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 static inline void totalram_pages_set(long val)
75 atomic_long_set(&_totalram_pages, val);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/pgtable.h>
100 #include <asm/processor.h>
103 * Architectures that support memory tagging (assigning tags to memory regions,
104 * embedding these tags into addresses that point to these memory regions, and
105 * checking that the memory and the pointer tags match on memory accesses)
106 * redefine this macro to strip tags from pointers.
107 * It's defined as noop for arcitectures that don't support memory tagging.
109 #ifndef untagged_addr
110 #define untagged_addr(addr) (addr)
114 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
118 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
122 #define lm_alias(x) __va(__pa_symbol(x))
126 * To prevent common memory management code establishing
127 * a zero page mapping on a read fault.
128 * This macro should be defined within <asm/pgtable.h>.
129 * s390 does this to prevent multiplexing of hardware bits
130 * related to the physical page in case of virtualization.
132 #ifndef mm_forbids_zeropage
133 #define mm_forbids_zeropage(X) (0)
137 * On some architectures it is expensive to call memset() for small sizes.
138 * If an architecture decides to implement their own version of
139 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
140 * define their own version of this macro in <asm/pgtable.h>
142 #if BITS_PER_LONG == 64
143 /* This function must be updated when the size of struct page grows above 80
144 * or reduces below 56. The idea that compiler optimizes out switch()
145 * statement, and only leaves move/store instructions. Also the compiler can
146 * combine write statments if they are both assignments and can be reordered,
147 * this can result in several of the writes here being dropped.
149 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
150 static inline void __mm_zero_struct_page(struct page *page)
152 unsigned long *_pp = (void *)page;
154 /* Check that struct page is either 56, 64, 72, or 80 bytes */
155 BUILD_BUG_ON(sizeof(struct page) & 7);
156 BUILD_BUG_ON(sizeof(struct page) < 56);
157 BUILD_BUG_ON(sizeof(struct page) > 80);
159 switch (sizeof(struct page)) {
161 _pp[9] = 0; /* fallthrough */
163 _pp[8] = 0; /* fallthrough */
165 _pp[7] = 0; /* fallthrough */
177 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
181 * Default maximum number of active map areas, this limits the number of vmas
182 * per mm struct. Users can overwrite this number by sysctl but there is a
185 * When a program's coredump is generated as ELF format, a section is created
186 * per a vma. In ELF, the number of sections is represented in unsigned short.
187 * This means the number of sections should be smaller than 65535 at coredump.
188 * Because the kernel adds some informative sections to a image of program at
189 * generating coredump, we need some margin. The number of extra sections is
190 * 1-3 now and depends on arch. We use "5" as safe margin, here.
192 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
193 * not a hard limit any more. Although some userspace tools can be surprised by
196 #define MAPCOUNT_ELF_CORE_MARGIN (5)
197 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
199 extern int sysctl_max_map_count;
201 extern unsigned long sysctl_user_reserve_kbytes;
202 extern unsigned long sysctl_admin_reserve_kbytes;
204 extern int sysctl_overcommit_memory;
205 extern int sysctl_overcommit_ratio;
206 extern unsigned long sysctl_overcommit_kbytes;
208 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
210 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
213 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
215 /* to align the pointer to the (next) page boundary */
216 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
218 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
219 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
221 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
224 * Linux kernel virtual memory manager primitives.
225 * The idea being to have a "virtual" mm in the same way
226 * we have a virtual fs - giving a cleaner interface to the
227 * mm details, and allowing different kinds of memory mappings
228 * (from shared memory to executable loading to arbitrary
232 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
233 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
234 void vm_area_free(struct vm_area_struct *);
237 extern struct rb_root nommu_region_tree;
238 extern struct rw_semaphore nommu_region_sem;
240 extern unsigned int kobjsize(const void *objp);
244 * vm_flags in vm_area_struct, see mm_types.h.
245 * When changing, update also include/trace/events/mmflags.h
247 #define VM_NONE 0x00000000
249 #define VM_READ 0x00000001 /* currently active flags */
250 #define VM_WRITE 0x00000002
251 #define VM_EXEC 0x00000004
252 #define VM_SHARED 0x00000008
254 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
255 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
256 #define VM_MAYWRITE 0x00000020
257 #define VM_MAYEXEC 0x00000040
258 #define VM_MAYSHARE 0x00000080
260 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
261 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
262 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
263 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
264 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
266 #define VM_LOCKED 0x00002000
267 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
269 /* Used by sys_madvise() */
270 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
271 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
273 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
274 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
275 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
276 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
277 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
278 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
279 #define VM_SYNC 0x00800000 /* Synchronous page faults */
280 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
281 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
282 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
284 #ifdef CONFIG_MEM_SOFT_DIRTY
285 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
287 # define VM_SOFTDIRTY 0
290 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
291 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
292 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
293 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
295 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
296 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
302 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
303 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
304 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
305 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
306 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
308 #ifdef CONFIG_ARCH_HAS_PKEYS
309 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
310 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
311 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
312 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
313 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
315 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
317 # define VM_PKEY_BIT4 0
319 #endif /* CONFIG_ARCH_HAS_PKEYS */
321 #if defined(CONFIG_X86)
322 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
323 #elif defined(CONFIG_PPC)
324 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
325 #elif defined(CONFIG_PARISC)
326 # define VM_GROWSUP VM_ARCH_1
327 #elif defined(CONFIG_IA64)
328 # define VM_GROWSUP VM_ARCH_1
329 #elif defined(CONFIG_SPARC64)
330 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
331 # define VM_ARCH_CLEAR VM_SPARC_ADI
332 #elif !defined(CONFIG_MMU)
333 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
336 #if defined(CONFIG_X86_INTEL_MPX)
337 /* MPX specific bounds table or bounds directory */
338 # define VM_MPX VM_HIGH_ARCH_4
340 # define VM_MPX VM_NONE
344 # define VM_GROWSUP VM_NONE
347 /* Bits set in the VMA until the stack is in its final location */
348 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
350 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
351 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
354 #ifdef CONFIG_STACK_GROWSUP
355 #define VM_STACK VM_GROWSUP
357 #define VM_STACK VM_GROWSDOWN
360 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
363 * Special vmas that are non-mergable, non-mlock()able.
364 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
366 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
368 /* This mask defines which mm->def_flags a process can inherit its parent */
369 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
371 /* This mask is used to clear all the VMA flags used by mlock */
372 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
374 /* Arch-specific flags to clear when updating VM flags on protection change */
375 #ifndef VM_ARCH_CLEAR
376 # define VM_ARCH_CLEAR VM_NONE
378 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
381 * mapping from the currently active vm_flags protection bits (the
382 * low four bits) to a page protection mask..
384 extern pgprot_t protection_map[16];
386 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
387 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
388 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
389 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
390 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
391 #define FAULT_FLAG_TRIED 0x20 /* Second try */
392 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
393 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
394 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
396 #define FAULT_FLAG_TRACE \
397 { FAULT_FLAG_WRITE, "WRITE" }, \
398 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
399 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
400 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
401 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
402 { FAULT_FLAG_TRIED, "TRIED" }, \
403 { FAULT_FLAG_USER, "USER" }, \
404 { FAULT_FLAG_REMOTE, "REMOTE" }, \
405 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
408 * vm_fault is filled by the the pagefault handler and passed to the vma's
409 * ->fault function. The vma's ->fault is responsible for returning a bitmask
410 * of VM_FAULT_xxx flags that give details about how the fault was handled.
412 * MM layer fills up gfp_mask for page allocations but fault handler might
413 * alter it if its implementation requires a different allocation context.
415 * pgoff should be used in favour of virtual_address, if possible.
418 struct vm_area_struct *vma; /* Target VMA */
419 unsigned int flags; /* FAULT_FLAG_xxx flags */
420 gfp_t gfp_mask; /* gfp mask to be used for allocations */
421 pgoff_t pgoff; /* Logical page offset based on vma */
422 unsigned long address; /* Faulting virtual address */
423 pmd_t *pmd; /* Pointer to pmd entry matching
425 pud_t *pud; /* Pointer to pud entry matching
428 pte_t orig_pte; /* Value of PTE at the time of fault */
430 struct page *cow_page; /* Page handler may use for COW fault */
431 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
432 struct page *page; /* ->fault handlers should return a
433 * page here, unless VM_FAULT_NOPAGE
434 * is set (which is also implied by
437 /* These three entries are valid only while holding ptl lock */
438 pte_t *pte; /* Pointer to pte entry matching
439 * the 'address'. NULL if the page
440 * table hasn't been allocated.
442 spinlock_t *ptl; /* Page table lock.
443 * Protects pte page table if 'pte'
444 * is not NULL, otherwise pmd.
446 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
447 * vm_ops->map_pages() calls
448 * alloc_set_pte() from atomic context.
449 * do_fault_around() pre-allocates
450 * page table to avoid allocation from
455 /* page entry size for vm->huge_fault() */
456 enum page_entry_size {
463 * These are the virtual MM functions - opening of an area, closing and
464 * unmapping it (needed to keep files on disk up-to-date etc), pointer
465 * to the functions called when a no-page or a wp-page exception occurs.
467 struct vm_operations_struct {
468 void (*open)(struct vm_area_struct * area);
469 void (*close)(struct vm_area_struct * area);
470 int (*split)(struct vm_area_struct * area, unsigned long addr);
471 int (*mremap)(struct vm_area_struct * area);
472 vm_fault_t (*fault)(struct vm_fault *vmf);
473 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
474 enum page_entry_size pe_size);
475 void (*map_pages)(struct vm_fault *vmf,
476 pgoff_t start_pgoff, pgoff_t end_pgoff);
477 unsigned long (*pagesize)(struct vm_area_struct * area);
479 /* notification that a previously read-only page is about to become
480 * writable, if an error is returned it will cause a SIGBUS */
481 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
483 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
484 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
486 /* called by access_process_vm when get_user_pages() fails, typically
487 * for use by special VMAs that can switch between memory and hardware
489 int (*access)(struct vm_area_struct *vma, unsigned long addr,
490 void *buf, int len, int write);
492 /* Called by the /proc/PID/maps code to ask the vma whether it
493 * has a special name. Returning non-NULL will also cause this
494 * vma to be dumped unconditionally. */
495 const char *(*name)(struct vm_area_struct *vma);
499 * set_policy() op must add a reference to any non-NULL @new mempolicy
500 * to hold the policy upon return. Caller should pass NULL @new to
501 * remove a policy and fall back to surrounding context--i.e. do not
502 * install a MPOL_DEFAULT policy, nor the task or system default
505 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
508 * get_policy() op must add reference [mpol_get()] to any policy at
509 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
510 * in mm/mempolicy.c will do this automatically.
511 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
512 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
513 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
514 * must return NULL--i.e., do not "fallback" to task or system default
517 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
521 * Called by vm_normal_page() for special PTEs to find the
522 * page for @addr. This is useful if the default behavior
523 * (using pte_page()) would not find the correct page.
525 struct page *(*find_special_page)(struct vm_area_struct *vma,
529 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
531 static const struct vm_operations_struct dummy_vm_ops = {};
533 memset(vma, 0, sizeof(*vma));
535 vma->vm_ops = &dummy_vm_ops;
536 INIT_LIST_HEAD(&vma->anon_vma_chain);
539 static inline void vma_set_anonymous(struct vm_area_struct *vma)
544 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
551 * The vma_is_shmem is not inline because it is used only by slow
552 * paths in userfault.
554 bool vma_is_shmem(struct vm_area_struct *vma);
556 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
559 int vma_is_stack_for_current(struct vm_area_struct *vma);
561 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
562 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
568 * FIXME: take this include out, include page-flags.h in
569 * files which need it (119 of them)
571 #include <linux/page-flags.h>
572 #include <linux/huge_mm.h>
575 * Methods to modify the page usage count.
577 * What counts for a page usage:
578 * - cache mapping (page->mapping)
579 * - private data (page->private)
580 * - page mapped in a task's page tables, each mapping
581 * is counted separately
583 * Also, many kernel routines increase the page count before a critical
584 * routine so they can be sure the page doesn't go away from under them.
588 * Drop a ref, return true if the refcount fell to zero (the page has no users)
590 static inline int put_page_testzero(struct page *page)
592 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
593 return page_ref_dec_and_test(page);
597 * Try to grab a ref unless the page has a refcount of zero, return false if
599 * This can be called when MMU is off so it must not access
600 * any of the virtual mappings.
602 static inline int get_page_unless_zero(struct page *page)
604 return page_ref_add_unless(page, 1, 0);
607 extern int page_is_ram(unsigned long pfn);
615 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
618 /* Support for virtually mapped pages */
619 struct page *vmalloc_to_page(const void *addr);
620 unsigned long vmalloc_to_pfn(const void *addr);
623 * Determine if an address is within the vmalloc range
625 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
626 * is no special casing required.
629 #ifndef is_ioremap_addr
630 #define is_ioremap_addr(x) is_vmalloc_addr(x)
634 extern bool is_vmalloc_addr(const void *x);
635 extern int is_vmalloc_or_module_addr(const void *x);
637 static inline bool is_vmalloc_addr(const void *x)
641 static inline int is_vmalloc_or_module_addr(const void *x)
647 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
648 static inline void *kvmalloc(size_t size, gfp_t flags)
650 return kvmalloc_node(size, flags, NUMA_NO_NODE);
652 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
654 return kvmalloc_node(size, flags | __GFP_ZERO, node);
656 static inline void *kvzalloc(size_t size, gfp_t flags)
658 return kvmalloc(size, flags | __GFP_ZERO);
661 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
665 if (unlikely(check_mul_overflow(n, size, &bytes)))
668 return kvmalloc(bytes, flags);
671 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
673 return kvmalloc_array(n, size, flags | __GFP_ZERO);
676 extern void kvfree(const void *addr);
678 static inline int compound_mapcount(struct page *page)
680 VM_BUG_ON_PAGE(!PageCompound(page), page);
681 page = compound_head(page);
682 return atomic_read(compound_mapcount_ptr(page)) + 1;
686 * The atomic page->_mapcount, starts from -1: so that transitions
687 * both from it and to it can be tracked, using atomic_inc_and_test
688 * and atomic_add_negative(-1).
690 static inline void page_mapcount_reset(struct page *page)
692 atomic_set(&(page)->_mapcount, -1);
695 int __page_mapcount(struct page *page);
697 static inline int page_mapcount(struct page *page)
699 VM_BUG_ON_PAGE(PageSlab(page), page);
701 if (unlikely(PageCompound(page)))
702 return __page_mapcount(page);
703 return atomic_read(&page->_mapcount) + 1;
706 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
707 int total_mapcount(struct page *page);
708 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
710 static inline int total_mapcount(struct page *page)
712 return page_mapcount(page);
714 static inline int page_trans_huge_mapcount(struct page *page,
717 int mapcount = page_mapcount(page);
719 *total_mapcount = mapcount;
724 static inline struct page *virt_to_head_page(const void *x)
726 struct page *page = virt_to_page(x);
728 return compound_head(page);
731 void __put_page(struct page *page);
733 void put_pages_list(struct list_head *pages);
735 void split_page(struct page *page, unsigned int order);
738 * Compound pages have a destructor function. Provide a
739 * prototype for that function and accessor functions.
740 * These are _only_ valid on the head of a compound page.
742 typedef void compound_page_dtor(struct page *);
744 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
745 enum compound_dtor_id {
748 #ifdef CONFIG_HUGETLB_PAGE
751 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
756 extern compound_page_dtor * const compound_page_dtors[];
758 static inline void set_compound_page_dtor(struct page *page,
759 enum compound_dtor_id compound_dtor)
761 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
762 page[1].compound_dtor = compound_dtor;
765 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
767 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
768 return compound_page_dtors[page[1].compound_dtor];
771 static inline unsigned int compound_order(struct page *page)
775 return page[1].compound_order;
778 static inline void set_compound_order(struct page *page, unsigned int order)
780 page[1].compound_order = order;
783 /* Returns the number of pages in this potentially compound page. */
784 static inline unsigned long compound_nr(struct page *page)
786 return 1UL << compound_order(page);
789 /* Returns the number of bytes in this potentially compound page. */
790 static inline unsigned long page_size(struct page *page)
792 return PAGE_SIZE << compound_order(page);
795 /* Returns the number of bits needed for the number of bytes in a page */
796 static inline unsigned int page_shift(struct page *page)
798 return PAGE_SHIFT + compound_order(page);
801 void free_compound_page(struct page *page);
805 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
806 * servicing faults for write access. In the normal case, do always want
807 * pte_mkwrite. But get_user_pages can cause write faults for mappings
808 * that do not have writing enabled, when used by access_process_vm.
810 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
812 if (likely(vma->vm_flags & VM_WRITE))
813 pte = pte_mkwrite(pte);
817 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
819 vm_fault_t finish_fault(struct vm_fault *vmf);
820 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
824 * Multiple processes may "see" the same page. E.g. for untouched
825 * mappings of /dev/null, all processes see the same page full of
826 * zeroes, and text pages of executables and shared libraries have
827 * only one copy in memory, at most, normally.
829 * For the non-reserved pages, page_count(page) denotes a reference count.
830 * page_count() == 0 means the page is free. page->lru is then used for
831 * freelist management in the buddy allocator.
832 * page_count() > 0 means the page has been allocated.
834 * Pages are allocated by the slab allocator in order to provide memory
835 * to kmalloc and kmem_cache_alloc. In this case, the management of the
836 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
837 * unless a particular usage is carefully commented. (the responsibility of
838 * freeing the kmalloc memory is the caller's, of course).
840 * A page may be used by anyone else who does a __get_free_page().
841 * In this case, page_count still tracks the references, and should only
842 * be used through the normal accessor functions. The top bits of page->flags
843 * and page->virtual store page management information, but all other fields
844 * are unused and could be used privately, carefully. The management of this
845 * page is the responsibility of the one who allocated it, and those who have
846 * subsequently been given references to it.
848 * The other pages (we may call them "pagecache pages") are completely
849 * managed by the Linux memory manager: I/O, buffers, swapping etc.
850 * The following discussion applies only to them.
852 * A pagecache page contains an opaque `private' member, which belongs to the
853 * page's address_space. Usually, this is the address of a circular list of
854 * the page's disk buffers. PG_private must be set to tell the VM to call
855 * into the filesystem to release these pages.
857 * A page may belong to an inode's memory mapping. In this case, page->mapping
858 * is the pointer to the inode, and page->index is the file offset of the page,
859 * in units of PAGE_SIZE.
861 * If pagecache pages are not associated with an inode, they are said to be
862 * anonymous pages. These may become associated with the swapcache, and in that
863 * case PG_swapcache is set, and page->private is an offset into the swapcache.
865 * In either case (swapcache or inode backed), the pagecache itself holds one
866 * reference to the page. Setting PG_private should also increment the
867 * refcount. The each user mapping also has a reference to the page.
869 * The pagecache pages are stored in a per-mapping radix tree, which is
870 * rooted at mapping->i_pages, and indexed by offset.
871 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
872 * lists, we instead now tag pages as dirty/writeback in the radix tree.
874 * All pagecache pages may be subject to I/O:
875 * - inode pages may need to be read from disk,
876 * - inode pages which have been modified and are MAP_SHARED may need
877 * to be written back to the inode on disk,
878 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
879 * modified may need to be swapped out to swap space and (later) to be read
884 * The zone field is never updated after free_area_init_core()
885 * sets it, so none of the operations on it need to be atomic.
888 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
889 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
890 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
891 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
892 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
893 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
896 * Define the bit shifts to access each section. For non-existent
897 * sections we define the shift as 0; that plus a 0 mask ensures
898 * the compiler will optimise away reference to them.
900 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
901 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
902 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
903 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
904 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
906 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
907 #ifdef NODE_NOT_IN_PAGE_FLAGS
908 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
909 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
910 SECTIONS_PGOFF : ZONES_PGOFF)
912 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
913 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
914 NODES_PGOFF : ZONES_PGOFF)
917 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
919 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
920 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
923 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
924 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
925 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
926 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
927 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
928 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
930 static inline enum zone_type page_zonenum(const struct page *page)
932 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
935 #ifdef CONFIG_ZONE_DEVICE
936 static inline bool is_zone_device_page(const struct page *page)
938 return page_zonenum(page) == ZONE_DEVICE;
940 extern void memmap_init_zone_device(struct zone *, unsigned long,
941 unsigned long, struct dev_pagemap *);
943 static inline bool is_zone_device_page(const struct page *page)
949 #ifdef CONFIG_DEV_PAGEMAP_OPS
950 void __put_devmap_managed_page(struct page *page);
951 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
952 static inline bool put_devmap_managed_page(struct page *page)
954 if (!static_branch_unlikely(&devmap_managed_key))
956 if (!is_zone_device_page(page))
958 switch (page->pgmap->type) {
959 case MEMORY_DEVICE_PRIVATE:
960 case MEMORY_DEVICE_FS_DAX:
961 __put_devmap_managed_page(page);
969 #else /* CONFIG_DEV_PAGEMAP_OPS */
970 static inline bool put_devmap_managed_page(struct page *page)
974 #endif /* CONFIG_DEV_PAGEMAP_OPS */
976 static inline bool is_device_private_page(const struct page *page)
978 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
979 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
980 is_zone_device_page(page) &&
981 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
984 static inline bool is_pci_p2pdma_page(const struct page *page)
986 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
987 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
988 is_zone_device_page(page) &&
989 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
992 /* 127: arbitrary random number, small enough to assemble well */
993 #define page_ref_zero_or_close_to_overflow(page) \
994 ((unsigned int) page_ref_count(page) + 127u <= 127u)
996 static inline void get_page(struct page *page)
998 page = compound_head(page);
1000 * Getting a normal page or the head of a compound page
1001 * requires to already have an elevated page->_refcount.
1003 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1007 static inline __must_check bool try_get_page(struct page *page)
1009 page = compound_head(page);
1010 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1016 static inline void put_page(struct page *page)
1018 page = compound_head(page);
1021 * For devmap managed pages we need to catch refcount transition from
1022 * 2 to 1, when refcount reach one it means the page is free and we
1023 * need to inform the device driver through callback. See
1024 * include/linux/memremap.h and HMM for details.
1026 if (put_devmap_managed_page(page))
1029 if (put_page_testzero(page))
1034 * put_user_page() - release a gup-pinned page
1035 * @page: pointer to page to be released
1037 * Pages that were pinned via get_user_pages*() must be released via
1038 * either put_user_page(), or one of the put_user_pages*() routines
1039 * below. This is so that eventually, pages that are pinned via
1040 * get_user_pages*() can be separately tracked and uniquely handled. In
1041 * particular, interactions with RDMA and filesystems need special
1044 * put_user_page() and put_page() are not interchangeable, despite this early
1045 * implementation that makes them look the same. put_user_page() calls must
1046 * be perfectly matched up with get_user_page() calls.
1048 static inline void put_user_page(struct page *page)
1053 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1056 void put_user_pages(struct page **pages, unsigned long npages);
1058 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1059 #define SECTION_IN_PAGE_FLAGS
1063 * The identification function is mainly used by the buddy allocator for
1064 * determining if two pages could be buddies. We are not really identifying
1065 * the zone since we could be using the section number id if we do not have
1066 * node id available in page flags.
1067 * We only guarantee that it will return the same value for two combinable
1070 static inline int page_zone_id(struct page *page)
1072 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1075 #ifdef NODE_NOT_IN_PAGE_FLAGS
1076 extern int page_to_nid(const struct page *page);
1078 static inline int page_to_nid(const struct page *page)
1080 struct page *p = (struct page *)page;
1082 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1086 #ifdef CONFIG_NUMA_BALANCING
1087 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1089 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1092 static inline int cpupid_to_pid(int cpupid)
1094 return cpupid & LAST__PID_MASK;
1097 static inline int cpupid_to_cpu(int cpupid)
1099 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1102 static inline int cpupid_to_nid(int cpupid)
1104 return cpu_to_node(cpupid_to_cpu(cpupid));
1107 static inline bool cpupid_pid_unset(int cpupid)
1109 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1112 static inline bool cpupid_cpu_unset(int cpupid)
1114 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1117 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1119 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1122 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1123 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1124 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1126 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1129 static inline int page_cpupid_last(struct page *page)
1131 return page->_last_cpupid;
1133 static inline void page_cpupid_reset_last(struct page *page)
1135 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1138 static inline int page_cpupid_last(struct page *page)
1140 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1143 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1145 static inline void page_cpupid_reset_last(struct page *page)
1147 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1149 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1150 #else /* !CONFIG_NUMA_BALANCING */
1151 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1153 return page_to_nid(page); /* XXX */
1156 static inline int page_cpupid_last(struct page *page)
1158 return page_to_nid(page); /* XXX */
1161 static inline int cpupid_to_nid(int cpupid)
1166 static inline int cpupid_to_pid(int cpupid)
1171 static inline int cpupid_to_cpu(int cpupid)
1176 static inline int cpu_pid_to_cpupid(int nid, int pid)
1181 static inline bool cpupid_pid_unset(int cpupid)
1186 static inline void page_cpupid_reset_last(struct page *page)
1190 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1194 #endif /* CONFIG_NUMA_BALANCING */
1196 #ifdef CONFIG_KASAN_SW_TAGS
1197 static inline u8 page_kasan_tag(const struct page *page)
1199 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1202 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1204 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1205 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1208 static inline void page_kasan_tag_reset(struct page *page)
1210 page_kasan_tag_set(page, 0xff);
1213 static inline u8 page_kasan_tag(const struct page *page)
1218 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1219 static inline void page_kasan_tag_reset(struct page *page) { }
1222 static inline struct zone *page_zone(const struct page *page)
1224 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1227 static inline pg_data_t *page_pgdat(const struct page *page)
1229 return NODE_DATA(page_to_nid(page));
1232 #ifdef SECTION_IN_PAGE_FLAGS
1233 static inline void set_page_section(struct page *page, unsigned long section)
1235 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1236 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1239 static inline unsigned long page_to_section(const struct page *page)
1241 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1245 static inline void set_page_zone(struct page *page, enum zone_type zone)
1247 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1248 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1251 static inline void set_page_node(struct page *page, unsigned long node)
1253 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1254 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1257 static inline void set_page_links(struct page *page, enum zone_type zone,
1258 unsigned long node, unsigned long pfn)
1260 set_page_zone(page, zone);
1261 set_page_node(page, node);
1262 #ifdef SECTION_IN_PAGE_FLAGS
1263 set_page_section(page, pfn_to_section_nr(pfn));
1268 static inline struct mem_cgroup *page_memcg(struct page *page)
1270 return page->mem_cgroup;
1272 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1274 WARN_ON_ONCE(!rcu_read_lock_held());
1275 return READ_ONCE(page->mem_cgroup);
1278 static inline struct mem_cgroup *page_memcg(struct page *page)
1282 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1284 WARN_ON_ONCE(!rcu_read_lock_held());
1290 * Some inline functions in vmstat.h depend on page_zone()
1292 #include <linux/vmstat.h>
1294 static __always_inline void *lowmem_page_address(const struct page *page)
1296 return page_to_virt(page);
1299 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1300 #define HASHED_PAGE_VIRTUAL
1303 #if defined(WANT_PAGE_VIRTUAL)
1304 static inline void *page_address(const struct page *page)
1306 return page->virtual;
1308 static inline void set_page_address(struct page *page, void *address)
1310 page->virtual = address;
1312 #define page_address_init() do { } while(0)
1315 #if defined(HASHED_PAGE_VIRTUAL)
1316 void *page_address(const struct page *page);
1317 void set_page_address(struct page *page, void *virtual);
1318 void page_address_init(void);
1321 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1322 #define page_address(page) lowmem_page_address(page)
1323 #define set_page_address(page, address) do { } while(0)
1324 #define page_address_init() do { } while(0)
1327 extern void *page_rmapping(struct page *page);
1328 extern struct anon_vma *page_anon_vma(struct page *page);
1329 extern struct address_space *page_mapping(struct page *page);
1331 extern struct address_space *__page_file_mapping(struct page *);
1334 struct address_space *page_file_mapping(struct page *page)
1336 if (unlikely(PageSwapCache(page)))
1337 return __page_file_mapping(page);
1339 return page->mapping;
1342 extern pgoff_t __page_file_index(struct page *page);
1345 * Return the pagecache index of the passed page. Regular pagecache pages
1346 * use ->index whereas swapcache pages use swp_offset(->private)
1348 static inline pgoff_t page_index(struct page *page)
1350 if (unlikely(PageSwapCache(page)))
1351 return __page_file_index(page);
1355 bool page_mapped(struct page *page);
1356 struct address_space *page_mapping(struct page *page);
1357 struct address_space *page_mapping_file(struct page *page);
1360 * Return true only if the page has been allocated with
1361 * ALLOC_NO_WATERMARKS and the low watermark was not
1362 * met implying that the system is under some pressure.
1364 static inline bool page_is_pfmemalloc(struct page *page)
1367 * Page index cannot be this large so this must be
1368 * a pfmemalloc page.
1370 return page->index == -1UL;
1374 * Only to be called by the page allocator on a freshly allocated
1377 static inline void set_page_pfmemalloc(struct page *page)
1382 static inline void clear_page_pfmemalloc(struct page *page)
1388 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1390 extern void pagefault_out_of_memory(void);
1392 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1395 * Flags passed to show_mem() and show_free_areas() to suppress output in
1398 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1400 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1403 extern bool can_do_mlock(void);
1405 static inline bool can_do_mlock(void) { return false; }
1407 extern int user_shm_lock(size_t, struct user_struct *);
1408 extern void user_shm_unlock(size_t, struct user_struct *);
1411 * Parameter block passed down to zap_pte_range in exceptional cases.
1413 struct zap_details {
1414 struct address_space *check_mapping; /* Check page->mapping if set */
1415 pgoff_t first_index; /* Lowest page->index to unmap */
1416 pgoff_t last_index; /* Highest page->index to unmap */
1419 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1421 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1424 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1425 unsigned long size);
1426 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1427 unsigned long size);
1428 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1429 unsigned long start, unsigned long end);
1431 struct mmu_notifier_range;
1433 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1434 unsigned long end, unsigned long floor, unsigned long ceiling);
1435 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1436 struct vm_area_struct *vma);
1437 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1438 struct mmu_notifier_range *range,
1439 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1440 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1441 unsigned long *pfn);
1442 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1443 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1444 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1445 void *buf, int len, int write);
1447 extern void truncate_pagecache(struct inode *inode, loff_t new);
1448 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1449 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1450 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1451 int truncate_inode_page(struct address_space *mapping, struct page *page);
1452 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1453 int invalidate_inode_page(struct page *page);
1456 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1457 unsigned long address, unsigned int flags);
1458 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1459 unsigned long address, unsigned int fault_flags,
1461 void unmap_mapping_pages(struct address_space *mapping,
1462 pgoff_t start, pgoff_t nr, bool even_cows);
1463 void unmap_mapping_range(struct address_space *mapping,
1464 loff_t const holebegin, loff_t const holelen, int even_cows);
1466 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1467 unsigned long address, unsigned int flags)
1469 /* should never happen if there's no MMU */
1471 return VM_FAULT_SIGBUS;
1473 static inline int fixup_user_fault(struct task_struct *tsk,
1474 struct mm_struct *mm, unsigned long address,
1475 unsigned int fault_flags, bool *unlocked)
1477 /* should never happen if there's no MMU */
1481 static inline void unmap_mapping_pages(struct address_space *mapping,
1482 pgoff_t start, pgoff_t nr, bool even_cows) { }
1483 static inline void unmap_mapping_range(struct address_space *mapping,
1484 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1487 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1488 loff_t const holebegin, loff_t const holelen)
1490 unmap_mapping_range(mapping, holebegin, holelen, 0);
1493 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1494 void *buf, int len, unsigned int gup_flags);
1495 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1496 void *buf, int len, unsigned int gup_flags);
1497 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1498 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1500 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1501 unsigned long start, unsigned long nr_pages,
1502 unsigned int gup_flags, struct page **pages,
1503 struct vm_area_struct **vmas, int *locked);
1504 long get_user_pages(unsigned long start, unsigned long nr_pages,
1505 unsigned int gup_flags, struct page **pages,
1506 struct vm_area_struct **vmas);
1507 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1508 unsigned int gup_flags, struct page **pages, int *locked);
1509 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1510 struct page **pages, unsigned int gup_flags);
1512 int get_user_pages_fast(unsigned long start, int nr_pages,
1513 unsigned int gup_flags, struct page **pages);
1515 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1516 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1517 struct task_struct *task, bool bypass_rlim);
1519 /* Container for pinned pfns / pages */
1520 struct frame_vector {
1521 unsigned int nr_allocated; /* Number of frames we have space for */
1522 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1523 bool got_ref; /* Did we pin pages by getting page ref? */
1524 bool is_pfns; /* Does array contain pages or pfns? */
1525 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1526 * pfns_vector_pages() or pfns_vector_pfns()
1530 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1531 void frame_vector_destroy(struct frame_vector *vec);
1532 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1533 unsigned int gup_flags, struct frame_vector *vec);
1534 void put_vaddr_frames(struct frame_vector *vec);
1535 int frame_vector_to_pages(struct frame_vector *vec);
1536 void frame_vector_to_pfns(struct frame_vector *vec);
1538 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1540 return vec->nr_frames;
1543 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1546 int err = frame_vector_to_pages(vec);
1549 return ERR_PTR(err);
1551 return (struct page **)(vec->ptrs);
1554 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1557 frame_vector_to_pfns(vec);
1558 return (unsigned long *)(vec->ptrs);
1562 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1563 struct page **pages);
1564 int get_kernel_page(unsigned long start, int write, struct page **pages);
1565 struct page *get_dump_page(unsigned long addr);
1567 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1568 extern void do_invalidatepage(struct page *page, unsigned int offset,
1569 unsigned int length);
1571 void __set_page_dirty(struct page *, struct address_space *, int warn);
1572 int __set_page_dirty_nobuffers(struct page *page);
1573 int __set_page_dirty_no_writeback(struct page *page);
1574 int redirty_page_for_writepage(struct writeback_control *wbc,
1576 void account_page_dirtied(struct page *page, struct address_space *mapping);
1577 void account_page_cleaned(struct page *page, struct address_space *mapping,
1578 struct bdi_writeback *wb);
1579 int set_page_dirty(struct page *page);
1580 int set_page_dirty_lock(struct page *page);
1581 void __cancel_dirty_page(struct page *page);
1582 static inline void cancel_dirty_page(struct page *page)
1584 /* Avoid atomic ops, locking, etc. when not actually needed. */
1585 if (PageDirty(page))
1586 __cancel_dirty_page(page);
1588 int clear_page_dirty_for_io(struct page *page);
1590 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1592 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1593 unsigned long old_addr, struct vm_area_struct *new_vma,
1594 unsigned long new_addr, unsigned long len,
1595 bool need_rmap_locks);
1596 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1597 unsigned long end, pgprot_t newprot,
1598 int dirty_accountable, int prot_numa);
1599 extern int mprotect_fixup(struct vm_area_struct *vma,
1600 struct vm_area_struct **pprev, unsigned long start,
1601 unsigned long end, unsigned long newflags);
1604 * doesn't attempt to fault and will return short.
1606 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1607 struct page **pages);
1609 * per-process(per-mm_struct) statistics.
1611 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1613 long val = atomic_long_read(&mm->rss_stat.count[member]);
1615 #ifdef SPLIT_RSS_COUNTING
1617 * counter is updated in asynchronous manner and may go to minus.
1618 * But it's never be expected number for users.
1623 return (unsigned long)val;
1626 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1628 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1630 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1632 mm_trace_rss_stat(mm, member, count);
1635 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1637 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1639 mm_trace_rss_stat(mm, member, count);
1642 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1644 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1646 mm_trace_rss_stat(mm, member, count);
1649 /* Optimized variant when page is already known not to be PageAnon */
1650 static inline int mm_counter_file(struct page *page)
1652 if (PageSwapBacked(page))
1653 return MM_SHMEMPAGES;
1654 return MM_FILEPAGES;
1657 static inline int mm_counter(struct page *page)
1660 return MM_ANONPAGES;
1661 return mm_counter_file(page);
1664 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1666 return get_mm_counter(mm, MM_FILEPAGES) +
1667 get_mm_counter(mm, MM_ANONPAGES) +
1668 get_mm_counter(mm, MM_SHMEMPAGES);
1671 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1673 return max(mm->hiwater_rss, get_mm_rss(mm));
1676 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1678 return max(mm->hiwater_vm, mm->total_vm);
1681 static inline void update_hiwater_rss(struct mm_struct *mm)
1683 unsigned long _rss = get_mm_rss(mm);
1685 if ((mm)->hiwater_rss < _rss)
1686 (mm)->hiwater_rss = _rss;
1689 static inline void update_hiwater_vm(struct mm_struct *mm)
1691 if (mm->hiwater_vm < mm->total_vm)
1692 mm->hiwater_vm = mm->total_vm;
1695 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1697 mm->hiwater_rss = get_mm_rss(mm);
1700 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1701 struct mm_struct *mm)
1703 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1705 if (*maxrss < hiwater_rss)
1706 *maxrss = hiwater_rss;
1709 #if defined(SPLIT_RSS_COUNTING)
1710 void sync_mm_rss(struct mm_struct *mm);
1712 static inline void sync_mm_rss(struct mm_struct *mm)
1717 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1718 static inline int pte_devmap(pte_t pte)
1724 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1726 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1728 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1732 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1736 #ifdef __PAGETABLE_P4D_FOLDED
1737 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1738 unsigned long address)
1743 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1746 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1747 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1748 unsigned long address)
1752 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1753 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1756 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1758 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1760 if (mm_pud_folded(mm))
1762 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1765 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1767 if (mm_pud_folded(mm))
1769 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1773 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1774 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1775 unsigned long address)
1780 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1781 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1784 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1786 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1788 if (mm_pmd_folded(mm))
1790 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1793 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1795 if (mm_pmd_folded(mm))
1797 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1802 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1804 atomic_long_set(&mm->pgtables_bytes, 0);
1807 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1809 return atomic_long_read(&mm->pgtables_bytes);
1812 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1814 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1817 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1819 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1823 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1824 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1829 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1830 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1833 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1834 int __pte_alloc_kernel(pmd_t *pmd);
1836 #if defined(CONFIG_MMU)
1839 * The following ifdef needed to get the 5level-fixup.h header to work.
1840 * Remove it when 5level-fixup.h has been removed.
1842 #ifndef __ARCH_HAS_5LEVEL_HACK
1843 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1844 unsigned long address)
1846 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1847 NULL : p4d_offset(pgd, address);
1850 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1851 unsigned long address)
1853 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1854 NULL : pud_offset(p4d, address);
1856 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1858 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1860 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1861 NULL: pmd_offset(pud, address);
1863 #endif /* CONFIG_MMU */
1865 #if USE_SPLIT_PTE_PTLOCKS
1866 #if ALLOC_SPLIT_PTLOCKS
1867 void __init ptlock_cache_init(void);
1868 extern bool ptlock_alloc(struct page *page);
1869 extern void ptlock_free(struct page *page);
1871 static inline spinlock_t *ptlock_ptr(struct page *page)
1875 #else /* ALLOC_SPLIT_PTLOCKS */
1876 static inline void ptlock_cache_init(void)
1880 static inline bool ptlock_alloc(struct page *page)
1885 static inline void ptlock_free(struct page *page)
1889 static inline spinlock_t *ptlock_ptr(struct page *page)
1893 #endif /* ALLOC_SPLIT_PTLOCKS */
1895 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1897 return ptlock_ptr(pmd_page(*pmd));
1900 static inline bool ptlock_init(struct page *page)
1903 * prep_new_page() initialize page->private (and therefore page->ptl)
1904 * with 0. Make sure nobody took it in use in between.
1906 * It can happen if arch try to use slab for page table allocation:
1907 * slab code uses page->slab_cache, which share storage with page->ptl.
1909 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1910 if (!ptlock_alloc(page))
1912 spin_lock_init(ptlock_ptr(page));
1916 #else /* !USE_SPLIT_PTE_PTLOCKS */
1918 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1920 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1922 return &mm->page_table_lock;
1924 static inline void ptlock_cache_init(void) {}
1925 static inline bool ptlock_init(struct page *page) { return true; }
1926 static inline void ptlock_free(struct page *page) {}
1927 #endif /* USE_SPLIT_PTE_PTLOCKS */
1929 static inline void pgtable_init(void)
1931 ptlock_cache_init();
1932 pgtable_cache_init();
1935 static inline bool pgtable_pte_page_ctor(struct page *page)
1937 if (!ptlock_init(page))
1939 __SetPageTable(page);
1940 inc_zone_page_state(page, NR_PAGETABLE);
1944 static inline void pgtable_pte_page_dtor(struct page *page)
1947 __ClearPageTable(page);
1948 dec_zone_page_state(page, NR_PAGETABLE);
1951 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1953 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1954 pte_t *__pte = pte_offset_map(pmd, address); \
1960 #define pte_unmap_unlock(pte, ptl) do { \
1965 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1967 #define pte_alloc_map(mm, pmd, address) \
1968 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1970 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1971 (pte_alloc(mm, pmd) ? \
1972 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1974 #define pte_alloc_kernel(pmd, address) \
1975 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1976 NULL: pte_offset_kernel(pmd, address))
1978 #if USE_SPLIT_PMD_PTLOCKS
1980 static struct page *pmd_to_page(pmd_t *pmd)
1982 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1983 return virt_to_page((void *)((unsigned long) pmd & mask));
1986 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1988 return ptlock_ptr(pmd_to_page(pmd));
1991 static inline bool pgtable_pmd_page_ctor(struct page *page)
1993 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1994 page->pmd_huge_pte = NULL;
1996 return ptlock_init(page);
1999 static inline void pgtable_pmd_page_dtor(struct page *page)
2001 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2002 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2007 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2011 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2013 return &mm->page_table_lock;
2016 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2017 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2019 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2023 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2025 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2031 * No scalability reason to split PUD locks yet, but follow the same pattern
2032 * as the PMD locks to make it easier if we decide to. The VM should not be
2033 * considered ready to switch to split PUD locks yet; there may be places
2034 * which need to be converted from page_table_lock.
2036 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2038 return &mm->page_table_lock;
2041 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2043 spinlock_t *ptl = pud_lockptr(mm, pud);
2049 extern void __init pagecache_init(void);
2050 extern void free_area_init(unsigned long * zones_size);
2051 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2052 unsigned long zone_start_pfn, unsigned long *zholes_size);
2053 extern void free_initmem(void);
2056 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2057 * into the buddy system. The freed pages will be poisoned with pattern
2058 * "poison" if it's within range [0, UCHAR_MAX].
2059 * Return pages freed into the buddy system.
2061 extern unsigned long free_reserved_area(void *start, void *end,
2062 int poison, const char *s);
2064 #ifdef CONFIG_HIGHMEM
2066 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2067 * and totalram_pages.
2069 extern void free_highmem_page(struct page *page);
2072 extern void adjust_managed_page_count(struct page *page, long count);
2073 extern void mem_init_print_info(const char *str);
2075 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2077 /* Free the reserved page into the buddy system, so it gets managed. */
2078 static inline void __free_reserved_page(struct page *page)
2080 ClearPageReserved(page);
2081 init_page_count(page);
2085 static inline void free_reserved_page(struct page *page)
2087 __free_reserved_page(page);
2088 adjust_managed_page_count(page, 1);
2091 static inline void mark_page_reserved(struct page *page)
2093 SetPageReserved(page);
2094 adjust_managed_page_count(page, -1);
2098 * Default method to free all the __init memory into the buddy system.
2099 * The freed pages will be poisoned with pattern "poison" if it's within
2100 * range [0, UCHAR_MAX].
2101 * Return pages freed into the buddy system.
2103 static inline unsigned long free_initmem_default(int poison)
2105 extern char __init_begin[], __init_end[];
2107 return free_reserved_area(&__init_begin, &__init_end,
2108 poison, "unused kernel");
2111 static inline unsigned long get_num_physpages(void)
2114 unsigned long phys_pages = 0;
2116 for_each_online_node(nid)
2117 phys_pages += node_present_pages(nid);
2122 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2124 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2125 * zones, allocate the backing mem_map and account for memory holes in a more
2126 * architecture independent manner. This is a substitute for creating the
2127 * zone_sizes[] and zholes_size[] arrays and passing them to
2128 * free_area_init_node()
2130 * An architecture is expected to register range of page frames backed by
2131 * physical memory with memblock_add[_node]() before calling
2132 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2133 * usage, an architecture is expected to do something like
2135 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2137 * for_each_valid_physical_page_range()
2138 * memblock_add_node(base, size, nid)
2139 * free_area_init_nodes(max_zone_pfns);
2141 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2142 * registered physical page range. Similarly
2143 * sparse_memory_present_with_active_regions() calls memory_present() for
2144 * each range when SPARSEMEM is enabled.
2146 * See mm/page_alloc.c for more information on each function exposed by
2147 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2149 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2150 unsigned long node_map_pfn_alignment(void);
2151 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2152 unsigned long end_pfn);
2153 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2154 unsigned long end_pfn);
2155 extern void get_pfn_range_for_nid(unsigned int nid,
2156 unsigned long *start_pfn, unsigned long *end_pfn);
2157 extern unsigned long find_min_pfn_with_active_regions(void);
2158 extern void free_bootmem_with_active_regions(int nid,
2159 unsigned long max_low_pfn);
2160 extern void sparse_memory_present_with_active_regions(int nid);
2162 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2164 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2165 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2166 static inline int __early_pfn_to_nid(unsigned long pfn,
2167 struct mminit_pfnnid_cache *state)
2172 /* please see mm/page_alloc.c */
2173 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2174 /* there is a per-arch backend function. */
2175 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2176 struct mminit_pfnnid_cache *state);
2179 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2180 void zero_resv_unavail(void);
2182 static inline void zero_resv_unavail(void) {}
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_mkwrite(struct vm_area_struct *vma,
2533 unsigned long addr, pfn_t pfn);
2534 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2536 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2537 unsigned long addr, struct page *page)
2539 int err = vm_insert_page(vma, addr, page);
2542 return VM_FAULT_OOM;
2543 if (err < 0 && err != -EBUSY)
2544 return VM_FAULT_SIGBUS;
2546 return VM_FAULT_NOPAGE;
2549 static inline vm_fault_t vmf_error(int err)
2552 return VM_FAULT_OOM;
2553 return VM_FAULT_SIGBUS;
2556 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2557 unsigned int foll_flags);
2559 #define FOLL_WRITE 0x01 /* check pte is writable */
2560 #define FOLL_TOUCH 0x02 /* mark page accessed */
2561 #define FOLL_GET 0x04 /* do get_page on page */
2562 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2563 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2564 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2565 * and return without waiting upon it */
2566 #define FOLL_POPULATE 0x40 /* fault in page */
2567 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2568 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2569 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2570 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2571 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2572 #define FOLL_MLOCK 0x1000 /* lock present pages */
2573 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2574 #define FOLL_COW 0x4000 /* internal GUP flag */
2575 #define FOLL_ANON 0x8000 /* don't do file mappings */
2576 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2577 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2580 * NOTE on FOLL_LONGTERM:
2582 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2583 * period _often_ under userspace control. This is contrasted with
2584 * iov_iter_get_pages() where usages which are transient.
2586 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2587 * lifetime enforced by the filesystem and we need guarantees that longterm
2588 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2589 * the filesystem. Ideas for this coordination include revoking the longterm
2590 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2591 * added after the problem with filesystems was found FS DAX VMAs are
2592 * specifically failed. Filesystem pages are still subject to bugs and use of
2593 * FOLL_LONGTERM should be avoided on those pages.
2595 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2596 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2597 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2598 * is due to an incompatibility with the FS DAX check and
2599 * FAULT_FLAG_ALLOW_RETRY
2601 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2602 * that region. And so CMA attempts to migrate the page before pinning when
2603 * FOLL_LONGTERM is specified.
2606 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2608 if (vm_fault & VM_FAULT_OOM)
2610 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2611 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2612 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2617 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2618 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2619 unsigned long size, pte_fn_t fn, void *data);
2620 extern int apply_to_existing_page_range(struct mm_struct *mm,
2621 unsigned long address, unsigned long size,
2622 pte_fn_t fn, void *data);
2624 #ifdef CONFIG_PAGE_POISONING
2625 extern bool page_poisoning_enabled(void);
2626 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2628 static inline bool page_poisoning_enabled(void) { return false; }
2629 static inline void kernel_poison_pages(struct page *page, int numpages,
2633 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2634 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2636 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2638 static inline bool want_init_on_alloc(gfp_t flags)
2640 if (static_branch_unlikely(&init_on_alloc) &&
2641 !page_poisoning_enabled())
2643 return flags & __GFP_ZERO;
2646 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2647 DECLARE_STATIC_KEY_TRUE(init_on_free);
2649 DECLARE_STATIC_KEY_FALSE(init_on_free);
2651 static inline bool want_init_on_free(void)
2653 return static_branch_unlikely(&init_on_free) &&
2654 !page_poisoning_enabled();
2657 #ifdef CONFIG_DEBUG_PAGEALLOC
2658 extern void init_debug_pagealloc(void);
2660 static inline void init_debug_pagealloc(void) {}
2662 extern bool _debug_pagealloc_enabled_early;
2663 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2665 static inline bool debug_pagealloc_enabled(void)
2667 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2668 _debug_pagealloc_enabled_early;
2672 * For use in fast paths after init_debug_pagealloc() has run, or when a
2673 * false negative result is not harmful when called too early.
2675 static inline bool debug_pagealloc_enabled_static(void)
2677 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2680 return static_branch_unlikely(&_debug_pagealloc_enabled);
2683 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2684 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2687 kernel_map_pages(struct page *page, int numpages, int enable)
2689 __kernel_map_pages(page, numpages, enable);
2691 #ifdef CONFIG_HIBERNATION
2692 extern bool kernel_page_present(struct page *page);
2693 #endif /* CONFIG_HIBERNATION */
2694 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2696 kernel_map_pages(struct page *page, int numpages, int enable) {}
2697 #ifdef CONFIG_HIBERNATION
2698 static inline bool kernel_page_present(struct page *page) { return true; }
2699 #endif /* CONFIG_HIBERNATION */
2700 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2702 #ifdef __HAVE_ARCH_GATE_AREA
2703 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2704 extern int in_gate_area_no_mm(unsigned long addr);
2705 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2707 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2711 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2712 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2716 #endif /* __HAVE_ARCH_GATE_AREA */
2718 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2720 #ifdef CONFIG_SYSCTL
2721 extern int sysctl_drop_caches;
2722 int drop_caches_sysctl_handler(struct ctl_table *, int,
2723 void __user *, size_t *, loff_t *);
2726 void drop_slab(void);
2727 void drop_slab_node(int nid);
2730 #define randomize_va_space 0
2732 extern int randomize_va_space;
2735 const char * arch_vma_name(struct vm_area_struct *vma);
2737 void print_vma_addr(char *prefix, unsigned long rip);
2739 static inline void print_vma_addr(char *prefix, unsigned long rip)
2744 void *sparse_buffer_alloc(unsigned long size);
2745 struct page * __populate_section_memmap(unsigned long pfn,
2746 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2747 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2748 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2749 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2750 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2751 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2752 void *vmemmap_alloc_block(unsigned long size, int node);
2754 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2755 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2756 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2757 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2759 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2760 struct vmem_altmap *altmap);
2761 void vmemmap_populate_print_last(void);
2762 #ifdef CONFIG_MEMORY_HOTPLUG
2763 void vmemmap_free(unsigned long start, unsigned long end,
2764 struct vmem_altmap *altmap);
2766 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2767 unsigned long nr_pages);
2770 MF_COUNT_INCREASED = 1 << 0,
2771 MF_ACTION_REQUIRED = 1 << 1,
2772 MF_MUST_KILL = 1 << 2,
2773 MF_SOFT_OFFLINE = 1 << 3,
2775 extern int memory_failure(unsigned long pfn, int flags);
2776 extern void memory_failure_queue(unsigned long pfn, int flags);
2777 extern int unpoison_memory(unsigned long pfn);
2778 extern int get_hwpoison_page(struct page *page);
2779 #define put_hwpoison_page(page) put_page(page)
2780 extern int sysctl_memory_failure_early_kill;
2781 extern int sysctl_memory_failure_recovery;
2782 extern void shake_page(struct page *p, int access);
2783 extern atomic_long_t num_poisoned_pages __read_mostly;
2784 extern int soft_offline_page(unsigned long pfn, int flags);
2788 * Error handlers for various types of pages.
2791 MF_IGNORED, /* Error: cannot be handled */
2792 MF_FAILED, /* Error: handling failed */
2793 MF_DELAYED, /* Will be handled later */
2794 MF_RECOVERED, /* Successfully recovered */
2797 enum mf_action_page_type {
2799 MF_MSG_KERNEL_HIGH_ORDER,
2801 MF_MSG_DIFFERENT_COMPOUND,
2802 MF_MSG_POISONED_HUGE,
2805 MF_MSG_NON_PMD_HUGE,
2806 MF_MSG_UNMAP_FAILED,
2807 MF_MSG_DIRTY_SWAPCACHE,
2808 MF_MSG_CLEAN_SWAPCACHE,
2809 MF_MSG_DIRTY_MLOCKED_LRU,
2810 MF_MSG_CLEAN_MLOCKED_LRU,
2811 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2812 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2815 MF_MSG_TRUNCATED_LRU,
2822 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2823 extern void clear_huge_page(struct page *page,
2824 unsigned long addr_hint,
2825 unsigned int pages_per_huge_page);
2826 extern void copy_user_huge_page(struct page *dst, struct page *src,
2827 unsigned long addr_hint,
2828 struct vm_area_struct *vma,
2829 unsigned int pages_per_huge_page);
2830 extern long copy_huge_page_from_user(struct page *dst_page,
2831 const void __user *usr_src,
2832 unsigned int pages_per_huge_page,
2833 bool allow_pagefault);
2834 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2836 #ifdef CONFIG_DEBUG_PAGEALLOC
2837 extern unsigned int _debug_guardpage_minorder;
2838 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2840 static inline unsigned int debug_guardpage_minorder(void)
2842 return _debug_guardpage_minorder;
2845 static inline bool debug_guardpage_enabled(void)
2847 return static_branch_unlikely(&_debug_guardpage_enabled);
2850 static inline bool page_is_guard(struct page *page)
2852 if (!debug_guardpage_enabled())
2855 return PageGuard(page);
2858 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2859 static inline bool debug_guardpage_enabled(void) { return false; }
2860 static inline bool page_is_guard(struct page *page) { return false; }
2861 #endif /* CONFIG_DEBUG_PAGEALLOC */
2863 #if MAX_NUMNODES > 1
2864 void __init setup_nr_node_ids(void);
2866 static inline void setup_nr_node_ids(void) {}
2869 extern int memcmp_pages(struct page *page1, struct page *page2);
2871 static inline int pages_identical(struct page *page1, struct page *page2)
2873 return !memcmp_pages(page1, page2);
2876 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
2877 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
2878 pgoff_t first_index, pgoff_t nr,
2879 pgoff_t bitmap_pgoff,
2880 unsigned long *bitmap,
2884 unsigned long wp_shared_mapping_range(struct address_space *mapping,
2885 pgoff_t first_index, pgoff_t nr);
2888 #endif /* __KERNEL__ */
2889 #endif /* _LINUX_MM_H */