDOWN
403 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
405 /* VMA basic access permission flags */
406 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
410 * Special vmas that are non-mergable, non-mlock()able.
412 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
414 /* This mask prevents VMA from being scanned with khugepaged */
415 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
417 /* This mask defines which mm->def_flags a process can inherit its parent */
418 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
420 /* This mask is used to clear all the VMA flags used by mlock */
421 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
423 /* Arch-specific flags to clear when updating VM flags on protection change */
424 #ifndef VM_ARCH_CLEAR
425 # define VM_ARCH_CLEAR VM_NONE
427 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
430 * mapping from the currently active vm_flags protection bits (the
431 * low four bits) to a page protection mask..
433 extern pgprot_t protection_map[16];
436 * enum fault_flag - Fault flag definitions.
437 * @FAULT_FLAG_WRITE: Fault was a write fault.
438 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
439 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
440 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
441 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
442 * @FAULT_FLAG_TRIED: The fault has been tried once.
443 * @FAULT_FLAG_USER: The fault originated in userspace.
444 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
445 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
446 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
448 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
449 * whether we would allow page faults to retry by specifying these two
450 * fault flags correctly. Currently there can be three legal combinations:
452 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
453 * this is the first try
455 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
456 * we've already tried at least once
458 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
460 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
461 * be used. Note that page faults can be allowed to retry for multiple times,
462 * in which case we'll have an initial fault with flags (a) then later on
463 * continuous faults with flags (b). We should always try to detect pending
464 * signals before a retry to make sure the continuous page faults can still be
465 * interrupted if necessary.
468 FAULT_FLAG_WRITE = 1 << 0,
469 FAULT_FLAG_MKWRITE = 1 << 1,
470 FAULT_FLAG_ALLOW_RETRY = 1 << 2,
471 FAULT_FLAG_RETRY_NOWAIT = 1 << 3,
472 FAULT_FLAG_KILLABLE = 1 << 4,
473 FAULT_FLAG_TRIED = 1 << 5,
474 FAULT_FLAG_USER = 1 << 6,
475 FAULT_FLAG_REMOTE = 1 << 7,
476 FAULT_FLAG_INSTRUCTION = 1 << 8,
477 FAULT_FLAG_INTERRUPTIBLE = 1 << 9,
481 * The default fault flags that should be used by most of the
482 * arch-specific page fault handlers.
484 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
485 FAULT_FLAG_KILLABLE | \
486 FAULT_FLAG_INTERRUPTIBLE)
489 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
490 * @flags: Fault flags.
492 * This is mostly used for places where we want to try to avoid taking
493 * the mmap_lock for too long a time when waiting for another condition
494 * to change, in which case we can try to be polite to release the
495 * mmap_lock in the first round to avoid potential starvation of other
496 * processes that would also want the mmap_lock.
498 * Return: true if the page fault allows retry and this is the first
499 * attempt of the fault handling; false otherwise.
501 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
503 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
504 (!(flags & FAULT_FLAG_TRIED));
507 #define FAULT_FLAG_TRACE \
508 { FAULT_FLAG_WRITE, "WRITE" }, \
509 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
510 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
511 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
512 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
513 { FAULT_FLAG_TRIED, "TRIED" }, \
514 { FAULT_FLAG_USER, "USER" }, \
515 { FAULT_FLAG_REMOTE, "REMOTE" }, \
516 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
517 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
520 * vm_fault is filled by the pagefault handler and passed to the vma's
521 * ->fault function. The vma's ->fault is responsible for returning a bitmask
522 * of VM_FAULT_xxx flags that give details about how the fault was handled.
524 * MM layer fills up gfp_mask for page allocations but fault handler might
525 * alter it if its implementation requires a different allocation context.
527 * pgoff should be used in favour of virtual_address, if possible.
531 struct vm_area_struct *vma; /* Target VMA */
532 gfp_t gfp_mask; /* gfp mask to be used for allocations */
533 pgoff_t pgoff; /* Logical page offset based on vma */
534 unsigned long address; /* Faulting virtual address */
536 enum fault_flag flags; /* FAULT_FLAG_xxx flags
537 * XXX: should really be 'const' */
538 pmd_t *pmd; /* Pointer to pmd entry matching
540 pud_t *pud; /* Pointer to pud entry matching
543 pte_t orig_pte; /* Value of PTE at the time of fault */
545 struct page *cow_page; /* Page handler may use for COW fault */
546 struct page *page; /* ->fault handlers should return a
547 * page here, unless VM_FAULT_NOPAGE
548 * is set (which is also implied by
551 /* These three entries are valid only while holding ptl lock */
552 pte_t *pte; /* Pointer to pte entry matching
553 * the 'address'. NULL if the page
554 * table hasn't been allocated.
556 spinlock_t *ptl; /* Page table lock.
557 * Protects pte page table if 'pte'
558 * is not NULL, otherwise pmd.
560 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
561 * vm_ops->map_pages() sets up a page
562 * table from atomic context.
563 * do_fault_around() pre-allocates
564 * page table to avoid allocation from
569 /* page entry size for vm->huge_fault() */
570 enum page_entry_size {
577 * These are the virtual MM functions - opening of an area, closing and
578 * unmapping it (needed to keep files on disk up-to-date etc), pointer
579 * to the functions called when a no-page or a wp-page exception occurs.
581 struct vm_operations_struct {
582 void (*open)(struct vm_area_struct * area);
583 void (*close)(struct vm_area_struct * area);
584 /* Called any time before splitting to check if it's allowed */
585 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
586 int (*mremap)(struct vm_area_struct *area);
588 * Called by mprotect() to make driver-specific permission
589 * checks before mprotect() is finalised. The VMA must not
590 * be modified. Returns 0 if eprotect() can proceed.
592 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
593 unsigned long end, unsigned long newflags);
594 vm_fault_t (*fault)(struct vm_fault *vmf);
595 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
596 enum page_entry_size pe_size);
597 vm_fault_t (*map_pages)(struct vm_fault *vmf,
598 pgoff_t start_pgoff, pgoff_t end_pgoff);
599 unsigned long (*pagesize)(struct vm_area_struct * area);
601 /* notification that a previously read-only page is about to become
602 * writable, if an error is returned it will cause a SIGBUS */
603 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
605 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
606 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
608 /* called by access_process_vm when get_user_pages() fails, typically
609 * for use by special VMAs. See also generic_access_phys() for a generic
610 * implementation useful for any iomem mapping.
612 int (*access)(struct vm_area_struct *vma, unsigned long addr,
613 void *buf, int len, int write);
615 /* Called by the /proc/PID/maps code to ask the vma whether it
616 * has a special name. Returning non-NULL will also cause this
617 * vma to be dumped unconditionally. */
618 const char *(*name)(struct vm_area_struct *vma);
622 * set_policy() op must add a reference to any non-NULL @new mempolicy
623 * to hold the policy upon return. Caller should pass NULL @new to
624 * remove a policy and fall back to surrounding context--i.e. do not
625 * install a MPOL_DEFAULT policy, nor the task or system default
628 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
631 * get_policy() op must add reference [mpol_get()] to any policy at
632 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
633 * in mm/mempolicy.c will do this automatically.
634 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
635 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
636 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
637 * must return NULL--i.e., do not "fallback" to task or system default
640 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
644 * Called by vm_normal_page() for special PTEs to find the
645 * page for @addr. This is useful if the default behavior
646 * (using pte_page()) would not find the correct page.
648 struct page *(*find_special_page)(struct vm_area_struct *vma,
652 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
654 static const struct vm_operations_struct dummy_vm_ops = {};
656 memset(vma, 0, sizeof(*vma));
658 vma->vm_ops = &dummy_vm_ops;
659 INIT_LIST_HEAD(&vma->anon_vma_chain);
662 static inline void vma_set_anonymous(struct vm_area_struct *vma)
667 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
672 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
674 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
679 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
680 VM_STACK_INCOMPLETE_SETUP)
686 static inline bool vma_is_foreign(struct vm_area_struct *vma)
691 if (current->mm != vma->vm_mm)
697 static inline bool vma_is_accessible(struct vm_area_struct *vma)
699 return vma->vm_flags & VM_ACCESS_FLAGS;
704 * The vma_is_shmem is not inline because it is used only by slow
705 * paths in userfault.
707 bool vma_is_shmem(struct vm_area_struct *vma);
709 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
712 int vma_is_stack_for_current(struct vm_area_struct *vma);
714 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
715 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
720 #include <linux/huge_mm.h>
723 * Methods to modify the page usage count.
725 * What counts for a page usage:
726 * - cache mapping (page->mapping)
727 * - private data (page->private)
728 * - page mapped in a task's page tables, each mapping
729 * is counted separately
731 * Also, many kernel routines increase the page count before a critical
732 * routine so they can be sure the page doesn't go away from under them.
736 * Drop a ref, return true if the refcount fell to zero (the page has no users)
738 static inline int put_page_testzero(struct page *page)
740 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
741 return page_ref_dec_and_test(page);
745 * Try to grab a ref unless the page has a refcount of zero, return false if
747 * This can be called when MMU is off so it must not access
748 * any of the virtual mappings.
750 static inline int get_page_unless_zero(struct page *page)
752 return page_ref_add_unless(page, 1, 0);
755 extern int page_is_ram(unsigned long pfn);
763 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
766 /* Support for virtually mapped pages */
767 struct page *vmalloc_to_page(const void *addr);
768 unsigned long vmalloc_to_pfn(const void *addr);
771 * Determine if an address is within the vmalloc range
773 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
774 * is no special casing required.
777 #ifndef is_ioremap_addr
778 #define is_ioremap_addr(x) is_vmalloc_addr(x)
782 extern bool is_vmalloc_addr(const void *x);
783 extern int is_vmalloc_or_module_addr(const void *x);
785 static inline bool is_vmalloc_addr(const void *x)
789 static inline int is_vmalloc_or_module_addr(const void *x)
795 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
796 static inline void *kvmalloc(size_t size, gfp_t flags)
798 return kvmalloc_node(size, flags, NUMA_NO_NODE);
800 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
802 return kvmalloc_node(size, flags | __GFP_ZERO, node);
804 static inline void *kvzalloc(size_t size, gfp_t flags)
806 return kvmalloc(size, flags | __GFP_ZERO);
809 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
813 if (unlikely(check_mul_overflow(n, size, &bytes)))
816 return kvmalloc(bytes, flags);
819 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
821 return kvmalloc_array(n, size, flags | __GFP_ZERO);
824 extern void kvfree(const void *addr);
825 extern void kvfree_sensitive(const void *addr, size_t len);
827 static inline int head_compound_mapcount(struct page *head)
829 return atomic_read(compound_mapcount_ptr(head)) + 1;
833 * Mapcount of compound page as a whole, does not include mapped sub-pages.
835 * Must be called only for compound pages or any their tail sub-pages.
837 static inline int compound_mapcount(struct page *page)
839 VM_BUG_ON_PAGE(!PageCompound(page), page);
840 page = compound_head(page);
841 return head_compound_mapcount(page);
845 * The atomic page->_mapcount, starts from -1: so that transitions
846 * both from it and to it can be tracked, using atomic_inc_and_test
847 * and atomic_add_negative(-1).
849 static inline void page_mapcount_reset(struct page *page)
851 atomic_set(&(page)->_mapcount, -1);
854 int __page_mapcount(struct page *page);
857 * Mapcount of 0-order page; when compound sub-page, includes
858 * compound_mapcount().
860 * Result is undefined for pages which cannot be mapped into userspace.
861 * For example SLAB or special types of pages. See function page_has_type().
862 * They use this place in struct page differently.
864 static inline int page_mapcount(struct page *page)
866 if (unlikely(PageCompound(page)))
867 return __page_mapcount(page);
868 return atomic_read(&page->_mapcount) + 1;
871 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
872 int total_mapcount(struct page *page);
873 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
875 static inline int total_mapcount(struct page *page)
877 return page_mapcount(page);
879 static inline int page_trans_huge_mapcount(struct page *page,
882 int mapcount = page_mapcount(page);
884 *total_mapcount = mapcount;
889 static inline struct page *virt_to_head_page(const void *x)
891 struct page *page = virt_to_page(x);
893 return compound_head(page);
896 void __put_page(struct page *page);
898 void put_pages_list(struct list_head *pages);
900 void split_page(struct page *page, unsigned int order);
903 * Compound pages have a destructor function. Provide a
904 * prototype for that function and accessor functions.
905 * These are _only_ valid on the head of a compound page.
907 typedef void compound_page_dtor(struct page *);
909 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
910 enum compound_dtor_id {
913 #ifdef CONFIG_HUGETLB_PAGE
916 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
921 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
923 static inline void set_compound_page_dtor(struct page *page,
924 enum compound_dtor_id compound_dtor)
926 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
927 page[1].compound_dtor = compound_dtor;
930 static inline void destroy_compound_page(struct page *page)
932 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
933 compound_page_dtors[page[1].compound_dtor](page);
936 static inline unsigned int compound_order(struct page *page)
940 return page[1].compound_order;
943 static inline bool hpage_pincount_available(struct page *page)
946 * Can the page->hpage_pinned_refcount field be used? That field is in
947 * the 3rd page of the compound page, so the smallest (2-page) compound
948 * pages cannot support it.
950 page = compound_head(page);
951 return PageCompound(page) && compound_order(page) > 1;
954 static inline int head_compound_pincount(struct page *head)
956 return atomic_read(compound_pincount_ptr(head));
959 static inline int compound_pincount(struct page *page)
961 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
962 page = compound_head(page);
963 return head_compound_pincount(page);
966 static inline void set_compound_order(struct page *page, unsigned int order)
968 page[1].compound_order = order;
969 page[1].compound_nr = 1U << order;
972 /* Returns the number of pages in this potentially compound page. */
973 static inline unsigned long compound_nr(struct page *page)
977 return page[1].compound_nr;
980 /* Returns the number of bytes in this potentially compound page. */
981 static inline unsigned long page_size(struct page *page)
983 return PAGE_SIZE << compound_order(page);
986 /* Returns the number of bits needed for the number of bytes in a page */
987 static inline unsigned int page_shift(struct page *page)
989 return PAGE_SHIFT + compound_order(page);
992 void free_compound_page(struct page *page);
996 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
997 * servicing faults for write access. In the normal case, do always want
998 * pte_mkwrite. But get_user_pages can cause write faults for mappings
999 * that do not have writing enabled, when used by access_process_vm.
1001 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1003 if (likely(vma->vm_flags & VM_WRITE))
1004 pte = pte_mkwrite(pte);
1008 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1009 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1011 vm_fault_t finish_fault(struct vm_fault *vmf);
1012 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1016 * Multiple processes may "see" the same page. E.g. for untouched
1017 * mappings of /dev/null, all processes see the same page full of
1018 * zeroes, and text pages of executables and shared libraries have
1019 * only one copy in memory, at most, normally.
1021 * For the non-reserved pages, page_count(page) denotes a reference count.
1022 * page_count() == 0 means the page is free. page->lru is then used for
1023 * freelist management in the buddy allocator.
1024 * page_count() > 0 means the page has been allocated.
1026 * Pages are allocated by the slab allocator in order to provide memory
1027 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1028 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1029 * unless a particular usage is carefully commented. (the responsibility of
1030 * freeing the kmalloc memory is the caller's, of course).
1032 * A page may be used by anyone else who does a __get_free_page().
1033 * In this case, page_count still tracks the references, and should only
1034 * be used through the normal accessor functions. The top bits of page->flags
1035 * and page->virtual store page management information, but all other fields
1036 * are unused and could be used privately, carefully. The management of this
1037 * page is the responsibility of the one who allocated it, and those who have
1038 * subsequently been given references to it.
1040 * The other pages (we may call them "pagecache pages") are completely
1041 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1042 * The following discussion applies only to them.
1044 * A pagecache page contains an opaque `private' member, which belongs to the
1045 * page's address_space. Usually, this is the address of a circular list of
1046 * the page's disk buffers. PG_private must be set to tell the VM to call
1047 * into the filesystem to release these pages.
1049 * A page may belong to an inode's memory mapping. In this case, page->mapping
1050 * is the pointer to the inode, and page->index is the file offset of the page,
1051 * in units of PAGE_SIZE.
1053 * If pagecache pages are not associated with an inode, they are said to be
1054 * anonymous pages. These may become associated with the swapcache, and in that
1055 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1057 * In either case (swapcache or inode backed), the pagecache itself holds one
1058 * reference to the page. Setting PG_private should also increment the
1059 * refcount. The each user mapping also has a reference to the page.
1061 * The pagecache pages are stored in a per-mapping radix tree, which is
1062 * rooted at mapping->i_pages, and indexed by offset.
1063 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1064 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1066 * All pagecache pages may be subject to I/O:
1067 * - inode pages may need to be read from disk,
1068 * - inode pages which have been modified and are MAP_SHARED may need
1069 * to be written back to the inode on disk,
1070 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1071 * modified may need to be swapped out to swap space and (later) to be read
1076 * The zone field is never updated after free_area_init_core()
1077 * sets it, so none of the operations on it need to be atomic.
1080 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1081 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1082 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1083 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1084 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1085 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1088 * Define the bit shifts to access each section. For non-existent
1089 * sections we define the shift as 0; that plus a 0 mask ensures
1090 * the compiler will optimise away reference to them.
1092 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1093 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1094 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1095 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1096 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1098 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1099 #ifdef NODE_NOT_IN_PAGE_FLAGS
1100 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1101 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1102 SECTIONS_PGOFF : ZONES_PGOFF)
1104 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1105 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1106 NODES_PGOFF : ZONES_PGOFF)
1109 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1111 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1112 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1113 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1114 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1115 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1116 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1118 static inline enum zone_type page_zonenum(const struct page *page)
1120 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1121 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1124 #ifdef CONFIG_ZONE_DEVICE
1125 static inline bool is_zone_device_page(const struct page *page)
1127 return page_zonenum(page) == ZONE_DEVICE;
1129 extern void memmap_init_zone_device(struct zone *, unsigned long,
1130 unsigned long, struct dev_pagemap *);
1132 static inline bool is_zone_device_page(const struct page *page)
1138 static inline bool is_zone_movable_page(const struct page *page)
1140 return page_zonenum(page) == ZONE_MOVABLE;
1143 #ifdef CONFIG_DEV_PAGEMAP_OPS
1144 void free_devmap_managed_page(struct page *page);
1145 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1147 static inline bool page_is_devmap_managed(struct page *page)
1149 if (!static_branch_unlikely(&devmap_managed_key))
1151 if (!is_zone_device_page(page))
1153 switch (page->pgmap->type) {
1154 case MEMORY_DEVICE_PRIVATE:
1155 case MEMORY_DEVICE_FS_DAX:
1163 void put_devmap_managed_page(struct page *page);
1165 #else /* CONFIG_DEV_PAGEMAP_OPS */
1166 static inline bool page_is_devmap_managed(struct page *page)
1171 static inline void put_devmap_managed_page(struct page *page)
1174 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1176 static inline bool is_device_private_page(const struct page *page)
1178 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1179 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1180 is_zone_device_page(page) &&
1181 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1184 static inline bool is_pci_p2pdma_page(const struct page *page)
1186 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1187 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1188 is_zone_device_page(page) &&
1189 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1192 /* 127: arbitrary random number, small enough to assemble well */
1193 #define page_ref_zero_or_close_to_overflow(page) \
1194 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1196 static inline void get_page(struct page *page)
1198 page = compound_head(page);
1200 * Getting a normal page or the head of a compound page
1201 * requires to already have an elevated page->_refcount.
1203 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1207 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1208 __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs,
1209 unsigned int flags);
1212 static inline __must_check bool try_get_page(struct page *page)
1214 page = compound_head(page);
1215 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1221 static inline void put_page(struct page *page)
1223 page = compound_head(page);
1226 * For devmap managed pages we need to catch refcount transition from
1227 * 2 to 1, when refcount reach one it means the page is free and we
1228 * need to inform the device driver through callback. See
1229 * include/linux/memremap.h and HMM for details.
1231 if (page_is_devmap_managed(page)) {
1232 put_devmap_managed_page(page);
1236 if (put_page_testzero(page))
1241 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1242 * the page's refcount so that two separate items are tracked: the original page
1243 * reference count, and also a new count of how many pin_user_pages() calls were
1244 * made against the page. ("gup-pinned" is another term for the latter).
1246 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1247 * distinct from normal pages. As such, the unpin_user_page() call (and its
1248 * variants) must be used in order to release gup-pinned pages.
1252 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1253 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1254 * simpler, due to the fact that adding an even power of two to the page
1255 * refcount has the effect of using only the upper N bits, for the code that
1256 * counts up using the bias value. This means that the lower bits are left for
1257 * the exclusive use of the original code that increments and decrements by one
1258 * (or at least, by much smaller values than the bias value).
1260 * Of course, once the lower bits overflow into the upper bits (and this is
1261 * OK, because subtraction recovers the original values), then visual inspection
1262 * no longer suffices to directly view the separate counts. However, for normal
1263 * applications that don't have huge page reference counts, this won't be an
1266 * Locking: the lockless algorithm described in page_cache_get_speculative()
1267 * and page_cache_gup_pin_speculative() provides safe operation for
1268 * get_user_pages and page_mkclean and other calls that race to set up page
1271 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1273 void unpin_user_page(struct page *page);
1274 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1276 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1278 void unpin_user_pages(struct page **pages, unsigned long npages);
1281 * page_maybe_dma_pinned - Report if a page is pinned for DMA.
1284 * This function checks if a page has been pinned via a call to
1285 * a function in the pin_user_pages() family.
1287 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1288 * because it means "definitely not pinned for DMA", but true means "probably
1289 * pinned for DMA, but possibly a false positive due to having at least
1290 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1292 * False positives are OK, because: a) it's unlikely for a page to get that many
1293 * refcounts, and b) all the callers of this routine are expected to be able to
1294 * deal gracefully with a false positive.
1296 * For huge pages, the result will be exactly correct. That's because we have
1297 * more tracking data available: the 3rd struct page in the compound page is
1298 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1301 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1303 * Return: True, if it is likely that the page has been "dma-pinned".
1304 * False, if the page is definitely not dma-pinned.
1306 static inline bool page_maybe_dma_pinned(struct page *page)
1308 if (hpage_pincount_available(page))
1309 return compound_pincount(page) > 0;
1312 * page_ref_count() is signed. If that refcount overflows, then
1313 * page_ref_count() returns a negative value, and callers will avoid
1314 * further incrementing the refcount.
1316 * Here, for that overflow case, use the signed bit to count a little
1317 * bit higher via unsigned math, and thus still get an accurate result.
1319 return ((unsigned int)page_ref_count(compound_head(page))) >=
1320 GUP_PIN_COUNTING_BIAS;
1323 static inline bool is_cow_mapping(vm_flags_t flags)
1325 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1329 * This should most likely only be called during fork() to see whether we
1330 * should break the cow immediately for a page on the src mm.
1332 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1335 if (!is_cow_mapping(vma->vm_flags))
1338 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1341 return page_maybe_dma_pinned(page);
1344 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1345 #define SECTION_IN_PAGE_FLAGS
1349 * The identification function is mainly used by the buddy allocator for
1350 * determining if two pages could be buddies. We are not really identifying
1351 * the zone since we could be using the section number id if we do not have
1352 * node id available in page flags.
1353 * We only guarantee that it will return the same value for two combinable
1356 static inline int page_zone_id(struct page *page)
1358 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1361 #ifdef NODE_NOT_IN_PAGE_FLAGS
1362 extern int page_to_nid(const struct page *page);
1364 static inline int page_to_nid(const struct page *page)
1366 struct page *p = (struct page *)page;
1368 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1372 #ifdef CONFIG_NUMA_BALANCING
1373 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1375 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1378 static inline int cpupid_to_pid(int cpupid)
1380 return cpupid & LAST__PID_MASK;
1383 static inline int cpupid_to_cpu(int cpupid)
1385 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1388 static inline int cpupid_to_nid(int cpupid)
1390 return cpu_to_node(cpupid_to_cpu(cpupid));
1393 static inline bool cpupid_pid_unset(int cpupid)
1395 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1398 static inline bool cpupid_cpu_unset(int cpupid)
1400 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1403 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1405 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1408 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1409 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1410 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1412 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1415 static inline int page_cpupid_last(struct page *page)
1417 return page->_last_cpupid;
1419 static inline void page_cpupid_reset_last(struct page *page)
1421 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1424 static inline int page_cpupid_last(struct page *page)
1426 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1429 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1431 static inline void page_cpupid_reset_last(struct page *page)
1433 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1435 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1436 #else /* !CONFIG_NUMA_BALANCING */
1437 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1439 return page_to_nid(page); /* XXX */
1442 static inline int page_cpupid_last(struct page *page)
1444 return page_to_nid(page); /* XXX */
1447 static inline int cpupid_to_nid(int cpupid)
1452 static inline int cpupid_to_pid(int cpupid)
1457 static inline int cpupid_to_cpu(int cpupid)
1462 static inline int cpu_pid_to_cpupid(int nid, int pid)
1467 static inline bool cpupid_pid_unset(int cpupid)
1472 static inline void page_cpupid_reset_last(struct page *page)
1476 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1480 #endif /* CONFIG_NUMA_BALANCING */
1482 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1485 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1486 * setting tags for all pages to native kernel tag value 0xff, as the default
1487 * value 0x00 maps to 0xff.
1490 static inline u8 page_kasan_tag(const struct page *page)
1494 if (kasan_enabled()) {
1495 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1502 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1504 if (kasan_enabled()) {
1506 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1507 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1511 static inline void page_kasan_tag_reset(struct page *page)
1513 if (kasan_enabled())
1514 page_kasan_tag_set(page, 0xff);
1517 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1519 static inline u8 page_kasan_tag(const struct page *page)
1524 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1525 static inline void page_kasan_tag_reset(struct page *page) { }
1527 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1529 static inline struct zone *page_zone(const struct page *page)
1531 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1534 static inline pg_data_t *page_pgdat(const struct page *page)
1536 return NODE_DATA(page_to_nid(page));
1539 #ifdef SECTION_IN_PAGE_FLAGS
1540 static inline void set_page_section(struct page *page, unsigned long section)
1542 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1543 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1546 static inline unsigned long page_to_section(const struct page *page)
1548 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1552 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1553 #ifdef CONFIG_MIGRATION
1554 static inline bool is_pinnable_page(struct page *page)
1556 return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) ||
1557 is_zero_pfn(page_to_pfn(page));
1560 static inline bool is_pinnable_page(struct page *page)
1566 static inline void set_page_zone(struct page *page, enum zone_type zone)
1568 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1569 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1572 static inline void set_page_node(struct page *page, unsigned long node)
1574 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1575 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1578 static inline void set_page_links(struct page *page, enum zone_type zone,
1579 unsigned long node, unsigned long pfn)
1581 set_page_zone(page, zone);
1582 set_page_node(page, node);
1583 #ifdef SECTION_IN_PAGE_FLAGS
1584 set_page_section(page, pfn_to_section_nr(pfn));
1589 * Some inline functions in vmstat.h depend on page_zone()
1591 #include <linux/vmstat.h>
1593 static __always_inline void *lowmem_page_address(const struct page *page)
1595 return page_to_virt(page);
1598 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1599 #define HASHED_PAGE_VIRTUAL
1602 #if defined(WANT_PAGE_VIRTUAL)
1603 static inline void *page_address(const struct page *page)
1605 return page->virtual;
1607 static inline void set_page_address(struct page *page, void *address)
1609 page->virtual = address;
1611 #define page_address_init() do { } while(0)
1614 #if defined(HASHED_PAGE_VIRTUAL)
1615 void *page_address(const struct page *page);
1616 void set_page_address(struct page *page, void *virtual);
1617 void page_address_init(void);
1620 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1621 #define page_address(page) lowmem_page_address(page)
1622 #define set_page_address(page, address) do { } while(0)
1623 #define page_address_init() do { } while(0)
1626 extern void *page_rmapping(struct page *page);
1627 extern struct anon_vma *page_anon_vma(struct page *page);
1628 extern struct address_space *page_mapping(struct page *page);
1630 extern struct address_space *__page_file_mapping(struct page *);
1633 struct address_space *page_file_mapping(struct page *page)
1635 if (unlikely(PageSwapCache(page)))
1636 return __page_file_mapping(page);
1638 return page->mapping;
1641 extern pgoff_t __page_file_index(struct page *page);
1644 * Return the pagecache index of the passed page. Regular pagecache pages
1645 * use ->index whereas swapcache pages use swp_offset(->private)
1647 static inline pgoff_t page_index(struct page *page)
1649 if (unlikely(PageSwapCache(page)))
1650 return __page_file_index(page);
1654 bool page_mapped(struct page *page);
1655 struct address_space *page_mapping(struct page *page);
1658 * Return true only if the page has been allocated with
1659 * ALLOC_NO_WATERMARKS and the low watermark was not
1660 * met implying that the system is under some pressure.
1662 static inline bool page_is_pfmemalloc(const struct page *page)
1665 * Page index cannot be this large so this must be
1666 * a pfmemalloc page.
1668 return page->index == -1UL;
1672 * Only to be called by the page allocator on a freshly allocated
1675 static inline void set_page_pfmemalloc(struct page *page)
1680 static inline void clear_page_pfmemalloc(struct page *page)
1686 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1688 extern void pagefault_out_of_memory(void);
1690 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1691 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1694 * Flags passed to show_mem() and show_free_areas() to suppress output in
1697 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1699 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1702 extern bool can_do_mlock(void);
1704 static inline bool can_do_mlock(void) { return false; }
1706 extern int user_shm_lock(size_t, struct ucounts *);
1707 extern void user_shm_unlock(size_t, struct ucounts *);
1710 * Parameter block passed down to zap_pte_range in exceptional cases.
1712 struct zap_details {
1713 struct address_space *check_mapping; /* Check page->mapping if set */
1714 pgoff_t first_index; /* Lowest page->index to unmap */
1715 pgoff_t last_index; /* Highest page->index to unmap */
1716 struct page *single_page; /* Locked page to be unmapped */
1719 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1721 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1724 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1725 unsigned long size);
1726 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1727 unsigned long size);
1728 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1729 unsigned long start, unsigned long end);
1731 struct mmu_notifier_range;
1733 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1734 unsigned long end, unsigned long floor, unsigned long ceiling);
1736 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1737 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1738 struct mmu_notifier_range *range, pte_t **ptepp,
1739 pmd_t **pmdpp, spinlock_t **ptlp);
1740 int follow_pte(struct mm_struct *mm, unsigned long address,
1741 pte_t **ptepp, spinlock_t **ptlp);
1742 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1743 unsigned long *pfn);
1744 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1745 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1746 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1747 void *buf, int len, int write);
1749 extern void truncate_pagecache(struct inode *inode, loff_t new);
1750 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1751 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1752 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1753 int truncate_inode_page(struct address_space *mapping, struct page *page);
1754 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1755 int invalidate_inode_page(struct page *page);
1758 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1759 unsigned long address, unsigned int flags,
1760 struct pt_regs *regs);
1761 extern int fixup_user_fault(struct mm_struct *mm,
1762 unsigned long address, unsigned int fault_flags,
1764 void unmap_mapping_page(struct page *page);
1765 void unmap_mapping_pages(struct address_space *mapping,
1766 pgoff_t start, pgoff_t nr, bool even_cows);
1767 void unmap_mapping_range(struct address_space *mapping,
1768 loff_t const holebegin, loff_t const holelen, int even_cows);
1770 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1771 unsigned long address, unsigned int flags,
1772 struct pt_regs *regs)
1774 /* should never happen if there's no MMU */
1776 return VM_FAULT_SIGBUS;
1778 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1779 unsigned int fault_flags, bool *unlocked)
1781 /* should never happen if there's no MMU */
1785 static inline void unmap_mapping_page(struct page *page) { }
1786 static inline void unmap_mapping_pages(struct address_space *mapping,
1787 pgoff_t start, pgoff_t nr, bool even_cows) { }
1788 static inline void unmap_mapping_range(struct address_space *mapping,
1789 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1792 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1793 loff_t const holebegin, loff_t const holelen)
1795 unmap_mapping_range(mapping, holebegin, holelen, 0);
1798 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1799 void *buf, int len, unsigned int gup_flags);
1800 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1801 void *buf, int len, unsigned int gup_flags);
1802 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1803 void *buf, int len, unsigned int gup_flags);
1805 long get_user_pages_remote(struct mm_struct *mm,
1806 unsigned long start, unsigned long nr_pages,
1807 unsigned int gup_flags, struct page **pages,
1808 struct vm_area_struct **vmas, int *locked);
1809 long pin_user_pages_remote(struct mm_struct *mm,
1810 unsigned long start, unsigned long nr_pages,
1811 unsigned int gup_flags, struct page **pages,
1812 struct vm_area_struct **vmas, int *locked);
1813 long get_user_pages(unsigned long start, unsigned long nr_pages,
1814 unsigned int gup_flags, struct page **pages,
1815 struct vm_area_struct **vmas);
1816 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1817 unsigned int gup_flags, struct page **pages,
1818 struct vm_area_struct **vmas);
1819 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1820 unsigned int gup_flags, struct page **pages, int *locked);
1821 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1822 unsigned int gup_flags, struct page **pages, int *locked);
1823 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1824 struct page **pages, unsigned int gup_flags);
1825 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1826 struct page **pages, unsigned int gup_flags);
1828 int get_user_pages_fast(unsigned long start, int nr_pages,
1829 unsigned int gup_flags, struct page **pages);
1830 int pin_user_pages_fast(unsigned long start, int nr_pages,
1831 unsigned int gup_flags, struct page **pages);
1833 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1834 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1835 struct task_struct *task, bool bypass_rlim);
1838 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1839 struct page **pages);
1840 int get_kernel_page(unsigned long start, int write, struct page **pages);
1841 struct page *get_dump_page(unsigned long addr);
1843 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1844 extern void do_invalidatepage(struct page *page, unsigned int offset,
1845 unsigned int length);
1847 int redirty_page_for_writepage(struct writeback_control *wbc,
1849 void account_page_cleaned(struct page *page, struct address_space *mapping,
1850 struct bdi_writeback *wb);
1851 int set_page_dirty(struct page *page);
1852 int set_page_dirty_lock(struct page *page);
1853 void __cancel_dirty_page(struct page *page);
1854 static inline void cancel_dirty_page(struct page *page)
1856 /* Avoid atomic ops, locking, etc. when not actually needed. */
1857 if (PageDirty(page))
1858 __cancel_dirty_page(page);
1860 int clear_page_dirty_for_io(struct page *page);
1862 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1864 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1865 unsigned long old_addr, struct vm_area_struct *new_vma,
1866 unsigned long new_addr, unsigned long len,
1867 bool need_rmap_locks);
1870 * Flags used by change_protection(). For now we make it a bitmap so
1871 * that we can pass in multiple flags just like parameters. However
1872 * for now all the callers are only use one of the flags at the same
1875 /* Whether we should allow dirty bit accounting */
1876 #define MM_CP_DIRTY_ACCT (1UL << 0)
1877 /* Whether this protection change is for NUMA hints */
1878 #define MM_CP_PROT_NUMA (1UL << 1)
1879 /* Whether this change is for write protecting */
1880 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1881 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1882 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1883 MM_CP_UFFD_WP_RESOLVE)
1885 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1886 unsigned long end, pgprot_t newprot,
1887 unsigned long cp_flags);
1888 extern int mprotect_fixup(struct vm_area_struct *vma,
1889 struct vm_area_struct **pprev, unsigned long start,
1890 unsigned long end, unsigned long newflags);
1893 * doesn't attempt to fault and will return short.
1895 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1896 unsigned int gup_flags, struct page **pages);
1897 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1898 unsigned int gup_flags, struct page **pages);
1900 static inline bool get_user_page_fast_only(unsigned long addr,
1901 unsigned int gup_flags, struct page **pagep)
1903 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1906 * per-process(per-mm_struct) statistics.
1908 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1910 long val = atomic_long_read(&mm->rss_stat.count[member]);
1912 #ifdef SPLIT_RSS_COUNTING
1914 * counter is updated in asynchronous manner and may go to minus.
1915 * But it's never be expected number for users.
1920 return (unsigned long)val;
1923 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1925 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1927 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1929 mm_trace_rss_stat(mm, member, count);
1932 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1934 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1936 mm_trace_rss_stat(mm, member, count);
1939 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1941 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1943 mm_trace_rss_stat(mm, member, count);
1946 /* Optimized variant when page is already known not to be PageAnon */
1947 static inline int mm_counter_file(struct page *page)
1949 if (PageSwapBacked(page))
1950 return MM_SHMEMPAGES;
1951 return MM_FILEPAGES;
1954 static inline int mm_counter(struct page *page)
1957 return MM_ANONPAGES;
1958 return mm_counter_file(page);
1961 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1963 return get_mm_counter(mm, MM_FILEPAGES) +
1964 get_mm_counter(mm, MM_ANONPAGES) +
1965 get_mm_counter(mm, MM_SHMEMPAGES);
1968 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1970 return max(mm->hiwater_rss, get_mm_rss(mm));
1973 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1975 return max(mm->hiwater_vm, mm->total_vm);
1978 static inline void update_hiwater_rss(struct mm_struct *mm)
1980 unsigned long _rss = get_mm_rss(mm);
1982 if ((mm)->hiwater_rss < _rss)
1983 (mm)->hiwater_rss = _rss;
1986 static inline void update_hiwater_vm(struct mm_struct *mm)
1988 if (mm->hiwater_vm < mm->total_vm)
1989 mm->hiwater_vm = mm->total_vm;
1992 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1994 mm->hiwater_rss = get_mm_rss(mm);
1997 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1998 struct mm_struct *mm)
2000 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2002 if (*maxrss < hiwater_rss)
2003 *maxrss = hiwater_rss;
2006 #if defined(SPLIT_RSS_COUNTING)
2007 void sync_mm_rss(struct mm_struct *mm);
2009 static inline void sync_mm_rss(struct mm_struct *mm)
2014 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2015 static inline int pte_special(pte_t pte)
2020 static inline pte_t pte_mkspecial(pte_t pte)
2026 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2027 static inline int pte_devmap(pte_t pte)
2033 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2035 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2037 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2041 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2045 #ifdef __PAGETABLE_P4D_FOLDED
2046 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2047 unsigned long address)
2052 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2055 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2056 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2057 unsigned long address)
2061 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2062 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2065 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2067 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2069 if (mm_pud_folded(mm))
2071 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2074 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2076 if (mm_pud_folded(mm))
2078 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2082 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2083 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2084 unsigned long address)
2089 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2090 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2093 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2095 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2097 if (mm_pmd_folded(mm))
2099 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2102 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2104 if (mm_pmd_folded(mm))
2106 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2111 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2113 atomic_long_set(&mm->pgtables_bytes, 0);
2116 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2118 return atomic_long_read(&mm->pgtables_bytes);
2121 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2123 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2126 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2128 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2132 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2133 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2138 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2139 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2142 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2143 int __pte_alloc_kernel(pmd_t *pmd);
2145 #if defined(CONFIG_MMU)
2147 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2148 unsigned long address)
2150 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2151 NULL : p4d_offset(pgd, address);
2154 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2155 unsigned long address)
2157 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2158 NULL : pud_offset(p4d, address);
2161 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2163 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2164 NULL: pmd_offset(pud, address);
2166 #endif /* CONFIG_MMU */
2168 #if USE_SPLIT_PTE_PTLOCKS
2169 #if ALLOC_SPLIT_PTLOCKS
2170 void __init ptlock_cache_init(void);
2171 extern bool ptlock_alloc(struct page *page);
2172 extern void ptlock_free(struct page *page);
2174 static inline spinlock_t *ptlock_ptr(struct page *page)
2178 #else /* ALLOC_SPLIT_PTLOCKS */
2179 static inline void ptlock_cache_init(void)
2183 static inline bool ptlock_alloc(struct page *page)
2188 static inline void ptlock_free(struct page *page)
2192 static inline spinlock_t *ptlock_ptr(struct page *page)
2196 #endif /* ALLOC_SPLIT_PTLOCKS */
2198 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2200 return ptlock_ptr(pmd_page(*pmd));
2203 static inline bool ptlock_init(struct page *page)
2206 * prep_new_page() initialize page->private (and therefore page->ptl)
2207 * with 0. Make sure nobody took it in use in between.
2209 * It can happen if arch try to use slab for page table allocation:
2210 * slab code uses page->slab_cache, which share storage with page->ptl.
2212 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2213 if (!ptlock_alloc(page))
2215 spin_lock_init(ptlock_ptr(page));
2219 #else /* !USE_SPLIT_PTE_PTLOCKS */
2221 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2223 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2225 return &mm->page_table_lock;
2227 static inline void ptlock_cache_init(void) {}
2228 static inline bool ptlock_init(struct page *page) { return true; }
2229 static inline void ptlock_free(struct page *page) {}
2230 #endif /* USE_SPLIT_PTE_PTLOCKS */
2232 static inline void pgtable_init(void)
2234 ptlock_cache_init();
2235 pgtable_cache_init();
2238 static inline bool pgtable_pte_page_ctor(struct page *page)
2240 if (!ptlock_init(page))
2242 __SetPageTable(page);
2243 inc_lruvec_page_state(page, NR_PAGETABLE);
2247 static inline void pgtable_pte_page_dtor(struct page *page)
2250 __ClearPageTable(page);
2251 dec_lruvec_page_state(page, NR_PAGETABLE);
2254 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2256 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2257 pte_t *__pte = pte_offset_map(pmd, address); \
2263 #define pte_unmap_unlock(pte, ptl) do { \
2268 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2270 #define pte_alloc_map(mm, pmd, address) \
2271 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2273 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2274 (pte_alloc(mm, pmd) ? \
2275 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2277 #define pte_alloc_kernel(pmd, address) \
2278 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2279 NULL: pte_offset_kernel(pmd, address))
2281 #if USE_SPLIT_PMD_PTLOCKS
2283 static struct page *pmd_to_page(pmd_t *pmd)
2285 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2286 return virt_to_page((void *)((unsigned long) pmd & mask));
2289 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2291 return ptlock_ptr(pmd_to_page(pmd));
2294 static inline bool pmd_ptlock_init(struct page *page)
2296 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2297 page->pmd_huge_pte = NULL;
2299 return ptlock_init(page);
2302 static inline void pmd_ptlock_free(struct page *page)
2304 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2305 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2310 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2314 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2316 return &mm->page_table_lock;
2319 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2320 static inline void pmd_ptlock_free(struct page *page) {}
2322 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2326 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2328 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2333 static inline bool pgtable_pmd_page_ctor(struct page *page)
2335 if (!pmd_ptlock_init(page))
2337 __SetPageTable(page);
2338 inc_lruvec_page_state(page, NR_PAGETABLE);
2342 static inline void pgtable_pmd_page_dtor(struct page *page)
2344 pmd_ptlock_free(page);
2345 __ClearPageTable(page);
2346 dec_lruvec_page_state(page, NR_PAGETABLE);
2350 * No scalability reason to split PUD locks yet, but follow the same pattern
2351 * as the PMD locks to make it easier if we decide to. The VM should not be
2352 * considered ready to switch to split PUD locks yet; there may be places
2353 * which need to be converted from page_table_lock.
2355 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2357 return &mm->page_table_lock;
2360 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2362 spinlock_t *ptl = pud_lockptr(mm, pud);
2368 extern void __init pagecache_init(void);
2369 extern void __init free_area_init_memoryless_node(int nid);
2370 extern void free_initmem(void);
2373 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2374 * into the buddy system. The freed pages will be poisoned with pattern
2375 * "poison" if it's within range [0, UCHAR_MAX].
2376 * Return pages freed into the buddy system.
2378 extern unsigned long free_reserved_area(void *start, void *end,
2379 int poison, const char *s);
2381 extern void adjust_managed_page_count(struct page *page, long count);
2382 extern void mem_init_print_info(void);
2384 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2386 /* Free the reserved page into the buddy system, so it gets managed. */
2387 static inline void free_reserved_page(struct page *page)
2389 ClearPageReserved(page);
2390 init_page_count(page);
2392 adjust_managed_page_count(page, 1);
2394 #define free_highmem_page(page) free_reserved_page(page)
2396 static inline void mark_page_reserved(struct page *page)
2398 SetPageReserved(page);
2399 adjust_managed_page_count(page, -1);
2403 * Default method to free all the __init memory into the buddy system.
2404 * The freed pages will be poisoned with pattern "poison" if it's within
2405 * range [0, UCHAR_MAX].
2406 * Return pages freed into the buddy system.
2408 static inline unsigned long free_initmem_default(int poison)
2410 extern char __init_begin[], __init_end[];
2412 return free_reserved_area(&__init_begin, &__init_end,
2413 poison, "unused kernel image (initmem)");
2416 static inline unsigned long get_num_physpages(void)
2419 unsigned long phys_pages = 0;
2421 for_each_online_node(nid)
2422 phys_pages += node_present_pages(nid);
2428 * Using memblock node mappings, an architecture may initialise its
2429 * zones, allocate the backing mem_map and account for memory holes in an
2430 * architecture independent manner.
2432 * An architecture is expected to register range of page frames backed by
2433 * physical memory with memblock_add[_node]() before calling
2434 * free_area_init() passing in the PFN each zone ends at. At a basic
2435 * usage, an architecture is expected to do something like
2437 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2439 * for_each_valid_physical_page_range()
2440 * memblock_add_node(base, size, nid)
2441 * free_area_init(max_zone_pfns);
2443 void free_area_init(unsigned long *max_zone_pfn);
2444 unsigned long node_map_pfn_alignment(void);
2445 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2446 unsigned long end_pfn);
2447 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2448 unsigned long end_pfn);
2449 extern void get_pfn_range_for_nid(unsigned int nid,
2450 unsigned long *start_pfn, unsigned long *end_pfn);
2451 extern unsigned long find_min_pfn_with_active_regions(void);
2454 static inline int early_pfn_to_nid(unsigned long pfn)
2459 /* please see mm/page_alloc.c */
2460 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2463 extern void set_dma_reserve(unsigned long new_dma_reserve);
2464 extern void memmap_init_range(unsigned long, int, unsigned long,
2465 unsigned long, unsigned long, enum meminit_context,
2466 struct vmem_altmap *, int migratetype);
2467 extern void setup_per_zone_wmarks(void);
2468 extern int __meminit init_per_zone_wmark_min(void);
2469 extern void mem_init(void);
2470 extern void __init mmap_init(void);
2471 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2472 extern long si_mem_available(void);
2473 extern void si_meminfo(struct sysinfo * val);
2474 extern void si_meminfo_node(struct sysinfo *val, int nid);
2475 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2476 extern unsigned long arch_reserved_kernel_pages(void);
2479 extern __printf(3, 4)
2480 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2482 extern void setup_per_cpu_pageset(void);
2485 extern int min_free_kbytes;
2486 extern int watermark_boost_factor;
2487 extern int watermark_scale_factor;
2488 extern bool arch_has_descending_max_zone_pfns(void);
2491 extern atomic_long_t mmap_pages_allocated;
2492 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2494 /* interval_tree.c */
2495 void vma_interval_tree_insert(struct vm_area_struct *node,
2496 struct rb_root_cached *root);
2497 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2498 struct vm_area_struct *prev,
2499 struct rb_root_cached *root);
2500 void vma_interval_tree_remove(struct vm_area_struct *node,
2501 struct rb_root_cached *root);
2502 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2503 unsigned long start, unsigned long last);
2504 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2505 unsigned long start, unsigned long last);
2507 #define vma_interval_tree_foreach(vma, root, start, last) \
2508 for (vma = vma_interval_tree_iter_first(root, start, last); \
2509 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2511 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2512 struct rb_root_cached *root);
2513 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2514 struct rb_root_cached *root);
2515 struct anon_vma_chain *
2516 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2517 unsigned long start, unsigned long last);
2518 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2519 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2520 #ifdef CONFIG_DEBUG_VM_RB
2521 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2524 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2525 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2526 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2529 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2530 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2531 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2532 struct vm_area_struct *expand);
2533 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2534 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2536 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2538 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2539 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2540 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2541 struct mempolicy *, struct vm_userfaultfd_ctx);
2542 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2543 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2544 unsigned long addr, int new_below);
2545 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2546 unsigned long addr, int new_below);
2547 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2548 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2549 struct rb_node **, struct rb_node *);
2550 extern void unlink_file_vma(struct vm_area_struct *);
2551 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2552 unsigned long addr, unsigned long len, pgoff_t pgoff,
2553 bool *need_rmap_locks);
2554 extern void exit_mmap(struct mm_struct *);
2556 static inline int check_data_rlimit(unsigned long rlim,
2558 unsigned long start,
2559 unsigned long end_data,
2560 unsigned long start_data)
2562 if (rlim < RLIM_INFINITY) {
2563 if (((new - start) + (end_data - start_data)) > rlim)
2570 extern int mm_take_all_locks(struct mm_struct *mm);
2571 extern void mm_drop_all_locks(struct mm_struct *mm);
2573 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2574 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2575 extern struct file *get_task_exe_file(struct task_struct *task);
2577 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2578 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2580 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2581 const struct vm_special_mapping *sm);
2582 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2583 unsigned long addr, unsigned long len,
2584 unsigned long flags,
2585 const struct vm_special_mapping *spec);
2586 /* This is an obsolete alternative to _install_special_mapping. */
2587 extern int install_special_mapping(struct mm_struct *mm,
2588 unsigned long addr, unsigned long len,
2589 unsigned long flags, struct page **pages);
2591 unsigned long randomize_stack_top(unsigned long stack_top);
2593 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2595 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2596 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2597 struct list_head *uf);
2598 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2599 unsigned long len, unsigned long prot, unsigned long flags,
2600 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2601 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2602 struct list_head *uf, bool downgrade);
2603 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2604 struct list_head *uf);
2605 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2608 extern int __mm_populate(unsigned long addr, unsigned long len,
2610 static inline void mm_populate(unsigned long addr, unsigned long len)
2613 (void) __mm_populate(addr, len, 1);
2616 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2619 /* These take the mm semaphore themselves */
2620 extern int __must_check vm_brk(unsigned long, unsigned long);
2621 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2622 extern int vm_munmap(unsigned long, size_t);
2623 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2624 unsigned long, unsigned long,
2625 unsigned long, unsigned long);
2627 struct vm_unmapped_area_info {
2628 #define VM_UNMAPPED_AREA_TOPDOWN 1
2629 unsigned long flags;
2630 unsigned long length;
2631 unsigned long low_limit;
2632 unsigned long high_limit;
2633 unsigned long align_mask;
2634 unsigned long align_offset;
2637 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2640 extern void truncate_inode_pages(struct address_space *, loff_t);
2641 extern void truncate_inode_pages_range(struct address_space *,
2642 loff_t lstart, loff_t lend);
2643 extern void truncate_inode_pages_final(struct address_space *);
2645 /* generic vm_area_ops exported for stackable file systems */
2646 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2647 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2648 pgoff_t start_pgoff, pgoff_t end_pgoff);
2649 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2651 /* mm/page-writeback.c */
2652 int __must_check write_one_page(struct page *page);
2653 void task_dirty_inc(struct task_struct *tsk);
2655 extern unsigned long stack_guard_gap;
2656 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2657 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2659 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2660 extern int expand_downwards(struct vm_area_struct *vma,
2661 unsigned long address);
2663 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2665 #define expand_upwards(vma, address) (0)
2668 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2669 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2670 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2671 struct vm_area_struct **pprev);
2674 * find_vma_intersection() - Look up the first VMA which intersects the interval
2675 * @mm: The process address space.
2676 * @start_addr: The inclusive start user address.
2677 * @end_addr: The exclusive end user address.
2679 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2680 * start_addr < end_addr.
2683 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2684 unsigned long start_addr,
2685 unsigned long end_addr)
2687 struct vm_area_struct *vma = find_vma(mm, start_addr);
2689 if (vma && end_addr <= vma->vm_start)
2695 * vma_lookup() - Find a VMA at a specific address
2696 * @mm: The process address space.
2697 * @addr: The user address.
2699 * Return: The vm_area_struct at the given address, %NULL otherwise.
2702 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2704 struct vm_area_struct *vma = find_vma(mm, addr);
2706 if (vma && addr < vma->vm_start)
2712 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2714 unsigned long vm_start = vma->vm_start;
2716 if (vma->vm_flags & VM_GROWSDOWN) {
2717 vm_start -= stack_guard_gap;
2718 if (vm_start > vma->vm_start)
2724 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2726 unsigned long vm_end = vma->vm_end;
2728 if (vma->vm_flags & VM_GROWSUP) {
2729 vm_end += stack_guard_gap;
2730 if (vm_end < vma->vm_end)
2731 vm_end = -PAGE_SIZE;
2736 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2738 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2741 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2742 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2743 unsigned long vm_start, unsigned long vm_end)
2745 struct vm_area_struct *vma = find_vma(mm, vm_start);
2747 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2753 static inline bool range_in_vma(struct vm_area_struct *vma,
2754 unsigned long start, unsigned long end)
2756 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2760 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2761 void vma_set_page_prot(struct vm_area_struct *vma);
2763 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2767 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2769 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2773 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2775 #ifdef CONFIG_NUMA_BALANCING
2776 unsigned long change_prot_numa(struct vm_area_struct *vma,
2777 unsigned long start, unsigned long end);
2780 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2781 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2782 unsigned long pfn, unsigned long size, pgprot_t);
2783 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2784 unsigned long pfn, unsigned long size, pgprot_t prot);
2785 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2786 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2787 struct page **pages, unsigned long *num);
2788 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2790 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2792 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2794 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2795 unsigned long pfn, pgprot_t pgprot);
2796 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2798 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2799 pfn_t pfn, pgprot_t pgprot);
2800 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2801 unsigned long addr, pfn_t pfn);
2802 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2804 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2805 unsigned long addr, struct page *page)
2807 int err = vm_insert_page(vma, addr, page);
2810 return VM_FAULT_OOM;
2811 if (err < 0 && err != -EBUSY)
2812 return VM_FAULT_SIGBUS;
2814 return VM_FAULT_NOPAGE;
2817 #ifndef io_remap_pfn_range
2818 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2819 unsigned long addr, unsigned long pfn,
2820 unsigned long size, pgprot_t prot)
2822 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2826 static inline vm_fault_t vmf_error(int err)
2829 return VM_FAULT_OOM;
2830 return VM_FAULT_SIGBUS;
2833 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2834 unsigned int foll_flags);
2836 #define FOLL_WRITE 0x01 /* check pte is writable */
2837 #define FOLL_TOUCH 0x02 /* mark page accessed */
2838 #define FOLL_GET 0x04 /* do get_page on page */
2839 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2840 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2841 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2842 * and return without waiting upon it */
2843 #define FOLL_POPULATE 0x40 /* fault in page */
2844 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2845 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2846 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2847 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2848 #define FOLL_MLOCK 0x1000 /* lock present pages */
2849 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2850 #define FOLL_COW 0x4000 /* internal GUP flag */
2851 #define FOLL_ANON 0x8000 /* don't do file mappings */
2852 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2853 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2854 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2855 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2858 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2859 * other. Here is what they mean, and how to use them:
2861 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2862 * period _often_ under userspace control. This is in contrast to
2863 * iov_iter_get_pages(), whose usages are transient.
2865 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2866 * lifetime enforced by the filesystem and we need guarantees that longterm
2867 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2868 * the filesystem. Ideas for this coordination include revoking the longterm
2869 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2870 * added after the problem with filesystems was found FS DAX VMAs are
2871 * specifically failed. Filesystem pages are still subject to bugs and use of
2872 * FOLL_LONGTERM should be avoided on those pages.
2874 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2875 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2876 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2877 * is due to an incompatibility with the FS DAX check and
2878 * FAULT_FLAG_ALLOW_RETRY.
2880 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2881 * that region. And so, CMA attempts to migrate the page before pinning, when
2882 * FOLL_LONGTERM is specified.
2884 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2885 * but an additional pin counting system) will be invoked. This is intended for
2886 * anything that gets a page reference and then touches page data (for example,
2887 * Direct IO). This lets the filesystem know that some non-file-system entity is
2888 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2889 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2890 * a call to unpin_user_page().
2892 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2893 * and separate refcounting mechanisms, however, and that means that each has
2894 * its own acquire and release mechanisms:
2896 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2898 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2900 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2901 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2902 * calls applied to them, and that's perfectly OK. This is a constraint on the
2903 * callers, not on the pages.)
2905 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2906 * directly by the caller. That's in order to help avoid mismatches when
2907 * releasing pages: get_user_pages*() pages must be released via put_page(),
2908 * while pin_user_pages*() pages must be released via unpin_user_page().
2910 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2913 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2915 if (vm_fault & VM_FAULT_OOM)
2917 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2918 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2919 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2924 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2925 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2926 unsigned long size, pte_fn_t fn, void *data);
2927 extern int apply_to_existing_page_range(struct mm_struct *mm,
2928 unsigned long address, unsigned long size,
2929 pte_fn_t fn, void *data);
2931 extern void init_mem_debugging_and_hardening(void);
2932 #ifdef CONFIG_PAGE_POISONING
2933 extern void __kernel_poison_pages(struct page *page, int numpages);
2934 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2935 extern bool _page_poisoning_enabled_early;
2936 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2937 static inline bool page_poisoning_enabled(void)
2939 return _page_poisoning_enabled_early;
2942 * For use in fast paths after init_mem_debugging() has run, or when a
2943 * false negative result is not harmful when called too early.
2945 static inline bool page_poisoning_enabled_static(void)
2947 return static_branch_unlikely(&_page_poisoning_enabled);
2949 static inline void kernel_poison_pages(struct page *page, int numpages)
2951 if (page_poisoning_enabled_static())
2952 __kernel_poison_pages(page, numpages);
2954 static inline void kernel_unpoison_pages(struct page *page, int numpages)
2956 if (page_poisoning_enabled_static())
2957 __kernel_unpoison_pages(page, numpages);
2960 static inline bool page_poisoning_enabled(void) { return false; }
2961 static inline bool page_poisoning_enabled_static(void) { return false; }
2962 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2963 static inline void kernel_poison_pages(struct page *page, int numpages) { }
2964 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2967 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
2968 static inline bool want_init_on_alloc(gfp_t flags)
2970 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
2973 return flags & __GFP_ZERO;
2976 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
2977 static inline bool want_init_on_free(void)
2979 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
2983 extern bool _debug_pagealloc_enabled_early;
2984 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2986 static inline bool debug_pagealloc_enabled(void)
2988 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2989 _debug_pagealloc_enabled_early;
2993 * For use in fast paths after init_debug_pagealloc() has run, or when a
2994 * false negative result is not harmful when called too early.
2996 static inline bool debug_pagealloc_enabled_static(void)
2998 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3001 return static_branch_unlikely(&_debug_pagealloc_enabled);
3004 #ifdef CONFIG_DEBUG_PAGEALLOC
3006 * To support DEBUG_PAGEALLOC architecture must ensure that
3007 * __kernel_map_pages() never fails
3009 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3011 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3013 if (debug_pagealloc_enabled_static())
3014 __kernel_map_pages(page, numpages, 1);
3017 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3019 if (debug_pagealloc_enabled_static())
3020 __kernel_map_pages(page, numpages, 0);
3022 #else /* CONFIG_DEBUG_PAGEALLOC */
3023 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3024 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3025 #endif /* CONFIG_DEBUG_PAGEALLOC */
3027 #ifdef __HAVE_ARCH_GATE_AREA
3028 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3029 extern int in_gate_area_no_mm(unsigned long addr);
3030 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3032 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3036 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3037 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3041 #endif /* __HAVE_ARCH_GATE_AREA */
3043 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3045 #ifdef CONFIG_SYSCTL
3046 extern int sysctl_drop_caches;
3047 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3051 void drop_slab(void);
3052 void drop_slab_node(int nid);
3055 #define randomize_va_space 0
3057 extern int randomize_va_space;
3060 const char * arch_vma_name(struct vm_area_struct *vma);
3062 void print_vma_addr(char *prefix, unsigned long rip);
3064 static inline void print_vma_addr(char *prefix, unsigned long rip)
3069 void *sparse_buffer_alloc(unsigned long size);
3070 struct page * __populate_section_memmap(unsigned long pfn,
3071 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3072 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3073 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3074 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3075 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3076 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3077 struct vmem_altmap *altmap);
3078 void *vmemmap_alloc_block(unsigned long size, int node);
3080 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3081 struct vmem_altmap *altmap);
3082 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3083 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3084 int node, struct vmem_altmap *altmap);
3085 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3086 struct vmem_altmap *altmap);
3087 void vmemmap_populate_print_last(void);
3088 #ifdef CONFIG_MEMORY_HOTPLUG
3089 void vmemmap_free(unsigned long start, unsigned long end,
3090 struct vmem_altmap *altmap);
3092 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3093 unsigned long nr_pages);
3096 MF_COUNT_INCREASED = 1 << 0,
3097 MF_ACTION_REQUIRED = 1 << 1,
3098 MF_MUST_KILL = 1 << 2,
3099 MF_SOFT_OFFLINE = 1 << 3,
3101 extern int memory_failure(unsigned long pfn, int flags);
3102 extern void memory_failure_queue(unsigned long pfn, int flags);
3103 extern void memory_failure_queue_kick(int cpu);
3104 extern int unpoison_memory(unsigned long pfn);
3105 extern int sysctl_memory_failure_early_kill;
3106 extern int sysctl_memory_failure_recovery;
3107 extern void shake_page(struct page *p, int access);
3108 extern atomic_long_t num_poisoned_pages __read_mostly;
3109 extern int soft_offline_page(unsigned long pfn, int flags);
3113 * Error handlers for various types of pages.
3116 MF_IGNORED, /* Error: cannot be handled */
3117 MF_FAILED, /* Error: handling failed */
3118 MF_DELAYED, /* Will be handled later */
3119 MF_RECOVERED, /* Successfully recovered */
3122 enum mf_action_page_type {
3124 MF_MSG_KERNEL_HIGH_ORDER,
3126 MF_MSG_DIFFERENT_COMPOUND,
3127 MF_MSG_POISONED_HUGE,
3130 MF_MSG_NON_PMD_HUGE,
3131 MF_MSG_UNMAP_FAILED,
3132 MF_MSG_DIRTY_SWAPCACHE,
3133 MF_MSG_CLEAN_SWAPCACHE,
3134 MF_MSG_DIRTY_MLOCKED_LRU,
3135 MF_MSG_CLEAN_MLOCKED_LRU,
3136 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3137 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3140 MF_MSG_TRUNCATED_LRU,
3148 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3149 extern void clear_huge_page(struct page *page,
3150 unsigned long addr_hint,
3151 unsigned int pages_per_huge_page);
3152 extern void copy_user_huge_page(struct page *dst, struct page *src,
3153 unsigned long addr_hint,
3154 struct vm_area_struct *vma,
3155 unsigned int pages_per_huge_page);
3156 extern long copy_huge_page_from_user(struct page *dst_page,
3157 const void __user *usr_src,
3158 unsigned int pages_per_huge_page,
3159 bool allow_pagefault);
3162 * vma_is_special_huge - Are transhuge page-table entries considered special?
3163 * @vma: Pointer to the struct vm_area_struct to consider
3165 * Whether transhuge page-table entries are considered "special" following
3166 * the definition in vm_normal_page().
3168 * Return: true if transhuge page-table entries should be considered special,
3171 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3173 return vma_is_dax(vma) || (vma->vm_file &&
3174 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3177 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3179 #ifdef CONFIG_DEBUG_PAGEALLOC
3180 extern unsigned int _debug_guardpage_minorder;
3181 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3183 static inline unsigned int debug_guardpage_minorder(void)
3185 return _debug_guardpage_minorder;
3188 static inline bool debug_guardpage_enabled(void)
3190 return static_branch_unlikely(&_debug_guardpage_enabled);
3193 static inline bool page_is_guard(struct page *page)
3195 if (!debug_guardpage_enabled())
3198 return PageGuard(page);
3201 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3202 static inline bool debug_guardpage_enabled(void) { return false; }
3203 static inline bool page_is_guard(struct page *page) { return false; }
3204 #endif /* CONFIG_DEBUG_PAGEALLOC */
3206 #if MAX_NUMNODES > 1
3207 void __init setup_nr_node_ids(void);
3209 static inline void setup_nr_node_ids(void) {}
3212 extern int memcmp_pages(struct page *page1, struct page *page2);
3214 static inline int pages_identical(struct page *page1, struct page *page2)
3216 return !memcmp_pages(page1, page2);
3219 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3220 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3221 pgoff_t first_index, pgoff_t nr,
3222 pgoff_t bitmap_pgoff,
3223 unsigned long *bitmap,
3227 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3228 pgoff_t first_index, pgoff_t nr);
3231 extern int sysctl_nr_trim_pages;
3233 #ifdef CONFIG_PRINTK
3234 void mem_dump_obj(void *object);
3236 static inline void mem_dump_obj(void *object) {}
3240 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3241 * @seals: the seals to check
3242 * @vma: the vma to operate on
3244 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3245 * the vma flags. Return 0 if check pass, or <0 for errors.
3247 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3249 if (seals & F_SEAL_FUTURE_WRITE) {
3251 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3252 * "future write" seal active.
3254 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3258 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3259 * MAP_SHARED and read-only, take care to not allow mprotect to
3260 * revert protections on such mappings. Do this only for shared
3261 * mappings. For private mappings, don't need to mask
3262 * VM_MAYWRITE as we still want them to be COW-writable.
3264 if (vma->vm_flags & VM_SHARED)
3265 vma->vm_flags &= ~(VM_MAYWRITE);
3271 #endif /* __KERNEL__ */
3272 #endif /* _LINUX_MM_H */