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 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
107 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
111 #define lm_alias(x) __va(__pa_symbol(x))
115 * To prevent common memory management code establishing
116 * a zero page mapping on a read fault.
117 * This macro should be defined within <asm/pgtable.h>.
118 * s390 does this to prevent multiplexing of hardware bits
119 * related to the physical page in case of virtualization.
121 #ifndef mm_forbids_zeropage
122 #define mm_forbids_zeropage(X) (0)
126 * On some architectures it is expensive to call memset() for small sizes.
127 * If an architecture decides to implement their own version of
128 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
129 * define their own version of this macro in <asm/pgtable.h>
131 #if BITS_PER_LONG == 64
132 /* This function must be updated when the size of struct page grows above 80
133 * or reduces below 56. The idea that compiler optimizes out switch()
134 * statement, and only leaves move/store instructions. Also the compiler can
135 * combine write statments if they are both assignments and can be reordered,
136 * this can result in several of the writes here being dropped.
138 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
139 static inline void __mm_zero_struct_page(struct page *page)
141 unsigned long *_pp = (void *)page;
143 /* Check that struct page is either 56, 64, 72, or 80 bytes */
144 BUILD_BUG_ON(sizeof(struct page) & 7);
145 BUILD_BUG_ON(sizeof(struct page) < 56);
146 BUILD_BUG_ON(sizeof(struct page) > 80);
148 switch (sizeof(struct page)) {
150 _pp[9] = 0; /* fallthrough */
152 _pp[8] = 0; /* fallthrough */
154 _pp[7] = 0; /* fallthrough */
166 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
170 * Default maximum number of active map areas, this limits the number of vmas
171 * per mm struct. Users can overwrite this number by sysctl but there is a
174 * When a program's coredump is generated as ELF format, a section is created
175 * per a vma. In ELF, the number of sections is represented in unsigned short.
176 * This means the number of sections should be smaller than 65535 at coredump.
177 * Because the kernel adds some informative sections to a image of program at
178 * generating coredump, we need some margin. The number of extra sections is
179 * 1-3 now and depends on arch. We use "5" as safe margin, here.
181 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
182 * not a hard limit any more. Although some userspace tools can be surprised by
185 #define MAPCOUNT_ELF_CORE_MARGIN (5)
186 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
188 extern int sysctl_max_map_count;
190 extern unsigned long sysctl_user_reserve_kbytes;
191 extern unsigned long sysctl_admin_reserve_kbytes;
193 extern int sysctl_overcommit_memory;
194 extern int sysctl_overcommit_ratio;
195 extern unsigned long sysctl_overcommit_kbytes;
197 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
199 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
202 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
204 /* to align the pointer to the (next) page boundary */
205 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
207 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
208 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
210 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
213 * Linux kernel virtual memory manager primitives.
214 * The idea being to have a "virtual" mm in the same way
215 * we have a virtual fs - giving a cleaner interface to the
216 * mm details, and allowing different kinds of memory mappings
217 * (from shared memory to executable loading to arbitrary
221 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
222 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
223 void vm_area_free(struct vm_area_struct *);
226 extern struct rb_root nommu_region_tree;
227 extern struct rw_semaphore nommu_region_sem;
229 extern unsigned int kobjsize(const void *objp);
233 * vm_flags in vm_area_struct, see mm_types.h.
234 * When changing, update also include/trace/events/mmflags.h
236 #define VM_NONE 0x00000000
238 #define VM_READ 0x00000001 /* currently active flags */
239 #define VM_WRITE 0x00000002
240 #define VM_EXEC 0x00000004
241 #define VM_SHARED 0x00000008
243 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
244 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
245 #define VM_MAYWRITE 0x00000020
246 #define VM_MAYEXEC 0x00000040
247 #define VM_MAYSHARE 0x00000080
249 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
250 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
251 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
252 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
253 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
255 #define VM_LOCKED 0x00002000
256 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
258 /* Used by sys_madvise() */
259 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
260 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
262 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
263 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
264 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
265 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
266 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
267 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
268 #define VM_SYNC 0x00800000 /* Synchronous page faults */
269 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
270 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
271 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
273 #ifdef CONFIG_MEM_SOFT_DIRTY
274 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
276 # define VM_SOFTDIRTY 0
279 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
280 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
281 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
282 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
284 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
285 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
286 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
287 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
288 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
289 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
290 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
291 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
292 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
293 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
294 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
295 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
297 #ifdef CONFIG_ARCH_HAS_PKEYS
298 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
299 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
300 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
301 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
302 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
304 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
306 # define VM_PKEY_BIT4 0
308 #endif /* CONFIG_ARCH_HAS_PKEYS */
310 #if defined(CONFIG_X86)
311 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
312 #elif defined(CONFIG_PPC)
313 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
314 #elif defined(CONFIG_PARISC)
315 # define VM_GROWSUP VM_ARCH_1
316 #elif defined(CONFIG_IA64)
317 # define VM_GROWSUP VM_ARCH_1
318 #elif defined(CONFIG_SPARC64)
319 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
320 # define VM_ARCH_CLEAR VM_SPARC_ADI
321 #elif !defined(CONFIG_MMU)
322 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
325 #if defined(CONFIG_X86_INTEL_MPX)
326 /* MPX specific bounds table or bounds directory */
327 # define VM_MPX VM_HIGH_ARCH_4
329 # define VM_MPX VM_NONE
333 # define VM_GROWSUP VM_NONE
336 /* Bits set in the VMA until the stack is in its final location */
337 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
339 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
340 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
343 #ifdef CONFIG_STACK_GROWSUP
344 #define VM_STACK VM_GROWSUP
346 #define VM_STACK VM_GROWSDOWN
349 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
352 * Special vmas that are non-mergable, non-mlock()able.
353 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
355 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
357 /* This mask defines which mm->def_flags a process can inherit its parent */
358 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
360 /* This mask is used to clear all the VMA flags used by mlock */
361 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
363 /* Arch-specific flags to clear when updating VM flags on protection change */
364 #ifndef VM_ARCH_CLEAR
365 # define VM_ARCH_CLEAR VM_NONE
367 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
370 * mapping from the currently active vm_flags protection bits (the
371 * low four bits) to a page protection mask..
373 extern pgprot_t protection_map[16];
375 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
376 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
377 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
378 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
379 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
380 #define FAULT_FLAG_TRIED 0x20 /* Second try */
381 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
382 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
383 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
385 #define FAULT_FLAG_TRACE \
386 { FAULT_FLAG_WRITE, "WRITE" }, \
387 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
388 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
389 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
390 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
391 { FAULT_FLAG_TRIED, "TRIED" }, \
392 { FAULT_FLAG_USER, "USER" }, \
393 { FAULT_FLAG_REMOTE, "REMOTE" }, \
394 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
397 * vm_fault is filled by the the pagefault handler and passed to the vma's
398 * ->fault function. The vma's ->fault is responsible for returning a bitmask
399 * of VM_FAULT_xxx flags that give details about how the fault was handled.
401 * MM layer fills up gfp_mask for page allocations but fault handler might
402 * alter it if its implementation requires a different allocation context.
404 * pgoff should be used in favour of virtual_address, if possible.
407 struct vm_area_struct *vma; /* Target VMA */
408 unsigned int flags; /* FAULT_FLAG_xxx flags */
409 gfp_t gfp_mask; /* gfp mask to be used for allocations */
410 pgoff_t pgoff; /* Logical page offset based on vma */
411 unsigned long address; /* Faulting virtual address */
412 pmd_t *pmd; /* Pointer to pmd entry matching
414 pud_t *pud; /* Pointer to pud entry matching
417 pte_t orig_pte; /* Value of PTE at the time of fault */
419 struct page *cow_page; /* Page handler may use for COW fault */
420 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
421 struct page *page; /* ->fault handlers should return a
422 * page here, unless VM_FAULT_NOPAGE
423 * is set (which is also implied by
426 /* These three entries are valid only while holding ptl lock */
427 pte_t *pte; /* Pointer to pte entry matching
428 * the 'address'. NULL if the page
429 * table hasn't been allocated.
431 spinlock_t *ptl; /* Page table lock.
432 * Protects pte page table if 'pte'
433 * is not NULL, otherwise pmd.
435 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
436 * vm_ops->map_pages() calls
437 * alloc_set_pte() from atomic context.
438 * do_fault_around() pre-allocates
439 * page table to avoid allocation from
444 /* page entry size for vm->huge_fault() */
445 enum page_entry_size {
452 * These are the virtual MM functions - opening of an area, closing and
453 * unmapping it (needed to keep files on disk up-to-date etc), pointer
454 * to the functions called when a no-page or a wp-page exception occurs.
456 struct vm_operations_struct {
457 void (*open)(struct vm_area_struct * area);
458 void (*close)(struct vm_area_struct * area);
459 int (*split)(struct vm_area_struct * area, unsigned long addr);
460 int (*mremap)(struct vm_area_struct * area);
461 vm_fault_t (*fault)(struct vm_fault *vmf);
462 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
463 enum page_entry_size pe_size);
464 void (*map_pages)(struct vm_fault *vmf,
465 pgoff_t start_pgoff, pgoff_t end_pgoff);
466 unsigned long (*pagesize)(struct vm_area_struct * area);
468 /* notification that a previously read-only page is about to become
469 * writable, if an error is returned it will cause a SIGBUS */
470 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
472 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
473 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
475 /* called by access_process_vm when get_user_pages() fails, typically
476 * for use by special VMAs that can switch between memory and hardware
478 int (*access)(struct vm_area_struct *vma, unsigned long addr,
479 void *buf, int len, int write);
481 /* Called by the /proc/PID/maps code to ask the vma whether it
482 * has a special name. Returning non-NULL will also cause this
483 * vma to be dumped unconditionally. */
484 const char *(*name)(struct vm_area_struct *vma);
488 * set_policy() op must add a reference to any non-NULL @new mempolicy
489 * to hold the policy upon return. Caller should pass NULL @new to
490 * remove a policy and fall back to surrounding context--i.e. do not
491 * install a MPOL_DEFAULT policy, nor the task or system default
494 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
497 * get_policy() op must add reference [mpol_get()] to any policy at
498 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
499 * in mm/mempolicy.c will do this automatically.
500 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
501 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
502 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
503 * must return NULL--i.e., do not "fallback" to task or system default
506 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
510 * Called by vm_normal_page() for special PTEs to find the
511 * page for @addr. This is useful if the default behavior
512 * (using pte_page()) would not find the correct page.
514 struct page *(*find_special_page)(struct vm_area_struct *vma,
518 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
520 static const struct vm_operations_struct dummy_vm_ops = {};
522 memset(vma, 0, sizeof(*vma));
524 vma->vm_ops = &dummy_vm_ops;
525 INIT_LIST_HEAD(&vma->anon_vma_chain);
528 static inline void vma_set_anonymous(struct vm_area_struct *vma)
533 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
534 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
539 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
540 static inline int pmd_devmap(pmd_t pmd)
544 static inline int pud_devmap(pud_t pud)
548 static inline int pgd_devmap(pgd_t pgd)
555 * FIXME: take this include out, include page-flags.h in
556 * files which need it (119 of them)
558 #include <linux/page-flags.h>
559 #include <linux/huge_mm.h>
562 * Methods to modify the page usage count.
564 * What counts for a page usage:
565 * - cache mapping (page->mapping)
566 * - private data (page->private)
567 * - page mapped in a task's page tables, each mapping
568 * is counted separately
570 * Also, many kernel routines increase the page count before a critical
571 * routine so they can be sure the page doesn't go away from under them.
575 * Drop a ref, return true if the refcount fell to zero (the page has no users)
577 static inline int put_page_testzero(struct page *page)
579 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
580 return page_ref_dec_and_test(page);
584 * Try to grab a ref unless the page has a refcount of zero, return false if
586 * This can be called when MMU is off so it must not access
587 * any of the virtual mappings.
589 static inline int get_page_unless_zero(struct page *page)
591 return page_ref_add_unless(page, 1, 0);
594 extern int page_is_ram(unsigned long pfn);
602 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
605 /* Support for virtually mapped pages */
606 struct page *vmalloc_to_page(const void *addr);
607 unsigned long vmalloc_to_pfn(const void *addr);
610 * Determine if an address is within the vmalloc range
612 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
613 * is no special casing required.
615 static inline bool is_vmalloc_addr(const void *x)
618 unsigned long addr = (unsigned long)x;
620 return addr >= VMALLOC_START && addr < VMALLOC_END;
626 extern int is_vmalloc_or_module_addr(const void *x);
628 static inline int is_vmalloc_or_module_addr(const void *x)
634 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
635 static inline void *kvmalloc(size_t size, gfp_t flags)
637 return kvmalloc_node(size, flags, NUMA_NO_NODE);
639 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
641 return kvmalloc_node(size, flags | __GFP_ZERO, node);
643 static inline void *kvzalloc(size_t size, gfp_t flags)
645 return kvmalloc(size, flags | __GFP_ZERO);
648 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
652 if (unlikely(check_mul_overflow(n, size, &bytes)))
655 return kvmalloc(bytes, flags);
658 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
660 return kvmalloc_array(n, size, flags | __GFP_ZERO);
663 extern void kvfree(const void *addr);
665 static inline atomic_t *compound_mapcount_ptr(struct page *page)
667 return &page[1].compound_mapcount;
670 static inline int compound_mapcount(struct page *page)
672 VM_BUG_ON_PAGE(!PageCompound(page), page);
673 page = compound_head(page);
674 return atomic_read(compound_mapcount_ptr(page)) + 1;
678 * The atomic page->_mapcount, starts from -1: so that transitions
679 * both from it and to it can be tracked, using atomic_inc_and_test
680 * and atomic_add_negative(-1).
682 static inline void page_mapcount_reset(struct page *page)
684 atomic_set(&(page)->_mapcount, -1);
687 int __page_mapcount(struct page *page);
689 static inline int page_mapcount(struct page *page)
691 VM_BUG_ON_PAGE(PageSlab(page), page);
693 if (unlikely(PageCompound(page)))
694 return __page_mapcount(page);
695 return atomic_read(&page->_mapcount) + 1;
698 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
699 int total_mapcount(struct page *page);
700 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
702 static inline int total_mapcount(struct page *page)
704 return page_mapcount(page);
706 static inline int page_trans_huge_mapcount(struct page *page,
709 int mapcount = page_mapcount(page);
711 *total_mapcount = mapcount;
716 static inline struct page *virt_to_head_page(const void *x)
718 struct page *page = virt_to_page(x);
720 return compound_head(page);
723 void __put_page(struct page *page);
725 void put_pages_list(struct list_head *pages);
727 void split_page(struct page *page, unsigned int order);
730 * Compound pages have a destructor function. Provide a
731 * prototype for that function and accessor functions.
732 * These are _only_ valid on the head of a compound page.
734 typedef void compound_page_dtor(struct page *);
736 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
737 enum compound_dtor_id {
740 #ifdef CONFIG_HUGETLB_PAGE
743 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
748 extern compound_page_dtor * const compound_page_dtors[];
750 static inline void set_compound_page_dtor(struct page *page,
751 enum compound_dtor_id compound_dtor)
753 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
754 page[1].compound_dtor = compound_dtor;
757 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
759 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
760 return compound_page_dtors[page[1].compound_dtor];
763 static inline unsigned int compound_order(struct page *page)
767 return page[1].compound_order;
770 static inline void set_compound_order(struct page *page, unsigned int order)
772 page[1].compound_order = order;
775 void free_compound_page(struct page *page);
779 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
780 * servicing faults for write access. In the normal case, do always want
781 * pte_mkwrite. But get_user_pages can cause write faults for mappings
782 * that do not have writing enabled, when used by access_process_vm.
784 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
786 if (likely(vma->vm_flags & VM_WRITE))
787 pte = pte_mkwrite(pte);
791 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
793 vm_fault_t finish_fault(struct vm_fault *vmf);
794 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
798 * Multiple processes may "see" the same page. E.g. for untouched
799 * mappings of /dev/null, all processes see the same page full of
800 * zeroes, and text pages of executables and shared libraries have
801 * only one copy in memory, at most, normally.
803 * For the non-reserved pages, page_count(page) denotes a reference count.
804 * page_count() == 0 means the page is free. page->lru is then used for
805 * freelist management in the buddy allocator.
806 * page_count() > 0 means the page has been allocated.
808 * Pages are allocated by the slab allocator in order to provide memory
809 * to kmalloc and kmem_cache_alloc. In this case, the management of the
810 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
811 * unless a particular usage is carefully commented. (the responsibility of
812 * freeing the kmalloc memory is the caller's, of course).
814 * A page may be used by anyone else who does a __get_free_page().
815 * In this case, page_count still tracks the references, and should only
816 * be used through the normal accessor functions. The top bits of page->flags
817 * and page->virtual store page management information, but all other fields
818 * are unused and could be used privately, carefully. The management of this
819 * page is the responsibility of the one who allocated it, and those who have
820 * subsequently been given references to it.
822 * The other pages (we may call them "pagecache pages") are completely
823 * managed by the Linux memory manager: I/O, buffers, swapping etc.
824 * The following discussion applies only to them.
826 * A pagecache page contains an opaque `private' member, which belongs to the
827 * page's address_space. Usually, this is the address of a circular list of
828 * the page's disk buffers. PG_private must be set to tell the VM to call
829 * into the filesystem to release these pages.
831 * A page may belong to an inode's memory mapping. In this case, page->mapping
832 * is the pointer to the inode, and page->index is the file offset of the page,
833 * in units of PAGE_SIZE.
835 * If pagecache pages are not associated with an inode, they are said to be
836 * anonymous pages. These may become associated with the swapcache, and in that
837 * case PG_swapcache is set, and page->private is an offset into the swapcache.
839 * In either case (swapcache or inode backed), the pagecache itself holds one
840 * reference to the page. Setting PG_private should also increment the
841 * refcount. The each user mapping also has a reference to the page.
843 * The pagecache pages are stored in a per-mapping radix tree, which is
844 * rooted at mapping->i_pages, and indexed by offset.
845 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
846 * lists, we instead now tag pages as dirty/writeback in the radix tree.
848 * All pagecache pages may be subject to I/O:
849 * - inode pages may need to be read from disk,
850 * - inode pages which have been modified and are MAP_SHARED may need
851 * to be written back to the inode on disk,
852 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
853 * modified may need to be swapped out to swap space and (later) to be read
858 * The zone field is never updated after free_area_init_core()
859 * sets it, so none of the operations on it need to be atomic.
862 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
863 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
864 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
865 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
866 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
867 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
870 * Define the bit shifts to access each section. For non-existent
871 * sections we define the shift as 0; that plus a 0 mask ensures
872 * the compiler will optimise away reference to them.
874 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
875 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
876 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
877 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
878 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
880 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
881 #ifdef NODE_NOT_IN_PAGE_FLAGS
882 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
883 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
884 SECTIONS_PGOFF : ZONES_PGOFF)
886 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
887 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
888 NODES_PGOFF : ZONES_PGOFF)
891 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
893 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
894 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
897 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
898 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
899 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
900 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
901 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
902 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
904 static inline enum zone_type page_zonenum(const struct page *page)
906 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
909 #ifdef CONFIG_ZONE_DEVICE
910 static inline bool is_zone_device_page(const struct page *page)
912 return page_zonenum(page) == ZONE_DEVICE;
914 extern void memmap_init_zone_device(struct zone *, unsigned long,
915 unsigned long, struct dev_pagemap *);
917 static inline bool is_zone_device_page(const struct page *page)
923 #ifdef CONFIG_DEV_PAGEMAP_OPS
924 void dev_pagemap_get_ops(void);
925 void dev_pagemap_put_ops(void);
926 void __put_devmap_managed_page(struct page *page);
927 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
928 static inline bool put_devmap_managed_page(struct page *page)
930 if (!static_branch_unlikely(&devmap_managed_key))
932 if (!is_zone_device_page(page))
934 switch (page->pgmap->type) {
935 case MEMORY_DEVICE_PRIVATE:
936 case MEMORY_DEVICE_PUBLIC:
937 case MEMORY_DEVICE_FS_DAX:
938 __put_devmap_managed_page(page);
946 static inline bool is_device_private_page(const struct page *page)
948 return is_zone_device_page(page) &&
949 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
952 static inline bool is_device_public_page(const struct page *page)
954 return is_zone_device_page(page) &&
955 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
958 #ifdef CONFIG_PCI_P2PDMA
959 static inline bool is_pci_p2pdma_page(const struct page *page)
961 return is_zone_device_page(page) &&
962 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
964 #else /* CONFIG_PCI_P2PDMA */
965 static inline bool is_pci_p2pdma_page(const struct page *page)
969 #endif /* CONFIG_PCI_P2PDMA */
971 #else /* CONFIG_DEV_PAGEMAP_OPS */
972 static inline void dev_pagemap_get_ops(void)
976 static inline void dev_pagemap_put_ops(void)
980 static inline bool put_devmap_managed_page(struct page *page)
985 static inline bool is_device_private_page(const struct page *page)
990 static inline bool is_device_public_page(const struct page *page)
995 static inline bool is_pci_p2pdma_page(const struct page *page)
999 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1001 /* 127: arbitrary random number, small enough to assemble well */
1002 #define page_ref_zero_or_close_to_overflow(page) \
1003 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1005 static inline void get_page(struct page *page)
1007 page = compound_head(page);
1009 * Getting a normal page or the head of a compound page
1010 * requires to already have an elevated page->_refcount.
1012 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1016 static inline __must_check bool try_get_page(struct page *page)
1018 page = compound_head(page);
1019 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1025 static inline void put_page(struct page *page)
1027 page = compound_head(page);
1030 * For devmap managed pages we need to catch refcount transition from
1031 * 2 to 1, when refcount reach one it means the page is free and we
1032 * need to inform the device driver through callback. See
1033 * include/linux/memremap.h and HMM for details.
1035 if (put_devmap_managed_page(page))
1038 if (put_page_testzero(page))
1043 * put_user_page() - release a gup-pinned page
1044 * @page: pointer to page to be released
1046 * Pages that were pinned via get_user_pages*() must be released via
1047 * either put_user_page(), or one of the put_user_pages*() routines
1048 * below. This is so that eventually, pages that are pinned via
1049 * get_user_pages*() can be separately tracked and uniquely handled. In
1050 * particular, interactions with RDMA and filesystems need special
1053 * put_user_page() and put_page() are not interchangeable, despite this early
1054 * implementation that makes them look the same. put_user_page() calls must
1055 * be perfectly matched up with get_user_page() calls.
1057 static inline void put_user_page(struct page *page)
1062 void put_user_pages_dirty(struct page **pages, unsigned long npages);
1063 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages);
1064 void put_user_pages(struct page **pages, unsigned long npages);
1066 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1067 #define SECTION_IN_PAGE_FLAGS
1071 * The identification function is mainly used by the buddy allocator for
1072 * determining if two pages could be buddies. We are not really identifying
1073 * the zone since we could be using the section number id if we do not have
1074 * node id available in page flags.
1075 * We only guarantee that it will return the same value for two combinable
1078 static inline int page_zone_id(struct page *page)
1080 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1083 #ifdef NODE_NOT_IN_PAGE_FLAGS
1084 extern int page_to_nid(const struct page *page);
1086 static inline int page_to_nid(const struct page *page)
1088 struct page *p = (struct page *)page;
1090 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1094 #ifdef CONFIG_NUMA_BALANCING
1095 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1097 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1100 static inline int cpupid_to_pid(int cpupid)
1102 return cpupid & LAST__PID_MASK;
1105 static inline int cpupid_to_cpu(int cpupid)
1107 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1110 static inline int cpupid_to_nid(int cpupid)
1112 return cpu_to_node(cpupid_to_cpu(cpupid));
1115 static inline bool cpupid_pid_unset(int cpupid)
1117 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1120 static inline bool cpupid_cpu_unset(int cpupid)
1122 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1125 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1127 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1130 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1131 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1132 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1134 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1137 static inline int page_cpupid_last(struct page *page)
1139 return page->_last_cpupid;
1141 static inline void page_cpupid_reset_last(struct page *page)
1143 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1146 static inline int page_cpupid_last(struct page *page)
1148 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1151 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1153 static inline void page_cpupid_reset_last(struct page *page)
1155 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1157 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1158 #else /* !CONFIG_NUMA_BALANCING */
1159 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1161 return page_to_nid(page); /* XXX */
1164 static inline int page_cpupid_last(struct page *page)
1166 return page_to_nid(page); /* XXX */
1169 static inline int cpupid_to_nid(int cpupid)
1174 static inline int cpupid_to_pid(int cpupid)
1179 static inline int cpupid_to_cpu(int cpupid)
1184 static inline int cpu_pid_to_cpupid(int nid, int pid)
1189 static inline bool cpupid_pid_unset(int cpupid)
1194 static inline void page_cpupid_reset_last(struct page *page)
1198 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1202 #endif /* CONFIG_NUMA_BALANCING */
1204 #ifdef CONFIG_KASAN_SW_TAGS
1205 static inline u8 page_kasan_tag(const struct page *page)
1207 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1210 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1212 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1213 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1216 static inline void page_kasan_tag_reset(struct page *page)
1218 page_kasan_tag_set(page, 0xff);
1221 static inline u8 page_kasan_tag(const struct page *page)
1226 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1227 static inline void page_kasan_tag_reset(struct page *page) { }
1230 static inline struct zone *page_zone(const struct page *page)
1232 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1235 static inline pg_data_t *page_pgdat(const struct page *page)
1237 return NODE_DATA(page_to_nid(page));
1240 #ifdef SECTION_IN_PAGE_FLAGS
1241 static inline void set_page_section(struct page *page, unsigned long section)
1243 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1244 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1247 static inline unsigned long page_to_section(const struct page *page)
1249 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1253 static inline void set_page_zone(struct page *page, enum zone_type zone)
1255 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1256 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1259 static inline void set_page_node(struct page *page, unsigned long node)
1261 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1262 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1265 static inline void set_page_links(struct page *page, enum zone_type zone,
1266 unsigned long node, unsigned long pfn)
1268 set_page_zone(page, zone);
1269 set_page_node(page, node);
1270 #ifdef SECTION_IN_PAGE_FLAGS
1271 set_page_section(page, pfn_to_section_nr(pfn));
1276 static inline struct mem_cgroup *page_memcg(struct page *page)
1278 return page->mem_cgroup;
1280 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1282 WARN_ON_ONCE(!rcu_read_lock_held());
1283 return READ_ONCE(page->mem_cgroup);
1286 static inline struct mem_cgroup *page_memcg(struct page *page)
1290 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1292 WARN_ON_ONCE(!rcu_read_lock_held());
1298 * Some inline functions in vmstat.h depend on page_zone()
1300 #include <linux/vmstat.h>
1302 static __always_inline void *lowmem_page_address(const struct page *page)
1304 return page_to_virt(page);
1307 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1308 #define HASHED_PAGE_VIRTUAL
1311 #if defined(WANT_PAGE_VIRTUAL)
1312 static inline void *page_address(const struct page *page)
1314 return page->virtual;
1316 static inline void set_page_address(struct page *page, void *address)
1318 page->virtual = address;
1320 #define page_address_init() do { } while(0)
1323 #if defined(HASHED_PAGE_VIRTUAL)
1324 void *page_address(const struct page *page);
1325 void set_page_address(struct page *page, void *virtual);
1326 void page_address_init(void);
1329 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1330 #define page_address(page) lowmem_page_address(page)
1331 #define set_page_address(page, address) do { } while(0)
1332 #define page_address_init() do { } while(0)
1335 extern void *page_rmapping(struct page *page);
1336 extern struct anon_vma *page_anon_vma(struct page *page);
1337 extern struct address_space *page_mapping(struct page *page);
1339 extern struct address_space *__page_file_mapping(struct page *);
1342 struct address_space *page_file_mapping(struct page *page)
1344 if (unlikely(PageSwapCache(page)))
1345 return __page_file_mapping(page);
1347 return page->mapping;
1350 extern pgoff_t __page_file_index(struct page *page);
1353 * Return the pagecache index of the passed page. Regular pagecache pages
1354 * use ->index whereas swapcache pages use swp_offset(->private)
1356 static inline pgoff_t page_index(struct page *page)
1358 if (unlikely(PageSwapCache(page)))
1359 return __page_file_index(page);
1363 bool page_mapped(struct page *page);
1364 struct address_space *page_mapping(struct page *page);
1365 struct address_space *page_mapping_file(struct page *page);
1368 * Return true only if the page has been allocated with
1369 * ALLOC_NO_WATERMARKS and the low watermark was not
1370 * met implying that the system is under some pressure.
1372 static inline bool page_is_pfmemalloc(struct page *page)
1375 * Page index cannot be this large so this must be
1376 * a pfmemalloc page.
1378 return page->index == -1UL;
1382 * Only to be called by the page allocator on a freshly allocated
1385 static inline void set_page_pfmemalloc(struct page *page)
1390 static inline void clear_page_pfmemalloc(struct page *page)
1396 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1398 extern void pagefault_out_of_memory(void);
1400 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1403 * Flags passed to show_mem() and show_free_areas() to suppress output in
1406 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1408 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1410 extern bool can_do_mlock(void);
1411 extern int user_shm_lock(size_t, struct user_struct *);
1412 extern void user_shm_unlock(size_t, struct user_struct *);
1415 * Parameter block passed down to zap_pte_range in exceptional cases.
1417 struct zap_details {
1418 struct address_space *check_mapping; /* Check page->mapping if set */
1419 pgoff_t first_index; /* Lowest page->index to unmap */
1420 pgoff_t last_index; /* Highest page->index to unmap */
1423 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1424 pte_t pte, bool with_public_device);
1425 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1427 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1430 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1431 unsigned long size);
1432 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1433 unsigned long size);
1434 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1435 unsigned long start, unsigned long end);
1438 * mm_walk - callbacks for walk_page_range
1439 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1440 * this handler should only handle pud_trans_huge() puds.
1441 * the pmd_entry or pte_entry callbacks will be used for
1443 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1444 * this handler is required to be able to handle
1445 * pmd_trans_huge() pmds. They may simply choose to
1446 * split_huge_page() instead of handling it explicitly.
1447 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1448 * @pte_hole: if set, called for each hole at all levels
1449 * @hugetlb_entry: if set, called for each hugetlb entry
1450 * @test_walk: caller specific callback function to determine whether
1451 * we walk over the current vma or not. Returning 0
1452 * value means "do page table walk over the current vma,"
1453 * and a negative one means "abort current page table walk
1454 * right now." 1 means "skip the current vma."
1455 * @mm: mm_struct representing the target process of page table walk
1456 * @vma: vma currently walked (NULL if walking outside vmas)
1457 * @private: private data for callbacks' usage
1459 * (see the comment on walk_page_range() for more details)
1462 int (*pud_entry)(pud_t *pud, unsigned long addr,
1463 unsigned long next, struct mm_walk *walk);
1464 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1465 unsigned long next, struct mm_walk *walk);
1466 int (*pte_entry)(pte_t *pte, unsigned long addr,
1467 unsigned long next, struct mm_walk *walk);
1468 int (*pte_hole)(unsigned long addr, unsigned long next,
1469 struct mm_walk *walk);
1470 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1471 unsigned long addr, unsigned long next,
1472 struct mm_walk *walk);
1473 int (*test_walk)(unsigned long addr, unsigned long next,
1474 struct mm_walk *walk);
1475 struct mm_struct *mm;
1476 struct vm_area_struct *vma;
1480 struct mmu_notifier_range;
1482 int walk_page_range(unsigned long addr, unsigned long end,
1483 struct mm_walk *walk);
1484 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1485 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1486 unsigned long end, unsigned long floor, unsigned long ceiling);
1487 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1488 struct vm_area_struct *vma);
1489 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1490 struct mmu_notifier_range *range,
1491 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1492 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1493 unsigned long *pfn);
1494 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1495 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1496 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1497 void *buf, int len, int write);
1499 extern void truncate_pagecache(struct inode *inode, loff_t new);
1500 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1501 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1502 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1503 int truncate_inode_page(struct address_space *mapping, struct page *page);
1504 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1505 int invalidate_inode_page(struct page *page);
1508 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1509 unsigned long address, unsigned int flags);
1510 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1511 unsigned long address, unsigned int fault_flags,
1513 void unmap_mapping_pages(struct address_space *mapping,
1514 pgoff_t start, pgoff_t nr, bool even_cows);
1515 void unmap_mapping_range(struct address_space *mapping,
1516 loff_t const holebegin, loff_t const holelen, int even_cows);
1518 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1519 unsigned long address, unsigned int flags)
1521 /* should never happen if there's no MMU */
1523 return VM_FAULT_SIGBUS;
1525 static inline int fixup_user_fault(struct task_struct *tsk,
1526 struct mm_struct *mm, unsigned long address,
1527 unsigned int fault_flags, bool *unlocked)
1529 /* should never happen if there's no MMU */
1533 static inline void unmap_mapping_pages(struct address_space *mapping,
1534 pgoff_t start, pgoff_t nr, bool even_cows) { }
1535 static inline void unmap_mapping_range(struct address_space *mapping,
1536 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1539 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1540 loff_t const holebegin, loff_t const holelen)
1542 unmap_mapping_range(mapping, holebegin, holelen, 0);
1545 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1546 void *buf, int len, unsigned int gup_flags);
1547 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1548 void *buf, int len, unsigned int gup_flags);
1549 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1550 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1552 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1553 unsigned long start, unsigned long nr_pages,
1554 unsigned int gup_flags, struct page **pages,
1555 struct vm_area_struct **vmas, int *locked);
1556 long get_user_pages(unsigned long start, unsigned long nr_pages,
1557 unsigned int gup_flags, struct page **pages,
1558 struct vm_area_struct **vmas);
1559 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1560 unsigned int gup_flags, struct page **pages, int *locked);
1561 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1562 struct page **pages, unsigned int gup_flags);
1564 int get_user_pages_fast(unsigned long start, int nr_pages,
1565 unsigned int gup_flags, struct page **pages);
1567 /* Container for pinned pfns / pages */
1568 struct frame_vector {
1569 unsigned int nr_allocated; /* Number of frames we have space for */
1570 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1571 bool got_ref; /* Did we pin pages by getting page ref? */
1572 bool is_pfns; /* Does array contain pages or pfns? */
1573 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1574 * pfns_vector_pages() or pfns_vector_pfns()
1578 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1579 void frame_vector_destroy(struct frame_vector *vec);
1580 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1581 unsigned int gup_flags, struct frame_vector *vec);
1582 void put_vaddr_frames(struct frame_vector *vec);
1583 int frame_vector_to_pages(struct frame_vector *vec);
1584 void frame_vector_to_pfns(struct frame_vector *vec);
1586 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1588 return vec->nr_frames;
1591 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1594 int err = frame_vector_to_pages(vec);
1597 return ERR_PTR(err);
1599 return (struct page **)(vec->ptrs);
1602 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1605 frame_vector_to_pfns(vec);
1606 return (unsigned long *)(vec->ptrs);
1610 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1611 struct page **pages);
1612 int get_kernel_page(unsigned long start, int write, struct page **pages);
1613 struct page *get_dump_page(unsigned long addr);
1615 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1616 extern void do_invalidatepage(struct page *page, unsigned int offset,
1617 unsigned int length);
1619 void __set_page_dirty(struct page *, struct address_space *, int warn);
1620 int __set_page_dirty_nobuffers(struct page *page);
1621 int __set_page_dirty_no_writeback(struct page *page);
1622 int redirty_page_for_writepage(struct writeback_control *wbc,
1624 void account_page_dirtied(struct page *page, struct address_space *mapping);
1625 void account_page_cleaned(struct page *page, struct address_space *mapping,
1626 struct bdi_writeback *wb);
1627 int set_page_dirty(struct page *page);
1628 int set_page_dirty_lock(struct page *page);
1629 void __cancel_dirty_page(struct page *page);
1630 static inline void cancel_dirty_page(struct page *page)
1632 /* Avoid atomic ops, locking, etc. when not actually needed. */
1633 if (PageDirty(page))
1634 __cancel_dirty_page(page);
1636 int clear_page_dirty_for_io(struct page *page);
1638 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1640 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1642 return !vma->vm_ops;
1647 * The vma_is_shmem is not inline because it is used only by slow
1648 * paths in userfault.
1650 bool vma_is_shmem(struct vm_area_struct *vma);
1652 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1655 int vma_is_stack_for_current(struct vm_area_struct *vma);
1657 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1658 unsigned long old_addr, struct vm_area_struct *new_vma,
1659 unsigned long new_addr, unsigned long len,
1660 bool need_rmap_locks);
1661 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1662 unsigned long end, pgprot_t newprot,
1663 int dirty_accountable, int prot_numa);
1664 extern int mprotect_fixup(struct vm_area_struct *vma,
1665 struct vm_area_struct **pprev, unsigned long start,
1666 unsigned long end, unsigned long newflags);
1669 * doesn't attempt to fault and will return short.
1671 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1672 struct page **pages);
1674 * per-process(per-mm_struct) statistics.
1676 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1678 long val = atomic_long_read(&mm->rss_stat.count[member]);
1680 #ifdef SPLIT_RSS_COUNTING
1682 * counter is updated in asynchronous manner and may go to minus.
1683 * But it's never be expected number for users.
1688 return (unsigned long)val;
1691 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1693 atomic_long_add(value, &mm->rss_stat.count[member]);
1696 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1698 atomic_long_inc(&mm->rss_stat.count[member]);
1701 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1703 atomic_long_dec(&mm->rss_stat.count[member]);
1706 /* Optimized variant when page is already known not to be PageAnon */
1707 static inline int mm_counter_file(struct page *page)
1709 if (PageSwapBacked(page))
1710 return MM_SHMEMPAGES;
1711 return MM_FILEPAGES;
1714 static inline int mm_counter(struct page *page)
1717 return MM_ANONPAGES;
1718 return mm_counter_file(page);
1721 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1723 return get_mm_counter(mm, MM_FILEPAGES) +
1724 get_mm_counter(mm, MM_ANONPAGES) +
1725 get_mm_counter(mm, MM_SHMEMPAGES);
1728 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1730 return max(mm->hiwater_rss, get_mm_rss(mm));
1733 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1735 return max(mm->hiwater_vm, mm->total_vm);
1738 static inline void update_hiwater_rss(struct mm_struct *mm)
1740 unsigned long _rss = get_mm_rss(mm);
1742 if ((mm)->hiwater_rss < _rss)
1743 (mm)->hiwater_rss = _rss;
1746 static inline void update_hiwater_vm(struct mm_struct *mm)
1748 if (mm->hiwater_vm < mm->total_vm)
1749 mm->hiwater_vm = mm->total_vm;
1752 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1754 mm->hiwater_rss = get_mm_rss(mm);
1757 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1758 struct mm_struct *mm)
1760 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1762 if (*maxrss < hiwater_rss)
1763 *maxrss = hiwater_rss;
1766 #if defined(SPLIT_RSS_COUNTING)
1767 void sync_mm_rss(struct mm_struct *mm);
1769 static inline void sync_mm_rss(struct mm_struct *mm)
1774 #ifndef __HAVE_ARCH_PTE_DEVMAP
1775 static inline int pte_devmap(pte_t pte)
1781 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1783 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1785 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1789 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1793 #ifdef __PAGETABLE_P4D_FOLDED
1794 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1795 unsigned long address)
1800 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1803 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1804 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1805 unsigned long address)
1809 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1810 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1813 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1815 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1817 if (mm_pud_folded(mm))
1819 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1822 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1824 if (mm_pud_folded(mm))
1826 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1830 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1831 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1832 unsigned long address)
1837 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1838 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1841 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1843 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1845 if (mm_pmd_folded(mm))
1847 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1850 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1852 if (mm_pmd_folded(mm))
1854 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1859 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1861 atomic_long_set(&mm->pgtables_bytes, 0);
1864 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1866 return atomic_long_read(&mm->pgtables_bytes);
1869 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1871 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1874 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1876 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1880 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1881 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1886 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1887 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1890 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1891 int __pte_alloc_kernel(pmd_t *pmd);
1894 * The following ifdef needed to get the 4level-fixup.h header to work.
1895 * Remove it when 4level-fixup.h has been removed.
1897 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1899 #ifndef __ARCH_HAS_5LEVEL_HACK
1900 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1901 unsigned long address)
1903 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1904 NULL : p4d_offset(pgd, address);
1907 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1908 unsigned long address)
1910 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1911 NULL : pud_offset(p4d, address);
1913 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1915 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1917 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1918 NULL: pmd_offset(pud, address);
1920 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1922 #if USE_SPLIT_PTE_PTLOCKS
1923 #if ALLOC_SPLIT_PTLOCKS
1924 void __init ptlock_cache_init(void);
1925 extern bool ptlock_alloc(struct page *page);
1926 extern void ptlock_free(struct page *page);
1928 static inline spinlock_t *ptlock_ptr(struct page *page)
1932 #else /* ALLOC_SPLIT_PTLOCKS */
1933 static inline void ptlock_cache_init(void)
1937 static inline bool ptlock_alloc(struct page *page)
1942 static inline void ptlock_free(struct page *page)
1946 static inline spinlock_t *ptlock_ptr(struct page *page)
1950 #endif /* ALLOC_SPLIT_PTLOCKS */
1952 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1954 return ptlock_ptr(pmd_page(*pmd));
1957 static inline bool ptlock_init(struct page *page)
1960 * prep_new_page() initialize page->private (and therefore page->ptl)
1961 * with 0. Make sure nobody took it in use in between.
1963 * It can happen if arch try to use slab for page table allocation:
1964 * slab code uses page->slab_cache, which share storage with page->ptl.
1966 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1967 if (!ptlock_alloc(page))
1969 spin_lock_init(ptlock_ptr(page));
1973 #else /* !USE_SPLIT_PTE_PTLOCKS */
1975 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1977 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1979 return &mm->page_table_lock;
1981 static inline void ptlock_cache_init(void) {}
1982 static inline bool ptlock_init(struct page *page) { return true; }
1983 static inline void ptlock_free(struct page *page) {}
1984 #endif /* USE_SPLIT_PTE_PTLOCKS */
1986 static inline void pgtable_init(void)
1988 ptlock_cache_init();
1989 pgtable_cache_init();
1992 static inline bool pgtable_page_ctor(struct page *page)
1994 if (!ptlock_init(page))
1996 __SetPageTable(page);
1997 inc_zone_page_state(page, NR_PAGETABLE);
2001 static inline void pgtable_page_dtor(struct page *page)
2004 __ClearPageTable(page);
2005 dec_zone_page_state(page, NR_PAGETABLE);
2008 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2010 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2011 pte_t *__pte = pte_offset_map(pmd, address); \
2017 #define pte_unmap_unlock(pte, ptl) do { \
2022 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2024 #define pte_alloc_map(mm, pmd, address) \
2025 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2027 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2028 (pte_alloc(mm, pmd) ? \
2029 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2031 #define pte_alloc_kernel(pmd, address) \
2032 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2033 NULL: pte_offset_kernel(pmd, address))
2035 #if USE_SPLIT_PMD_PTLOCKS
2037 static struct page *pmd_to_page(pmd_t *pmd)
2039 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2040 return virt_to_page((void *)((unsigned long) pmd & mask));
2043 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2045 return ptlock_ptr(pmd_to_page(pmd));
2048 static inline bool pgtable_pmd_page_ctor(struct page *page)
2050 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2051 page->pmd_huge_pte = NULL;
2053 return ptlock_init(page);
2056 static inline void pgtable_pmd_page_dtor(struct page *page)
2058 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2059 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2064 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2068 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2070 return &mm->page_table_lock;
2073 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2074 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2076 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2080 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2082 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2088 * No scalability reason to split PUD locks yet, but follow the same pattern
2089 * as the PMD locks to make it easier if we decide to. The VM should not be
2090 * considered ready to switch to split PUD locks yet; there may be places
2091 * which need to be converted from page_table_lock.
2093 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2095 return &mm->page_table_lock;
2098 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2100 spinlock_t *ptl = pud_lockptr(mm, pud);
2106 extern void __init pagecache_init(void);
2107 extern void free_area_init(unsigned long * zones_size);
2108 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2109 unsigned long zone_start_pfn, unsigned long *zholes_size);
2110 extern void free_initmem(void);
2113 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2114 * into the buddy system. The freed pages will be poisoned with pattern
2115 * "poison" if it's within range [0, UCHAR_MAX].
2116 * Return pages freed into the buddy system.
2118 extern unsigned long free_reserved_area(void *start, void *end,
2119 int poison, const char *s);
2121 #ifdef CONFIG_HIGHMEM
2123 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2124 * and totalram_pages.
2126 extern void free_highmem_page(struct page *page);
2129 extern void adjust_managed_page_count(struct page *page, long count);
2130 extern void mem_init_print_info(const char *str);
2132 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2134 /* Free the reserved page into the buddy system, so it gets managed. */
2135 static inline void __free_reserved_page(struct page *page)
2137 ClearPageReserved(page);
2138 init_page_count(page);
2142 static inline void free_reserved_page(struct page *page)
2144 __free_reserved_page(page);
2145 adjust_managed_page_count(page, 1);
2148 static inline void mark_page_reserved(struct page *page)
2150 SetPageReserved(page);
2151 adjust_managed_page_count(page, -1);
2155 * Default method to free all the __init memory into the buddy system.
2156 * The freed pages will be poisoned with pattern "poison" if it's within
2157 * range [0, UCHAR_MAX].
2158 * Return pages freed into the buddy system.
2160 static inline unsigned long free_initmem_default(int poison)
2162 extern char __init_begin[], __init_end[];
2164 return free_reserved_area(&__init_begin, &__init_end,
2165 poison, "unused kernel");
2168 static inline unsigned long get_num_physpages(void)
2171 unsigned long phys_pages = 0;
2173 for_each_online_node(nid)
2174 phys_pages += node_present_pages(nid);
2179 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2181 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2182 * zones, allocate the backing mem_map and account for memory holes in a more
2183 * architecture independent manner. This is a substitute for creating the
2184 * zone_sizes[] and zholes_size[] arrays and passing them to
2185 * free_area_init_node()
2187 * An architecture is expected to register range of page frames backed by
2188 * physical memory with memblock_add[_node]() before calling
2189 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2190 * usage, an architecture is expected to do something like
2192 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2194 * for_each_valid_physical_page_range()
2195 * memblock_add_node(base, size, nid)
2196 * free_area_init_nodes(max_zone_pfns);
2198 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2199 * registered physical page range. Similarly
2200 * sparse_memory_present_with_active_regions() calls memory_present() for
2201 * each range when SPARSEMEM is enabled.
2203 * See mm/page_alloc.c for more information on each function exposed by
2204 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2206 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2207 unsigned long node_map_pfn_alignment(void);
2208 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2209 unsigned long end_pfn);
2210 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2211 unsigned long end_pfn);
2212 extern void get_pfn_range_for_nid(unsigned int nid,
2213 unsigned long *start_pfn, unsigned long *end_pfn);
2214 extern unsigned long find_min_pfn_with_active_regions(void);
2215 extern void free_bootmem_with_active_regions(int nid,
2216 unsigned long max_low_pfn);
2217 extern void sparse_memory_present_with_active_regions(int nid);
2219 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2221 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2222 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2223 static inline int __early_pfn_to_nid(unsigned long pfn,
2224 struct mminit_pfnnid_cache *state)
2229 /* please see mm/page_alloc.c */
2230 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2231 /* there is a per-arch backend function. */
2232 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2233 struct mminit_pfnnid_cache *state);
2236 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2237 void zero_resv_unavail(void);
2239 static inline void zero_resv_unavail(void) {}
2242 extern void set_dma_reserve(unsigned long new_dma_reserve);
2243 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2244 enum memmap_context, struct vmem_altmap *);
2245 extern void setup_per_zone_wmarks(void);
2246 extern int __meminit init_per_zone_wmark_min(void);
2247 extern void mem_init(void);
2248 extern void __init mmap_init(void);
2249 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2250 extern long si_mem_available(void);
2251 extern void si_meminfo(struct sysinfo * val);
2252 extern void si_meminfo_node(struct sysinfo *val, int nid);
2253 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2254 extern unsigned long arch_reserved_kernel_pages(void);
2257 extern __printf(3, 4)
2258 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2260 extern void setup_per_cpu_pageset(void);
2262 extern void zone_pcp_update(struct zone *zone);
2263 extern void zone_pcp_reset(struct zone *zone);
2266 extern int min_free_kbytes;
2267 extern int watermark_boost_factor;
2268 extern int watermark_scale_factor;
2271 extern atomic_long_t mmap_pages_allocated;
2272 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2274 /* interval_tree.c */
2275 void vma_interval_tree_insert(struct vm_area_struct *node,
2276 struct rb_root_cached *root);
2277 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2278 struct vm_area_struct *prev,
2279 struct rb_root_cached *root);
2280 void vma_interval_tree_remove(struct vm_area_struct *node,
2281 struct rb_root_cached *root);
2282 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2283 unsigned long start, unsigned long last);
2284 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2285 unsigned long start, unsigned long last);
2287 #define vma_interval_tree_foreach(vma, root, start, last) \
2288 for (vma = vma_interval_tree_iter_first(root, start, last); \
2289 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2291 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2292 struct rb_root_cached *root);
2293 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2294 struct rb_root_cached *root);
2295 struct anon_vma_chain *
2296 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2297 unsigned long start, unsigned long last);
2298 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2299 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2300 #ifdef CONFIG_DEBUG_VM_RB
2301 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2304 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2305 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2306 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2309 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2310 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2311 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2312 struct vm_area_struct *expand);
2313 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2314 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2316 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2318 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2319 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2320 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2321 struct mempolicy *, struct vm_userfaultfd_ctx);
2322 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2323 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2324 unsigned long addr, int new_below);
2325 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2326 unsigned long addr, int new_below);
2327 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2328 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2329 struct rb_node **, struct rb_node *);
2330 extern void unlink_file_vma(struct vm_area_struct *);
2331 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2332 unsigned long addr, unsigned long len, pgoff_t pgoff,
2333 bool *need_rmap_locks);
2334 extern void exit_mmap(struct mm_struct *);
2336 static inline int check_data_rlimit(unsigned long rlim,
2338 unsigned long start,
2339 unsigned long end_data,
2340 unsigned long start_data)
2342 if (rlim < RLIM_INFINITY) {
2343 if (((new - start) + (end_data - start_data)) > rlim)
2350 extern int mm_take_all_locks(struct mm_struct *mm);
2351 extern void mm_drop_all_locks(struct mm_struct *mm);
2353 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2354 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2355 extern struct file *get_task_exe_file(struct task_struct *task);
2357 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2358 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2360 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2361 const struct vm_special_mapping *sm);
2362 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2363 unsigned long addr, unsigned long len,
2364 unsigned long flags,
2365 const struct vm_special_mapping *spec);
2366 /* This is an obsolete alternative to _install_special_mapping. */
2367 extern int install_special_mapping(struct mm_struct *mm,
2368 unsigned long addr, unsigned long len,
2369 unsigned long flags, struct page **pages);
2371 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2373 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2374 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2375 struct list_head *uf);
2376 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2377 unsigned long len, unsigned long prot, unsigned long flags,
2378 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2379 struct list_head *uf);
2380 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2381 struct list_head *uf, bool downgrade);
2382 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2383 struct list_head *uf);
2385 static inline unsigned long
2386 do_mmap_pgoff(struct file *file, unsigned long addr,
2387 unsigned long len, unsigned long prot, unsigned long flags,
2388 unsigned long pgoff, unsigned long *populate,
2389 struct list_head *uf)
2391 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2395 extern int __mm_populate(unsigned long addr, unsigned long len,
2397 static inline void mm_populate(unsigned long addr, unsigned long len)
2400 (void) __mm_populate(addr, len, 1);
2403 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2406 /* These take the mm semaphore themselves */
2407 extern int __must_check vm_brk(unsigned long, unsigned long);
2408 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2409 extern int vm_munmap(unsigned long, size_t);
2410 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2411 unsigned long, unsigned long,
2412 unsigned long, unsigned long);
2414 struct vm_unmapped_area_info {
2415 #define VM_UNMAPPED_AREA_TOPDOWN 1
2416 unsigned long flags;
2417 unsigned long length;
2418 unsigned long low_limit;
2419 unsigned long high_limit;
2420 unsigned long align_mask;
2421 unsigned long align_offset;
2424 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2425 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2428 * Search for an unmapped address range.
2430 * We are looking for a range that:
2431 * - does not intersect with any VMA;
2432 * - is contained within the [low_limit, high_limit) interval;
2433 * - is at least the desired size.
2434 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2436 static inline unsigned long
2437 vm_unmapped_area(struct vm_unmapped_area_info *info)
2439 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2440 return unmapped_area_topdown(info);
2442 return unmapped_area(info);
2446 extern void truncate_inode_pages(struct address_space *, loff_t);
2447 extern void truncate_inode_pages_range(struct address_space *,
2448 loff_t lstart, loff_t lend);
2449 extern void truncate_inode_pages_final(struct address_space *);
2451 /* generic vm_area_ops exported for stackable file systems */
2452 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2453 extern void filemap_map_pages(struct vm_fault *vmf,
2454 pgoff_t start_pgoff, pgoff_t end_pgoff);
2455 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2457 /* mm/page-writeback.c */
2458 int __must_check write_one_page(struct page *page);
2459 void task_dirty_inc(struct task_struct *tsk);
2462 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2464 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2465 pgoff_t offset, unsigned long nr_to_read);
2467 void page_cache_sync_readahead(struct address_space *mapping,
2468 struct file_ra_state *ra,
2471 unsigned long size);
2473 void page_cache_async_readahead(struct address_space *mapping,
2474 struct file_ra_state *ra,
2478 unsigned long size);
2480 extern unsigned long stack_guard_gap;
2481 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2482 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2484 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2485 extern int expand_downwards(struct vm_area_struct *vma,
2486 unsigned long address);
2488 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2490 #define expand_upwards(vma, address) (0)
2493 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2494 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2495 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2496 struct vm_area_struct **pprev);
2498 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2499 NULL if none. Assume start_addr < end_addr. */
2500 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2502 struct vm_area_struct * vma = find_vma(mm,start_addr);
2504 if (vma && end_addr <= vma->vm_start)
2509 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2511 unsigned long vm_start = vma->vm_start;
2513 if (vma->vm_flags & VM_GROWSDOWN) {
2514 vm_start -= stack_guard_gap;
2515 if (vm_start > vma->vm_start)
2521 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2523 unsigned long vm_end = vma->vm_end;
2525 if (vma->vm_flags & VM_GROWSUP) {
2526 vm_end += stack_guard_gap;
2527 if (vm_end < vma->vm_end)
2528 vm_end = -PAGE_SIZE;
2533 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2535 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2538 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2539 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2540 unsigned long vm_start, unsigned long vm_end)
2542 struct vm_area_struct *vma = find_vma(mm, vm_start);
2544 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2550 static inline bool range_in_vma(struct vm_area_struct *vma,
2551 unsigned long start, unsigned long end)
2553 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2557 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2558 void vma_set_page_prot(struct vm_area_struct *vma);
2560 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2564 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2566 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2570 #ifdef CONFIG_NUMA_BALANCING
2571 unsigned long change_prot_numa(struct vm_area_struct *vma,
2572 unsigned long start, unsigned long end);
2575 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2576 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2577 unsigned long pfn, unsigned long size, pgprot_t);
2578 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2579 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2581 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2583 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2585 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2586 unsigned long pfn, pgprot_t pgprot);
2587 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2589 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2590 unsigned long addr, pfn_t pfn);
2591 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2593 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2594 unsigned long addr, struct page *page)
2596 int err = vm_insert_page(vma, addr, page);
2599 return VM_FAULT_OOM;
2600 if (err < 0 && err != -EBUSY)
2601 return VM_FAULT_SIGBUS;
2603 return VM_FAULT_NOPAGE;
2606 static inline vm_fault_t vmf_error(int err)
2609 return VM_FAULT_OOM;
2610 return VM_FAULT_SIGBUS;
2613 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2614 unsigned int foll_flags);
2616 #define FOLL_WRITE 0x01 /* check pte is writable */
2617 #define FOLL_TOUCH 0x02 /* mark page accessed */
2618 #define FOLL_GET 0x04 /* do get_page on page */
2619 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2620 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2621 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2622 * and return without waiting upon it */
2623 #define FOLL_POPULATE 0x40 /* fault in page */
2624 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2625 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2626 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2627 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2628 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2629 #define FOLL_MLOCK 0x1000 /* lock present pages */
2630 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2631 #define FOLL_COW 0x4000 /* internal GUP flag */
2632 #define FOLL_ANON 0x8000 /* don't do file mappings */
2633 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2636 * NOTE on FOLL_LONGTERM:
2638 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2639 * period _often_ under userspace control. This is contrasted with
2640 * iov_iter_get_pages() where usages which are transient.
2642 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2643 * lifetime enforced by the filesystem and we need guarantees that longterm
2644 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2645 * the filesystem. Ideas for this coordination include revoking the longterm
2646 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2647 * added after the problem with filesystems was found FS DAX VMAs are
2648 * specifically failed. Filesystem pages are still subject to bugs and use of
2649 * FOLL_LONGTERM should be avoided on those pages.
2651 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2652 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2653 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2654 * is due to an incompatibility with the FS DAX check and
2655 * FAULT_FLAG_ALLOW_RETRY
2657 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2658 * that region. And so CMA attempts to migrate the page before pinning when
2659 * FOLL_LONGTERM is specified.
2662 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2664 if (vm_fault & VM_FAULT_OOM)
2666 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2667 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2668 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2673 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2675 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2676 unsigned long size, pte_fn_t fn, void *data);
2679 #ifdef CONFIG_PAGE_POISONING
2680 extern bool page_poisoning_enabled(void);
2681 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2683 static inline bool page_poisoning_enabled(void) { return false; }
2684 static inline void kernel_poison_pages(struct page *page, int numpages,
2688 extern bool _debug_pagealloc_enabled;
2690 static inline bool debug_pagealloc_enabled(void)
2692 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) && _debug_pagealloc_enabled;
2695 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2696 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2699 kernel_map_pages(struct page *page, int numpages, int enable)
2701 __kernel_map_pages(page, numpages, enable);
2703 #ifdef CONFIG_HIBERNATION
2704 extern bool kernel_page_present(struct page *page);
2705 #endif /* CONFIG_HIBERNATION */
2706 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2708 kernel_map_pages(struct page *page, int numpages, int enable) {}
2709 #ifdef CONFIG_HIBERNATION
2710 static inline bool kernel_page_present(struct page *page) { return true; }
2711 #endif /* CONFIG_HIBERNATION */
2712 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2714 #ifdef __HAVE_ARCH_GATE_AREA
2715 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2716 extern int in_gate_area_no_mm(unsigned long addr);
2717 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2719 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2723 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2724 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2728 #endif /* __HAVE_ARCH_GATE_AREA */
2730 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2732 #ifdef CONFIG_SYSCTL
2733 extern int sysctl_drop_caches;
2734 int drop_caches_sysctl_handler(struct ctl_table *, int,
2735 void __user *, size_t *, loff_t *);
2738 void drop_slab(void);
2739 void drop_slab_node(int nid);
2742 #define randomize_va_space 0
2744 extern int randomize_va_space;
2747 const char * arch_vma_name(struct vm_area_struct *vma);
2748 void print_vma_addr(char *prefix, unsigned long rip);
2750 void *sparse_buffer_alloc(unsigned long size);
2751 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2752 struct vmem_altmap *altmap);
2753 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2754 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2755 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2756 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2757 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2758 void *vmemmap_alloc_block(unsigned long size, int node);
2760 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2761 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2762 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2763 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2765 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2766 struct vmem_altmap *altmap);
2767 void vmemmap_populate_print_last(void);
2768 #ifdef CONFIG_MEMORY_HOTPLUG
2769 void vmemmap_free(unsigned long start, unsigned long end,
2770 struct vmem_altmap *altmap);
2772 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2773 unsigned long nr_pages);
2776 MF_COUNT_INCREASED = 1 << 0,
2777 MF_ACTION_REQUIRED = 1 << 1,
2778 MF_MUST_KILL = 1 << 2,
2779 MF_SOFT_OFFLINE = 1 << 3,
2781 extern int memory_failure(unsigned long pfn, int flags);
2782 extern void memory_failure_queue(unsigned long pfn, int flags);
2783 extern int unpoison_memory(unsigned long pfn);
2784 extern int get_hwpoison_page(struct page *page);
2785 #define put_hwpoison_page(page) put_page(page)
2786 extern int sysctl_memory_failure_early_kill;
2787 extern int sysctl_memory_failure_recovery;
2788 extern void shake_page(struct page *p, int access);
2789 extern atomic_long_t num_poisoned_pages __read_mostly;
2790 extern int soft_offline_page(struct page *page, int flags);
2794 * Error handlers for various types of pages.
2797 MF_IGNORED, /* Error: cannot be handled */
2798 MF_FAILED, /* Error: handling failed */
2799 MF_DELAYED, /* Will be handled later */
2800 MF_RECOVERED, /* Successfully recovered */
2803 enum mf_action_page_type {
2805 MF_MSG_KERNEL_HIGH_ORDER,
2807 MF_MSG_DIFFERENT_COMPOUND,
2808 MF_MSG_POISONED_HUGE,
2811 MF_MSG_NON_PMD_HUGE,
2812 MF_MSG_UNMAP_FAILED,
2813 MF_MSG_DIRTY_SWAPCACHE,
2814 MF_MSG_CLEAN_SWAPCACHE,
2815 MF_MSG_DIRTY_MLOCKED_LRU,
2816 MF_MSG_CLEAN_MLOCKED_LRU,
2817 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2818 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2821 MF_MSG_TRUNCATED_LRU,
2828 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2829 extern void clear_huge_page(struct page *page,
2830 unsigned long addr_hint,
2831 unsigned int pages_per_huge_page);
2832 extern void copy_user_huge_page(struct page *dst, struct page *src,
2833 unsigned long addr_hint,
2834 struct vm_area_struct *vma,
2835 unsigned int pages_per_huge_page);
2836 extern long copy_huge_page_from_user(struct page *dst_page,
2837 const void __user *usr_src,
2838 unsigned int pages_per_huge_page,
2839 bool allow_pagefault);
2840 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2842 extern struct page_ext_operations debug_guardpage_ops;
2844 #ifdef CONFIG_DEBUG_PAGEALLOC
2845 extern unsigned int _debug_guardpage_minorder;
2846 extern bool _debug_guardpage_enabled;
2848 static inline unsigned int debug_guardpage_minorder(void)
2850 return _debug_guardpage_minorder;
2853 static inline bool debug_guardpage_enabled(void)
2855 return _debug_guardpage_enabled;
2858 static inline bool page_is_guard(struct page *page)
2860 struct page_ext *page_ext;
2862 if (!debug_guardpage_enabled())
2865 page_ext = lookup_page_ext(page);
2866 if (unlikely(!page_ext))
2869 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2872 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2873 static inline bool debug_guardpage_enabled(void) { return false; }
2874 static inline bool page_is_guard(struct page *page) { return false; }
2875 #endif /* CONFIG_DEBUG_PAGEALLOC */
2877 #if MAX_NUMNODES > 1
2878 void __init setup_nr_node_ids(void);
2880 static inline void setup_nr_node_ids(void) {}
2883 #endif /* __KERNEL__ */
2884 #endif /* _LINUX_MM_H */