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/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 #include <linux/overflow.h>
29 #include <linux/sizes.h>
30 #include <linux/sched.h>
34 struct anon_vma_chain;
37 struct writeback_control;
40 void init_mm_internals(void);
42 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
43 extern unsigned long max_mapnr;
45 static inline void set_max_mapnr(unsigned long limit)
50 static inline void set_max_mapnr(unsigned long limit) { }
53 extern atomic_long_t _totalram_pages;
54 static inline unsigned long totalram_pages(void)
56 return (unsigned long)atomic_long_read(&_totalram_pages);
59 static inline void totalram_pages_inc(void)
61 atomic_long_inc(&_totalram_pages);
64 static inline void totalram_pages_dec(void)
66 atomic_long_dec(&_totalram_pages);
69 static inline void totalram_pages_add(long count)
71 atomic_long_add(count, &_totalram_pages);
74 extern void * high_memory;
75 extern int page_cluster;
78 extern int sysctl_legacy_va_layout;
80 #define sysctl_legacy_va_layout 0
83 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
84 extern const int mmap_rnd_bits_min;
85 extern const int mmap_rnd_bits_max;
86 extern int mmap_rnd_bits __read_mostly;
88 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
89 extern const int mmap_rnd_compat_bits_min;
90 extern const int mmap_rnd_compat_bits_max;
91 extern int mmap_rnd_compat_bits __read_mostly;
95 #include <asm/pgtable.h>
96 #include <asm/processor.h>
99 * Architectures that support memory tagging (assigning tags to memory regions,
100 * embedding these tags into addresses that point to these memory regions, and
101 * checking that the memory and the pointer tags match on memory accesses)
102 * redefine this macro to strip tags from pointers.
103 * It's defined as noop for arcitectures that don't support memory tagging.
105 #ifndef untagged_addr
106 #define untagged_addr(addr) (addr)
110 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
114 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
118 #define lm_alias(x) __va(__pa_symbol(x))
122 * To prevent common memory management code establishing
123 * a zero page mapping on a read fault.
124 * This macro should be defined within <asm/pgtable.h>.
125 * s390 does this to prevent multiplexing of hardware bits
126 * related to the physical page in case of virtualization.
128 #ifndef mm_forbids_zeropage
129 #define mm_forbids_zeropage(X) (0)
133 * On some architectures it is expensive to call memset() for small sizes.
134 * If an architecture decides to implement their own version of
135 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
136 * define their own version of this macro in <asm/pgtable.h>
138 #if BITS_PER_LONG == 64
139 /* This function must be updated when the size of struct page grows above 80
140 * or reduces below 56. The idea that compiler optimizes out switch()
141 * statement, and only leaves move/store instructions. Also the compiler can
142 * combine write statments if they are both assignments and can be reordered,
143 * this can result in several of the writes here being dropped.
145 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
146 static inline void __mm_zero_struct_page(struct page *page)
148 unsigned long *_pp = (void *)page;
150 /* Check that struct page is either 56, 64, 72, or 80 bytes */
151 BUILD_BUG_ON(sizeof(struct page) & 7);
152 BUILD_BUG_ON(sizeof(struct page) < 56);
153 BUILD_BUG_ON(sizeof(struct page) > 80);
155 switch (sizeof(struct page)) {
157 _pp[9] = 0; /* fallthrough */
159 _pp[8] = 0; /* fallthrough */
161 _pp[7] = 0; /* fallthrough */
173 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
177 * Default maximum number of active map areas, this limits the number of vmas
178 * per mm struct. Users can overwrite this number by sysctl but there is a
181 * When a program's coredump is generated as ELF format, a section is created
182 * per a vma. In ELF, the number of sections is represented in unsigned short.
183 * This means the number of sections should be smaller than 65535 at coredump.
184 * Because the kernel adds some informative sections to a image of program at
185 * generating coredump, we need some margin. The number of extra sections is
186 * 1-3 now and depends on arch. We use "5" as safe margin, here.
188 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
189 * not a hard limit any more. Although some userspace tools can be surprised by
192 #define MAPCOUNT_ELF_CORE_MARGIN (5)
193 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
195 extern int sysctl_max_map_count;
197 extern unsigned long sysctl_user_reserve_kbytes;
198 extern unsigned long sysctl_admin_reserve_kbytes;
200 extern int sysctl_overcommit_memory;
201 extern int sysctl_overcommit_ratio;
202 extern unsigned long sysctl_overcommit_kbytes;
204 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
206 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
209 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
211 /* to align the pointer to the (next) page boundary */
212 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
214 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
215 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
217 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
220 * Linux kernel virtual memory manager primitives.
221 * The idea being to have a "virtual" mm in the same way
222 * we have a virtual fs - giving a cleaner interface to the
223 * mm details, and allowing different kinds of memory mappings
224 * (from shared memory to executable loading to arbitrary
228 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
229 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
230 void vm_area_free(struct vm_area_struct *);
233 extern struct rb_root nommu_region_tree;
234 extern struct rw_semaphore nommu_region_sem;
236 extern unsigned int kobjsize(const void *objp);
240 * vm_flags in vm_area_struct, see mm_types.h.
241 * When changing, update also include/trace/events/mmflags.h
243 #define VM_NONE 0x00000000
245 #define VM_READ 0x00000001 /* currently active flags */
246 #define VM_WRITE 0x00000002
247 #define VM_EXEC 0x00000004
248 #define VM_SHARED 0x00000008
250 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
251 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
252 #define VM_MAYWRITE 0x00000020
253 #define VM_MAYEXEC 0x00000040
254 #define VM_MAYSHARE 0x00000080
256 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
257 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
258 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
259 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
260 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
262 #define VM_LOCKED 0x00002000
263 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
265 /* Used by sys_madvise() */
266 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
267 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
269 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
270 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
271 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
272 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
273 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
274 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
275 #define VM_SYNC 0x00800000 /* Synchronous page faults */
276 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
277 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
278 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
280 #ifdef CONFIG_MEM_SOFT_DIRTY
281 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
283 # define VM_SOFTDIRTY 0
286 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
287 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
288 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
289 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
291 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
292 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
293 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
294 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
295 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
296 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
298 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
299 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
300 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
301 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
302 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
304 #ifdef CONFIG_ARCH_HAS_PKEYS
305 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
306 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
307 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
308 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
309 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
311 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
313 # define VM_PKEY_BIT4 0
315 #endif /* CONFIG_ARCH_HAS_PKEYS */
317 #if defined(CONFIG_X86)
318 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
319 #elif defined(CONFIG_PPC)
320 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
321 #elif defined(CONFIG_PARISC)
322 # define VM_GROWSUP VM_ARCH_1
323 #elif defined(CONFIG_IA64)
324 # define VM_GROWSUP VM_ARCH_1
325 #elif defined(CONFIG_SPARC64)
326 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
327 # define VM_ARCH_CLEAR VM_SPARC_ADI
328 #elif !defined(CONFIG_MMU)
329 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
332 #if defined(CONFIG_X86_INTEL_MPX)
333 /* MPX specific bounds table or bounds directory */
334 # define VM_MPX VM_HIGH_ARCH_4
336 # define VM_MPX VM_NONE
340 # define VM_GROWSUP VM_NONE
343 /* Bits set in the VMA until the stack is in its final location */
344 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
346 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
347 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
350 #ifdef CONFIG_STACK_GROWSUP
351 #define VM_STACK VM_GROWSUP
353 #define VM_STACK VM_GROWSDOWN
356 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
359 * Special vmas that are non-mergable, non-mlock()able.
361 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
363 /* This mask prevents VMA from being scanned with khugepaged */
364 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
366 /* This mask defines which mm->def_flags a process can inherit its parent */
367 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
369 /* This mask is used to clear all the VMA flags used by mlock */
370 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
372 /* Arch-specific flags to clear when updating VM flags on protection change */
373 #ifndef VM_ARCH_CLEAR
374 # define VM_ARCH_CLEAR VM_NONE
376 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
379 * mapping from the currently active vm_flags protection bits (the
380 * low four bits) to a page protection mask..
382 extern pgprot_t protection_map[16];
385 * Fault flag definitions.
387 * @FAULT_FLAG_WRITE: Fault was a write fault.
388 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
389 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
390 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_sem and wait when retrying.
391 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
392 * @FAULT_FLAG_TRIED: The fault has been tried once.
393 * @FAULT_FLAG_USER: The fault originated in userspace.
394 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
395 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
396 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
398 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
399 * whether we would allow page faults to retry by specifying these two
400 * fault flags correctly. Currently there can be three legal combinations:
402 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
403 * this is the first try
405 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
406 * we've already tried at least once
408 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
410 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
411 * be used. Note that page faults can be allowed to retry for multiple times,
412 * in which case we'll have an initial fault with flags (a) then later on
413 * continuous faults with flags (b). We should always try to detect pending
414 * signals before a retry to make sure the continuous page faults can still be
415 * interrupted if necessary.
417 #define FAULT_FLAG_WRITE 0x01
418 #define FAULT_FLAG_MKWRITE 0x02
419 #define FAULT_FLAG_ALLOW_RETRY 0x04
420 #define FAULT_FLAG_RETRY_NOWAIT 0x08
421 #define FAULT_FLAG_KILLABLE 0x10
422 #define FAULT_FLAG_TRIED 0x20
423 #define FAULT_FLAG_USER 0x40
424 #define FAULT_FLAG_REMOTE 0x80
425 #define FAULT_FLAG_INSTRUCTION 0x100
426 #define FAULT_FLAG_INTERRUPTIBLE 0x200
429 * The default fault flags that should be used by most of the
430 * arch-specific page fault handlers.
432 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
433 FAULT_FLAG_KILLABLE | \
434 FAULT_FLAG_INTERRUPTIBLE)
437 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
439 * This is mostly used for places where we want to try to avoid taking
440 * the mmap_sem for too long a time when waiting for another condition
441 * to change, in which case we can try to be polite to release the
442 * mmap_sem in the first round to avoid potential starvation of other
443 * processes that would also want the mmap_sem.
445 * Return: true if the page fault allows retry and this is the first
446 * attempt of the fault handling; false otherwise.
448 static inline bool fault_flag_allow_retry_first(unsigned int flags)
450 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
451 (!(flags & FAULT_FLAG_TRIED));
454 #define FAULT_FLAG_TRACE \
455 { FAULT_FLAG_WRITE, "WRITE" }, \
456 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
457 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
458 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
459 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
460 { FAULT_FLAG_TRIED, "TRIED" }, \
461 { FAULT_FLAG_USER, "USER" }, \
462 { FAULT_FLAG_REMOTE, "REMOTE" }, \
463 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
464 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
467 * vm_fault is filled by the the pagefault handler and passed to the vma's
468 * ->fault function. The vma's ->fault is responsible for returning a bitmask
469 * of VM_FAULT_xxx flags that give details about how the fault was handled.
471 * MM layer fills up gfp_mask for page allocations but fault handler might
472 * alter it if its implementation requires a different allocation context.
474 * pgoff should be used in favour of virtual_address, if possible.
477 struct vm_area_struct *vma; /* Target VMA */
478 unsigned int flags; /* FAULT_FLAG_xxx flags */
479 gfp_t gfp_mask; /* gfp mask to be used for allocations */
480 pgoff_t pgoff; /* Logical page offset based on vma */
481 unsigned long address; /* Faulting virtual address */
482 pmd_t *pmd; /* Pointer to pmd entry matching
484 pud_t *pud; /* Pointer to pud entry matching
487 pte_t orig_pte; /* Value of PTE at the time of fault */
489 struct page *cow_page; /* Page handler may use for COW fault */
490 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
491 struct page *page; /* ->fault handlers should return a
492 * page here, unless VM_FAULT_NOPAGE
493 * is set (which is also implied by
496 /* These three entries are valid only while holding ptl lock */
497 pte_t *pte; /* Pointer to pte entry matching
498 * the 'address'. NULL if the page
499 * table hasn't been allocated.
501 spinlock_t *ptl; /* Page table lock.
502 * Protects pte page table if 'pte'
503 * is not NULL, otherwise pmd.
505 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
506 * vm_ops->map_pages() calls
507 * alloc_set_pte() from atomic context.
508 * do_fault_around() pre-allocates
509 * page table to avoid allocation from
514 /* page entry size for vm->huge_fault() */
515 enum page_entry_size {
522 * These are the virtual MM functions - opening of an area, closing and
523 * unmapping it (needed to keep files on disk up-to-date etc), pointer
524 * to the functions called when a no-page or a wp-page exception occurs.
526 struct vm_operations_struct {
527 void (*open)(struct vm_area_struct * area);
528 void (*close)(struct vm_area_struct * area);
529 int (*split)(struct vm_area_struct * area, unsigned long addr);
530 int (*mremap)(struct vm_area_struct * area);
531 vm_fault_t (*fault)(struct vm_fault *vmf);
532 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
533 enum page_entry_size pe_size);
534 void (*map_pages)(struct vm_fault *vmf,
535 pgoff_t start_pgoff, pgoff_t end_pgoff);
536 unsigned long (*pagesize)(struct vm_area_struct * area);
538 /* notification that a previously read-only page is about to become
539 * writable, if an error is returned it will cause a SIGBUS */
540 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
542 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
543 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
545 /* called by access_process_vm when get_user_pages() fails, typically
546 * for use by special VMAs that can switch between memory and hardware
548 int (*access)(struct vm_area_struct *vma, unsigned long addr,
549 void *buf, int len, int write);
551 /* Called by the /proc/PID/maps code to ask the vma whether it
552 * has a special name. Returning non-NULL will also cause this
553 * vma to be dumped unconditionally. */
554 const char *(*name)(struct vm_area_struct *vma);
558 * set_policy() op must add a reference to any non-NULL @new mempolicy
559 * to hold the policy upon return. Caller should pass NULL @new to
560 * remove a policy and fall back to surrounding context--i.e. do not
561 * install a MPOL_DEFAULT policy, nor the task or system default
564 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
567 * get_policy() op must add reference [mpol_get()] to any policy at
568 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
569 * in mm/mempolicy.c will do this automatically.
570 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
571 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
572 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
573 * must return NULL--i.e., do not "fallback" to task or system default
576 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
580 * Called by vm_normal_page() for special PTEs to find the
581 * page for @addr. This is useful if the default behavior
582 * (using pte_page()) would not find the correct page.
584 struct page *(*find_special_page)(struct vm_area_struct *vma,
588 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
590 static const struct vm_operations_struct dummy_vm_ops = {};
592 memset(vma, 0, sizeof(*vma));
594 vma->vm_ops = &dummy_vm_ops;
595 INIT_LIST_HEAD(&vma->anon_vma_chain);
598 static inline void vma_set_anonymous(struct vm_area_struct *vma)
603 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
608 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
610 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
615 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
616 VM_STACK_INCOMPLETE_SETUP)
622 static inline bool vma_is_foreign(struct vm_area_struct *vma)
627 if (current->mm != vma->vm_mm)
634 * The vma_is_shmem is not inline because it is used only by slow
635 * paths in userfault.
637 bool vma_is_shmem(struct vm_area_struct *vma);
639 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
642 int vma_is_stack_for_current(struct vm_area_struct *vma);
644 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
645 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
651 * FIXME: take this include out, include page-flags.h in
652 * files which need it (119 of them)
654 #include <linux/page-flags.h>
655 #include <linux/huge_mm.h>
658 * Methods to modify the page usage count.
660 * What counts for a page usage:
661 * - cache mapping (page->mapping)
662 * - private data (page->private)
663 * - page mapped in a task's page tables, each mapping
664 * is counted separately
666 * Also, many kernel routines increase the page count before a critical
667 * routine so they can be sure the page doesn't go away from under them.
671 * Drop a ref, return true if the refcount fell to zero (the page has no users)
673 static inline int put_page_testzero(struct page *page)
675 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
676 return page_ref_dec_and_test(page);
680 * Try to grab a ref unless the page has a refcount of zero, return false if
682 * This can be called when MMU is off so it must not access
683 * any of the virtual mappings.
685 static inline int get_page_unless_zero(struct page *page)
687 return page_ref_add_unless(page, 1, 0);
690 extern int page_is_ram(unsigned long pfn);
698 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
701 /* Support for virtually mapped pages */
702 struct page *vmalloc_to_page(const void *addr);
703 unsigned long vmalloc_to_pfn(const void *addr);
706 * Determine if an address is within the vmalloc range
708 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
709 * is no special casing required.
712 #ifndef is_ioremap_addr
713 #define is_ioremap_addr(x) is_vmalloc_addr(x)
717 extern bool is_vmalloc_addr(const void *x);
718 extern int is_vmalloc_or_module_addr(const void *x);
720 static inline bool is_vmalloc_addr(const void *x)
724 static inline int is_vmalloc_or_module_addr(const void *x)
730 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
731 static inline void *kvmalloc(size_t size, gfp_t flags)
733 return kvmalloc_node(size, flags, NUMA_NO_NODE);
735 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
737 return kvmalloc_node(size, flags | __GFP_ZERO, node);
739 static inline void *kvzalloc(size_t size, gfp_t flags)
741 return kvmalloc(size, flags | __GFP_ZERO);
744 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
748 if (unlikely(check_mul_overflow(n, size, &bytes)))
751 return kvmalloc(bytes, flags);
754 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
756 return kvmalloc_array(n, size, flags | __GFP_ZERO);
759 extern void kvfree(const void *addr);
761 static inline int compound_mapcount(struct page *page)
763 VM_BUG_ON_PAGE(!PageCompound(page), page);
764 page = compound_head(page);
765 return atomic_read(compound_mapcount_ptr(page)) + 1;
769 * The atomic page->_mapcount, starts from -1: so that transitions
770 * both from it and to it can be tracked, using atomic_inc_and_test
771 * and atomic_add_negative(-1).
773 static inline void page_mapcount_reset(struct page *page)
775 atomic_set(&(page)->_mapcount, -1);
778 int __page_mapcount(struct page *page);
780 static inline int page_mapcount(struct page *page)
782 VM_BUG_ON_PAGE(PageSlab(page), page);
784 if (unlikely(PageCompound(page)))
785 return __page_mapcount(page);
786 return atomic_read(&page->_mapcount) + 1;
789 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
790 int total_mapcount(struct page *page);
791 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
793 static inline int total_mapcount(struct page *page)
795 return page_mapcount(page);
797 static inline int page_trans_huge_mapcount(struct page *page,
800 int mapcount = page_mapcount(page);
802 *total_mapcount = mapcount;
807 static inline struct page *virt_to_head_page(const void *x)
809 struct page *page = virt_to_page(x);
811 return compound_head(page);
814 void __put_page(struct page *page);
816 void put_pages_list(struct list_head *pages);
818 void split_page(struct page *page, unsigned int order);
821 * Compound pages have a destructor function. Provide a
822 * prototype for that function and accessor functions.
823 * These are _only_ valid on the head of a compound page.
825 typedef void compound_page_dtor(struct page *);
827 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
828 enum compound_dtor_id {
831 #ifdef CONFIG_HUGETLB_PAGE
834 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
839 extern compound_page_dtor * const compound_page_dtors[];
841 static inline void set_compound_page_dtor(struct page *page,
842 enum compound_dtor_id compound_dtor)
844 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
845 page[1].compound_dtor = compound_dtor;
848 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
850 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
851 return compound_page_dtors[page[1].compound_dtor];
854 static inline unsigned int compound_order(struct page *page)
858 return page[1].compound_order;
861 static inline bool hpage_pincount_available(struct page *page)
864 * Can the page->hpage_pinned_refcount field be used? That field is in
865 * the 3rd page of the compound page, so the smallest (2-page) compound
866 * pages cannot support it.
868 page = compound_head(page);
869 return PageCompound(page) && compound_order(page) > 1;
872 static inline int compound_pincount(struct page *page)
874 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
875 page = compound_head(page);
876 return atomic_read(compound_pincount_ptr(page));
879 static inline void set_compound_order(struct page *page, unsigned int order)
881 page[1].compound_order = order;
884 /* Returns the number of pages in this potentially compound page. */
885 static inline unsigned long compound_nr(struct page *page)
887 return 1UL << compound_order(page);
890 /* Returns the number of bytes in this potentially compound page. */
891 static inline unsigned long page_size(struct page *page)
893 return PAGE_SIZE << compound_order(page);
896 /* Returns the number of bits needed for the number of bytes in a page */
897 static inline unsigned int page_shift(struct page *page)
899 return PAGE_SHIFT + compound_order(page);
902 void free_compound_page(struct page *page);
906 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
907 * servicing faults for write access. In the normal case, do always want
908 * pte_mkwrite. But get_user_pages can cause write faults for mappings
909 * that do not have writing enabled, when used by access_process_vm.
911 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
913 if (likely(vma->vm_flags & VM_WRITE))
914 pte = pte_mkwrite(pte);
918 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
920 vm_fault_t finish_fault(struct vm_fault *vmf);
921 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
925 * Multiple processes may "see" the same page. E.g. for untouched
926 * mappings of /dev/null, all processes see the same page full of
927 * zeroes, and text pages of executables and shared libraries have
928 * only one copy in memory, at most, normally.
930 * For the non-reserved pages, page_count(page) denotes a reference count.
931 * page_count() == 0 means the page is free. page->lru is then used for
932 * freelist management in the buddy allocator.
933 * page_count() > 0 means the page has been allocated.
935 * Pages are allocated by the slab allocator in order to provide memory
936 * to kmalloc and kmem_cache_alloc. In this case, the management of the
937 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
938 * unless a particular usage is carefully commented. (the responsibility of
939 * freeing the kmalloc memory is the caller's, of course).
941 * A page may be used by anyone else who does a __get_free_page().
942 * In this case, page_count still tracks the references, and should only
943 * be used through the normal accessor functions. The top bits of page->flags
944 * and page->virtual store page management information, but all other fields
945 * are unused and could be used privately, carefully. The management of this
946 * page is the responsibility of the one who allocated it, and those who have
947 * subsequently been given references to it.
949 * The other pages (we may call them "pagecache pages") are completely
950 * managed by the Linux memory manager: I/O, buffers, swapping etc.
951 * The following discussion applies only to them.
953 * A pagecache page contains an opaque `private' member, which belongs to the
954 * page's address_space. Usually, this is the address of a circular list of
955 * the page's disk buffers. PG_private must be set to tell the VM to call
956 * into the filesystem to release these pages.
958 * A page may belong to an inode's memory mapping. In this case, page->mapping
959 * is the pointer to the inode, and page->index is the file offset of the page,
960 * in units of PAGE_SIZE.
962 * If pagecache pages are not associated with an inode, they are said to be
963 * anonymous pages. These may become associated with the swapcache, and in that
964 * case PG_swapcache is set, and page->private is an offset into the swapcache.
966 * In either case (swapcache or inode backed), the pagecache itself holds one
967 * reference to the page. Setting PG_private should also increment the
968 * refcount. The each user mapping also has a reference to the page.
970 * The pagecache pages are stored in a per-mapping radix tree, which is
971 * rooted at mapping->i_pages, and indexed by offset.
972 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
973 * lists, we instead now tag pages as dirty/writeback in the radix tree.
975 * All pagecache pages may be subject to I/O:
976 * - inode pages may need to be read from disk,
977 * - inode pages which have been modified and are MAP_SHARED may need
978 * to be written back to the inode on disk,
979 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
980 * modified may need to be swapped out to swap space and (later) to be read
985 * The zone field is never updated after free_area_init_core()
986 * sets it, so none of the operations on it need to be atomic.
989 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
990 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
991 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
992 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
993 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
994 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
997 * Define the bit shifts to access each section. For non-existent
998 * sections we define the shift as 0; that plus a 0 mask ensures
999 * the compiler will optimise away reference to them.
1001 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1002 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1003 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1004 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1005 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1007 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1008 #ifdef NODE_NOT_IN_PAGE_FLAGS
1009 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1010 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1011 SECTIONS_PGOFF : ZONES_PGOFF)
1013 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1014 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1015 NODES_PGOFF : ZONES_PGOFF)
1018 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1020 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1021 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1022 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1023 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1024 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1025 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1027 static inline enum zone_type page_zonenum(const struct page *page)
1029 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1032 #ifdef CONFIG_ZONE_DEVICE
1033 static inline bool is_zone_device_page(const struct page *page)
1035 return page_zonenum(page) == ZONE_DEVICE;
1037 extern void memmap_init_zone_device(struct zone *, unsigned long,
1038 unsigned long, struct dev_pagemap *);
1040 static inline bool is_zone_device_page(const struct page *page)
1046 #ifdef CONFIG_DEV_PAGEMAP_OPS
1047 void free_devmap_managed_page(struct page *page);
1048 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1050 static inline bool page_is_devmap_managed(struct page *page)
1052 if (!static_branch_unlikely(&devmap_managed_key))
1054 if (!is_zone_device_page(page))
1056 switch (page->pgmap->type) {
1057 case MEMORY_DEVICE_PRIVATE:
1058 case MEMORY_DEVICE_FS_DAX:
1066 void put_devmap_managed_page(struct page *page);
1068 #else /* CONFIG_DEV_PAGEMAP_OPS */
1069 static inline bool page_is_devmap_managed(struct page *page)
1074 static inline void put_devmap_managed_page(struct page *page)
1077 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1079 static inline bool is_device_private_page(const struct page *page)
1081 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1082 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1083 is_zone_device_page(page) &&
1084 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1087 static inline bool is_pci_p2pdma_page(const struct page *page)
1089 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1090 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1091 is_zone_device_page(page) &&
1092 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1095 /* 127: arbitrary random number, small enough to assemble well */
1096 #define page_ref_zero_or_close_to_overflow(page) \
1097 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1099 static inline void get_page(struct page *page)
1101 page = compound_head(page);
1103 * Getting a normal page or the head of a compound page
1104 * requires to already have an elevated page->_refcount.
1106 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1110 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1112 static inline __must_check bool try_get_page(struct page *page)
1114 page = compound_head(page);
1115 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1121 static inline void put_page(struct page *page)
1123 page = compound_head(page);
1126 * For devmap managed pages we need to catch refcount transition from
1127 * 2 to 1, when refcount reach one it means the page is free and we
1128 * need to inform the device driver through callback. See
1129 * include/linux/memremap.h and HMM for details.
1131 if (page_is_devmap_managed(page)) {
1132 put_devmap_managed_page(page);
1136 if (put_page_testzero(page))
1141 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1142 * the page's refcount so that two separate items are tracked: the original page
1143 * reference count, and also a new count of how many pin_user_pages() calls were
1144 * made against the page. ("gup-pinned" is another term for the latter).
1146 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1147 * distinct from normal pages. As such, the unpin_user_page() call (and its
1148 * variants) must be used in order to release gup-pinned pages.
1152 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1153 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1154 * simpler, due to the fact that adding an even power of two to the page
1155 * refcount has the effect of using only the upper N bits, for the code that
1156 * counts up using the bias value. This means that the lower bits are left for
1157 * the exclusive use of the original code that increments and decrements by one
1158 * (or at least, by much smaller values than the bias value).
1160 * Of course, once the lower bits overflow into the upper bits (and this is
1161 * OK, because subtraction recovers the original values), then visual inspection
1162 * no longer suffices to directly view the separate counts. However, for normal
1163 * applications that don't have huge page reference counts, this won't be an
1166 * Locking: the lockless algorithm described in page_cache_get_speculative()
1167 * and page_cache_gup_pin_speculative() provides safe operation for
1168 * get_user_pages and page_mkclean and other calls that race to set up page
1171 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1173 void unpin_user_page(struct page *page);
1174 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1176 void unpin_user_pages(struct page **pages, unsigned long npages);
1179 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1181 * This function checks if a page has been pinned via a call to
1182 * pin_user_pages*().
1184 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1185 * because it means "definitely not pinned for DMA", but true means "probably
1186 * pinned for DMA, but possibly a false positive due to having at least
1187 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1189 * False positives are OK, because: a) it's unlikely for a page to get that many
1190 * refcounts, and b) all the callers of this routine are expected to be able to
1191 * deal gracefully with a false positive.
1193 * For huge pages, the result will be exactly correct. That's because we have
1194 * more tracking data available: the 3rd struct page in the compound page is
1195 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1198 * For more information, please see Documentation/vm/pin_user_pages.rst.
1200 * @page: pointer to page to be queried.
1201 * @Return: True, if it is likely that the page has been "dma-pinned".
1202 * False, if the page is definitely not dma-pinned.
1204 static inline bool page_maybe_dma_pinned(struct page *page)
1206 if (hpage_pincount_available(page))
1207 return compound_pincount(page) > 0;
1210 * page_ref_count() is signed. If that refcount overflows, then
1211 * page_ref_count() returns a negative value, and callers will avoid
1212 * further incrementing the refcount.
1214 * Here, for that overflow case, use the signed bit to count a little
1215 * bit higher via unsigned math, and thus still get an accurate result.
1217 return ((unsigned int)page_ref_count(compound_head(page))) >=
1218 GUP_PIN_COUNTING_BIAS;
1221 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1222 #define SECTION_IN_PAGE_FLAGS
1226 * The identification function is mainly used by the buddy allocator for
1227 * determining if two pages could be buddies. We are not really identifying
1228 * the zone since we could be using the section number id if we do not have
1229 * node id available in page flags.
1230 * We only guarantee that it will return the same value for two combinable
1233 static inline int page_zone_id(struct page *page)
1235 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1238 #ifdef NODE_NOT_IN_PAGE_FLAGS
1239 extern int page_to_nid(const struct page *page);
1241 static inline int page_to_nid(const struct page *page)
1243 struct page *p = (struct page *)page;
1245 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1249 #ifdef CONFIG_NUMA_BALANCING
1250 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1252 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1255 static inline int cpupid_to_pid(int cpupid)
1257 return cpupid & LAST__PID_MASK;
1260 static inline int cpupid_to_cpu(int cpupid)
1262 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1265 static inline int cpupid_to_nid(int cpupid)
1267 return cpu_to_node(cpupid_to_cpu(cpupid));
1270 static inline bool cpupid_pid_unset(int cpupid)
1272 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1275 static inline bool cpupid_cpu_unset(int cpupid)
1277 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1280 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1282 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1285 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1286 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1287 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1289 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1292 static inline int page_cpupid_last(struct page *page)
1294 return page->_last_cpupid;
1296 static inline void page_cpupid_reset_last(struct page *page)
1298 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1301 static inline int page_cpupid_last(struct page *page)
1303 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1306 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1308 static inline void page_cpupid_reset_last(struct page *page)
1310 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1312 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1313 #else /* !CONFIG_NUMA_BALANCING */
1314 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1316 return page_to_nid(page); /* XXX */
1319 static inline int page_cpupid_last(struct page *page)
1321 return page_to_nid(page); /* XXX */
1324 static inline int cpupid_to_nid(int cpupid)
1329 static inline int cpupid_to_pid(int cpupid)
1334 static inline int cpupid_to_cpu(int cpupid)
1339 static inline int cpu_pid_to_cpupid(int nid, int pid)
1344 static inline bool cpupid_pid_unset(int cpupid)
1349 static inline void page_cpupid_reset_last(struct page *page)
1353 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1357 #endif /* CONFIG_NUMA_BALANCING */
1359 #ifdef CONFIG_KASAN_SW_TAGS
1360 static inline u8 page_kasan_tag(const struct page *page)
1362 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1365 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1367 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1368 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1371 static inline void page_kasan_tag_reset(struct page *page)
1373 page_kasan_tag_set(page, 0xff);
1376 static inline u8 page_kasan_tag(const struct page *page)
1381 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1382 static inline void page_kasan_tag_reset(struct page *page) { }
1385 static inline struct zone *page_zone(const struct page *page)
1387 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1390 static inline pg_data_t *page_pgdat(const struct page *page)
1392 return NODE_DATA(page_to_nid(page));
1395 #ifdef SECTION_IN_PAGE_FLAGS
1396 static inline void set_page_section(struct page *page, unsigned long section)
1398 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1399 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1402 static inline unsigned long page_to_section(const struct page *page)
1404 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1408 static inline void set_page_zone(struct page *page, enum zone_type zone)
1410 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1411 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1414 static inline void set_page_node(struct page *page, unsigned long node)
1416 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1417 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1420 static inline void set_page_links(struct page *page, enum zone_type zone,
1421 unsigned long node, unsigned long pfn)
1423 set_page_zone(page, zone);
1424 set_page_node(page, node);
1425 #ifdef SECTION_IN_PAGE_FLAGS
1426 set_page_section(page, pfn_to_section_nr(pfn));
1431 static inline struct mem_cgroup *page_memcg(struct page *page)
1433 return page->mem_cgroup;
1435 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1437 WARN_ON_ONCE(!rcu_read_lock_held());
1438 return READ_ONCE(page->mem_cgroup);
1441 static inline struct mem_cgroup *page_memcg(struct page *page)
1445 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1447 WARN_ON_ONCE(!rcu_read_lock_held());
1453 * Some inline functions in vmstat.h depend on page_zone()
1455 #include <linux/vmstat.h>
1457 static __always_inline void *lowmem_page_address(const struct page *page)
1459 return page_to_virt(page);
1462 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1463 #define HASHED_PAGE_VIRTUAL
1466 #if defined(WANT_PAGE_VIRTUAL)
1467 static inline void *page_address(const struct page *page)
1469 return page->virtual;
1471 static inline void set_page_address(struct page *page, void *address)
1473 page->virtual = address;
1475 #define page_address_init() do { } while(0)
1478 #if defined(HASHED_PAGE_VIRTUAL)
1479 void *page_address(const struct page *page);
1480 void set_page_address(struct page *page, void *virtual);
1481 void page_address_init(void);
1484 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1485 #define page_address(page) lowmem_page_address(page)
1486 #define set_page_address(page, address) do { } while(0)
1487 #define page_address_init() do { } while(0)
1490 extern void *page_rmapping(struct page *page);
1491 extern struct anon_vma *page_anon_vma(struct page *page);
1492 extern struct address_space *page_mapping(struct page *page);
1494 extern struct address_space *__page_file_mapping(struct page *);
1497 struct address_space *page_file_mapping(struct page *page)
1499 if (unlikely(PageSwapCache(page)))
1500 return __page_file_mapping(page);
1502 return page->mapping;
1505 extern pgoff_t __page_file_index(struct page *page);
1508 * Return the pagecache index of the passed page. Regular pagecache pages
1509 * use ->index whereas swapcache pages use swp_offset(->private)
1511 static inline pgoff_t page_index(struct page *page)
1513 if (unlikely(PageSwapCache(page)))
1514 return __page_file_index(page);
1518 bool page_mapped(struct page *page);
1519 struct address_space *page_mapping(struct page *page);
1520 struct address_space *page_mapping_file(struct page *page);
1523 * Return true only if the page has been allocated with
1524 * ALLOC_NO_WATERMARKS and the low watermark was not
1525 * met implying that the system is under some pressure.
1527 static inline bool page_is_pfmemalloc(struct page *page)
1530 * Page index cannot be this large so this must be
1531 * a pfmemalloc page.
1533 return page->index == -1UL;
1537 * Only to be called by the page allocator on a freshly allocated
1540 static inline void set_page_pfmemalloc(struct page *page)
1545 static inline void clear_page_pfmemalloc(struct page *page)
1551 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1553 extern void pagefault_out_of_memory(void);
1555 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1558 * Flags passed to show_mem() and show_free_areas() to suppress output in
1561 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1563 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1566 extern bool can_do_mlock(void);
1568 static inline bool can_do_mlock(void) { return false; }
1570 extern int user_shm_lock(size_t, struct user_struct *);
1571 extern void user_shm_unlock(size_t, struct user_struct *);
1574 * Parameter block passed down to zap_pte_range in exceptional cases.
1576 struct zap_details {
1577 struct address_space *check_mapping; /* Check page->mapping if set */
1578 pgoff_t first_index; /* Lowest page->index to unmap */
1579 pgoff_t last_index; /* Highest page->index to unmap */
1582 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1584 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1587 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1588 unsigned long size);
1589 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1590 unsigned long size);
1591 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1592 unsigned long start, unsigned long end);
1594 struct mmu_notifier_range;
1596 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1597 unsigned long end, unsigned long floor, unsigned long ceiling);
1598 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1599 struct vm_area_struct *vma);
1600 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1601 struct mmu_notifier_range *range,
1602 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1603 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1604 unsigned long *pfn);
1605 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1606 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1607 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1608 void *buf, int len, int write);
1610 extern void truncate_pagecache(struct inode *inode, loff_t new);
1611 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1612 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1613 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1614 int truncate_inode_page(struct address_space *mapping, struct page *page);
1615 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1616 int invalidate_inode_page(struct page *page);
1619 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1620 unsigned long address, unsigned int flags);
1621 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1622 unsigned long address, unsigned int fault_flags,
1624 void unmap_mapping_pages(struct address_space *mapping,
1625 pgoff_t start, pgoff_t nr, bool even_cows);
1626 void unmap_mapping_range(struct address_space *mapping,
1627 loff_t const holebegin, loff_t const holelen, int even_cows);
1629 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1630 unsigned long address, unsigned int flags)
1632 /* should never happen if there's no MMU */
1634 return VM_FAULT_SIGBUS;
1636 static inline int fixup_user_fault(struct task_struct *tsk,
1637 struct mm_struct *mm, unsigned long address,
1638 unsigned int fault_flags, bool *unlocked)
1640 /* should never happen if there's no MMU */
1644 static inline void unmap_mapping_pages(struct address_space *mapping,
1645 pgoff_t start, pgoff_t nr, bool even_cows) { }
1646 static inline void unmap_mapping_range(struct address_space *mapping,
1647 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1650 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1651 loff_t const holebegin, loff_t const holelen)
1653 unmap_mapping_range(mapping, holebegin, holelen, 0);
1656 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1657 void *buf, int len, unsigned int gup_flags);
1658 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1659 void *buf, int len, unsigned int gup_flags);
1660 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1661 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1663 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1664 unsigned long start, unsigned long nr_pages,
1665 unsigned int gup_flags, struct page **pages,
1666 struct vm_area_struct **vmas, int *locked);
1667 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1668 unsigned long start, unsigned long nr_pages,
1669 unsigned int gup_flags, struct page **pages,
1670 struct vm_area_struct **vmas, int *locked);
1671 long get_user_pages(unsigned long start, unsigned long nr_pages,
1672 unsigned int gup_flags, struct page **pages,
1673 struct vm_area_struct **vmas);
1674 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1675 unsigned int gup_flags, struct page **pages,
1676 struct vm_area_struct **vmas);
1677 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1678 unsigned int gup_flags, struct page **pages, int *locked);
1679 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1680 struct page **pages, unsigned int gup_flags);
1682 int get_user_pages_fast(unsigned long start, int nr_pages,
1683 unsigned int gup_flags, struct page **pages);
1684 int pin_user_pages_fast(unsigned long start, int nr_pages,
1685 unsigned int gup_flags, struct page **pages);
1687 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1688 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1689 struct task_struct *task, bool bypass_rlim);
1691 /* Container for pinned pfns / pages */
1692 struct frame_vector {
1693 unsigned int nr_allocated; /* Number of frames we have space for */
1694 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1695 bool got_ref; /* Did we pin pages by getting page ref? */
1696 bool is_pfns; /* Does array contain pages or pfns? */
1697 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1698 * pfns_vector_pages() or pfns_vector_pfns()
1702 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1703 void frame_vector_destroy(struct frame_vector *vec);
1704 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1705 unsigned int gup_flags, struct frame_vector *vec);
1706 void put_vaddr_frames(struct frame_vector *vec);
1707 int frame_vector_to_pages(struct frame_vector *vec);
1708 void frame_vector_to_pfns(struct frame_vector *vec);
1710 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1712 return vec->nr_frames;
1715 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1718 int err = frame_vector_to_pages(vec);
1721 return ERR_PTR(err);
1723 return (struct page **)(vec->ptrs);
1726 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1729 frame_vector_to_pfns(vec);
1730 return (unsigned long *)(vec->ptrs);
1734 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1735 struct page **pages);
1736 int get_kernel_page(unsigned long start, int write, struct page **pages);
1737 struct page *get_dump_page(unsigned long addr);
1739 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1740 extern void do_invalidatepage(struct page *page, unsigned int offset,
1741 unsigned int length);
1743 void __set_page_dirty(struct page *, struct address_space *, int warn);
1744 int __set_page_dirty_nobuffers(struct page *page);
1745 int __set_page_dirty_no_writeback(struct page *page);
1746 int redirty_page_for_writepage(struct writeback_control *wbc,
1748 void account_page_dirtied(struct page *page, struct address_space *mapping);
1749 void account_page_cleaned(struct page *page, struct address_space *mapping,
1750 struct bdi_writeback *wb);
1751 int set_page_dirty(struct page *page);
1752 int set_page_dirty_lock(struct page *page);
1753 void __cancel_dirty_page(struct page *page);
1754 static inline void cancel_dirty_page(struct page *page)
1756 /* Avoid atomic ops, locking, etc. when not actually needed. */
1757 if (PageDirty(page))
1758 __cancel_dirty_page(page);
1760 int clear_page_dirty_for_io(struct page *page);
1762 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1764 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1765 unsigned long old_addr, struct vm_area_struct *new_vma,
1766 unsigned long new_addr, unsigned long len,
1767 bool need_rmap_locks);
1768 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1769 unsigned long end, pgprot_t newprot,
1770 int dirty_accountable, int prot_numa);
1771 extern int mprotect_fixup(struct vm_area_struct *vma,
1772 struct vm_area_struct **pprev, unsigned long start,
1773 unsigned long end, unsigned long newflags);
1776 * doesn't attempt to fault and will return short.
1778 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1779 struct page **pages);
1781 * per-process(per-mm_struct) statistics.
1783 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1785 long val = atomic_long_read(&mm->rss_stat.count[member]);
1787 #ifdef SPLIT_RSS_COUNTING
1789 * counter is updated in asynchronous manner and may go to minus.
1790 * But it's never be expected number for users.
1795 return (unsigned long)val;
1798 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1800 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1802 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1804 mm_trace_rss_stat(mm, member, count);
1807 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1809 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1811 mm_trace_rss_stat(mm, member, count);
1814 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1816 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1818 mm_trace_rss_stat(mm, member, count);
1821 /* Optimized variant when page is already known not to be PageAnon */
1822 static inline int mm_counter_file(struct page *page)
1824 if (PageSwapBacked(page))
1825 return MM_SHMEMPAGES;
1826 return MM_FILEPAGES;
1829 static inline int mm_counter(struct page *page)
1832 return MM_ANONPAGES;
1833 return mm_counter_file(page);
1836 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1838 return get_mm_counter(mm, MM_FILEPAGES) +
1839 get_mm_counter(mm, MM_ANONPAGES) +
1840 get_mm_counter(mm, MM_SHMEMPAGES);
1843 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1845 return max(mm->hiwater_rss, get_mm_rss(mm));
1848 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1850 return max(mm->hiwater_vm, mm->total_vm);
1853 static inline void update_hiwater_rss(struct mm_struct *mm)
1855 unsigned long _rss = get_mm_rss(mm);
1857 if ((mm)->hiwater_rss < _rss)
1858 (mm)->hiwater_rss = _rss;
1861 static inline void update_hiwater_vm(struct mm_struct *mm)
1863 if (mm->hiwater_vm < mm->total_vm)
1864 mm->hiwater_vm = mm->total_vm;
1867 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1869 mm->hiwater_rss = get_mm_rss(mm);
1872 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1873 struct mm_struct *mm)
1875 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1877 if (*maxrss < hiwater_rss)
1878 *maxrss = hiwater_rss;
1881 #if defined(SPLIT_RSS_COUNTING)
1882 void sync_mm_rss(struct mm_struct *mm);
1884 static inline void sync_mm_rss(struct mm_struct *mm)
1889 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1890 static inline int pte_devmap(pte_t pte)
1896 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1898 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1900 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1904 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1908 #ifdef __PAGETABLE_P4D_FOLDED
1909 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1910 unsigned long address)
1915 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1918 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1919 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1920 unsigned long address)
1924 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1925 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1928 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1930 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1932 if (mm_pud_folded(mm))
1934 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1937 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1939 if (mm_pud_folded(mm))
1941 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1945 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1946 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1947 unsigned long address)
1952 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1953 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1956 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1958 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1960 if (mm_pmd_folded(mm))
1962 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1965 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1967 if (mm_pmd_folded(mm))
1969 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1974 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1976 atomic_long_set(&mm->pgtables_bytes, 0);
1979 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1981 return atomic_long_read(&mm->pgtables_bytes);
1984 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1986 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1989 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1991 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1995 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1996 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2001 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2002 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2005 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2006 int __pte_alloc_kernel(pmd_t *pmd);
2008 #if defined(CONFIG_MMU)
2011 * The following ifdef needed to get the 5level-fixup.h header to work.
2012 * Remove it when 5level-fixup.h has been removed.
2014 #ifndef __ARCH_HAS_5LEVEL_HACK
2015 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2016 unsigned long address)
2018 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2019 NULL : p4d_offset(pgd, address);
2022 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2023 unsigned long address)
2025 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2026 NULL : pud_offset(p4d, address);
2028 #endif /* !__ARCH_HAS_5LEVEL_HACK */
2030 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2032 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2033 NULL: pmd_offset(pud, address);
2035 #endif /* CONFIG_MMU */
2037 #if USE_SPLIT_PTE_PTLOCKS
2038 #if ALLOC_SPLIT_PTLOCKS
2039 void __init ptlock_cache_init(void);
2040 extern bool ptlock_alloc(struct page *page);
2041 extern void ptlock_free(struct page *page);
2043 static inline spinlock_t *ptlock_ptr(struct page *page)
2047 #else /* ALLOC_SPLIT_PTLOCKS */
2048 static inline void ptlock_cache_init(void)
2052 static inline bool ptlock_alloc(struct page *page)
2057 static inline void ptlock_free(struct page *page)
2061 static inline spinlock_t *ptlock_ptr(struct page *page)
2065 #endif /* ALLOC_SPLIT_PTLOCKS */
2067 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2069 return ptlock_ptr(pmd_page(*pmd));
2072 static inline bool ptlock_init(struct page *page)
2075 * prep_new_page() initialize page->private (and therefore page->ptl)
2076 * with 0. Make sure nobody took it in use in between.
2078 * It can happen if arch try to use slab for page table allocation:
2079 * slab code uses page->slab_cache, which share storage with page->ptl.
2081 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2082 if (!ptlock_alloc(page))
2084 spin_lock_init(ptlock_ptr(page));
2088 #else /* !USE_SPLIT_PTE_PTLOCKS */
2090 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2092 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2094 return &mm->page_table_lock;
2096 static inline void ptlock_cache_init(void) {}
2097 static inline bool ptlock_init(struct page *page) { return true; }
2098 static inline void ptlock_free(struct page *page) {}
2099 #endif /* USE_SPLIT_PTE_PTLOCKS */
2101 static inline void pgtable_init(void)
2103 ptlock_cache_init();
2104 pgtable_cache_init();
2107 static inline bool pgtable_pte_page_ctor(struct page *page)
2109 if (!ptlock_init(page))
2111 __SetPageTable(page);
2112 inc_zone_page_state(page, NR_PAGETABLE);
2116 static inline void pgtable_pte_page_dtor(struct page *page)
2119 __ClearPageTable(page);
2120 dec_zone_page_state(page, NR_PAGETABLE);
2123 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2125 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2126 pte_t *__pte = pte_offset_map(pmd, address); \
2132 #define pte_unmap_unlock(pte, ptl) do { \
2137 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2139 #define pte_alloc_map(mm, pmd, address) \
2140 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2142 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2143 (pte_alloc(mm, pmd) ? \
2144 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2146 #define pte_alloc_kernel(pmd, address) \
2147 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2148 NULL: pte_offset_kernel(pmd, address))
2150 #if USE_SPLIT_PMD_PTLOCKS
2152 static struct page *pmd_to_page(pmd_t *pmd)
2154 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2155 return virt_to_page((void *)((unsigned long) pmd & mask));
2158 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2160 return ptlock_ptr(pmd_to_page(pmd));
2163 static inline bool pgtable_pmd_page_ctor(struct page *page)
2165 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2166 page->pmd_huge_pte = NULL;
2168 return ptlock_init(page);
2171 static inline void pgtable_pmd_page_dtor(struct page *page)
2173 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2174 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2179 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2183 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2185 return &mm->page_table_lock;
2188 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2189 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2191 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2195 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2197 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2203 * No scalability reason to split PUD locks yet, but follow the same pattern
2204 * as the PMD locks to make it easier if we decide to. The VM should not be
2205 * considered ready to switch to split PUD locks yet; there may be places
2206 * which need to be converted from page_table_lock.
2208 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2210 return &mm->page_table_lock;
2213 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2215 spinlock_t *ptl = pud_lockptr(mm, pud);
2221 extern void __init pagecache_init(void);
2222 extern void free_area_init(unsigned long * zones_size);
2223 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2224 unsigned long zone_start_pfn, unsigned long *zholes_size);
2225 extern void free_initmem(void);
2228 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2229 * into the buddy system. The freed pages will be poisoned with pattern
2230 * "poison" if it's within range [0, UCHAR_MAX].
2231 * Return pages freed into the buddy system.
2233 extern unsigned long free_reserved_area(void *start, void *end,
2234 int poison, const char *s);
2236 #ifdef CONFIG_HIGHMEM
2238 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2239 * and totalram_pages.
2241 extern void free_highmem_page(struct page *page);
2244 extern void adjust_managed_page_count(struct page *page, long count);
2245 extern void mem_init_print_info(const char *str);
2247 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2249 /* Free the reserved page into the buddy system, so it gets managed. */
2250 static inline void __free_reserved_page(struct page *page)
2252 ClearPageReserved(page);
2253 init_page_count(page);
2257 static inline void free_reserved_page(struct page *page)
2259 __free_reserved_page(page);
2260 adjust_managed_page_count(page, 1);
2263 static inline void mark_page_reserved(struct page *page)
2265 SetPageReserved(page);
2266 adjust_managed_page_count(page, -1);
2270 * Default method to free all the __init memory into the buddy system.
2271 * The freed pages will be poisoned with pattern "poison" if it's within
2272 * range [0, UCHAR_MAX].
2273 * Return pages freed into the buddy system.
2275 static inline unsigned long free_initmem_default(int poison)
2277 extern char __init_begin[], __init_end[];
2279 return free_reserved_area(&__init_begin, &__init_end,
2280 poison, "unused kernel");
2283 static inline unsigned long get_num_physpages(void)
2286 unsigned long phys_pages = 0;
2288 for_each_online_node(nid)
2289 phys_pages += node_present_pages(nid);
2294 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2296 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2297 * zones, allocate the backing mem_map and account for memory holes in a more
2298 * architecture independent manner. This is a substitute for creating the
2299 * zone_sizes[] and zholes_size[] arrays and passing them to
2300 * free_area_init_node()
2302 * An architecture is expected to register range of page frames backed by
2303 * physical memory with memblock_add[_node]() before calling
2304 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2305 * usage, an architecture is expected to do something like
2307 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2309 * for_each_valid_physical_page_range()
2310 * memblock_add_node(base, size, nid)
2311 * free_area_init_nodes(max_zone_pfns);
2313 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2314 * registered physical page range. Similarly
2315 * sparse_memory_present_with_active_regions() calls memory_present() for
2316 * each range when SPARSEMEM is enabled.
2318 * See mm/page_alloc.c for more information on each function exposed by
2319 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2321 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2322 unsigned long node_map_pfn_alignment(void);
2323 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2324 unsigned long end_pfn);
2325 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2326 unsigned long end_pfn);
2327 extern void get_pfn_range_for_nid(unsigned int nid,
2328 unsigned long *start_pfn, unsigned long *end_pfn);
2329 extern unsigned long find_min_pfn_with_active_regions(void);
2330 extern void free_bootmem_with_active_regions(int nid,
2331 unsigned long max_low_pfn);
2332 extern void sparse_memory_present_with_active_regions(int nid);
2334 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2336 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2337 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2338 static inline int __early_pfn_to_nid(unsigned long pfn,
2339 struct mminit_pfnnid_cache *state)
2344 /* please see mm/page_alloc.c */
2345 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2346 /* there is a per-arch backend function. */
2347 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2348 struct mminit_pfnnid_cache *state);
2351 extern void set_dma_reserve(unsigned long new_dma_reserve);
2352 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2353 enum memmap_context, struct vmem_altmap *);
2354 extern void setup_per_zone_wmarks(void);
2355 extern int __meminit init_per_zone_wmark_min(void);
2356 extern void mem_init(void);
2357 extern void __init mmap_init(void);
2358 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2359 extern long si_mem_available(void);
2360 extern void si_meminfo(struct sysinfo * val);
2361 extern void si_meminfo_node(struct sysinfo *val, int nid);
2362 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2363 extern unsigned long arch_reserved_kernel_pages(void);
2366 extern __printf(3, 4)
2367 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2369 extern void setup_per_cpu_pageset(void);
2372 extern int min_free_kbytes;
2373 extern int watermark_boost_factor;
2374 extern int watermark_scale_factor;
2377 extern atomic_long_t mmap_pages_allocated;
2378 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2380 /* interval_tree.c */
2381 void vma_interval_tree_insert(struct vm_area_struct *node,
2382 struct rb_root_cached *root);
2383 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2384 struct vm_area_struct *prev,
2385 struct rb_root_cached *root);
2386 void vma_interval_tree_remove(struct vm_area_struct *node,
2387 struct rb_root_cached *root);
2388 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2389 unsigned long start, unsigned long last);
2390 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2391 unsigned long start, unsigned long last);
2393 #define vma_interval_tree_foreach(vma, root, start, last) \
2394 for (vma = vma_interval_tree_iter_first(root, start, last); \
2395 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2397 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2398 struct rb_root_cached *root);
2399 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2400 struct rb_root_cached *root);
2401 struct anon_vma_chain *
2402 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2403 unsigned long start, unsigned long last);
2404 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2405 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2406 #ifdef CONFIG_DEBUG_VM_RB
2407 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2410 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2411 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2412 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2415 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2416 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2417 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2418 struct vm_area_struct *expand);
2419 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2420 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2422 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2424 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2425 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2426 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2427 struct mempolicy *, struct vm_userfaultfd_ctx);
2428 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2429 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2430 unsigned long addr, int new_below);
2431 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2432 unsigned long addr, int new_below);
2433 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2434 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2435 struct rb_node **, struct rb_node *);
2436 extern void unlink_file_vma(struct vm_area_struct *);
2437 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2438 unsigned long addr, unsigned long len, pgoff_t pgoff,
2439 bool *need_rmap_locks);
2440 extern void exit_mmap(struct mm_struct *);
2442 static inline int check_data_rlimit(unsigned long rlim,
2444 unsigned long start,
2445 unsigned long end_data,
2446 unsigned long start_data)
2448 if (rlim < RLIM_INFINITY) {
2449 if (((new - start) + (end_data - start_data)) > rlim)
2456 extern int mm_take_all_locks(struct mm_struct *mm);
2457 extern void mm_drop_all_locks(struct mm_struct *mm);
2459 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2460 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2461 extern struct file *get_task_exe_file(struct task_struct *task);
2463 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2464 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2466 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2467 const struct vm_special_mapping *sm);
2468 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2469 unsigned long addr, unsigned long len,
2470 unsigned long flags,
2471 const struct vm_special_mapping *spec);
2472 /* This is an obsolete alternative to _install_special_mapping. */
2473 extern int install_special_mapping(struct mm_struct *mm,
2474 unsigned long addr, unsigned long len,
2475 unsigned long flags, struct page **pages);
2477 unsigned long randomize_stack_top(unsigned long stack_top);
2479 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2481 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2482 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2483 struct list_head *uf);
2484 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2485 unsigned long len, unsigned long prot, unsigned long flags,
2486 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2487 struct list_head *uf);
2488 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2489 struct list_head *uf, bool downgrade);
2490 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2491 struct list_head *uf);
2492 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2494 static inline unsigned long
2495 do_mmap_pgoff(struct file *file, unsigned long addr,
2496 unsigned long len, unsigned long prot, unsigned long flags,
2497 unsigned long pgoff, unsigned long *populate,
2498 struct list_head *uf)
2500 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2504 extern int __mm_populate(unsigned long addr, unsigned long len,
2506 static inline void mm_populate(unsigned long addr, unsigned long len)
2509 (void) __mm_populate(addr, len, 1);
2512 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2515 /* These take the mm semaphore themselves */
2516 extern int __must_check vm_brk(unsigned long, unsigned long);
2517 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2518 extern int vm_munmap(unsigned long, size_t);
2519 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2520 unsigned long, unsigned long,
2521 unsigned long, unsigned long);
2523 struct vm_unmapped_area_info {
2524 #define VM_UNMAPPED_AREA_TOPDOWN 1
2525 unsigned long flags;
2526 unsigned long length;
2527 unsigned long low_limit;
2528 unsigned long high_limit;
2529 unsigned long align_mask;
2530 unsigned long align_offset;
2533 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2536 extern void truncate_inode_pages(struct address_space *, loff_t);
2537 extern void truncate_inode_pages_range(struct address_space *,
2538 loff_t lstart, loff_t lend);
2539 extern void truncate_inode_pages_final(struct address_space *);
2541 /* generic vm_area_ops exported for stackable file systems */
2542 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2543 extern void filemap_map_pages(struct vm_fault *vmf,
2544 pgoff_t start_pgoff, pgoff_t end_pgoff);
2545 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2547 /* mm/page-writeback.c */
2548 int __must_check write_one_page(struct page *page);
2549 void task_dirty_inc(struct task_struct *tsk);
2552 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2554 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2555 pgoff_t offset, unsigned long nr_to_read);
2557 void page_cache_sync_readahead(struct address_space *mapping,
2558 struct file_ra_state *ra,
2561 unsigned long size);
2563 void page_cache_async_readahead(struct address_space *mapping,
2564 struct file_ra_state *ra,
2568 unsigned long size);
2570 extern unsigned long stack_guard_gap;
2571 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2572 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2574 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2575 extern int expand_downwards(struct vm_area_struct *vma,
2576 unsigned long address);
2578 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2580 #define expand_upwards(vma, address) (0)
2583 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2584 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2585 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2586 struct vm_area_struct **pprev);
2588 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2589 NULL if none. Assume start_addr < end_addr. */
2590 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2592 struct vm_area_struct * vma = find_vma(mm,start_addr);
2594 if (vma && end_addr <= vma->vm_start)
2599 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2601 unsigned long vm_start = vma->vm_start;
2603 if (vma->vm_flags & VM_GROWSDOWN) {
2604 vm_start -= stack_guard_gap;
2605 if (vm_start > vma->vm_start)
2611 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2613 unsigned long vm_end = vma->vm_end;
2615 if (vma->vm_flags & VM_GROWSUP) {
2616 vm_end += stack_guard_gap;
2617 if (vm_end < vma->vm_end)
2618 vm_end = -PAGE_SIZE;
2623 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2625 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2628 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2629 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2630 unsigned long vm_start, unsigned long vm_end)
2632 struct vm_area_struct *vma = find_vma(mm, vm_start);
2634 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2640 static inline bool range_in_vma(struct vm_area_struct *vma,
2641 unsigned long start, unsigned long end)
2643 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2647 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2648 void vma_set_page_prot(struct vm_area_struct *vma);
2650 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2654 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2656 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2660 #ifdef CONFIG_NUMA_BALANCING
2661 unsigned long change_prot_numa(struct vm_area_struct *vma,
2662 unsigned long start, unsigned long end);
2665 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2666 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2667 unsigned long pfn, unsigned long size, pgprot_t);
2668 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2669 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2671 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2673 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2675 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2676 unsigned long pfn, pgprot_t pgprot);
2677 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2679 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2680 pfn_t pfn, pgprot_t pgprot);
2681 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2682 unsigned long addr, pfn_t pfn);
2683 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2685 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2686 unsigned long addr, struct page *page)
2688 int err = vm_insert_page(vma, addr, page);
2691 return VM_FAULT_OOM;
2692 if (err < 0 && err != -EBUSY)
2693 return VM_FAULT_SIGBUS;
2695 return VM_FAULT_NOPAGE;
2698 static inline vm_fault_t vmf_error(int err)
2701 return VM_FAULT_OOM;
2702 return VM_FAULT_SIGBUS;
2705 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2706 unsigned int foll_flags);
2708 #define FOLL_WRITE 0x01 /* check pte is writable */
2709 #define FOLL_TOUCH 0x02 /* mark page accessed */
2710 #define FOLL_GET 0x04 /* do get_page on page */
2711 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2712 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2713 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2714 * and return without waiting upon it */
2715 #define FOLL_POPULATE 0x40 /* fault in page */
2716 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2717 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2718 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2719 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2720 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2721 #define FOLL_MLOCK 0x1000 /* lock present pages */
2722 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2723 #define FOLL_COW 0x4000 /* internal GUP flag */
2724 #define FOLL_ANON 0x8000 /* don't do file mappings */
2725 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2726 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2727 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2730 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2731 * other. Here is what they mean, and how to use them:
2733 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2734 * period _often_ under userspace control. This is in contrast to
2735 * iov_iter_get_pages(), whose usages are transient.
2737 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2738 * lifetime enforced by the filesystem and we need guarantees that longterm
2739 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2740 * the filesystem. Ideas for this coordination include revoking the longterm
2741 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2742 * added after the problem with filesystems was found FS DAX VMAs are
2743 * specifically failed. Filesystem pages are still subject to bugs and use of
2744 * FOLL_LONGTERM should be avoided on those pages.
2746 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2747 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2748 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2749 * is due to an incompatibility with the FS DAX check and
2750 * FAULT_FLAG_ALLOW_RETRY.
2752 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2753 * that region. And so, CMA attempts to migrate the page before pinning, when
2754 * FOLL_LONGTERM is specified.
2756 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2757 * but an additional pin counting system) will be invoked. This is intended for
2758 * anything that gets a page reference and then touches page data (for example,
2759 * Direct IO). This lets the filesystem know that some non-file-system entity is
2760 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2761 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2762 * a call to unpin_user_page().
2764 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2765 * and separate refcounting mechanisms, however, and that means that each has
2766 * its own acquire and release mechanisms:
2768 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2770 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2772 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2773 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2774 * calls applied to them, and that's perfectly OK. This is a constraint on the
2775 * callers, not on the pages.)
2777 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2778 * directly by the caller. That's in order to help avoid mismatches when
2779 * releasing pages: get_user_pages*() pages must be released via put_page(),
2780 * while pin_user_pages*() pages must be released via unpin_user_page().
2782 * Please see Documentation/vm/pin_user_pages.rst for more information.
2785 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2787 if (vm_fault & VM_FAULT_OOM)
2789 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2790 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2791 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2796 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2797 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2798 unsigned long size, pte_fn_t fn, void *data);
2799 extern int apply_to_existing_page_range(struct mm_struct *mm,
2800 unsigned long address, unsigned long size,
2801 pte_fn_t fn, void *data);
2803 #ifdef CONFIG_PAGE_POISONING
2804 extern bool page_poisoning_enabled(void);
2805 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2807 static inline bool page_poisoning_enabled(void) { return false; }
2808 static inline void kernel_poison_pages(struct page *page, int numpages,
2812 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2813 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2815 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2817 static inline bool want_init_on_alloc(gfp_t flags)
2819 if (static_branch_unlikely(&init_on_alloc) &&
2820 !page_poisoning_enabled())
2822 return flags & __GFP_ZERO;
2825 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2826 DECLARE_STATIC_KEY_TRUE(init_on_free);
2828 DECLARE_STATIC_KEY_FALSE(init_on_free);
2830 static inline bool want_init_on_free(void)
2832 return static_branch_unlikely(&init_on_free) &&
2833 !page_poisoning_enabled();
2836 #ifdef CONFIG_DEBUG_PAGEALLOC
2837 extern void init_debug_pagealloc(void);
2839 static inline void init_debug_pagealloc(void) {}
2841 extern bool _debug_pagealloc_enabled_early;
2842 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2844 static inline bool debug_pagealloc_enabled(void)
2846 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2847 _debug_pagealloc_enabled_early;
2851 * For use in fast paths after init_debug_pagealloc() has run, or when a
2852 * false negative result is not harmful when called too early.
2854 static inline bool debug_pagealloc_enabled_static(void)
2856 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2859 return static_branch_unlikely(&_debug_pagealloc_enabled);
2862 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2863 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2866 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2867 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2870 kernel_map_pages(struct page *page, int numpages, int enable)
2872 __kernel_map_pages(page, numpages, enable);
2874 #ifdef CONFIG_HIBERNATION
2875 extern bool kernel_page_present(struct page *page);
2876 #endif /* CONFIG_HIBERNATION */
2877 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2879 kernel_map_pages(struct page *page, int numpages, int enable) {}
2880 #ifdef CONFIG_HIBERNATION
2881 static inline bool kernel_page_present(struct page *page) { return true; }
2882 #endif /* CONFIG_HIBERNATION */
2883 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2885 #ifdef __HAVE_ARCH_GATE_AREA
2886 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2887 extern int in_gate_area_no_mm(unsigned long addr);
2888 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2890 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2894 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2895 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2899 #endif /* __HAVE_ARCH_GATE_AREA */
2901 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2903 #ifdef CONFIG_SYSCTL
2904 extern int sysctl_drop_caches;
2905 int drop_caches_sysctl_handler(struct ctl_table *, int,
2906 void __user *, size_t *, loff_t *);
2909 void drop_slab(void);
2910 void drop_slab_node(int nid);
2913 #define randomize_va_space 0
2915 extern int randomize_va_space;
2918 const char * arch_vma_name(struct vm_area_struct *vma);
2920 void print_vma_addr(char *prefix, unsigned long rip);
2922 static inline void print_vma_addr(char *prefix, unsigned long rip)
2927 void *sparse_buffer_alloc(unsigned long size);
2928 struct page * __populate_section_memmap(unsigned long pfn,
2929 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2930 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2931 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2932 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2933 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2934 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2935 void *vmemmap_alloc_block(unsigned long size, int node);
2937 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2938 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2939 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2940 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2942 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2943 struct vmem_altmap *altmap);
2944 void vmemmap_populate_print_last(void);
2945 #ifdef CONFIG_MEMORY_HOTPLUG
2946 void vmemmap_free(unsigned long start, unsigned long end,
2947 struct vmem_altmap *altmap);
2949 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2950 unsigned long nr_pages);
2953 MF_COUNT_INCREASED = 1 << 0,
2954 MF_ACTION_REQUIRED = 1 << 1,
2955 MF_MUST_KILL = 1 << 2,
2956 MF_SOFT_OFFLINE = 1 << 3,
2958 extern int memory_failure(unsigned long pfn, int flags);
2959 extern void memory_failure_queue(unsigned long pfn, int flags);
2960 extern int unpoison_memory(unsigned long pfn);
2961 extern int get_hwpoison_page(struct page *page);
2962 #define put_hwpoison_page(page) put_page(page)
2963 extern int sysctl_memory_failure_early_kill;
2964 extern int sysctl_memory_failure_recovery;
2965 extern void shake_page(struct page *p, int access);
2966 extern atomic_long_t num_poisoned_pages __read_mostly;
2967 extern int soft_offline_page(unsigned long pfn, int flags);
2971 * Error handlers for various types of pages.
2974 MF_IGNORED, /* Error: cannot be handled */
2975 MF_FAILED, /* Error: handling failed */
2976 MF_DELAYED, /* Will be handled later */
2977 MF_RECOVERED, /* Successfully recovered */
2980 enum mf_action_page_type {
2982 MF_MSG_KERNEL_HIGH_ORDER,
2984 MF_MSG_DIFFERENT_COMPOUND,
2985 MF_MSG_POISONED_HUGE,
2988 MF_MSG_NON_PMD_HUGE,
2989 MF_MSG_UNMAP_FAILED,
2990 MF_MSG_DIRTY_SWAPCACHE,
2991 MF_MSG_CLEAN_SWAPCACHE,
2992 MF_MSG_DIRTY_MLOCKED_LRU,
2993 MF_MSG_CLEAN_MLOCKED_LRU,
2994 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2995 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2998 MF_MSG_TRUNCATED_LRU,
3005 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3006 extern void clear_huge_page(struct page *page,
3007 unsigned long addr_hint,
3008 unsigned int pages_per_huge_page);
3009 extern void copy_user_huge_page(struct page *dst, struct page *src,
3010 unsigned long addr_hint,
3011 struct vm_area_struct *vma,
3012 unsigned int pages_per_huge_page);
3013 extern long copy_huge_page_from_user(struct page *dst_page,
3014 const void __user *usr_src,
3015 unsigned int pages_per_huge_page,
3016 bool allow_pagefault);
3017 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3019 #ifdef CONFIG_DEBUG_PAGEALLOC
3020 extern unsigned int _debug_guardpage_minorder;
3021 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3023 static inline unsigned int debug_guardpage_minorder(void)
3025 return _debug_guardpage_minorder;
3028 static inline bool debug_guardpage_enabled(void)
3030 return static_branch_unlikely(&_debug_guardpage_enabled);
3033 static inline bool page_is_guard(struct page *page)
3035 if (!debug_guardpage_enabled())
3038 return PageGuard(page);
3041 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3042 static inline bool debug_guardpage_enabled(void) { return false; }
3043 static inline bool page_is_guard(struct page *page) { return false; }
3044 #endif /* CONFIG_DEBUG_PAGEALLOC */
3046 #if MAX_NUMNODES > 1
3047 void __init setup_nr_node_ids(void);
3049 static inline void setup_nr_node_ids(void) {}
3052 extern int memcmp_pages(struct page *page1, struct page *page2);
3054 static inline int pages_identical(struct page *page1, struct page *page2)
3056 return !memcmp_pages(page1, page2);
3059 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3060 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3061 pgoff_t first_index, pgoff_t nr,
3062 pgoff_t bitmap_pgoff,
3063 unsigned long *bitmap,
3067 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3068 pgoff_t first_index, pgoff_t nr);
3071 #endif /* __KERNEL__ */
3072 #endif /* _LINUX_MM_H */