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>
33 struct anon_vma_chain;
36 struct writeback_control;
39 void init_mm_internals(void);
41 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
42 extern unsigned long max_mapnr;
44 static inline void set_max_mapnr(unsigned long limit)
49 static inline void set_max_mapnr(unsigned long limit) { }
52 extern atomic_long_t _totalram_pages;
53 static inline unsigned long totalram_pages(void)
55 return (unsigned long)atomic_long_read(&_totalram_pages);
58 static inline void totalram_pages_inc(void)
60 atomic_long_inc(&_totalram_pages);
63 static inline void totalram_pages_dec(void)
65 atomic_long_dec(&_totalram_pages);
68 static inline void totalram_pages_add(long count)
70 atomic_long_add(count, &_totalram_pages);
73 extern void * high_memory;
74 extern int page_cluster;
77 extern int sysctl_legacy_va_layout;
79 #define sysctl_legacy_va_layout 0
82 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
83 extern const int mmap_rnd_bits_min;
84 extern const int mmap_rnd_bits_max;
85 extern int mmap_rnd_bits __read_mostly;
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
88 extern const int mmap_rnd_compat_bits_min;
89 extern const int mmap_rnd_compat_bits_max;
90 extern int mmap_rnd_compat_bits __read_mostly;
94 #include <asm/pgtable.h>
95 #include <asm/processor.h>
98 * Architectures that support memory tagging (assigning tags to memory regions,
99 * embedding these tags into addresses that point to these memory regions, and
100 * checking that the memory and the pointer tags match on memory accesses)
101 * redefine this macro to strip tags from pointers.
102 * It's defined as noop for arcitectures that don't support memory tagging.
104 #ifndef untagged_addr
105 #define untagged_addr(addr) (addr)
109 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
113 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
117 #define lm_alias(x) __va(__pa_symbol(x))
121 * To prevent common memory management code establishing
122 * a zero page mapping on a read fault.
123 * This macro should be defined within <asm/pgtable.h>.
124 * s390 does this to prevent multiplexing of hardware bits
125 * related to the physical page in case of virtualization.
127 #ifndef mm_forbids_zeropage
128 #define mm_forbids_zeropage(X) (0)
132 * On some architectures it is expensive to call memset() for small sizes.
133 * If an architecture decides to implement their own version of
134 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
135 * define their own version of this macro in <asm/pgtable.h>
137 #if BITS_PER_LONG == 64
138 /* This function must be updated when the size of struct page grows above 80
139 * or reduces below 56. The idea that compiler optimizes out switch()
140 * statement, and only leaves move/store instructions. Also the compiler can
141 * combine write statments if they are both assignments and can be reordered,
142 * this can result in several of the writes here being dropped.
144 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
145 static inline void __mm_zero_struct_page(struct page *page)
147 unsigned long *_pp = (void *)page;
149 /* Check that struct page is either 56, 64, 72, or 80 bytes */
150 BUILD_BUG_ON(sizeof(struct page) & 7);
151 BUILD_BUG_ON(sizeof(struct page) < 56);
152 BUILD_BUG_ON(sizeof(struct page) > 80);
154 switch (sizeof(struct page)) {
156 _pp[9] = 0; /* fallthrough */
158 _pp[8] = 0; /* fallthrough */
160 _pp[7] = 0; /* fallthrough */
172 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
176 * Default maximum number of active map areas, this limits the number of vmas
177 * per mm struct. Users can overwrite this number by sysctl but there is a
180 * When a program's coredump is generated as ELF format, a section is created
181 * per a vma. In ELF, the number of sections is represented in unsigned short.
182 * This means the number of sections should be smaller than 65535 at coredump.
183 * Because the kernel adds some informative sections to a image of program at
184 * generating coredump, we need some margin. The number of extra sections is
185 * 1-3 now and depends on arch. We use "5" as safe margin, here.
187 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
188 * not a hard limit any more. Although some userspace tools can be surprised by
191 #define MAPCOUNT_ELF_CORE_MARGIN (5)
192 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
194 extern int sysctl_max_map_count;
196 extern unsigned long sysctl_user_reserve_kbytes;
197 extern unsigned long sysctl_admin_reserve_kbytes;
199 extern int sysctl_overcommit_memory;
200 extern int sysctl_overcommit_ratio;
201 extern unsigned long sysctl_overcommit_kbytes;
203 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
205 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
208 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
210 /* to align the pointer to the (next) page boundary */
211 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
213 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
214 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
216 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
219 * Linux kernel virtual memory manager primitives.
220 * The idea being to have a "virtual" mm in the same way
221 * we have a virtual fs - giving a cleaner interface to the
222 * mm details, and allowing different kinds of memory mappings
223 * (from shared memory to executable loading to arbitrary
227 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
228 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
229 void vm_area_free(struct vm_area_struct *);
232 extern struct rb_root nommu_region_tree;
233 extern struct rw_semaphore nommu_region_sem;
235 extern unsigned int kobjsize(const void *objp);
239 * vm_flags in vm_area_struct, see mm_types.h.
240 * When changing, update also include/trace/events/mmflags.h
242 #define VM_NONE 0x00000000
244 #define VM_READ 0x00000001 /* currently active flags */
245 #define VM_WRITE 0x00000002
246 #define VM_EXEC 0x00000004
247 #define VM_SHARED 0x00000008
249 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
250 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
251 #define VM_MAYWRITE 0x00000020
252 #define VM_MAYEXEC 0x00000040
253 #define VM_MAYSHARE 0x00000080
255 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
256 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
257 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
258 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
259 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
261 #define VM_LOCKED 0x00002000
262 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
264 /* Used by sys_madvise() */
265 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
266 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
268 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
269 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
270 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
271 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
272 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
273 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
274 #define VM_SYNC 0x00800000 /* Synchronous page faults */
275 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
276 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
277 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
279 #ifdef CONFIG_MEM_SOFT_DIRTY
280 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
282 # define VM_SOFTDIRTY 0
285 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
286 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
287 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
288 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
290 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
291 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
292 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
293 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
294 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
295 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
296 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
297 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
298 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
299 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
300 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
301 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
303 #ifdef CONFIG_ARCH_HAS_PKEYS
304 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
305 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
306 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
307 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
308 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
310 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
312 # define VM_PKEY_BIT4 0
314 #endif /* CONFIG_ARCH_HAS_PKEYS */
316 #if defined(CONFIG_X86)
317 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
318 #elif defined(CONFIG_PPC)
319 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
320 #elif defined(CONFIG_PARISC)
321 # define VM_GROWSUP VM_ARCH_1
322 #elif defined(CONFIG_IA64)
323 # define VM_GROWSUP VM_ARCH_1
324 #elif defined(CONFIG_SPARC64)
325 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
326 # define VM_ARCH_CLEAR VM_SPARC_ADI
327 #elif !defined(CONFIG_MMU)
328 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
331 #if defined(CONFIG_X86_INTEL_MPX)
332 /* MPX specific bounds table or bounds directory */
333 # define VM_MPX VM_HIGH_ARCH_4
335 # define VM_MPX VM_NONE
339 # define VM_GROWSUP VM_NONE
342 /* Bits set in the VMA until the stack is in its final location */
343 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
345 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
346 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
349 #ifdef CONFIG_STACK_GROWSUP
350 #define VM_STACK VM_GROWSUP
352 #define VM_STACK VM_GROWSDOWN
355 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
358 * Special vmas that are non-mergable, non-mlock()able.
359 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
361 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
363 /* This mask defines which mm->def_flags a process can inherit its parent */
364 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
366 /* This mask is used to clear all the VMA flags used by mlock */
367 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
369 /* Arch-specific flags to clear when updating VM flags on protection change */
370 #ifndef VM_ARCH_CLEAR
371 # define VM_ARCH_CLEAR VM_NONE
373 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
376 * mapping from the currently active vm_flags protection bits (the
377 * low four bits) to a page protection mask..
379 extern pgprot_t protection_map[16];
381 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
382 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
383 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
384 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
385 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
386 #define FAULT_FLAG_TRIED 0x20 /* Second try */
387 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
388 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
389 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
391 #define FAULT_FLAG_TRACE \
392 { FAULT_FLAG_WRITE, "WRITE" }, \
393 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
394 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
395 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
396 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
397 { FAULT_FLAG_TRIED, "TRIED" }, \
398 { FAULT_FLAG_USER, "USER" }, \
399 { FAULT_FLAG_REMOTE, "REMOTE" }, \
400 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
403 * vm_fault is filled by the the pagefault handler and passed to the vma's
404 * ->fault function. The vma's ->fault is responsible for returning a bitmask
405 * of VM_FAULT_xxx flags that give details about how the fault was handled.
407 * MM layer fills up gfp_mask for page allocations but fault handler might
408 * alter it if its implementation requires a different allocation context.
410 * pgoff should be used in favour of virtual_address, if possible.
413 struct vm_area_struct *vma; /* Target VMA */
414 unsigned int flags; /* FAULT_FLAG_xxx flags */
415 gfp_t gfp_mask; /* gfp mask to be used for allocations */
416 pgoff_t pgoff; /* Logical page offset based on vma */
417 unsigned long address; /* Faulting virtual address */
418 pmd_t *pmd; /* Pointer to pmd entry matching
420 pud_t *pud; /* Pointer to pud entry matching
423 pte_t orig_pte; /* Value of PTE at the time of fault */
425 struct page *cow_page; /* Page handler may use for COW fault */
426 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
427 struct page *page; /* ->fault handlers should return a
428 * page here, unless VM_FAULT_NOPAGE
429 * is set (which is also implied by
432 /* These three entries are valid only while holding ptl lock */
433 pte_t *pte; /* Pointer to pte entry matching
434 * the 'address'. NULL if the page
435 * table hasn't been allocated.
437 spinlock_t *ptl; /* Page table lock.
438 * Protects pte page table if 'pte'
439 * is not NULL, otherwise pmd.
441 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
442 * vm_ops->map_pages() calls
443 * alloc_set_pte() from atomic context.
444 * do_fault_around() pre-allocates
445 * page table to avoid allocation from
450 /* page entry size for vm->huge_fault() */
451 enum page_entry_size {
458 * These are the virtual MM functions - opening of an area, closing and
459 * unmapping it (needed to keep files on disk up-to-date etc), pointer
460 * to the functions called when a no-page or a wp-page exception occurs.
462 struct vm_operations_struct {
463 void (*open)(struct vm_area_struct * area);
464 void (*close)(struct vm_area_struct * area);
465 int (*split)(struct vm_area_struct * area, unsigned long addr);
466 int (*mremap)(struct vm_area_struct * area);
467 vm_fault_t (*fault)(struct vm_fault *vmf);
468 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
469 enum page_entry_size pe_size);
470 void (*map_pages)(struct vm_fault *vmf,
471 pgoff_t start_pgoff, pgoff_t end_pgoff);
472 unsigned long (*pagesize)(struct vm_area_struct * area);
474 /* notification that a previously read-only page is about to become
475 * writable, if an error is returned it will cause a SIGBUS */
476 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
478 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
479 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
481 /* called by access_process_vm when get_user_pages() fails, typically
482 * for use by special VMAs that can switch between memory and hardware
484 int (*access)(struct vm_area_struct *vma, unsigned long addr,
485 void *buf, int len, int write);
487 /* Called by the /proc/PID/maps code to ask the vma whether it
488 * has a special name. Returning non-NULL will also cause this
489 * vma to be dumped unconditionally. */
490 const char *(*name)(struct vm_area_struct *vma);
494 * set_policy() op must add a reference to any non-NULL @new mempolicy
495 * to hold the policy upon return. Caller should pass NULL @new to
496 * remove a policy and fall back to surrounding context--i.e. do not
497 * install a MPOL_DEFAULT policy, nor the task or system default
500 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
503 * get_policy() op must add reference [mpol_get()] to any policy at
504 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
505 * in mm/mempolicy.c will do this automatically.
506 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
507 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
508 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
509 * must return NULL--i.e., do not "fallback" to task or system default
512 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
516 * Called by vm_normal_page() for special PTEs to find the
517 * page for @addr. This is useful if the default behavior
518 * (using pte_page()) would not find the correct page.
520 struct page *(*find_special_page)(struct vm_area_struct *vma,
524 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
526 static const struct vm_operations_struct dummy_vm_ops = {};
528 memset(vma, 0, sizeof(*vma));
530 vma->vm_ops = &dummy_vm_ops;
531 INIT_LIST_HEAD(&vma->anon_vma_chain);
534 static inline void vma_set_anonymous(struct vm_area_struct *vma)
539 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
546 * The vma_is_shmem is not inline because it is used only by slow
547 * paths in userfault.
549 bool vma_is_shmem(struct vm_area_struct *vma);
551 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
554 int vma_is_stack_for_current(struct vm_area_struct *vma);
556 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
557 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
563 * FIXME: take this include out, include page-flags.h in
564 * files which need it (119 of them)
566 #include <linux/page-flags.h>
567 #include <linux/huge_mm.h>
570 * Methods to modify the page usage count.
572 * What counts for a page usage:
573 * - cache mapping (page->mapping)
574 * - private data (page->private)
575 * - page mapped in a task's page tables, each mapping
576 * is counted separately
578 * Also, many kernel routines increase the page count before a critical
579 * routine so they can be sure the page doesn't go away from under them.
583 * Drop a ref, return true if the refcount fell to zero (the page has no users)
585 static inline int put_page_testzero(struct page *page)
587 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
588 return page_ref_dec_and_test(page);
592 * Try to grab a ref unless the page has a refcount of zero, return false if
594 * This can be called when MMU is off so it must not access
595 * any of the virtual mappings.
597 static inline int get_page_unless_zero(struct page *page)
599 return page_ref_add_unless(page, 1, 0);
602 extern int page_is_ram(unsigned long pfn);
610 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
613 /* Support for virtually mapped pages */
614 struct page *vmalloc_to_page(const void *addr);
615 unsigned long vmalloc_to_pfn(const void *addr);
618 * Determine if an address is within the vmalloc range
620 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
621 * is no special casing required.
624 #ifndef is_ioremap_addr
625 #define is_ioremap_addr(x) is_vmalloc_addr(x)
629 extern bool is_vmalloc_addr(const void *x);
630 extern int is_vmalloc_or_module_addr(const void *x);
632 static inline bool is_vmalloc_addr(const void *x)
636 static inline int is_vmalloc_or_module_addr(const void *x)
642 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
643 static inline void *kvmalloc(size_t size, gfp_t flags)
645 return kvmalloc_node(size, flags, NUMA_NO_NODE);
647 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
649 return kvmalloc_node(size, flags | __GFP_ZERO, node);
651 static inline void *kvzalloc(size_t size, gfp_t flags)
653 return kvmalloc(size, flags | __GFP_ZERO);
656 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
660 if (unlikely(check_mul_overflow(n, size, &bytes)))
663 return kvmalloc(bytes, flags);
666 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
668 return kvmalloc_array(n, size, flags | __GFP_ZERO);
671 extern void kvfree(const void *addr);
673 static inline int compound_mapcount(struct page *page)
675 VM_BUG_ON_PAGE(!PageCompound(page), page);
676 page = compound_head(page);
677 return atomic_read(compound_mapcount_ptr(page)) + 1;
681 * The atomic page->_mapcount, starts from -1: so that transitions
682 * both from it and to it can be tracked, using atomic_inc_and_test
683 * and atomic_add_negative(-1).
685 static inline void page_mapcount_reset(struct page *page)
687 atomic_set(&(page)->_mapcount, -1);
690 int __page_mapcount(struct page *page);
692 static inline int page_mapcount(struct page *page)
694 VM_BUG_ON_PAGE(PageSlab(page), page);
696 if (unlikely(PageCompound(page)))
697 return __page_mapcount(page);
698 return atomic_read(&page->_mapcount) + 1;
701 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
702 int total_mapcount(struct page *page);
703 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
705 static inline int total_mapcount(struct page *page)
707 return page_mapcount(page);
709 static inline int page_trans_huge_mapcount(struct page *page,
712 int mapcount = page_mapcount(page);
714 *total_mapcount = mapcount;
719 static inline struct page *virt_to_head_page(const void *x)
721 struct page *page = virt_to_page(x);
723 return compound_head(page);
726 void __put_page(struct page *page);
728 void put_pages_list(struct list_head *pages);
730 void split_page(struct page *page, unsigned int order);
733 * Compound pages have a destructor function. Provide a
734 * prototype for that function and accessor functions.
735 * These are _only_ valid on the head of a compound page.
737 typedef void compound_page_dtor(struct page *);
739 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
740 enum compound_dtor_id {
743 #ifdef CONFIG_HUGETLB_PAGE
746 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
751 extern compound_page_dtor * const compound_page_dtors[];
753 static inline void set_compound_page_dtor(struct page *page,
754 enum compound_dtor_id compound_dtor)
756 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
757 page[1].compound_dtor = compound_dtor;
760 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
762 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
763 return compound_page_dtors[page[1].compound_dtor];
766 static inline unsigned int compound_order(struct page *page)
770 return page[1].compound_order;
773 static inline void set_compound_order(struct page *page, unsigned int order)
775 page[1].compound_order = order;
778 /* Returns the number of pages in this potentially compound page. */
779 static inline unsigned long compound_nr(struct page *page)
781 return 1UL << compound_order(page);
784 /* Returns the number of bytes in this potentially compound page. */
785 static inline unsigned long page_size(struct page *page)
787 return PAGE_SIZE << compound_order(page);
790 /* Returns the number of bits needed for the number of bytes in a page */
791 static inline unsigned int page_shift(struct page *page)
793 return PAGE_SHIFT + compound_order(page);
796 void free_compound_page(struct page *page);
800 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
801 * servicing faults for write access. In the normal case, do always want
802 * pte_mkwrite. But get_user_pages can cause write faults for mappings
803 * that do not have writing enabled, when used by access_process_vm.
805 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
807 if (likely(vma->vm_flags & VM_WRITE))
808 pte = pte_mkwrite(pte);
812 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
814 vm_fault_t finish_fault(struct vm_fault *vmf);
815 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
819 * Multiple processes may "see" the same page. E.g. for untouched
820 * mappings of /dev/null, all processes see the same page full of
821 * zeroes, and text pages of executables and shared libraries have
822 * only one copy in memory, at most, normally.
824 * For the non-reserved pages, page_count(page) denotes a reference count.
825 * page_count() == 0 means the page is free. page->lru is then used for
826 * freelist management in the buddy allocator.
827 * page_count() > 0 means the page has been allocated.
829 * Pages are allocated by the slab allocator in order to provide memory
830 * to kmalloc and kmem_cache_alloc. In this case, the management of the
831 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
832 * unless a particular usage is carefully commented. (the responsibility of
833 * freeing the kmalloc memory is the caller's, of course).
835 * A page may be used by anyone else who does a __get_free_page().
836 * In this case, page_count still tracks the references, and should only
837 * be used through the normal accessor functions. The top bits of page->flags
838 * and page->virtual store page management information, but all other fields
839 * are unused and could be used privately, carefully. The management of this
840 * page is the responsibility of the one who allocated it, and those who have
841 * subsequently been given references to it.
843 * The other pages (we may call them "pagecache pages") are completely
844 * managed by the Linux memory manager: I/O, buffers, swapping etc.
845 * The following discussion applies only to them.
847 * A pagecache page contains an opaque `private' member, which belongs to the
848 * page's address_space. Usually, this is the address of a circular list of
849 * the page's disk buffers. PG_private must be set to tell the VM to call
850 * into the filesystem to release these pages.
852 * A page may belong to an inode's memory mapping. In this case, page->mapping
853 * is the pointer to the inode, and page->index is the file offset of the page,
854 * in units of PAGE_SIZE.
856 * If pagecache pages are not associated with an inode, they are said to be
857 * anonymous pages. These may become associated with the swapcache, and in that
858 * case PG_swapcache is set, and page->private is an offset into the swapcache.
860 * In either case (swapcache or inode backed), the pagecache itself holds one
861 * reference to the page. Setting PG_private should also increment the
862 * refcount. The each user mapping also has a reference to the page.
864 * The pagecache pages are stored in a per-mapping radix tree, which is
865 * rooted at mapping->i_pages, and indexed by offset.
866 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
867 * lists, we instead now tag pages as dirty/writeback in the radix tree.
869 * All pagecache pages may be subject to I/O:
870 * - inode pages may need to be read from disk,
871 * - inode pages which have been modified and are MAP_SHARED may need
872 * to be written back to the inode on disk,
873 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
874 * modified may need to be swapped out to swap space and (later) to be read
879 * The zone field is never updated after free_area_init_core()
880 * sets it, so none of the operations on it need to be atomic.
883 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
884 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
885 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
886 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
887 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
888 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
891 * Define the bit shifts to access each section. For non-existent
892 * sections we define the shift as 0; that plus a 0 mask ensures
893 * the compiler will optimise away reference to them.
895 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
896 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
897 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
898 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
899 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
901 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
902 #ifdef NODE_NOT_IN_PAGE_FLAGS
903 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
904 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
905 SECTIONS_PGOFF : ZONES_PGOFF)
907 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
908 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
909 NODES_PGOFF : ZONES_PGOFF)
912 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
914 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
915 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
916 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
917 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
918 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
919 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
921 static inline enum zone_type page_zonenum(const struct page *page)
923 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
926 #ifdef CONFIG_ZONE_DEVICE
927 static inline bool is_zone_device_page(const struct page *page)
929 return page_zonenum(page) == ZONE_DEVICE;
931 extern void memmap_init_zone_device(struct zone *, unsigned long,
932 unsigned long, struct dev_pagemap *);
934 static inline bool is_zone_device_page(const struct page *page)
940 #ifdef CONFIG_DEV_PAGEMAP_OPS
941 void free_devmap_managed_page(struct page *page);
942 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
944 static inline bool page_is_devmap_managed(struct page *page)
946 if (!static_branch_unlikely(&devmap_managed_key))
948 if (!is_zone_device_page(page))
950 switch (page->pgmap->type) {
951 case MEMORY_DEVICE_PRIVATE:
952 case MEMORY_DEVICE_FS_DAX:
960 void put_devmap_managed_page(struct page *page);
962 #else /* CONFIG_DEV_PAGEMAP_OPS */
963 static inline bool page_is_devmap_managed(struct page *page)
968 static inline void put_devmap_managed_page(struct page *page)
971 #endif /* CONFIG_DEV_PAGEMAP_OPS */
973 static inline bool is_device_private_page(const struct page *page)
975 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
976 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
977 is_zone_device_page(page) &&
978 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
981 static inline bool is_pci_p2pdma_page(const struct page *page)
983 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
984 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
985 is_zone_device_page(page) &&
986 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
989 /* 127: arbitrary random number, small enough to assemble well */
990 #define page_ref_zero_or_close_to_overflow(page) \
991 ((unsigned int) page_ref_count(page) + 127u <= 127u)
993 static inline void get_page(struct page *page)
995 page = compound_head(page);
997 * Getting a normal page or the head of a compound page
998 * requires to already have an elevated page->_refcount.
1000 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1004 static inline __must_check bool try_get_page(struct page *page)
1006 page = compound_head(page);
1007 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1013 static inline void put_page(struct page *page)
1015 page = compound_head(page);
1018 * For devmap managed pages we need to catch refcount transition from
1019 * 2 to 1, when refcount reach one it means the page is free and we
1020 * need to inform the device driver through callback. See
1021 * include/linux/memremap.h and HMM for details.
1023 if (page_is_devmap_managed(page)) {
1024 put_devmap_managed_page(page);
1028 if (put_page_testzero(page))
1033 * unpin_user_page() - release a gup-pinned page
1034 * @page: pointer to page to be released
1036 * Pages that were pinned via pin_user_pages*() must be released via either
1037 * unpin_user_page(), or one of the unpin_user_pages*() routines. This is so
1038 * that eventually such pages can be separately tracked and uniquely handled. In
1039 * particular, interactions with RDMA and filesystems need special handling.
1041 * unpin_user_page() and put_page() are not interchangeable, despite this early
1042 * implementation that makes them look the same. unpin_user_page() calls must
1043 * be perfectly matched up with pin*() calls.
1045 static inline void unpin_user_page(struct page *page)
1050 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1053 void unpin_user_pages(struct page **pages, unsigned long npages);
1055 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1056 #define SECTION_IN_PAGE_FLAGS
1060 * The identification function is mainly used by the buddy allocator for
1061 * determining if two pages could be buddies. We are not really identifying
1062 * the zone since we could be using the section number id if we do not have
1063 * node id available in page flags.
1064 * We only guarantee that it will return the same value for two combinable
1067 static inline int page_zone_id(struct page *page)
1069 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1072 #ifdef NODE_NOT_IN_PAGE_FLAGS
1073 extern int page_to_nid(const struct page *page);
1075 static inline int page_to_nid(const struct page *page)
1077 struct page *p = (struct page *)page;
1079 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1083 #ifdef CONFIG_NUMA_BALANCING
1084 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1086 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1089 static inline int cpupid_to_pid(int cpupid)
1091 return cpupid & LAST__PID_MASK;
1094 static inline int cpupid_to_cpu(int cpupid)
1096 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1099 static inline int cpupid_to_nid(int cpupid)
1101 return cpu_to_node(cpupid_to_cpu(cpupid));
1104 static inline bool cpupid_pid_unset(int cpupid)
1106 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1109 static inline bool cpupid_cpu_unset(int cpupid)
1111 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1114 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1116 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1119 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1120 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1121 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1123 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1126 static inline int page_cpupid_last(struct page *page)
1128 return page->_last_cpupid;
1130 static inline void page_cpupid_reset_last(struct page *page)
1132 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1135 static inline int page_cpupid_last(struct page *page)
1137 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1140 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1142 static inline void page_cpupid_reset_last(struct page *page)
1144 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1146 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1147 #else /* !CONFIG_NUMA_BALANCING */
1148 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1150 return page_to_nid(page); /* XXX */
1153 static inline int page_cpupid_last(struct page *page)
1155 return page_to_nid(page); /* XXX */
1158 static inline int cpupid_to_nid(int cpupid)
1163 static inline int cpupid_to_pid(int cpupid)
1168 static inline int cpupid_to_cpu(int cpupid)
1173 static inline int cpu_pid_to_cpupid(int nid, int pid)
1178 static inline bool cpupid_pid_unset(int cpupid)
1183 static inline void page_cpupid_reset_last(struct page *page)
1187 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1191 #endif /* CONFIG_NUMA_BALANCING */
1193 #ifdef CONFIG_KASAN_SW_TAGS
1194 static inline u8 page_kasan_tag(const struct page *page)
1196 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1199 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1201 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1202 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1205 static inline void page_kasan_tag_reset(struct page *page)
1207 page_kasan_tag_set(page, 0xff);
1210 static inline u8 page_kasan_tag(const struct page *page)
1215 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1216 static inline void page_kasan_tag_reset(struct page *page) { }
1219 static inline struct zone *page_zone(const struct page *page)
1221 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1224 static inline pg_data_t *page_pgdat(const struct page *page)
1226 return NODE_DATA(page_to_nid(page));
1229 #ifdef SECTION_IN_PAGE_FLAGS
1230 static inline void set_page_section(struct page *page, unsigned long section)
1232 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1233 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1236 static inline unsigned long page_to_section(const struct page *page)
1238 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1242 static inline void set_page_zone(struct page *page, enum zone_type zone)
1244 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1245 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1248 static inline void set_page_node(struct page *page, unsigned long node)
1250 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1251 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1254 static inline void set_page_links(struct page *page, enum zone_type zone,
1255 unsigned long node, unsigned long pfn)
1257 set_page_zone(page, zone);
1258 set_page_node(page, node);
1259 #ifdef SECTION_IN_PAGE_FLAGS
1260 set_page_section(page, pfn_to_section_nr(pfn));
1265 static inline struct mem_cgroup *page_memcg(struct page *page)
1267 return page->mem_cgroup;
1269 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1271 WARN_ON_ONCE(!rcu_read_lock_held());
1272 return READ_ONCE(page->mem_cgroup);
1275 static inline struct mem_cgroup *page_memcg(struct page *page)
1279 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1281 WARN_ON_ONCE(!rcu_read_lock_held());
1287 * Some inline functions in vmstat.h depend on page_zone()
1289 #include <linux/vmstat.h>
1291 static __always_inline void *lowmem_page_address(const struct page *page)
1293 return page_to_virt(page);
1296 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1297 #define HASHED_PAGE_VIRTUAL
1300 #if defined(WANT_PAGE_VIRTUAL)
1301 static inline void *page_address(const struct page *page)
1303 return page->virtual;
1305 static inline void set_page_address(struct page *page, void *address)
1307 page->virtual = address;
1309 #define page_address_init() do { } while(0)
1312 #if defined(HASHED_PAGE_VIRTUAL)
1313 void *page_address(const struct page *page);
1314 void set_page_address(struct page *page, void *virtual);
1315 void page_address_init(void);
1318 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1319 #define page_address(page) lowmem_page_address(page)
1320 #define set_page_address(page, address) do { } while(0)
1321 #define page_address_init() do { } while(0)
1324 extern void *page_rmapping(struct page *page);
1325 extern struct anon_vma *page_anon_vma(struct page *page);
1326 extern struct address_space *page_mapping(struct page *page);
1328 extern struct address_space *__page_file_mapping(struct page *);
1331 struct address_space *page_file_mapping(struct page *page)
1333 if (unlikely(PageSwapCache(page)))
1334 return __page_file_mapping(page);
1336 return page->mapping;
1339 extern pgoff_t __page_file_index(struct page *page);
1342 * Return the pagecache index of the passed page. Regular pagecache pages
1343 * use ->index whereas swapcache pages use swp_offset(->private)
1345 static inline pgoff_t page_index(struct page *page)
1347 if (unlikely(PageSwapCache(page)))
1348 return __page_file_index(page);
1352 bool page_mapped(struct page *page);
1353 struct address_space *page_mapping(struct page *page);
1354 struct address_space *page_mapping_file(struct page *page);
1357 * Return true only if the page has been allocated with
1358 * ALLOC_NO_WATERMARKS and the low watermark was not
1359 * met implying that the system is under some pressure.
1361 static inline bool page_is_pfmemalloc(struct page *page)
1364 * Page index cannot be this large so this must be
1365 * a pfmemalloc page.
1367 return page->index == -1UL;
1371 * Only to be called by the page allocator on a freshly allocated
1374 static inline void set_page_pfmemalloc(struct page *page)
1379 static inline void clear_page_pfmemalloc(struct page *page)
1385 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1387 extern void pagefault_out_of_memory(void);
1389 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1392 * Flags passed to show_mem() and show_free_areas() to suppress output in
1395 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1397 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1400 extern bool can_do_mlock(void);
1402 static inline bool can_do_mlock(void) { return false; }
1404 extern int user_shm_lock(size_t, struct user_struct *);
1405 extern void user_shm_unlock(size_t, struct user_struct *);
1408 * Parameter block passed down to zap_pte_range in exceptional cases.
1410 struct zap_details {
1411 struct address_space *check_mapping; /* Check page->mapping if set */
1412 pgoff_t first_index; /* Lowest page->index to unmap */
1413 pgoff_t last_index; /* Highest page->index to unmap */
1416 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1418 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1421 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1422 unsigned long size);
1423 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1424 unsigned long size);
1425 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1426 unsigned long start, unsigned long end);
1428 struct mmu_notifier_range;
1430 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1431 unsigned long end, unsigned long floor, unsigned long ceiling);
1432 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1433 struct vm_area_struct *vma);
1434 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1435 struct mmu_notifier_range *range,
1436 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1437 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1438 unsigned long *pfn);
1439 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1440 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1441 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1442 void *buf, int len, int write);
1444 extern void truncate_pagecache(struct inode *inode, loff_t new);
1445 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1446 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1447 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1448 int truncate_inode_page(struct address_space *mapping, struct page *page);
1449 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1450 int invalidate_inode_page(struct page *page);
1453 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1454 unsigned long address, unsigned int flags);
1455 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1456 unsigned long address, unsigned int fault_flags,
1458 void unmap_mapping_pages(struct address_space *mapping,
1459 pgoff_t start, pgoff_t nr, bool even_cows);
1460 void unmap_mapping_range(struct address_space *mapping,
1461 loff_t const holebegin, loff_t const holelen, int even_cows);
1463 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1464 unsigned long address, unsigned int flags)
1466 /* should never happen if there's no MMU */
1468 return VM_FAULT_SIGBUS;
1470 static inline int fixup_user_fault(struct task_struct *tsk,
1471 struct mm_struct *mm, unsigned long address,
1472 unsigned int fault_flags, bool *unlocked)
1474 /* should never happen if there's no MMU */
1478 static inline void unmap_mapping_pages(struct address_space *mapping,
1479 pgoff_t start, pgoff_t nr, bool even_cows) { }
1480 static inline void unmap_mapping_range(struct address_space *mapping,
1481 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1484 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1485 loff_t const holebegin, loff_t const holelen)
1487 unmap_mapping_range(mapping, holebegin, holelen, 0);
1490 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1491 void *buf, int len, unsigned int gup_flags);
1492 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1493 void *buf, int len, unsigned int gup_flags);
1494 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1495 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1497 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1498 unsigned long start, unsigned long nr_pages,
1499 unsigned int gup_flags, struct page **pages,
1500 struct vm_area_struct **vmas, int *locked);
1501 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1502 unsigned long start, unsigned long nr_pages,
1503 unsigned int gup_flags, struct page **pages,
1504 struct vm_area_struct **vmas, int *locked);
1505 long get_user_pages(unsigned long start, unsigned long nr_pages,
1506 unsigned int gup_flags, struct page **pages,
1507 struct vm_area_struct **vmas);
1508 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1509 unsigned int gup_flags, struct page **pages,
1510 struct vm_area_struct **vmas);
1511 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1512 unsigned int gup_flags, struct page **pages, int *locked);
1513 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1514 struct page **pages, unsigned int gup_flags);
1516 int get_user_pages_fast(unsigned long start, int nr_pages,
1517 unsigned int gup_flags, struct page **pages);
1518 int pin_user_pages_fast(unsigned long start, int nr_pages,
1519 unsigned int gup_flags, struct page **pages);
1521 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1522 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1523 struct task_struct *task, bool bypass_rlim);
1525 /* Container for pinned pfns / pages */
1526 struct frame_vector {
1527 unsigned int nr_allocated; /* Number of frames we have space for */
1528 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1529 bool got_ref; /* Did we pin pages by getting page ref? */
1530 bool is_pfns; /* Does array contain pages or pfns? */
1531 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1532 * pfns_vector_pages() or pfns_vector_pfns()
1536 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1537 void frame_vector_destroy(struct frame_vector *vec);
1538 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1539 unsigned int gup_flags, struct frame_vector *vec);
1540 void put_vaddr_frames(struct frame_vector *vec);
1541 int frame_vector_to_pages(struct frame_vector *vec);
1542 void frame_vector_to_pfns(struct frame_vector *vec);
1544 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1546 return vec->nr_frames;
1549 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1552 int err = frame_vector_to_pages(vec);
1555 return ERR_PTR(err);
1557 return (struct page **)(vec->ptrs);
1560 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1563 frame_vector_to_pfns(vec);
1564 return (unsigned long *)(vec->ptrs);
1568 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1569 struct page **pages);
1570 int get_kernel_page(unsigned long start, int write, struct page **pages);
1571 struct page *get_dump_page(unsigned long addr);
1573 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1574 extern void do_invalidatepage(struct page *page, unsigned int offset,
1575 unsigned int length);
1577 void __set_page_dirty(struct page *, struct address_space *, int warn);
1578 int __set_page_dirty_nobuffers(struct page *page);
1579 int __set_page_dirty_no_writeback(struct page *page);
1580 int redirty_page_for_writepage(struct writeback_control *wbc,
1582 void account_page_dirtied(struct page *page, struct address_space *mapping);
1583 void account_page_cleaned(struct page *page, struct address_space *mapping,
1584 struct bdi_writeback *wb);
1585 int set_page_dirty(struct page *page);
1586 int set_page_dirty_lock(struct page *page);
1587 void __cancel_dirty_page(struct page *page);
1588 static inline void cancel_dirty_page(struct page *page)
1590 /* Avoid atomic ops, locking, etc. when not actually needed. */
1591 if (PageDirty(page))
1592 __cancel_dirty_page(page);
1594 int clear_page_dirty_for_io(struct page *page);
1596 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1598 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1599 unsigned long old_addr, struct vm_area_struct *new_vma,
1600 unsigned long new_addr, unsigned long len,
1601 bool need_rmap_locks);
1602 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1603 unsigned long end, pgprot_t newprot,
1604 int dirty_accountable, int prot_numa);
1605 extern int mprotect_fixup(struct vm_area_struct *vma,
1606 struct vm_area_struct **pprev, unsigned long start,
1607 unsigned long end, unsigned long newflags);
1610 * doesn't attempt to fault and will return short.
1612 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1613 struct page **pages);
1615 * per-process(per-mm_struct) statistics.
1617 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1619 long val = atomic_long_read(&mm->rss_stat.count[member]);
1621 #ifdef SPLIT_RSS_COUNTING
1623 * counter is updated in asynchronous manner and may go to minus.
1624 * But it's never be expected number for users.
1629 return (unsigned long)val;
1632 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1634 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1636 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1638 mm_trace_rss_stat(mm, member, count);
1641 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1643 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1645 mm_trace_rss_stat(mm, member, count);
1648 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1650 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1652 mm_trace_rss_stat(mm, member, count);
1655 /* Optimized variant when page is already known not to be PageAnon */
1656 static inline int mm_counter_file(struct page *page)
1658 if (PageSwapBacked(page))
1659 return MM_SHMEMPAGES;
1660 return MM_FILEPAGES;
1663 static inline int mm_counter(struct page *page)
1666 return MM_ANONPAGES;
1667 return mm_counter_file(page);
1670 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1672 return get_mm_counter(mm, MM_FILEPAGES) +
1673 get_mm_counter(mm, MM_ANONPAGES) +
1674 get_mm_counter(mm, MM_SHMEMPAGES);
1677 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1679 return max(mm->hiwater_rss, get_mm_rss(mm));
1682 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1684 return max(mm->hiwater_vm, mm->total_vm);
1687 static inline void update_hiwater_rss(struct mm_struct *mm)
1689 unsigned long _rss = get_mm_rss(mm);
1691 if ((mm)->hiwater_rss < _rss)
1692 (mm)->hiwater_rss = _rss;
1695 static inline void update_hiwater_vm(struct mm_struct *mm)
1697 if (mm->hiwater_vm < mm->total_vm)
1698 mm->hiwater_vm = mm->total_vm;
1701 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1703 mm->hiwater_rss = get_mm_rss(mm);
1706 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1707 struct mm_struct *mm)
1709 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1711 if (*maxrss < hiwater_rss)
1712 *maxrss = hiwater_rss;
1715 #if defined(SPLIT_RSS_COUNTING)
1716 void sync_mm_rss(struct mm_struct *mm);
1718 static inline void sync_mm_rss(struct mm_struct *mm)
1723 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1724 static inline int pte_devmap(pte_t pte)
1730 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1732 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1734 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1738 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1742 #ifdef __PAGETABLE_P4D_FOLDED
1743 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1744 unsigned long address)
1749 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1752 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1753 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1754 unsigned long address)
1758 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1759 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1762 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1764 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1766 if (mm_pud_folded(mm))
1768 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1771 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1773 if (mm_pud_folded(mm))
1775 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1779 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1780 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1781 unsigned long address)
1786 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1787 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1790 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1792 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1794 if (mm_pmd_folded(mm))
1796 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1799 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1801 if (mm_pmd_folded(mm))
1803 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1808 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1810 atomic_long_set(&mm->pgtables_bytes, 0);
1813 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1815 return atomic_long_read(&mm->pgtables_bytes);
1818 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1820 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1823 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1825 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1829 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1830 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1835 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1836 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1839 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1840 int __pte_alloc_kernel(pmd_t *pmd);
1842 #if defined(CONFIG_MMU)
1845 * The following ifdef needed to get the 5level-fixup.h header to work.
1846 * Remove it when 5level-fixup.h has been removed.
1848 #ifndef __ARCH_HAS_5LEVEL_HACK
1849 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1850 unsigned long address)
1852 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1853 NULL : p4d_offset(pgd, address);
1856 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1857 unsigned long address)
1859 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1860 NULL : pud_offset(p4d, address);
1862 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1864 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1866 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1867 NULL: pmd_offset(pud, address);
1869 #endif /* CONFIG_MMU */
1871 #if USE_SPLIT_PTE_PTLOCKS
1872 #if ALLOC_SPLIT_PTLOCKS
1873 void __init ptlock_cache_init(void);
1874 extern bool ptlock_alloc(struct page *page);
1875 extern void ptlock_free(struct page *page);
1877 static inline spinlock_t *ptlock_ptr(struct page *page)
1881 #else /* ALLOC_SPLIT_PTLOCKS */
1882 static inline void ptlock_cache_init(void)
1886 static inline bool ptlock_alloc(struct page *page)
1891 static inline void ptlock_free(struct page *page)
1895 static inline spinlock_t *ptlock_ptr(struct page *page)
1899 #endif /* ALLOC_SPLIT_PTLOCKS */
1901 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1903 return ptlock_ptr(pmd_page(*pmd));
1906 static inline bool ptlock_init(struct page *page)
1909 * prep_new_page() initialize page->private (and therefore page->ptl)
1910 * with 0. Make sure nobody took it in use in between.
1912 * It can happen if arch try to use slab for page table allocation:
1913 * slab code uses page->slab_cache, which share storage with page->ptl.
1915 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1916 if (!ptlock_alloc(page))
1918 spin_lock_init(ptlock_ptr(page));
1922 #else /* !USE_SPLIT_PTE_PTLOCKS */
1924 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1926 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1928 return &mm->page_table_lock;
1930 static inline void ptlock_cache_init(void) {}
1931 static inline bool ptlock_init(struct page *page) { return true; }
1932 static inline void ptlock_free(struct page *page) {}
1933 #endif /* USE_SPLIT_PTE_PTLOCKS */
1935 static inline void pgtable_init(void)
1937 ptlock_cache_init();
1938 pgtable_cache_init();
1941 static inline bool pgtable_pte_page_ctor(struct page *page)
1943 if (!ptlock_init(page))
1945 __SetPageTable(page);
1946 inc_zone_page_state(page, NR_PAGETABLE);
1950 static inline void pgtable_pte_page_dtor(struct page *page)
1953 __ClearPageTable(page);
1954 dec_zone_page_state(page, NR_PAGETABLE);
1957 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1959 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1960 pte_t *__pte = pte_offset_map(pmd, address); \
1966 #define pte_unmap_unlock(pte, ptl) do { \
1971 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1973 #define pte_alloc_map(mm, pmd, address) \
1974 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1976 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1977 (pte_alloc(mm, pmd) ? \
1978 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1980 #define pte_alloc_kernel(pmd, address) \
1981 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1982 NULL: pte_offset_kernel(pmd, address))
1984 #if USE_SPLIT_PMD_PTLOCKS
1986 static struct page *pmd_to_page(pmd_t *pmd)
1988 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1989 return virt_to_page((void *)((unsigned long) pmd & mask));
1992 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1994 return ptlock_ptr(pmd_to_page(pmd));
1997 static inline bool pgtable_pmd_page_ctor(struct page *page)
1999 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2000 page->pmd_huge_pte = NULL;
2002 return ptlock_init(page);
2005 static inline void pgtable_pmd_page_dtor(struct page *page)
2007 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2008 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2013 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2017 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2019 return &mm->page_table_lock;
2022 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2023 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2025 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2029 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2031 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2037 * No scalability reason to split PUD locks yet, but follow the same pattern
2038 * as the PMD locks to make it easier if we decide to. The VM should not be
2039 * considered ready to switch to split PUD locks yet; there may be places
2040 * which need to be converted from page_table_lock.
2042 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2044 return &mm->page_table_lock;
2047 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2049 spinlock_t *ptl = pud_lockptr(mm, pud);
2055 extern void __init pagecache_init(void);
2056 extern void free_area_init(unsigned long * zones_size);
2057 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2058 unsigned long zone_start_pfn, unsigned long *zholes_size);
2059 extern void free_initmem(void);
2062 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2063 * into the buddy system. The freed pages will be poisoned with pattern
2064 * "poison" if it's within range [0, UCHAR_MAX].
2065 * Return pages freed into the buddy system.
2067 extern unsigned long free_reserved_area(void *start, void *end,
2068 int poison, const char *s);
2070 #ifdef CONFIG_HIGHMEM
2072 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2073 * and totalram_pages.
2075 extern void free_highmem_page(struct page *page);
2078 extern void adjust_managed_page_count(struct page *page, long count);
2079 extern void mem_init_print_info(const char *str);
2081 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2083 /* Free the reserved page into the buddy system, so it gets managed. */
2084 static inline void __free_reserved_page(struct page *page)
2086 ClearPageReserved(page);
2087 init_page_count(page);
2091 static inline void free_reserved_page(struct page *page)
2093 __free_reserved_page(page);
2094 adjust_managed_page_count(page, 1);
2097 static inline void mark_page_reserved(struct page *page)
2099 SetPageReserved(page);
2100 adjust_managed_page_count(page, -1);
2104 * Default method to free all the __init memory into the buddy system.
2105 * The freed pages will be poisoned with pattern "poison" if it's within
2106 * range [0, UCHAR_MAX].
2107 * Return pages freed into the buddy system.
2109 static inline unsigned long free_initmem_default(int poison)
2111 extern char __init_begin[], __init_end[];
2113 return free_reserved_area(&__init_begin, &__init_end,
2114 poison, "unused kernel");
2117 static inline unsigned long get_num_physpages(void)
2120 unsigned long phys_pages = 0;
2122 for_each_online_node(nid)
2123 phys_pages += node_present_pages(nid);
2128 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2130 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2131 * zones, allocate the backing mem_map and account for memory holes in a more
2132 * architecture independent manner. This is a substitute for creating the
2133 * zone_sizes[] and zholes_size[] arrays and passing them to
2134 * free_area_init_node()
2136 * An architecture is expected to register range of page frames backed by
2137 * physical memory with memblock_add[_node]() before calling
2138 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2139 * usage, an architecture is expected to do something like
2141 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2143 * for_each_valid_physical_page_range()
2144 * memblock_add_node(base, size, nid)
2145 * free_area_init_nodes(max_zone_pfns);
2147 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2148 * registered physical page range. Similarly
2149 * sparse_memory_present_with_active_regions() calls memory_present() for
2150 * each range when SPARSEMEM is enabled.
2152 * See mm/page_alloc.c for more information on each function exposed by
2153 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2155 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2156 unsigned long node_map_pfn_alignment(void);
2157 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2158 unsigned long end_pfn);
2159 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2160 unsigned long end_pfn);
2161 extern void get_pfn_range_for_nid(unsigned int nid,
2162 unsigned long *start_pfn, unsigned long *end_pfn);
2163 extern unsigned long find_min_pfn_with_active_regions(void);
2164 extern void free_bootmem_with_active_regions(int nid,
2165 unsigned long max_low_pfn);
2166 extern void sparse_memory_present_with_active_regions(int nid);
2168 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2170 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2171 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2172 static inline int __early_pfn_to_nid(unsigned long pfn,
2173 struct mminit_pfnnid_cache *state)
2178 /* please see mm/page_alloc.c */
2179 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2180 /* there is a per-arch backend function. */
2181 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2182 struct mminit_pfnnid_cache *state);
2185 extern void set_dma_reserve(unsigned long new_dma_reserve);
2186 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2187 enum memmap_context, struct vmem_altmap *);
2188 extern void setup_per_zone_wmarks(void);
2189 extern int __meminit init_per_zone_wmark_min(void);
2190 extern void mem_init(void);
2191 extern void __init mmap_init(void);
2192 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2193 extern long si_mem_available(void);
2194 extern void si_meminfo(struct sysinfo * val);
2195 extern void si_meminfo_node(struct sysinfo *val, int nid);
2196 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2197 extern unsigned long arch_reserved_kernel_pages(void);
2200 extern __printf(3, 4)
2201 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2203 extern void setup_per_cpu_pageset(void);
2206 extern int min_free_kbytes;
2207 extern int watermark_boost_factor;
2208 extern int watermark_scale_factor;
2211 extern atomic_long_t mmap_pages_allocated;
2212 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2214 /* interval_tree.c */
2215 void vma_interval_tree_insert(struct vm_area_struct *node,
2216 struct rb_root_cached *root);
2217 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2218 struct vm_area_struct *prev,
2219 struct rb_root_cached *root);
2220 void vma_interval_tree_remove(struct vm_area_struct *node,
2221 struct rb_root_cached *root);
2222 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2223 unsigned long start, unsigned long last);
2224 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2225 unsigned long start, unsigned long last);
2227 #define vma_interval_tree_foreach(vma, root, start, last) \
2228 for (vma = vma_interval_tree_iter_first(root, start, last); \
2229 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2231 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2232 struct rb_root_cached *root);
2233 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2234 struct rb_root_cached *root);
2235 struct anon_vma_chain *
2236 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2237 unsigned long start, unsigned long last);
2238 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2239 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2240 #ifdef CONFIG_DEBUG_VM_RB
2241 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2244 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2245 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2246 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2249 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2250 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2251 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2252 struct vm_area_struct *expand);
2253 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2254 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2256 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2258 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2259 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2260 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2261 struct mempolicy *, struct vm_userfaultfd_ctx);
2262 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2263 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2264 unsigned long addr, int new_below);
2265 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2266 unsigned long addr, int new_below);
2267 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2268 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2269 struct rb_node **, struct rb_node *);
2270 extern void unlink_file_vma(struct vm_area_struct *);
2271 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2272 unsigned long addr, unsigned long len, pgoff_t pgoff,
2273 bool *need_rmap_locks);
2274 extern void exit_mmap(struct mm_struct *);
2276 static inline int check_data_rlimit(unsigned long rlim,
2278 unsigned long start,
2279 unsigned long end_data,
2280 unsigned long start_data)
2282 if (rlim < RLIM_INFINITY) {
2283 if (((new - start) + (end_data - start_data)) > rlim)
2290 extern int mm_take_all_locks(struct mm_struct *mm);
2291 extern void mm_drop_all_locks(struct mm_struct *mm);
2293 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2294 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2295 extern struct file *get_task_exe_file(struct task_struct *task);
2297 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2298 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2300 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2301 const struct vm_special_mapping *sm);
2302 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2303 unsigned long addr, unsigned long len,
2304 unsigned long flags,
2305 const struct vm_special_mapping *spec);
2306 /* This is an obsolete alternative to _install_special_mapping. */
2307 extern int install_special_mapping(struct mm_struct *mm,
2308 unsigned long addr, unsigned long len,
2309 unsigned long flags, struct page **pages);
2311 unsigned long randomize_stack_top(unsigned long stack_top);
2313 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2315 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2316 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2317 struct list_head *uf);
2318 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2319 unsigned long len, unsigned long prot, unsigned long flags,
2320 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2321 struct list_head *uf);
2322 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2323 struct list_head *uf, bool downgrade);
2324 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2325 struct list_head *uf);
2326 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2328 static inline unsigned long
2329 do_mmap_pgoff(struct file *file, unsigned long addr,
2330 unsigned long len, unsigned long prot, unsigned long flags,
2331 unsigned long pgoff, unsigned long *populate,
2332 struct list_head *uf)
2334 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2338 extern int __mm_populate(unsigned long addr, unsigned long len,
2340 static inline void mm_populate(unsigned long addr, unsigned long len)
2343 (void) __mm_populate(addr, len, 1);
2346 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2349 /* These take the mm semaphore themselves */
2350 extern int __must_check vm_brk(unsigned long, unsigned long);
2351 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2352 extern int vm_munmap(unsigned long, size_t);
2353 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2354 unsigned long, unsigned long,
2355 unsigned long, unsigned long);
2357 struct vm_unmapped_area_info {
2358 #define VM_UNMAPPED_AREA_TOPDOWN 1
2359 unsigned long flags;
2360 unsigned long length;
2361 unsigned long low_limit;
2362 unsigned long high_limit;
2363 unsigned long align_mask;
2364 unsigned long align_offset;
2367 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2368 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2371 * Search for an unmapped address range.
2373 * We are looking for a range that:
2374 * - does not intersect with any VMA;
2375 * - is contained within the [low_limit, high_limit) interval;
2376 * - is at least the desired size.
2377 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2379 static inline unsigned long
2380 vm_unmapped_area(struct vm_unmapped_area_info *info)
2382 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2383 return unmapped_area_topdown(info);
2385 return unmapped_area(info);
2389 extern void truncate_inode_pages(struct address_space *, loff_t);
2390 extern void truncate_inode_pages_range(struct address_space *,
2391 loff_t lstart, loff_t lend);
2392 extern void truncate_inode_pages_final(struct address_space *);
2394 /* generic vm_area_ops exported for stackable file systems */
2395 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2396 extern void filemap_map_pages(struct vm_fault *vmf,
2397 pgoff_t start_pgoff, pgoff_t end_pgoff);
2398 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2400 /* mm/page-writeback.c */
2401 int __must_check write_one_page(struct page *page);
2402 void task_dirty_inc(struct task_struct *tsk);
2405 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2407 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2408 pgoff_t offset, unsigned long nr_to_read);
2410 void page_cache_sync_readahead(struct address_space *mapping,
2411 struct file_ra_state *ra,
2414 unsigned long size);
2416 void page_cache_async_readahead(struct address_space *mapping,
2417 struct file_ra_state *ra,
2421 unsigned long size);
2423 extern unsigned long stack_guard_gap;
2424 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2425 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2427 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2428 extern int expand_downwards(struct vm_area_struct *vma,
2429 unsigned long address);
2431 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2433 #define expand_upwards(vma, address) (0)
2436 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2437 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2438 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2439 struct vm_area_struct **pprev);
2441 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2442 NULL if none. Assume start_addr < end_addr. */
2443 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2445 struct vm_area_struct * vma = find_vma(mm,start_addr);
2447 if (vma && end_addr <= vma->vm_start)
2452 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2454 unsigned long vm_start = vma->vm_start;
2456 if (vma->vm_flags & VM_GROWSDOWN) {
2457 vm_start -= stack_guard_gap;
2458 if (vm_start > vma->vm_start)
2464 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2466 unsigned long vm_end = vma->vm_end;
2468 if (vma->vm_flags & VM_GROWSUP) {
2469 vm_end += stack_guard_gap;
2470 if (vm_end < vma->vm_end)
2471 vm_end = -PAGE_SIZE;
2476 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2478 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2481 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2482 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2483 unsigned long vm_start, unsigned long vm_end)
2485 struct vm_area_struct *vma = find_vma(mm, vm_start);
2487 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2493 static inline bool range_in_vma(struct vm_area_struct *vma,
2494 unsigned long start, unsigned long end)
2496 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2500 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2501 void vma_set_page_prot(struct vm_area_struct *vma);
2503 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2507 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2509 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2513 #ifdef CONFIG_NUMA_BALANCING
2514 unsigned long change_prot_numa(struct vm_area_struct *vma,
2515 unsigned long start, unsigned long end);
2518 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2519 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2520 unsigned long pfn, unsigned long size, pgprot_t);
2521 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2522 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2524 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2526 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2528 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2529 unsigned long pfn, pgprot_t pgprot);
2530 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2532 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2533 pfn_t pfn, pgprot_t pgprot);
2534 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2535 unsigned long addr, pfn_t pfn);
2536 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2538 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2539 unsigned long addr, struct page *page)
2541 int err = vm_insert_page(vma, addr, page);
2544 return VM_FAULT_OOM;
2545 if (err < 0 && err != -EBUSY)
2546 return VM_FAULT_SIGBUS;
2548 return VM_FAULT_NOPAGE;
2551 static inline vm_fault_t vmf_error(int err)
2554 return VM_FAULT_OOM;
2555 return VM_FAULT_SIGBUS;
2558 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2559 unsigned int foll_flags);
2561 #define FOLL_WRITE 0x01 /* check pte is writable */
2562 #define FOLL_TOUCH 0x02 /* mark page accessed */
2563 #define FOLL_GET 0x04 /* do get_page on page */
2564 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2565 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2566 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2567 * and return without waiting upon it */
2568 #define FOLL_POPULATE 0x40 /* fault in page */
2569 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2570 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2571 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2572 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2573 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2574 #define FOLL_MLOCK 0x1000 /* lock present pages */
2575 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2576 #define FOLL_COW 0x4000 /* internal GUP flag */
2577 #define FOLL_ANON 0x8000 /* don't do file mappings */
2578 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2579 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2580 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2583 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2584 * other. Here is what they mean, and how to use them:
2586 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2587 * period _often_ under userspace control. This is in contrast to
2588 * iov_iter_get_pages(), whose usages are transient.
2590 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2591 * lifetime enforced by the filesystem and we need guarantees that longterm
2592 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2593 * the filesystem. Ideas for this coordination include revoking the longterm
2594 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2595 * added after the problem with filesystems was found FS DAX VMAs are
2596 * specifically failed. Filesystem pages are still subject to bugs and use of
2597 * FOLL_LONGTERM should be avoided on those pages.
2599 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2600 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2601 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2602 * is due to an incompatibility with the FS DAX check and
2603 * FAULT_FLAG_ALLOW_RETRY.
2605 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2606 * that region. And so, CMA attempts to migrate the page before pinning, when
2607 * FOLL_LONGTERM is specified.
2609 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2610 * but an additional pin counting system) will be invoked. This is intended for
2611 * anything that gets a page reference and then touches page data (for example,
2612 * Direct IO). This lets the filesystem know that some non-file-system entity is
2613 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2614 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2615 * a call to unpin_user_page().
2617 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2618 * and separate refcounting mechanisms, however, and that means that each has
2619 * its own acquire and release mechanisms:
2621 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2623 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2625 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2626 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2627 * calls applied to them, and that's perfectly OK. This is a constraint on the
2628 * callers, not on the pages.)
2630 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2631 * directly by the caller. That's in order to help avoid mismatches when
2632 * releasing pages: get_user_pages*() pages must be released via put_page(),
2633 * while pin_user_pages*() pages must be released via unpin_user_page().
2635 * Please see Documentation/vm/pin_user_pages.rst for more information.
2638 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2640 if (vm_fault & VM_FAULT_OOM)
2642 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2643 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2644 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2649 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2650 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2651 unsigned long size, pte_fn_t fn, void *data);
2652 extern int apply_to_existing_page_range(struct mm_struct *mm,
2653 unsigned long address, unsigned long size,
2654 pte_fn_t fn, void *data);
2656 #ifdef CONFIG_PAGE_POISONING
2657 extern bool page_poisoning_enabled(void);
2658 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2660 static inline bool page_poisoning_enabled(void) { return false; }
2661 static inline void kernel_poison_pages(struct page *page, int numpages,
2665 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2666 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2668 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2670 static inline bool want_init_on_alloc(gfp_t flags)
2672 if (static_branch_unlikely(&init_on_alloc) &&
2673 !page_poisoning_enabled())
2675 return flags & __GFP_ZERO;
2678 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2679 DECLARE_STATIC_KEY_TRUE(init_on_free);
2681 DECLARE_STATIC_KEY_FALSE(init_on_free);
2683 static inline bool want_init_on_free(void)
2685 return static_branch_unlikely(&init_on_free) &&
2686 !page_poisoning_enabled();
2689 #ifdef CONFIG_DEBUG_PAGEALLOC
2690 extern void init_debug_pagealloc(void);
2692 static inline void init_debug_pagealloc(void) {}
2694 extern bool _debug_pagealloc_enabled_early;
2695 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2697 static inline bool debug_pagealloc_enabled(void)
2699 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2700 _debug_pagealloc_enabled_early;
2704 * For use in fast paths after init_debug_pagealloc() has run, or when a
2705 * false negative result is not harmful when called too early.
2707 static inline bool debug_pagealloc_enabled_static(void)
2709 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2712 return static_branch_unlikely(&_debug_pagealloc_enabled);
2715 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2716 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2719 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2720 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2723 kernel_map_pages(struct page *page, int numpages, int enable)
2725 __kernel_map_pages(page, numpages, enable);
2727 #ifdef CONFIG_HIBERNATION
2728 extern bool kernel_page_present(struct page *page);
2729 #endif /* CONFIG_HIBERNATION */
2730 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2732 kernel_map_pages(struct page *page, int numpages, int enable) {}
2733 #ifdef CONFIG_HIBERNATION
2734 static inline bool kernel_page_present(struct page *page) { return true; }
2735 #endif /* CONFIG_HIBERNATION */
2736 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2738 #ifdef __HAVE_ARCH_GATE_AREA
2739 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2740 extern int in_gate_area_no_mm(unsigned long addr);
2741 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2743 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2747 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2748 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2752 #endif /* __HAVE_ARCH_GATE_AREA */
2754 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2756 #ifdef CONFIG_SYSCTL
2757 extern int sysctl_drop_caches;
2758 int drop_caches_sysctl_handler(struct ctl_table *, int,
2759 void __user *, size_t *, loff_t *);
2762 void drop_slab(void);
2763 void drop_slab_node(int nid);
2766 #define randomize_va_space 0
2768 extern int randomize_va_space;
2771 const char * arch_vma_name(struct vm_area_struct *vma);
2773 void print_vma_addr(char *prefix, unsigned long rip);
2775 static inline void print_vma_addr(char *prefix, unsigned long rip)
2780 void *sparse_buffer_alloc(unsigned long size);
2781 struct page * __populate_section_memmap(unsigned long pfn,
2782 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2783 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2784 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2785 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2786 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2787 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2788 void *vmemmap_alloc_block(unsigned long size, int node);
2790 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2791 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2792 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2793 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2795 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2796 struct vmem_altmap *altmap);
2797 void vmemmap_populate_print_last(void);
2798 #ifdef CONFIG_MEMORY_HOTPLUG
2799 void vmemmap_free(unsigned long start, unsigned long end,
2800 struct vmem_altmap *altmap);
2802 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2803 unsigned long nr_pages);
2806 MF_COUNT_INCREASED = 1 << 0,
2807 MF_ACTION_REQUIRED = 1 << 1,
2808 MF_MUST_KILL = 1 << 2,
2809 MF_SOFT_OFFLINE = 1 << 3,
2811 extern int memory_failure(unsigned long pfn, int flags);
2812 extern void memory_failure_queue(unsigned long pfn, int flags);
2813 extern int unpoison_memory(unsigned long pfn);
2814 extern int get_hwpoison_page(struct page *page);
2815 #define put_hwpoison_page(page) put_page(page)
2816 extern int sysctl_memory_failure_early_kill;
2817 extern int sysctl_memory_failure_recovery;
2818 extern void shake_page(struct page *p, int access);
2819 extern atomic_long_t num_poisoned_pages __read_mostly;
2820 extern int soft_offline_page(unsigned long pfn, int flags);
2824 * Error handlers for various types of pages.
2827 MF_IGNORED, /* Error: cannot be handled */
2828 MF_FAILED, /* Error: handling failed */
2829 MF_DELAYED, /* Will be handled later */
2830 MF_RECOVERED, /* Successfully recovered */
2833 enum mf_action_page_type {
2835 MF_MSG_KERNEL_HIGH_ORDER,
2837 MF_MSG_DIFFERENT_COMPOUND,
2838 MF_MSG_POISONED_HUGE,
2841 MF_MSG_NON_PMD_HUGE,
2842 MF_MSG_UNMAP_FAILED,
2843 MF_MSG_DIRTY_SWAPCACHE,
2844 MF_MSG_CLEAN_SWAPCACHE,
2845 MF_MSG_DIRTY_MLOCKED_LRU,
2846 MF_MSG_CLEAN_MLOCKED_LRU,
2847 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2848 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2851 MF_MSG_TRUNCATED_LRU,
2858 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2859 extern void clear_huge_page(struct page *page,
2860 unsigned long addr_hint,
2861 unsigned int pages_per_huge_page);
2862 extern void copy_user_huge_page(struct page *dst, struct page *src,
2863 unsigned long addr_hint,
2864 struct vm_area_struct *vma,
2865 unsigned int pages_per_huge_page);
2866 extern long copy_huge_page_from_user(struct page *dst_page,
2867 const void __user *usr_src,
2868 unsigned int pages_per_huge_page,
2869 bool allow_pagefault);
2870 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2872 #ifdef CONFIG_DEBUG_PAGEALLOC
2873 extern unsigned int _debug_guardpage_minorder;
2874 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2876 static inline unsigned int debug_guardpage_minorder(void)
2878 return _debug_guardpage_minorder;
2881 static inline bool debug_guardpage_enabled(void)
2883 return static_branch_unlikely(&_debug_guardpage_enabled);
2886 static inline bool page_is_guard(struct page *page)
2888 if (!debug_guardpage_enabled())
2891 return PageGuard(page);
2894 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2895 static inline bool debug_guardpage_enabled(void) { return false; }
2896 static inline bool page_is_guard(struct page *page) { return false; }
2897 #endif /* CONFIG_DEBUG_PAGEALLOC */
2899 #if MAX_NUMNODES > 1
2900 void __init setup_nr_node_ids(void);
2902 static inline void setup_nr_node_ids(void) {}
2905 extern int memcmp_pages(struct page *page1, struct page *page2);
2907 static inline int pages_identical(struct page *page1, struct page *page2)
2909 return !memcmp_pages(page1, page2);
2912 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
2913 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
2914 pgoff_t first_index, pgoff_t nr,
2915 pgoff_t bitmap_pgoff,
2916 unsigned long *bitmap,
2920 unsigned long wp_shared_mapping_range(struct address_space *mapping,
2921 pgoff_t first_index, pgoff_t nr);
2924 #endif /* __KERNEL__ */
2925 #endif /* _LINUX_MM_H */