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>
37 struct anon_vma_chain;
40 struct writeback_control;
44 void init_mm_internals(void);
46 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
47 extern unsigned long max_mapnr;
49 static inline void set_max_mapnr(unsigned long limit)
54 static inline void set_max_mapnr(unsigned long limit) { }
57 extern atomic_long_t _totalram_pages;
58 static inline unsigned long totalram_pages(void)
60 return (unsigned long)atomic_long_read(&_totalram_pages);
63 static inline void totalram_pages_inc(void)
65 atomic_long_inc(&_totalram_pages);
68 static inline void totalram_pages_dec(void)
70 atomic_long_dec(&_totalram_pages);
73 static inline void totalram_pages_add(long count)
75 atomic_long_add(count, &_totalram_pages);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/processor.h>
102 * Architectures that support memory tagging (assigning tags to memory regions,
103 * embedding these tags into addresses that point to these memory regions, and
104 * checking that the memory and the pointer tags match on memory accesses)
105 * redefine this macro to strip tags from pointers.
106 * It's defined as noop for arcitectures that don't support memory tagging.
108 #ifndef untagged_addr
109 #define untagged_addr(addr) (addr)
113 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 #define lm_alias(x) __va(__pa_symbol(x))
125 * To prevent common memory management code establishing
126 * a zero page mapping on a read fault.
127 * This macro should be defined within <asm/pgtable.h>.
128 * s390 does this to prevent multiplexing of hardware bits
129 * related to the physical page in case of virtualization.
131 #ifndef mm_forbids_zeropage
132 #define mm_forbids_zeropage(X) (0)
136 * On some architectures it is expensive to call memset() for small sizes.
137 * If an architecture decides to implement their own version of
138 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
139 * define their own version of this macro in <asm/pgtable.h>
141 #if BITS_PER_LONG == 64
142 /* This function must be updated when the size of struct page grows above 80
143 * or reduces below 56. The idea that compiler optimizes out switch()
144 * statement, and only leaves move/store instructions. Also the compiler can
145 * combine write statments if they are both assignments and can be reordered,
146 * this can result in several of the writes here being dropped.
148 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
149 static inline void __mm_zero_struct_page(struct page *page)
151 unsigned long *_pp = (void *)page;
153 /* Check that struct page is either 56, 64, 72, or 80 bytes */
154 BUILD_BUG_ON(sizeof(struct page) & 7);
155 BUILD_BUG_ON(sizeof(struct page) < 56);
156 BUILD_BUG_ON(sizeof(struct page) > 80);
158 switch (sizeof(struct page)) {
179 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
183 * Default maximum number of active map areas, this limits the number of vmas
184 * per mm struct. Users can overwrite this number by sysctl but there is a
187 * When a program's coredump is generated as ELF format, a section is created
188 * per a vma. In ELF, the number of sections is represented in unsigned short.
189 * This means the number of sections should be smaller than 65535 at coredump.
190 * Because the kernel adds some informative sections to a image of program at
191 * generating coredump, we need some margin. The number of extra sections is
192 * 1-3 now and depends on arch. We use "5" as safe margin, here.
194 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
195 * not a hard limit any more. Although some userspace tools can be surprised by
198 #define MAPCOUNT_ELF_CORE_MARGIN (5)
199 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
201 extern int sysctl_max_map_count;
203 extern unsigned long sysctl_user_reserve_kbytes;
204 extern unsigned long sysctl_admin_reserve_kbytes;
206 extern int sysctl_overcommit_memory;
207 extern int sysctl_overcommit_ratio;
208 extern unsigned long sysctl_overcommit_kbytes;
210 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
212 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
214 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
217 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
219 /* to align the pointer to the (next) page boundary */
220 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
222 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
223 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
225 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
228 * Linux kernel virtual memory manager primitives.
229 * The idea being to have a "virtual" mm in the same way
230 * we have a virtual fs - giving a cleaner interface to the
231 * mm details, and allowing different kinds of memory mappings
232 * (from shared memory to executable loading to arbitrary
236 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
237 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
238 void vm_area_free(struct vm_area_struct *);
241 extern struct rb_root nommu_region_tree;
242 extern struct rw_semaphore nommu_region_sem;
244 extern unsigned int kobjsize(const void *objp);
248 * vm_flags in vm_area_struct, see mm_types.h.
249 * When changing, update also include/trace/events/mmflags.h
251 #define VM_NONE 0x00000000
253 #define VM_READ 0x00000001 /* currently active flags */
254 #define VM_WRITE 0x00000002
255 #define VM_EXEC 0x00000004
256 #define VM_SHARED 0x00000008
258 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
259 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
260 #define VM_MAYWRITE 0x00000020
261 #define VM_MAYEXEC 0x00000040
262 #define VM_MAYSHARE 0x00000080
264 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
265 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
266 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
267 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
268 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
270 #define VM_LOCKED 0x00002000
271 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
273 /* Used by sys_madvise() */
274 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
275 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
277 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
278 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
279 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
280 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
281 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
282 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
283 #define VM_SYNC 0x00800000 /* Synchronous page faults */
284 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
285 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
286 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
288 #ifdef CONFIG_MEM_SOFT_DIRTY
289 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
291 # define VM_SOFTDIRTY 0
294 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
295 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
296 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
297 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
299 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
300 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
302 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
303 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
304 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
305 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
306 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
307 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
308 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
309 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
310 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
312 #ifdef CONFIG_ARCH_HAS_PKEYS
313 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
314 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
315 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
316 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
317 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
319 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
321 # define VM_PKEY_BIT4 0
323 #endif /* CONFIG_ARCH_HAS_PKEYS */
325 #if defined(CONFIG_X86)
326 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
327 #elif defined(CONFIG_PPC)
328 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
329 #elif defined(CONFIG_PARISC)
330 # define VM_GROWSUP VM_ARCH_1
331 #elif defined(CONFIG_IA64)
332 # define VM_GROWSUP VM_ARCH_1
333 #elif defined(CONFIG_SPARC64)
334 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
335 # define VM_ARCH_CLEAR VM_SPARC_ADI
336 #elif defined(CONFIG_ARM64)
337 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
338 # define VM_ARCH_CLEAR VM_ARM64_BTI
339 #elif !defined(CONFIG_MMU)
340 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
344 # define VM_GROWSUP VM_NONE
347 /* Bits set in the VMA until the stack is in its final location */
348 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
350 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
352 /* Common data flag combinations */
353 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
354 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
355 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
356 VM_MAYWRITE | VM_MAYEXEC)
357 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
358 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
360 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
361 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
364 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
365 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
368 #ifdef CONFIG_STACK_GROWSUP
369 #define VM_STACK VM_GROWSUP
371 #define VM_STACK VM_GROWSDOWN
374 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
376 /* VMA basic access permission flags */
377 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
381 * Special vmas that are non-mergable, non-mlock()able.
383 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
385 /* This mask prevents VMA from being scanned with khugepaged */
386 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
388 /* This mask defines which mm->def_flags a process can inherit its parent */
389 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
391 /* This mask is used to clear all the VMA flags used by mlock */
392 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
394 /* Arch-specific flags to clear when updating VM flags on protection change */
395 #ifndef VM_ARCH_CLEAR
396 # define VM_ARCH_CLEAR VM_NONE
398 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
401 * mapping from the currently active vm_flags protection bits (the
402 * low four bits) to a page protection mask..
404 extern pgprot_t protection_map[16];
407 * Fault flag definitions.
409 * @FAULT_FLAG_WRITE: Fault was a write fault.
410 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
411 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
412 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
413 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
414 * @FAULT_FLAG_TRIED: The fault has been tried once.
415 * @FAULT_FLAG_USER: The fault originated in userspace.
416 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
417 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
418 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
420 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
421 * whether we would allow page faults to retry by specifying these two
422 * fault flags correctly. Currently there can be three legal combinations:
424 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
425 * this is the first try
427 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
428 * we've already tried at least once
430 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
432 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
433 * be used. Note that page faults can be allowed to retry for multiple times,
434 * in which case we'll have an initial fault with flags (a) then later on
435 * continuous faults with flags (b). We should always try to detect pending
436 * signals before a retry to make sure the continuous page faults can still be
437 * interrupted if necessary.
439 #define FAULT_FLAG_WRITE 0x01
440 #define FAULT_FLAG_MKWRITE 0x02
441 #define FAULT_FLAG_ALLOW_RETRY 0x04
442 #define FAULT_FLAG_RETRY_NOWAIT 0x08
443 #define FAULT_FLAG_KILLABLE 0x10
444 #define FAULT_FLAG_TRIED 0x20
445 #define FAULT_FLAG_USER 0x40
446 #define FAULT_FLAG_REMOTE 0x80
447 #define FAULT_FLAG_INSTRUCTION 0x100
448 #define FAULT_FLAG_INTERRUPTIBLE 0x200
451 * The default fault flags that should be used by most of the
452 * arch-specific page fault handlers.
454 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
455 FAULT_FLAG_KILLABLE | \
456 FAULT_FLAG_INTERRUPTIBLE)
459 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
461 * This is mostly used for places where we want to try to avoid taking
462 * the mmap_lock for too long a time when waiting for another condition
463 * to change, in which case we can try to be polite to release the
464 * mmap_lock in the first round to avoid potential starvation of other
465 * processes that would also want the mmap_lock.
467 * Return: true if the page fault allows retry and this is the first
468 * attempt of the fault handling; false otherwise.
470 static inline bool fault_flag_allow_retry_first(unsigned int flags)
472 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
473 (!(flags & FAULT_FLAG_TRIED));
476 #define FAULT_FLAG_TRACE \
477 { FAULT_FLAG_WRITE, "WRITE" }, \
478 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
479 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
480 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
481 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
482 { FAULT_FLAG_TRIED, "TRIED" }, \
483 { FAULT_FLAG_USER, "USER" }, \
484 { FAULT_FLAG_REMOTE, "REMOTE" }, \
485 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
486 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
489 * vm_fault is filled by the pagefault handler and passed to the vma's
490 * ->fault function. The vma's ->fault is responsible for returning a bitmask
491 * of VM_FAULT_xxx flags that give details about how the fault was handled.
493 * MM layer fills up gfp_mask for page allocations but fault handler might
494 * alter it if its implementation requires a different allocation context.
496 * pgoff should be used in favour of virtual_address, if possible.
499 struct vm_area_struct *vma; /* Target VMA */
500 unsigned int flags; /* FAULT_FLAG_xxx flags */
501 gfp_t gfp_mask; /* gfp mask to be used for allocations */
502 pgoff_t pgoff; /* Logical page offset based on vma */
503 unsigned long address; /* Faulting virtual address */
504 pmd_t *pmd; /* Pointer to pmd entry matching
506 pud_t *pud; /* Pointer to pud entry matching
509 pte_t orig_pte; /* Value of PTE at the time of fault */
511 struct page *cow_page; /* Page handler may use for COW fault */
512 struct page *page; /* ->fault handlers should return a
513 * page here, unless VM_FAULT_NOPAGE
514 * is set (which is also implied by
517 /* These three entries are valid only while holding ptl lock */
518 pte_t *pte; /* Pointer to pte entry matching
519 * the 'address'. NULL if the page
520 * table hasn't been allocated.
522 spinlock_t *ptl; /* Page table lock.
523 * Protects pte page table if 'pte'
524 * is not NULL, otherwise pmd.
526 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
527 * vm_ops->map_pages() calls
528 * alloc_set_pte() from atomic context.
529 * do_fault_around() pre-allocates
530 * page table to avoid allocation from
535 /* page entry size for vm->huge_fault() */
536 enum page_entry_size {
543 * These are the virtual MM functions - opening of an area, closing and
544 * unmapping it (needed to keep files on disk up-to-date etc), pointer
545 * to the functions called when a no-page or a wp-page exception occurs.
547 struct vm_operations_struct {
548 void (*open)(struct vm_area_struct * area);
549 void (*close)(struct vm_area_struct * area);
550 int (*split)(struct vm_area_struct * area, unsigned long addr);
551 int (*mremap)(struct vm_area_struct * area);
552 vm_fault_t (*fault)(struct vm_fault *vmf);
553 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
554 enum page_entry_size pe_size);
555 void (*map_pages)(struct vm_fault *vmf,
556 pgoff_t start_pgoff, pgoff_t end_pgoff);
557 unsigned long (*pagesize)(struct vm_area_struct * area);
559 /* notification that a previously read-only page is about to become
560 * writable, if an error is returned it will cause a SIGBUS */
561 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
563 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
564 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
566 /* called by access_process_vm when get_user_pages() fails, typically
567 * for use by special VMAs that can switch between memory and hardware
569 int (*access)(struct vm_area_struct *vma, unsigned long addr,
570 void *buf, int len, int write);
572 /* Called by the /proc/PID/maps code to ask the vma whether it
573 * has a special name. Returning non-NULL will also cause this
574 * vma to be dumped unconditionally. */
575 const char *(*name)(struct vm_area_struct *vma);
579 * set_policy() op must add a reference to any non-NULL @new mempolicy
580 * to hold the policy upon return. Caller should pass NULL @new to
581 * remove a policy and fall back to surrounding context--i.e. do not
582 * install a MPOL_DEFAULT policy, nor the task or system default
585 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
588 * get_policy() op must add reference [mpol_get()] to any policy at
589 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
590 * in mm/mempolicy.c will do this automatically.
591 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
592 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
593 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
594 * must return NULL--i.e., do not "fallback" to task or system default
597 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
601 * Called by vm_normal_page() for special PTEs to find the
602 * page for @addr. This is useful if the default behavior
603 * (using pte_page()) would not find the correct page.
605 struct page *(*find_special_page)(struct vm_area_struct *vma,
609 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
611 static const struct vm_operations_struct dummy_vm_ops = {};
613 memset(vma, 0, sizeof(*vma));
615 vma->vm_ops = &dummy_vm_ops;
616 INIT_LIST_HEAD(&vma->anon_vma_chain);
619 static inline void vma_set_anonymous(struct vm_area_struct *vma)
624 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
629 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
631 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
636 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
637 VM_STACK_INCOMPLETE_SETUP)
643 static inline bool vma_is_foreign(struct vm_area_struct *vma)
648 if (current->mm != vma->vm_mm)
654 static inline bool vma_is_accessible(struct vm_area_struct *vma)
656 return vma->vm_flags & VM_ACCESS_FLAGS;
661 * The vma_is_shmem is not inline because it is used only by slow
662 * paths in userfault.
664 bool vma_is_shmem(struct vm_area_struct *vma);
666 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
669 int vma_is_stack_for_current(struct vm_area_struct *vma);
671 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
672 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
677 #include <linux/huge_mm.h>
680 * Methods to modify the page usage count.
682 * What counts for a page usage:
683 * - cache mapping (page->mapping)
684 * - private data (page->private)
685 * - page mapped in a task's page tables, each mapping
686 * is counted separately
688 * Also, many kernel routines increase the page count before a critical
689 * routine so they can be sure the page doesn't go away from under them.
693 * Drop a ref, return true if the refcount fell to zero (the page has no users)
695 static inline int put_page_testzero(struct page *page)
697 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
698 return page_ref_dec_and_test(page);
702 * Try to grab a ref unless the page has a refcount of zero, return false if
704 * This can be called when MMU is off so it must not access
705 * any of the virtual mappings.
707 static inline int get_page_unless_zero(struct page *page)
709 return page_ref_add_unless(page, 1, 0);
712 extern int page_is_ram(unsigned long pfn);
720 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
723 /* Support for virtually mapped pages */
724 struct page *vmalloc_to_page(const void *addr);
725 unsigned long vmalloc_to_pfn(const void *addr);
728 * Determine if an address is within the vmalloc range
730 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
731 * is no special casing required.
734 #ifndef is_ioremap_addr
735 #define is_ioremap_addr(x) is_vmalloc_addr(x)
739 extern bool is_vmalloc_addr(const void *x);
740 extern int is_vmalloc_or_module_addr(const void *x);
742 static inline bool is_vmalloc_addr(const void *x)
746 static inline int is_vmalloc_or_module_addr(const void *x)
752 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
753 static inline void *kvmalloc(size_t size, gfp_t flags)
755 return kvmalloc_node(size, flags, NUMA_NO_NODE);
757 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
759 return kvmalloc_node(size, flags | __GFP_ZERO, node);
761 static inline void *kvzalloc(size_t size, gfp_t flags)
763 return kvmalloc(size, flags | __GFP_ZERO);
766 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
770 if (unlikely(check_mul_overflow(n, size, &bytes)))
773 return kvmalloc(bytes, flags);
776 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
778 return kvmalloc_array(n, size, flags | __GFP_ZERO);
781 extern void kvfree(const void *addr);
782 extern void kvfree_sensitive(const void *addr, size_t len);
784 static inline int head_mapcount(struct page *head)
786 return atomic_read(compound_mapcount_ptr(head)) + 1;
790 * Mapcount of compound page as a whole, does not include mapped sub-pages.
792 * Must be called only for compound pages or any their tail sub-pages.
794 static inline int compound_mapcount(struct page *page)
796 VM_BUG_ON_PAGE(!PageCompound(page), page);
797 page = compound_head(page);
798 return head_mapcount(page);
802 * The atomic page->_mapcount, starts from -1: so that transitions
803 * both from it and to it can be tracked, using atomic_inc_and_test
804 * and atomic_add_negative(-1).
806 static inline void page_mapcount_reset(struct page *page)
808 atomic_set(&(page)->_mapcount, -1);
811 int __page_mapcount(struct page *page);
814 * Mapcount of 0-order page; when compound sub-page, includes
815 * compound_mapcount().
817 * Result is undefined for pages which cannot be mapped into userspace.
818 * For example SLAB or special types of pages. See function page_has_type().
819 * They use this place in struct page differently.
821 static inline int page_mapcount(struct page *page)
823 if (unlikely(PageCompound(page)))
824 return __page_mapcount(page);
825 return atomic_read(&page->_mapcount) + 1;
828 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
829 int total_mapcount(struct page *page);
830 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
832 static inline int total_mapcount(struct page *page)
834 return page_mapcount(page);
836 static inline int page_trans_huge_mapcount(struct page *page,
839 int mapcount = page_mapcount(page);
841 *total_mapcount = mapcount;
846 static inline struct page *virt_to_head_page(const void *x)
848 struct page *page = virt_to_page(x);
850 return compound_head(page);
853 void __put_page(struct page *page);
855 void put_pages_list(struct list_head *pages);
857 void split_page(struct page *page, unsigned int order);
860 * Compound pages have a destructor function. Provide a
861 * prototype for that function and accessor functions.
862 * These are _only_ valid on the head of a compound page.
864 typedef void compound_page_dtor(struct page *);
866 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
867 enum compound_dtor_id {
870 #ifdef CONFIG_HUGETLB_PAGE
873 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
878 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
880 static inline void set_compound_page_dtor(struct page *page,
881 enum compound_dtor_id compound_dtor)
883 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
884 page[1].compound_dtor = compound_dtor;
887 static inline void destroy_compound_page(struct page *page)
889 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
890 compound_page_dtors[page[1].compound_dtor](page);
893 static inline unsigned int compound_order(struct page *page)
897 return page[1].compound_order;
900 static inline bool hpage_pincount_available(struct page *page)
903 * Can the page->hpage_pinned_refcount field be used? That field is in
904 * the 3rd page of the compound page, so the smallest (2-page) compound
905 * pages cannot support it.
907 page = compound_head(page);
908 return PageCompound(page) && compound_order(page) > 1;
911 static inline int head_pincount(struct page *head)
913 return atomic_read(compound_pincount_ptr(head));
916 static inline int compound_pincount(struct page *page)
918 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
919 page = compound_head(page);
920 return head_pincount(page);
923 static inline void set_compound_order(struct page *page, unsigned int order)
925 page[1].compound_order = order;
926 page[1].compound_nr = 1U << order;
929 /* Returns the number of pages in this potentially compound page. */
930 static inline unsigned long compound_nr(struct page *page)
934 return page[1].compound_nr;
937 /* Returns the number of bytes in this potentially compound page. */
938 static inline unsigned long page_size(struct page *page)
940 return PAGE_SIZE << compound_order(page);
943 /* Returns the number of bits needed for the number of bytes in a page */
944 static inline unsigned int page_shift(struct page *page)
946 return PAGE_SHIFT + compound_order(page);
949 void free_compound_page(struct page *page);
953 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
954 * servicing faults for write access. In the normal case, do always want
955 * pte_mkwrite. But get_user_pages can cause write faults for mappings
956 * that do not have writing enabled, when used by access_process_vm.
958 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
960 if (likely(vma->vm_flags & VM_WRITE))
961 pte = pte_mkwrite(pte);
965 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct page *page);
966 vm_fault_t finish_fault(struct vm_fault *vmf);
967 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
971 * Multiple processes may "see" the same page. E.g. for untouched
972 * mappings of /dev/null, all processes see the same page full of
973 * zeroes, and text pages of executables and shared libraries have
974 * only one copy in memory, at most, normally.
976 * For the non-reserved pages, page_count(page) denotes a reference count.
977 * page_count() == 0 means the page is free. page->lru is then used for
978 * freelist management in the buddy allocator.
979 * page_count() > 0 means the page has been allocated.
981 * Pages are allocated by the slab allocator in order to provide memory
982 * to kmalloc and kmem_cache_alloc. In this case, the management of the
983 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
984 * unless a particular usage is carefully commented. (the responsibility of
985 * freeing the kmalloc memory is the caller's, of course).
987 * A page may be used by anyone else who does a __get_free_page().
988 * In this case, page_count still tracks the references, and should only
989 * be used through the normal accessor functions. The top bits of page->flags
990 * and page->virtual store page management information, but all other fields
991 * are unused and could be used privately, carefully. The management of this
992 * page is the responsibility of the one who allocated it, and those who have
993 * subsequently been given references to it.
995 * The other pages (we may call them "pagecache pages") are completely
996 * managed by the Linux memory manager: I/O, buffers, swapping etc.
997 * The following discussion applies only to them.
999 * A pagecache page contains an opaque `private' member, which belongs to the
1000 * page's address_space. Usually, this is the address of a circular list of
1001 * the page's disk buffers. PG_private must be set to tell the VM to call
1002 * into the filesystem to release these pages.
1004 * A page may belong to an inode's memory mapping. In this case, page->mapping
1005 * is the pointer to the inode, and page->index is the file offset of the page,
1006 * in units of PAGE_SIZE.
1008 * If pagecache pages are not associated with an inode, they are said to be
1009 * anonymous pages. These may become associated with the swapcache, and in that
1010 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1012 * In either case (swapcache or inode backed), the pagecache itself holds one
1013 * reference to the page. Setting PG_private should also increment the
1014 * refcount. The each user mapping also has a reference to the page.
1016 * The pagecache pages are stored in a per-mapping radix tree, which is
1017 * rooted at mapping->i_pages, and indexed by offset.
1018 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1019 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1021 * All pagecache pages may be subject to I/O:
1022 * - inode pages may need to be read from disk,
1023 * - inode pages which have been modified and are MAP_SHARED may need
1024 * to be written back to the inode on disk,
1025 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1026 * modified may need to be swapped out to swap space and (later) to be read
1031 * The zone field is never updated after free_area_init_core()
1032 * sets it, so none of the operations on it need to be atomic.
1035 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1036 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1037 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1038 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1039 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1040 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1043 * Define the bit shifts to access each section. For non-existent
1044 * sections we define the shift as 0; that plus a 0 mask ensures
1045 * the compiler will optimise away reference to them.
1047 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1048 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1049 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1050 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1051 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1053 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1054 #ifdef NODE_NOT_IN_PAGE_FLAGS
1055 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1056 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1057 SECTIONS_PGOFF : ZONES_PGOFF)
1059 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1060 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1061 NODES_PGOFF : ZONES_PGOFF)
1064 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1066 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1067 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1068 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1069 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1070 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1071 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1073 static inline enum zone_type page_zonenum(const struct page *page)
1075 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1076 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1079 #ifdef CONFIG_ZONE_DEVICE
1080 static inline bool is_zone_device_page(const struct page *page)
1082 return page_zonenum(page) == ZONE_DEVICE;
1084 extern void memmap_init_zone_device(struct zone *, unsigned long,
1085 unsigned long, struct dev_pagemap *);
1087 static inline bool is_zone_device_page(const struct page *page)
1093 #ifdef CONFIG_DEV_PAGEMAP_OPS
1094 void free_devmap_managed_page(struct page *page);
1095 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1097 static inline bool page_is_devmap_managed(struct page *page)
1099 if (!static_branch_unlikely(&devmap_managed_key))
1101 if (!is_zone_device_page(page))
1103 switch (page->pgmap->type) {
1104 case MEMORY_DEVICE_PRIVATE:
1105 case MEMORY_DEVICE_FS_DAX:
1113 void put_devmap_managed_page(struct page *page);
1115 #else /* CONFIG_DEV_PAGEMAP_OPS */
1116 static inline bool page_is_devmap_managed(struct page *page)
1121 static inline void put_devmap_managed_page(struct page *page)
1124 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1126 static inline bool is_device_private_page(const struct page *page)
1128 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1129 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1130 is_zone_device_page(page) &&
1131 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1134 static inline bool is_pci_p2pdma_page(const struct page *page)
1136 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1137 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1138 is_zone_device_page(page) &&
1139 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1142 /* 127: arbitrary random number, small enough to assemble well */
1143 #define page_ref_zero_or_close_to_overflow(page) \
1144 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1146 static inline void get_page(struct page *page)
1148 page = compound_head(page);
1150 * Getting a normal page or the head of a compound page
1151 * requires to already have an elevated page->_refcount.
1153 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1157 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1159 static inline __must_check bool try_get_page(struct page *page)
1161 page = compound_head(page);
1162 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1168 static inline void put_page(struct page *page)
1170 page = compound_head(page);
1173 * For devmap managed pages we need to catch refcount transition from
1174 * 2 to 1, when refcount reach one it means the page is free and we
1175 * need to inform the device driver through callback. See
1176 * include/linux/memremap.h and HMM for details.
1178 if (page_is_devmap_managed(page)) {
1179 put_devmap_managed_page(page);
1183 if (put_page_testzero(page))
1188 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1189 * the page's refcount so that two separate items are tracked: the original page
1190 * reference count, and also a new count of how many pin_user_pages() calls were
1191 * made against the page. ("gup-pinned" is another term for the latter).
1193 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1194 * distinct from normal pages. As such, the unpin_user_page() call (and its
1195 * variants) must be used in order to release gup-pinned pages.
1199 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1200 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1201 * simpler, due to the fact that adding an even power of two to the page
1202 * refcount has the effect of using only the upper N bits, for the code that
1203 * counts up using the bias value. This means that the lower bits are left for
1204 * the exclusive use of the original code that increments and decrements by one
1205 * (or at least, by much smaller values than the bias value).
1207 * Of course, once the lower bits overflow into the upper bits (and this is
1208 * OK, because subtraction recovers the original values), then visual inspection
1209 * no longer suffices to directly view the separate counts. However, for normal
1210 * applications that don't have huge page reference counts, this won't be an
1213 * Locking: the lockless algorithm described in page_cache_get_speculative()
1214 * and page_cache_gup_pin_speculative() provides safe operation for
1215 * get_user_pages and page_mkclean and other calls that race to set up page
1218 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1220 void unpin_user_page(struct page *page);
1221 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1223 void unpin_user_pages(struct page **pages, unsigned long npages);
1226 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1228 * This function checks if a page has been pinned via a call to
1229 * pin_user_pages*().
1231 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1232 * because it means "definitely not pinned for DMA", but true means "probably
1233 * pinned for DMA, but possibly a false positive due to having at least
1234 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1236 * False positives are OK, because: a) it's unlikely for a page to get that many
1237 * refcounts, and b) all the callers of this routine are expected to be able to
1238 * deal gracefully with a false positive.
1240 * For huge pages, the result will be exactly correct. That's because we have
1241 * more tracking data available: the 3rd struct page in the compound page is
1242 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1245 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1247 * @page: pointer to page to be queried.
1248 * @Return: True, if it is likely that the page has been "dma-pinned".
1249 * False, if the page is definitely not dma-pinned.
1251 static inline bool page_maybe_dma_pinned(struct page *page)
1253 if (hpage_pincount_available(page))
1254 return compound_pincount(page) > 0;
1257 * page_ref_count() is signed. If that refcount overflows, then
1258 * page_ref_count() returns a negative value, and callers will avoid
1259 * further incrementing the refcount.
1261 * Here, for that overflow case, use the signed bit to count a little
1262 * bit higher via unsigned math, and thus still get an accurate result.
1264 return ((unsigned int)page_ref_count(compound_head(page))) >=
1265 GUP_PIN_COUNTING_BIAS;
1268 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1269 #define SECTION_IN_PAGE_FLAGS
1273 * The identification function is mainly used by the buddy allocator for
1274 * determining if two pages could be buddies. We are not really identifying
1275 * the zone since we could be using the section number id if we do not have
1276 * node id available in page flags.
1277 * We only guarantee that it will return the same value for two combinable
1280 static inline int page_zone_id(struct page *page)
1282 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1285 #ifdef NODE_NOT_IN_PAGE_FLAGS
1286 extern int page_to_nid(const struct page *page);
1288 static inline int page_to_nid(const struct page *page)
1290 struct page *p = (struct page *)page;
1292 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1296 #ifdef CONFIG_NUMA_BALANCING
1297 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1299 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1302 static inline int cpupid_to_pid(int cpupid)
1304 return cpupid & LAST__PID_MASK;
1307 static inline int cpupid_to_cpu(int cpupid)
1309 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1312 static inline int cpupid_to_nid(int cpupid)
1314 return cpu_to_node(cpupid_to_cpu(cpupid));
1317 static inline bool cpupid_pid_unset(int cpupid)
1319 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1322 static inline bool cpupid_cpu_unset(int cpupid)
1324 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1327 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1329 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1332 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1333 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1334 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1336 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1339 static inline int page_cpupid_last(struct page *page)
1341 return page->_last_cpupid;
1343 static inline void page_cpupid_reset_last(struct page *page)
1345 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1348 static inline int page_cpupid_last(struct page *page)
1350 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1353 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1355 static inline void page_cpupid_reset_last(struct page *page)
1357 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1359 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1360 #else /* !CONFIG_NUMA_BALANCING */
1361 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1363 return page_to_nid(page); /* XXX */
1366 static inline int page_cpupid_last(struct page *page)
1368 return page_to_nid(page); /* XXX */
1371 static inline int cpupid_to_nid(int cpupid)
1376 static inline int cpupid_to_pid(int cpupid)
1381 static inline int cpupid_to_cpu(int cpupid)
1386 static inline int cpu_pid_to_cpupid(int nid, int pid)
1391 static inline bool cpupid_pid_unset(int cpupid)
1396 static inline void page_cpupid_reset_last(struct page *page)
1400 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1404 #endif /* CONFIG_NUMA_BALANCING */
1406 #ifdef CONFIG_KASAN_SW_TAGS
1407 static inline u8 page_kasan_tag(const struct page *page)
1409 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1412 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1414 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1415 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1418 static inline void page_kasan_tag_reset(struct page *page)
1420 page_kasan_tag_set(page, 0xff);
1423 static inline u8 page_kasan_tag(const struct page *page)
1428 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1429 static inline void page_kasan_tag_reset(struct page *page) { }
1432 static inline struct zone *page_zone(const struct page *page)
1434 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1437 static inline pg_data_t *page_pgdat(const struct page *page)
1439 return NODE_DATA(page_to_nid(page));
1442 #ifdef SECTION_IN_PAGE_FLAGS
1443 static inline void set_page_section(struct page *page, unsigned long section)
1445 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1446 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1449 static inline unsigned long page_to_section(const struct page *page)
1451 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1455 static inline void set_page_zone(struct page *page, enum zone_type zone)
1457 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1458 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1461 static inline void set_page_node(struct page *page, unsigned long node)
1463 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1464 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1467 static inline void set_page_links(struct page *page, enum zone_type zone,
1468 unsigned long node, unsigned long pfn)
1470 set_page_zone(page, zone);
1471 set_page_node(page, node);
1472 #ifdef SECTION_IN_PAGE_FLAGS
1473 set_page_section(page, pfn_to_section_nr(pfn));
1478 static inline struct mem_cgroup *page_memcg(struct page *page)
1480 return page->mem_cgroup;
1482 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1484 WARN_ON_ONCE(!rcu_read_lock_held());
1485 return READ_ONCE(page->mem_cgroup);
1488 static inline struct mem_cgroup *page_memcg(struct page *page)
1492 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1494 WARN_ON_ONCE(!rcu_read_lock_held());
1500 * Some inline functions in vmstat.h depend on page_zone()
1502 #include <linux/vmstat.h>
1504 static __always_inline void *lowmem_page_address(const struct page *page)
1506 return page_to_virt(page);
1509 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1510 #define HASHED_PAGE_VIRTUAL
1513 #if defined(WANT_PAGE_VIRTUAL)
1514 static inline void *page_address(const struct page *page)
1516 return page->virtual;
1518 static inline void set_page_address(struct page *page, void *address)
1520 page->virtual = address;
1522 #define page_address_init() do { } while(0)
1525 #if defined(HASHED_PAGE_VIRTUAL)
1526 void *page_address(const struct page *page);
1527 void set_page_address(struct page *page, void *virtual);
1528 void page_address_init(void);
1531 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1532 #define page_address(page) lowmem_page_address(page)
1533 #define set_page_address(page, address) do { } while(0)
1534 #define page_address_init() do { } while(0)
1537 extern void *page_rmapping(struct page *page);
1538 extern struct anon_vma *page_anon_vma(struct page *page);
1539 extern struct address_space *page_mapping(struct page *page);
1541 extern struct address_space *__page_file_mapping(struct page *);
1544 struct address_space *page_file_mapping(struct page *page)
1546 if (unlikely(PageSwapCache(page)))
1547 return __page_file_mapping(page);
1549 return page->mapping;
1552 extern pgoff_t __page_file_index(struct page *page);
1555 * Return the pagecache index of the passed page. Regular pagecache pages
1556 * use ->index whereas swapcache pages use swp_offset(->private)
1558 static inline pgoff_t page_index(struct page *page)
1560 if (unlikely(PageSwapCache(page)))
1561 return __page_file_index(page);
1565 bool page_mapped(struct page *page);
1566 struct address_space *page_mapping(struct page *page);
1567 struct address_space *page_mapping_file(struct page *page);
1570 * Return true only if the page has been allocated with
1571 * ALLOC_NO_WATERMARKS and the low watermark was not
1572 * met implying that the system is under some pressure.
1574 static inline bool page_is_pfmemalloc(struct page *page)
1577 * Page index cannot be this large so this must be
1578 * a pfmemalloc page.
1580 return page->index == -1UL;
1584 * Only to be called by the page allocator on a freshly allocated
1587 static inline void set_page_pfmemalloc(struct page *page)
1592 static inline void clear_page_pfmemalloc(struct page *page)
1598 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1600 extern void pagefault_out_of_memory(void);
1602 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1603 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1606 * Flags passed to show_mem() and show_free_areas() to suppress output in
1609 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1611 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1614 extern bool can_do_mlock(void);
1616 static inline bool can_do_mlock(void) { return false; }
1618 extern int user_shm_lock(size_t, struct user_struct *);
1619 extern void user_shm_unlock(size_t, struct user_struct *);
1622 * Parameter block passed down to zap_pte_range in exceptional cases.
1624 struct zap_details {
1625 struct address_space *check_mapping; /* Check page->mapping if set */
1626 pgoff_t first_index; /* Lowest page->index to unmap */
1627 pgoff_t last_index; /* Highest page->index to unmap */
1630 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1632 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1635 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1636 unsigned long size);
1637 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1638 unsigned long size);
1639 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1640 unsigned long start, unsigned long end);
1642 struct mmu_notifier_range;
1644 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1645 unsigned long end, unsigned long floor, unsigned long ceiling);
1646 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1647 struct vm_area_struct *vma);
1648 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1649 struct mmu_notifier_range *range,
1650 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1651 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1652 unsigned long *pfn);
1653 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1654 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1655 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1656 void *buf, int len, int write);
1658 extern void truncate_pagecache(struct inode *inode, loff_t new);
1659 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1660 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1661 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1662 int truncate_inode_page(struct address_space *mapping, struct page *page);
1663 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1664 int invalidate_inode_page(struct page *page);
1667 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1668 unsigned long address, unsigned int flags,
1669 struct pt_regs *regs);
1670 extern int fixup_user_fault(struct mm_struct *mm,
1671 unsigned long address, unsigned int fault_flags,
1673 void unmap_mapping_pages(struct address_space *mapping,
1674 pgoff_t start, pgoff_t nr, bool even_cows);
1675 void unmap_mapping_range(struct address_space *mapping,
1676 loff_t const holebegin, loff_t const holelen, int even_cows);
1678 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1679 unsigned long address, unsigned int flags,
1680 struct pt_regs *regs)
1682 /* should never happen if there's no MMU */
1684 return VM_FAULT_SIGBUS;
1686 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1687 unsigned int fault_flags, bool *unlocked)
1689 /* should never happen if there's no MMU */
1693 static inline void unmap_mapping_pages(struct address_space *mapping,
1694 pgoff_t start, pgoff_t nr, bool even_cows) { }
1695 static inline void unmap_mapping_range(struct address_space *mapping,
1696 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1699 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1700 loff_t const holebegin, loff_t const holelen)
1702 unmap_mapping_range(mapping, holebegin, holelen, 0);
1705 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1706 void *buf, int len, unsigned int gup_flags);
1707 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1708 void *buf, int len, unsigned int gup_flags);
1709 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1710 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1712 long get_user_pages_remote(struct mm_struct *mm,
1713 unsigned long start, unsigned long nr_pages,
1714 unsigned int gup_flags, struct page **pages,
1715 struct vm_area_struct **vmas, int *locked);
1716 long pin_user_pages_remote(struct mm_struct *mm,
1717 unsigned long start, unsigned long nr_pages,
1718 unsigned int gup_flags, struct page **pages,
1719 struct vm_area_struct **vmas, int *locked);
1720 long get_user_pages(unsigned long start, unsigned long nr_pages,
1721 unsigned int gup_flags, struct page **pages,
1722 struct vm_area_struct **vmas);
1723 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1724 unsigned int gup_flags, struct page **pages,
1725 struct vm_area_struct **vmas);
1726 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1727 unsigned int gup_flags, struct page **pages, int *locked);
1728 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1729 unsigned int gup_flags, struct page **pages, int *locked);
1730 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1731 struct page **pages, unsigned int gup_flags);
1732 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1733 struct page **pages, unsigned int gup_flags);
1735 int get_user_pages_fast(unsigned long start, int nr_pages,
1736 unsigned int gup_flags, struct page **pages);
1737 int pin_user_pages_fast(unsigned long start, int nr_pages,
1738 unsigned int gup_flags, struct page **pages);
1740 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1741 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1742 struct task_struct *task, bool bypass_rlim);
1744 /* Container for pinned pfns / pages */
1745 struct frame_vector {
1746 unsigned int nr_allocated; /* Number of frames we have space for */
1747 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1748 bool got_ref; /* Did we pin pages by getting page ref? */
1749 bool is_pfns; /* Does array contain pages or pfns? */
1750 void *ptrs[]; /* Array of pinned pfns / pages. Use
1751 * pfns_vector_pages() or pfns_vector_pfns()
1755 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1756 void frame_vector_destroy(struct frame_vector *vec);
1757 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1758 unsigned int gup_flags, struct frame_vector *vec);
1759 void put_vaddr_frames(struct frame_vector *vec);
1760 int frame_vector_to_pages(struct frame_vector *vec);
1761 void frame_vector_to_pfns(struct frame_vector *vec);
1763 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1765 return vec->nr_frames;
1768 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1771 int err = frame_vector_to_pages(vec);
1774 return ERR_PTR(err);
1776 return (struct page **)(vec->ptrs);
1779 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1782 frame_vector_to_pfns(vec);
1783 return (unsigned long *)(vec->ptrs);
1787 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1788 struct page **pages);
1789 int get_kernel_page(unsigned long start, int write, struct page **pages);
1790 struct page *get_dump_page(unsigned long addr);
1792 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1793 extern void do_invalidatepage(struct page *page, unsigned int offset,
1794 unsigned int length);
1796 void __set_page_dirty(struct page *, struct address_space *, int warn);
1797 int __set_page_dirty_nobuffers(struct page *page);
1798 int __set_page_dirty_no_writeback(struct page *page);
1799 int redirty_page_for_writepage(struct writeback_control *wbc,
1801 void account_page_dirtied(struct page *page, struct address_space *mapping);
1802 void account_page_cleaned(struct page *page, struct address_space *mapping,
1803 struct bdi_writeback *wb);
1804 int set_page_dirty(struct page *page);
1805 int set_page_dirty_lock(struct page *page);
1806 void __cancel_dirty_page(struct page *page);
1807 static inline void cancel_dirty_page(struct page *page)
1809 /* Avoid atomic ops, locking, etc. when not actually needed. */
1810 if (PageDirty(page))
1811 __cancel_dirty_page(page);
1813 int clear_page_dirty_for_io(struct page *page);
1815 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1817 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1818 unsigned long old_addr, struct vm_area_struct *new_vma,
1819 unsigned long new_addr, unsigned long len,
1820 bool need_rmap_locks);
1823 * Flags used by change_protection(). For now we make it a bitmap so
1824 * that we can pass in multiple flags just like parameters. However
1825 * for now all the callers are only use one of the flags at the same
1828 /* Whether we should allow dirty bit accounting */
1829 #define MM_CP_DIRTY_ACCT (1UL << 0)
1830 /* Whether this protection change is for NUMA hints */
1831 #define MM_CP_PROT_NUMA (1UL << 1)
1832 /* Whether this change is for write protecting */
1833 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1834 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1835 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1836 MM_CP_UFFD_WP_RESOLVE)
1838 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1839 unsigned long end, pgprot_t newprot,
1840 unsigned long cp_flags);
1841 extern int mprotect_fixup(struct vm_area_struct *vma,
1842 struct vm_area_struct **pprev, unsigned long start,
1843 unsigned long end, unsigned long newflags);
1846 * doesn't attempt to fault and will return short.
1848 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1849 unsigned int gup_flags, struct page **pages);
1850 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1851 unsigned int gup_flags, struct page **pages);
1853 static inline bool get_user_page_fast_only(unsigned long addr,
1854 unsigned int gup_flags, struct page **pagep)
1856 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1859 * per-process(per-mm_struct) statistics.
1861 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1863 long val = atomic_long_read(&mm->rss_stat.count[member]);
1865 #ifdef SPLIT_RSS_COUNTING
1867 * counter is updated in asynchronous manner and may go to minus.
1868 * But it's never be expected number for users.
1873 return (unsigned long)val;
1876 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1878 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1880 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1882 mm_trace_rss_stat(mm, member, count);
1885 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1887 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1889 mm_trace_rss_stat(mm, member, count);
1892 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1894 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1896 mm_trace_rss_stat(mm, member, count);
1899 /* Optimized variant when page is already known not to be PageAnon */
1900 static inline int mm_counter_file(struct page *page)
1902 if (PageSwapBacked(page))
1903 return MM_SHMEMPAGES;
1904 return MM_FILEPAGES;
1907 static inline int mm_counter(struct page *page)
1910 return MM_ANONPAGES;
1911 return mm_counter_file(page);
1914 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1916 return get_mm_counter(mm, MM_FILEPAGES) +
1917 get_mm_counter(mm, MM_ANONPAGES) +
1918 get_mm_counter(mm, MM_SHMEMPAGES);
1921 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1923 return max(mm->hiwater_rss, get_mm_rss(mm));
1926 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1928 return max(mm->hiwater_vm, mm->total_vm);
1931 static inline void update_hiwater_rss(struct mm_struct *mm)
1933 unsigned long _rss = get_mm_rss(mm);
1935 if ((mm)->hiwater_rss < _rss)
1936 (mm)->hiwater_rss = _rss;
1939 static inline void update_hiwater_vm(struct mm_struct *mm)
1941 if (mm->hiwater_vm < mm->total_vm)
1942 mm->hiwater_vm = mm->total_vm;
1945 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1947 mm->hiwater_rss = get_mm_rss(mm);
1950 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1951 struct mm_struct *mm)
1953 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1955 if (*maxrss < hiwater_rss)
1956 *maxrss = hiwater_rss;
1959 #if defined(SPLIT_RSS_COUNTING)
1960 void sync_mm_rss(struct mm_struct *mm);
1962 static inline void sync_mm_rss(struct mm_struct *mm)
1967 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1968 static inline int pte_special(pte_t pte)
1973 static inline pte_t pte_mkspecial(pte_t pte)
1979 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1980 static inline int pte_devmap(pte_t pte)
1986 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1988 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1990 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1994 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1998 #ifdef __PAGETABLE_P4D_FOLDED
1999 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2000 unsigned long address)
2005 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2008 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2009 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2010 unsigned long address)
2014 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2015 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2018 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2020 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2022 if (mm_pud_folded(mm))
2024 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2027 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2029 if (mm_pud_folded(mm))
2031 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2035 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2036 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2037 unsigned long address)
2042 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2043 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2046 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2048 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2050 if (mm_pmd_folded(mm))
2052 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2055 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2057 if (mm_pmd_folded(mm))
2059 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2064 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2066 atomic_long_set(&mm->pgtables_bytes, 0);
2069 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2071 return atomic_long_read(&mm->pgtables_bytes);
2074 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2076 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2079 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2081 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2085 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2086 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2091 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2092 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2095 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2096 int __pte_alloc_kernel(pmd_t *pmd);
2098 #if defined(CONFIG_MMU)
2100 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2101 unsigned long address)
2103 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2104 NULL : p4d_offset(pgd, address);
2107 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2108 unsigned long address)
2110 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2111 NULL : pud_offset(p4d, address);
2114 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2116 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2117 NULL: pmd_offset(pud, address);
2119 #endif /* CONFIG_MMU */
2121 #if USE_SPLIT_PTE_PTLOCKS
2122 #if ALLOC_SPLIT_PTLOCKS
2123 void __init ptlock_cache_init(void);
2124 extern bool ptlock_alloc(struct page *page);
2125 extern void ptlock_free(struct page *page);
2127 static inline spinlock_t *ptlock_ptr(struct page *page)
2131 #else /* ALLOC_SPLIT_PTLOCKS */
2132 static inline void ptlock_cache_init(void)
2136 static inline bool ptlock_alloc(struct page *page)
2141 static inline void ptlock_free(struct page *page)
2145 static inline spinlock_t *ptlock_ptr(struct page *page)
2149 #endif /* ALLOC_SPLIT_PTLOCKS */
2151 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2153 return ptlock_ptr(pmd_page(*pmd));
2156 static inline bool ptlock_init(struct page *page)
2159 * prep_new_page() initialize page->private (and therefore page->ptl)
2160 * with 0. Make sure nobody took it in use in between.
2162 * It can happen if arch try to use slab for page table allocation:
2163 * slab code uses page->slab_cache, which share storage with page->ptl.
2165 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2166 if (!ptlock_alloc(page))
2168 spin_lock_init(ptlock_ptr(page));
2172 #else /* !USE_SPLIT_PTE_PTLOCKS */
2174 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2176 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2178 return &mm->page_table_lock;
2180 static inline void ptlock_cache_init(void) {}
2181 static inline bool ptlock_init(struct page *page) { return true; }
2182 static inline void ptlock_free(struct page *page) {}
2183 #endif /* USE_SPLIT_PTE_PTLOCKS */
2185 static inline void pgtable_init(void)
2187 ptlock_cache_init();
2188 pgtable_cache_init();
2191 static inline bool pgtable_pte_page_ctor(struct page *page)
2193 if (!ptlock_init(page))
2195 __SetPageTable(page);
2196 inc_zone_page_state(page, NR_PAGETABLE);
2200 static inline void pgtable_pte_page_dtor(struct page *page)
2203 __ClearPageTable(page);
2204 dec_zone_page_state(page, NR_PAGETABLE);
2207 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2209 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2210 pte_t *__pte = pte_offset_map(pmd, address); \
2216 #define pte_unmap_unlock(pte, ptl) do { \
2221 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2223 #define pte_alloc_map(mm, pmd, address) \
2224 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2226 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2227 (pte_alloc(mm, pmd) ? \
2228 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2230 #define pte_alloc_kernel(pmd, address) \
2231 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2232 NULL: pte_offset_kernel(pmd, address))
2234 #if USE_SPLIT_PMD_PTLOCKS
2236 static struct page *pmd_to_page(pmd_t *pmd)
2238 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2239 return virt_to_page((void *)((unsigned long) pmd & mask));
2242 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2244 return ptlock_ptr(pmd_to_page(pmd));
2247 static inline bool pgtable_pmd_page_ctor(struct page *page)
2249 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2250 page->pmd_huge_pte = NULL;
2252 return ptlock_init(page);
2255 static inline void pgtable_pmd_page_dtor(struct page *page)
2257 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2258 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2263 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2267 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2269 return &mm->page_table_lock;
2272 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2273 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2275 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2279 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2281 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2287 * No scalability reason to split PUD locks yet, but follow the same pattern
2288 * as the PMD locks to make it easier if we decide to. The VM should not be
2289 * considered ready to switch to split PUD locks yet; there may be places
2290 * which need to be converted from page_table_lock.
2292 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2294 return &mm->page_table_lock;
2297 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2299 spinlock_t *ptl = pud_lockptr(mm, pud);
2305 extern void __init pagecache_init(void);
2306 extern void __init free_area_init_memoryless_node(int nid);
2307 extern void free_initmem(void);
2310 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2311 * into the buddy system. The freed pages will be poisoned with pattern
2312 * "poison" if it's within range [0, UCHAR_MAX].
2313 * Return pages freed into the buddy system.
2315 extern unsigned long free_reserved_area(void *start, void *end,
2316 int poison, const char *s);
2318 #ifdef CONFIG_HIGHMEM
2320 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2321 * and totalram_pages.
2323 extern void free_highmem_page(struct page *page);
2326 extern void adjust_managed_page_count(struct page *page, long count);
2327 extern void mem_init_print_info(const char *str);
2329 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2331 /* Free the reserved page into the buddy system, so it gets managed. */
2332 static inline void __free_reserved_page(struct page *page)
2334 ClearPageReserved(page);
2335 init_page_count(page);
2339 static inline void free_reserved_page(struct page *page)
2341 __free_reserved_page(page);
2342 adjust_managed_page_count(page, 1);
2345 static inline void mark_page_reserved(struct page *page)
2347 SetPageReserved(page);
2348 adjust_managed_page_count(page, -1);
2352 * Default method to free all the __init memory into the buddy system.
2353 * The freed pages will be poisoned with pattern "poison" if it's within
2354 * range [0, UCHAR_MAX].
2355 * Return pages freed into the buddy system.
2357 static inline unsigned long free_initmem_default(int poison)
2359 extern char __init_begin[], __init_end[];
2361 return free_reserved_area(&__init_begin, &__init_end,
2362 poison, "unused kernel");
2365 static inline unsigned long get_num_physpages(void)
2368 unsigned long phys_pages = 0;
2370 for_each_online_node(nid)
2371 phys_pages += node_present_pages(nid);
2377 * Using memblock node mappings, an architecture may initialise its
2378 * zones, allocate the backing mem_map and account for memory holes in an
2379 * architecture independent manner.
2381 * An architecture is expected to register range of page frames backed by
2382 * physical memory with memblock_add[_node]() before calling
2383 * free_area_init() passing in the PFN each zone ends at. At a basic
2384 * usage, an architecture is expected to do something like
2386 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2388 * for_each_valid_physical_page_range()
2389 * memblock_add_node(base, size, nid)
2390 * free_area_init(max_zone_pfns);
2392 void free_area_init(unsigned long *max_zone_pfn);
2393 unsigned long node_map_pfn_alignment(void);
2394 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2395 unsigned long end_pfn);
2396 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2397 unsigned long end_pfn);
2398 extern void get_pfn_range_for_nid(unsigned int nid,
2399 unsigned long *start_pfn, unsigned long *end_pfn);
2400 extern unsigned long find_min_pfn_with_active_regions(void);
2402 #ifndef CONFIG_NEED_MULTIPLE_NODES
2403 static inline int early_pfn_to_nid(unsigned long pfn)
2408 /* please see mm/page_alloc.c */
2409 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2410 /* there is a per-arch backend function. */
2411 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2412 struct mminit_pfnnid_cache *state);
2415 extern void set_dma_reserve(unsigned long new_dma_reserve);
2416 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2417 enum memmap_context, struct vmem_altmap *);
2418 extern void setup_per_zone_wmarks(void);
2419 extern int __meminit init_per_zone_wmark_min(void);
2420 extern void mem_init(void);
2421 extern void __init mmap_init(void);
2422 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2423 extern long si_mem_available(void);
2424 extern void si_meminfo(struct sysinfo * val);
2425 extern void si_meminfo_node(struct sysinfo *val, int nid);
2426 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2427 extern unsigned long arch_reserved_kernel_pages(void);
2430 extern __printf(3, 4)
2431 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2433 extern void setup_per_cpu_pageset(void);
2436 extern int min_free_kbytes;
2437 extern int watermark_boost_factor;
2438 extern int watermark_scale_factor;
2439 extern bool arch_has_descending_max_zone_pfns(void);
2442 extern atomic_long_t mmap_pages_allocated;
2443 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2445 /* interval_tree.c */
2446 void vma_interval_tree_insert(struct vm_area_struct *node,
2447 struct rb_root_cached *root);
2448 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2449 struct vm_area_struct *prev,
2450 struct rb_root_cached *root);
2451 void vma_interval_tree_remove(struct vm_area_struct *node,
2452 struct rb_root_cached *root);
2453 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2454 unsigned long start, unsigned long last);
2455 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2456 unsigned long start, unsigned long last);
2458 #define vma_interval_tree_foreach(vma, root, start, last) \
2459 for (vma = vma_interval_tree_iter_first(root, start, last); \
2460 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2462 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2463 struct rb_root_cached *root);
2464 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2465 struct rb_root_cached *root);
2466 struct anon_vma_chain *
2467 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2468 unsigned long start, unsigned long last);
2469 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2470 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2471 #ifdef CONFIG_DEBUG_VM_RB
2472 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2475 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2476 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2477 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2480 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2481 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2482 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2483 struct vm_area_struct *expand);
2484 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2485 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2487 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2489 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2490 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2491 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2492 struct mempolicy *, struct vm_userfaultfd_ctx);
2493 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2494 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2495 unsigned long addr, int new_below);
2496 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2497 unsigned long addr, int new_below);
2498 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2499 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2500 struct rb_node **, struct rb_node *);
2501 extern void unlink_file_vma(struct vm_area_struct *);
2502 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2503 unsigned long addr, unsigned long len, pgoff_t pgoff,
2504 bool *need_rmap_locks);
2505 extern void exit_mmap(struct mm_struct *);
2507 static inline int check_data_rlimit(unsigned long rlim,
2509 unsigned long start,
2510 unsigned long end_data,
2511 unsigned long start_data)
2513 if (rlim < RLIM_INFINITY) {
2514 if (((new - start) + (end_data - start_data)) > rlim)
2521 extern int mm_take_all_locks(struct mm_struct *mm);
2522 extern void mm_drop_all_locks(struct mm_struct *mm);
2524 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2525 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2526 extern struct file *get_task_exe_file(struct task_struct *task);
2528 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2529 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2531 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2532 const struct vm_special_mapping *sm);
2533 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2534 unsigned long addr, unsigned long len,
2535 unsigned long flags,
2536 const struct vm_special_mapping *spec);
2537 /* This is an obsolete alternative to _install_special_mapping. */
2538 extern int install_special_mapping(struct mm_struct *mm,
2539 unsigned long addr, unsigned long len,
2540 unsigned long flags, struct page **pages);
2542 unsigned long randomize_stack_top(unsigned long stack_top);
2544 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2546 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2547 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2548 struct list_head *uf);
2549 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2550 unsigned long len, unsigned long prot, unsigned long flags,
2551 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2552 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2553 struct list_head *uf, bool downgrade);
2554 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2555 struct list_head *uf);
2556 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2559 extern int __mm_populate(unsigned long addr, unsigned long len,
2561 static inline void mm_populate(unsigned long addr, unsigned long len)
2564 (void) __mm_populate(addr, len, 1);
2567 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2570 /* These take the mm semaphore themselves */
2571 extern int __must_check vm_brk(unsigned long, unsigned long);
2572 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2573 extern int vm_munmap(unsigned long, size_t);
2574 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2575 unsigned long, unsigned long,
2576 unsigned long, unsigned long);
2578 struct vm_unmapped_area_info {
2579 #define VM_UNMAPPED_AREA_TOPDOWN 1
2580 unsigned long flags;
2581 unsigned long length;
2582 unsigned long low_limit;
2583 unsigned long high_limit;
2584 unsigned long align_mask;
2585 unsigned long align_offset;
2588 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2591 extern void truncate_inode_pages(struct address_space *, loff_t);
2592 extern void truncate_inode_pages_range(struct address_space *,
2593 loff_t lstart, loff_t lend);
2594 extern void truncate_inode_pages_final(struct address_space *);
2596 /* generic vm_area_ops exported for stackable file systems */
2597 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2598 extern void filemap_map_pages(struct vm_fault *vmf,
2599 pgoff_t start_pgoff, pgoff_t end_pgoff);
2600 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2602 /* mm/page-writeback.c */
2603 int __must_check write_one_page(struct page *page);
2604 void task_dirty_inc(struct task_struct *tsk);
2606 extern unsigned long stack_guard_gap;
2607 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2608 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2610 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2611 extern int expand_downwards(struct vm_area_struct *vma,
2612 unsigned long address);
2614 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2616 #define expand_upwards(vma, address) (0)
2619 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2620 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2621 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2622 struct vm_area_struct **pprev);
2624 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2625 NULL if none. Assume start_addr < end_addr. */
2626 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2628 struct vm_area_struct * vma = find_vma(mm,start_addr);
2630 if (vma && end_addr <= vma->vm_start)
2635 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2637 unsigned long vm_start = vma->vm_start;
2639 if (vma->vm_flags & VM_GROWSDOWN) {
2640 vm_start -= stack_guard_gap;
2641 if (vm_start > vma->vm_start)
2647 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2649 unsigned long vm_end = vma->vm_end;
2651 if (vma->vm_flags & VM_GROWSUP) {
2652 vm_end += stack_guard_gap;
2653 if (vm_end < vma->vm_end)
2654 vm_end = -PAGE_SIZE;
2659 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2661 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2664 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2665 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2666 unsigned long vm_start, unsigned long vm_end)
2668 struct vm_area_struct *vma = find_vma(mm, vm_start);
2670 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2676 static inline bool range_in_vma(struct vm_area_struct *vma,
2677 unsigned long start, unsigned long end)
2679 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2683 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2684 void vma_set_page_prot(struct vm_area_struct *vma);
2686 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2690 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2692 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2696 #ifdef CONFIG_NUMA_BALANCING
2697 unsigned long change_prot_numa(struct vm_area_struct *vma,
2698 unsigned long start, unsigned long end);
2701 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2702 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2703 unsigned long pfn, unsigned long size, pgprot_t);
2704 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2705 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2706 struct page **pages, unsigned long *num);
2707 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2709 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2711 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2713 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2714 unsigned long pfn, pgprot_t pgprot);
2715 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2717 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2718 pfn_t pfn, pgprot_t pgprot);
2719 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2720 unsigned long addr, pfn_t pfn);
2721 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2723 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2724 unsigned long addr, struct page *page)
2726 int err = vm_insert_page(vma, addr, page);
2729 return VM_FAULT_OOM;
2730 if (err < 0 && err != -EBUSY)
2731 return VM_FAULT_SIGBUS;
2733 return VM_FAULT_NOPAGE;
2736 static inline vm_fault_t vmf_error(int err)
2739 return VM_FAULT_OOM;
2740 return VM_FAULT_SIGBUS;
2743 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2744 unsigned int foll_flags);
2746 #define FOLL_WRITE 0x01 /* check pte is writable */
2747 #define FOLL_TOUCH 0x02 /* mark page accessed */
2748 #define FOLL_GET 0x04 /* do get_page on page */
2749 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2750 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2751 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2752 * and return without waiting upon it */
2753 #define FOLL_POPULATE 0x40 /* fault in page */
2754 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2755 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2756 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2757 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2758 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2759 #define FOLL_MLOCK 0x1000 /* lock present pages */
2760 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2761 #define FOLL_COW 0x4000 /* internal GUP flag */
2762 #define FOLL_ANON 0x8000 /* don't do file mappings */
2763 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2764 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2765 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2766 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2769 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2770 * other. Here is what they mean, and how to use them:
2772 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2773 * period _often_ under userspace control. This is in contrast to
2774 * iov_iter_get_pages(), whose usages are transient.
2776 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2777 * lifetime enforced by the filesystem and we need guarantees that longterm
2778 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2779 * the filesystem. Ideas for this coordination include revoking the longterm
2780 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2781 * added after the problem with filesystems was found FS DAX VMAs are
2782 * specifically failed. Filesystem pages are still subject to bugs and use of
2783 * FOLL_LONGTERM should be avoided on those pages.
2785 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2786 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2787 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2788 * is due to an incompatibility with the FS DAX check and
2789 * FAULT_FLAG_ALLOW_RETRY.
2791 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2792 * that region. And so, CMA attempts to migrate the page before pinning, when
2793 * FOLL_LONGTERM is specified.
2795 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2796 * but an additional pin counting system) will be invoked. This is intended for
2797 * anything that gets a page reference and then touches page data (for example,
2798 * Direct IO). This lets the filesystem know that some non-file-system entity is
2799 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2800 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2801 * a call to unpin_user_page().
2803 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2804 * and separate refcounting mechanisms, however, and that means that each has
2805 * its own acquire and release mechanisms:
2807 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2809 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2811 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2812 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2813 * calls applied to them, and that's perfectly OK. This is a constraint on the
2814 * callers, not on the pages.)
2816 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2817 * directly by the caller. That's in order to help avoid mismatches when
2818 * releasing pages: get_user_pages*() pages must be released via put_page(),
2819 * while pin_user_pages*() pages must be released via unpin_user_page().
2821 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2824 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2826 if (vm_fault & VM_FAULT_OOM)
2828 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2829 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2830 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2835 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2836 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2837 unsigned long size, pte_fn_t fn, void *data);
2838 extern int apply_to_existing_page_range(struct mm_struct *mm,
2839 unsigned long address, unsigned long size,
2840 pte_fn_t fn, void *data);
2842 #ifdef CONFIG_PAGE_POISONING
2843 extern bool page_poisoning_enabled(void);
2844 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2846 static inline bool page_poisoning_enabled(void) { return false; }
2847 static inline void kernel_poison_pages(struct page *page, int numpages,
2851 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2852 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2854 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2856 static inline bool want_init_on_alloc(gfp_t flags)
2858 if (static_branch_unlikely(&init_on_alloc) &&
2859 !page_poisoning_enabled())
2861 return flags & __GFP_ZERO;
2864 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2865 DECLARE_STATIC_KEY_TRUE(init_on_free);
2867 DECLARE_STATIC_KEY_FALSE(init_on_free);
2869 static inline bool want_init_on_free(void)
2871 return static_branch_unlikely(&init_on_free) &&
2872 !page_poisoning_enabled();
2875 #ifdef CONFIG_DEBUG_PAGEALLOC
2876 extern void init_debug_pagealloc(void);
2878 static inline void init_debug_pagealloc(void) {}
2880 extern bool _debug_pagealloc_enabled_early;
2881 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2883 static inline bool debug_pagealloc_enabled(void)
2885 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2886 _debug_pagealloc_enabled_early;
2890 * For use in fast paths after init_debug_pagealloc() has run, or when a
2891 * false negative result is not harmful when called too early.
2893 static inline bool debug_pagealloc_enabled_static(void)
2895 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2898 return static_branch_unlikely(&_debug_pagealloc_enabled);
2901 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2902 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2905 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2906 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2909 kernel_map_pages(struct page *page, int numpages, int enable)
2911 __kernel_map_pages(page, numpages, enable);
2913 #ifdef CONFIG_HIBERNATION
2914 extern bool kernel_page_present(struct page *page);
2915 #endif /* CONFIG_HIBERNATION */
2916 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2918 kernel_map_pages(struct page *page, int numpages, int enable) {}
2919 #ifdef CONFIG_HIBERNATION
2920 static inline bool kernel_page_present(struct page *page) { return true; }
2921 #endif /* CONFIG_HIBERNATION */
2922 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2924 #ifdef __HAVE_ARCH_GATE_AREA
2925 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2926 extern int in_gate_area_no_mm(unsigned long addr);
2927 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2929 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2933 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2934 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2938 #endif /* __HAVE_ARCH_GATE_AREA */
2940 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2942 #ifdef CONFIG_SYSCTL
2943 extern int sysctl_drop_caches;
2944 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2948 void drop_slab(void);
2949 void drop_slab_node(int nid);
2952 #define randomize_va_space 0
2954 extern int randomize_va_space;
2957 const char * arch_vma_name(struct vm_area_struct *vma);
2959 void print_vma_addr(char *prefix, unsigned long rip);
2961 static inline void print_vma_addr(char *prefix, unsigned long rip)
2966 void *sparse_buffer_alloc(unsigned long size);
2967 struct page * __populate_section_memmap(unsigned long pfn,
2968 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2969 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2970 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2971 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2972 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2973 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
2974 struct vmem_altmap *altmap);
2975 void *vmemmap_alloc_block(unsigned long size, int node);
2977 void *vmemmap_alloc_block_buf(unsigned long size, int node,
2978 struct vmem_altmap *altmap);
2979 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2980 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2981 int node, struct vmem_altmap *altmap);
2982 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2983 struct vmem_altmap *altmap);
2984 void vmemmap_populate_print_last(void);
2985 #ifdef CONFIG_MEMORY_HOTPLUG
2986 void vmemmap_free(unsigned long start, unsigned long end,
2987 struct vmem_altmap *altmap);
2989 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2990 unsigned long nr_pages);
2993 MF_COUNT_INCREASED = 1 << 0,
2994 MF_ACTION_REQUIRED = 1 << 1,
2995 MF_MUST_KILL = 1 << 2,
2996 MF_SOFT_OFFLINE = 1 << 3,
2998 extern int memory_failure(unsigned long pfn, int flags);
2999 extern void memory_failure_queue(unsigned long pfn, int flags);
3000 extern void memory_failure_queue_kick(int cpu);
3001 extern int unpoison_memory(unsigned long pfn);
3002 extern int get_hwpoison_page(struct page *page);
3003 #define put_hwpoison_page(page) put_page(page)
3004 extern int sysctl_memory_failure_early_kill;
3005 extern int sysctl_memory_failure_recovery;
3006 extern void shake_page(struct page *p, int access);
3007 extern atomic_long_t num_poisoned_pages __read_mostly;
3008 extern int soft_offline_page(unsigned long pfn, int flags);
3012 * Error handlers for various types of pages.
3015 MF_IGNORED, /* Error: cannot be handled */
3016 MF_FAILED, /* Error: handling failed */
3017 MF_DELAYED, /* Will be handled later */
3018 MF_RECOVERED, /* Successfully recovered */
3021 enum mf_action_page_type {
3023 MF_MSG_KERNEL_HIGH_ORDER,
3025 MF_MSG_DIFFERENT_COMPOUND,
3026 MF_MSG_POISONED_HUGE,
3029 MF_MSG_NON_PMD_HUGE,
3030 MF_MSG_UNMAP_FAILED,
3031 MF_MSG_DIRTY_SWAPCACHE,
3032 MF_MSG_CLEAN_SWAPCACHE,
3033 MF_MSG_DIRTY_MLOCKED_LRU,
3034 MF_MSG_CLEAN_MLOCKED_LRU,
3035 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3036 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3039 MF_MSG_TRUNCATED_LRU,
3046 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3047 extern void clear_huge_page(struct page *page,
3048 unsigned long addr_hint,
3049 unsigned int pages_per_huge_page);
3050 extern void copy_user_huge_page(struct page *dst, struct page *src,
3051 unsigned long addr_hint,
3052 struct vm_area_struct *vma,
3053 unsigned int pages_per_huge_page);
3054 extern long copy_huge_page_from_user(struct page *dst_page,
3055 const void __user *usr_src,
3056 unsigned int pages_per_huge_page,
3057 bool allow_pagefault);
3060 * vma_is_special_huge - Are transhuge page-table entries considered special?
3061 * @vma: Pointer to the struct vm_area_struct to consider
3063 * Whether transhuge page-table entries are considered "special" following
3064 * the definition in vm_normal_page().
3066 * Return: true if transhuge page-table entries should be considered special,
3069 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3071 return vma_is_dax(vma) || (vma->vm_file &&
3072 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3075 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3077 #ifdef CONFIG_DEBUG_PAGEALLOC
3078 extern unsigned int _debug_guardpage_minorder;
3079 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3081 static inline unsigned int debug_guardpage_minorder(void)
3083 return _debug_guardpage_minorder;
3086 static inline bool debug_guardpage_enabled(void)
3088 return static_branch_unlikely(&_debug_guardpage_enabled);
3091 static inline bool page_is_guard(struct page *page)
3093 if (!debug_guardpage_enabled())
3096 return PageGuard(page);
3099 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3100 static inline bool debug_guardpage_enabled(void) { return false; }
3101 static inline bool page_is_guard(struct page *page) { return false; }
3102 #endif /* CONFIG_DEBUG_PAGEALLOC */
3104 #if MAX_NUMNODES > 1
3105 void __init setup_nr_node_ids(void);
3107 static inline void setup_nr_node_ids(void) {}
3110 extern int memcmp_pages(struct page *page1, struct page *page2);
3112 static inline int pages_identical(struct page *page1, struct page *page2)
3114 return !memcmp_pages(page1, page2);
3117 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3118 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3119 pgoff_t first_index, pgoff_t nr,
3120 pgoff_t bitmap_pgoff,
3121 unsigned long *bitmap,
3125 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3126 pgoff_t first_index, pgoff_t nr);
3129 extern int sysctl_nr_trim_pages;
3131 #endif /* __KERNEL__ */
3132 #endif /* _LINUX_MM_H */