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) }
567 #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
568 static inline int pmd_devmap(pmd_t pmd)
572 static inline int pud_devmap(pud_t pud)
576 static inline int pgd_devmap(pgd_t pgd)
583 * FIXME: take this include out, include page-flags.h in
584 * files which need it (119 of them)
586 #include <linux/page-flags.h>
587 #include <linux/huge_mm.h>
590 * Methods to modify the page usage count.
592 * What counts for a page usage:
593 * - cache mapping (page->mapping)
594 * - private data (page->private)
595 * - page mapped in a task's page tables, each mapping
596 * is counted separately
598 * Also, many kernel routines increase the page count before a critical
599 * routine so they can be sure the page doesn't go away from under them.
603 * Drop a ref, return true if the refcount fell to zero (the page has no users)
605 static inline int put_page_testzero(struct page *page)
607 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
608 return page_ref_dec_and_test(page);
612 * Try to grab a ref unless the page has a refcount of zero, return false if
614 * This can be called when MMU is off so it must not access
615 * any of the virtual mappings.
617 static inline int get_page_unless_zero(struct page *page)
619 return page_ref_add_unless(page, 1, 0);
622 extern int page_is_ram(unsigned long pfn);
630 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
633 /* Support for virtually mapped pages */
634 struct page *vmalloc_to_page(const void *addr);
635 unsigned long vmalloc_to_pfn(const void *addr);
638 * Determine if an address is within the vmalloc range
640 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
641 * is no special casing required.
643 static inline bool is_vmalloc_addr(const void *x)
646 unsigned long addr = (unsigned long)x;
648 return addr >= VMALLOC_START && addr < VMALLOC_END;
654 #ifndef is_ioremap_addr
655 #define is_ioremap_addr(x) is_vmalloc_addr(x)
659 extern int is_vmalloc_or_module_addr(const void *x);
661 static inline int is_vmalloc_or_module_addr(const void *x)
667 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
668 static inline void *kvmalloc(size_t size, gfp_t flags)
670 return kvmalloc_node(size, flags, NUMA_NO_NODE);
672 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
674 return kvmalloc_node(size, flags | __GFP_ZERO, node);
676 static inline void *kvzalloc(size_t size, gfp_t flags)
678 return kvmalloc(size, flags | __GFP_ZERO);
681 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
685 if (unlikely(check_mul_overflow(n, size, &bytes)))
688 return kvmalloc(bytes, flags);
691 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
693 return kvmalloc_array(n, size, flags | __GFP_ZERO);
696 extern void kvfree(const void *addr);
698 static inline atomic_t *compound_mapcount_ptr(struct page *page)
700 return &page[1].compound_mapcount;
703 static inline int compound_mapcount(struct page *page)
705 VM_BUG_ON_PAGE(!PageCompound(page), page);
706 page = compound_head(page);
707 return atomic_read(compound_mapcount_ptr(page)) + 1;
711 * The atomic page->_mapcount, starts from -1: so that transitions
712 * both from it and to it can be tracked, using atomic_inc_and_test
713 * and atomic_add_negative(-1).
715 static inline void page_mapcount_reset(struct page *page)
717 atomic_set(&(page)->_mapcount, -1);
720 int __page_mapcount(struct page *page);
722 static inline int page_mapcount(struct page *page)
724 VM_BUG_ON_PAGE(PageSlab(page), page);
726 if (unlikely(PageCompound(page)))
727 return __page_mapcount(page);
728 return atomic_read(&page->_mapcount) + 1;
731 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
732 int total_mapcount(struct page *page);
733 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
735 static inline int total_mapcount(struct page *page)
737 return page_mapcount(page);
739 static inline int page_trans_huge_mapcount(struct page *page,
742 int mapcount = page_mapcount(page);
744 *total_mapcount = mapcount;
749 static inline struct page *virt_to_head_page(const void *x)
751 struct page *page = virt_to_page(x);
753 return compound_head(page);
756 void __put_page(struct page *page);
758 void put_pages_list(struct list_head *pages);
760 void split_page(struct page *page, unsigned int order);
763 * Compound pages have a destructor function. Provide a
764 * prototype for that function and accessor functions.
765 * These are _only_ valid on the head of a compound page.
767 typedef void compound_page_dtor(struct page *);
769 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
770 enum compound_dtor_id {
773 #ifdef CONFIG_HUGETLB_PAGE
776 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
781 extern compound_page_dtor * const compound_page_dtors[];
783 static inline void set_compound_page_dtor(struct page *page,
784 enum compound_dtor_id compound_dtor)
786 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
787 page[1].compound_dtor = compound_dtor;
790 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
792 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
793 return compound_page_dtors[page[1].compound_dtor];
796 static inline unsigned int compound_order(struct page *page)
800 return page[1].compound_order;
803 static inline void set_compound_order(struct page *page, unsigned int order)
805 page[1].compound_order = order;
808 /* Returns the number of pages in this potentially compound page. */
809 static inline unsigned long compound_nr(struct page *page)
811 return 1UL << compound_order(page);
814 /* Returns the number of bytes in this potentially compound page. */
815 static inline unsigned long page_size(struct page *page)
817 return PAGE_SIZE << compound_order(page);
820 /* Returns the number of bits needed for the number of bytes in a page */
821 static inline unsigned int page_shift(struct page *page)
823 return PAGE_SHIFT + compound_order(page);
826 void free_compound_page(struct page *page);
830 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
831 * servicing faults for write access. In the normal case, do always want
832 * pte_mkwrite. But get_user_pages can cause write faults for mappings
833 * that do not have writing enabled, when used by access_process_vm.
835 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
837 if (likely(vma->vm_flags & VM_WRITE))
838 pte = pte_mkwrite(pte);
842 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
844 vm_fault_t finish_fault(struct vm_fault *vmf);
845 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
849 * Multiple processes may "see" the same page. E.g. for untouched
850 * mappings of /dev/null, all processes see the same page full of
851 * zeroes, and text pages of executables and shared libraries have
852 * only one copy in memory, at most, normally.
854 * For the non-reserved pages, page_count(page) denotes a reference count.
855 * page_count() == 0 means the page is free. page->lru is then used for
856 * freelist management in the buddy allocator.
857 * page_count() > 0 means the page has been allocated.
859 * Pages are allocated by the slab allocator in order to provide memory
860 * to kmalloc and kmem_cache_alloc. In this case, the management of the
861 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
862 * unless a particular usage is carefully commented. (the responsibility of
863 * freeing the kmalloc memory is the caller's, of course).
865 * A page may be used by anyone else who does a __get_free_page().
866 * In this case, page_count still tracks the references, and should only
867 * be used through the normal accessor functions. The top bits of page->flags
868 * and page->virtual store page management information, but all other fields
869 * are unused and could be used privately, carefully. The management of this
870 * page is the responsibility of the one who allocated it, and those who have
871 * subsequently been given references to it.
873 * The other pages (we may call them "pagecache pages") are completely
874 * managed by the Linux memory manager: I/O, buffers, swapping etc.
875 * The following discussion applies only to them.
877 * A pagecache page contains an opaque `private' member, which belongs to the
878 * page's address_space. Usually, this is the address of a circular list of
879 * the page's disk buffers. PG_private must be set to tell the VM to call
880 * into the filesystem to release these pages.
882 * A page may belong to an inode's memory mapping. In this case, page->mapping
883 * is the pointer to the inode, and page->index is the file offset of the page,
884 * in units of PAGE_SIZE.
886 * If pagecache pages are not associated with an inode, they are said to be
887 * anonymous pages. These may become associated with the swapcache, and in that
888 * case PG_swapcache is set, and page->private is an offset into the swapcache.
890 * In either case (swapcache or inode backed), the pagecache itself holds one
891 * reference to the page. Setting PG_private should also increment the
892 * refcount. The each user mapping also has a reference to the page.
894 * The pagecache pages are stored in a per-mapping radix tree, which is
895 * rooted at mapping->i_pages, and indexed by offset.
896 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
897 * lists, we instead now tag pages as dirty/writeback in the radix tree.
899 * All pagecache pages may be subject to I/O:
900 * - inode pages may need to be read from disk,
901 * - inode pages which have been modified and are MAP_SHARED may need
902 * to be written back to the inode on disk,
903 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
904 * modified may need to be swapped out to swap space and (later) to be read
909 * The zone field is never updated after free_area_init_core()
910 * sets it, so none of the operations on it need to be atomic.
913 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
914 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
915 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
916 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
917 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
918 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
921 * Define the bit shifts to access each section. For non-existent
922 * sections we define the shift as 0; that plus a 0 mask ensures
923 * the compiler will optimise away reference to them.
925 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
926 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
927 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
928 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
929 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
931 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
932 #ifdef NODE_NOT_IN_PAGE_FLAGS
933 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
934 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
935 SECTIONS_PGOFF : ZONES_PGOFF)
937 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
938 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
939 NODES_PGOFF : ZONES_PGOFF)
942 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
944 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
945 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
948 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
949 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
950 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
951 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
952 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
953 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
955 static inline enum zone_type page_zonenum(const struct page *page)
957 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
960 #ifdef CONFIG_ZONE_DEVICE
961 static inline bool is_zone_device_page(const struct page *page)
963 return page_zonenum(page) == ZONE_DEVICE;
965 extern void memmap_init_zone_device(struct zone *, unsigned long,
966 unsigned long, struct dev_pagemap *);
968 static inline bool is_zone_device_page(const struct page *page)
974 #ifdef CONFIG_DEV_PAGEMAP_OPS
975 void __put_devmap_managed_page(struct page *page);
976 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
977 static inline bool put_devmap_managed_page(struct page *page)
979 if (!static_branch_unlikely(&devmap_managed_key))
981 if (!is_zone_device_page(page))
983 switch (page->pgmap->type) {
984 case MEMORY_DEVICE_PRIVATE:
985 case MEMORY_DEVICE_FS_DAX:
986 __put_devmap_managed_page(page);
994 #else /* CONFIG_DEV_PAGEMAP_OPS */
995 static inline bool put_devmap_managed_page(struct page *page)
999 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1001 static inline bool is_device_private_page(const struct page *page)
1003 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1004 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1005 is_zone_device_page(page) &&
1006 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1009 static inline bool is_pci_p2pdma_page(const struct page *page)
1011 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1012 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1013 is_zone_device_page(page) &&
1014 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1017 /* 127: arbitrary random number, small enough to assemble well */
1018 #define page_ref_zero_or_close_to_overflow(page) \
1019 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1021 static inline void get_page(struct page *page)
1023 page = compound_head(page);
1025 * Getting a normal page or the head of a compound page
1026 * requires to already have an elevated page->_refcount.
1028 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1032 static inline __must_check bool try_get_page(struct page *page)
1034 page = compound_head(page);
1035 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1041 static inline void put_page(struct page *page)
1043 page = compound_head(page);
1046 * For devmap managed pages we need to catch refcount transition from
1047 * 2 to 1, when refcount reach one it means the page is free and we
1048 * need to inform the device driver through callback. See
1049 * include/linux/memremap.h and HMM for details.
1051 if (put_devmap_managed_page(page))
1054 if (put_page_testzero(page))
1059 * put_user_page() - release a gup-pinned page
1060 * @page: pointer to page to be released
1062 * Pages that were pinned via get_user_pages*() must be released via
1063 * either put_user_page(), or one of the put_user_pages*() routines
1064 * below. This is so that eventually, pages that are pinned via
1065 * get_user_pages*() can be separately tracked and uniquely handled. In
1066 * particular, interactions with RDMA and filesystems need special
1069 * put_user_page() and put_page() are not interchangeable, despite this early
1070 * implementation that makes them look the same. put_user_page() calls must
1071 * be perfectly matched up with get_user_page() calls.
1073 static inline void put_user_page(struct page *page)
1078 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1081 void put_user_pages(struct page **pages, unsigned long npages);
1083 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1084 #define SECTION_IN_PAGE_FLAGS
1088 * The identification function is mainly used by the buddy allocator for
1089 * determining if two pages could be buddies. We are not really identifying
1090 * the zone since we could be using the section number id if we do not have
1091 * node id available in page flags.
1092 * We only guarantee that it will return the same value for two combinable
1095 static inline int page_zone_id(struct page *page)
1097 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1100 #ifdef NODE_NOT_IN_PAGE_FLAGS
1101 extern int page_to_nid(const struct page *page);
1103 static inline int page_to_nid(const struct page *page)
1105 struct page *p = (struct page *)page;
1107 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1111 #ifdef CONFIG_NUMA_BALANCING
1112 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1114 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1117 static inline int cpupid_to_pid(int cpupid)
1119 return cpupid & LAST__PID_MASK;
1122 static inline int cpupid_to_cpu(int cpupid)
1124 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1127 static inline int cpupid_to_nid(int cpupid)
1129 return cpu_to_node(cpupid_to_cpu(cpupid));
1132 static inline bool cpupid_pid_unset(int cpupid)
1134 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1137 static inline bool cpupid_cpu_unset(int cpupid)
1139 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1142 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1144 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1147 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1148 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1149 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1151 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1154 static inline int page_cpupid_last(struct page *page)
1156 return page->_last_cpupid;
1158 static inline void page_cpupid_reset_last(struct page *page)
1160 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1163 static inline int page_cpupid_last(struct page *page)
1165 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1168 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1170 static inline void page_cpupid_reset_last(struct page *page)
1172 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1174 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1175 #else /* !CONFIG_NUMA_BALANCING */
1176 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1178 return page_to_nid(page); /* XXX */
1181 static inline int page_cpupid_last(struct page *page)
1183 return page_to_nid(page); /* XXX */
1186 static inline int cpupid_to_nid(int cpupid)
1191 static inline int cpupid_to_pid(int cpupid)
1196 static inline int cpupid_to_cpu(int cpupid)
1201 static inline int cpu_pid_to_cpupid(int nid, int pid)
1206 static inline bool cpupid_pid_unset(int cpupid)
1211 static inline void page_cpupid_reset_last(struct page *page)
1215 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1219 #endif /* CONFIG_NUMA_BALANCING */
1221 #ifdef CONFIG_KASAN_SW_TAGS
1222 static inline u8 page_kasan_tag(const struct page *page)
1224 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1227 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1229 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1230 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1233 static inline void page_kasan_tag_reset(struct page *page)
1235 page_kasan_tag_set(page, 0xff);
1238 static inline u8 page_kasan_tag(const struct page *page)
1243 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1244 static inline void page_kasan_tag_reset(struct page *page) { }
1247 static inline struct zone *page_zone(const struct page *page)
1249 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1252 static inline pg_data_t *page_pgdat(const struct page *page)
1254 return NODE_DATA(page_to_nid(page));
1257 #ifdef SECTION_IN_PAGE_FLAGS
1258 static inline void set_page_section(struct page *page, unsigned long section)
1260 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1261 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1264 static inline unsigned long page_to_section(const struct page *page)
1266 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1270 static inline void set_page_zone(struct page *page, enum zone_type zone)
1272 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1273 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1276 static inline void set_page_node(struct page *page, unsigned long node)
1278 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1279 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1282 static inline void set_page_links(struct page *page, enum zone_type zone,
1283 unsigned long node, unsigned long pfn)
1285 set_page_zone(page, zone);
1286 set_page_node(page, node);
1287 #ifdef SECTION_IN_PAGE_FLAGS
1288 set_page_section(page, pfn_to_section_nr(pfn));
1293 static inline struct mem_cgroup *page_memcg(struct page *page)
1295 return page->mem_cgroup;
1297 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1299 WARN_ON_ONCE(!rcu_read_lock_held());
1300 return READ_ONCE(page->mem_cgroup);
1303 static inline struct mem_cgroup *page_memcg(struct page *page)
1307 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1309 WARN_ON_ONCE(!rcu_read_lock_held());
1315 * Some inline functions in vmstat.h depend on page_zone()
1317 #include <linux/vmstat.h>
1319 static __always_inline void *lowmem_page_address(const struct page *page)
1321 return page_to_virt(page);
1324 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1325 #define HASHED_PAGE_VIRTUAL
1328 #if defined(WANT_PAGE_VIRTUAL)
1329 static inline void *page_address(const struct page *page)
1331 return page->virtual;
1333 static inline void set_page_address(struct page *page, void *address)
1335 page->virtual = address;
1337 #define page_address_init() do { } while(0)
1340 #if defined(HASHED_PAGE_VIRTUAL)
1341 void *page_address(const struct page *page);
1342 void set_page_address(struct page *page, void *virtual);
1343 void page_address_init(void);
1346 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1347 #define page_address(page) lowmem_page_address(page)
1348 #define set_page_address(page, address) do { } while(0)
1349 #define page_address_init() do { } while(0)
1352 extern void *page_rmapping(struct page *page);
1353 extern struct anon_vma *page_anon_vma(struct page *page);
1354 extern struct address_space *page_mapping(struct page *page);
1356 extern struct address_space *__page_file_mapping(struct page *);
1359 struct address_space *page_file_mapping(struct page *page)
1361 if (unlikely(PageSwapCache(page)))
1362 return __page_file_mapping(page);
1364 return page->mapping;
1367 extern pgoff_t __page_file_index(struct page *page);
1370 * Return the pagecache index of the passed page. Regular pagecache pages
1371 * use ->index whereas swapcache pages use swp_offset(->private)
1373 static inline pgoff_t page_index(struct page *page)
1375 if (unlikely(PageSwapCache(page)))
1376 return __page_file_index(page);
1380 bool page_mapped(struct page *page);
1381 struct address_space *page_mapping(struct page *page);
1382 struct address_space *page_mapping_file(struct page *page);
1385 * Return true only if the page has been allocated with
1386 * ALLOC_NO_WATERMARKS and the low watermark was not
1387 * met implying that the system is under some pressure.
1389 static inline bool page_is_pfmemalloc(struct page *page)
1392 * Page index cannot be this large so this must be
1393 * a pfmemalloc page.
1395 return page->index == -1UL;
1399 * Only to be called by the page allocator on a freshly allocated
1402 static inline void set_page_pfmemalloc(struct page *page)
1407 static inline void clear_page_pfmemalloc(struct page *page)
1413 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1415 extern void pagefault_out_of_memory(void);
1417 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1420 * Flags passed to show_mem() and show_free_areas() to suppress output in
1423 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1425 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1428 extern bool can_do_mlock(void);
1430 static inline bool can_do_mlock(void) { return false; }
1432 extern int user_shm_lock(size_t, struct user_struct *);
1433 extern void user_shm_unlock(size_t, struct user_struct *);
1436 * Parameter block passed down to zap_pte_range in exceptional cases.
1438 struct zap_details {
1439 struct address_space *check_mapping; /* Check page->mapping if set */
1440 pgoff_t first_index; /* Lowest page->index to unmap */
1441 pgoff_t last_index; /* Highest page->index to unmap */
1444 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1446 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1449 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1450 unsigned long size);
1451 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1452 unsigned long size);
1453 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1454 unsigned long start, unsigned long end);
1456 struct mmu_notifier_range;
1458 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1459 unsigned long end, unsigned long floor, unsigned long ceiling);
1460 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1461 struct vm_area_struct *vma);
1462 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1463 struct mmu_notifier_range *range,
1464 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1465 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1466 unsigned long *pfn);
1467 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1468 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1469 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1470 void *buf, int len, int write);
1472 extern void truncate_pagecache(struct inode *inode, loff_t new);
1473 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1474 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1475 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1476 int truncate_inode_page(struct address_space *mapping, struct page *page);
1477 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1478 int invalidate_inode_page(struct page *page);
1481 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1482 unsigned long address, unsigned int flags);
1483 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1484 unsigned long address, unsigned int fault_flags,
1486 void unmap_mapping_pages(struct address_space *mapping,
1487 pgoff_t start, pgoff_t nr, bool even_cows);
1488 void unmap_mapping_range(struct address_space *mapping,
1489 loff_t const holebegin, loff_t const holelen, int even_cows);
1491 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1492 unsigned long address, unsigned int flags)
1494 /* should never happen if there's no MMU */
1496 return VM_FAULT_SIGBUS;
1498 static inline int fixup_user_fault(struct task_struct *tsk,
1499 struct mm_struct *mm, unsigned long address,
1500 unsigned int fault_flags, bool *unlocked)
1502 /* should never happen if there's no MMU */
1506 static inline void unmap_mapping_pages(struct address_space *mapping,
1507 pgoff_t start, pgoff_t nr, bool even_cows) { }
1508 static inline void unmap_mapping_range(struct address_space *mapping,
1509 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1512 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1513 loff_t const holebegin, loff_t const holelen)
1515 unmap_mapping_range(mapping, holebegin, holelen, 0);
1518 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1519 void *buf, int len, unsigned int gup_flags);
1520 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1521 void *buf, int len, unsigned int gup_flags);
1522 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1523 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1525 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1526 unsigned long start, unsigned long nr_pages,
1527 unsigned int gup_flags, struct page **pages,
1528 struct vm_area_struct **vmas, int *locked);
1529 long get_user_pages(unsigned long start, unsigned long nr_pages,
1530 unsigned int gup_flags, struct page **pages,
1531 struct vm_area_struct **vmas);
1532 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1533 unsigned int gup_flags, struct page **pages, int *locked);
1534 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1535 struct page **pages, unsigned int gup_flags);
1537 int get_user_pages_fast(unsigned long start, int nr_pages,
1538 unsigned int gup_flags, struct page **pages);
1540 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1541 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1542 struct task_struct *task, bool bypass_rlim);
1544 /* Container for pinned pfns / pages */
1545 struct frame_vector {
1546 unsigned int nr_allocated; /* Number of frames we have space for */
1547 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1548 bool got_ref; /* Did we pin pages by getting page ref? */
1549 bool is_pfns; /* Does array contain pages or pfns? */
1550 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1551 * pfns_vector_pages() or pfns_vector_pfns()
1555 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1556 void frame_vector_destroy(struct frame_vector *vec);
1557 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1558 unsigned int gup_flags, struct frame_vector *vec);
1559 void put_vaddr_frames(struct frame_vector *vec);
1560 int frame_vector_to_pages(struct frame_vector *vec);
1561 void frame_vector_to_pfns(struct frame_vector *vec);
1563 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1565 return vec->nr_frames;
1568 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1571 int err = frame_vector_to_pages(vec);
1574 return ERR_PTR(err);
1576 return (struct page **)(vec->ptrs);
1579 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1582 frame_vector_to_pfns(vec);
1583 return (unsigned long *)(vec->ptrs);
1587 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1588 struct page **pages);
1589 int get_kernel_page(unsigned long start, int write, struct page **pages);
1590 struct page *get_dump_page(unsigned long addr);
1592 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1593 extern void do_invalidatepage(struct page *page, unsigned int offset,
1594 unsigned int length);
1596 void __set_page_dirty(struct page *, struct address_space *, int warn);
1597 int __set_page_dirty_nobuffers(struct page *page);
1598 int __set_page_dirty_no_writeback(struct page *page);
1599 int redirty_page_for_writepage(struct writeback_control *wbc,
1601 void account_page_dirtied(struct page *page, struct address_space *mapping);
1602 void account_page_cleaned(struct page *page, struct address_space *mapping,
1603 struct bdi_writeback *wb);
1604 int set_page_dirty(struct page *page);
1605 int set_page_dirty_lock(struct page *page);
1606 void __cancel_dirty_page(struct page *page);
1607 static inline void cancel_dirty_page(struct page *page)
1609 /* Avoid atomic ops, locking, etc. when not actually needed. */
1610 if (PageDirty(page))
1611 __cancel_dirty_page(page);
1613 int clear_page_dirty_for_io(struct page *page);
1615 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1617 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1618 unsigned long old_addr, struct vm_area_struct *new_vma,
1619 unsigned long new_addr, unsigned long len,
1620 bool need_rmap_locks);
1621 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1622 unsigned long end, pgprot_t newprot,
1623 int dirty_accountable, int prot_numa);
1624 extern int mprotect_fixup(struct vm_area_struct *vma,
1625 struct vm_area_struct **pprev, unsigned long start,
1626 unsigned long end, unsigned long newflags);
1629 * doesn't attempt to fault and will return short.
1631 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1632 struct page **pages);
1634 * per-process(per-mm_struct) statistics.
1636 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1638 long val = atomic_long_read(&mm->rss_stat.count[member]);
1640 #ifdef SPLIT_RSS_COUNTING
1642 * counter is updated in asynchronous manner and may go to minus.
1643 * But it's never be expected number for users.
1648 return (unsigned long)val;
1651 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1653 atomic_long_add(value, &mm->rss_stat.count[member]);
1656 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1658 atomic_long_inc(&mm->rss_stat.count[member]);
1661 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1663 atomic_long_dec(&mm->rss_stat.count[member]);
1666 /* Optimized variant when page is already known not to be PageAnon */
1667 static inline int mm_counter_file(struct page *page)
1669 if (PageSwapBacked(page))
1670 return MM_SHMEMPAGES;
1671 return MM_FILEPAGES;
1674 static inline int mm_counter(struct page *page)
1677 return MM_ANONPAGES;
1678 return mm_counter_file(page);
1681 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1683 return get_mm_counter(mm, MM_FILEPAGES) +
1684 get_mm_counter(mm, MM_ANONPAGES) +
1685 get_mm_counter(mm, MM_SHMEMPAGES);
1688 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1690 return max(mm->hiwater_rss, get_mm_rss(mm));
1693 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1695 return max(mm->hiwater_vm, mm->total_vm);
1698 static inline void update_hiwater_rss(struct mm_struct *mm)
1700 unsigned long _rss = get_mm_rss(mm);
1702 if ((mm)->hiwater_rss < _rss)
1703 (mm)->hiwater_rss = _rss;
1706 static inline void update_hiwater_vm(struct mm_struct *mm)
1708 if (mm->hiwater_vm < mm->total_vm)
1709 mm->hiwater_vm = mm->total_vm;
1712 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1714 mm->hiwater_rss = get_mm_rss(mm);
1717 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1718 struct mm_struct *mm)
1720 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1722 if (*maxrss < hiwater_rss)
1723 *maxrss = hiwater_rss;
1726 #if defined(SPLIT_RSS_COUNTING)
1727 void sync_mm_rss(struct mm_struct *mm);
1729 static inline void sync_mm_rss(struct mm_struct *mm)
1734 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1735 static inline int pte_devmap(pte_t pte)
1741 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1743 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1745 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1749 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1753 #ifdef __PAGETABLE_P4D_FOLDED
1754 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1755 unsigned long address)
1760 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1763 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1764 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1765 unsigned long address)
1769 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1770 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1773 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1775 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1777 if (mm_pud_folded(mm))
1779 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1782 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1784 if (mm_pud_folded(mm))
1786 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1790 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1791 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1792 unsigned long address)
1797 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1798 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1801 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1803 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1805 if (mm_pmd_folded(mm))
1807 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1810 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1812 if (mm_pmd_folded(mm))
1814 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1819 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1821 atomic_long_set(&mm->pgtables_bytes, 0);
1824 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1826 return atomic_long_read(&mm->pgtables_bytes);
1829 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1831 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1834 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1836 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1840 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1841 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1846 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1847 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1850 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1851 int __pte_alloc_kernel(pmd_t *pmd);
1854 * The following ifdef needed to get the 4level-fixup.h header to work.
1855 * Remove it when 4level-fixup.h has been removed.
1857 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1859 #ifndef __ARCH_HAS_5LEVEL_HACK
1860 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1861 unsigned long address)
1863 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1864 NULL : p4d_offset(pgd, address);
1867 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1868 unsigned long address)
1870 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1871 NULL : pud_offset(p4d, address);
1873 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1875 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1877 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1878 NULL: pmd_offset(pud, address);
1880 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1882 #if USE_SPLIT_PTE_PTLOCKS
1883 #if ALLOC_SPLIT_PTLOCKS
1884 void __init ptlock_cache_init(void);
1885 extern bool ptlock_alloc(struct page *page);
1886 extern void ptlock_free(struct page *page);
1888 static inline spinlock_t *ptlock_ptr(struct page *page)
1892 #else /* ALLOC_SPLIT_PTLOCKS */
1893 static inline void ptlock_cache_init(void)
1897 static inline bool ptlock_alloc(struct page *page)
1902 static inline void ptlock_free(struct page *page)
1906 static inline spinlock_t *ptlock_ptr(struct page *page)
1910 #endif /* ALLOC_SPLIT_PTLOCKS */
1912 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1914 return ptlock_ptr(pmd_page(*pmd));
1917 static inline bool ptlock_init(struct page *page)
1920 * prep_new_page() initialize page->private (and therefore page->ptl)
1921 * with 0. Make sure nobody took it in use in between.
1923 * It can happen if arch try to use slab for page table allocation:
1924 * slab code uses page->slab_cache, which share storage with page->ptl.
1926 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1927 if (!ptlock_alloc(page))
1929 spin_lock_init(ptlock_ptr(page));
1933 #else /* !USE_SPLIT_PTE_PTLOCKS */
1935 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1937 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1939 return &mm->page_table_lock;
1941 static inline void ptlock_cache_init(void) {}
1942 static inline bool ptlock_init(struct page *page) { return true; }
1943 static inline void ptlock_free(struct page *page) {}
1944 #endif /* USE_SPLIT_PTE_PTLOCKS */
1946 static inline void pgtable_init(void)
1948 ptlock_cache_init();
1949 pgtable_cache_init();
1952 static inline bool pgtable_page_ctor(struct page *page)
1954 if (!ptlock_init(page))
1956 __SetPageTable(page);
1957 inc_zone_page_state(page, NR_PAGETABLE);
1961 static inline void pgtable_page_dtor(struct page *page)
1964 __ClearPageTable(page);
1965 dec_zone_page_state(page, NR_PAGETABLE);
1968 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1970 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1971 pte_t *__pte = pte_offset_map(pmd, address); \
1977 #define pte_unmap_unlock(pte, ptl) do { \
1982 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1984 #define pte_alloc_map(mm, pmd, address) \
1985 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1987 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1988 (pte_alloc(mm, pmd) ? \
1989 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1991 #define pte_alloc_kernel(pmd, address) \
1992 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1993 NULL: pte_offset_kernel(pmd, address))
1995 #if USE_SPLIT_PMD_PTLOCKS
1997 static struct page *pmd_to_page(pmd_t *pmd)
1999 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2000 return virt_to_page((void *)((unsigned long) pmd & mask));
2003 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2005 return ptlock_ptr(pmd_to_page(pmd));
2008 static inline bool pgtable_pmd_page_ctor(struct page *page)
2010 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2011 page->pmd_huge_pte = NULL;
2013 return ptlock_init(page);
2016 static inline void pgtable_pmd_page_dtor(struct page *page)
2018 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2019 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2024 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2028 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2030 return &mm->page_table_lock;
2033 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2034 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2036 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2040 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2042 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2048 * No scalability reason to split PUD locks yet, but follow the same pattern
2049 * as the PMD locks to make it easier if we decide to. The VM should not be
2050 * considered ready to switch to split PUD locks yet; there may be places
2051 * which need to be converted from page_table_lock.
2053 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2055 return &mm->page_table_lock;
2058 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2060 spinlock_t *ptl = pud_lockptr(mm, pud);
2066 extern void __init pagecache_init(void);
2067 extern void free_area_init(unsigned long * zones_size);
2068 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2069 unsigned long zone_start_pfn, unsigned long *zholes_size);
2070 extern void free_initmem(void);
2073 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2074 * into the buddy system. The freed pages will be poisoned with pattern
2075 * "poison" if it's within range [0, UCHAR_MAX].
2076 * Return pages freed into the buddy system.
2078 extern unsigned long free_reserved_area(void *start, void *end,
2079 int poison, const char *s);
2081 #ifdef CONFIG_HIGHMEM
2083 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2084 * and totalram_pages.
2086 extern void free_highmem_page(struct page *page);
2089 extern void adjust_managed_page_count(struct page *page, long count);
2090 extern void mem_init_print_info(const char *str);
2092 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2094 /* Free the reserved page into the buddy system, so it gets managed. */
2095 static inline void __free_reserved_page(struct page *page)
2097 ClearPageReserved(page);
2098 init_page_count(page);
2102 static inline void free_reserved_page(struct page *page)
2104 __free_reserved_page(page);
2105 adjust_managed_page_count(page, 1);
2108 static inline void mark_page_reserved(struct page *page)
2110 SetPageReserved(page);
2111 adjust_managed_page_count(page, -1);
2115 * Default method to free all the __init memory into the buddy system.
2116 * The freed pages will be poisoned with pattern "poison" if it's within
2117 * range [0, UCHAR_MAX].
2118 * Return pages freed into the buddy system.
2120 static inline unsigned long free_initmem_default(int poison)
2122 extern char __init_begin[], __init_end[];
2124 return free_reserved_area(&__init_begin, &__init_end,
2125 poison, "unused kernel");
2128 static inline unsigned long get_num_physpages(void)
2131 unsigned long phys_pages = 0;
2133 for_each_online_node(nid)
2134 phys_pages += node_present_pages(nid);
2139 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2141 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2142 * zones, allocate the backing mem_map and account for memory holes in a more
2143 * architecture independent manner. This is a substitute for creating the
2144 * zone_sizes[] and zholes_size[] arrays and passing them to
2145 * free_area_init_node()
2147 * An architecture is expected to register range of page frames backed by
2148 * physical memory with memblock_add[_node]() before calling
2149 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2150 * usage, an architecture is expected to do something like
2152 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2154 * for_each_valid_physical_page_range()
2155 * memblock_add_node(base, size, nid)
2156 * free_area_init_nodes(max_zone_pfns);
2158 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2159 * registered physical page range. Similarly
2160 * sparse_memory_present_with_active_regions() calls memory_present() for
2161 * each range when SPARSEMEM is enabled.
2163 * See mm/page_alloc.c for more information on each function exposed by
2164 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2166 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2167 unsigned long node_map_pfn_alignment(void);
2168 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2169 unsigned long end_pfn);
2170 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2171 unsigned long end_pfn);
2172 extern void get_pfn_range_for_nid(unsigned int nid,
2173 unsigned long *start_pfn, unsigned long *end_pfn);
2174 extern unsigned long find_min_pfn_with_active_regions(void);
2175 extern void free_bootmem_with_active_regions(int nid,
2176 unsigned long max_low_pfn);
2177 extern void sparse_memory_present_with_active_regions(int nid);
2179 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2181 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2182 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2183 static inline int __early_pfn_to_nid(unsigned long pfn,
2184 struct mminit_pfnnid_cache *state)
2189 /* please see mm/page_alloc.c */
2190 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2191 /* there is a per-arch backend function. */
2192 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2193 struct mminit_pfnnid_cache *state);
2196 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2197 void zero_resv_unavail(void);
2199 static inline void zero_resv_unavail(void) {}
2202 extern void set_dma_reserve(unsigned long new_dma_reserve);
2203 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2204 enum memmap_context, struct vmem_altmap *);
2205 extern void setup_per_zone_wmarks(void);
2206 extern int __meminit init_per_zone_wmark_min(void);
2207 extern void mem_init(void);
2208 extern void __init mmap_init(void);
2209 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2210 extern long si_mem_available(void);
2211 extern void si_meminfo(struct sysinfo * val);
2212 extern void si_meminfo_node(struct sysinfo *val, int nid);
2213 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2214 extern unsigned long arch_reserved_kernel_pages(void);
2217 extern __printf(3, 4)
2218 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2220 extern void setup_per_cpu_pageset(void);
2222 extern void zone_pcp_update(struct zone *zone);
2223 extern void zone_pcp_reset(struct zone *zone);
2226 extern int min_free_kbytes;
2227 extern int watermark_boost_factor;
2228 extern int watermark_scale_factor;
2231 extern atomic_long_t mmap_pages_allocated;
2232 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2234 /* interval_tree.c */
2235 void vma_interval_tree_insert(struct vm_area_struct *node,
2236 struct rb_root_cached *root);
2237 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2238 struct vm_area_struct *prev,
2239 struct rb_root_cached *root);
2240 void vma_interval_tree_remove(struct vm_area_struct *node,
2241 struct rb_root_cached *root);
2242 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2243 unsigned long start, unsigned long last);
2244 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2245 unsigned long start, unsigned long last);
2247 #define vma_interval_tree_foreach(vma, root, start, last) \
2248 for (vma = vma_interval_tree_iter_first(root, start, last); \
2249 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2251 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2252 struct rb_root_cached *root);
2253 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2254 struct rb_root_cached *root);
2255 struct anon_vma_chain *
2256 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2257 unsigned long start, unsigned long last);
2258 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2259 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2260 #ifdef CONFIG_DEBUG_VM_RB
2261 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2264 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2265 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2266 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2269 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2270 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2271 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2272 struct vm_area_struct *expand);
2273 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2274 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2276 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2278 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2279 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2280 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2281 struct mempolicy *, struct vm_userfaultfd_ctx);
2282 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2283 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2284 unsigned long addr, int new_below);
2285 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2286 unsigned long addr, int new_below);
2287 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2288 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2289 struct rb_node **, struct rb_node *);
2290 extern void unlink_file_vma(struct vm_area_struct *);
2291 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2292 unsigned long addr, unsigned long len, pgoff_t pgoff,
2293 bool *need_rmap_locks);
2294 extern void exit_mmap(struct mm_struct *);
2296 static inline int check_data_rlimit(unsigned long rlim,
2298 unsigned long start,
2299 unsigned long end_data,
2300 unsigned long start_data)
2302 if (rlim < RLIM_INFINITY) {
2303 if (((new - start) + (end_data - start_data)) > rlim)
2310 extern int mm_take_all_locks(struct mm_struct *mm);
2311 extern void mm_drop_all_locks(struct mm_struct *mm);
2313 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2314 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2315 extern struct file *get_task_exe_file(struct task_struct *task);
2317 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2318 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2320 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2321 const struct vm_special_mapping *sm);
2322 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2323 unsigned long addr, unsigned long len,
2324 unsigned long flags,
2325 const struct vm_special_mapping *spec);
2326 /* This is an obsolete alternative to _install_special_mapping. */
2327 extern int install_special_mapping(struct mm_struct *mm,
2328 unsigned long addr, unsigned long len,
2329 unsigned long flags, struct page **pages);
2331 unsigned long randomize_stack_top(unsigned long stack_top);
2333 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2335 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2336 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2337 struct list_head *uf);
2338 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2339 unsigned long len, unsigned long prot, unsigned long flags,
2340 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2341 struct list_head *uf);
2342 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2343 struct list_head *uf, bool downgrade);
2344 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2345 struct list_head *uf);
2347 static inline unsigned long
2348 do_mmap_pgoff(struct file *file, unsigned long addr,
2349 unsigned long len, unsigned long prot, unsigned long flags,
2350 unsigned long pgoff, unsigned long *populate,
2351 struct list_head *uf)
2353 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2357 extern int __mm_populate(unsigned long addr, unsigned long len,
2359 static inline void mm_populate(unsigned long addr, unsigned long len)
2362 (void) __mm_populate(addr, len, 1);
2365 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2368 /* These take the mm semaphore themselves */
2369 extern int __must_check vm_brk(unsigned long, unsigned long);
2370 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2371 extern int vm_munmap(unsigned long, size_t);
2372 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2373 unsigned long, unsigned long,
2374 unsigned long, unsigned long);
2376 struct vm_unmapped_area_info {
2377 #define VM_UNMAPPED_AREA_TOPDOWN 1
2378 unsigned long flags;
2379 unsigned long length;
2380 unsigned long low_limit;
2381 unsigned long high_limit;
2382 unsigned long align_mask;
2383 unsigned long align_offset;
2386 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2387 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2390 * Search for an unmapped address range.
2392 * We are looking for a range that:
2393 * - does not intersect with any VMA;
2394 * - is contained within the [low_limit, high_limit) interval;
2395 * - is at least the desired size.
2396 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2398 static inline unsigned long
2399 vm_unmapped_area(struct vm_unmapped_area_info *info)
2401 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2402 return unmapped_area_topdown(info);
2404 return unmapped_area(info);
2408 extern void truncate_inode_pages(struct address_space *, loff_t);
2409 extern void truncate_inode_pages_range(struct address_space *,
2410 loff_t lstart, loff_t lend);
2411 extern void truncate_inode_pages_final(struct address_space *);
2413 /* generic vm_area_ops exported for stackable file systems */
2414 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2415 extern void filemap_map_pages(struct vm_fault *vmf,
2416 pgoff_t start_pgoff, pgoff_t end_pgoff);
2417 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2419 /* mm/page-writeback.c */
2420 int __must_check write_one_page(struct page *page);
2421 void task_dirty_inc(struct task_struct *tsk);
2424 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2426 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2427 pgoff_t offset, unsigned long nr_to_read);
2429 void page_cache_sync_readahead(struct address_space *mapping,
2430 struct file_ra_state *ra,
2433 unsigned long size);
2435 void page_cache_async_readahead(struct address_space *mapping,
2436 struct file_ra_state *ra,
2440 unsigned long size);
2442 extern unsigned long stack_guard_gap;
2443 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2444 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2446 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2447 extern int expand_downwards(struct vm_area_struct *vma,
2448 unsigned long address);
2450 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2452 #define expand_upwards(vma, address) (0)
2455 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2456 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2457 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2458 struct vm_area_struct **pprev);
2460 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2461 NULL if none. Assume start_addr < end_addr. */
2462 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2464 struct vm_area_struct * vma = find_vma(mm,start_addr);
2466 if (vma && end_addr <= vma->vm_start)
2471 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2473 unsigned long vm_start = vma->vm_start;
2475 if (vma->vm_flags & VM_GROWSDOWN) {
2476 vm_start -= stack_guard_gap;
2477 if (vm_start > vma->vm_start)
2483 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2485 unsigned long vm_end = vma->vm_end;
2487 if (vma->vm_flags & VM_GROWSUP) {
2488 vm_end += stack_guard_gap;
2489 if (vm_end < vma->vm_end)
2490 vm_end = -PAGE_SIZE;
2495 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2497 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2500 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2501 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2502 unsigned long vm_start, unsigned long vm_end)
2504 struct vm_area_struct *vma = find_vma(mm, vm_start);
2506 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2512 static inline bool range_in_vma(struct vm_area_struct *vma,
2513 unsigned long start, unsigned long end)
2515 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2519 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2520 void vma_set_page_prot(struct vm_area_struct *vma);
2522 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2526 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2528 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2532 #ifdef CONFIG_NUMA_BALANCING
2533 unsigned long change_prot_numa(struct vm_area_struct *vma,
2534 unsigned long start, unsigned long end);
2537 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2538 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2539 unsigned long pfn, unsigned long size, pgprot_t);
2540 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2541 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2543 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2545 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2547 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2548 unsigned long pfn, pgprot_t pgprot);
2549 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2551 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2552 unsigned long addr, pfn_t pfn);
2553 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2555 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2556 unsigned long addr, struct page *page)
2558 int err = vm_insert_page(vma, addr, page);
2561 return VM_FAULT_OOM;
2562 if (err < 0 && err != -EBUSY)
2563 return VM_FAULT_SIGBUS;
2565 return VM_FAULT_NOPAGE;
2568 static inline vm_fault_t vmf_error(int err)
2571 return VM_FAULT_OOM;
2572 return VM_FAULT_SIGBUS;
2575 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2576 unsigned int foll_flags);
2578 #define FOLL_WRITE 0x01 /* check pte is writable */
2579 #define FOLL_TOUCH 0x02 /* mark page accessed */
2580 #define FOLL_GET 0x04 /* do get_page on page */
2581 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2582 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2583 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2584 * and return without waiting upon it */
2585 #define FOLL_POPULATE 0x40 /* fault in page */
2586 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2587 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2588 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2589 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2590 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2591 #define FOLL_MLOCK 0x1000 /* lock present pages */
2592 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2593 #define FOLL_COW 0x4000 /* internal GUP flag */
2594 #define FOLL_ANON 0x8000 /* don't do file mappings */
2595 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2596 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2599 * NOTE on FOLL_LONGTERM:
2601 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2602 * period _often_ under userspace control. This is contrasted with
2603 * iov_iter_get_pages() where usages which are transient.
2605 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2606 * lifetime enforced by the filesystem and we need guarantees that longterm
2607 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2608 * the filesystem. Ideas for this coordination include revoking the longterm
2609 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2610 * added after the problem with filesystems was found FS DAX VMAs are
2611 * specifically failed. Filesystem pages are still subject to bugs and use of
2612 * FOLL_LONGTERM should be avoided on those pages.
2614 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2615 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2616 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2617 * is due to an incompatibility with the FS DAX check and
2618 * FAULT_FLAG_ALLOW_RETRY
2620 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2621 * that region. And so CMA attempts to migrate the page before pinning when
2622 * FOLL_LONGTERM is specified.
2625 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2627 if (vm_fault & VM_FAULT_OOM)
2629 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2630 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2631 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2636 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2637 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2638 unsigned long size, pte_fn_t fn, void *data);
2641 #ifdef CONFIG_PAGE_POISONING
2642 extern bool page_poisoning_enabled(void);
2643 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2645 static inline bool page_poisoning_enabled(void) { return false; }
2646 static inline void kernel_poison_pages(struct page *page, int numpages,
2650 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2651 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2653 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2655 static inline bool want_init_on_alloc(gfp_t flags)
2657 if (static_branch_unlikely(&init_on_alloc) &&
2658 !page_poisoning_enabled())
2660 return flags & __GFP_ZERO;
2663 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2664 DECLARE_STATIC_KEY_TRUE(init_on_free);
2666 DECLARE_STATIC_KEY_FALSE(init_on_free);
2668 static inline bool want_init_on_free(void)
2670 return static_branch_unlikely(&init_on_free) &&
2671 !page_poisoning_enabled();
2674 #ifdef CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT
2675 DECLARE_STATIC_KEY_TRUE(_debug_pagealloc_enabled);
2677 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2680 static inline bool debug_pagealloc_enabled(void)
2682 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2685 return static_branch_unlikely(&_debug_pagealloc_enabled);
2688 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2689 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2692 kernel_map_pages(struct page *page, int numpages, int enable)
2694 __kernel_map_pages(page, numpages, enable);
2696 #ifdef CONFIG_HIBERNATION
2697 extern bool kernel_page_present(struct page *page);
2698 #endif /* CONFIG_HIBERNATION */
2699 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2701 kernel_map_pages(struct page *page, int numpages, int enable) {}
2702 #ifdef CONFIG_HIBERNATION
2703 static inline bool kernel_page_present(struct page *page) { return true; }
2704 #endif /* CONFIG_HIBERNATION */
2705 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2707 #ifdef __HAVE_ARCH_GATE_AREA
2708 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2709 extern int in_gate_area_no_mm(unsigned long addr);
2710 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2712 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2716 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2717 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2721 #endif /* __HAVE_ARCH_GATE_AREA */
2723 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2725 #ifdef CONFIG_SYSCTL
2726 extern int sysctl_drop_caches;
2727 int drop_caches_sysctl_handler(struct ctl_table *, int,
2728 void __user *, size_t *, loff_t *);
2731 void drop_slab(void);
2732 void drop_slab_node(int nid);
2735 #define randomize_va_space 0
2737 extern int randomize_va_space;
2740 const char * arch_vma_name(struct vm_area_struct *vma);
2742 void print_vma_addr(char *prefix, unsigned long rip);
2744 static inline void print_vma_addr(char *prefix, unsigned long rip)
2749 void *sparse_buffer_alloc(unsigned long size);
2750 struct page * __populate_section_memmap(unsigned long pfn,
2751 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2752 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2753 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2754 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2755 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2756 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2757 void *vmemmap_alloc_block(unsigned long size, int node);
2759 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2760 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2761 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2762 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2764 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2765 struct vmem_altmap *altmap);
2766 void vmemmap_populate_print_last(void);
2767 #ifdef CONFIG_MEMORY_HOTPLUG
2768 void vmemmap_free(unsigned long start, unsigned long end,
2769 struct vmem_altmap *altmap);
2771 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2772 unsigned long nr_pages);
2775 MF_COUNT_INCREASED = 1 << 0,
2776 MF_ACTION_REQUIRED = 1 << 1,
2777 MF_MUST_KILL = 1 << 2,
2778 MF_SOFT_OFFLINE = 1 << 3,
2780 extern int memory_failure(unsigned long pfn, int flags);
2781 extern void memory_failure_queue(unsigned long pfn, int flags);
2782 extern int unpoison_memory(unsigned long pfn);
2783 extern int get_hwpoison_page(struct page *page);
2784 #define put_hwpoison_page(page) put_page(page)
2785 extern int sysctl_memory_failure_early_kill;
2786 extern int sysctl_memory_failure_recovery;
2787 extern void shake_page(struct page *p, int access);
2788 extern atomic_long_t num_poisoned_pages __read_mostly;
2789 extern int soft_offline_page(struct page *page, int flags);
2793 * Error handlers for various types of pages.
2796 MF_IGNORED, /* Error: cannot be handled */
2797 MF_FAILED, /* Error: handling failed */
2798 MF_DELAYED, /* Will be handled later */
2799 MF_RECOVERED, /* Successfully recovered */
2802 enum mf_action_page_type {
2804 MF_MSG_KERNEL_HIGH_ORDER,
2806 MF_MSG_DIFFERENT_COMPOUND,
2807 MF_MSG_POISONED_HUGE,
2810 MF_MSG_NON_PMD_HUGE,
2811 MF_MSG_UNMAP_FAILED,
2812 MF_MSG_DIRTY_SWAPCACHE,
2813 MF_MSG_CLEAN_SWAPCACHE,
2814 MF_MSG_DIRTY_MLOCKED_LRU,
2815 MF_MSG_CLEAN_MLOCKED_LRU,
2816 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2817 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2820 MF_MSG_TRUNCATED_LRU,
2827 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2828 extern void clear_huge_page(struct page *page,
2829 unsigned long addr_hint,
2830 unsigned int pages_per_huge_page);
2831 extern void copy_user_huge_page(struct page *dst, struct page *src,
2832 unsigned long addr_hint,
2833 struct vm_area_struct *vma,
2834 unsigned int pages_per_huge_page);
2835 extern long copy_huge_page_from_user(struct page *dst_page,
2836 const void __user *usr_src,
2837 unsigned int pages_per_huge_page,
2838 bool allow_pagefault);
2839 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2841 #ifdef CONFIG_DEBUG_PAGEALLOC
2842 extern unsigned int _debug_guardpage_minorder;
2843 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2845 static inline unsigned int debug_guardpage_minorder(void)
2847 return _debug_guardpage_minorder;
2850 static inline bool debug_guardpage_enabled(void)
2852 return static_branch_unlikely(&_debug_guardpage_enabled);
2855 static inline bool page_is_guard(struct page *page)
2857 if (!debug_guardpage_enabled())
2860 return PageGuard(page);
2863 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2864 static inline bool debug_guardpage_enabled(void) { return false; }
2865 static inline bool page_is_guard(struct page *page) { return false; }
2866 #endif /* CONFIG_DEBUG_PAGEALLOC */
2868 #if MAX_NUMNODES > 1
2869 void __init setup_nr_node_ids(void);
2871 static inline void setup_nr_node_ids(void) {}
2874 extern int memcmp_pages(struct page *page1, struct page *page2);
2876 static inline int pages_identical(struct page *page1, struct page *page2)
2878 return !memcmp_pages(page1, page2);
2881 #endif /* __KERNEL__ */
2882 #endif /* _LINUX_MM_H */